Geographic variation, interactions with animals and sexual dimorphism: glancing at range limits in a gynodioecious plant.

Tesis Doctoral Antonio R. Castilla Álvarez Sevilla, Diciembre 2011 Supervisor Dr. Conchita Alonso Menéndez Researcher Department of Evolutionary Ecology Doñana Biological Station -CSIC (EBD-CSIC) Seville-Spain Supervisor Dr. Carlos M. Herrera Maliani Professor of Research Department of Evolutionary Ecology Doñana Biological Station -CSIC (EBD-CSIC) Seville-Spain Tutor Dr. Salvador Talavera Lozano Professor Department of Ecology and Plant Biology University of Seville Seville-Spain A mis padres y a Mercedes Según la teoría del mosaico geográfi co de coevolución, las variaciones espaciales en la magnitud del fl ujo génico y en el signo e intensidad de la selección ejercida por animales sobre los rasgos de las plantas, son los dos elementos esenciales que dan lugar a la aparición de mosaicos de selección y determinan el grado de adaptación entre las especies de plantas y animales. Hay ciertas situaciones naturales donde cabe esperar que se produzcan, de manera predecible y a lo largo de extensiones geográfi cas relativamente pequeñas, gradientes en la magnitud del fl ujo génico, así como en la forma e intensidad de selección por probables variaciones en la composición y/o abundancia de los polinizadores y herbívoros. Uno de estos lugares dotados de una singularidad ecológica propia son los márgenes de distribución de las especies vegetales. Además de sus bien conocidas peculiaridades ecológicas, los márgenes de distribución también presentan peculiaridades genéticas que los convierten en lugares ideales para verifi car las predicciones de la teoría del mosaico geográfi co de coevolución. El trabajo de esta tesis doctoral se diseñó para verifi car la siguiente predicción general, basada en los postulados de la teoría del mosaico geográfi co de coevolución: gradientes en la magnitud del fl ujo génico que sean consistentes con gradientes de selección por animales (polinizadores y herbívoros) sobre algunos caracteres vegetales, darán lugar a diferenciación intraespecífi ca en los caracteres sujetos a selección. Esta predicción se abordó mediante el estudio de patrones de selección por animales (polinizadores y herbívoros), variación fenotípica y fl ujo génico replicado en grupos de poblaciones centrales y marginales del arbusto ginodioico Daphne laureola. A continuación se presenta un resumen de cada uno de los capítulos incluidos en esta tesis doctoral, señalando los resultados principales de cada uno de ellos. Geographic variation, interactions with animals and sexual dimorphism: glancing at range limits in a gynodioecious plant. Geographic variation, interactions with animals and sexual dimorphism: glancing at range limits in a gynodioecious plant. - 18 - En el Capítulo 1 se comparan poblaciones centrales y marginales de D. laureola en las Sierras Béticas con respecto a su aislamiento espacial, tamaño poblacional, diversidad y diferenciación genética. Se comprobó que las poblaciones marginales no tienen un tamaño menor que las poblaciones centrales pero sí están más aisladas espacialmente. Sin embargo, el grado de aislamiento espacial de las poblaciones del margen occidental fue mucho mayor. Estas diferencias en la estructura geográfica de las poblaciones dieron lugar a un patrón diferente de estructuración genética en los dos márgenes de distribución estudiados y definieron dos grupos genéticos dentro de la especie en el área de estudio. En el Capítulo 2 aplicamos un enfoque genómico para detectar marcadores (AFLP loci) cuyas frecuencias mostrasen un patrón alejado del patrón neutral (outliers), una evidencia de selección natural sobre tales loci o loci ligados a los mismos. Encontramos que existe selección diferencial en poblaciones centrales y marginales de D. laureola. Una vez detectados los loci outliers, los retiramos del análisis de la estructura genética de las poblaciones para observar cuánto de la diferenciación genética entre los grupos disjuntos de poblaciones se podía atribuir a selección diferencial. Sin embargo, la fuerte diferenciación genética entre las poblaciones disjuntas marginales y el resto de poblaciones persistió. Este resultado demostró que factores selectivos y no selectivos son relevantes en la fuerte diferenciación genética de las poblaciones de D. laureola. Además, individuos hembra y hermafroditas mostraron patrones de selección diferentes, con hermafroditas presentando siempre un mayor número de loci bajo selección. Esta diferencia entre sexos fue más marcada en las poblaciones disjuntas marginales y sugirió una mayor presión selectiva sobre individuos hermafroditas que podría relacionarse con el ambiente de polinización. El Capítulo 3 analiza si los individuos de poblaciones disjuntas marginales y centrales continuas se diferencian en el tamaño y crecimiento de los individuos, su éxito reproductivo y ambiente de polinización. Los individuos de poblaciones centrales continuas presentaron tamaños mayores y produjeron más flores que dieron lugar a cosechas de frutos mayores. Sin embargo, la proporción de flores que dieron frutos (“fruit set”) fue mayor en las plantas de las poblaciones marginales, Geographic variation, interactions with animals and sexual dimorphism: glancing at range limits in a gynodioecious plant. - 19 - sugiriendo un servicio de polinización más eficiente en estas poblaciones. Además no detectamos diferencias en la densidad de individuos reproductivos en poblaciones centrales y marginales. Nuestros resultados no apoyaron por tanto fuertes diferencias de estrés ecológico entre poblaciones centrales continuas y marginales disjuntas, pero indicaron ciertos cambios en las interacciones planta-polinizador que podrían tener una especial relevancia para el polimorfismo sexual de esta especie ginodioica. Entre ellas cabe destacar que el escarabajo Meligethes elongatus, el principal polinizador de la especie en las poblaciones centrales continuas, no está presente en las poblaciones marginales disjuntas. En el Capítulo 4 comparamos el éxito de polinización en individuos de distinto sexo en poblaciones centrales y marginales, diferenciando sus componentes cuantitativo y cualitativo. Encontramos una reducción en la divergencia entre sexos en las poblaciones marginales disjuntas en la cantidad y calidad del polen recibido. Además, analizamos la variación fenotípica en algunos rasgos florales así como los patrones de selección por polinizadores sobre dichos rasgos. Encontramos una reducción en el dimorfismo sexual más evidente, la longitud del tubo de la corola, por reducción del tamaño de los hermafroditas. Además el único rasgo fenotípico estudiado que presentó un patrón de selección por polinizadores concordante con la variación geográfica observada fue la longitud del tubo de la corola: individuos hermafroditas con flores más largas tuvieron mayor éxito reproductivo únicamente en poblaciones centrales. Por tanto, selección diferencial por polinizadores junto con mecanismos de selección post-polinización parecen estar contribuyendo a las diferencias en el nivel de dimorfismo sexual floral entre poblaciones centrales continuas y marginales disjuntas de D. laureola. Por último, el Capítulo 5 se centra en el papel de los herbívoros sobre la diferenciación intraespecífica de los rasgos vegetativos de D. laureola. Los resultados de este capítulo demostraron que los herbívoros no parecen jugar un papel decisivo en este sentido. No encontramos diferencias entre poblaciones centrales continuas y marginales disjuntas en el nivel medio de herbivoría, descartando una mayor presión por herbívoros en los márgenes dedistribución. Además las diferencias Geographic variation, interactions with animals and sexual dimorphism: glancing at range limits in a gynodioecious plant. - 20 - en los niveles de herbivoría entre individuos no se correspondieron con las distancias genéticas entre ellos. Por tanto, estas diferencias parecen ser fruto de la variación espacial en el consumo por herbívoros debido a variación (micro)ambiental dentro de las poblaciones. Rasgos vegetativos como el área específi ca foliar o el contenido en fenoles de las hojas fueron diferentes en poblaciones centrales continuas y marginales disjuntas de D. laureola, pero esta diferenciación regional no se correspondió con una selección diferencial por herbívoros y podría tener relación con cambios regionales en la composición del dosel arbóreo y el ambiente lumínico del sotobosque. Lejos de ser entes aislados dentro de los ecosistemas, las plantas constituyen elementos dinámicos que continuamente interaccionan con otros organismos vivos. La naturaleza sésil de las plantas limita decisivamente dos fases críticas en su ciclo reproductivo como son la reproducción sexual y la dispersión de las semillas (Herrera 2002). Estos dos procesos requieren el movimiento en el espacio de estructuras reproductivas. Durante la reproducción sexual, los granos de polen deben ser transportados desde el individuo distancias variables hasta alcanzar los gametofi tos femeninos de otras plantas, en la población o fuera de ella. Del mismo modo, la dispersión de semillas implica el movimiento de éstas lejos de la planta madre. Ya que la capacidad de movimiento es una cualidad ampliamente extendida dentro del reino animal, no es de extrañar que las plantas hayan desarrollado una enorme variedad de mecanismos destinados a usar a los animales como vectores que permitan el movimiento de su polen y semillas. De hecho, actualmente está ampliamente aceptado que esta enorme explotación del movimiento animal para la dispersión del polen y de las semillas constituye uno de los factores decisivos en la tremenda diversifi cación y éxito ecológico de las angiospermas (Price 2002; Kay & Sargent 2009). También las interacciones de tipo antagonista han contribuido a la diversifi cación dentro de las angiospermas. La herbivoría o consumo de las plantas por animales engloba tipos muy diferentes de interacción que van desde las de tipo depredación (p.ej. consumo de semillas o plántulas) a las que no implican la muerte de individuo (p.ej. consumo de hojas). Debido a su naturaleza sésil, las plantas no pueden escapar del consumo por herbívoros y por tanto han tenido que desarrollar una enorme variedad de mecanismos destinados a evitar y/o tolerar el daño producido por los herbívoros. Por lo tanto, interacciones mutualistas y antagonistas entre plantas y animales han contribuido y continúan contribuyendo a la enorme diversifi cación dentro de las angiospermas (Labandeira 2002). Geographic variation, interactions with animals and sexual dimorphism: glancing at range limits in a gynodioecious plant. Variación geográfi ca en las interacciones planta-animal Geographic variation, interactions with animals and sexual dimorphism: glancing at range limits in a gynodioecious plant. - 22 - Una de la ideas centrales del pensamiento evolucionista actual es que los cambios macroevolutivos son el resultado acumulado de los procesos microevolutivos que tienen lugar a nivel intraespecífico (Simpson 1953, Bock 1970). Por este motivo, una de las maneras tradicionales de abordar el estudio de la adaptación ha sido investigar la variación geográfica en los caracteres de los organismos y su relación con los posibles factores selectivos que pudieran ser responsables de esa variación (Endler 1986). Esta idea la expresan de forma sintética Gould & Johnston (1972, p. 457) cuando afirman: “Since we view transpecific evolution as an extension of events at the species level, the foundation of most evolutionary theory rests upon inferences drawn from geographic variation or upon the verification of predictions made about it”. En el caso concreto de las plantas, se ha prestado especial atención a la variación geográfica en los rasgos relevantes para sus interacciones con animales (p.ej. polinizadores, herbívoros, dispersantes de semillas, etc) dada la importancia ya comentada previamente de este tipo interacciones en los procesos de diferenciación intraespecífica. En este sentido, la comunidad animal con la que una determinada especie vegetal interactúa se espera que varíe geográficamente. Por tanto, el estudio de la variación geográfica intraespecífica en rasgos relevantes para las interacciones planta-animal y su posible relación con una selección divergente efectuada por animales, es crucial para entender el vínculo entre micro y macroevolución en estos caracteres y el papel de las interacciones planta-animal en la evolución de los sistemas sexuales en plantas (Barrett et al. 2001; Zangerl & Berenbaum 2003; Herrera et al. 2006). Sin embargo, la mayoría de estas investigaciones sobre las relaciones entre plantas y animales se han llevado a cabo sin atender a su contexto geográfico, a pesar de su importancia para entender los procesos de evolución adaptativa (Herrera et al. 2006). La teoría del mosaico geográfico de coevolución (TMGC en adelante) propuesta por Thompson (1994, 2005) y los modelos genéticos asociados a ella (Gomulkiewicz et al. 2000; AlleaumeBenharira et al. 2006) proporcionan un modelo teórico que permite estudiar el papel causal de los animales en la aparición y mantenimiento de diferencias intraespecíficas en los rasgos fenotípicos importantes para las interacciones planta-animal. Según la TMGC, las dinámicas coevolutivas globales de las interacciones planta-animal vienen determinadas por tres componentes de la es- Geographic variation, interactions with animals and sexual dimorphism: glancing at range limits in a gynodioecious plant. - 23 - tructura geográfica de las poblaciones de las especies interactuantes: los mosaicos de selección, los puntos calientes coevolutivos y la remezcla de caracteres (“trait remixing”; Thompson 1994; 2005). Estos tres conceptos se describen brevemente a continuación: “Mosaicos de selección”: el resultado de las interacciones a menudo difiere entre poblaciones de las especies que interaccionan. Esto es debido principalmente a variaciones locales en el ambiente abiótico, en la estructura demográfica y genética y en el contexto de la comunidad ecológica donde se desarrolla la interacción. Estas diferencias entre poblaciones en el resultado de la interacción crean un mosaico geográfico en la interacción que es el material básico para la dinámica de la coevolución. “Puntos coevolutivos calientes y fríos”: consecuencia de la variación geográfica en el resultado de las interacciones, una interacción puede coevolucionar en algunas poblaciones (puntos calientes coevolutivos), mientras que en otras puede afectar a sólo una de las especies participantes o no tener ningún efecto (puntos fríos coevolutivos). Incluso pueden existir poblaciones de una especie participante en la interacción que se localicen fuera del área de distribución geográfica de la otra especie participante. “Remezcla de caracteres”: el mosaico geográfico se reajusta continuamente debido a los efectos combinados del flujo génico entre poblaciones, a la deriva genética y a las dinámicas de extinción/ colonización local. Especies sexualmente polimórficas: importancia de las variaciones geográficas en las interacciones planta-animal Las condiciones ecológicas en las que se desarrolla una especie varían tanto espacialmente como temporalmente. Poblaciones en diferentes regiones dentro del rango de una especie a menudo se enfrentan a escenarios ecológicos diferentes que dan lugar a cambios en la demografía, Geographic variation, interactions with animals and sexual dimorphism: glancing at range limits in a gynodioecious plant. - 24 - la estructura genética y las características fenotípicas de las poblaciones (Thompson 1994, 2005; Herrera et al. 2006). Tal diferenciación geográfica puede tener importantes consecuencias micro y macroevolutivas que en última instancia pueden llevar al aislamiento reproductivo y diversificación de linajes de plantas (Baker 1959; Lloyd 1965; Strid 1969; Grant 1971; McNeill & Jain 1983; Wyatt 1988; Olmstead 1990; Barrett 1995). Esta variación geográfica tanto en las condiciones abióticas como bióticas se piensa que ha contribuido a diversificar los sistemas reproductivos y sexuales de las Angiospermas (Barrett et al. 2001; Barrett 2002, 2010). Dentro de la impresionante diversidad de sistemas sexuales en plantas podemos encontrar desde especies monomórficas donde cada individuo contribuye reproductivamente a través de sus funciones sexuales masculina y femenina, hasta especies polimórficas donde los individuos están agrupados en dos o más morfotipos sexuales distintos que contribuyen a la siguiente generación principalmente a través de la función masculina o femenina (Lloyd 1980). Dentro de los sistemas monomórficos tenemos el hermafroditismo, donde los individuos portan flores con la función masculina y femenina desarrolladas y la monoecia donde las funciones masculinas y femeninas están separadas en flores distintas pero dentro del mismo individuo. En el caso de las especies dimórficas podemos encontrar la dioecia donde las funciones sexuales masculina y femenina aparecen en flores distintas que además son portadas por individuos distintos, la ginodioecia con individuos hembras y hermafroditas coexistiendo en sus poblaciones o la androdioecia con individuos machos y hermafroditas. Existen además sistemas hermafroditas pero polimórficos, como la distilia, a los que no nos vamos a referir. La dioecia no es muy frecuente en angiospermas a diferencia del reino animal donde es la condición más abundante (6 % de las especies de angiospermas; Renner & Ricklefs 1995). No obstante, la dioecia ha evolucionado desde el hermafroditismo en multitud de ocasiones, al menos hay 100 transiciones descritas (Barrett 2002; Charlesworth 2002). En concreto, la ruta evolutiva desde el hermafroditismo a la dioecia vía ginodioecia está apoyada teóricamente y empíricamente (Charlesworth 1999; Webb 1999; Weiblen et al. 2000). Esta ruta evolutiva implica dos fases con- Geographic variation, interactions with animals and sexual dimorphism: glancing at range limits in a gynodioecious plant. - 25 - secutivas, primero la invasión y expansión de mutantes estériles masculinos (i.e. hembras) dentro de poblaciones hermafroditas y, segundo, la acción de la selección natural para incrementar la función masculina en los hermafroditas. Sin embargo, apenas hay información sobre los mecanismos ecológicos que promueven estas dos transiciones, un aspecto clave para determinar si la ginodioecia puede ser un sistema evolutivamente estable (Ashman 2006). Darwin (1877) fue el primero que estableció una asociación entre dimorfismo sexual y estrés abiótico, señalando “a very dry station apparently favours the presence of the female form” (p. 301). Esta asociación entre dimorfismo sexual y estrés abiótico ha sido documentado en varios estudios filogenéticos (p.e. Hart 1985; Weller et al. 1995) y también por numerosos estudios de variación entre poblaciones (Weller et al. 1998; Asikainen & Mutikainen 2003; Vaughton & Ramsey 2004; Case & Barrett 2004). Sin embargo, aunque se piensa que los factores bióticos tienen un papel relevante en la aparición de sistemas polimórficos sexuales, apenas hay estudios que hayan examinado su papel causal. Por ejemplo se piensa que una reducción en la calidad de la polinización promueve diversas estrategias reproductivas como la transición desde la fecundación cruzada a la autofecundación (Schoen et al. 1996; Herlihy & Eckert 2002; Moeller & Geber 2005) o la evolución desde el monomorfismo al dimorfismo sexual (Sakai & Weller 1999; Barrett et al. 2001; Ashman 2006). En el caso concreto de la ginodioecia, se supone que los polinizadores pueden afectar a la evolución del sistema sexual de dos maneras contrapuestas (Ashman 2006). Por un lado, reducciones o cambios en la composición de la fauna de polinizadores podrían llevar a insuficiente o menor polinización en los individuos, respectivamente (ver Harder & Barrett 1996). En especies ginodioicas con hermafroditas autocompatibles, este peor ambiente de polinización puede llevar a un aumento de la autofecundación en los hermafroditas con el consiguiente incremento de la endogamia en su descendencia. El aumento de la endogamia en la descendencia de los hermafroditas puede suponer una desventaja a la hora de competir con la descendencia de los individuos hembra, producida exclusivamente por fecundación cruzada, promoviendo así la expansión de individuos hembra en las poblaciones (Charlesworth & Charlesworth 1978). Por otro lado, la propagación de los individuos hembras dentro de las poblaciones puede ser impedida por Geographic variation, interactions with animals and sexual dimorphism: glancing at range limits in a gynodioecious plant. - 32 - especie en el área de estudio con tres poblaciones localizadas en el margen disjunto suroccidental (Capítulos 3, 4 y 5). En el sureste peninsular la especie es frecuente en el macizo formado por las sierras CazorlaSegura-Castril-Alcaraz donde ocupa un amplio gradiente altitudinal (700-1900 m.s.n.m.). En Cazorla-Segura, en la mayoría de las localidades, D. laureola forma parte del sotobosque de pinares de Pinus nigra subsp salzmanii, donde también se encuentran otras especies leñosas como Quercus faginea, Prunus mahaleb, Taxus baccata, Acer monspessulanum, Juniperus communis o Crataegus monogyna. Mención especial merecen las poblaciones de D. laureola localizadas en la sierra de Segura, en el extremo nororiental de las Cordilleras Béticas, donde frecuentemente aparece asociada a acebedas. En estas poblaciones la vegetación se caracteriza por un dosel arbóreo de Pinus nigra subsp. salzmanii bajo el que se desarrollan abundantes los acebos (Ilex aquifolium), y otras especies tales como Corylus avellana, Quercus pyrenaica o Acer opalus subsp. granatense. En el suroeste peninsular D. laureola presenta un grupo más reducido de poblaciones localizadas en distintos macizos montañosos. En varias de estas poblaciones la especie que caracteriza el estrato arbóreo es Abies pinsapo, que da lugar a una de las formaciones boscosas más singulares de la Península Ibérica (Blanco et al. 1998). Bajo la cobertura de A. pinsapo crece una comunidad bastante pobre en términos de diversidad en la que destacan especies como Hedera helix, Ruscus aculeatus, Rubia peregrina o Iris foetidissima (Arista 1995). En zonas donde la cobertura de A. pinsapo es menor se establecen especies como Quercus faginea, Ulex baeticus, Crataegus monogyna o Rosa canina. Cada una de las poblaciones de estudio fue clasificada siguiendo el sistema de clasificación climático propuesto por Köppen en su última versión de 1936, conocida también como clasificación de Köppen-Geiger (Köppen 1931, 1936). A pesar de que esta clasificación se definió hace unos 100 años, sigue siendo una de las clasificaciones más utilizadas en estudios climatológicos a nivel mundial (Agencia Estatal de Meteorología & Instituto de Meteorología 2011). La clasificación de Geographic variation, interactions with animals and sexual dimorphism: glancing at range limits in a gynodioecious plant. - 33 - Köppen define distintos tipos de clima a partir de los valores medios mensuales de precipitación y temperatura. Los valores de precipitación y temperaturas para cada una de las poblaciones de estudio fueron estimados a partir de modelos de interpolación basados en registros meteorológicos de 30 años (Bustamante 2003; Tabla 1). Basándonos en los datos obtenidos, todas las poblaciones de estudio se encuandran en la categoría climática Csb propuesta por Köppen. Es decir, las poblaciones de D. laureola del sur de la Península ocupan áreas donde la temperatura media del mes más frío está comprendida entre 0 y 18 ºC, observándose un período estival marcadamente seco pero templado (temperatura media del mes más cálido ≤ 22 ºC y con cuatro meses o más con una temperatura media superior a 10 ºC). Poblacion X Y Altitud (m.s.n.m.) P (mm) T(ºC) m(ºC) M(ºC) Margen nororiental Acebeas 537013 4241386 1320 1067 11.5 3.9 22.0 Navalasna 539809 4240677 1202 969 11.5 3.8 21.9 Río Madera 534676 4236195 1220 1009 11.5 3.8 21.9 La Cumbre 528903 4225256 1480 912 11.7 4.1 22.0 Acebeas de los Jabalises 536391 4229352 1430 907 11.9 4.3 22.3 Región central Espinarea 503818 4192104 1173 870 11.2 3.8 21.5 Cañada del Espino 510134 4197123 1575 1068 11.0 3.7 21.3 Fuente Bermejo 514050 4198162 1513 1007 10.2 2.9 20.6 Valdecuevas 511627 4196202 1380 1051 10.7 3.4 21.1 Roblehondo 511209 4200234 1235 1036 11.1 3.6 21.3 Margen suroccidental Almijara 424124 4079200 1453 655 12.0 4.4 20.5 Algeciras 274886 3996149 506 887 16.9 8.3 22.8 Cañada de las Ánimas 319005 4063911 1333 1030 11.6 4.7 20.9 Fuente Molina 318354 4063324 1380 1046 11.5 4.6 20.8 Grazalema 283308 4072648 1229 1270 13.2 6.2 22.3 Sierra del Reloj 288021 4067469 1126 1665 13.4 6.4 22.5 Tabla 1. Caracterización de las poblaciones de estudio. Se indican las coordenadas geográficas (X,Y), la altitud, la precipitación media anual P, la temperatura media anual T, la temperatura media del mes más frío m y la temperatura media del mes más cálido M. Los valores de temperatura y precipitación para cada localidad fueron estimados a partir de modelos de interpolación basados en registros meteorológicos de 30 años (Bustamante 2003). Geographic variation, interactions with animals and sexual dimorphism: glancing at range limits in a gynodioecious plant. - 34 - Estructura de la Tesis Doctoral Esta Tesis doctoral consta de cinco capítulos cuyo contenido y finalidad se detallan a continuación. Todos los capítulos de esta tesis corresponden a trabajos originales, algunos de ellos publicados, otros en vías de publicación. En el Capítulo 1 (“Genetic structure of the shrub Daphne laureola across the Baetic Ranges, a Mediterranean glacial refugium and biodiversity hotspot”) se examina en primer lugar si en el área de estudio hay una reducción del tamaño poblacional y un incremento del aislamiento geográfico de las poblaciones hacia los márgenes de distribución de D. laureola. En un segundo paso, se analiza la estructura genética de las poblaciones en el área de estudio y concretamente se evalúa si hay una reducción de la diversidad genética y un aumento de la diferenciación genética en poblaciones marginales. Publicado en Plant Biology (en prensa). En el Capítulo 2 (“Differential selection between sexes and regions in a gynodioecious shrub revealed by genemic scanning”) se adopta un enfoque basado en el escaneo genómico (“genomic scanning”) para detectar evidencia de selección natural a partir de las frecuencias alélicas de un elevado número de marcadores genéticos (AFLP loci). Concretamente se examina si individuos de poblaciones centrales continuas y marginales disjuntas presentan patrones de selección diferencial, y si dentro de cada uno de estos grupos poblacionales, individuos con distinto sexo muestran también evidencia de selección diferencial. En revisión. El Capítulo 3 (“Exploring local borders of distribution in the shrub Daphne laureola: individual and population traits”) está enfocado en la comparación entre poblaciones centrales continuas y marginales disjuntas en relación al tamaño y éxito reproductivo de sus individuos. Tanto la identidad del polinizador principal como la abundancia de otras especies en floración como potenciales competidoras por los polinizadores son tenidas en cuenta en dicha comparación. Publicado en 2011, Acta Oecologica 37: 269-276. Geographic variation, interactions with animals and sexual dimorphism: glancing at range limits in a gynodioecious plant. - 35 - En el Capítulo 4 (“Reduced gender divergence in floral traits and pollination success at disjunct marginal populations of a gynodioecious shrub”) se investiga cómo un cambio en la identidad del polinizador principal afecta al éxito de polinización y a los patrones de selección floral en poblaciones centrales continuas y marginales disjuntas de D. laureola. En revisión. Por último, el Capítulo 5 (“Loose herbivory interactions in a long-lived understory plant: nongenetic basis for adult defoliation”) investiga las diferencias en los niveles de herbivoría en poblaciones centrales continuas y marginales disjuntas de D. laureola. También se exploran la existencia de diferencias entre estos grupos poblacionales en el efecto de la herbivoría sobre la fitness así como los patrones de selección fenotípica de los herbívoros sobre rasgos vegetativos. En revisión. Castilla, A.R., Alonso, C. and Herrera, C.M. In press. Genetic structure of the shrub Daphne laureola across the Baetic Ranges, a Mediterranean refugium and biodiversity hotspot. Plant Biology. Chapter 1.- Genetic structure of the shrub Daphne laureola across the Baetic Ranges, a Mediterranean glacial refugium and biodiversity hotspot Abstract: Distribution margins constitute areas particularly prone to random and/ or adaptive intraspecifi c diff erentiation in plants. This trend may be particularly marked in species discontinuously distributed across mountainous ranges, where sharp geographic isolation gradients and habitat boundaries will enhance genetic isolation among populations. In this study we analysed the level of neutral genetic diff erentiation among populations of the long-lived shrub Daphne laureola (Thymelaeaceae) across the Baetic Ranges, a glacial refugium and biodiversity hotspot in the western Mediterranean Basin. Within area, core and marginal populations of D. laureola were compared with regard to their spatial isolation, size, genetic diversity and diff erentiation. A spatially explicit analysis conducted on the vast majority of the species’ known populations in the study area (N = 111) showed that marginal populations (western and eastern) present larger spatial isolation than core populations, but are not smaller. We compared genetic diversity and diff erentiation between core and marginal populations using a subsample of fi fteen populations and 225 amplifi ed fragment length polymorphism (AFLP) markers. Core and marginal populations did not diff er in genetic diversity, probably because of the occurrence of large populations on the local margins. Western populations were strongly diff erentiated from the other populations. In addition, spatial and genetic diff erentiation among populations was larger on the western margin. Eastern populations constituted a genetically homogeneous group closely related to core populations, despite their greater spatial isolation. Results suggest that studies on phenotypic diff erentiation between core and marginal populations of D. laureola, and presumably other species likewise having discontinuous distributions across the Baetic ranges, should take into account geographical diff erences in level of genetic diff erentiation between the diff erent distribution borders. Genetic structure of the shrub Daphne laureola across the Baetic Ranges, a Mediterranean glacial refugium and biodiversity hotspot. - 38 - Resumen Los márgenes en la distribución de las especies vegetales constituyen áreas donde es particularmente factible encontrar diferenciación intraespecífi ca neutral y/o adaptativa. Esta tendencia debería ser especialmente marcada en especies distribuidas discontinuamente en distintas cordilleras montañosas, donde fuertes gradientes de aislamiento geográfi co y heterogeneidad de hábitat aumentarán el aislamiento genético de las poblaciones. En este capítulo analizamos el nivel de diferenciación genética neutral entre poblaciones del arbusto Daphne laureola (Thymelaeaceae) en las Cordilleras Béticas, un refugio glaciar y un punto caliente de biodiversidad en la cuenca occidental del Mediterráneo. Poblaciones centrales y marginales de D. laureola en el área de estudio fueron comparadas con respecto a su aislamiento espacial, tamaño poblacional, diversidad y diferenciación genética. Un análisis espacialmente explícito llevado a cabo usando la mayor parte de las poblaciones conocidas de la especie en el área de estudio (N = 111) mostró que las poblaciones marginales (occidentales y orientales) presentaron un mayor aislamiento espacial que las poblaciones centrales, pero no presentaron un menor tamaño poblacional. Usando una submuestra de 15 poblaciones y 225 marcadores de polimorfi smo en la longitud de fragmentos amplifi cados (AFLPs) comparamos a las poblaciones centrales y marginales en relación a su diversidad y diferenciación genética. Poblaciones centrales y marginales presentaron similares niveles de diversidad genética, probablemente como consecuencia de la existencia de poblaciones de gran tamaño en los márgenes locales de distribución. Las poblaciones marginales occidentales estuvieron genéticamente muy diferenciadas del resto de poblaciones. Además, tanto el aislamiento espacial como genético fue mayor en el borde occidental. Las poblaciones orientales constituyeron un grupo genéticamente homogéneo muy próximo a las poblaciones centrales, a pesar de su mayor aislamiento espacial. Los resultados de este estudio sugieren que los estudios sobre diferenciación fenotípica entre poblaciones centrales y marginales de D. laureola, y probablemente de otras especies con distribuciones disjuntas en sistemas montañosos, deberían tener en cuenta diferencias en el nivel de diferenciación genética entre los distintos bordes de distribución de la especie. Genetic structure of the shrub Daphne laureola across the Baetic Ranges, a Mediterranean glacial refugium and biodiversity hotspot. - 39 - Introduction High mountains are continental islands from a biogeographic perspective, each one surrounded by low-altitude environments characterized by unsuitable present-day climatic conditions for cold-adapted mountain plants (although see Birks & Willis 2008). Consequently many mountain plants often show disjunct geographical distributions. Within each distribution patch the species colonises a geographical gradient of environmental conditions and becomes more abundant where individual survival, reproduction and hence population growth are highest, and increasingly less abundant as conditions depart from this optimum (Hengeveld & Haeck 1982; Brown 1984). Therefore, within each distribution patch, a species is expected to exhibit a local center of abundance or core region, with populations becoming progressively smaller and more spatially isolated towards the local range limits (Brussard 1984; Vucetich & Waite 2003). Such a combination of geographic variation in population size and spatial isolation is expected to have important consequences for the genetic structure of plant populations (Vucetich & Waite 2003; Alleaume-Benharira et al. 2006). Genetic models of ecological margins pertain to local range margins rather than species global distribution margins because they consider a continuous spatial area that is not much larger than the dispersal distance (Alleaume-Benharira et al. 2006; Bridle & Vines 2007; Bridle et al. 2010). These genetic models predict, within an ecological gradient, (i) reduced genetic diversity of marginal populations because of founder eff ects, bottlenecks, inbreeding, genetic drift or directional selection; and (ii) increased genetic diff erentiation among marginal populations through reduced gene fl ow (Young et al. 1996; Lammi et al. 1999; Lowe et al. 2005). Furthermore, marginal and isolated populations could be genetically distinct as a result of the reduced arrival of maladapted genes from core populations (“gene swamping”; García-Ramos & Kirkpatrick 1997; Alleaume-Benharira et al. 2006), favouring adaptation to local environments and thus becoming particularly valuable for species maintenance at the global scale (Lesica & Allendorf 1995). Therefore, determining the level of spatial and genetic isolation of marginal populations with regard to core populations, as well as genetic diversity within populations, genetic diff erentiation among populations and its relationship with spatial isolation is necessary to evaluate the distinctiveness Genetic structure of the shrub Daphne laureola across the Baetic Ranges, a Mediterranean glacial refugium and biodiversity hotspot. - 40 - and adaptive potential of marginal populations, particularly in plants associated with mountainous environments. Recently Eckert et al. (2008) reviewed 84 studies involving 67 plant species that tested for declines in within-population genetic diversity and /or increases in among-population diff erentiation towards range margins using nuclear molecular genetic markers. They found that 64.2 % and 70.3 % of studies detected a decline of genetic diversity and an increase of genetic diff erentiation towards range margins, respectively, the two trends being positively associated. However, the generality of both trends is questionable given the substantial biogeographic and taxonomic biases in the available studies. First, most studies assume that marginal populations are smaller and more spatially isolated than core populations, although only a few investigations quantifi ed the spatial isolation of populations and its relationship with genetic diff erentiation (but see Eckstein et al. 2006; Medrano & Herrera 2008). Furthermore, most studies have focused on northern borders of a species yet diff erent selective pressures could operate in contracting versus expanding margins (Hampe & Petit 2005; Cassel-Lundhagen 2010). For most temperate woody species, the Mediterranean mountains constitute the southernmost distribution limits. These areas, in contrast to northernmost expanding distribution limits, represent long-term stable margins which have acted as glacial refugia during the last glacial stage (Benett et al. 1991; Carrión 2002; Carrión et al. 2003; Médail & Diadema 2009) . However, the number of molecular analyses of plant populations in these areas is still limited compared with more northern European mountain ranges such as the Alps, Pyrenees or Carpathians (Kropf et al. 2008; but see Kropf et al. 2006; Herrera & Bazaga 2008 a,b; Medrano & Herrera 2008). In addition, Mediterranean mountains constitute one of the World biodiversity hotspots because of their high complexity in terms of geology, climate and history (Thompson 2005; Blondel et al. 2010). Therefore, analysing the current genetic structure of plant populations in Mediterranean mountains may be useful to understand how the orography of heterogeneous landscapes contributes to genetic isolation of populations promoting intraspecifi c diff erentiation (Médail & Diadema 2009; Thompson 2005; pp: 77-80). Genetic structure of the shrub Daphne laureola across the Baetic Ranges, a Mediterranean glacial refugium and biodiversity hotspot. - 41 - In this study, we tested for diff erences in population size, spatial isolation and genetic diversity and diff erentiation of marginal and core populations of the temperate evergreen shrub Daphne laureola in the Baetic Ranges, a glacial refugium and one of the most biologically diverse regions in the Mediterranean Basin, located in the southern Iberian Peninsula. This area constitutes the southernmost distribution patch of the species in Europe, encompassing > 100 populations with a maximum distance among them ca. 350 km, and largely isolated from other distribution patches (Alonso et al. 2007). Specifi cally, we addressed the following questions: (i) Are marginal populations smaller and more spatially isolated than core populations? (ii) Do core populations have larger genetic diversity than marginal ones? (iii) Are marginal populations more genetically diff erentiated than core populations? (iv) Do marginal populations form distinct genetic groups to core populations at this southern distribution patch? Material and Methods Study Species Daphne laureola L. (Thymelaeaceae) has a Palaearctic distribution, extending from the Atlas Mountains in Morocco to the UK and Hungary northwards (Meusel et al. 1978; Fig. 1a). The species grows in the understory of coniferous and mixed forests, associated with the main calcareous ranges of Europe and North Africa. In the Iberian Peninsula, it shows a disjunct distribution, being abundant in the northern Cantabrian Range and the Pyrenees, and also in the southern Baetic Ranges, but absent in central areas (Alonso et al. 2007 and references therein). The species fl owers in winter, and the pollen beetle Meligethes elongatus Rosenhauer and, to a much smaller extent, small solitary bees and noctuid moths are its main pollinators in southeast Spanish populations (Alonso 2004). However, a recent study has shown that M. elongatus does not visit D. laureola fl owers in populations on the southwest edge (Castilla et al. 2011). Single-seeded black drupes ripen in June and are dispersed by several bird species (Obeso 1985; Hulme 1992). Genetic structure of the shrub Daphne laureola across the Baetic Ranges, a Mediterranean glacial refugium and biodiversity hotspot. - 48 - marginal populations were also explored with Wilcoxon exact tests. The infl uence of population size and connectivity on genetic diversity (PLP and Hj) and diff erentiation (DW and Fst) was analysed using Spearman correlations. A three-level hierarchical analysis of molecular variance (AMOVA; Excoffi er et al. 1992) was calculated with the program GenAlEx 6.3. We analyzed the following levels: (i) region (eastern, core and western); (ii) populations within region; and (iii) individuals within populations. Additionally, separate two level AMOVAs were calculated for each region to analyse the partitioning of molecular variance among and within populations. A standard test for genetic isolation by geographic distance was ruled out for two reasons. On the one hand, the low number of populations per region (N = 5) did not allow testing for genetic isolation by distance in each region. On the other hand, western populations were so distant from the remaining populations (> 138 km) that it introduces a large discontinuity better described by the categorical variable “region” than by the continuous variable “geographic distance”. Relationships among populations were visualized through neighbour-joining clustering of pairwise FST, and the signifi cance of the branches was assessed through 10,000 bootstrap replicates (Felsenstein 1985). A second approach based on statistical inference with Bayesian clustering methods using STRUCTURE 2.2.3. was also used to further elucidate the uppermost level of genetic structure in the study region (Pritchard et al. 2000; Falush et al. 2007). This program probabilistically assigns genotyped individuals into genetic groups in order to minimise departures from Hardy-Weinberg equilibrium and linkage equilibrium. The number of genetic groups was explored by performing 20 replicates of each simulation from K = 1 to K = 17, with a burn-in of 50,000 and MCMC of 100,000, assuming admixture and correlated allele frequencies as recommended in Pritchard et al. (2000). We applied Evanno et al.’s (2005) modal ΔK parameter as the choice criterion to detect the true number of genetic groups in the set of N = 533 individuals assayed. Genetic structure of the shrub Daphne laureola across the Baetic Ranges, a Mediterranean glacial refugium and biodiversity hotspot. - 49 - Results Are marginal populations more spatially isolated and smaller than core populations? We found diff erences between regions in connectivity (�2 = 1745.08, df = 2, P < 0.0001; KruskalWallis test), with density of neighbour populations being largest in the core region (Table 3). Core populations had more than four times the number of neighbour populations within 25 km than eastern populations (W = 43.29, P < 0.0001; Table 3). The diff erence with western populations was even larger. Core populations had more than 80 times the number of populations within 25 km than western populations (W = 17.20, P < 0.0001; Table 3). There was a signifi cant diff erence in population size among regions (�2 = 8.52, df = 2, P < 0.05; Kruskal-Wallis test). Unexpectedly, the core region had the lowest population sizes, with only 17 % of populations having more than 100 reproductive individuals, whereas the more isolated populations in the western region had the largest average size, with 83 % of populations having more than 100 reproductive individuals (Table 3). Eastern and core populations did not diff er in population size (W = 605.5, P = 0.12) but western populations had larger average population size than core populations (W = 122, P < 0.01). Connectivity was not correlated with population size(rs = - 0.15, P = 0.11; N = 111). Table 3. Average values of connectivity in eastern, central and western populations of D. laureola in the south of Iberian Peninsula. N indicates the number of populations in each region. Connectivity refers to the number of neighbour populations in a distance radius of 25 km. Population size was categorized according to the abundance of reproductive individuals (1, ≤30; 2, 31-100; 3, ≥ 100). The table shows the percentage of populations of each category per region. N Connectivity Population Size Category (%) 1 2 3 Eastern 17 20.8 ± 2.5 12 47 41 Central 88 82.6 ± 0.8 8 75 17 Western 6 1.7 ± 0.2 17 0 83 Genetic structure of the shrub Daphne laureola across the Baetic Ranges, a Mediterranean glacial refugium and biodiversity hotspot. - 50 - Do Core Populations Have Larger Genetic Diversity Than Marginal Ones? Average gene diversity of core populations did not diff er signifi cantly from eastern (W= 27, df = 8, P = 0.5) or western populations (W = 33, df = 8, P = 0.1), although mean Hj tended to be slightly larger in core (0.140) than in marginal populations (0.126 and 0.113 in eastern and western populations, respectively; Table 1). Western populations exhibited the broadest variation range of gene diversity, presenting the minimum and maximum values (0.068 and 0.195) of all studied populations. Similarly, core populations did not show statistically signifi cant diff erences in the percentage of polymorphic loci (PLP) relative to eastern (W = 27, df = 8, P = 0.5) or western populations (W = 33.5, df = 8, P = 0.1), although mean PLP tended to be larger in core than in marginal populations (33.9 %, 32.0 % and 26.7 % in core, eastern and western populations, respectively; Table 1). Again, the maximum and minimum PLP values (48.9 % and 13.3 %) were found in western populations. The FML population had the largest average values of both gene diversity and PLP in contrast to the rest of western populations with low values for both estimates of genetic diversity. In fact, excluding FML from analysis led to an important decrease in genetic diversity of western region (Hj = 0.093 and PLP = 21.1 %). Population size and PLP showed a positive relationship, although it was only marginally signifi cant (rs = 0.50, P = 0.06, N = 15). No signifi cant relationship was found between population size and gene diversity (rs= 0.29, P = 0.30, N = 15). PLP and gene diversity were both positively related to connectivity, although the relationships were only marginally signifi - cant (rs = 0.47, P = 0.07, N = 15 in both cases). Are Marginal Populations More Genetically Diff erentiated Than Core Ones? On average, western populations had more private fragments (2.0) than either eastern or core populations (0.6 and 0.2 private fragments respectively; Table 1). The westernmost population (ALG) had the highest number of private fragments despite its low values for mean gene diversity and percentage of polymorphic loci (Table 1). Mean pairwise FST was larger in western than in core populations (W = 55, df = 8, P < 0.0001; Fig. 2), which were in turn more genetically diff erentiated than eastern ones (W = 64, df = 8, P < 0.001; Fig. 2). Genetic structure of the shrub Daphne laureola across the Baetic Ranges, a Mediterranean glacial refugium and biodiversity hotspot. - 51 - Figure 2. Genetic diff erentiation of the populations in core and marginal regions. Population pairwise FST and genetic rarity (DW) per region (eastern, core and western) are represented by means (±SE). Statistically signifi cant variation between marginal and core populations is indicated with asterisks (* P <0.01; ** P <0.001; *** P <0.0001 ). Furthermore, the genetic rarity index DW was signifi cantly larger in western populations than in core populations (W = 16, df = 8, P < 0.01; Fig. 2). In contrast, there were not diff erences in DW between core and eastern populations (W = 26.5, df = 8, P = 0.4; Fig. 2). Interestingly, genetic rarity were negatively related to connectivity (rs = -0.59, P < 0.05, N = 15). Population pair-wise FST also was negatively related to connectivity of populations, although the relationship was only marginally signifi cant (rs = -0.47, P = 0.08, N = 15). Population size was unrelated to genetic rarity and population pair-wise FST (rs = 0.03, P = 0.90 and rs = 0.08, P = 0.79 respectively; N = 15). Genetic structure of the shrub Daphne laureola across the Baetic Ranges, a Mediterranean glacial refugium and biodiversity hotspot. - 52 - The three-level AMOVA revealed a strong genetic diff erentiation among regions (�ST = 0.45). Most molecular variance was accounted for diff erences among regions and within populations (46 and 40 % respectively), whereas diff erences among populations within a region accounted for a reduced amount of molecular diversity (16 %). Results of the AMOVA analysis also showed that western populations were more genetically diff erentiated among them, with larger values of molecular variance explained among populations, and an average �ST higher than eastern and core populations (Table 4). Eastern populations presented the lowest values of molecular variance explained among populations and �ST (Table 4). Source of variation d.f. % Variance ФST P-value 1. Three-level AMOVA Among regions 2 45 0.455 0.0001 Among populations 12 14 Within populations 518 40 2. Eastern Among populations 4 9 0.090 0.0001 EWithin population 170 91 3. Core Among populations 4 13 0.125 0.0001 Within population 169 87 4. Western Among populations 4 47 0.472 0.0001 Within population 179 53 Table 4. Genetic diff erentiation (�ST) and percentage of molecular variance distributed among regions, among populations and within populations in the three regions defi ned in this study. 9999 permutations were used in the analysis. Genetic structure of the shrub Daphne laureola across the Baetic Ranges, a Mediterranean glacial refugium and biodiversity hotspot. - 53 - Do Marginal Populations Form Genetically Distinct Groups? Two distinct genetic clusters, one represented by the western populations and the other grouping together core and eastern populations, were supported by both neighbour-joining and Bayesian approaches (Fig. 3 and 4). The dendrogram based on neighbour-joining approach showed a split between western and the other populations with a support of 100 % (Fig. 3). Within the largest group, two populations, VCV and FBJ, both located in a particular watershed (Guadalentín Valley; Fig. 4b), appeared diff erentiated from each other and from the rest of the core and eastern populations, with a support of 97 % in both cases (Fig. 3). In addition, the Bayesian analysis showed a distinct modal maximum of ΔK at K = 2 genetic groups. Membership assignments to the two genetic groups exhibited a distinct geographic pattern coincident with the results of the neighbour-joining approach (Fig. 4a), supporting that western populations formed a genetic group diff erentiated from the rest of populations analysed. Figure 3. Neighbour-joining clustering based on FST, showing the relationship among the 15 sampled populations of D. laureola. Bootstrap percentage values are indicated above the branches. Populations coded as in Table 2 and regions as in Fig. 1. Genetic structure of the shrub Daphne laureola across the Baetic Ranges, a Mediterranean glacial refugium and biodiversity hotspot. - 54 - Figure 4. (a) Genetic population structure of Daphne laureola in the Baetic Ranges inferred using the model-based Bayesian clustering implemented in STRUCTURE which does not require a priori categories. The individual pie charts indicate the mean proportion of membership of each local population for the inferred number of K = 2 genetic groups. (b) Detailed population structure within the largest genetic group that combines core and eastern populations. Dashed line indicates the highest elevation of the mountainous range. Populations coded as in Table. Discussion Genetic characteristics of populations may be strongly infl uenced by the size and spatial distribution of populations through their relationships with genetic drift and gene fl ow. Thus, these characteristics are expected to vary across species’ geographic ranges (Eckert et al. 2008). In the following paragraphs, we will discuss our results supporting, fi rst, the increased spatial isolation of marginal populations and, second, the genetic diff erentiation of the most spatially isolated local margin of the shrub D. laureola in the Mediterranean Baetic Ranges, an area close to the species southern distribution range. Genetic structure of the shrub Daphne laureola across the Baetic Ranges, a Mediterranean glacial refugium and biodiversity hotspot. - 55 - Spatial Isolation and Size of D. laureola Populations in a Southern Distribution Patch Our results, based on an extensive sampling across the Baetic Ranges, showed that populations of this temperate shrub were fewer and more spatially isolated towards local eastern and western margins, supporting the hypothesis of increased isolation at local range margins (Brown 1984; Vucetich & Waite 2003; Bridle et al. 2010), and highlighting the suitability of the spatially explicit analysis adopted here to identify meaningful geographic discontinuities (Table 2). The western region comprised a few highly isolated populations located in diff erent mountain ranges (Fig. 1b), the core region with abundant and very connected populations at all distances comprised populations mainly located in the Sierras de Cazorla and Las Villas, and the eastern region associated with those populations in Sierras de Segura and Alcaraz, where the species is still common but populations exhibit a reduction in connectivity among them (Fig. 1c; Table 2). The predicted decline in population size towards local borders of distribution was not confi rmed in D. laureola populations across the Baetic Ranges (see also Jump & Woodward 2003; Herlihy & Eckert 2005). In fact, core populations were on average smaller than marginal populations, although the trend was only signifi cant in the comparison between core and western populations. The wider altitudinal range of core populations could likely contribute to variability in population size in this region through altitudinal gradients in ecological conditions (Herrera and Bazaga 2008b). The presence of large marginal populations with more than 100 reproductive individuals suggests the existence of some suitable locations without evidence of ecological stress for D. laureola towards the local borders of distribution (Castilla et al. 2011). Genetic Diversity, Diff erentiation and Structure, and their Relationships with the Spatial Isolation of Populations Core and marginal populations of D. laureola in the Baetic Ranges exhibited similar levels of genetic diversity, and regions did not diff er signifi cantly in either gene diversity or percentage of polymorphic loci within a population. These results contrast with those found for other plants (e.g., Lammi et al. 1999; Lönn & Prentice 2002; Arnaud-Haond et al. 2006; Eckert et al. 2008). The absence of clear diff erences between core and marginal populations in genetic diversity could be Genetic structure of the shrub Daphne laureola across the Baetic Ranges, a Mediterranean glacial refugium and biodiversity hotspot. - 56 - related to the infl uence of population size (Lammi et al. 1999; Leimu et al. 2006) because marginal populations, contrary to expectations, were not smaller than core populations. Further, our results supported that PLP was marginally related to population size. Therefore, the existence of large populations on D. laureola’s local borders of distribution could prevent genetic erosion through mechanisms of genetic rescue (Young et al. 1996; Garant et al. 2007). Interestingly, the western margin was internally very heterogeneous with regard to the genetic diversity of the populations. With the exception of FML population that exhibited the highest genetic diversity among all study populations, the other four tended to have a much more reduced genetic diversity than core and eastern populations, mostly supporting the expectation of a reduced genetic diversity in the most spatially isolated populations of the western margin, with the single exception already mentioned. Strong genetic diff erentiation among D. laureola populations across the Baetic Ranges was indicated in our results using AFLP markers (�ST = 0.45). These results are consistent with observed genetic diff erentiation in DNA sequence in a much reduced number of sampled populations and individuals (Alonso and Herrera 2011). In addition, the two estimates of genetic diff erentiation (FST and DW) were positively related to the spatial isolation of populations. In particular, results showed that the highly isolated western populations constitute a genetic group strongly diff erentiated from core and eastern populations, supporting a positive relationship between geographic isolation and genetic diff erentiation in our study system (Eckstein et al. 2006; Medrano & Herrera 2008). Also, our results suggest the existence of two types of local borders of distribution in the study area from a genetic viewpoint: a western margin with populations being both more spatially and genetically isolated among them, as usually happens in the rear edge of species distributions (Hampe & Petit 2005); and an eastern margin comprising a group of more spatially isolated but genetically homogeneous populations, closely related to core populations and therefore more typical of a expanding edge (Schönswetter et al. 2002; Csërgo et al. 2009). The genetic similarity among core and eastern populations suggests extensive gene fl ow between the two regions. Eastern populations are distant by only about 20-30 km from core populations, all located along an extensive mountainous range, which could facilitate local gene fl ow through natural corridors Genetic structure of the shrub Daphne laureola across the Baetic Ranges, a Mediterranean glacial refugium and biodiversity hotspot. - 57 - for the dispersal vectors of the species (see e.g. Gaudeul et al. 2000; Barrett et al. 2004). However, western populations are located far away from the core populations (130 km approximately), limiting the gene fl ow among western and core populations to long-distance infrequent dispersal events (Kropf et al. 2006; Kropf et al. 2008). In addition, western populations are located in diff erent mountainous ranges separated by dozens to hundreds of kilometres of unsuitable habitats for the species, which could also reduce gene fl ow among populations on the western edge, thus promoting their genetic diff erentiation. Supporting this notion, western populations also had the highest values of genetic rarity estimates and recent studies have also revealed some ecological particularities of these populations. For instance, the female frequency decreases in southwest populations, the ALG population being constituted exclusively by hermaphrodite plants in the most extreme case (Alonso et al. 2007). In addition, individuals of western populations present lower fl oral displays than those of core populations, coincident with a change in the pollinator fauna (Castilla et al. 2011). The ecological particularities of western populations and their strong genetic isolation highlight the potential role that local adaptation events could have in these marginal disjunct populations (Lesica & Allendorf 1995; García-Ramos & Kirkpatrick 1997; Arnaud-Haond et al. 2006; Cassel-Lundhagen 2010). Concluding Remarks Among-population phenotypic divergence is relatively common feature in plant species across Mediterranean mountain ranges, geographic margins of species’ ranges being particularly prone to the evolution of this random or adaptive diff erentiation. Background information on patterns of neutral diff erentiation is particularly interesting for disentangling the role of natural selection on this phenotypic divergence. This approach is especially useful in plant species of Mediterranean mountain ranges because of the strong relevance of historical processes on the genetic structure of their populations. In this sense, genetic structure analyses comparing core and marginal populations based on neutral markers emerge as a useful tool to estimate the level of neutral diff erentiation in populations where among-population phenotypic diff erentiation is expected. Results of our study show that neutral diff erentiation varies on diff erent local margins of distribution. In Diff erential selection between sexes and regions in a gynodioecious shrub revealed by genomic scanning. - 64 - phism in reproductive traits contrasting with the modest sexual dimorphism in vegetative traits (Ashman, 2005). Therefore, diff erential selective regimes on reproductive traits (e.g. fl oral traits) of hermaphrodite and female individuals are expected to occur. Pollinators constitute important selective agents on reproductive traits and thus, they may have a relevant role on determining the extent of sexual dimorphism in these traits, particularly because pollinator fauna often vary geographically (Herrera et al., 2006; Ashman, 2000; Ashman & Diefenderfer, 2001; Ashman, 2006 and references therein). Daphne laureola L. (Thymelaeaceae) is a sexually polymorphic shrub that presents gynodioecious populations with hermaphrodite and female individuals, and monomorphic populations made up exclusively of hermaphrodites (Alonso et al., 2007). In gynodioecious populations, D. laureola shows sexual dimorphism in traits related to its fl oral morphology and chemical defences. Hermaphrodites have larger fl owers than females (Alonso, 2005), a widespread pattern in gynodioecious species (Eckhart, 1999; Shykoff et al., 2003). Furthermore, females present larger concentration of specifi c chemical defences (coumarins) than hermaphrodites (Alonso et al., 2005). We studied patterns of selection in populations of D. laureola located across the Baetic Ranges in southern Iberian Peninsula. Genetic structure analyses have shown that D. laureola presents two genetically diff erentiated population groups (SE and SW populations hereafter; Alonso & Herrera, 2011; Castilla et al., unpublished data). However, the inclusion of only a few outlier loci under selection could greatly bias the estimates of population evolutionary history as determinant of genetic population structure (Luikart et al., 2003; Herrera & Bazaga, 2008). Therefore, the analysis of the same polymorphic markers discarding the outlier loci may reveal how much of the genetic divergence between SE and SW populations of D. laureola across the Baetic Ranges is accounted by diff erences in the selective regimes. Furthermore, SE and SW populations diff er also in the pollinator fauna and their overall female frequency (Alonso et al., 2007; Castilla et al., 2011). Diff erences in reproductive success and phenotypic traits (e.g. fl oral display) suggest the existence of local adaptation in the most isolated populations of the SW region (Castilla et al., 2011). In this study, we adopted a genomic scan approach to detect signatures of diff erential selection between Diff erential selection between sexes and regions in a gynodioecious shrub revealed by genomic scanning. - 65 - female and hermaphrodite individuals of D. laureola, and to asses whether patterns of sex-specifi c selection remained consistent between the SE and SW populations. Specifi cally, we addressed the following questions: (1) are there genetic signatures of diff erential selection between the two study regions? (2) does the genetic divergence between SE and SW populations of D. laureola remain after excluding loci under selection? (3) are there contrasting selective regimes on individuals of diff erent sex? (4) do diff erences between sexes in the selective patterns remain consistent across regions? Material and methods Study Species and Populations Daphne laureola L. (Thymelaeaceae) is an evergreen shrub with a disjunct Palaearctic distribution, with distribution patches on the main calcareous mountainous ranges of Europe and northern Africa, including the Baetic Ranges in Southern Spain where this study was conducted (Fig. 1). Within the Baetic Ranges, D. laureola is less frequent in the SW region, where populations are characteristically more isolated and genetically diff erentiated from those in the SE region, which represents the core area of distribution (Alonso & Herrera, 2011; Castilla et al., unpublished data). The two regions also diff er in the composition of pollinator fauna, with SE populations presenting as main pollinator the pollen beetle Meligethes elongatus Rosenhauer, which is absent in SW populations (Alonso, 2004; Castilla et al., 2011). Furthermore, overall female frequency is lower in SW populations (Alonso et al., 2007). Diff erential selection between sexes and regions in a gynodioecious shrub revealed by genomic scanning. - 66 - Figure 1. Distribution of Daphne laureola populations in the Baetic Ranges (Southern Spain), and location of the SE and SW populations defi ned in this study, here denoted by fi lled circles and black triangles, respectively. The grey scale corresponds to diff erent intervals of altitude (m.a.s.l.). AFLP analyses Were sampled 15 populations of D. laureola, ten in the most densely populated SE region and fi ve in the marginal SW region (Fig. 1). In March 2007, fresh leaf material was collected from 25-40 randomly chosen reproductive individuals per population, placed in small paper envelopes and dried immediately at ambient temperature in sealed containers with abundant silica gel. Dried leaves were homogenized to a fi ne powder using a Retsch MM 200 mill. Total genomic DNA was extracted from approximately 15 mg of ground leaf material using DNeasy Plant Mini Kit (Qiagen) and following the manufacturer protocol. DNA concentration of extracts was estimated by running electrophoreses of 5 μl aliquots on 0.8 % agarose gels. The AFLP analysis was performed essentially as originally described by Vos et al. (1995), with modifi cations involving the use of fl uo- Diff erential selection between sexes and regions in a gynodioecious shrub revealed by genomic scanning. - 67 - rescent dye-labeled selective primers following Applied Biosystems (2005). Restriction-ligation was conducted using EcoRI / MseI and PstI / MseI endonuclease mixtures and doubled-stranded adaptors. A total of 32 EcoRI + 3 MseI + 3 and 32 PstI + 3 MseI + 3 primer pairs were fi rst screened for selective amplifi cation in a pilot study conducted with a random subsample of 8 individuals from four widely scattered populations. We fi nally selected four EcoRI + 3 / MseI + 3 and four PstI / MseI + 3 primer combinations on the basis of their higher polymorphism and ease of interpretation of band pattern (see Table 2 in Chapter 1). Each plant was fi ngerprinted using the eight combinations chosen. Fragment separation and detection was made using an ABI PRISM 3100 DNA sequencer. The presence or absence of each marker in each individual plant was scored manually by visualizing electrophoregrams with GeneMapper 3.7 software. All scoring was done by the same person (A.R.C.). Prior to statistical analyses, we excluded markers present in < 1 or > 99 % of individuals. Furthermore, only fragments ≥ 150 base pairs in size were considered, as a way of reducing the potential impact of size homoplasy (Vekemans et al., 2002; Caballero et al., 2008). In addition, fi ve plants that produced noisy sequencer electrophoregrams for some primer combinations after running the analyses several times, were also excluded from the sample. Our fi nal data set consisted of a total of 493 individuals, each scored for presence / absence of 225 polymorphic loci. Genotyping error rates were determined for each primer combination by running repeated, independent analyses for a total of 32 randomly chosen individual plants, and estimated as the ratio of the total number of loci with discordant scores (all individuals combined) to a product of the number of individuals by the total number of scored loci (Bonin et al., 2004; Pompanon et al., 2005). Error rates varied among primers combinations, being larger in EcoRI combinations than in PstI combinations (see Table 2 in Chapter 1). The mean error rate (± SE) for the whole set of eight combinations was 0.9 ± 0.4%. Diff erential selection between sexes and regions in a gynodioecious shrub revealed by genomic scanning. - 68 - Detection of Loci Under Selection Analyses to identify loci under selection were conducted separately for SE and SW populations of D. laureola. In addition, we analysed separately female and hermaphrodite individuals within each population group (57 females and 141 hermaphrodites in SW populations; 102 females and 193 hermaphrodites in SE populations). The software package BayeScan version 2.01 (Foll & Gagiotti, 2008; Foll et al., 2010) was used to test for the neutrality of the AFLP markers. BayeScan extends Beaumont & Balding’s (2004) Bayesian method that implements the multinomial-Dirichlet likelihood, which arises in a wide range of neutral population genetic models. The methodology consists of identifying loci that present population diff erentiation coeffi cients (FST) that are distinct from those under neutral expectations. A locus is suspected to be under selection if the locus-specifi c eff ect is signifi cantly diff erent from zero. The method developed by Foll & Gagiotti (2008) uses a hierarchical Bayesian approach to estimate the posterior probabilities of two alternative models, one including the eff ects of the selection and one excluding them. In the context of the multiple testing, we also need to incorporate our scepticism related to statistical error type I about the chance that each locus is under selection. This is done by BayeScan setting the prior odds of the neutral model and using the Posterior Odds instead of Bayes factors to make decisions (Foll et al., 2010). Posterior odds are simply the ratio of posterior probabilities, and indicate how more likely the model with selection is compared to the neutral model. Posterior probabilities directly allow the control of the False Discovery Rate (FDR). In our analyses, we calculated the posterior probabilities threshold to detect loci under selection leading to a False Discovery Rate ≤ 0.001 and we used a value of prior odds of 10, which is reasonable for the identifi cation of candidate loci within a few hundreds of markers (Foll et al., 2010). We ran BayeScan with 10 pilot runs, burn-in of 50000, total of 550000 iterations, sample size of 10000 and thinning interval of 50. Diff erential selection between sexes and regions in a gynodioecious shrub revealed by genomic scanning. - 69 - Genetic Structure of Populations We removed outlier loci from the analysis of the genetic structure to detect how much of the genetic divergence between SE and SW populations of D. laureola across the Baetic Ranges was accounted by diff erences in the selective regimes. The inclusion of only a few outlier loci among many neutral loci can greatly bias estimates of the eff ects of population evolutionary history (Luikart et al., 2003; Herrera & Bazaga, 2008). Therefore, to avoid making erroneous conclusions from using outlier loci, population parameters should be estimated without outliers (Luikart et al., 2003; Herrera, 2011). The loci departing signifi cantly from neutral expectation were removed from further analyses of genetic structure and gene fl ow, altogether 13 loci (Supplementary material). An approach based on statistical inference with Bayesian clustering methods using STRUCTURE 2.2.3. was then used to further elucidate the uppermost level of genetic structure in the study region (Pritchard et al., 2000; Falush et al., 2007). This program probabilistically assigns genotyped individuals into genetic groups in order to minimize departures from Hardy-Weinberg equilibrium and linkage disequilibrium. The number of genetic groups was explored by performing 20 replicates of each simulation from K=1 to K=17, with a burning of 50000 and MCMC of 100000, assuming admixture and correlated allele frequencies as recommended by Pritchard et al. (2000). We applied Evanno et al.’s (2005) modal ΔK parameter as the choice criterion to detect the true number of genetic groups in the set of individuals assayed. Results Loci Under Selection and Neutral Genetic Structure of Populations BayeScan identifi ed six and seven AFLP markers with evidence of purifying selection in the SE and SW populations, respectively (3 % of all markers approximately in both cases; Fig. 2a). Note that loci under purifying selection in one group of populations were not under detectable selection in the other group (Fig. 2b). Diff erential selection between sexes and regions in a gynodioecious shrub revealed by genomic scanning. - 70 - Figure 2. Detection of selection using a Bayesian genomic scan approach in SE and SW populations of Daphne laureola (black and grey dots respectively). Each dot corresponds to an AFLP marker (N = 225). Dashed lines indicate the threshold value of the log (PO) beyond which there is evidence for a locus to be subject to purifying selection at a False Discovery Rate of 0.001. Note that a threshold value was calculated independently for SE and SW populations. (a) Plot of the estimated FST values against the (log PO) for each locus with labels of the outlier loci. (b) Comparison of loci-specifi c patterns of purifying selection (log PO) in SE and SW populations. Solid lines join values of log (PO) of the same locus in SE and SW populations. Only loci under purifying selection in at least one group of studied populations were represented. Diff erential selection between sexes and regions in a gynodioecious shrub revealed by genomic scanning. - 71 - Bayesian analysis of putatively neutral genetic variation (i.e. after removing selected loci from the data), which did not rely on predetermined assignment of samples to regions or populations, showed a distinct modal maximum of ΔK at K = 2 genetic groups. Note that SW populations formed a genetic group diff erentiated from the rest of populations with nearly all populations being assigned exclusively to the SW genetic group (Fig. 3). Figure 3. Genetic population structure of Daphne laureola in the Baetic Ranges inferred with the model-based Bayesian clustering implemented in STRUCTURE from data on putatively neutral loci alone. The individual pie charts indicate the mean proportion of membership of each local population for the inferred number of K = 2 genetic groups. Diff erential Selection on Sexes Very few loci were subject to purifying selection in female plants. In SE populations, hermaphrodite individuals presented three loci under purifying selection in comparison to only one in females (Fig. 4 a-c). The diff erence between sexes was even sharper in SW populations, where hermaphrodite and female individuals exhibited seven and one only locus under purifying selection, respectively (Fig. 4d-f). The separate analyses by sex confi rmed that the identity of the loci under selection on each sex was diff erent in SE and SW populations (Fig. 4). Diff erential selection between sexes and regions in a gynodioecious shrub revealed by genomic scanning. - 72 - Figure 4. Sex-related selection patterns revealed by Bayesian genomic scan in the SE (upper row) and SW (lower row) populations of the gynodioecious shrub Daphne laureola. Within each row, from left to right, the fi rst two graphs plot estimated FST of each locus against its corresponding value of log (PO) with labels of the loci outliers, separately for hermaphrodites (a, d) and females (b, e), respectively. The third graph illustrates diff erences between sexes in the log (PO) values of loci found subject to purifying selection in at least one sex (c, f). Solid lines join the values of log (PO) of the same locus in hermaphrodites and females. Each dot corresponds to an AFLP marker (N = 225; white and black dots, hermaphrodites and females, respectively). On each panel, dashed lines indicate the threshold value of the log (PO) beyond which there is evidence for a locus to be subject to purifying selection at a False Discovery Rate of 0.001. Independent threshold values were calculated for each dataset (SE hermaphrodites, SE females, SW hermaphrodites, SW females). Discussion Diff erential Selection Between Regions and Neutral Genetic Structure Strong genetic diff erentiation among disjunct populations is a common feature in plant species of Mediterranean mountainous ranges (Schönswetter et al., 2006; Ronikier et al., 2008; Csergö et al., 2009). This genetic divergence at intraspecifi c level could be the result of non-selective mechanisms such as random fi xation of diff erent alleles in disjunct distribution patches because of founder events during colonization or genetic drift arising from reduction of population sizes during warmer stages in the Holocene (Young et al., 1996; Thompson, 2005). Contrasting selective regimes in disjunct parts of the species distribution range can also lead to strong genetic diff erentiation among disjunct distribution patches (Kawecki, 2008; Sexton et al., 2009). Divergent selective regimes in disjunct groups of populations because of variation in abiotic and/or biotic factors have Diff erential selection between sexes and regions in a gynodioecious shrub revealed by genomic scanning. - 73 - been documented in some plant species (Allan & Pannell, 2009; Hereford & Winn, 2008; Parisod & Joost, 2010). Populations of D. laureola in southern Iberian Peninsula present a strong genetic diff erentiation, exhibiting a clear split between SW and SE populations (Alonso & Herrera, 2011; Castilla et al., unpublished data). Results of the present study suggest that both non-selective and selective mechanisms have most likely contributed to the strong genetic divergence of D. laureola populations across the Baetic Ranges. In particular, we found that SE and SW populations of D. laureola present a similar number of loci under purifying selection, which may tentatively be interpreted as an indication of equivalent levels of selective pressure. This result contradicts to the expectation of marginal populations representing areas with stronger selective pressures than core populations (Kawecki, 2008; Sexton et al., 2009). However, the non-overlapping identity of the loci under selection in both groups of populations suggests diff erent selective regimes in SE and SW populations. Reduced fl oral display sizes and increased relative fruit set associated to a change in the pollinator fauna has been reported in SW populations (Castilla et al., 2011). Studies of phenotypic selection by pollinators on fl oral traits in these two disjunct groups of populations are necessary to understand the role of geographic variation in plant-pollinator interactions on the intraspecifi c diff erentiation of D. laureola populations. Our results also supported a role of non-selective mechanisms in contributing to the strong genetic diff erentiation of the D. laureola across the Baetic Ranges. The clear genetic split between SW and SE populations of species previously reported (Alonso & Herrera, 2011; Castilla et al., unpubl. data) remained after removing all the loci under purifying selection. Range shifts associated with climatic oscillations during the Quaternary have led to random fi xation of diff erent alleles in disjunct groups of populations contributing to genetic intraspecifi c diff erentiation of plant species (Bittkau & Comes, 2005; Schönswetter et al., 2006). In this sense, Baetic Ranges emerge as a mountainous range less aff ected by the glaciations during the Quaternary than more northern mountain systems (Benett et al., 1991; Médail & Diadema, 2009) and thus the strong genetic divergence of D. laureola populations could refl ect a long-term history of genetic isolation between the two currently geographically disjunct groups of populations here studied. Exploring local borders of distribution in the shrub Daphne laureola: individual and population traits. - 80 - Plant-pollinator interactions are an example of selection pressures expected to vary geographically following spatial changes in the abundance and identity of the fl oral visitors (Moeller, 2005, 2006; Thompson, 2005). In general, small and/or isolated populations attract fewer and less diverse pollinators (Groom, 1998; Sih and Baltus, 1987). Also, marginal populations have been found to have poorer pollinator faunas than central populations (Silva-Montellano and Eguiarte, 2003; Stone and Jenkins, 2008). Pollination regimes diff ering in composition and abundance of pollinators can in turn lead to diff erential reproductive success (Herrera et al., 2006; Pérez-Barrales et al., 2007), and in the specifi c context of the comparison between central and marginal populations, could lead to a higher reproductive success in central populations. In the present study, we compared central continuous and marginal disjunct populations of D. laureola within its distribution range in the south of the Iberian Peninsula. We used a broad array of population features and plant fi tness correlates, in addition to consideration of the main pollinator identity and abundance of co-fl owering species. Daphne laureola is an evergreen shrub with a wide distribution range within Europe, growing in the understory of mountain forests and presenting distribution patches on the main calcareous mountain ranges of southern Europe and north of Africa (Meusel et al., 1978). In our study area, most populations are gynodioecious with hermaphrodite and female plants coexisting within populations (Alonso et al., 2007). The main pollinator of D. laureola in the central area is a small pollen beetle, Meligethes elongatus Rosenhauer (Alonso, 2004), but its role as pollinator in marginal populations was uncertain. We analyzed population features such as population size, density of reproductive and non-reproductive individuals within population, identity of main pollinators and alternative fl oral resources for the pollinators, and individual life-history traits including plant size, shoot growth, fl ower and fruit production, fruit-set, and mortality. We further studied diff erences between females and hermaphrodites in size and reproductive traits presuming the existence of more stressful conditions in marginal populations that would allow to test the “sex-diff erential plasticity hypothesis” (Case and Ashman, 2007; Dorken and Mitchard, 2008). A considerable number of studies support that females are more abundant under stressful conditions because by investing in one only sex function, females Exploring local borders of distribution in the shrub Daphne laureola: individual and population traits. - 81 - have more available resources for other functions thus increasing their relative fi tness compared to hermaphrodites (reviewed in Ashman, 2006). However, according to the sex-diff erential plasticity hypothesis (Dorken and Mitchard, 2008), if hermaphrodites are plastic in their investment in female reproduction, they will allocate fewer resources to seed production under stressful conditions counteracting the females’ advantages. Specifi cally, we addressed the following questions: (1) do population size and density of individuals diff er between central continuous and marginal disjunct populations? (2) do plants in central continuous and marginal disjunct populations diff er in size, growth, fl ower and fruit production? (3) do the identity of main pollinator and the abundance and diversity of other co-fl owering species vary in central continuous and marginal disjunct populations? and (4) do hermaphrodite plants present a lower seed output than females in the marginal disjunct populations? Material and methods Study Species D. laureola L. (Thymelaeaceae) is an evergreen shrub with a disjunct Palaearctic distribution associated with the main calcareous mountain ranges of Europe, and north of Africa. In the Iberian Peninsula it shows a disjunct distribution, being abundant in the northern Cantabrian Range and the Pyrenees, and also in the southern Baetic Ranges (Alonso et al., 2007; Fig. 1a). This study was conducted in 2007-2009 in the Baetic Ranges. Plants of D. laureola present a variable number of branches that elongate over several consecutive growing seasons without any secondary branching. Leaves occur only at the distal end of branches, where they form a well-defi ned rosette, termed “leaf whorl” hereafter. It has a broad winter-fl owering period (January-April) and produces a large number of small, tubular, greenyellowish fl owers aggregated into several compact axillary infl orescences per stem. D. laureola presents gynodioecious populations where female and hermaphrodite individuals coexist, and populations having exclusively hermaphrodite plants, the last ones being rare in our study area (Alonso et al., 2007). Fructifi cation begins in June, presenting one only seed per fruit. Exploring local borders of distribution in the shrub Daphne laureola: individual and population traits. - 82 - Figure 1. (a) Distribution range of Daphne laureola in the Iberian Peninsula. (b)More detailed distribution of populations in the study area and location of central and marginal regions defi ned in this study. Intervals of altitude (m.a.s.l.) in the study area are represented with grey scale (see legend). Study area Across the Baetic Ranges, D. laureola is especially abundant in the Sierra de Cazorla, with at least 88 populations highly connected among them located along a broad altitudinal range. In fact, considering population as a discrete group of individuals distant from other groups of individuals by at least 1 km, most populations of Cazorla have > 60 neighbouring populations in a radius of 25 km (Castilla et al., 2011). The geographic isolation among population increases towards eastern and western regions of the Baetic Ranges, being isolation more marked in the western populations here analysed (Castilla et al., 2011). In particular, western populations of D. laureola have on average < 3 neighbouring populations in a radius of 25 km. Therefore, in this study we The main pollinator of D. laureola in populations of the Sierra de Cazorla, which constitute the center of abundance of the species within the studied southern Iberian region (Fig. 1b), is M. elongatus Rosenhauer, a small pollen beetle that moves frequently among fl owers of the same individual (Alonso, 2004). Exploring local borders of distribution in the shrub Daphne laureola: individual and population traits. - 83 - considered the Sierra de Cazorla as the local continuous center (“central region”, hereafter) and populations at the western edge as a local disjunct margin (“marginal region”, hereafter) of the species’ distribution patch in southern Iberian Peninsula (Fig. 1b). We studied three accessible populations per region. In the central region we used Valdecuevas, Cañada del Espino and Fuente Bermejo, the distance among them being 2-5 km. In the marginal region we used two diff erent mountain ranges. Two populations (Fuente Molina and Cañada de las Animas) were located in the Sierra de las Nieves, with a distance between them < 3 km. The third marginal population was located in the Sierra de Grazalema, distant from the other two populations about 35 km. All study populations were gynodioecious, and the proportion of female and hermaphrodite individuals determined during the fl owering period of 2007 by examining 100 randomly chosen individuals was ca. 20 % in all of them (Table 1). At the beginning of the study we haphazardly marked 25 hermaphrodite and 15 female plants in each study population, to avoid potential artefacts due to uncontrolled variation on sample sex ratio. Population Features: Size, Plant Density and Mortality of Individuals Populations were classed into one of the following size categories: <500, 500-1000, 1000-1500, 1500-2000, >2000 reproductive individuals. Densities of reproductive and non-reproductive individuals within each population were estimated using 20 randomly located circular areas of 10 m diameter in which the number of reproductive and non-reproductive individuals were counted. In 2009 summer, the mortality of marked individuals was recorded. Individual Features: Size, Annual Shoot Growth, Floral Display and Seed Production Four measures related to individual size were obtained for every marked plant: height of the tallest branch (“plant height” hereafter), total number of leaf whorls (“number of leaf whorls” hereafter) and the major and the minor diameter of the plant vertical projection, used to estimate the area of the plant assimilating it to an ellipse (“plant area” hereafter). Exploring local borders of distribution in the shrub Daphne laureola: individual and population traits. - 84 - Three measures related to fl oral display were also obtained for every marked plant: proportion of stems bearing fl owers; number of infl orescences/stem (based on counts of fi ve stems per plant) and number of fl owers/infl orescence (counted in one infl orescence per sampled stem). Total fl ower production per plant was then estimated as the number of fl owering stems x average number of infl orescences/stem x average number of fl owers/infl orescence. The annual shoot growth was estimated by measuring the shoot elongation at the end of the growing season in the fi ve stems per plant used for the fl oral display (“shoot growth” hereafter). In addition, the total number of fruits produced per individual was counted (“fruit production” hereafter) and the proportion of fruits per fl ower (“fruit set” hereafter) was estimated using the fi ve marked infl orescences. Main Pollinator and Alternative Floral Resources Previous studies showed that the pollen beetle M. elongatus (Nitidulidae) is the main pollinator of D. laureola in Sierra de Cazorla, determining important aspects of species’ reproductive biology (Alonso, 2004; Medrano et al., 2005). However, occasional observations on D. laureola fl owers during the fl owering period of 2007 (A.R.C. personal observation) indicated that pollen beetles were apparently absent from marginal populations, where several species of Hymenoptera, mainly Bombus terrestris L, Xylocopa violacea L. and Apis mellifera L., and another beetle Tachyporus nitidulus Fabricius (Staphylinidae) were observed (Figure 2). Therefore, in 2009 censuses were conducted to determine the abundance of M. elongatus in marginal populations of D. laureola. The censuses were done in sunny days. In each census, the observer was close to the plant and counted the number of M. elongatus individuals systematically by examining all the fl owering stems of the plant during a maximum interval of three minutes. The number of fl owering stems, the time spent and whether the plant was in sun or shade during the observation were recorded. Forty individuals per population were censused in all cases. In marginal populations, censuses were conducted in three diff erent dates covering all the fl owering period (early, peak and late fl owering) to exclude the possibility of a mismatch between the presence of M. elongatus and the D. laureola’s fl owering peak. For comparison within the same season, plants in central populations were similarly censused although only in a single date at fl owering peak. We censused the same Exploring local borders of distribution in the shrub Daphne laureola: individual and population traits. - 85 - number of individuals in marginal and central populations, investing 168 minutes in marginal populations during the fl owering peak and altogether 539 and 304 minutes in marginal and central populations, respectively. In addition, the abundance of other fl owering species which could potentially share pollinators with D. laureola was estimated by counting the number of fl owering individuals of each species within fi ve meters around every censused plant. Figure 2. Pollinator assemblage in marginal populations of Daphne laureola in southern Iberian Peninsula. (a) Bombus terrestris; (b) Xylocopa violacea; (c) Apis mellifera. a b c Data Analyses All statistical analyses were performed using the SAS statistical package (SAS, Institute 2002). In order to compare marginal and central populations, tests considering the eff ects of region (central vs. marginal) and sex (female vs. hermaphrodite) as fi xed factors and the eff ect of population as random, were conducted. Density of reproductive and non-reproductive individuals, plant area, plant height, number of leaf whorls, fl ower production, shoot growth, fruit production and fruit set were the response variables analysed separately. Fruit set was normally distributed and plant height was normalized after being log transformed. Both responses were analyzed by mixed models (Procedure MIXED). We analyzed all the other variables with generalized linear mixed models (Procedure GLIMMIX), using the negative binomial function. Diff erences among regions in the re- Exploring local borders of distribution in the shrub Daphne laureola: individual and population traits. - 86 - lationship between fl ower production and individual size (plant area and number of leaf whorls per plant) were studied with an analysis of covariance (ANCOVA). Flower production was the response variable, plant area, and number of leaf whorls were the explanatory variables and region was the categorical eff ect. In addition, we explored potential trade-off s between reproduction and shoot growth with an analysis of covariance, distinguishing shoot growth as the response variable, fl ower and fruit production as the explanatory variables and region as the categorical eff ect. Finally we used Fisher’s exact test to analyze diff erences in the frequency of dead individuals between central and marginal populations. Means are given ± 1 SE. Results Population Features: Size, Plant Density and Mortality of Individuals All the populations studied had more than 500 reproductive individuals (Table 1). The two largest populations occurred in the central region, with > 2000 reproductive plants. The two smallest populations, with 500-1000 reproductive individuals were found one in the central region and the other in the marginal region. Table 1. Population features of the six Daphne laureola populations studied. Population Coordinates Region Altitude (m.a.s.l.) Size (No. reproductive individuals) Female frequency (%) Valdecuevas 37º 54’ N 2º 52’ W Central 1575 >2000 25.5 Cañada del Espino 37º 55’ N 2º 53’ W Central 1513 500-1000 20 Fuente Bermejo 37º 55’ N 2º 50’ W Central 1380 >2000 21 Grazalema 36º 46’ N 5º 25’ W Marginal 1229 1500-2000 10 Fuente Molina 36º 41’ N 5º 1’ W Marginal 1380 1000-1500 20 Cañada de las Animas 36º 42’ N 5º 1’ W Marginal 1333 500-1000 18.5 Exploring local borders of distribution in the shrub Daphne laureola: individual and population traits. - 87 - There was marginally signifi cant variation among regions for plant density within population (F1,106 = 3.86, P = 0.05). Marginal populations were denser than central populations (0.09 ± 0.01 and 0.06 ± 0.01 individuals / 100 m2 respectively). Density of reproductive individuals did not diff er signifi cantly between regions (F1,118 = 2.48, P > 0.1; Fig. 2). However, density of non-reproductive individuals in marginal populations was larger than in central populations (F1,118 = 13.84, P < 0.001; Fig. 2). The percentage of dead individuals after two years was also larger in marginal than in central populations (17 % and 2 % respectively; Fisher’s exact test: P < 0.0001). Figure 3. Regional diff erences in the density of reproductive (a) and non-reproductive individuals of Daphne laureola based on data of three populations per region and 20 plots per populations. In each plot, reproductive and non-reproductive individuals were counted in a circular area with a diameter of fi ve meters. Bars represent the least-square means (± s.e.) Exploring local borders of distribution in the shrub Daphne laureola: individual and population traits. - 88 - Individual Features: Plant Size, Shoot Growth, Floral Display and Fruit Production Plants of central populations were larger than those growing in marginal populations. The average area of individuals in central populations was more than twice the area of plants in marginal populations (F1,232 = 17.65, P < 0.0001; Fig. 3a). Also the number of leaf whorls was more than three times larger in central populations (F1,232 = 16.92, P < 0.0001; Fig. 3b). Plants in marginal populations tended to be taller than those in central populations but their height averages did not diff er signifi cantly (84.72 ± 2.44 cm vs. 80.22 ± 2.28 cm; F1,232 = 0.71, P = 0.40). Shoot growth was signifi cantly larger in marginal populations (5.9 ± 0.2 cm vs. 3.9 ± 0.1 cm; F1,229 = 24.07, P < 0.0001). Also, shoot growth was negatively related to fl ower and fruit production (F1,235 = 36.21, P < 0.0001 and F1,235 = 9.01, P < 0.01, respectively). The ANCOVA revealed that negative relationships between growth and fl ower and fruit production were consistent in the two study regions (P = 0.28 and P = 0.29 for the interaction between region and fl ower and fruit production, respectively). Furthermore, there were not signifi cant diff erences between hermaphrodite and female plants in any size-related variable (P > 0.5 in all cases). Absence of diff erential size between sexes was consistent between the two study regions (P > 0.5 in the interaction region × sex for all size variables). Exploring local borders of distribution in the shrub Daphne laureola: individual and population traits. - 89 - Figure 4. Diff erences in average plant area (a), number of leaf whorls (b) and fl ower production per plant (c) in central and marginal populations of Daphne laureola, based on data of three populations per region and 40 individuals per population. Bars represent the least-square (+ s.e.) after accounting for sex (fi xed eff ect) and population (random eff ect) variation. Exploring local borders of distribution in the shrub Daphne laureola: individual and population traits. - 96 - adults and higher density of non-reproductive individuals in marginal populations would suggest a higher turnover in these populations. A higher demographic variability in marginal populations of plant species has been found by several studies (Lönn and Prentice, 2002; Moore, 2009; Nantel and Gagnon, 1999). However, more demographic analyses are required to determine if this is a common pattern in plants and to investigate the possible underlying factors. Individual Features: Plant Size, Annual Growth, Floral Display and Reproductive Output Adult plants of D. laureola vary widely in size and architecture through the Iberian Peninsula (Alonso et al., 2007). The present study found that in its southern distribution area, individuals of central populations were larger than those in marginal populations, which had as a direct consequence a huge variation in fl ower production. Plants in central populations produced one order of magnitude more fl owers per individual than in marginal populations, because they consistently produced more reproductive stems, more infl orescences per stem and more fl owers per infl orescence. These results suggested more optimal conditions for the species in the central locations, further supported by the steeper slope of the relationship between fl ower production and individual size in central populations. We must point out, however, that shoot growth was lower in central populations, likely showing a trade-off between reproduction and growth within the season with diff erent outcomes in central and marginal populations (Obeso, 2002 and references therein; but see also Knops et al., 2007). Shoot growth was negatively related to fl ower and fruit productions. Plants of central populations produced more fl owers and fruits than plants of marginal populations, which could have constrained more the shoot growth in individuals of central populations. Ecological stress in distribution borders can also result in diff erential reproduction between central and marginal populations (Hengeveld and Haeck, 1982; Brown, 1984; Lawton, 1993). We found that D. laureola individuals of central populations produced more fruits than those on marginal populations, associated with the diff erences in size and fl ower production mentioned above. The connection between the number and density of adult individuals in populations and their re- Exploring local borders of distribution in the shrub Daphne laureola: individual and population traits. - 97 - productive output was not, however, supported because central and marginal populations did not diff er in size or density of adult plants (see also Samis and Eckert, 2007). The expectation that population size and density of individuals are mainly determined by individual seed production, however, could be unrealistic in some circumstances. Likely there are other factors constraining seedling establishment, such as abundance and quality of available sites for establishment and further postdispersal factors that aff ect seed survival, germination and seedling survival (Harper, 1977; Clark et al., 2007) that could be more determinant for population size and density of individuals. Finally, according to the sex-diff erential plasticity hypothesis (Dorken and Mitchard, 2008), if the hermaphrodites are plastic in their investment on female reproduction they will allocate fewer resources to the seed production under more stressful conditions, and subsequently their relative female fi tness should change. Our results, however, indicated that hermaphrodite individuals did not produce less fruits than female individuals in D. laureola central or marginal populations, suggesting either the absence of ecological stress in both study regions or a similar response of the two sexes. Pollination Environment Given the synchronic fl owering of D. laureola individuals (Alonso, 2004), the much higher fl ower production recorded in central populations may have consequences for the interaction with pollinators, since fl oral display can infl uence both the attraction and the behaviour of the pollinators (Eckhart, 1991; Ashman and King, 2005). In addition, poorer pollinator faunas are expected in marginal populations (Moeller, 2006; Silva-Montellano and Eguiarte, 2003; Stone and Jenkins, 2008). Previous studies conducted in the central region documented both high selfi ng rates in hermaphrodites and strong inbreeding depression from seedling to reproductive stage in D. laureola populations characterized by individuals with large fl oral displays (Medrano et al., 2005) and the pollen beetle M. elongatus as the main pollinator (Alonso, 2004). The reduction of individual size and fl oral display in marginal populations could therefore have some consequences for hermaphrodite Exploring local borders of distribution in the shrub Daphne laureola: individual and population traits. - 98 - reproduction through seed production, associated with a decrease in geitonogamy (Klinkhamer and de Jong, 1993). The increased fruit-set recorded in marginal populations suggested that pollinator services might be improved in these sites. A possible explanation could be related to diff erences in the identity of the main pollinator in central and marginal populations. Pollinator censuses in 2009 confi rmed that M. elongatus was the main pollinator in the central populations studied, detecting its presence in all populations, with an average of 2 beetles/plant. In contrast, M. elongatus was not found in any census at the marginal populations studied, despite considerably higher and phenologically wider sampling eff ort than in central populations. Additionally, several species of Hymenoptera as Bombus terrestris, Xylocopa violacea and Apis mellifera were observed visiting fl owers of D. laureola at the marginal populations. Thus, the absence of a key pollinator that promotes geitonogamy as M. elongatus and the existence of other more dynamic visitors did not support the prediction of poorer pollinator faunas at marginal populations and suggested that the larger fruit set in marginal populations could be related to changes in the pollinator fauna potentially leading to reduced geitonogamy. A more exhaustive sampling eff ort is required to quantify the abundance of bees and bumblebees as pollinators of marginal populations of D. laureola and their consequences for realized mating system. Furthermore, the diversity of alternative fl oral resources for insects was higher in central populations, with up to six diff erent co-fl owering species in the same population. In this context, D. laureola fl owers appear to be a low-valued resource for most day-fl ying fl oral visitors in central populations, where previous studies and anecdotal observations showed that bumblebees and bees only exceptionally visited D. laureola fl owers and preferred visiting other coexisting fl owering species (Alonso, 2004). In contrast, pollinators in marginal populations scarcely have alternative fl oral resources, mainly Helleborus foetidus in low abundance, which could drive them to visit a less attractive species such as D. laureola. This shift of pollinators, together with the increased fruit set in marginal populations despite a lower fl oral display, suggest a more eff ective pollination environment in marginal populations. Interestingly, one of the central populations presented a very poor fl owering community, similar to marginal populations. Further studies in populations with poor winter co-fl owering communities could be useful to test if individuals of D. laureola are visited more frequently by bees and bumblebees in them. Exploring local borders of distribution in the shrub Daphne laureola: individual and population traits. - 99 - To summarize, marginal populations of D. laureola in the southern Iberian Peninsula did not diff er signifi cantly from central ones in size and density of individuals. In addition, the lower reproductive investment of marginal populations and their higher fruit set could be related to observed variation in plant-pollinator interactions, which could be relevant for the sexual polymorphism in this gynodioecious species if levels of geitonogamy were reduced in marginal populations. Diff erences in reproductive success between central and marginal populations of D. laureola seemed therefore more closely related to variation in biotic selective pressures than to changes in the species’ abiotic niche. Additional work is currently underway focusing on the eff ects of changes in the pollinator community on fl oral traits and the genetic structure of marginal and central populations, which may inform about the importance of other mechanisms potentially involved in the delimitation of distribution ranges (e.g. limited gene fl ow, unstable metapopulation dynamics). These studies should contribute to a better understanding of the evolutionary potential of marginal disjunct populations. Acknowledgments We are grateful to Mercedes Pérez for her enthusiastic support in the fi eld work. Mónica Medrano, José L. Garrido, Clara de Vega and two anonymous reviewers provided useful comments and criticisms on the manuscript. Raimundo Outerelo identifi ed Tachyporus nitidulus. Also we are grateful to Consejería de Medio Ambiente, Junta de Andalucía for permission to work within Natural Parks and to the environment agents, especially to Manolo Calvillo, Paco and David, for facilities provided during fi eld work. This study was funded by the Spanish Ministerio de Educación y Ciencia through research project CGL2006-01355, the Consejo Superior de Investigaciones Científi cas (CSIC) through an I3P fellowship to A.C., and the Consejería de Innovación Ciencia y Empresa, Junta de Andalucía, through the research project RNM156-2005. This paper is dedicated to the memory of Miguel Blázquez. Exploring local borders of distribution in the shrub Daphne laureola: individual and population traits. - 100 - Castilla, A.R., Alonso, C. and Herrera, C.M. In review. Reduced gender divergence in fl oral traits and pollination success at disjunct marginal populations of a gynodioecious shrub. Chapter 4.- Reduced gender divergence in fl oral traits and pollination success at disjunct marginal populations of a gynodioecious shrub Abstract: In sexually polymorphic plant species the extent of gender divergence in fl oral morphology and phenology are expected to be infl uenced by gender diff erences in pollination success and phenotypic selection by pollinators. Margins of distribution constitute natural areas where changes in the pollinator fauna and thus variation in gender divergence of fl oral traits are expected. Here we indirectly assessed the eff ects of a change in the identity of main pollinator occurring in a disjunct group of marginal populations of the gynodioecious shrub Daphne laureola by analyzing gender relative pollination success and the gender-specifi c selection patterns of pollinators on fl oral traits and fl owering phenology. We found that pollinators deposited larger pollen loads on individuals of marginal populations, contrary to the expectation of less effi cient pollinator service on distribution margins, and diff erences between genders in the frequency of visited fl owers also vanished. Furthermore, hermaphrodites and females received similar quality pollen in marginal populations contrasting with core populations, where females received higher quality pollen than hermaphrodites. Pollinators selected longer fl owers only on hermaphrodite individuals of core populations. In addition, hermaphrodites of core populations with longer fl owers presented greater fruit sets and a marginally signifi cant negative association with a locus under purifying selection. Consistent with these selection patterns, the gender divergence in the longitude of fl ower tube was greater in core populations. On the other hand, pollinators selected also towards late fl owering on hermaphrodite individuals of marginal populations, being this phenological trait negatively related to another locus under purifying selection. However, a geographic variation in this phenological trait congruent with this phenotypic selection pattern was not observed. Our results supported that geographic variation in the composition of the pollinator faunas could lead to changes in both qualitative and quantitative components of pollination success and to diff erences in the gender-specifi c selection patterns by pollinators. These changes can have a relevant role in the geographic variation of the gender divergence of fl oral traits. Therefore, the results of the present study supported that marginal disjunct populations. Reduced gender divergence in fl oral traits and pollination success at disjunct marginal populations of a gynodioecious shrub constitute natural areas where changes in the evolutionary dynamics of plantpollinator interactions can be expected, especially in sexually polymorphic plant species. Resumen Las diferencias entre plantas de distinto sexo en el éxito de la polinización y en los patrones de selección fenotípica ejercidos por los polinizadores pueden determinar el grado de divergencia entre sexos en los rasgos florales y en la fenología de la floración en especies vegetales con sistemas sexuales polimórficos. Los márgenes de distribución constituyen áreas naturales donde se esperan cambios en la fauna de polinizadores y por tanto, donde se espera divergencia entre plantas de distinto sexo en los rasgos florales. En este estudio, nosotros hemos evaluado indirectamente el efecto del cambio en la identidad del polinizador principal en poblaciones marginales del arbusto ginodioico Daphne laureola. Para ello, hemos analizado las diferencias entre plantas de distinto sexo en el éxito de polinización y su relación con los rasgos florales y la fenología de floración. En este trabajo hemos encontrado que los polinizadores en las poblaciones marginales depositaron cargas polínicas mayores en los individuos, contrario a la expectativa de un servicio de polinización menos eficiente en los márgenes de distribución. Los polinizadores visitaron con similar frecuencia las flores de individuos hembras y hermafroditas. Los individuos hermafroditas y hembra recibieron polen de similar calidad en las poblaciones marginales en contraste con las poblaciones centrales, donde los individuos hembras recibieron polen de mejor calidad que los hermafroditas. Por otro lado, los polinizadores seleccionaron positivamente flores más largas sólo en los hermafroditas de las poblaciones centrales. Además, los hermafroditas de las poblaciones centrales con flores más largas también presentaron mayor proporción de flores produciendo frutos. La variación en la longitud del tubo floral estuvo relacionada negativamente, de manera marginal, con la presencia de un locus con signos de estar bajo selección purificadora. Consistente con estos patrones de selección, encontramos que la divergencia entre plantas de distinto sexo con respecto a la longitud del tubo floral fue mayor en las poblaciones centrales. En las poblaciones marginales, los polinizadores seleccionaron positivamente fenologías de floración más tempranas sólo en los individuos hermafroditas. Además la variación en este rasgo fenológico estuvo relacionada negativamente con otro locus con signos de selección purificadora. Sin embargo, la variación geográfica en este rasgo no fue congruente con el patrón de selección fenotípica observado. Nuestros resultados apoyaron la idea de que los cambios en la composición de la fauna de polinizadores dan lugar a variación en los componentes cuantitativo y cualitativo del éxito de la polinización; y a diferencias en los patrones de selección ejercidos por los polinizadores entre plantas de distinto sexo. Estos cambios podrían tener una especial relevancia sobre la variación geográfica de la divergencia de los rasgos florales entre plantas de distinto sexo. Por tanto, los resultados de este estudio señalan las poblaciones marginales disjuntas de las especies vegetales como áreas naturales importantes para el estudio de los cambios en las dinámicas coevolutivas de las interacciones planta-animal, especialmente en plantas con sistemas polimórficos sexuales. Introduction Geographic variation in the composition of the pollinator fauna in plant populations may lead to important changes in the pollination success of individuals through variation in quantity and/ or quality of pollen received (Herrera 1987; Herrera 2004; Gómez et al. 2010; Alonso et al. 2011). Variation in the pollination success may have a relevant role in the evolution of sexually polymorphic systems in plant species (Barrett 2002; Ashman 2006). For instance, divergence in pollination success between sex morphs in gynodioecious species can facilitate the spreading of female individuals through increased number and/or better quality of their offspring (Charlesworth & Charlesworth 1978). Many gynodioecious species present self-compatible hermaphrodites (Meagher 2007; Ehlers & Schierup 2008). As a consequence, higher quality seed production by females can result from the inability of that morph to self-fertilize, thereby limiting inbreeding in their offspring (e.g. Thompson & Tarayre 2000; Delph 2004; Chang 2007), that linked to strong inbreeding depression may contribute to female persistence within populations (Charlesworth & Charlesworth 1978; Medrano et al. 2005; Ramsey et al. 2006). Habitat fragmentation simultaneously affects a variety of abiotic and biotic factors, which directly or indirectly can change the abundance and composition of pollinators (Steffan-Dewenter & Tscharntke 1999; Knight et al. 2005; González-Varo et al. 2009). In this sense, ecological margins in plant distributions constitute areas where populations are geographically more isolated and, thus, where shifts towards poorer pollinator fauna may be expected (Silva-Montellano & Eguiarte 2003; Stone & Jenkins 2008). Different pollinator faunas may exert different patterns of phenotypic selection on floral or phenological traits (Herrera et al. 2006; Harder & Johnson 2009). This differential selection could lead to intraspecific phenotypic differentiation and ultimately to speciation events (Johnson 2006). However, few studies have analyzed specifically the causal role of the spatially variable selection by pollinators on the geographical differentiation in floral traits of plant species (Herrera et al. 2006), despite most of them have found a relevant role of divergent selection by pollinators on intraspecific floral differentiation of plant species (e.g. Totland 2001; Reduced gender divergence in fl oral traits and pollination success at disjunct marginal populations of a gynodioecious shrub - 104 - Herrera et al. 2006; Gómez & Perfectti 2010). Clarifi cation of the role of pollinators in the evolution of sexually polymorphic systems awaits studies that document if changes in pollinator fauna composition aff ect diff erently to individuals with diff erent sex in their pollination success and their phenotypic traits (Ashman 2006). Flowers of angiosperms with unisexual fl owers frequently exhibit dimorphism in secondary sex characters (Eckhart 1999), including gender diff erences in fl ower size, fl ower shape, nectar production, fl ower longevity, fl oral scent or fl owering phenology (Eckhart 1999; Asikainen & Mutikainen 2005; Collin & Shykoff 2010). In gynodioecious species, where hermaphrodite and female plants occur in populations, fl ower size is almost universally larger in hermaphrodites than females (Eckhart 1999; Shykoff et al. 2003; but see Delph & Lively 1992). Gender divergence has been also confi rmed in fl owering phenology of gynodieocious species. Female individuals frequently fl ower somewhat earlier than hermaphrodites (Uno 1982; Ashman & Stanton 1991; Asikainen & Mutikainen 2005). Both fl ower size and fl owering phenology may be largely infl uenced by pollinator’s selection patterns (Campbell 1989; Galen 1989; Johnston 1991; Conner 1996). Therefore, the levels of gender divergence in fl ower size and fl owering phenology are expected to vary in populations with diff erent pollinator faunas. In this study, we analyzed the geographic variation in pollination success and gender divergence in fl oral morphology and fl owering phenology of the long-lived gynodioecious shrub Daphne laureola. We compared core continuous and marginal disjunct populations in southern Iberian Peninsula which diff er in the identity of their main pollinator (Alonso 2004; Chapter 3). The persistence and frequency of female individuals in core continuous populations is accounted for a combination of extensive selfi ng in hermaphrodites and strong inbreeding depression in their off spring (Medrano et al. 2005). In addition, hermaphrodites of D. laureola present longer fl owers and start their fl owering earlier than females in core populations (Alonso 2004; Alonso 2005), but the level of spatial variation of gender divergence has not been analysed in a broader geographic context. In addition, a previous work has shown the existence of diff erent loci under purifying selection Reduced gender divergence in fl oral traits and pollination success at disjunct marginal populations of a gynodioecious shrub - 105 - between marginal disjunct and the remaining populations of southern Iberian Peninsula (Chapter 2). In the present study, we addressed specifi cally the following questions: (1) is the pollination success in its quantitative and qualitative components lower in marginal disjunct populations of D. laureola in southern Iberian Peninsula?, (2) is the level of gender divergence in fl oral and phenological traits diff erent in core and marginal disjunct populations?, (3) is there diff erential phenotypic selection on fl oral and phenological traits in continuous core and marginal disjunct populations of D. laureola?, (4) do patterns of phenotypic selection by pollinators match to geographic variation in the level of gender divergence in fl oral traits?, and (5) are fl oral and phenological traits associated to loci under purifying selection? Material and Methods Study Species Daphne laureola L. (Thymelaeaceae) is an evergreen shrub with a disjunct Palaearctic distribution associated with the main calcareous mountainous ranges of Europe and northern Africa. In the Iberian Peninsula, it shows a disjunct distribution, being abundant in the northern Cantabrian Range and the Pyrenees and also in the southern Baetic Ranges (Fig. 1a; Alonso et al. 2007). This study was conducted in the Baetic Ranges (Fig. 1b), where the species fl owers from January to April, a period characterized by low temperatures, frequent rains and irregular snowfalls. Each plant produces a large number of small, tubular, green-yellowish fl owers aggregated into several compact infl orescences per stem. The two sexes are similar in size, fl ower and fruit production (Alonso & Herrera 2001). Individual fl owers of both genders have a single ovule and remain open continuously during approximately one month. Flowers consist of a corolla tube (4-12 mm) with four lobes and bear eight stamens arranged in two whorls of four stamens each. Hermaphrodite plants are fully self-compatible, but fruit production requires fl ower visitation by pollinators and an excess of pollen may clog-up the stigma of hermaphrodites, reducing their fruit set (Alonso & Herrera 2001). The small beetle Meligethes elongatus Rosenhauer constitute the main pollinator in Reduced gender divergence in fl oral traits and pollination success at disjunct marginal populations of a gynodioecious shrub - 112 - Results Quantity and Quality Components of Pollination Success Hermaphrodite fl owers were quantitatively more successful than females but the magnitude of divergence between the two sexes was highly diff erent in the two regions (Fig. 2). Flowers of hermaphrodite individuals received larger mean pollen loads than females and the magnitude of the diff erence was higher in marginal populations (F1,1865 = 16.14, P < 0.0001 region x sex interaction; Figure 2a). Furthermore the diff erence between sexes in the mean percentage of visited fl owers was widely diff erent among regions (F1,227 = 89.20, P < 0.0001 region x sex interaction), because hermaphrodites had larger mean percentage of visited fl owers than females only in core populations (Fig. 2b). Figure 2. Variation in average pollination success (± SD/SE) between genders (hermaphrodites vs. females) and regions (core vs. marginal) of Daphne laureola in southern Iberian Peninsula. (a) (a) Pollen load refers to the average number of pollen grains deposited on the stigma. (b) Visited fl owers refer to the percentage of fl owers with at least one pollen grain recorded from a sample of 12 fl owers per plant. In both panels N = X females and XX hermaphrodites per region. Reduced gender divergence in fl oral traits and pollination success at disjunct marginal populations of a gynodioecious shrub - 113 - Piecewise analyses further indicated that visited female fl owers were more frequently limited by quantity than hermaphrodite fl owers (�22 = 114.39, P < 0.0001 and �22 = 1226.25, P < 0.0001, for core and marginal populations), although with some relevant diff erences between the two regions. In core populations quantitative limitation was more frequent in females but relatively common in the two sexes, with 64.6 and 36.8 % samples falling below the lower BCa. In marginal populations female fl owers were similarly aff ected by quantity limitation (66.75 %) but the frequency of quantity limitation in hermaphrodite fl owers was much lower (6.43 %). As regards limitation by quality, only 11.8 and 11.2 % of visited female fl owers fall above the higher BCa in core and marginal populations, respectively. However, 85.4 % of hermaphrodite fl owers in marginal populations were limited by quality, in contrast to 50.6 % recorded in core populations. Furthermore, in core populations the slope of the initial relationship (b1) was nearly twice as high in females as in hermaphrodites whereas in marginal populations it was similar between sexes and substantially lower in the females, thus, indicating a reduction in the quality of pollen received by females of marginal populations (Table 1 and Figure 3). Finally, the slope of the second relationship (b2) did not diff er from 0 in any case (Table 1 and Figure 3). This last result suggests that the styles were fully saturated with pollen tubes, although the number of tubes corresponding to the breakpoint was 16 in hermaphrodites and 6 in females of core populations, and 8 in hermaphrodites and 13 in females of marginal populations; always higher than the single ovule per fl ower. Region Sex b1 Breakpoint (BCa) b2 Core Female 0.55 11 (4.51-19.34) 0.12 Hermaphrodite 0.34 45.96 (37.06-55.37) -0.001 Marginal Female 0.26 50.52 (29.27-71.26) -0.058 Hermaphrodite 0.31 24.80 (19.74-43.34) 0.002 Table 3. Parameters of piecewise regression analysis of the relationship between numbers of pollen grains and pollen tubes for hermaphrodite and female individuals in core and marginal populations of Daphne laureola. The breakpoint indicates the number of pollen grains at which the slope of the relationship changes. Bootstrapping Confi dence intervals (BCa) were estimated with n = 1000 bootstraps. The slopes of the relationship before and after the breakpoint are represented by b1 and b2, respectively. Reduced gender divergence in fl oral traits and pollination success at disjunct marginal populations of a gynodioecious shrub - 114 - Figure 3. Piecewise regression of the relationship between the number of pollen grains and pollen tubes for females and hermaphrodites of D. laureola in core and marginal populations (solid and dashed lines respectively). For comparison, the four piecewise models are presented on a common scale. Reduced gender divergence in fl oral traits and pollination success at disjunct marginal populations of a gynodioecious shrub - 115 - Intraspecifi c Floral Traits Diff erentiation Hermaphrodite fl owers were longer and wider than female fl owers. Floral tubes were longer in hermaphrodites than females, although the magnitude of the gender divergence was lower at marginal populations (F1,2392 = 30.63, P < 0.0001; region x sex interaction; Fig. 4a). In particular, hermaphrodite fl owers of marginal populations had shorter tubes than those of core populations (Fig. 4a), whereas female fl owers were similar between regions (Fig. 4a). Flowers of hermaphrodites had greater mean outer diameter than those of females (F1,2321 = 174.08, P < 0.0001; Fig. 4b). The magnitude of the diff erence between both sexes was similar in core and marginal populations (P = 0.8; region x sex interaction; Fig. 4b). Flowers of hermaphrodites showed larger mean aperture of tube than those of females in both core and marginal populations (F1,2377 = 491.46, P < 0.0001; Fig. 4c). The average values of aperture of tube and the magnitude of the diff erence between both sexes was similar between core and marginal populations (F1,2377 = 2.22, P = 0.1; region x sex interaction). In addition, fl owers of the two sexes had greater mean aperture of tube in marginal than in core populations (F1,2377 = 52.95, P < 0.0001; Fig. 4c). Gender divergence in phenological traits was diff erent in core and marginal populations of D. laureola. Sexes diff ered in the time of fl owering in marginal populations, where females were delayed, but not in core populations (F1,232 = 25.66, P < 0.0001, interaction region x sex; Fig. 4d). Also, individuals in core populations had more open fl owers per infl orescence at fl owering peak than those of marginal populations. Within-plant asynchrony was higher in core populations and the sign of gender divergence changed between regions (F1,232 = 22.06, P < 0.0001; interaction region x sex). Hermaphrodites presented greater variation in the number of open fl owers during the fl owering peak in core populations, in contrast to the marginal populations were females were more variable (Fig. 4e). Reduced gender divergence in fl oral traits and pollination success at disjunct marginal populations of a gynodioecious shrub - 116 - Fig 4. Variation in average of fl oral phenotypic traits between genders (hermaphrodites vs. females) and regions (core vs. marginal) of Daphne laureola in southern Iberian Peninsula. Three morphological traits related to fl ower size were studied: (a) longitude of tube; (b) outer corolla diameter; and (c) aperture of tube. Additionally, two phenological traits were studied: percentage of open fl owers during fl owering peak, fl owering time; and coeffi cient of variation of the number of open fl owers during fl owering peak, fl owering asynchrony. Values plotted are the model-adjusted least-squares means (± SE). Phenotypic Selection Estimated From Pollination Success Only some fl oral traits of hermaphrodite individuals were infl uential for the mean number of pollen tubes recorded per plant, our estimate of individual relative pollination success. In core populations, mean number of pollen tubes was signifi cantly related to longitude of tube and aperture of tube in hermaphrodites, being the level of statistic signifi cance variable among populations (Table 2). In particular, the mean number of pollen tubes was positively related to longitude of tube in hermaphrodites of two core populations (CDE and VCV; Fig. 5). Furthermore, mean number of pollen tubes showed a marginally signifi cant positive relationship with aperture of tube in hermaphrodites of VCV population (t = 1.85, P = 0.07; b = 0.14 ± 0.08). However, there was not any signifi cant relationship in the other two core populations (P = 0.12 and P = 0.43 for hermaphrodites of CDE and FBJ respectively). Reduced gender divergence in fl oral traits and pollination success at disjunct marginal populations of a gynodioecious shrub - 117 - Core Hermaphrodites Core Females Marginal Hermaphrodites Marginal Females Eff ect d.f. F(62) Pb (SE) F(33) Pb (SE) F(62) Pb (SE) F(36) Pb (SE) LT 1 6.16 < 0.05 0.20 (0.08) 0.41 0.52 0.09 (0.14) 0.00 0.96 0.01 (0.09) 0.12 0.74 -0.04 (0.11) OD 1 0.98 0.33 0.05 (0.05) 0.07 0.79 -0.04 (0.14) 0.28 0.60 0.05 (0.10) 0.21 0.65 0.05 (0.12) AT 1 0.50 0.48 0.14 (0.08) 0.03 0.87 0.03 (0.15) 0.44 0.51 0.07 (0.10) 0.47 0.50 0.08 (0.11) FT 1 0.48 0.49 0.07 (0.11) 0.93 0.34 -0.25 (0.26) 5.30 < 0.05 -0.25 (0.11) 1.30 0.26 -0.14 (0.12) FA 1 0.02 0.90 0.01 (0.10) 1.12 0.30 -0.22 (0.21) 0.05 0.82 -0.03 (0.11) 2.16 0.15 0.16 (0.11) Pop 2 0.79 0.46 0.41 0.66 0.59 0.56 0.05 0.95 LT x Pop 2 3.31 < 0.05 AT x Pop 2 3.33 < 0.05 In addition, in marginal populations the pollination success was not related to fl oral morphology but it was larger in late fl owering hermaphrodite individuals (i.e. with lower fl owering time; Table 2). Neither fl oral morphological nor phenological traits were related to the mean number of pollen tubes in females of the two study regions (Table 2). Table 2. ANCOVA analyses comparing patterns of selection by pollinators on fl oral traits in females and hermaphrodites of core and marginal populations of Daphne laureola. Dependent variable was the relative mean number of pollen tubes. Independent variables were standardized to mean 0 and variance 1. Signifi cant relationships (P ≤ 0.05) are in boldface. LT, OD, AT, FT and FA refer to longitude of tube, outer diameter, aperture tube, fl owering time and fl owering asynchrony respectively. We fi rst ran the model with all of the interactions and then excluded the nonsignifi cant ones. Only linear selection gradients were considered (see Material and Methods). Results show the standardized selection coeffi cient (b) and its standard error mean (SE). Reduced gender divergence in fl oral traits and pollination success at disjunct marginal populations of a gynodioecious shrub - 118 - Fig 5. Standardized linear phenotypic selection gradients for longitude of fl ower tube of D. laureola, based on number of pollen tubes per style as an estimate of pollination success (upper panels) and fruit set as an estimate of the maternal component of female fi tness (bottom panels). Relative values of dependent variable, estimated by dividing the mean values per individual by the population means, were used. Independent variable was standardized to mean 1 and variance 0. Solid and dashed lines indicate signifi cant and non-signifi cant relationships respectively. Phenotypic Selection Estimated From Maternal Component of Fitness Again, phenotypic selection patterns were variable among sexes, populations and regions. In core populations, positive selection on longitude of tube variable among populations was detected in hermaphrodites (Table 3). As noted from pollination success (see previous section), hermaphrodites with greater fruit set presented also longer corolla tubes in CDE and VCV populations but not in FBJ (Fig. 5). Females of all core populations exhibited positive selection on outer diameter (Table 3). Also in females we detected selection on fl owering asynchrony that varied among populations (Table 3). Higher fruit set was positively related to fl owering asynchrony only in CDE (Fig. 6). In marginal populations, we did not detect any selection on phenotypic traits of hermaphrodite plants (Table 3). However, phenotypic selection on fl owering time and fl owering asynchrony Reduced gender divergence in fl oral traits and pollination success at disjunct marginal populations of a gynodioecious shrub - 119 - variable among populations in females of marginal populations was detected (Table 3). Fruit set was negatively related to fl owering time and fl owering asynchrony in females of GRZ population (Fig. 7). Any phenotypic selection on fl owering time was detected neither CDA nor FML (t = 0.19, P = 0.85 and t = -0.71, P = 0.48). Flowering asynchrony was not related to fruit set neither CDA nor FML (t = -0.23, P = 0.82 and t = -0.63, P = 0.53). Table 3. ANCOVA analyses comparing patterns of selection on fl oral traits in females and hermaphrodites of core and marginal populations of Daphne laureola. Dependent variable was relative fruit set, as an estimate of the maternal component of fi tness. Independent variables were standardized to mean 0 and variance 1. Signifi cant relationships (P ≤ 0.05 or better) are in boldface. LT, OD, AT, FT and FA refer to longitude of tube, outer diameter, aperture tube, fl owering time and fl owering asynchrony respectively. We fi rst ran the model with all of the interactions and then excluded the nonsignifi cant ones. Only linear selection gradients were considered (see Methods). Results show the standardized selection coeffi cient (b) and its standard error mean (SE). Core Hermaphrodites Core Females Marginal Hermaphrodites Marginal Females Eff ect d.f. F(64) Pb (SE) F(34) Pb (SE) F(62) Pb (SE) F(32) Pb (SE) LT 19.59 < 0.01 0.23 (0.11) 0.34 0.56 0.04 (0.08) 0.00 0.96 0.002 (0.05) 3.50 0.07 0.15 (0.08) OD 1 0.13 0.72 -0.02 (0.06) 8.49 < 0.01 0.24 (0.08) 0.11 0.74 -0.02 (0.05) 0.22 0.64 0.04 (0.09) AT 1 0.05 0.83 -0.01 (0.06) 2.58 0.12 0.17 (0.10) 0.32 0.57 0.03 (0.05) 0.65 0.42 -0.06 (0.08) FT 1 0.19 0.66 -0.06 (0.14) 3.03 0.09 0.41(0.24) 0.39 0.53 0.04 (0.06) 5.13 < 0.05 -0.56 (0.18) FA 1 0.03 0.87 -0.02 (0.14) 1.72 0.20 0.09 (0.21) 0.03 0.86 -0.01 (0.06) 12.84 < 0.01 -0.91 (0.20) Pop 2 1.39 0.26 0.25 0.78 0.34 0.71 2.59 0.09 LT x Pop 24.02 < 0.05 FT x Pop 23.34 < 0.05 FA x Pop 25.15 < 0.05 7.13 < 0.01 Reduced gender divergence in fl oral traits and pollination success at disjunct marginal populations of a gynodioecious shrub - 120 - Fig 6. Standardized linear phenotypic selection gradients for fl owering asynchrony (FA) in female individuals of core populations of Daphne laureola. Flowering asynchrony between infl orescences of the same individual was estimated through the coeffi cient of variation of the number of open fl owers per infl orescence (N = 5 per plant) during the population fl owering peak. Dependent variable was the relative fruit set, estimated by dividing the mean values per individual by the population means. Reduced gender divergence in fl oral traits and pollination success at disjunct marginal populations of a gynodioecious shrub - 121 - Fig 7. Standardized linear phenotypic selection gradients for fl owering time and fl owering asynchrony (FT and FA, respectively) in female individuals of the marginal population GRZ. Signifi cant relationships between these phenological variables and relative fruit set were only found in females of this marginal population (see results). Flowering time was estimated as the mean percentage of opened fl owers per infl orescence in each individual during the population fl owering peak. Flowering asynchrony was estimated through the variation coeffi cient of the number of opened fl owers per infl orescence in each individual during the population fl owering peak. Dependent variable was the relative fruit set, estimated by dividing the mean values per individual by the population means. Reduced gender divergence in fl oral traits and pollination success at disjunct marginal populations of a gynodioecious shrub - 128 - Acknowledgements We thank María del Mar Alonso and Rocío Requerey for their help in the measuring of fl owers and pollination success. Pilar Bazaga for assistance with AFLP analysis. L.F. Delph and T.-L. Ashman for helpful comments. This study was funded by the Spanish Ministerio de Educación y Ciencia through research project CGL 2006-01355/BOS, the Consejo Superior de Investigaciones Científi cas (CSIC) through an I3P fellowship to A.R.C. and the Consejería de Innovación Ciencia y Empresa, Junta de Andalucía, through the research project RNM156-2005. Castilla, A.R., Alonso, C. and Herrera, C.M. In review. Loose herbivory interactions in a long-lived understory plant: non-genetic basis for adult defoliation. Loose herbivory interactions in a long-lived understory plant: non-genetic basis for adult defoliation Chapter 5.- Abstract: Antagonist interactions are considered an important selective factor on plant species. Geographic variation in the phenotypic selection patterns by herbivores on plant traits could lead to intraspecifi c diff erentiation in plant traits relevant for plant-herbivore interaction. In the present study, we analyzed diff erences in the average herbivory level and its eff ect on maternal fi tness in core continuous and marginal disjunct populations of Daphne laureola. Furthermore, we investigated the existence of intraspecifi c diff erentiation in vegetative traits and its potential connection to diff erential selection by herbivores in both groups of populations. Genetic distance among individuals calculated from 225 polymorphic AFLP loci was related to diff erences in herbivory among individuals to investigate the potential genetic basis of the resistance or tolerance to herbivory. Our results did not support diff erential herbivory incidence but did a diff erential eff ect of herbivory on maternal fi tness between core continuous and marginal disjunct populations of D. laureola. However, diff erences in the herbivory level among individuals did not relate to genetic distance among them ruling out a genetic basis of resistance or tolerance to herbivory in D. laureola. Changes in the herbivore abundance because of spatial variation of abiotic conditions could explain diff erences in the level of herbivory among individuals. In addition, herbivores did not exert phenotypic selection consistent with the geographic variation in studied plant traits. Regional diff erences in the light environment of the understory because of changes in the composition of tree canopy could be related to regional diff erentiation of vegetative traits of D. laureola. Therefore, both among individual diff erences in the herbivory level within populations and geographic variation in vegetative traits of D. laureola seem to be consequence of environmental heterogeneity more than output of geographically diff erential selection by herbivores. Loose herbivory interactions in a long-lived understory plant: non-genetic basis for adult defoliation - 130 - Resumen Las interacciones antagonistas están consideradas un agente selectivo importante para las especies vegetales. La variación geográfi ca en los patrones de selección fenotípica de los herbívoros sobre rasgos fenotípicos de las plantas puede originar diferenciación intraespecífi ca en tales rasgos. En este estudio comparamos poblaciones centrales continuas y marginales disjuntas de Daphne laureola en relación a sus niveles medios de herbivoría y al efecto de ésta sobre la proporción de fl ores que producen frutos. Además investigamos la existencia de diferenciación intraespecífi ca en rasgos vegetativos y su potencial conexión con selección diferencial llevada a cabo por herbívoros en ambos grupos de poblaciones. Las distancias genéticas entre individuos fueron estimadas mediante el análisis de 225 AFLP loci polimórfi cos y relacionadas con las diferencias en el nivel de herbivoría de los individuos para investigar la posible base genética de la resistencia o tolerancia a la herbivoría. Nuestros resultados no apoyaron una incidencia diferencial pero sí un efecto diferencial sobre la proporción de fl ores que producen frutos en poblaciones centrales continuas y marginales disjuntas de D. laureola. Sin embargo, las diferencias en el nivel de herbivoría experimentado por los individuos no se relacionó con la distancia genética entre ellos descartando una base genética de la resistencia o tolerancia a la herbivoría en D. laureola. Cambios en la abundancia de los herbívoros debido a variación espacial de las condiciones abióticas dentro de las poblaciones de la planta podrían explicar las diferencias en los niveles de herbivoría entre individuos de una misma población. Además, los herbívoros no ejercieron selección fenotípica consistente con el patrón de diferenciación geográfi ca mostrado por los rasgos vegetativos estudiados. Diferencias regionales en el ambiente lumínico del sotobosque debido a cambios en la composición del dosel arbóreo podrían estar relacionadas con la diferenciación regional en los rasgos vegetativos de D. laureola. Por tanto, tanto las diferencias entre individuos dentro de las poblaciones en los niveles de herbivoría como la variación geográfi ca en los rasgos vegetativos de D. laureola parecen ser una consecuencia de la heterogeneidad ambiental más que un resultado de selección diferencial llevada a cabo por los herbívoros. Loose herbivory interactions in a long-lived understory plant: non-genetic basis for adult defoliation - 131 - Introduction Herbivory can reduce growth, fecundity and survival of plants (Crawley 1989; Strauss 1991). Geographic variation in plant-herbivore interactions is also commonly observed and thus both the magnitude and the eff ects of herbivory on plant populations should be geographically variable (Strauss and Zangerl 2002; Pennings and Silliman 2005; Adams and Zhang 2009). Spatial variation in the intensity of interactions is important because it establishes geographic selection mosaics where intraspecifi c diff erentiation processes could occur (Thompson 1994). Herbivore abundance can vary spatially and temporally because of abiotic (e.g. weather conditions) and biotic (e.g. abundance of their host plant) factors (Schoonhoven et al. 1998). Such variations can be spatially random but we can predict that several factors can lead to changes in plant-herbivore interactions associated with marginal populations (Maron and Chrone 2006; Gaston 2009). Marginal populations of plant species frequently present ecological and genetic peculiarities that could contribute to modify the relationships between plants and their herbivores (e.g. reduced genetic diversity could decrease the adaptative potential of marginal populations to herbivores). First, changes in herbivore fauna have been frequently reported in marginal populations of the host plant (Lewinsohn et al. 2005). Furthermore, the negative impact of herbivory on fi tness of individuals could be larger in marginal populations contributing to the maintenance of stable distributional limits (Bruelheide and Scheidel 1999; Maron and Chrone 2006). Therefore, margins of distribution emerge as natural locations where plant-herbivore interactions are expected to vary (Maron and Chrone 2006). However, few empirical studies have examined the infl uence of natural enemies at range limits (Gaston 2009). Furthermore, plant traits related to resistance to herbivory should be under diff erential selective pressure in a landscape where the herbivore-plant interactions vary spatially and temporarily (Zangerl and Berenbaum 2003; Muola et al. 2010a; Vergeer and Kunin 2011). Plant quality from the herbivore’s viewpoint depends on a wide variety of phenotypic traits encompassing from nutritio- Loose herbivory interactions in a long-lived understory plant: non-genetic basis for adult defoliation - 132 - nal and secondary chemicals to plant architectural traits. Variation in these traits can refl ect not only genetic variation among individuals, but also unevenly distributed resources (Karban 1992). Therefore, the analysis of the geographic variation in phenotypic traits related to herbivory, their connection to diff erential selection regimes by herbivores and the determination of the genetic or environmental control of such phenotypic variation is a matter of maximal relevance in evolutionary ecology. In the present study, we compared core continuous and marginal disjunct populations of the shrub Daphne laureola within its distribution range in the South of Iberian Peninsula. In this region, most of populations of the species are gynodioecious with female and hermaphrodite individuals coexisting within populations (Alonso et al. 2007). Four noctuid caterpillars constitute the main herbivores of the species in the study area (Alonso and Herrera 1996). Gender diff erences in resistance and tolerance to damage by herbivores has been reported in gynodioecious species. Previous study did not fi nd that D. laureola herbivores discriminate between leaves of female and hermaphrodite plants (Alonso 2003). However, marginal populations frequently present harsher conditions that could promote sex-diff erential resistance and tolerance to damage by herbivores (Ashman 2006; Kawecki 2008). Therefore, in the present study we analyze the eff ect of herbivory on plants with diff erent sex in both core continuous and marginal disjunct populations. Furthermore, intraspecifi c variation in defoliation levels has been related to plant architecture (Alonso and Herrera 1996) and leaf nutrient composition (Alonso and Herrera 2003). However, the infl uence of secondary metabolites on this plant-herbivore system is still not fully understood. Specifi cally we addressed the following questions: (1) do core continuous and marginal disjunct populations of D. laureola diff er in their level of herbivory damage? (2) does herbivory damage have stronger impact on fi tness in individuals of marginal disjunct populations? (3) is there diff erential phenotypic selection by herbivores in core continuous and marginal disjunct populations? and (4) do individuals genetically diff erentiated showed signifi cantly diff erent level of herbivory damage? Loose herbivory interactions in a long-lived understory plant: non-genetic basis for adult defoliation - 133 - Material and methods Study Species Daphne laureola L. (Thymelaeaceae) is a long-lived evergreen shrub with a disjunct Palaearctic distribution associated to the main calcareous mountainous ranges of Europe and north of Africa (Meusel et al. 1978). This study was conducted in 2007-2009 in the Baetic Ranges in southern Iberian Peninsula. The fl owering occurs in January-April shortly before new leaves are produced. D. laureola plants consist of a variable number of erect stems that rise at ground level from a common trunk. Leaves are found only at the distal end of each branch, forming a single, well-defi ned leaf whorl. Four polyphagus noctuid species are the main herbivores of the species in the study region (Trigonophora fl ammea Esper., Noctua janthe Bkh., Noctua fi mbriata Schreber, Pseudenargia ulicis Staud.). The period of activity of herbivores encompasses from April to mid-June. We only evaluated the importance of caterpillars as consumers of foliage, although they also consume fl owers and unripe fruits (Alonso and Herrera 1996). Study Area Across the Baetic Ranges, D. laureola presents a local continuous center in the Sierra de Cazorla with numerous and largely connected populations (Castilla et al. in press). The spatial isolation increases towards eastern and western margins, although the increase is sharper in western edge (Castilla et al. in press). Therefore, in this study we considered the Sierra de Cazorla as the local continuous center (“core region” hereafter) and populations at the western edge as a local disjunct margin (“marginal region” hereafter) of the species’ distribution patch in southern Iberian Peninsula. We studied three accessible populations per region (see Chapter 3). All study populations were gynodioecious, and the proportion of female and hermaphrodite individuals was determined during fl owering period of 2007 by examining 100 randomly chosen individuals. The frequency of females was ca. 20 % in all of them (see Chapter 3). At the beginning of the study, we haphazardly marked 25 hermaphrodite and 15 female plants in each study population. Loose herbivory interactions in a long-lived understory plant: non-genetic basis for adult defoliation - 134 - Plant Phenotypic and Genetic Traits In 2007, we measured the basal diameter, at ground level, of the supporting stem of each marked plant using a Mitutoyo digital calliper (basal diameter, hereafter). A previous study showed basal diameter is under negative phenotypic selection by herbivores in at least some populations of D. laureola in the core region (Alonso and Herrera 1996). In addition, we estimated proportion of fruits per fl ower using randomly marked fi ve infl orescences per individual (“fruit set” hereafter). In 2009 July, fi ve intact leafs (i.e. without signs of herbivory) were haphazardly removed with scissors of each marked plant, placed into sealed plastic bags and kept in a portable cooler to minimize water loss during their transport to the fi eld station. Afterwards, leaves were individually placed in small paper envelopes and dried immediately at ambient temperature in sealed containers with abundant silica gel. Dried leaves were individually weighted on an analytical balance Mettler Toledo PL 203.S and their area measured using a leaf area meter LI-3000C (LI-COR, Lincoln, USA). Then, dry weight and leaf area were used to calculate the specifi c leaf area mm2 g-1 for each leaf (SLA, hereafter). Dry leaves from the same plant (N = 5), were pooled into a single sample for chemical analyses and homogenised to a fi ne powder using a Retsch MM 301 mill. This fi ne powder was the starting material for the extraction procedure. 250 mg of fi ne powder of each individual were extracted with 100 ml of methanol/water (70:30, v/v) by agitation at room temperature for 24 hours. The extraction procedure was repeated twice for each individual. The mixtures were centrifugated at 4000 rpm, fi ltered through a Whatman fi lter paper and the fi ltrate was concentrated by evaporation of the solvent, using a rotatory evaporator (BÜCHI Rotavapor R-215, BÜCHI, Lausanne, Switzerland). The extract was then re-dissolved in 10 ml of methanol/water (80:20, v/v) Total phenolic concentration in plant extracts (Phenolic content, hereafter) were determined spectrophotometrically by the Folin-Ciocalteau assay (Singleton et al. 1999) using gallic acid as a standard. An aliquot of 20 µl diluted with 1580 µl water was mixed with 100 µl of Folin-Ciocalteau phenol reagent Loose herbivory interactions in a long-lived understory plant: non-genetic basis for adult defoliation - 135 - and allowed to react for 5 min. Afterwards, we added 300 µl of saturated Na2CO3 solution and allowed to stand for 120 min at room temperature before the absorbance of the reaction mixture was measured in triplicate at 765 nm. Phenolic content of the plant extracts was expressed as mg gallic acid equivalents per g of plant material. Moreover, we quantifi ed the concentration in leaves of the three most abundant coumarins in D. laureola using a subsample of 15 individuals (Alonso et al. 2005; Alonso et al. 2009). Phenolic content showed a strong positive correlation with concentration of the three coumarin compounds analyzed (Fig. 1). Figure 1. Relationship between total phenolic content and the three most abundant coumarins in leaves of D. laureola (N = 15). Pearson correlation was used. Loose herbivory interactions in a long-lived understory plant: non-genetic basis for adult defoliation - 136 - Amplifi ed fragment length polymorphism (AFLP) markers were used to characterize 240 study plants genetically. The AFLP analysis was performed essentially as originally described by Vos et al. (1995), with modifi cations involving the use of fl uorescent dye-labelled selective primers following Applied Biosystems (2005). Each plant was fi ngerprinted using four EcoRI + 3/MseI + 3 and four PstI + 2/MseI + 3 primer combinations. Fragment separation and detection were made using an ABI PRISM 3130xl DNA sequencer, and the presence/absence of each marker in each individual plant was scored manually by visualizing electrophoregrams with GeneMapper 3.7 software. Only fragments ≥ 150 base pairs in size were considered, as a way of reducing the potential impact of size homoplasy (Vekemans et al. 2002)., Details on primer combinations, number of markers, scoring error rates and levels of polymorphism can be found in Castilla et al. in press. Herbivory Incidence The proportion of leaf area removed by herbivores was quantifi ed in the leaf whorls of 20 % of total stems per plant, except for individuals with ≤ 12 stems, in which the proportion of leaf area removed was quantifi ed in all stems (Herbivory level, hereafter). Herbivory level of individuals was quantifi ed in two diff erent years (2007 and 2009). In each leaf whorl, each leave was classifi ed into one of 6 herbivory classes according to percentage of leaf area removed: 0, no signs of herbivory; 1, 1-5% area removed; 2, 6-25%; 3, 26-50%; 4, 51-75%; and 5, >75%. An overall estimate of herbivory level for each leaf whorl was obtained following the method proposed by Alonso and Herrera (1996). Finally we calculated the mean value of herbivoy level per individual averaging the herbivory values of all its leaf whorls. Data Analysis All statistical analyses were performed using SAS statistical package (SAS Institute 2002). Diff erences in the level of herbivory between regions and sexes were analyzed by generalized linear mixed models using the negative binomial function (Procedure GLIMMIX). Eff ects of region Loose herbivory interactions in a long-lived understory plant: non-genetic basis for adult defoliation - 137 - (core vs marginal), population and sex (hermaphrodite vs female) were treated as fi xed factors. Diff erences in the level of herbivory in 2007 and 2009 were analyzed with separated analyses. In addition, temporal variability in the level of herbivory of individuals was explored with spearman rank correlation. Diff erences between sexes and regions in basal diameter, phenolic content and SLA of individuals were analyzed by generalized linear mixed models (Procedure GLIMMIX) with normal error distribution for phenolic content and negative binomial distribution for both basal diameter and SLA. Sex, population and region eff ects were treated as fi xed. We studied the relationship between herbivory and fruit set in 2007, analyzing separately hermaphrodite and female individuals within each region, because it is expected that the eff ect of herbivory on fi tness varies between hermaphrodite and female individuals in gynodioecious species (Ashman 2006, and references therein). For each sex, we explored potential among population diff erences in the relationship between herbivory level and fruit set using an analysis of covariance (Procedure MIXED). Fruit set was the response variable, showing a normal distribution. Herbivory level was the continuous explanatory variable and population was the categorical eff ect. Phenotypic selection at each region was examined by Lande-Arnold selection gradient models (Lande and Arnold 1983). Within each region, we previously analyzed phenotypic selection patterns by herbivores on individuals of diff erent sex separately. However, the similarity between sexes in phenotypic selection analyses together with the absence of diff erences in the herbivory level and phenotypic traits lead us to analyzed hermaphrodite and female individuals of each region jointly to get larger statistical power in the analyses. Therefore, we conducted separated phenotypic selection analyses for individuals of core and marginal populations without tacking into account sex of individuals. We quantifi ed phenotypic selection only in 2009 because we only measured foliar traits (SLA and phenolic content) in that year and foliar traits frequently present among-year variation. We used relative values of the response variable, herbivory level (estima- Loose herbivory interactions in a long-lived understory plant: non-genetic basis for adult defoliation - 144 - Selection by Herbivores on Plant Traits According to the previous results, we could expect stronger phenotypic selection on individuals of marginal populations. In core populations plants with lower phenolic content presented marginally larger herbivory level, in contrast to marginal populations where the phenolic content had not any eff ect (Table 2). In marginal populations, herbivores also exerted phenotypic selection on basal diameter but this eff ect was variable among populations (Table 2). Individuals with larger basal diameter presented also larger herbivory level in FML population, but there were not any eff ect in the other two populations (Fig. 6). In core populations, there was not any relationship between herbivory level and basal diameter (Table 2). Finally, herbivores did not exert any phenotypic selection on SLA neither core nor marginal populations (Table 2). Core Marginal Eff ect d.f.F(108) Pb (SE) F(101) Pb (SE) BD 1 2.64 0.11 -0.18 (0.11) 0.71 0.40 -0.19 (0.12) PC 1 3.90 0.05 -0.21 (0.11) 2.71 0.10 -0.13 (0.08) SLA 1 0.11 0.74 -0.04 (0.11) 1.26 0.27 -0.09 (0.08) Pop 2 0.03 0.97 0.11 0.89 BD x Pop 27.31 < 0.01 Table 2. ANCOVA analyses comparing patterns of selection by herbivores on vegetative and foliar traits in core and marginal populations of D. laureola. Dependent variable was the level of herbivory. BD, PC and SLA represent basal diameter, phenolic content and specifi c leaf area respectively. Independent variables were standardized to mean 0 and variance 1. Signifi cant relationships (P < 0.05 or better) are in boldface. Marginally signifi cant relationships (0.05 ≤ P < 0.09) are in italics. Only linear selection gradients were considered (see Material and Methods). Results show the standardized selection coeffi cient (b) and its standard error mean (SE). Loose herbivory interactions in a long-lived understory plant: non-genetic basis for adult defoliation - 145 - Figure 6. Standardized linear phenotypic selection gradients for the basal diameter (BD) in individuals of marginal populations of D. laureola (FML, CDA, GRZ). Dependent variable was the relative fruit set, estimated by dividing the mean values per individual by the population means. Loose herbivory interactions in a long-lived understory plant: non-genetic basis for adult defoliation - 146 - Discussion Herbivory and Its Eff ect on Fitness in Core and Marginal Populations Antagonistic interactions are considered important agents limiting geographic distribution of species (Gomulkiewicz et al. 2000; Nuismer et al. 2000; Paterson et al. 2010). For instance, escape from predators or pathogens has been reported as a causal agent of geographic expansion in numerous invasive species (Maron and Vilá 2001; Torchin et al. 2003). In addition, a growing body of experimental literature demonstrating a key role of herbivory in the spatial structure of plant populations is emerging (Bruelheide and Scheidel 1999; Fine et al. 2004; Gómez et al. 2005; Miller et al. 2009). However, herbivory has not been traditionally considered an important factor limiting geographic distribution of plant species (Gaston 2009). Core populations of D. laureola presented larger herbivory level in 2009, whereas core and marginal populations did not diff er in the average herbivory level in 2007. Decrease of herbivory damage in marginal populations has been reported in other studies (Moore 2009; Vergeer and Kunin 2011). However, herbivory aff ected negatively to fruit set of individuals of both sexes in marginal populations, but did not in core populations. Interestingly, individuals of marginal populations had greater fruit set than those of core populations (Castilla et al. 2011a). Therefore, results of the present study give still more support to our hypothesis of a more effi cient pollination environment in marginal populations of D. laureola (Castilla et al. 2011a). Furthermore, we found low average levels of herbivory in populations of D. laureola, but the diff erences among individuals were temporarily consistent in line with previous investigations of the study species at other core populations (Alonso and Herrera 1996). Nevertheless, this result does not rule out a relevant role of D. laureola herbivores as selective agents because chronic levels of relatively insubstantial herbivory (that is, damage that results in relatively little tissue lost) can have signifi cant impacts on longlived plants over their lifetimes (Doak 1992; Miller et al. 2009). Loose herbivory interactions in a long-lived understory plant: non-genetic basis for adult defoliation - 147 - Selection by Herbivores in Core and Marginal Populations of D. laureola Herbivores frequently exert phenotypic selection on vegetative traits of plant species, although the intensity of such phenotypic selection frequently varies geographically (Gómez et al. 2005; Pennings et al. 2009; Muola et al. 2010a; Vergeer and Kunin 2011). This geographic variation in phenotypic selection could lead to intraspecifi c diff erentiation among plant populations in phenotypic traits important for such interactions (e.g. Herrera et al. 2006 for selection by pollinators). A rigorous demonstration of a relevant role of D. laureola’s herbivores as selective agents requires demonstrating that observed diff erences in the herbivory level among individuals have a genetic basis. Resistance and tolerance to herbivory are commonly genetically determined traits that show heritable genetic variation within species (e.g. Muola et al. 2010b; Kuzina et al. 2011; Evans et al. 2011). In model organisms and crop species, great strides have been made in understanding the evolution of plant defence, as well as expression changes in response to herbivory contrasting to non-model organisms where few studies have examined this issue (Anderson and Mitchell-Olds 2011). Our results did not support a predictable, and temporarily constant relationship between diff erences in the herbivory level among individuals and their genetic distance in wild populations of D. laureola. Furthermore, although all studied vegetative traits did diff er between core and marginal populations of D. laureola, herbivores did not promote such phenotypic diff erentiation through phenotypic selection. Herbivores exerted negative selection on the phenolic content of leaves in individuals of core populations, whereas herbivores did not exert any selection on this trait at marginal populations. This result seems to match with the geographic variation in the chemical composition of D. laureola populations, with an increase of the level of phenolic compounds of defence in populations where herbivores exert negative phenotypic selection on this trait. However, we must be cautious here because selection in core populations was only marginally signifi cant. On the other hand, neither SLA nor basal diameter geographic variations matched with the phenotypic selection patterns by herbivores in core and marginal populations. Therefore both diff erences in the level of herbivory among individuals within populations and regional diff erences in vegetative traits seem to be related to abiotic variation. On the one hand, Loose herbivory interactions in a long-lived understory plant: non-genetic basis for adult defoliation - 148 - abundance of terrestrial insect herbivores can be limited by abiotic conditions, particularly temperature and moisture (Kingsolver 1989; Hunter and Prince 1998; Ritchie 2000). Therefore, our results suggest that environmental heterogeneity among and within D. laureola populations lead to spatial variation in the abundance of herbivores, promoting diff erences in the herbivory level among individuals and populations. On the other hand, individuals of core populations presented lower average basal diameter and leafs with lower SLA but larger phenolic content than those individuals of marginal populations. Diff erences in SLA and phenolic content could be related to contrasting light environments in core and marginal populations of D. laureola. Plants growing in shaded environments frequently present leaves with larger SLA which are more effi cient in capturing a limiting light resource (Chen et al. 1996; Meziane and Shipley 1999; Barber and Marquis 2011). On the other hand, light exposure can alter leaf quality, especially in plants with carbonbased defences, resulting in plants with lower phenolic content in leaves in shaded environments (Nichols-Orions 1991; Dudt and Shure 1994; Barber and Marquis 2011). In core region, D. laureola grows in the understory of mixed forests characterized by shading heterogeneity (Alonso and Herrera 2008). In marginal areas, individuals of D. laureola are established in the shrub stratum of closed fi r forests, Abies pinsapo Boiss, characterized by more homogeneous and closed canopy (Arista 1995). Therefore, the larger SLA and lower phenolic content in leaves of individuals of marginal populations could be a consequence of living in more shaded environments. Therefore, results of the present study suggested that herbivores did not have a relevant role promoting intraspecifi c diff erentiation in vegetative traits of adult plants among D. laureola populations. However, we only have studied the eff ects of foliar herbivory on adult plants and some studies point out the relevance of herbivory diff ers among diff erent life stages in plant species. For instance, slugs also seem to have a relevant role in the survival of D. laureola seedlings (Alonso personal comm.), but we did not analyze their eff ect in the present study. In this sense, Bruelheide and Scheidel (1999) found that slug herbivory was an important factor limiting the geographical distribution of Arnica montana. On the other hand, foliar herbivores of D. laureola also consume fl owers and fruits but their eff ects are still waiting to be evaluated. Loose herbivory interactions in a long-lived understory plant: non-genetic basis for adult defoliation - 149 - Concluding Remarks Results of the present study showed important diff erences in the herbivory level among nearby populations more than clear regional divergence. Among population and among individual diff erences seem to be related to environmental heterogeneity determining spatial variation in the abundance of herbivores more than diff erences in the resistance of individual plants with a genetic basis. In addition, our results ruled out a regional diff erentiation in vegetative traits conducted by divergent phenotypic selection by herbivores in core and marginal populations of D. laureola. Rather contrasting light environments because of diff erences in the species composition of the tree canopy in core and marginal populations of D. laureola could be related to regional diff erences in foliar traits. In addition, sexes did diff er neither vegetative traits nor phenotypic selection by herbivores supporting the widely accepted issue of scarce gender diff erentiation of vegetative traits in contrast to reproductive traits in gynodioecious species. Acknowledgements We thank María del Mar Alonso for her invaluable help in the fi eld work. Isabel María García carried out the chemical analyses. This study was funded by the Spanish Ministerio de Educación y Ciencia through research project CGL 2006-01355/BOS, the Consejo Superior de Investigaciones Científi cas (CSIC) through an I3P fellowship to A.R.C. and the Consejería de Innovación Ciencia y Empresa, Junta de Andalucía, through the research project RNM156-2005. Geographic variation, interactions with animals and sexual dimorphism: glancing at range limits in a gynodioecious plant. - 151 - El estudio de la variación geográfi ca intraespecífi ca en rasgos relevantes para las interacciones planta-animal y su posible relación con una selección divergente efectuada por animales, es crucial para entender el vínculo entre micro y macroevolución en estos caracteres y el papel de las interacciones planta-animal en la evolución de los sistemas sexuales en plantas (Barrett et al. 2001; Zangerl & Berenbaum 2003; Herrera et al. 2006). La teoría del mosaico geográfi co de coevolución propuesta por Thompson (1994, 2005) y los modelos genéticos asociados a ella (Gomulkiewicz et al. 2000; Alleaume-Benharira et al. 2006) proporcionan un modelo teórico que permite el acercamiento al estudio del papel causal de los animales en la aparición y mantenimiento de diferencias intraespecífi cas en los rasgos fenotípicos importantes para las interacciones planta-animal (Vermeer et al. 2011). Sin embargo, este marco teórico no predice las áreas concretas donde el cambio de las dinámicas coevolutivas en las interacciones planta-animal sea esperable a priori. El objetivo general de esta Tesis Doctoral ha sido tratar de mostrar cómo los márgenes de distribución de las especies vegetales, debido a sus peculiaridades ecológicas y genéticas, constituyen áreas donde cabe esperar cambios en las interacciones planta-animal. La integración de información obtenida mediante muestreos de campo (cuantifi cación del éxito de polinización y reproducción, censos de polinizadores, estimas de herbivoría), técnicas moleculares (diversidad y diferenciación genética, detección de loci bajo selección natural) y medición de rasgos fenotípicos en laboratorio (morfología fl oral y foliar, fenología, análisis de la concentración de fenoles en hojas) han permitido inferir la importancia de polinizadores y herbívoros como agentes del cambio en los rasgos fenotípicos de D. laureola en márgenes de su distribución. La Estructuración Geográfi ca de las Poblaciones de D. laureola El análisis espacialmente explícito llevado a cabo con la mayor parte de las poblaciones de D. laureola existentes en las Cordilleras Béticas reveló que la estructuración geográfi ca de las po- Geographic variation, interactions with animals and sexual dimorphism: glancing at range limits in a gynodioecious plant. - 152 - blaciones varió hacia los márgenes de la distribución. Los resultados de este estudio mostraron un incremento del aislamiento espacial de las poblaciones hacia los márgenes de distribución de la especie en el área de estudio (Capítulo 1). Sin embargo, no hubo evidencias que apoyasen una reducción del tamaño poblacional hacia los márgenes de distribución (Capítulo 1). De hecho, las poblaciones marginales de D. laureola en la región de estudio se caracterizaron por presentar en muchos casos más de 100 individuos reproductivos. Estos resultados sugirieron la existencia de localidades ecológicamente óptimas para la especie en todo el área de estudio, aunque la abundancia de las mismas disminuyó hacia los márgenes de distribución (Eckert et al. 2008; Yakimowski and Eckert 2008; Sexton et al. 2009). Por otro lado, aunque el aislamiento espacial de las poblaciones aumentó tanto hacia el margen oriental como hacia el margen occidental, la magnitud del aislamiento no fue equivalente. Las poblaciones del margen occidental presentaron una nivel de aislamiento espacial mucho mayor que las poblaciones del margen oriental. El análisis de la estructura genética de las poblaciones demostró que estas diferencias en el aislamiento espacial de las poblaciones tienen importantes consecuencias sobre el patrón de diferenciación genética de las poblaciones (Capítulo 1). En esta Tesis Doctoral demostramos que la estructuración genética de las poblaciones puede ser diferente en distintos márgenes de distribución (Hampe and Petit 2005; Sexton et al. 2009). Las poblaciones béticas del margen oriental fueron genéticamente muy similares a las poblaciones centrales y además, constituyeron un grupo genéticamente homogéneo, constituyendo un límite en expansión (“expanding edge”). En cambio, las poblaciones disjuntas del margen occidental estuvieron genéticamente muy diferenciadas del resto de las poblaciones béticas y presentaron además una mayor heterogeneidad genética entre ellas, características típicas de un margen de distribución de retaguardia (“rear or trailing edge”). Esta distinción genética entre diferentes bordes de distribución de una especie vegetal ha sido propuesta a escala de la distribución geográfica global de las especies, donde la distancia geográfica entre bordes ocurre a escala continental (Hampe and Petit 2005). Sin embargo, nosotros hemos demostrado que la estructuración genética diferencial en distintos bordes de distribución puede producirse igualmente a escalas geográficas más reducidas, como son cada uno de los parches de distribución de las especies vegetales con distribuciones disjuntas. Este tipo de situaciones pueden ser especialmente Geographic variation, interactions with animals and sexual dimorphism: glancing at range limits in a gynodioecious plant. - 153 - frecuentes en sistemas montañosos que no se vieron afectados drásticamente por los períodos glaciares, como es el caso de las Cordilleras Béticas (Benett et al. 1991; Carrión 2002; Carrión et al. 2003; Médail & Diadema 2009). Este patrón diferencial de estructuración genética puede influir de manera diferente en las dinámicas planta-animal a través de la remezcla de caracteres vía flujo génico, deriva genética y dinámicas de extinción/colonización (Thompson 1994, 2005). La existencia de una intensa selección natural diferencial podría explicar la fuerte diferenciación genética entre las poblaciones disjuntas del margen occidental frente al resto de poblaciones. Por ello, analizamos los perfiles genéticos de los individuos siguiendo un enfoque genómico poblacional destinado a detectar la presencia de marcadores sujetos a selección entre los marcadores polimórficos usados (Capítulo 2). Nuestros resultados confirmaron la existencia de selección natural diferencial entre ambas regiones. Los loci que mostraron signos de selección natural en las poblaciones disjuntas del margen occidental no mostraron tales signos en el resto de poblaciones, y viceversa. Una vez detectados los loci outliers, se retiraron del análisis de la estructura genética de las poblaciones para discernir cuánto de la diferenciación genética entre los grupos disjuntos de poblaciones se podía atribuir a selección diferencial y cuánto a variación neutral afectada por efectos históricos de deriva y flujo génicos. La fuerte diferenciación genética entre las poblaciones disjuntas del margen occidental y el resto de poblaciones persistió después de retirar los loci outliers del análisis. Por tanto, nuestros resultados confirmaron que tanto la selección natural diferencial como el fuerte aislamiento genético parecen estar contribuyendo a la marcada diferenciación genética entre las poblaciones disjuntas del margen occidental y el resto de poblaciones béticas de D. laureola. El papel de la selección diferencial sobre plantas de distinto sexo en el mantenimiento del polimorfismo sexual en plantas ha sido demostrado en otras especies ginodioicas (Ashman 1999, 2003; Sakai et al. 2007). Nuestros resultados confirmaron un patrón diferencial entre individuos de distinto sexo en ambas regiones (Capítulo 2). Los individuos hermafroditas siempre presentaron un mayor número de loci con frecuencias alejadas del patrón neutral que los individuos hembra.