1 INTRODUCTION

Understanding the main drivers of species distribution patterns and the mechanisms of coexistence is the central goal of ecology. Competition for resources and other ecological interactions often lead to the divergence of clades into multiple niches and the appearance of novel traits (Gilbert et al., 2018; Rüber et al., 1999). Over the course of evolution, some taxonomic groups accumulate morphological and ecological variation among their constituent species, others produce more similar species, and others can show parallel evolution (Rüber et al., 1999; Sidlauskas, 2008). However, closely related taxa tend to show similarities in many characteristics, including morphological, trophic, reproductive, behavioural, or ecological traits, due to common ancestry (Harvey & Pagel, 1991; Kamilar & Cooper, 2013). This phylogenetic relatedness can be measured by the phylogenetic signal (PS), defined by Blomberg and Garland (2002) as the “tendency for related species to resemble each other more than they resemble species drawn at random from the tree”. Previous studies found that the PS varies substantially across trait types (Blomberg et al., 2003a; Freckleton et al., 2002; Kamilar & Cooper, 2013). Most but not all traits display significant PS, which tends to be strongest in morphological traits such as body size, intermediate in life-history and physiological traits, and low in behavioural traits (Blomberg et al., 2003a; Kamilar & Cooper, 2013). For instance, in primates, dietary traits and climatic niche were among the variables with lowest PS (Kamilar & Cooper, 2013). Comte et al. (2014) found that fish traits related to morphological attributes and trophic position showed stronger phylogenetic clustering than other reproductive and habitat use characteristics.

The retention of niche-related ecological traits over time, causing that closely related species are more ecologically similar than would be expected based on their phylogenetic relationships (Losos, 2008), is termed phylogenetic niche conservatism (PNC) and strong PS has often been interpreted as evidence of it (Wiens et al., 2010). Some degree of PNC is likely in many species and its understanding is important to inform potential responses to global warming or species introductions in new areas (Wiens et al., 2010; Wiens & Graham, 2005). For instance, species with little tolerance to encompass the new environmental conditions and with strong PNC must either migrate or go extinct, while species with more evolutionarily labile traits could potentially adapt (Holt, 1990; Wiens et al., 2010). Nevertheless, strong PS can result from PNC or from Brownian motion (BM)-like evolutionary change (e.g. due to genetic drift or randomly fluctuating natural selection) (Losos, 2008; Wiens et al., 2010). So, PS is seen as a necessary but insufficient condition for PNC (Losos, 2008) and their relationship is complex (Revell et al., 2008; Wiens et al., 2010). So far, there is no universal test for PNC (Wiens et al., 2010) but a recent, promising approach is to compare the relative fit of different evolutionary models to the data, including the BM model and models of stasis or stabilising selection such as Ornstein–Uhlenbeck (OU) models (Kozak & Wiens, 2010; Losos, 2008; Wiens et al., 2010). The BM model assumes that the correlation structure among trait values is proportional to the extent of shared ancestry for pairs of species (Felsenstein, 1973), and works reasonably well as a model of trait evolution (Beaulieu et al., 2012). The OU models incorporate both selection and drift and are more general than pure drift models based on BM (Butler & King, 2004). They have been proved useful in a variety of contexts as they can capture the heterogeneity in the evolutionary process (Beaulieu et al., 2012; Pennell et al., 2015). In fact, several OU models with different degrees of complexity have been proposed, allowing to translate hypotheses regarding evolution in different selective regimes into explicit models (see Beaulieu et al., 2012; Butler & King, 2004; Hansen, 1997).

If traits have PS, it is often useful to apply phylogenetic methods (i.e. comparative methods), which have become a standard ecological tool in recent decades (Losos, 2008). When used in combination with trait-based approaches, phylogenetic analysis can strengthen hypothesis testing and generate new insights (de Bello et al., 2015), as these methods account for the non-independence of species in statistical analyses due to shared evolutionary history (Felsenstein, 1985; Revell et al., 2008). For instance, it can prove useful to consider phylogeny when assessing evolutionary mechanisms underlying present trait-environment patterns (de Bello et al., 2015). However, non-phylogenetic analyses answer questions at different evolutionary scales (de Bello et al., 2015) and are also informative, particularly when well-resolved phylogenies are not available for study taxa (Losos, 1999). Comparing the results of comparative and non-phylogenetic analyses can also inform about the existence of PNC and thus, as a rule of thumb, it might be useful to apply both techniques to trait data (de Bello et al., 2015).

Elevational and longitudinal gradients (i.e. stream size or upstream–downstream) are well-studied in river ecosystems. Both spatial gradients covary and display variation in many environmental variables such as water temperature, stream flow, habitat features, and productivity (Jones et al., 2017; Vannote et al., 1980). They are also well known to shape fish communities, with changes in fish abundance, richness, species composition and traits (e.g. Cook et al., 2004). Several studies have already analysed the trait–environment relationships of freshwater fishes along these spatial gradients (Jones et al., 2017; Kennedy et al., 2003; Pease et al., 2012). For instance, some studies revealed that species from uppermost reaches have more fusiform bodies, larger egg sizes and longer spawning seasons, but smaller body sizes and smaller clutches than species from lower river reaches (Jones et al., 2017; Pease et al., 2012; Winemiller & Rose, 1992). Similarly, Kennedy et al. (2003) also revealed a significant intraspecific trait variability across the elevation gradient. However, few studies have addressed this issue from a phylogenetic perspective (Comte et al., 2014). Therefore, little is known about which traits or groups of traits (e.g., morphological, trophic, reproductive, and habitat-use traits) are the most conserved in inland fishes, and how their evolution correlates with their elevational and longitudinal distributions.

The Iberian Peninsula is well suited to study the evolutionary assembly of fish species and traits along spatial gradients because of its complex orography, diverse climate, and particular ichthyofauna. This region is a mountainous territory with a broad range of elevation rising from the sea level, over a large central plateau (Meseta Central) to the peaks of over 3000 m (e.g. Bayón & Vilà, 2019; Sabater et al., 2009). Moreover, Iberian freshwaters are inhabited by 68 native fish species, of which 41 are endemic and they have been subjected to very prolonged isolation and speciation processes (Doadrio, 2001), but they are also inhabited by 32 alien species, some of them widespread throughout the planet.

Our main objectives were: (1) to compare the PS of several morphological, trophic, habitat use, and reproductive traits in inland fishes (i.e. species from freshwater ecosystems, including diadromous fishes and a few marine species that enter rivers); and (2) to test for correlated evolution of these traits with elevational and longitudinal distribution (i.e. if traits and species niche tend to evolve together) under three models of niche evolution (i.e. BM, OU stasis, and OU trend models). We hypothesised that the majority of traits would show PS (Johnson & Stinchcombe, 2007) but its magnitude would vary among trait types (i.e. morphological, trophic, reproductive, or habitat use). Specifically, we predicted that fish body size and other morphological traits would show higher PS than reproductive or habitat traits as in other taxonomic groups (Blomberg et al., 2003a; Comte et al., 2014; Kamilar & Cooper, 2013). Finally, we hypothesised that fish traits would display correlated evolution with elevational and longitudinal gradients, since the functional trait composition of fish assemblages is known to change across the river continuum (Pease et al., 2012).

中文翻译：

---

内陆鱼类沿海拔和纵向梯度的系统发育信号和进化关系

1 简介

了解物种分布模式的主要驱动因素和共存机制是生态学的核心目标。资源竞争和其他生态相互作用往往导致进化枝分化为多个生态位和新特征的出现（Gilbert et al.,  2018 ; Rüber et al.,  1999）。在进化过程中，一些分类群在其组成物种之间积累了形态和生态变异，另一些则产生了更多相似的物种，而另一些则表现出平行进化（Rüber et al.,  1999 ; Sidlauskas,  2008）。然而，由于共同祖先，密切相关的分类群往往在许多特征上表现出相似性，包括形态、营养、生殖、行为或生态特征（Harvey & Pagel,  1991 ; Kamilar & Cooper,  2013）。这种系统发育相关性可以通过系统发育信号(PS) 来衡量，由 Blomberg 和 Garland ( 2002 ) 定义为“相关物种彼此相似的趋势，而不是它们与从树上随机抽取的物种相似的趋势”。先前的研究发现，PS 在不同性状类型之间存在很大差异（Blomberg 等，  2003a；Freckleton 等，  2002；Kamilar & Cooper，  2013）。大多数但并非所有性状都显示出显着的 PS，其往往在形态特征（如体型）中最强，在生活史和生理特征中居中，在行为特征中较低（Blomberg et al.,  2003a ; Kamilar & Cooper,  2013） . 例如，在灵长类动物中，饮食特征和气候生态位是 PS 最低的变量之一（Kamilar & Cooper，  2013 年）。孔特等人。( 2014 ) 发现与形态属性和营养位置相关的鱼类性状比其他生殖和栖息地利用特征表现出更强的系统发育聚类。

随着时间的推移，生态位相关生态特征的保留，导致密切相关的物种在生态上比基于它们的系统发育关系所预期的更相似（Losos，  2008 年），被称为系统发育生态位保守主义（PNC），强 PS 经常被解释为作为证据（Wiens 等人，  2010 年）。许多物种可能存在某种程度的 PNC，了解其对于全球变暖或新地区物种引进的潜在反应非常重要（Wiens 等，  2010；Wiens & Graham，  2005）。例如，对新环境条件几乎没有耐受性且具有强 PNC 的物种必须迁移或灭绝，而具有更多进化不稳定特征的物种可能会适应 (Holt,  1990 ; Wiens et al.,  2010 )。然而，强 PS 可能来自 PNC 或类似布朗运动 (BM) 的进化变化（例如，由于遗传漂移或随机波动的自然选择）（Losos，  2008 年；Wiens 等人，  2010 年）。因此，PS 被视为 PNC 的必要但不充分条件（Losos，  2008 年），它们之间的关系很复杂（Revell 等人，  2008 年；Wiens 等人，  2010 年））。到目前为止，还没有针对 PNC 的通用测试（Wiens 等人，  2010 年），但最近一种很有前途的方法是比较不同进化模型与数据的相对拟合，包括 BM 模型和停滞或稳定选择模型，例如作为 Ornstein-Uhlenbeck (OU) 模型 (Kozak & Wiens,  2010 ; Losos,  2008 ; Wiens et al.,  2010 )。BM 模型假设性状值之间的相关结构与物种对的共同祖先的程度成正比（Felsenstein，  1973 年），并且作为性状进化模型的效果相当好（Beaulieu 等人，  2012 年））。OU 模型结合了选择和漂移，比基于 BM 的纯漂移模型更通用（Butler & King,  2004）。它们已被证明在各种情况下都很有用，因为它们可以捕捉进化过程中的异质性（Beaulieu 等人，  2012 年；Pennell 等人，  2015 年）。事实上，已经提出了几种具有不同复杂程度的 OU 模型，允许将关于不同选择机制中进化的假设转化为显式模型（参见 Beaulieu 等人，  2012；Butler & King，  2004；Hansen，  1997）。

如果性状具有 PS，则应用系统发育方法（即比较方法）通常很有用，这已成为近几十年来的标准生态工具（Losos，  2008 年）。当与基于特征的方法结合使用时，系统发育分析可以加强假设检验并产生新的见解（de Bello 等人，  2015 年），因为这些方法在统计分析中解释了由于共享进化历史而导致物种的非独立性（ Felsenstein，  1985 年；Revell 等人，  2008 年）。例如，在评估当前特征-环境模式的进化机制时，考虑系统发育可能是有用的（de Bello et al.,  2015）。然而，非系统发育分析回答了不同进化尺度的问题（de Bello 等人，  2015 年）并且也提供了丰富的信息，特别是当研究分类群无法获得很好解决的系统发育时（Losos，  1999 年）。比较比较和非系统发育分析的结果也可以了解 PNC 的存在，因此，根据经验，将这两种技术应用于性状数据可能是有用的（de Bello 等人，  2015 年）。

海拔和纵向梯度（即河流大小或上游-下游）在河流生态系统中得到了充分研究。两个空间梯度在许多环境变量（例如水温、河流流量、栖息地特征和生产力）中都发生变化并显示变化（Jones 等人，  2017 年；Vannote 等人，  1980 年）。众所周知，它们可以塑造鱼类群落，改变鱼类的丰度、丰富度、物种组成和性状（例如 Cook 等，  2004）。一些研究已经分析了淡水鱼在这些空间梯度上的性状-环境关系（Jones 等人，  2017 年；肯尼迪等人，  2003 年；Pease 等人，  2012 年））。例如，一些研究表明，上游的物种比下游的物种具有更多的梭形体、更大的卵和更长的产卵季节，但体型更小，产卵期也更小（Jones 等人，  2017 年；Pease 等人，  2012 年；Winmiller & Rose，  1992 年）。同样，肯尼迪等人。( 2003 ) 还揭示了跨海拔梯度的显着种内性状变异性。然而，很少有研究从系统发育的角度解决这个问题（Comte et al.,  2014）。因此，关于哪些特征或特征组（例如，形态、营养、生殖和栖息地利用特征）在内陆鱼类中最为保守，以及它们的进化如何与它们的海拔和纵向分布相关，我们知之甚少。

伊比利亚半岛非常适合研究鱼类物种和性状沿空间梯度的进化组合，因为其地形复杂、气候多样，以及特殊的鱼类。该地区是一个多山地区，海拔范围从海平面上升到一个大的中央高原（Meseta Central）到超过 3000 米的山峰（例如 Bayón & Vilà，2019 年；Sabater 等人，  2009 年）。此外，伊比利亚淡水区居住着 68 种本地鱼类，其中 41 种是地方性鱼类，它们经历了非常长时间的隔离和物种形成过程（Doadrio，  2001 年），但它们也居住着 32 种外来物种，其中一些广泛分布于各地星球。

我们的主要目标是：（1）比较内陆鱼类（即来自淡水生态系统的物种，包括水生鱼类和一些进入河流的海洋物种）的几种形态、营养、栖息地利用和繁殖特征的 PS；(2) 在生态位进化的三种模型（即BM、OU 停滞和OU 趋势模型）下，检验这些性状与海拔和纵向分布（即性状和物种生态位是否倾向于一起进化）的相关进化。我们假设大多数性状会表现出 PS（Johnson & Stinchcombe,  2007) 但其大小会因性状类型（即形态、营养、生殖或栖息地使用）而异。具体来说，我们预测鱼类的体型和其他形态特征将显示出比其他分类群中的生殖或栖息地特征更高的 PS（Blomberg 等人，  2003a；Comte 等人，  2014 年；Kamilar & Cooper，  2013 年）。最后，我们假设鱼类性状会显示出与海拔和纵向梯度相关的进化，因为已知鱼类组合的功能性状组成会在河流连续体中发生变化（Pease et al.,  2012）。