Molecular Phylogenies Link Rates of Evolution and Speciation
Molecular Phylogenies Link Rates of Evolution and Speciation
复制标题
DOI：
10.1126/science.1083202
复制
发表时间：
2003-07
期刊：
影响因子：
56.9
通讯作者：
A. J. Webster;R. Payne;M. Pagel
中科院分区：
文献类型：
--
作者：
A. J. Webster;R. Payne;M. Pagel
自Mayr(1)提出物种形成的创始人效应模型以来，进化生物学家一直在寻找遗传进化速度与物种形成之间的相关性；事实上，这种联系构成了间断平衡理论的基础。然而，迄今为止，物种形成（物种形成减去灭绝）的净速率与遗传变化之间的相关性很少（2,3），也没有对这种关系的普遍性作出估计。我们利用56个已发表的从基因序列数据中推断出的系统发生，将物种形成事件的净数量与潜在的遗传变化进行了比较(4)，我们估计这两者在大约30 - 50%的情况下是相关的。如果物种形成事件和遗传进化速率相关联，那么从树根到树尖的总遗传距离（“路径长度”x）将与中间物种形成事件的净数量（“节点”n）相关。在发生了许多物种形成事件的地方，应该有更多的总遗传变化（图1A）。无论是恒定的分子钟，还是与物种形成无关的可变进化速率，都不会显示出这种效应。我们使用所有系统发育信息并控制共同祖先来评估相关性(5)。对于每棵树，我们比较了物种形成事件数量独立于路径长度的模型（“随机行走”模型）与路径长度和物种形成事件数量相关的高速率（“定向”）模型的对数似然（4,5）。我们发现，在56个系统发生中，有28个物种形成事件与路径长度之间存在显著关联（50.7%，图1B），涉及多个分类群(4)。物种的非随机抽样可能会影响一棵树内物种形成的表观速率，从而导致已知的系统发育重建的伪产物：在树木中取样的物种数量较多的区域，将推断出更多的遗传变化(4)。作为x的函数，伪影对n产生单调增加和向上弯曲的关系。因此，它可以通过nx的图来评估，当伪影存在时，我们期望1(4)。使用对数似然比检验(4)，我们发现28棵树中的2棵中有1棵显著大于1。更保守的标准是从28棵树中排除任何数字为1的树。该标准鉴定出13棵树，剔除后剩下43棵；其中15个（34.8%）显示出效果（图1B）。在分类群和树的大小上没有明显的偏差(4)。一个显著的关联可能偶然出现在任何单一的树，但可以排除样本作为一个整体。在我们最保守的排除标准下，似然比值分布的KolmogorovSmirnov检验的P值为0.000005(4)。我们的研究结果表明，快速的遗传进化经常参与物种形成。间断的分子钟与Mayr(1)的设想是一致的，尽管在瓶颈(1)期间的极端遗传漂变(1)或入侵物种的适应压力(6)中，快速遗传进化的爆发在多大程度上是有争议的。或者，观察到的相关性可能是物种形成率更高、进化速度更快的谱系的结果。但是，谱系水平上物种形成率的变化意味着在更快的谱系中有更多的物种代表，因此预计会导致我们控制的系统发育重建人工制品。因此，我们认为，从我们过滤的数据集中，排除了带有伪影的树的结果最有可能归因于间断的分子钟。在某种程度上，表型和基因型是耦合的，我们的研究结果提供了与化石记录中报道的形态特征间断平衡的观察相对应的基因和合理的联系(7)。
Evolutionary biologists have sought a correlation between rates of genetic evolution and speciation ever since Mayr (1) proposed his founder-effect model of speciation; indeed this link formed the basis of the theory of punctuated equilibrium. Yet to date few correlations between net rates of speciation (speciation minus extinction) and genetic change have been demonstrated (2, 3), nor has an estimate of the generality of this relationship become available. We compared the net number of speciation events to underlying genetic change using 56 published phylogenies inferred from gene-sequence data (4 ), and we estimate that the two are correlated in approximately 30 to 50% of cases. If speciation events and rates of genetic evolution are linked, the total genetic distance from the root of a tree to its tips (“path length,” x) will be correlated with the net number of intervening speciation events (“nodes,” n). Where many speciation events have occurred, there should be more total genetic change (Fig. 1A). Neither a constant molecular clock nor variable rates of evolution uncorrelated with speciation would show this effect. We assessed the correlation using all phylogenetic information and controlling for shared ancestry (5). For each tree, we compared the log-likelihood of a model in which the number of speciation events is independent of path length (“random walk” model) with the log-likelihood of an elevated rates (“directional”) model in which path length and number of speciation events are correlated (4, 5). We found a significant association between speciation events and path length in 28 of the 56 phylogenies (50 7%, Fig. 1B), over a range of taxa (4 ). Nonrandom sampling of species might bias apparent rates of speciation within a tree, leading to a known artifact of phylogeny reconstruction: that more genetic change will be inferred in regions of the tree in which a greater number of species has been sampled (4 ). The artifact produces a monotonically increasing and upward-curving relationship for n as a function of x. Therefore, it can be assessed by a plot of n x , where we expect 1 when the artifact is present (4 ). Using a log-likelihood ratio test (4 ), we found significantly greater than 1 in 2 of the 28 trees. A more conservative criterion is to exclude from the 28 trees any with numerically 1. This criterion identified 13 trees, the exclusion of which leaves 43 trees; 15 of these (34.8 7%) show the effect (Fig. 1B). There was no obvious bias in taxa or in tree size (4 ). A significant association could arise by chance in any single tree, but can be ruled out for the sample as a whole. KolmogorovSmirnov tests of the distribution of likelihood ratio values yielded P values 0.000005 under our most conservative exclusion criterion (4 ). Our findings indicate that rapid genetic evolution frequently attends speciation. A punctuated molecular clock is consistent with Mayr’s (1) scenario, although to what degree bouts of rapid genetic evolution could arise from extreme genetic drift during bottlenecks (1) or from adaptive pressures on invasive species (6) remains contentious. Alternatively, the observed correlation could be a consequence of faster-evolving lineages with higher rates of speciation. But lineage-level variation in speciation rates would imply greater species representation in faster lineages, and thus is expected to lead to the phylogeny reconstruction artifact for which we controlled. Therefore, we believe that the results from our filtered data sets in which trees with the artifact were excluded are most likely attributable to a punctuated molecular clock. To the extent that phenotypes and genotypes are coupled, our results provide both a genetic counterpart to and a plausible link with the observations of punctuated equilibrium of morphological traits reported in the fossil record (7 ).