Figure 4. Alignment and tree of nifHDKT and of dreped MAGs for reconcilation and ancestral reconstruction

Sequences of the <i>nifH</i>, <i>nifD</i>, <i>nifK </i>and <i>nifT</i> genes were chosen to infer the evolutionary history of nitrogen fixation because they form an operon with the minimum catalytic gene set for the reaction (Dos Santos et al. 2012). Sequences were extracted from the DRAM annotation of the dereplicated MAGs, including the genome of <i>Vibrio natrigensis,</i> which also served as the outgroup for the gene tree. MAGs that did not contain the full set of genes and were excluded from the analysis. Each renamed gene extraction (Rodriguez-R and Konstantinidis 2016) was aligned independently with FSA v1.15.9 (Bradley et al. 2009) using the <i>--fast</i> flag. A tree was inferred from each alignment to ensure that the genes have evolved together. Subsequently, the alignments were concatenated using SeqKit v2.3.0 (Shen et al. 2016) and a tree inferred from it. For the reconstruction of ancestral states and the phylogenetic reconciliation we inferred a de-replicated version of the global symbiont phylogenomic tree. For this, the diversity of MAGs was reduced by following the dRep v3.2.2 (Olm et al. 2017) dereplicate workflow with the S_ANI parameter set to 0.99. To maintain the topology of the original tree, MAGs classified as <i>Ca. </i>Thiodiazotropha endolucinida Caribbean had to be dereplicated independently with the S_ANI flag set to 0.995 (to increase the represented diversity of the clade). The genome of <i>Vibrio natrigensis </i>genome (GCF_001456255.1), which was used as an outgroup, was downloaded from the NCBI database.Both trees wereinferred with IQ-Tree v2.2.2.1 (Minh et al. 2020; Hoang et al. 2018; Kalyaanamoorthy et al. 2017) with auto substitution model detection, 1,000 ultrafast bootstrap (UFB) replicates and 1,000 samples for SH-aLRT branch testing. Nodes with values of UFB greater or equal to 95% and of SH-aLRT greater or equal to 80% were considered to be strongly supported. The nifHDKT best-fit model was TN+F+I+G4, and the one for the dreped tree was Q.insect+F+I+G4

选取<i>nifH</i>、<i>nifD</i>、<i>nifK</i>及<i>nifT</i>基因的序列以推断固氮作用的演化历史，原因在于这些基因构成了该反应所需的最小催化基因集合操纵子（Dos Santos等，2012）。序列从去冗余宏基因组组装基因组（Metagenome-Assembled Genome，简称MAGs）的DRAM注释结果中提取，其中包含<i>Vibrio natrigensis</i>的基因组，该基因组同时作为基因树的外类群。未携带完整基因集合的MAGs被排除在本次分析之外。每一个经Rodriguez-R与Konstantinidis（2016）提出的方法重命名的基因提取结果均使用FSA v1.15.9（Bradley等，2009）并添加`--fast`参数独立进行序列比对。针对每一条比对结果分别构建进化树，以验证这些基因是否协同演化。随后，使用SeqKit v2.3.0（Shen等，2016）将各比对序列进行拼接，并基于拼接后的序列构建进化树。为重建祖先状态并开展系统发育调和分析，我们构建了去冗余的全球共生体系统发育基因组树版本。为此，我们遵循dRep v3.2.2（Olm等，2017）的去冗余流程以降低MAGs的多样性，其中S_ANI参数设置为0.99。为保留原始树的拓扑结构，被归类为<i>Ca. Thiodiazotropha endolucinida</i>加勒比海演化支的MAGs需要单独以S_ANI参数设置为0.995进行去冗余，以提升该演化支的代表性多样性。作为外类群使用的<i>Vibrio natrigensis</i>基因组（GCF_001456255.1）从NCBI数据库下载获得。两棵进化树均使用IQ-Tree v2.2.2.1（Minh等，2020；Hoang等，2018；Kalyaanamoorthy等，2017）构建，采用自动替换模型选择、1000次超快速自助法（UFB）重复采样以及1000次SH-aLRT分支检验采样。UFB值≥95%且SH-aLRT值≥80%的节点被视为具有强支持度。nifHDKT基因集的最优拟合模型为TN+F+I+G4，而去冗余树的最优拟合模型为Q.insect+F+I+G4。