兽研1. 引言
冠状病毒是遗传多样性最为丰富的RNA病毒之一，可感染广泛的哺乳动物和禽类宿主。其进化成功在很大程度上归因于基因组可塑性与适应性灵活性的结合，这得益于其大型RNA基因组、复杂的复制机制以及频繁的基因交换[1]。与许多主要以核苷酸替换驱动多样化的RNA病毒不同，冠状病毒可通过重组获得大量遗传变异，从而产生新的基因组结构并出现不同的进化谱系[2]。理解重组驱动的创新如何与功能和宿主相关的进化约束相平衡，对于阐明冠状病毒多样化和适应的机制至关重要。
α冠状病毒是冠状病毒科中一个尤其活跃的谱系，存在众多宿主相关多样化和种间基因交换的实例[3,4]。猫冠状病毒（FCoVs）是全球分布的α冠状病毒，为研究病毒在长期宿主种群中的进化提供了重要模型。FCoVs在全球家猫和野猫中广泛流行，尽管其宿主范围相对有限，却表现出显著的遗传多样性[5]。虽然大多数感染为无症状或与轻度肠道疾病相关，但在持续流行过程中积累的进化变化偶尔可促成猫传染性腹膜炎病毒（FIPV）的出现，后者是导致致死性全身性疾病的致病性变异株[6]。尽管在理解与FCoV致病性转变相关的分子决定因素方面已取得重大进展，但在种群水平上塑造FCoV多样性的进化过程仍不甚明确。
FCoV进化的一个显著特征是其与犬冠状病毒（CCoVs）的密切进化关系[7]。II型FCoVs的出现被归因于涉及CCoVs的历史重组事件，尤其是影响刺突基因的重组，这提供了α冠状病毒内跨物种基因交换的一个特征明确的例子[8]。据报道，FCoVs和CCoVs均利用氨肽酶N（APN/CD13）作为功能性受体，尽管受体使用可能因病毒类型和宿主物种而异[9,10,11]。这种共享的受体使用为刺突蛋白的跨物种兼容性提供了合理的生物学背景，但FCoV–CCoV重组发生的具体宿主和流行病学情况仍不清楚。这一进化历史凸显了重组引入外源基因组物质并产生新病毒谱系的能力。然而，冠状病毒基因组并非简单的可互换遗传片段的集合。重组区域必须在新基因组背景中发挥功能，并始终受到蛋白质功能、复制需求和宿主相互作用所施加的约束。因此，重组的进化后果不仅取决于遗传多样性的产生，还取决于随后的选择和基因组兼容性。
尽管FCoV全基因组序列的可获得性不断增加，但支配FCoV多样化的进化过程仍未被完全理解。以往研究主要集中于特定基因组区域，尤其是刺突基因，因为其在受体识别、宿主适应和致病性中发挥核心作用[12]。然而，需要全基因组视角来确定FCoV基因组的不同区域如何促成谱系出现和长期多样化。特别是，理解FCoV进化是由新基因组组件的偶发性获取主导，还是由已建立谱系内的持续适应主导，仍是一个重要的未解问题。
病毒进化受到重组之外多种相互作用力量的塑造[13]。在宿主体内长期流行过程中，病毒积累遗传变化，同时始终受到病毒生物学和宿主免疫所施加的选择压力的约束。这些进化约束可影响基因组组成、编码序列变异以及适应性变化在病毒基因间的分布。这些过程对冠状病毒尤为重要，其大型基因组编码的蛋白质具有从复制和基因组维持到宿主相互作用和免疫调节等多种功能。因此，不同基因组区域可能遵循不同的进化轨迹，一些区域表现出广泛的多样化，而另一些区域则保持进化上的约束[14]。
因此，FCoVs 为研究基因组创新与进化约束如何在漫长的进化时间尺度上相互作用以塑造冠状病毒多样性提供了机会[7,15]。尽管先前的研究已描述了单个重组 FCoV 毒株和特定的重组事件，但这些观察结果大多是在特定毒株、基因组区域或地理种群的背景下加以考虑的。因此，需要进行更广泛的全基因组评估，以确定这些个体进化事件如何共同促进 FCoVs 的长期多样化。这种方法可以揭示遗传交换、核苷酸水平的适应和选择压力是独立起作用还是共同作用，从而界定流行 FCoVs 的进化格局。
在本研究中，我们分析了全球五十多年来收集的 127 个完整 FCoV 基因组，以探究 FCoV 多样化背后的进化机制。我们并非侧重于识别单个进化事件，而是整合了系统发育重建、重组分析、基因水平进化比较、基因组组成分析和选择评估，以考察多种进化过程如何在基因组尺度上共同塑造 FCoV 的多样化。这些发现为甲型冠状病毒的进化动力学提供了见解，并增进了我们对冠状病毒基因组如何在保持功能完整性的同时实现多样化的理解。
2. 材料与方法
2.1. 序列收集与数据集准备
使用包括“猫冠状病毒”和“猫传染性腹膜炎病毒”在内的相关关键词组合，从 NCBI GenBank 数据库中检索猫冠状病毒的完整基因组序列。根据相关元数据和基因组注释对检索到的序列进行筛选，以构建代表自然发生的 FCoVs 的数据集。仅保留注释为完整或近乎完整基因组且序列完整性足以进行全基因组进化分析的序列。缺失区域较多或基因组覆盖不完整的序列被排除。被鉴定为实验室来源、合成、工程化、嵌合、疫苗相关或以其他方式人工生成的病毒的序列被排除。经过质量控制和人工整理后，保留了 127 个完整 FCoV 基因组用于后续进化分析。所有 127 个基因组的登录号见补充表 S1。最终数据集代表了从多个地理区域采样、时间跨度超过五十年的病毒，从而提供了 FCoV 多样性在时间和地理上的广泛覆盖。
2.2. 多序列比对与系统发育分析
使用 MAFFT（v7.518）并采用自动算法选择策略对 127 个完整 FCoV 基因组序列进行比对[16]。随后使用 trimAl（v1.5.0）以 0.05 的空位阈值对比对结果进行修剪。在超过 95% 的序列中存在空位的比对列被从比对中排除[17]。所得比对用于最大似然系统发育重建。
使用 IQ-TREE 2 重建最大似然（ML）系统发育树[18]。对于全基因组系统发育分析，基于贝叶斯信息准则（BIC）使用集成的 ModelFinder 模块确定最佳核苷酸替换模型，鉴定出 UNREST + FO + R5 为最佳拟合模型。随后在该模型下推断 ML 树，并使用 1000 次自举重复评估分支支持度。
2.3. 重组分析
在全基因组和单个基因水平上研究了重组。进行全基因组分析以表征潜在的镶嵌基因组结构并识别推定的重组断点，而进行基因水平分析以评估重组信号在 FCoV 基因组中的分布。
使用 SimPlot（v3.5.1）研究全基因组水平上的潜在重组事件[19]。将来自 GIIa 和 GIIb 谱系的代表性 FCoV 基因组（分别包括 PQ192636.1 和 PQ133177.1）与代表 GI 来源 FCoVs 和犬冠状病毒的候选亲本基因组进行比较。候选亲本序列根据其与查询基因组的系统发育关系和序列相似性进行选择。
使用滑动窗口方法在整个基因组上计算每个查询基因组与候选亲本序列之间的成对核苷酸序列相似性。应用 500 个核苷酸的窗口大小和 50 个核苷酸的步长。沿基因组检查相似性图谱，以识别查询序列与不同候选亲本基因组相对相似性的变化。以主要相似性模式的突然转变来识别推定的重组区域并估计重组断点的大致位置。随后将推断的重组区域映射到基因组注释上，以确定相应的基因组和编码区。
对所有24个已注释的FCoV基因分别进行了基因特异性重组分析。对于每个基因，从全部127个完整FCoV基因组中提取相应的核苷酸序列，生成24个独立的基因特异性数据集，每个数据集包含127条序列。在进行重组分析之前，每个基因特异性数据集均使用MAFFT独立比对。
使用重组检测程序第5版（RDP5）[20]对所得比对结果进行分析。应用了RDP5中实现的七种互补的重组检测方法，包括RDP、GENECONV、BootScan、MaxChi、Chimaera、SiScan和3Seq。除另有说明外均使用默认设置。为提高重组检测的稳健性并减少方法特异性假阳性信号的影响，采用了严格的共识标准，即仅当重组事件得到七种方法中至少五种的支持时才被视为可靠。该标准要求多数独立检测方法的一致支持，而非依赖单一算法。对于每个基因，记录根据该标准鉴定为重组体的序列，并计算重组阳性序列的数量，以表征重组在FCoV基因组中沿基因分布的基因特异性特征。
对24个基因鉴定出的重组阳性序列数量进行汇总，并使用R（v4.3.1）中的ggplot2包可视化为热图。所得热图用于比较各FCoV基因之间重组阳性序列的相对分布。
为独立评估每个基因比对是否包含统计学显著的重组证据，使用SplitsTree（v6.3.30）[21]中实现的成对同塑性指数（PHI）检验对相同的24个基因特异性比对进行分析。PHI检验评估核苷酸位点之间的不相容性，并为比对序列数据集内的重组信号提供统计检验。p值<0.05被认为具有统计学显著性，并解释为相应基因内存在可检测重组信号的证据。PHI检验结果与RDP5结果一并评估，为FCoV基因组中的基因特异性重组提供互补证据。
2.4. 单倍型网络分析
对膜（M）基因和非结构蛋白10（NSP10）基因进行了单倍型分析，以表征FCoV之间基因特异性的遗传多样性和进化关系。选择这两个基因是因为它们在基因水平重组分析中均未显示可检测的重组证据，从而提供了主要基于核苷酸变异评估进化关系的非重组基因组区域。
对于每个基因，使用DnaSP（v6.12.03）从所得比对中鉴定独特的核苷酸单倍型，并根据携带相应核苷酸模式的序列数量计算每个单倍型的频率[22]。随后将鉴定出的单倍型导入PopART（v1.7）进行网络构建。
使用PopART中实现的邻接（NJ）法推断单倍型关系。所得单倍型网络经可视化以表示独特单倍型之间的遗传关系，其中每个节点对应一个单倍型，连接反映其相对遗传分化。根据全基因组系统发育谱系（GI、GIIa和GIIb）注释单倍型属性，以检查单倍型在主要FCoV谱系中的分布和遗传关系。
2.5. 二核苷酸组成和CpG相对丰度
对全部127个完整FCoV基因组的二核苷酸组成进行表征，以评估核苷酸使用模式和二核苷酸对的相对丰度。对于每个基因组，使用EMBOSS（v6.6.0）套件中实现的compseq程序计算所有16种可能二核苷酸的频率。字长设置为2，并根据相应基因组的核苷酸组成计算每种二核苷酸的观测频率和期望频率。
每种二核苷酸的相对丰度使用观测值与期望值（O/E）比值进行量化，计算为观测频率除以期望频率。O/E比值为1表示观测频率与基于 underlying 单核苷酸组成的期望频率一致，而低于或高于1的值分别表示相对匮乏或富集。对每个基因组独立计算所有16种二核苷酸的O/E比值，随后在GI、GIIa和GIIb谱系之间进行比较，以表征二核苷酸组成的谱系特异性模式和谱系内变异。
由于CpG在分析的所有基因组中表现出最强的相对耗竭，因此进一步研究了其时间动态。将每个基因组的CpG O/E比值与其对应的采样年份配对。使用R语言ggstatsplot包中的ggscatterstats函数进行Pearson相关分析，以评估CpG相对丰度与采样年份之间的关联。Pearson相关系数（r）及相应的p值用于量化时间关联的强度和统计显著性。仅对采样年份具有足够变异的谱系进行时间相关分析。
2.6. 选择压力分析
使用本研究纳入的127个FCoV全基因组序列（经序列整理后），评估了作用于FCoV蛋白编码基因的选择压力。从全基因组中提取了对应于24个注释FCoV基因各自的编码序列，并对每个基因独立进行密码子水平比对，以保持阅读框并确保同义和非同义替换的准确推断。
使用在Datamonkey网络平台上实现的快速无约束贝叶斯近似（FUBAR）方法对所得基因特异性密码子比对进行分析[23]。FUBAR通过在贝叶斯框架下估计位点特异性同义和非同义替换速率，用于鉴定受到普遍正选择或负选择的单个密码子位点。根据FUBAR实施的统计标准，密码子位点被认为是正选择。随后对24个蛋白编码基因中鉴定出的正选择位点进行汇总并在基因间进行比较，以表征FCoV基因组中选择压力的分布。
3. 结果
3.1. FCoV的进化关系与系统发育多样性
为研究猫冠状病毒之间的进化关系，使用127个全基因组序列重建了最大似然系统发育树。所得系统发育树解析出三个支持良好的分支，每个分支的自举值均为100，证明推断的进化关系具有稳健性（图1）。这三个簇被命名为GI、GIIa和GIIb。GIIa和GIIb谱系聚集在一起形成一个支持良好的单系群，自举值为100，表明它们彼此之间比与GI共享更近期的共同祖先（图1）。
[图：猫冠状病毒的全球系统发育关系与遗传多样性。最大似然系统发育树]
GI包含最多的FCoV基因组，包括1970年至2023年间采集的毒株。系统发育拓扑结构显示该谱系内无明显的地理聚集，表明不存在明显的地理结构（图1）。在相似采样期内，同一地区内鉴定出遗传上不同的病毒。例如，2017年前后在澳大利亚流行的病毒分布在GI内的不同位置，一个簇位于树的顶部，其他澳大利亚毒株占据更基部的分支（图1）。这一模式突显了同一地理区域内流行的FCoV具有相当大的遗传多样性。
在GI内，分支A由来自中国、泰国、荷兰、丹麦和英国的密切相关毒株组成，表明遗传相似的FCoV经历了广泛的国际传播（图1）。这一观察结果与FCoV在亚洲、欧洲、澳大利亚和美洲的广泛流行相一致。相比之下，分支B由分别于1988年、2015年和2024年采集的三株日本毒株组成（图1）。这些病毒形成一个支持良好的单系簇，表明密切相关的病毒谱系可在单一地理区域内维持数十年。总体而言，这些发现表明GI谱系经历了复杂的进化历史，既有广泛地理传播的证据，也有持续本地多样化的证据，导致时间和空间上存在显著的遗传异质性。
GIIa主要由2002年至2024年间采集的美国和中国的FCoV毒株组成。在该谱系内，分支C代表优势簇，其特征是极短的终端分支，表明流行毒株之间的遗传分化有限（图1）。尽管遗传相似度很高，分支C的成员已在多个国家被检测到，时间跨度超过二十年，表明其持续流行且进化多样化相对有限。
GIIb形成GIIa的姐妹谱系，主要由2023年在塞浦路斯采集的毒株组成，包括FCoV 23（图1）。与GIIa的分支C类似，大多数GIIb病毒也表现出非常短的终端分支，反映出该谱系内高度的基因组保守性。GIIb内观察到的高度相似的系统发育拓扑结构与密切相关的重组谱系的传播一致，而非反复发生的独立重组事件。
3.2. 重组塑造了FCoV的基因组多样性和进化历史
为探究重组对FCoV进化的贡献，在全基因组和单基因水平上开展了重组分析。选取GIIa和GIIb的代表性基因组以表征其重组模式。
对于GIIa，SimPlot分析鉴定出PQ192636.1为通过GI FCoV与犬冠状病毒重组产生的重组基因组（图2A）。大部分基因组骨架遗传自GI FCoV谱系，而大部分ORF1b连同完整的刺突基因及附属基因3a、3b和3c来源于犬冠状病毒（图2A）。尽管从犬冠状病毒获得了较大范围的基因组区域，但大多数基因组区域的序列相似性仍相对较高，而刺突基因的分化程度显著大于其余编码序列。该模式表明，在这些α冠状病毒的进化历史中，刺突基因积累的遗传变异远多于其他基因组区域。
[图：猫冠状病毒中的全基因组及基因特异性重组模式。（A）代表性序列的SimPlot分析]
在GIIb中鉴定出不同的重组模式。SimPlot分析显示，PQ133177.1同样源于GI FCoV与犬冠状病毒之间的重组，尽管重组断点与GIIa中观察到的不同（图2B）。在该基因组中，主要亲本骨架同样来源于GI FCoV，而大部分刺突基因连同编码NSP15和NSP16的ORF1b末端区域由犬冠状病毒贡献（图2B）。GIIa和GIIb中观察到的截然不同的重组架构表明，这两个谱系最可能源于涉及犬冠状病毒的独立重组事件，而非源于共同的重组祖先。
值得注意的是，刺突基因参与了产生GIIa和GIIb的重组事件，而两个谱系之间的主要区别在于ORF1b的基因组来源（图2）。该观察结果与系统发育关系一致，即GIIa和GIIb形成密切相关的姐妹谱系，同时通过不同的重组历史保留了独特的基因组架构（图1）。总体而言，这些发现表明FCoV与犬冠状病毒之间的反复重组在产生G2谱系的基因组多样性方面发挥了关键作用，同时维持了它们密切的进化关系。
为进一步表征FCoV基因组中的重组格局，使用SplitsTree连同RDP5中实现的七种重组检测算法对全部24个病毒基因进行了独立分析。仅由七种RDP5方法中至少五种支持的重组事件被视为可靠。联合分析揭示重组在整个FCoV基因组中广泛分布，并影响结构基因和非结构基因，尽管各基因间重组程度差异显著。
在所有基因中，NSP3的重组频率最高，鉴定出46条重组序列，其次是NSP12（31条序列）、NSP2（27条序列）和刺突基因（20条序列）（图2C）。与这些发现一致，这些基因的SplitsTree分析显示出广泛的网状网络结构，为频繁重组和复杂的进化关系提供了独立证据（图2C）。
相比之下，重组信号在基因组中分布不均。RDP5和SplitsTree均未在NSP1、NSP5、NSP7、NSP10或膜基因中检测到重组证据，表明重组并未对所有FCoV基因的进化历史做出同等贡献（图2C）。相反，重组表现出显著的基因特异性变异，若干基因经历频繁重组，而其他基因在其进化历史中保持高度保守。
综上所述，这些发现揭示了FCoV中重组的层级模式，即大基因组片段的反复交换促进谱系出现，而特定基因内的频繁重组进一步加速了病毒进化过程中的序列多样化。
3.3. 单倍型网络分析揭示FCoV中基因特异性进化模式
为进一步探究突变对FCoV进化的贡献，分别对代表结构基因和非结构基因的M基因和NSP10基因进行了单倍型网络分析，二者均未显示可检测到的重组证据（图2C）。
M基因从127条序列中产生了107个单倍型，未发现占主导地位的单倍型，反映出由累积突变驱动的广泛序列多样化（图3A）。与全基因组系统发育不同，M基因单倍型网络未能保留区分GI和GIIa的独立谱系结构。相反，GI由分布在整个网络中的众多遗传上分化的单倍型组成，表明谱系内存在大量多样性（图3A）。相比之下，GIIb单倍型形成一个遗传分化有限的紧凑簇（图3A）。尽管全基因组系统发育分析推断GIIa与GIIb之间存在密切的姐妹关系（图1），但GIIb膜单倍型与单个GI单倍型的关联比与GIIa单倍型的关联更密切（图3A）。此外，GIIa膜单倍型被划分为三个遗传上不同的组，各组之间被相对较大的突变距离分隔，提示该基因在该谱系内具有多个进化起源（图3A）。
[图：非重组FCoV基因的单倍型网络。（A）基于核苷酸序列构建的邻接法单倍型网络]
NSP10基因分析从相同的127个基因组中产生了86个单倍型，同样没有优势单倍型，证实了类似的高水平遗传多样性（图3B）。GI NSP10序列由集中在网络中心区域的众多密切相关的单倍型以及两个更为分化的外周簇所代表（图3B）。与膜基因一致，GIIa NSP10单倍型未被识别为独立谱系，而是散布在中心GI相关单倍型之间（图3B）。GIIb NSP10单倍型同样未能形成独立谱系，而是分布在外周GI相关簇中，同时彼此之间保持遗传相似性（图3B）。
值得注意的是，膜基因和NSP10单倍型网络均未能重现全基因组系统发育所推断的GIIa与GIIb之间的密切进化关系。相反，来自这两个谱系的单倍型经常显示与GI的关系比彼此之间更密切。总体而言，这些观察结果揭示了基因水平与全基因组水平进化关系之间的显著不一致，这与在FCoV基因组中发现的普遍重组现象一致。
3.4. FCoV进化过程中保守的二核苷酸组成偏好和渐进性CpG耗竭
为表征FCoV基因组中的二核苷酸组成，对每个基因组计算了16种二核苷酸各自的观测值与期望值之比。在无组成偏好情况下，观测值与期望值之比预计接近1。
在GI谱系中发现了明显的二核苷酸偏好。7种二核苷酸的观测值与期望值之比超过1，表现为过表达，其中TpG的平均比值最高，为1.38，其次是CpA为1.29，ApC为1.2（图4A）。相比之下，8种二核苷酸的观测值与期望值之比在0.75至1.00之间，表明存在不同程度的低表达（图4A）。尽管TpT的比值仍接近期望值，但TpC和TpA持续低表达。其中几种中度低表达的二核苷酸在GI基因组间表现出显著有限的变异，如箱线图中几乎重叠的四分位距所示，提示尽管序列持续进化，其相对频率仍保持高度稳定。
[图：FCoV中二核苷酸组成偏倚及CpG丰度的时间动态。（A–C）观测值与期望值（O/E）比值]
在所有二核苷酸中，CpG始终是低表达最显著的，平均观测值与期望值之比为0.51（图4A），而TpG是富集最显著的三核苷酸。CpG耗竭与TpG富集之间的显著对比凸显了FCoV基因组内二核苷酸使用的明显不平衡。此外，CpG在GI基因组间的变异性大于大多数其他二核苷酸，表明不同流行株之间CpG抑制的程度存在差异。
GIIa的整体二核苷酸组成与GI高度相似。8种二核苷酸过表达，其中TpG、CpA和ApC的观测值与期望值之比再次最高，分别为1.39、1.31和1.21（图4B）。CpG仍为低表达最显著的二核苷酸，平均比值为0.49（图4B）。然而，与GI相比，GIIa内的二核苷酸频率更为均一，在单个基因组之间仅检测到有限变异。
在GIIb中观察到类似的组成特征。约一半的二核苷酸过表达，其余则低表达。TpG、CpA和ApC再次为富集最显著的三核苷酸，平均观测值与期望值之比分别为1.39、1.34和1.24（图4C）。与GI和GIIa一致，CpG的观测值与期望值之比最低，为0.53（图4C）。尽管GIIa和GIIb具有重组起源，其二核苷酸组成仍与GI高度相似，表明二核苷酸使用的整体模式在α冠状病毒中保守，而非谱系特异性。
由于CpG在三个谱系中始终代表最耗竭的二核苷酸，因此进一步研究了其时间动态。在GI谱系内，CpG观察值与期望值的比值与采样年份呈显著负相关（r = −0.28，p = 7.12 × 10−3），表明CpG频率随时间逐渐降低（图4D）。在GIIa谱系中未检测到显著的时间关联（图4E），而未对GIIb进行相关性分析，因为所有可用的基因组均在同一年采集。综上所述，这些发现表明CpG抑制是FCoVs的一个保守基因组特征，并且CpG耗竭在GI谱系的整个进化历史中持续存在（图4）。
3.5. 不同的选择压力塑造了FCoV基因的进化
为了表征作用于FCoV基因的选择压力，使用Datamonkey平台中实现的FUBAR进行了基于密码子的选择分析。总体而言，这些分析揭示正选择在FCoV基因组中分布不均，并集中于一部分基因。
在所有基因中，NSP2和NSP3表现出最多的正选择密码子位点，FUBAR在每个基因中鉴定出十个位点（图5A）。这些结果表明NSP2和NSP3在FCoV进化过程中均经历了显著的正选择，突显了这两个基因作为FCoV基因组中选择压力的重要靶标。
[图：FCoV基因组中正选择的不均一模式。（A）正选择位点在基因组上的分布]
对于S基因，使用FUBAR将完整的FCoV刺突序列集作为单一数据集进行分析。七个密码子位点，对应于氨基酸位置D33、V65、S82、E219、K267、V425和E798，显示出显著的正选择证据（图5）。这些正选择位点分布在刺突蛋白的S1结构域中，表明选择压力并不局限于FCoV刺突蛋白的单一功能区域（图5）。
FUBAR还在另外11个FCoV基因中检测到正选择（图5A）。在这些基因中，基因7b包含最多的正选择密码子位点，鉴定出七个位点，其次是NSP12和M基因，分别有五个和四个位点。在NSP1、N基因、3a和3c中各自鉴定出三个正选择位点，而NSP6、NSP9、NSP14和NSP16各包含一个正选择位点。综上所述，这些结果表明正选择位点分布在FCoV基因组的多个结构基因和非结构基因中（图5A）。
总体而言，这些发现表明正选择在病毒进化过程中作用于相当比例的FCoV基因。值得注意的是，表现出最强正选择信号的基因在很大程度上与那些表现出频繁重组的基因重叠，将这些基因鉴定为FCoV基因组中的主要进化热点。这些基因中适应性替换的反复积累与病毒宿主相互作用期间施加的持续进化压力相一致。
4. 讨论
本研究提供了关于猫冠状病毒进化动力学的全基因组视角，并表明FCoV的多样化涉及在不同基因组尺度上运作的多种进化过程。在基因组水平上，同源重组与基因组结构的差异以及不同遗传谱系的出现相关。在基因水平上，序列变异和阵发性正选择促进了各个基因组区域的多样化。在更广泛的尺度上，尽管存在序列变异，保守基因组特征的持续存在表明存在长期的组织约束。综上所述，这些发现突显了FCoV进化过程中基因组多样化与进化约束的共存。
这种层级组织最明显的表现之一是全基因组与基因水平进化关系之间的显著不一致。全基因组系统发育分析一致解析出三个得到良好支持的谱系（图1），而从非重组基因重建的单倍型网络未能恢复相同的谱系关系（图3）。相反，来自不同基因组背景的基因通常显示出与祖先GI病毒比彼此之间更近的关系。尽管系统发育不一致通常被视为重组的证据，但我们的发现进一步表明它反映了猫冠状病毒进化的一个更基本特征：病毒基因组的不同组成部分可能遵循部分独立的进化轨迹。因此，冠状病毒基因组的进化历史不能总是从单个基因的历史中推断出来。
这一观察对冠状病毒的进化研究具有更广泛的意义。许多以往的研究依赖于单个基因，尤其是刺突基因，因其生物学重要性和序列数据的广泛可获取性，用以推断病毒的起源、传播历史或宿主适应性[8,24]。然而，冠状病毒以整合但模块化的基因组进行进化。各个基因组区域经历不同组合的重组、突变和选择约束，导致整个基因组呈现异质的进化历史。因此，基因树和基因组树不应被视为病毒进化的竞争性表示，而应被视为在不同生物学尺度上发生的进化过程的互补描述[25]。在我们的数据集中，全基因组系统发育关系与基因水平单倍型模式之间的不一致表明，不同基因组区域可能保留部分不同的进化历史。因此，全基因组分析和基于基因的分析可能为FCoV进化提供互补的视角，全基因组分析反映完整基因组背景之间的关系，而基因水平分析则突出单个基因组区域内的变异。
我们的结果进一步表明，突变和重组可能在猫冠状病毒多样化中发挥不同的进化作用。膜基因和NSP10基因内广泛的单倍型多样性表明，即使在缺乏可检测重组的基因组区域，核苷酸替换也会持续积累，表明突变过程在病毒长期流行过程中有助于遗传多样性的产生（图3）。然而，仅在这些基因中观察到的序列变异模式并不能完全重现从完整基因组推断的主要谱系关系。相反，全基因组系统发育和重组分析表明，重组与FCoV谱系间基因组构型的差异相关。综上所述，这些观察结果与突变和重组对FCoV多样化的互补贡献一致，但并未确定它们的相对贡献或两种过程之间的特定时间或因果关系。
上述提出的层级框架进一步表明，同源重组不仅仅是产生遗传多样性的机制，而是一个反复重组冠状病毒基因组架构的过程。冠状病毒在正链RNA病毒中拥有最高的同源重组频率之一，因为复制过程中的不连续RNA合成促进了病毒RNA依赖性RNA聚合酶的频繁模板切换[26,27]。因此，在单个复制周期内，密切相关的病毒之间可以交换大段基因组片段而非单个核苷酸。与点突变的逐渐积累相比，这种大规模遗传交换使FCoV能够在更短的进化时间尺度上探索新的基因组组合。
我们的分析表明，GIIa和GIIb谱系的出现与猫和犬冠状病毒之间反复而非单次的重组事件一致。尽管两个谱系共享涉及CCoV来源基因组区域的相似进化起源，但它们具有不同的重组架构和断点分布，表明它们是通过独立的进化事件产生的，而非从共同的重组祖先多样化而来。反复的种间重组同样已在其他几种α冠状病毒和β冠状病毒中被记录，突显了同源重组作为基因组创新的持续来源而非孤立的历史事件[28,29]。总之，这些观察结果表明，相关冠状病毒之间的遗传物质交换代表了一个持续的进化过程，能够反复产生新的病毒基因组背景。
这些发现的一个重要启示是，猫冠状病毒基因组似乎以相互作用的功能模块组装体的形式进化，而非不可分割的遗传单元。重组谱系中涵盖刺突基因和部分ORF1b的基因组区域的独立替换表明，某些基因组区域可以在保持剩余基因组功能的同时被交换。在其他冠状病毒中也观察到类似模式，其中与宿主相互作用相关的基因经常表现出与核心复制机器不同的进化历史[30,31]。这种模块化可能通过允许单个功能组件被替换或优化而无需同时修改整个基因组来提高进化灵活性。然而，FCoV基因组的模块化性质不应被解释为无限的可互换性。功能模块仍然嵌入在高度协调的复制网络中，成功的重组最终取决于新获得的基因组区域与预先存在的遗传背景之间的兼容性。
如果重组产生了新的基因组结构，一个同样重要的问题是为什么演化改变会反复集中在少数病毒基因上。我们的分析揭示，频繁发生重组的基因与经历反复正选择的基因之间存在显著重叠，尤其是NSP3、刺突基因、NSP2、NSP12和7b基因（图5）。这种重叠表明某些基因组区域经历了多种形式的演化变异，但并未确立重组与正选择之间的直接关系。换言之，FCoV的演化创新似乎集中在有限数量的功能模块内，这些模块介导病毒与宿主之间的关键相互作用。
这一解释与越来越多的证据一致，表明FCoV演化在整个基因组上高度异质。病毒蛋白在与宿主分子、免疫防御和变化的细胞环境相互作用程度上差异很大。位于病毒—宿主界面的蛋白预计会经历持续的选择压力，因为病毒在优化复制效率的同时逃避宿主抗病毒反应[32]。在此类条件下，任何能够提高病毒适应性的遗传改变——无论是通过突变还是重组引入——都有更高的概率被自然选择保留。相反，执行高度保守的结构或酶功能的蛋白受到更强的纯化选择，因为即使微小的序列改变也可能损害必要的病毒过程。因此，在FCoV基因组中观察到的异质演化格局可能反映的是功能约束的差异，而非内在突变率或重组率的差异。
在所有基因中，NSP3是这一演化原理最清晰的例证。NSP3不仅表现出最强的重组证据，而且包含本研究中鉴定出的最多正选择密码子。这种多功能蛋白在冠状病毒复制周期中占据核心位置，整合了蛋白水解加工、复制复合体组装、膜重塑以及多种先天免疫拮抗机制[33]。与专一于单一生物学功能的蛋白不同，NSP3持续暴露于来自病毒复制和宿主抗病毒防御的选择压力。因此，其异常的演化动力学可能反映了多种选择力量同时作用于单一蛋白的汇聚。
刺突蛋白的演化模式展示了一种互补但不同的适应策略。NSP3主要介导与宿主的细胞内相互作用，而刺突糖蛋白则定义了冠状病毒识别宿主受体并遭遇抗体介导免疫压力的细胞外界面。因此，正选择位点集中在S1亚基内——受体结合和抗原决定簇所在之处——而S2内的膜融合机制相对保守，这并不令人意外。值得注意的是，本研究中鉴定出的两个独立重组谱系均涉及刺突基因的替换，表明宿主相互作用模块的改造在FCoV演化过程中反复伴随谱系的出现。
在NSP2、NSP12和核衣壳蛋白中也检测到适应性演化，突显冠状病毒的适应远不止于受体识别（图5A）。尽管这些蛋白执行多样的生物学功能，它们都参与直接影响感染细胞内病毒适应性的过程，包括复制调控、基因组合成、细胞内组织以及宿主反应的调节[34]。因此，它们的适应性多样化强化了这一观点：成功的FCoV演化依赖于病毒生命周期多个组分的协调优化，而非任何单一蛋白的改造。总体而言，这些发现表明FCoV的适应性演化分布在一个相互连接的功能网络中，各个蛋白在不同的选择机制下演化，但共同促进病毒的适应。
重要的是，同一基因内反复重组与正选择的共存不应被解释为一个过程直接驱动另一个过程的证据。相反，这两个过程可能是同一潜在演化原理的结果：与宿主相互作用最强的基因同时也是演化创新最受青睐的基因。重组提供了获取替代遗传变异的途径，而正选择决定了这些变异中哪些连同新出现的突变能够赋予足够的适应性优势，从而在长期病毒传播中持续存在。这种重叠表明某些基因组区域经历了多种形式的演化变异，但并未确立重组与正选择之间的直接关系。
上述进化模式表明，FCoV中广泛的基因组创新伴随着同样普遍的进化约束。尽管重组反复重塑基因组结构，适应性进化持续修饰单个蛋白质，但若干基本基因组特征在病毒多样化过程中仍保持显著稳定。这一明显的悖论凸显了FCoV进化的一个重要特征：成功的适应并不需要不受限制的遗传改变，而是在长期功能和基因组约束所设定的边界内发生[30]。
在全部三个进化谱系中观察到的高度保守的二核苷酸组成体现了这些约束。尽管GI、GIIa和GIIb具有独立的进化历史和不同的重组结构，它们仍保留了几乎相同的二核苷酸使用模式，其特征为TpG和CpA富集以及CpG持续缺失（图4）。这些观察表明，仅靠大规模基因组交换不足以从根本上改变FCoV基因组的组成特性。相反，尽管初级核苷酸序列不断更替，基因组组成似乎仍保持显著韧性，这意味着组成性进化是在超越单个基因组区域遗传的约束下运行的。
在这些保守特征中，CpG缺失是脊椎动物RNA病毒最突出的基因组特征之一。尽管CpG频率降低被广泛解释为适应宿主抗病毒机制的证据，特别是ZAP介导的对富含CpG的病毒RNA的识别，但越来越多的证据表明，没有任何单一机制能充分解释这一现象[35,36]。突变偏倚、密码子使用、RNA结构需求、复制相关约束以及多种宿主选择压力都可能对长期CpG进化有所贡献。因此，在GI谱系内观察到的CpG逐渐减少与长期组成性变化一致，但应谨慎解释，因为可用序列在采样年份和地理区域上分布不均。特别是，在本数据集中时间和地理效应无法完全区分，观察到的CpG丰度与采样年份之间的负相关本身并不能确立纯粹的时间效应。尽管如此，观察到的模式与GI FCoV多样化过程中CpG组成逐渐变化的可能性相符。GIIa中缺乏类似的时间趋势应谨慎解释，考虑到其起源较晚、采样期较窄以及重组进化历史。
在FCoV中观察到的显著且持续的CpG缺失可能具有超出理解其进化历史的意义。在其他RNA病毒中，包括甲型流感病毒，实验性增加同义CpG二核苷酸已被证明可在保持编码蛋白序列的同时减弱病毒复制，凸显了CpG操作作为理性减毒潜在策略的价值[37,38]。类似方法是否可用于FCoV尚不清楚，但此处鉴定的保守CpG约束为研究同义CpG修饰作为FCoV理性减毒的潜在途径提供了进化依据。
这些基因组进化模式是否与FCoV感染向FIP的转变相关仍是一个重要但未解决的问题，因为本数据集可用的异质性表型信息不允许进行系统的表型分层分析。值得注意的是，先前的工作已将特定的重组FCoV–CCoV谱系与FIP暴发联系起来，凸显了基因组重组与致病表型之间的潜在相关性[7]。
综上所述，我们的发现支持一个FCoV进化的概念框架，即基因组多样化与进化约束同时发生。我们数据集中观察到的模式表明，重组与基因组结构变异相关，而突变和正选择促进单个基因的序列多样化。同时，尽管存在序列变异，保守基因组特征的持续存在提示了长期基因组约束的影响。尽管本研究中未定量评估这些过程之间的相对贡献和时间关系，但它们在不同基因组尺度上的对比模式为将FCoV多样化解释为基因组可塑性与进化约束之间的平衡提供了框架。需要进一步的比较和实验研究来确定这些过程如何影响病毒适应性、宿主适应和致病表型。
5. 结论
我们的研究结果揭示，猫冠状病毒的长期演化是由基因组创新与演化约束之间的动态平衡所塑造的。反复发生的重组，包括与犬冠状病毒的独立交换，促进了不同基因组谱系的出现，而突变和正选择则驱动单个基因内的持续多样化。这些过程尤其集中在NSP3和S等演化热点区域，凸显了复制相关蛋白和宿主互作蛋白在FCoV适应中的重要性。与此同时，主要谱系中CpG的持续耗竭及其随时间的逐渐下降表明，尽管存在广泛的序列更替和重组，更广泛的基因组约束仍然保持保守。总之，这些发现支持一个概念框架，即FCoV的多样化涉及基因组可塑性和演化约束，而非单一主导的演化过程。尽管本研究未定量评估重组、突变和选择之间的相对贡献及时间关系，但它们在不同基因组尺度上的对比模式为FCoV多样化提供了新的见解，并可能为未来猫及 related alphacoronaviruses的演化监测提供参考。
致谢
我们感谢扬州大学兽医学院高性能计算集群提供计算资源。
缩写
本手稿中使用以下缩写：
| FCoV | feline coronavirus |
| FIPV | feline infectious peritonitis virus |
| CCoV | canine coronavirus |
补充材料
以下支持信息可从以下网址下载：https://www.mdpi.com/article/10.3390/v18091037/s1，表S1：本研究中分析的FCoV基因组序列信息。
[内容请查看原文]
作者贡献
W.W.构思并设计了本研究。X.W.和Q.Z.进行了实验、分析了数据并起草了手稿。W.W.参与了手稿的修订。所有作者均已阅读并同意发表的手稿版本。
机构审查委员会声明
本研究使用公开可用的基因组数据，因此无需伦理批准。
知情同意声明
不适用。
数据可用性声明
本研究中呈现的原始贡献已包含在文章材料中。进一步咨询可联系通讯作者。
利益冲突
作者声明，他们没有已知的竞争性经济利益或个人关系，可能影响本文所报告的工作。
资助声明
本研究得到国家自然科学基金（32500538）、111计划D18007和江苏高校优势学科建设工程（PAPD）的资助。
脚注
[内容请查看原文]
参考文献
[内容请查看原文]
相关数据
补充材料
[内容请查看原文]
数据可用性声明
本研究中呈现的原始贡献已包含在文章材料中。进一步咨询可联系通讯作者。
查看英文原文▾
1. Introduction
Coronaviruses are among the most genetically diverse RNA viruses, infecting a broad range of mammalian and avian hosts. Their evolutionary success is largely attributed to a combination of genomic plasticity and adaptive flexibility, enabled by large RNA genomes, complex replication mechanisms, and frequent genetic exchange [1]. Unlike many RNA viruses whose diversification is primarily driven by nucleotide substitution, coronaviruses can acquire substantial genetic variation through recombination, allowing the generation of novel genomic architectures and the emergence of distinct evolutionary lineages [2]. Understanding how recombination-driven innovation is balanced by functional and host-associated evolutionary constraints is essential for elucidating the mechanisms underlying coronavirus diversification and adaptation.
Alphacoronaviruses represent a particularly dynamic lineage within the coronavirus family, with numerous examples of host-associated diversification and interspecies genetic exchange [3,4]. Feline coronaviruses (FCoVs) are globally distributed alphacoronaviruses that provide an important model for studying viral evolution within long-term host populations. FCoVs circulate extensively among domestic and wild felids worldwide and exhibit remarkable genetic diversity despite their relatively limited host range [5]. Although most infections are asymptomatic or associated with mild enteric disease, evolutionary changes accumulated during persistent circulation can occasionally contribute to the emergence of feline infectious peritonitis virus (FIPV), a pathogenic variant responsible for fatal systemic disease [6]. While substantial progress has been made in understanding the molecular determinants associated with FCoV pathogenic transition, the evolutionary processes shaping FCoV diversity at the population level remain less well defined.
A prominent feature of FCoV evolution is its close evolutionary relationship with canine coronaviruses (CCoVs) [7]. The emergence of type II FCoVs has been attributed to historical recombination events involving CCoVs, particularly affecting the spike gene, providing a well-characterized example of cross-species genetic exchange within alphacoronaviruses [8]. Both FCoVs and CCoVs have been reported to utilize aminopeptidase N (APN/CD13) as a functional receptor, although receptor usage can vary according to viral type and host species [9,10,11]. This shared receptor usage provides a plausible biological context for cross-species compatibility of spike proteins, but the specific host and epidemiological circumstances in which FCoV–CCoV recombination occurred remain unknown. This evolutionary history highlights the capacity of recombination to introduce foreign genomic material and generate new viral lineages. However, coronavirus genomes are not simply collections of interchangeable genetic segments. Recombined regions must function within new genomic contexts and remain subject to constraints imposed by protein function, replication requirements, and host interactions. Therefore, the evolutionary consequences of recombination depend not only on the generation of genetic diversity but also on subsequent selection and genomic compatibility.
Despite increasing availability of complete FCoV genome sequences, the evolutionary processes governing FCoV diversification remain incompletely understood. Previous studies have primarily focused on specific genomic regions, particularly the spike gene, because of its central role in receptor recognition, host adaptation, and pathogenicity [12]. However, a genome-wide perspective is required to determine how different regions of the FCoV genome contribute to lineage emergence and long-term diversification. In particular, understanding whether FCoV evolution is dominated by episodic acquisition of novel genomic components or by continuous adaptation within established lineages remains an important unresolved question.
Viral evolution is shaped by multiple interacting forces beyond recombination [13]. During long-term circulation within hosts, viruses accumulate genetic changes while remaining constrained by selective pressures imposed by viral biology and host immunity. These evolutionary constraints can influence genome composition, coding sequence variation, and the distribution of adaptive changes across viral genes. Such processes are especially relevant for coronaviruses, whose large genomes encode proteins with diverse functions ranging from replication and genome maintenance to host interaction and immune modulation. Consequently, different genomic regions may follow distinct evolutionary trajectories, with some regions exhibiting extensive diversification while others remain evolutionarily constrained [14].
FCoVs therefore provide an opportunity to investigate how genomic innovation and evolutionary constraints interact to shape coronavirus diversity over extended evolutionary timescales [7,15]. Although individual recombinant FCoV strains and specific recombination events have been described in previous studies, these observations have largely been considered in the context of particular strains, genomic regions, or geographic populations. A broader genome-wide assessment is therefore needed to determine how such individual evolutionary events collectively contribute to the long-term diversification of FCoVs. Such an approach can reveal whether genetic exchange, nucleotide-level adaptation, and selective forces act independently or collectively to define the evolutionary landscape of circulating FCoVs.
In this study, we analyzed 127 complete FCoV genomes collected globally over more than five decades to investigate the evolutionary mechanisms underlying FCoV diversification. Rather than focusing on the identification of individual evolutionary events, we integrated phylogenetic reconstruction, recombination analysis, gene-level evolutionary comparisons, genome composition analysis, and selection assessment to examine how multiple evolutionary processes jointly shape FCoV diversification across genomic scales. These findings provide insights into the evolutionary dynamics of alphacoronaviruses and improve our understanding of how coronavirus genomes diversify while maintaining functional integrity.
2. Materials and Methods
2.1. Sequence Collection and Dataset Preparation
Complete genome sequences of feline coronaviruses were retrieved from the NCBI GenBank database using combinations of relevant keywords, including “feline coronavirus” and “feline infectious peritonitis virus”. The retrieved sequences were screened based on their associated metadata and genomic annotations to construct a dataset representing naturally occurring FCoVs. Only sequences annotated as complete or near-complete genomes with sufficient sequence integrity for whole-genome evolutionary analysis were retained. Sequences with substantial missing regions or incomplete genome coverage were excluded. Sequences identified as laboratory-derived, synthetic, engineered, chimeric, vaccine-associated, or otherwise artificially generated viruses were excluded. After quality control and manual curation, 127 complete FCoV genomes were retained for subsequent evolutionary analyses. The accession numbers of all 127 genomes are provided in Supplementary Table S1. The final dataset represented viruses sampled from multiple geographic regions and spanned more than five decades, thereby providing broad temporal and geographic coverage of FCoV diversity.
2.2. Multiple Sequence Alignment and Phylogenetic Analyses
The 127 complete FCoV genome sequences were aligned using MAFFT (v7.518) with the automatic algorithm selection strategy [16]. The resulting multiple-sequence alignment was subsequently trimmed using trimAl (v1.5.0) with a gap threshold of 0.05. Alignment columns in which gaps were present in more than 95% of the sequences were excluded from the alignment [17]. The resulting alignment was used for maximum-likelihood phylogenetic reconstruction.
A maximum-likelihood (ML) phylogenetic tree was reconstructed using IQ-TREE 2 [18]. For the whole-genome phylogenetic analysis, the optimal nucleotide substitution model was determined using the integrated ModelFinder module based on the Bayesian information criterion (BIC), which identified UNREST + FO + R5 as the best-fitting model. The ML tree was subsequently inferred under this model, and branch support was assessed using 1000 bootstrap replicates.
2.3. Recombination Analyses
Recombination was investigated at both the whole-genome and individual-gene levels. Whole-genome analysis was performed to characterize potential mosaic genomic structures and identify putative recombination breakpoints, whereas gene-level analyses were conducted to assess the distribution of recombination signals across the FCoV genome.
Potential recombination events at the whole-genome level were investigated using SimPlot (v3.5.1) [19]. Representative FCoV genomes from the GIIa and GIIb lineages, including PQ192636.1 and PQ133177.1, respectively, were compared with candidate parental genomes representing FCoVs from GI and canine coronaviruses. Candidate parental sequences were selected based on their phylogenetic relationships and sequence similarity to the query genomes.
Pairwise nucleotide sequence similarity between each query genome and candidate parental sequences was calculated across the complete genome using a sliding-window approach. A window size of 500 nucleotides and a step size of 50 nucleotides were applied. Similarity profiles were examined along the genome to identify changes in the relative similarity of query sequences to different candidate parental genomes. Abrupt transitions in the predominant similarity pattern were used to identify putative recombination regions and estimate the approximate locations of recombination breakpoints. The inferred recombinant regions were subsequently mapped onto the genome annotation to determine the corresponding genomic and coding regions.
Gene-specific recombination analyses were performed independently for all 24 annotated FCoV genes. For each gene, the corresponding nucleotide sequences were extracted from all 127 complete FCoV genomes, generating 24 independent gene-specific datasets, each containing 127 sequences. Each gene-specific dataset was independently aligned using MAFFT before recombination analyses.
The resulting alignments were analyzed using the Recombination Detection Program version 5 (RDP5) [20]. Seven complementary recombination detection methods implemented in RDP5 were applied, including RDP, GENECONV, BootScan, MaxChi, Chimaera, SiScan, and 3Seq. Default settings were used unless otherwise specified. To increase the robustness of recombination detection and reduce the influence of method-specific false-positive signals, a stringent consensus criterion was applied, whereby only recombination events supported by at least five of the seven methods were considered reliable. This criterion required concordant support from the majority of the independent detection methods rather than relying on a single algorithm. For each gene, sequences identified as recombinant according to this criterion were recorded, and the number of recombination-positive sequences was calculated to characterize the gene-specific distribution of recombination across the FCoV genome.
The numbers of recombination-positive sequences identified for the 24 genes were summarized and visualized as a heatmap using the ggplot2 package in R (v4.3.1). The resulting heatmap was used to compare the relative distribution of recombination-positive sequences among individual FCoV genes.
To independently evaluate whether each gene alignment contained statistically significant evidence of recombination, the same 24 gene-specific alignments were analyzed using the Pairwise Homoplasy Index (PHI) test implemented in SplitsTree (v6.3.30) [21]. The PHI test assesses incompatibility among nucleotide sites and provides a statistical test for recombination signals within aligned sequence datasets. A p value < 0.05 was considered statistically significant and interpreted as evidence of detectable recombination signals within the corresponding gene. The PHI test results were evaluated alongside the RDP5 results to provide complementary evidence for gene-specific recombination across the FCoV genome.
2.4. Haplotype Network Analyses
Haplotype analyses were performed for the membrane (M) and nonstructural protein 10 (NSP10) genes to characterize gene-specific genetic diversity and evolutionary relationships among FCoVs. These two genes were selected because neither showed detectable evidence of recombination in the gene-level recombination analyses, thereby providing nonrecombinant genomic regions for assessing evolutionary relationships based primarily on nucleotide variation.
For each gene, unique nucleotide haplotypes were identified from the resulting alignments using DnaSP (v6.12.03), and the frequency of each haplotype was calculated based on the number of sequences carrying the corresponding nucleotide pattern [22]. The identified haplotypes were subsequently imported into PopART (v1.7) for network construction.
Haplotype relationships were inferred using the Neighbor-Joining (NJ) method implemented in PopART. The resulting haplotype networks were visualized to represent genetic relationships among unique haplotypes, with each node corresponding to an individual haplotype and connections reflecting their relative genetic divergence. Haplotype attributes were annotated according to the whole-genome phylogenetic lineages (GI, GIIa, and GIIb) to examine the distribution and genetic relationships of haplotypes across the major FCoV lineages.
2.5. Dinucleotide Composition and CpG Relative Abundance
Dinucleotide composition was characterized across the 127 complete FCoV genomes to assess nucleotide usage patterns and the relative abundance of dinucleotide pairs. For each genome, the frequencies of all 16 possible dinucleotides were calculated using the compseq program implemented in the EMBOSS (v6.6.0) suite. The word size was set to 2, and the observed and expected frequencies of each dinucleotide were calculated based on the nucleotide composition of the corresponding genome.
The relative abundance of each dinucleotide was quantified using the observed-to-expected (O/E) ratio, calculated as the observed frequency divided by the expected frequency. An O/E ratio of 1 indicates that the observed frequency is consistent with the expected frequency based on the underlying mononucleotide composition, whereas values below or above 1 indicate relative depletion or enrichment, respectively. The O/E ratios of all 16 dinucleotides were calculated independently for each genome and subsequently compared among the GI, GIIa, and GIIb lineages to characterize lineage-specific patterns and within-lineage variation in dinucleotide composition.
Because CpG exhibited the strongest relative depletion across the analyzed genomes, its temporal dynamics were further examined. The CpG O/E ratio for each genome was paired with its corresponding sampling year. Pearson correlation analysis was performed using the ggscatterstats function in the ggstatsplot package in R to assess the association between CpG relative abundance and sampling year. The Pearson correlation coefficient (r) and corresponding p value were used to quantify the strength and statistical significance of the temporal association. Temporal correlation analysis was performed only for lineages with sufficient variation in sampling years.
2.6. Selection Pressure Analyses
Selection pressures acting on FCoV protein-coding genes were evaluated using the complete set of 127 FCoV genomes included in this study after sequence curation. Coding sequences corresponding to each of the 24 annotated FCoV genes were extracted from the complete genomes, and each gene was independently aligned at the codon level to preserve the reading frame and ensure accurate inference of synonymous and nonsynonymous substitutions.
The resulting gene-specific codon alignments were analyzed using the Fast Unconstrained Bayesian AppRoximation (FUBAR) method implemented on the Datamonkey web platform [23]. FUBAR was used to identify individual codon sites subject to pervasive positive or negative selection by estimating site-specific synonymous and nonsynonymous substitution rates under a Bayesian framework. Codon sites were considered positively selected according to the statistical criteria implemented by FUBAR. The positively selected sites identified across the 24 protein-coding genes were subsequently summarized and compared among genes to characterize the distribution of selective pressures across the FCoV genome.
3. Results
3.1. Evolutionary Relationships and Phylogenetic Diversity of FCoVs
To investigate the evolutionary relationships among feline coronaviruses, a maximum likelihood phylogenetic tree was reconstructed using 127 complete genome sequences. The resulting phylogeny resolved three well-supported clades, each receiving a bootstrap value of 100, demonstrating the robustness of the inferred evolutionary relationships (Figure 1). These three clusters were designated GI, GIIa, and GIIb. The GIIa and GIIb lineages clustered together to form a strongly supported monophyletic group with a bootstrap value of 100, indicating that they share a more recent common ancestor with each other than with GI (Figure 1).
[图：Global phylogenetic relationships and genetic diversity of feline coronaviruses. Maximum-likelihood phylogenetic tree re]
GI contained the largest number of FCoV genomes and included strains collected between 1970 and 2023. The phylogenetic topology revealed no apparent geographic clustering within this lineage, suggesting the absence of obvious geographic structure (Figure 1). Genetically distinct viruses were identified within the same region during similar sampling periods. For example, viruses circulating in Australia around 2017 were distributed across different positions within GI, with one cluster located in the upper portion of the tree and additional Australian strains occupying more basal branches (Figure 1). This pattern highlights the considerable genetic diversity of FCoVs circulating within the same geographic region.
Within GI, clade A comprised closely related strains originating from China, Thailand, the Netherlands, Denmark, and the United Kingdom, indicating that genetically similar FCoVs have undergone extensive international dissemination (Figure 1). This observation is consistent with the widespread circulation of FCoVs across Asia, Europe, Australia, and the Americas. In contrast, clade B consisted of three Japanese strains collected in 1988, 2015, and 2024, respectively (Figure 1). These viruses formed a well-supported monophyletic cluster, suggesting that closely related viral lineages can be maintained within a single geographic region over several decades. Overall, these findings demonstrate that the GI lineage has undergone a complex evolutionary history, with evidence of both widespread geographic dissemination and sustained local diversification, resulting in substantial genetic heterogeneity across time and space.
GIIa primarily comprised FCoV strains from the United States and China that were collected between 2002 and 2024. Within this lineage, clade C represented the predominant cluster and was characterized by extremely short terminal branches, indicating limited genetic divergence among circulating strains (Figure 1). Despite this high degree of genetic similarity, members of clade C have been detected across multiple countries over a period exceeding two decades, suggesting sustained circulation with relatively limited evolutionary diversification.
GIIb formed the sister lineage of GIIa and predominantly consisted of strains collected in Cyprus in 2023, including FCoV 23 (Figure 1). Similar to clade C of GIIa, most GIIb viruses also exhibited very short terminal branches, reflecting a high level of genomic conservation within the lineage. The highly similar phylogenetic topology observed within GIIb is consistent with the dissemination of closely related recombinant lineages rather than repeated independent recombination events.
3.2. Recombination Shapes the Genomic Diversity and Evolutionary History of FCoVs
To investigate the contribution of recombination to FCoV evolution, recombination analyses were conducted at both the whole genome and individual gene levels. Representative genomes from GIIa and GIIb were selected to characterize their recombination patterns.
For GIIa, SimPlot analysis identified PQ192636.1 as a recombinant genome generated through recombination between a GI FCoV and a canine coronavirus (Figure 2A). Most of the genomic backbone was inherited from the GI FCoV lineage, whereas the majority of ORF1b together with the complete spike gene and accessory genes 3a, 3b, and 3c originated from canine coronavirus (Figure 2A). Although an extensive genomic region was acquired from canine coronavirus, sequence similarity remained relatively high across most genomic regions, while the spike gene exhibited substantially greater divergence than the remaining coding sequences. This pattern indicates that the spike gene has accumulated considerably greater genetic variation than other genomic regions during the evolutionary history of these alphacoronaviruses.
[图：Genome-wide and gene-specific recombination patterns in feline coronaviruses. ( A ) SimPlot analysis of the representati]
A distinct recombination pattern was identified in GIIb. SimPlot analysis showed that PQ133177.1 also originated through recombination between a GI FCoV and a canine coronavirus, although the recombination breakpoints differed from those observed in GIIa (Figure 2B). In this genome, the major parental backbone was likewise derived from GI FCoV, whereas most of the spike gene together with the terminal region of ORF1b encoding NSP15 and NSP16 was contributed by canine coronavirus (Figure 2B). The contrasting recombination architectures observed in GIIa and GIIb indicate that these two lineages most likely originated from independent recombination events involving canine coronavirus rather than from a common recombinant ancestor.
Notably, the spike gene was involved in the recombination events that gave rise to both GIIa and GIIb, whereas the principal distinction between the two lineages was the genomic origin of ORF1b (Figure 2). This observation is consistent with the phylogenetic relationship in which GIIa and GIIb form closely related sister lineages while retaining distinct genomic architectures generated through different recombination histories (Figure 1). Collectively, these findings demonstrate that repeated recombination between FCoVs and canine coronaviruses has played a pivotal role in generating the genomic diversity of G2 lineages while maintaining their close evolutionary relationship.
To further characterize the recombination landscape across the FCoV genome, all 24 viral genes were independently analyzed using SplitsTree together with the seven recombination detection algorithms implemented in RDP5. Only recombination events supported by at least five of the seven RDP5 methods were considered reliable. The combined analyses revealed that recombination was widespread throughout the FCoV genome and affected both structural and nonstructural genes, although the extent of recombination varied markedly among individual genes.
Among all genes, NSP3 exhibited the highest frequency of recombination, with 46 recombinant sequences identified, followed by NSP12 with 31 sequences, NSP2 with 27 sequences, and the spike gene with 20 sequences (Figure 2C). Consistent with these findings, SplitsTree analyses of these genes displayed extensive reticulate network structures, providing independent evidence for frequent recombination and complex evolutionary relationships (Figure 2C).
In contrast, recombination signals were unevenly distributed across the genome. No evidence of recombination was detected in NSP1, NSP5, NSP7, NSP10, or the membrane gene by either RDP5 or SplitsTree, indicating that recombination has not contributed equally to the evolutionary history of all FCoV genes (Figure 2C). Instead, recombination exhibited pronounced gene specific variation, with several genes undergoing frequent recombination, whereas others remained highly conserved throughout their evolutionary history.
Taken together, these findings reveal a hierarchical pattern of recombination in FCoVs, whereby recurrent exchange of large genomic fragments contributes to lineage emergence, whereas frequent recombination within specific genes further accelerates sequence diversification during viral evolution.
3.3. Haplotype Network Analyses Reveal Gene Specific Evolutionary Patterns in FCoVs
To further investigate the contribution of mutation to FCoV evolution, haplotype network analyses were performed for the M and NSP10 genes, representing structural and non-structural genes, respectively, both of which showed no detectable evidence of recombination (Figure 2C).
The M gene generated 107 haplotypes from 127 sequences, with no predominant haplotype identified, reflecting extensive sequence diversification driven by accumulated mutations (Figure 3A). Unlike the whole-genome phylogeny, the M gene haplotype network did not preserve the distinct lineage structure separating GI and GIIa. Instead, GI comprised numerous genetically divergent haplotypes distributed throughout the network, indicative of substantial within lineage diversity (Figure 3A). By comparison, the GIIb haplotypes formed a compact cluster with limited genetic divergence (Figure 3A). Despite the close sister relationship between GIIa and GIIb inferred from whole-genome phylogenetic analysis (Figure 1), the GIIb membrane haplotypes were more closely associated with a single GI haplotype than with those of GIIa (Figure 3A). Furthermore, the GIIa membrane haplotypes were partitioned into three genetically distinct groups separated by relatively large mutational distances, suggesting multiple evolutionary origins for this gene within the lineage (Figure 3A).
[图：Haplotype networks of nonrecombinant FCoV genes. ( A ) Neighbor-joining haplotype network constructed from nucleotide se]
Analysis of the NSP10 gene produced 86 haplotypes from the same 127 genomes, again with no dominant haplotype, confirming a similarly high level of genetic diversity (Figure 3B). The GI NSP10 sequences were represented by numerous closely related haplotypes concentrated in the central region of the network, together with two more divergent peripheral clusters (Figure 3B). Consistent with the membrane gene, the GIIa NSP10 haplotypes were not recovered as a discrete lineage but were instead interspersed among the central GI associated haplotypes (Figure 3B). The GIIb NSP10 haplotypes likewise failed to form an independent lineage and were distributed among peripheral GI associated clusters while remaining genetically similar to one another (Figure 3B).
Notably, neither the membrane nor the NSP10 haplotype network recapitulated the close evolutionary relationship between GIIa and GIIb inferred from the whole-genome phylogeny. Instead, haplotypes from both lineages frequently showed closer relationships with GI than with each other. Collectively, these observations reveal substantial discordance between gene-level and genome-wide evolutionary relationships, consistent with the pervasive recombination identified across the FCoV genome.
3.4. Conserved Dinucleotide Composition Bias and Progressive CpG Depletion During FCoV Evolution
To characterize dinucleotide composition across FCoV genomes, the observed to expected ratio of each of the sixteen dinucleotides was calculated for every genome. In the absence of compositional bias, the observed to expected ratio is expected to approximate one.
Pronounced dinucleotide bias was identified in the GI lineage. Seven dinucleotides were overrepresented with observed to expected ratios exceeding one, among which TpG exhibited the highest mean ratio of 1.38, followed by CpA at 1.29 and ApC at 1.2 (Figure 4A). In contrast, eight dinucleotides displayed observed to expected ratios between 0.75 and 1.00, indicating varying degrees of underrepresentation (Figure 4A). Although the ratio of TpT remained close to the expected value, TpC and TpA were consistently depleted. Several of these moderately underrepresented dinucleotides exhibited remarkably limited variation among GI genomes, as evidenced by the nearly overlapping interquartile ranges in the boxplots, suggesting that their relative frequencies have remained highly stable despite continuous sequence evolution.
[图：Dinucleotide composition bias and temporal dynamics of CpG abundance in FCoVs. ( A – C ) Observed-to-expected (O/E) rati]
Among all dinucleotides, CpG was consistently the most depleted, with a mean observed to expected ratio of 0.51 (Figure 4A), whereas TpG represented the most enriched dinucleotide. The marked contrast between CpG depletion and TpG enrichment highlights a pronounced imbalance in dinucleotide usage within FCoV genomes. Furthermore, CpG displayed greater variability among GI genomes than most other dinucleotides, indicating that the extent of CpG suppression differed among circulating strains.
The overall dinucleotide composition of GIIa closely resembled that of GI. Eight dinucleotides were overrepresented, with TpG, CpA, and ApC again exhibiting the highest observed to expected ratios of 1.39, 1.31, and 1.21, respectively (Figure 4B). CpG remained the most depleted dinucleotide with a mean ratio of 0.49 (Figure 4B). Compared with GI, however, dinucleotide frequencies within GIIa were considerably more homogeneous, with only limited variation detected among individual genomes.
A similar compositional profile was observed in GIIb. Approximately half of the dinucleotides were overrepresented, whereas the remainder were underrepresented. TpG, CpA, and ApC again represented the most enriched dinucleotides, with mean observed to expected ratios of 1.39, 1.34, and 1.24, respectively (Figure 4C). Consistent with both GI and GIIa, CpG exhibited the lowest observed to expected ratio at 0.53 (Figure 4C). Despite the recombinant origin of GIIa and GIIb, their dinucleotide composition remained highly similar to that of GI, indicating that the overall pattern of dinucleotide usage is conserved across alphacoronaviruses rather than being lineage-specific.
Because CpG consistently represented the most depleted dinucleotide in all three lineages, its temporal dynamics were further investigated. Within GI, the CpG observed to expected ratio exhibited a significant negative correlation with sampling year (r = −0.28, p = 7.12 × 10−3), demonstrating a gradual reduction in CpG frequency over time (Figure 4D). No significant temporal association was detected in GIIa (Figure 4E), whereas correlation analysis was not performed for GIIb because all available genomes were collected during the same year. Together, these findings demonstrate that CpG suppression represents a conserved genomic characteristic of FCoVs and that CpG depletion has continued throughout the evolutionary history of the GI lineage (Figure 4).
3.5. Distinct Selective Pressures Shape the Evolution of FCoV Genes
To characterize the selective pressures acting on FCoV genes, codon-based selection analyses were performed using FUBAR implemented in the Datamonkey platform. Collectively, these analyses revealed that positive selection was unevenly distributed across the FCoV genome and was concentrated in a subset of genes.
Among all genes, NSP2 and NSP3 exhibited the highest numbers of positively selected codon sites, with ten sites identified in each gene by FUBAR (Figure 5A). These results indicate that both NSP2 and NSP3 experienced substantial positive selection during FCoV evolution, highlighting these two genes as prominent targets of selective pressure within the FCoV genome.
[图：Heterogeneous patterns of positive selection across the FCoV genome. ( A ) Distribution of positively selected sites acr]
For the S gene, the complete set of FCoV spike sequences was analyzed as a single dataset using FUBAR. Seven codon sites, corresponding to amino acid positions D33, V65, S82, E219, K267, V425, and E798, showed significant evidence of positive selection (Figure 5). These positively selected sites were distributed across the S1 domain of the spike protein, indicating that selective pressures were not restricted to a single functional region of the FCoV spike protein (Figure 5).
Positive selection was also detected in 11 additional FCoV genes by FUBAR (Figure 5A). Gene 7b contained the highest number of positively selected codon sites among these genes, with seven sites identified, followed by NSP12 and the M gene, with five and four sites, respectively. Three positively selected sites were identified in each of NSP1, the N gene, 3a, and 3c, whereas NSP6, NSP9, NSP14, and NSP16 each contained one positively selected site. Together, these results indicate that positively selected sites were distributed across multiple structural and nonstructural genes of the FCoV genome (Figure 5A).
Collectively, these findings demonstrate that positive selection has acted on a substantial proportion of FCoV genes during viral evolution. Notably, the genes showing the strongest signatures of positive selection largely overlapped with those exhibiting frequent recombination, identifying these genes as major evolutionary hotspots within the FCoV genome. The recurrent accumulation of adaptive substitutions in these genes is consistent with persistent evolutionary pressures imposed during virus host interactions.
4. Discussion
The present study provides a genome-wide perspective on the evolutionary dynamics of feline coronaviruses and indicates that FCoV diversification involves multiple evolutionary processes operating across different genomic scales. At the genome level, homologous recombination is associated with differences in genomic architecture and the emergence of distinct genetic lineages. At the gene level, sequence variation and episodic positive selection contribute to diversification of individual genomic regions. At a broader scale, the persistence of conserved genomic characteristics despite sequence variation suggests the presence of longer-term compositional constraints. Together, these findings highlight the coexistence of genomic diversification and evolutionary constraint during FCoV evolution.
One of the clearest manifestations of this hierarchical organization is the marked discordance between genome-wide and gene-level evolutionary relationships. Whole-genome phylogenetic analyses consistently resolved three well-supported lineages (Figure 1), whereas haplotype networks reconstructed from nonrecombinant genes failed to recover the same lineage relationships (Figure 3). Instead, genes from different genomic backgrounds frequently displayed closer relationships to ancestral GI viruses than to each other. Although phylogenetic incongruence is commonly regarded as evidence of recombination, our findings further demonstrate that it reflects a more fundamental characteristic of feline coronavirus evolution: different components of the viral genome may follow partially independent evolutionary trajectories. Consequently, the evolutionary history of a coronavirus genome cannot always be inferred from the history of individual genes.
This observation has broader implications for evolutionary studies of coronaviruses. Many previous investigations have relied on single genes, particularly the spike gene, to infer viral origins, transmission histories, or host adaptation because of its biological importance and extensive sequence availability [8,24]. However, coronaviruses evolve as integrated yet modular genomes. Individual genomic regions experience different combinations of recombination, mutation, and selective constraint, resulting in heterogeneous evolutionary histories across the genome. As a consequence, gene trees and genome trees should not be viewed as competing representations of viral evolution, but rather as complementary descriptions of evolutionary processes occurring at different biological scales [25]. In our dataset, the discordance between whole-genome phylogenetic relationships and gene-level haplotype patterns suggests that different genomic regions may retain partially distinct evolutionary histories. Thus, whole-genome and gene-based analyses may provide complementary perspectives on FCoV evolution, with whole-genome analyses reflecting relationships among complete genomic backgrounds and gene-level analyses highlighting variation within individual genomic regions.
Our results further suggest that mutation and recombination may play distinct evolutionary roles in feline coronavirus diversification. Extensive haplotype diversity within the membrane and NSP10 genes demonstrates that nucleotide substitutions accumulate continuously even in genomic regions lacking detectable recombination, indicating that mutational processes contribute to the generation of genetic diversity during long-term viral circulation (Figure 3). Nevertheless, the patterns of sequence variation observed in these genes alone do not fully recapitulate the major lineage relationships inferred from complete genomes. Instead, the genome-wide phylogenetic and recombination analyses indicate that recombination is associated with differences in genomic configuration among FCoV lineages. Taken together, these observations are consistent with complementary contributions of mutation and recombination to FCoV diversification, but do not establish their relative contributions or a specific temporal or causal relationship between the two processes.
The hierarchical framework proposed above further suggests that homologous recombination is not simply a mechanism for generating genetic diversity but a process that repeatedly reorganizes coronavirus genome architecture. Coronaviruses possess one of the highest homologous recombination frequencies among positive-sense RNA viruses because discontinuous RNA synthesis during replication facilitates frequent template switching by the viral RNA-dependent RNA polymerase [26,27]. As a consequence, extensive genomic fragments rather than individual nucleotides can be exchanged between closely related viruses within a single replication cycle. Compared with the gradual accumulation of point mutations, such large-scale genetic exchange enables FCoVs to explore novel genomic combinations on a substantially shorter evolutionary timescale.
Our analyses indicate that the emergence of the GIIa and GIIb lineages is consistent with repeated, rather than singular, recombination events between feline and canine coronaviruses. Although both lineages share a similar evolutionary origin involving CCoV-derived genomic regions, they possess distinct recombination architectures and breakpoint distributions, indicating that they were generated through independent evolutionary events instead of diversification from a common recombinant ancestor. Recurrent interspecies recombination has likewise been documented in several other alphacoronaviruses and betacoronaviruses, highlighting homologous recombination as a continual source of genomic innovation rather than an isolated historical occurrence [28,29]. Together, these observations suggest that the exchange of genetic material between related coronaviruses represents an ongoing evolutionary process capable of repeatedly generating novel viral genomic backgrounds.
An important implication of these findings is that feline coronavirus genomes appear to evolve as assemblies of interacting functional modules rather than as indivisible genetic units. The independent replacement of genomic regions encompassing the spike gene and portions of ORF1b in the recombinant lineages indicates that certain genomic regions can be exchanged while preserving the functionality of the remaining genome. Similar patterns have been observed in other coronaviruses, where genes associated with host interaction frequently exhibit evolutionary histories distinct from those of the core replication machinery [30,31]. Such modularity likely increases evolutionary flexibility by allowing individual functional components to be replaced or optimized without requiring simultaneous modification of the entire genome. However, the modular nature of FCoV genomes should not be interpreted as unlimited interchangeability. Functional modules remain embedded within highly coordinated replication networks, and successful recombination ultimately depends on the compatibility between newly acquired genomic regions and the pre-existing genetic background.
If recombination generates new genomic architectures, an equally important question is why evolutionary change is repeatedly concentrated within only a subset of viral genes. Our analyses revealed a striking overlap between genes exhibiting frequent recombination and those undergoing recurrent positive selection, particularly NSP3, the spike gene, NSP2, NSP12, and 7b gene (Figure 5). This overlap indicates that some genomic regions experience multiple forms of evolutionary variation, but does not establish a direct relationship between recombination and positive selection. In other words, evolutionary innovation in FCoVs appears to be focused within a limited number of functional modules that mediate critical interactions between the virus and its host.
This interpretation is consistent with a growing body of evidence indicating that FCoV evolution is highly heterogeneous across the genome. Viral proteins differ substantially in the degree to which they interact with host molecules, immune defenses, and changing cellular environments. Proteins positioned at the virus–host interface are expected to experience continual selective pressure as viruses optimize replication efficiency while simultaneously evading host antiviral responses [32]. Under such conditions, any genetic change capable of improving viral fitness—whether introduced through mutation or recombination—has an increased probability of being retained by natural selection. Conversely, proteins performing highly conserved structural or enzymatic functions are subject to stronger purifying selection because even minor sequence alterations may compromise essential viral processes. The heterogeneous evolutionary landscape observed across the FCoV genome therefore likely reflects differences in functional constraint rather than differences in intrinsic mutation or recombination rates.
Among all genes, NSP3 represents the clearest example of this evolutionary principle. NSP3 not only exhibited the strongest evidence of recombination but also contained the largest number of positively selected codons identified in this study. This multifunctional protein occupies a central position within the coronavirus replication cycle, integrating proteolytic processing, replication complex assembly, membrane remodeling, and multiple mechanisms of innate immune antagonism [33]. Unlike proteins dedicated to a single biological function, NSP3 is continuously exposed to selective pressures arising from both viral replication and host antiviral defenses. Its exceptional evolutionary dynamics therefore likely reflect the convergence of multiple selective forces acting simultaneously on a single protein.
The evolutionary pattern of the spike protein illustrates a complementary but distinct adaptive strategy. Whereas NSP3 primarily mediates intracellular interactions with the host, the spike glycoprotein defines the extracellular interface through which coronaviruses recognize host receptors and encounter antibody-mediated immune pressure. It is therefore unsurprising that positively selected sites were concentrated within the S1 subunit, where receptor-binding and antigenic determinants are located, while the membrane-fusion machinery within S2 remained comparatively conserved. Notably, both independent recombinant lineages identified in this study involved replacement of the spike gene, indicating that modification of host interaction modules has repeatedly accompanied lineage emergence during FCoV evolution.
Adaptive evolution was also detected in NSP2, NSP12, and the nucleocapsid protein, highlighting that coronavirus adaptation extends well beyond receptor recognition (Figure 5A). Although these proteins perform diverse biological functions, they all contribute to processes that directly influence viral fitness within infected cells, including regulation of replication, genome synthesis, intracellular organization, and modulation of host responses [34]. Their adaptive diversification therefore reinforces the view that successful FCoV evolution depends on coordinated optimization of multiple components of the viral life cycle rather than modification of any single protein. Collectively, these findings suggest that adaptive evolution in FCoVs is distributed across an interconnected functional network, with individual proteins evolving under different selective regimes but contributing collectively to viral adaptation.
Importantly, the coexistence of recurrent recombination and positive selection within the same genes should not be interpreted as evidence that one process directly drives the other. Instead, both processes are likely consequences of the same underlying evolutionary principle: genes experiencing the strongest interactions with the host are simultaneously the genes in which evolutionary innovation is most strongly favored. Recombination provides access to alternative genetic variants, whereas positive selection determines which of these variants, together with newly arising mutations, confer sufficient fitness advantages to persist during long-term viral circulation. This overlap indicates that some genomic regions experience multiple forms of evolutionary variation, but does not establish a direct relationship between recombination and positive selection.
The evolutionary patterns described above indicate that extensive genomic innovation in FCoVs is accompanied by equally pervasive evolutionary constraint. While recombination repeatedly reorganizes genome architecture and adaptive evolution continuously modifies individual proteins, several fundamental genomic characteristics remain remarkably stable throughout viral diversification. This apparent paradox highlights an important feature of FCoV evolution: successful adaptation does not require unrestricted genetic change but instead occurs within boundaries imposed by long-term functional and genomic constraints [30].
The highly conserved dinucleotide composition observed across all three evolutionary lineages exemplifies these constraints. Despite their independent evolutionary histories and distinct recombinant architectures, GI, GIIa, and GIIb retained nearly identical patterns of dinucleotide usage, characterized by enrichment of TpG and CpA together with persistent depletion of CpG (Figure 4). These observations indicate that large-scale genomic exchange alone is insufficient to fundamentally alter the compositional properties of FCoV genomes. Instead, genome composition appears to remain remarkably resilient despite continual turnover of primary nucleotide sequence, implying that compositional evolution operates under constraints that extend beyond the inheritance of individual genomic regions.
Among these conserved features, CpG depletion represents one of the most prominent genomic signatures of vertebrate RNA viruses. Although reduced CpG frequencies are widely interpreted as evidence of adaptation to host antiviral mechanisms, particularly ZAP-mediated recognition of CpG-rich viral RNA, increasing evidence suggests that no single mechanism adequately explains this phenomenon [35,36]. Mutational biases, codon usage, RNA structural requirements, replication-associated constraints, and multiple host selective pressures are all likely to contribute to long-term CpG evolution. The gradual reduction in CpG observed within the GI lineage is therefore consistent with long-term compositional change, but should be interpreted cautiously because the available sequences are unevenly distributed across sampling years and geographic regions. In particular, temporal and geographic effects cannot be completely disentangled in the present dataset, and the observed negative correlation between CpG abundance and sampling year does not by itself establish a purely temporal effect. Nevertheless, the observed pattern is compatible with the possibility that CpG composition has gradually changed during the diversification of GI FCoVs. The absence of a similar temporal trend in GIIa should be interpreted cautiously, given its more recent origin, narrower sampling period, and recombinant evolutionary history.
The pronounced and persistent CpG depletion observed in FCoVs may also have implications beyond understanding their evolutionary history. In other RNA viruses, including influenza A virus, experimentally increasing synonymous CpG dinucleotides has been shown to attenuate viral replication while preserving the encoded protein sequences, highlighting CpG manipulation as a potential strategy for rational attenuation [37,38]. Whether a similar approach could be exploited for FCoVs remains unknown, but the conserved CpG constraint identified here provides an evolutionary rationale for investigating synonymous CpG modification as a potential route toward rational FCoV attenuation.
Whether these genomic evolutionary patterns are associated with the transition from FCoV infection to FIP remains an important but unresolved question, as the heterogeneous phenotypic information available for the present dataset does not allow a systematic phenotype-stratified analysis. Notably, previous work has linked specific recombinant FCoV–CCoV lineages to FIP outbreaks, highlighting the potential relevance of genomic recombination to pathogenic phenotypes [7].
Taken together, our findings support a conceptual framework for FCoV evolution in which genomic diversification occurs alongside evolutionary constraints. The patterns observed in our dataset indicate that recombination is associated with variation in genomic architecture, whereas mutation and positive selection contribute to sequence diversification across individual genes. At the same time, the persistence of conserved genomic characteristics despite sequence variation suggests the influence of longer-term genomic constraints. Although the relative contributions and temporal relationships among these processes were not quantitatively assessed in the present study, their contrasting patterns across genomic scales provide a framework for interpreting FCoV diversification as a balance between genomic plasticity and evolutionary constraint. Further comparative and experimental studies will be required to determine how these processes influence viral fitness, host adaptation, and pathogenic phenotypes.
5. Conclusions
Our findings reveal that the long-term evolution of feline coronaviruses is shaped by a dynamic balance between genomic innovation and evolutionary constraint. Recurrent recombination, including independent exchanges with canine coronaviruses, contributes to the emergence of distinct genomic lineages, whereas mutation and positive selection drive continued diversification within individual genes. These processes are particularly concentrated in evolutionary hotspots such as NSP3 and S, highlighting the importance of both replication-associated and host-interacting proteins in FCoV adaptation. At the same time, the persistent depletion of CpG across major lineages and its progressive decline over time indicate that broader genomic constraints remain conserved despite extensive sequence turnover and recombination. Together, these findings support a conceptual framework in which FCoV diversification involves both genomic plasticity and evolutionary constraint, rather than a single dominant evolutionary process. Although the relative contributions and temporal relationships among recombination, mutation, and selection were not quantitatively assessed in this study, their contrasting patterns across genomic scales provide new insights into FCoV diversification and may inform future evolutionary surveillance of feline and related alphacoronaviruses.
Acknowledgments
We thank the High Performance Computing Cluster of the College of Veterinary Medicine, Yangzhou University, for providing computational resources.
Abbreviations
The following abbreviations are used in this manuscript:
| FCoV | feline coronavirus |
| FIPV | feline infectious peritonitis virus |
| CCoV | canine coronavirus |
Supplementary Materials
The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/v18091037/s1, Table S1: Information on the FCoV genome sequences analyzed in this study.
[内容请查看原文]
Author Contributions
W.W. conceived and designed the study. X.W. and Q.Z. performed the experiments, analyzed the data, and drafted the manuscript. W.W., contributed to revise the manuscript. All authors have read and agreed to the published version of the manuscript.
Institutional Review Board Statement
No ethical approval was required as this study used publicly available genomic data.
Informed Consent Statement
Not applicable.
Data Availability Statement
The original contributions presented in this study are included in the article material. Further inquiries can be directed to the corresponding authors.
Conflicts of Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Funding Statement
This work was supported by the National Natural Science Foundation of China (32500538), the 111 Project D18007, and the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD).
Footnotes
[内容请查看原文]
References
[内容请查看原文]
Associated Data
Supplementary Materials
[内容请查看原文]
Data Availability Statement
The original contributions presented in this study are included in the article material. Further inquiries can be directed to the corresponding authors.