Phylogenomics of Psammodynastes and Buhoma (Elapoidea: Serpentes), with the outline of a brand new Asian snake household

Genome sequencing, processing of goal seize and entire genome sequence information

The ultraconserved components (UCE) of 47 taxa, together with 44 elapoids and three of the 4 outgroups, are from Das et al.⁷. These samples had been particularly enriched and sequenced for UCEs⁷.

UCEs of 11 extra elapoids, together with one of many focal taxa of this paper—Psammodynastes pulverulentus, and ten elapids, and one extra outgroup taxon belonging to Homalopsidae, Myanophis thanlyinensis, had been harvested in silico from entire genome sequences. The reference genome sequencing of P. pulverulentus and meeting had been carried out at Iridian Genomes (Bethesda, USA). Sequencing of reference genome, described on this part, and Sanger sequencing and CT scans, reported below different sections, utilised solely preserved museum specimens and no dwell animals had been collected/experimented upon/euthanised for our examine. Genomic DNA from a P. pulverulentus (FMNH 273629) from the gathering on the Field Museum of Natural History (Chicago, USA) was extracted with the Qiagen DNAEasy equipment following the usual directions. Library preparation was carried out utilizing Illumina TruSeq equipment with commonplace adapters and following the usual protocol. Sequencing at 105X protection was carried out on llumina X-Ten platform. Raw reads had been preassembled to scaffold (103,289 scaffolds) stage with SPAdes model 3.15.4³⁴ and Zanfona (https://github.com/zanfona734/zanfona) was used for genome ending. The meeting has been made available on the NCBI Genome database (https://www.ncbi.nlm.nih.gov/datasets/genome/GCA_025802295.1/). Rest of the genome assemblies had been obtained from the NCBI Genome database (NCBI Taxonomy since July, 2023). Reference meeting identify, hyperlink and related publications, if any, for all of the reference genomes, together with that of P. pulverulentus, are listed within the Supplementary Information.

Reference genome sequences for all of the elapoids and a homalopsid available within the NCBI Genome (NCBI Taxonomy) database had been downloaded. The meeting of Hydrophis hardwickii has meeting contamination and therefore, was not included within the downstream analyses. The phyluce pipeline, model 1.7.2,³⁵ was used to match the Tetrapods-UCE-5Kv1 bait set (https://www.ultraconserved.org/) to the genome sequences and create a database utilizing the phyluce_probe_run_multiple_lastzs_sqlite operate. Using the lastz database generated in that step, loci matching the UCE probe had been extracted with the phyluce_probe_slice_sequence_from_genomes operate, retaining 400 bp of downstream and upstream flanks. The outputs had been symlinked to the contigs generated from the dataset of Das et al.⁷. To generate the mandatory probe match database and configuration information for the downstream processing, the probe matching step was repeated with phyluce_assembly_match_contigs_to_probes command with the identical Tetrapods-UCE-5Kv1 probe set. Finally, the phyluce_assembly_get_match_counts and the phyluce_assembly_get_fastas_from_match_counts features had been used to extract the UCEs for all of the 59 taxa. MAFFT³⁶ was used, together with edge trimming, to align every individual UCEs locus dataset on phyluce. Preparation of the ultimate UCE a number of sequence alignments for phylogenomic analyses was carried out within the phyluce pipeline and contains three steps, specifically, removing of UCE locus identify from the taxon nomen, making ready datasets of three completely different ranges of completion (viz., 50, 75 and 95% full datasets) and producing concatenated datasets for mixed information analyses.

Sanger sequencing and preparation of Sanger sequence information

For this examine, we’ve sequenced mitochondrial cytochrome b (CYTB), NADH dehydrogenase subunit 4 (ND4) and 16S ribosomal RNA gene (16S) and nuclear brain-derived neurotrophic issue (BDNF), oocyte maturation issue mos (C-MOS) and recombination activating gene 1 (RAG1) of Buhoma depressiceps (UTEP 22599) and B. marlieri (UTEP 22598) specimen within the assortment of the University of Texas at El Paso (El Paso, USA). The PCR and sequencing primers for BDNF had been as in³⁷. The remainder of the primers had been from³⁸. The laboratory protocols, together with these for genomic DNA extraction, PCR amplification and Sanger sequencing comply with³⁸.

The uncooked Sanger sequencing reads had been assembled utilizing the Pearl programme of the Tracy toolkit³⁹ hosted at GEAR-GENOMICS (https://www.gear-genomics.com/). The meeting was visually checked in Pearl and base calls had been rectified, based mostly on thorough inspection of the chromatogram, if wanted. The ND4 and CYTB sequences of B. marlieri had been of low high quality and therefore, these had been discarded from subsequent analyses. Open Reading Frames (ORF) of the protein-coding sequences had been detected with NCBI ORF Finder (https://www.ncbi.nlm.nih.gov/orffinder/), with setting the ORF begin codon as any sense codon. Afterwards, we carried out a megablast on NCBI BLAST with default parameters to substantiate the identification of the sequences we generated.

We looked for and obtained sequences of mitochondrial CYTB, ND4, 16S and 12S (12S ribosomal RNA gene) and nuclear BDNF, C-MOS, RAG1 and RAG2 (recombination activating gene 2) for elapoids and outgroups for which we’ve UCE information from NCBI GenBank. While a lot of the Sanger sequences in our dataset belong to the identical species as within the UCE dataset, in a number of cases we’ve included the sequence of a intently associated congener if the sequence of a gene for a selected taxon is unavailable. This was carried out provided that the taxon is the only consultant of its genus in our dataset and the congeneric standing of the substitute is taxonomically uncontroversial. GenBank accession numbers of the sequences, together with those sequenced by us for this examine, and another related info are offered within the Supplementary Information (Table S1).

We would not have tissue pattern of Buhoma procterae and due to this fact used CYTB, ND4, 12S, 16S, C-MOS and RAG2 sequences from¹¹. We additionally obtained 12S and RAG2 sequences of B. depressiceps generated in the identical examine. The B. depressiceps specimen utilized in that work originated from Gabon and therefore, is unambiguously referable to B. depressiceps and never B. marlieri.

For taxa with a reference genome, we harvested nuclear genes from the genome meeting itself to alleviate the necessity for creating composites. To do that, we BLAST-ed conspecific/congeneric/confamilial sequences of the 4 nuclear loci individually towards a genome with megablast and default parameters. The matched area was extracted and megablast-ed towards the GenBank database, no matter whether or not the genome was annotated or not, to make sure that we’ve obtained the precise gene we had been on the lookout for. Nuclear loci sequences harvested on this approach had been added to our Sanger loci dataset. For the question sequence accession quantity, see Supplementary Information (Table S2–S5).

Sequences had been aligned with MUSCLE⁴⁰, carried out on MEGA 11⁴¹, with default parameters. Alignments had been scrutinised visually. Overhangs of a few very lengthy sequences on the ends that had been lacking in different sequences had been trimmed. External gaps had been coded as lacking information with SequenceMatrix 1.9⁴². SequenceMatrix 1.9 was used to concatenate the Sanger loci. We have ready the next concatenated datasets—(1) 4 nuclear loci, (2) 4 mitochondrial loci, (3) all eight Sanger loci, (4) 4 nuclear loci plus 50, 75 and 95% full, concatenated UCE datasets, and (5) all eight Sanger loci plus 50, 75 and 95% full, concatenated UCE datasets, utilizing SequenceMatrix 1.9. Prior to combining Sanger nuclear and UCE loci, we ran native blastn (blast-2.10.0 +) of unaligned multifasta question of Sanger nuclear loci, with e-values 1e⁻⁵⁰ and 0.05, towards a database (ready with makeblastdb [blast-2.10.0 +]) of unaligned multifasta of UCEs to make sure that UCEs flanks didn’t overlap any of the 4 nuclear genes. No match was detected within the native blastn searches.

Phylogenomic analyses

We inferred quartet-based multispecies coalescent species bushes from gene bushes of individual UCE loci, most probability (ML) phylogenies from concatenated UCE solely dataset, concatenated UCE + conventional marker dataset and mixed conventional marker dataset and Bayesian Inference (BI) phylogeny from concatenated UCE + conventional marker dataset.

For the multispecies coalescent (MSC) phylogenetic tree estimation, we used weighted ASTRAL hybrid (wASTRAL-h) methodology⁴³. Algorithms of weighted ASTRAL class weigh quartets based mostly on department assist, department size or each with out requiring any arbitrarily specified threshold. This has been demonstrated to lead to phylogenies extra correct than these produced by ASTRAL-III algorithm⁴³. Among the three completely different weighting approaches, the one which weighs based mostly on each the department size and assist (i.e., wASTRAL-h) performs the very best. ML gene bushes had been estimated for 50, 75 and 95% full UCE loci set with IQ-TREE 2.2.0.5⁴⁴ with 1000 ultrafast bootstrap⁴⁵ replicates per gene tree. Model choice for every locus was carried out with the default ModelFinder⁴⁶ possibility. MSC species tree was then inferred from the gene bushes set with wASTRAL-h with default settings.

We used IQ-TREE 2.2.0.5 to deduce ML species tree from the UCE-only concatenated matrices of fifty, 75 and 95% completeness. The default mannequin choice and optimum partitioning scheme discovering algorithm of IQ-TREE, i.e., ModelFinder plus partition merging (MPF + MERGE possibility), proved to be computationally infeasible on our cluster. Hence, we selected the TESTMERGE possibility, which features identically to PartitionFinder⁴⁷ and is much less computationally taxing. TESTMERGE differs from ModelFinder in that it doesn’t think about the FreeRate heterogeneity mannequin⁴⁶ carried out within the latter. Additionally, we used the relaxed hierarchical clustering methodology⁴⁸ to contemplate the highest 10% most comparable partition (individual UCE locus) pairs (by setting -rcluster to 10) that are then merged till no higher partition merging schemes may be estimated for additional computational effectivity. Branch assist was assessed with 1000 ultrafast bootstrap (UFBoot) replicates. To cut back the potential of flawed splits receiving excessive UFBoot assist⁴⁹, we resampled from inside the partitions and websites inside a partition by setting the –sampling to GENESITE.

We inferred ML phylogenies with UFBoot department assist from UCE + all eight Sanger loci and UCE + solely the Sanger nuclear loci mixed datasets with the identical software program and parameter settings as with the UCE-only ML analyses.

ML phylogenies had been additionally inferred from concatenated Sanger-sequenced mito-nuclear, mitochondrial and nuclear loci datasets individually with IQ-TREE. Small sizes of those datasets allowed a full MFP + MERGE mannequin choice and grasping seek for the very best partitioning scheme. For protein-coding genes, partitions had been inputted as individual codon positions. UFBoot assist, with within-partition and sites-within-partition resampling enabled, was computed for these phylogenies in the identical method because the UCE and UCE + conventional marker ML analyses.

Finally, we additionally computed a BI phylogeny from the 95% full UCE and conventional nuclear loci dataset with ExaBayes 1.5.1⁵⁰ to check if BI corroborates the end result from the ML analyses of UCE and conventional marker mixed information. We ran the MPI (message passing interface) model ExaBayes 1.5.1, i.e., exabayes, on unpartitioned UCE and nuclear loci matrix with 200 MPI processes. Two unbiased runs, every with 4 coupled chains (one chilly, three heated), had been run as parallel processes. Number of swap makes an attempt per era was set to 2. Sampling frequency was set to 500 generations. Diagnostics for convergence had been checked each 5000 generations. Other parameters had been left at default settings. Though the variety of generations was set to 100 M previous to the graduation of the evaluation, two consecutive runs (a second one from the checkpoint of the timed-out first run on the cluster) may attain solely 2.7 M generations. However, the typical commonplace deviation of the break up frequencies (ASDSF), the default divergence metric of ExaBayes, fell to three.88% which is decrease than the default 5% convergence diagnostic worth, thus indicating a suitable stage of convergence. A majority rule consensus topology was estimated from the posterior distribution of bushes after discarding the primary 25% as burn-in with the consense utility of ExaBayes.

All the phylogenomic analyses had been run on the Puhti supercomputing cluster of the Finnish CSC-IT Centre for Science Ltd. Phylogeny visualisation was carried out with iTOL 6.7.6⁵¹.

Analyses of battle in information and topology

To higher establish the potential causal components that complicate the inference of the fast radiation of elapoids, together with the focal taxa, we investigated web site concordance issue, quartet assist and examined for the presence of any deep reticulation occasions.

We have computed the location concordance issue (sCF)¹⁸ for the 50% full UCE wASTRAL-h and IQ-TREE species bushes, conventional nuclear loci phylogeny and the mitochondrial marker-based phylogeny with IQ-TREE 2.2.0.5. ASTRAL 5.7.8⁵² was used to compute the quartet assist for branches in 50% full UCE wASTRAL-h phylogeny.

We ready a 100% full UCE dataset consisting of a colubrid outgroup (Spalerosophis diadema), one of many focal genera, specifically Psammodynastes, and one consultant taxon from each different elapoid households/subfamilies. This 11-taxon dataset was ready with the phyluce pipeline. A set of unrooted ML gene bushes was estimated with IQ-TREE 2.2.0.5 with the identical parameters as with the gene bushes estimated for wASTRAL-h analyses within the earlier part. Another set of ML gene bushes, rooted by specifying the outgroup with -o command, was additionally inferred in the identical method. The unrooted gene bushes had been used as enter to PhyloNetworks⁵³ in Julia 1.9.1 to estimate the quartet concordance components (quartetCF). The quartetCFs had been used to deduce Maximum Pseudolikelihood (MPL) multispecies coalescent networks (MSCN) with the snaq! algorithm⁵⁴ of PhyloNetworks. Firstly, one MSCN with zero reticulations (therefore, basically a tree) was inferred with the quartetCFs from the unrooted gene bushes and a wASTRAL-h phylogeny computed from the identical set of gene bushes. For the following MSCN estimation, with 1 to six reticulations (hmax), we used the gene bushes set and the primary MSCN (hmax = 0) as enter. The greatest variety of hybridisations was decided by plotting the log-likelihoods from the aforementioned runs towards the hmax following the advice within the on-line guide. A second set of MPL-based MSCNs, with 0–6 reticulations, was estimated from the rooted gene bushes with PhyloNet 3.8.2⁵⁵. Unlike the snaq! of PhyloNetworks, the MPL algorithm of PhyloNet makes use of rooted triplets induced by the gene bushes to compute the MSCN⁵⁶. This methodology produced 5 MSCN per evaluation. The greatest variety of reticulations was decided by plotting the very best log-likelihood from every evaluation towards the variety of reticulations following⁵⁷.

All the aforementioned analyses had been performed on a neighborhood machine. All the MSCNs had been visualised with the PhyloPlots package deal on Julia 1.9.1.

Timetree evaluation

We estimated a time-calibrated phylogeny with the 95% UCE plus conventional nuclear loci dataset utilizing the RelTime-ML^58,59 as carried out on MEGA 11⁴¹. Log-normal calibration densities⁶⁰ used had been the identical as in⁷. We set GTR + G + I because the substitution mannequin for the evaluation. The ensuing Timetree was visualised with FigTree 1.4.4⁶¹.

Micro-CT scan and comparability of anatomical information

We have μ -CT scanned the heads of Buhoma depressiceps, B. marlieri, B. procterae and B. vauerocegae within the holdings of the Royal Belgian Institute of Natural Sciences (Brussels, Belgium) and the Royal Museum for Central Africa (Tervuren, Belgium). These embrace the holotype of B. marlieri (RMCA-VER-REP 18091). The μ-CT scan of the skull of Psammodynastes pulverulentus was obtained from MorphoSource. We additionally scanned 4 elapids, 5 lamprophiines, 4 psammophiines, one pseudaspidine and two prosymnines for comparability. We additionally examined extra μ-CT scans studied by⁷ to broaden the scope of the comparability. The μ-CT scans generated within the current examine have been uploaded to MorphoSource and are listed within the Supplementary Information (Table S6).

Several specimens had been scanned on the μCT facility of the Royal Belgian Institute of Natural Sciences. Depending on the dimensions of the specimen, scans had been carried out utilizing: a) an SimpleTom 150 (RX Solutions, Chavanod, France) with an aluminum filter at 10–30 W, 110 kV, 5.5–12.5frames/s, 1440 projections per rotation and 6–26 μm isotropic voxelsize; b) an XRE UniTom (Tescan XRE, Ghent, Belgium) at 10–22 W, 75 kV, 150–400 ms body charge, 1800 projections per rotation, and 6 to 11 μm isotropic voxelsize. Segmentation of the scans was carried out utilizing Dragonfly software program, Version 4.1 for Windows (Object Research Systems (ORS) Inc., Montreal, Canada, 2020, https://www.theobjects.com/dragonfly/index.html). Cleaning of the ensuing 3D fashions was carried out utilizing GOM Inspect (https://www.gom.com/). Terminology used for cranial bones and muscle attachment websites follows^62,63.