Rhodophyta (red algae) is one of three lineages of Archaeplastida, a supergroup that is united by the primary endosymbiotic origin of plastids in eukaryotes. Red algae are a diverse and species-rich group, members of which are typically photoautotrophic, but are united by a number of highly derived characteristics: they have relatively small intron-poor genomes, reduced metabolism and lack cytoskeletal structures that are associated with motility, flagella and centrioles. This suggests that marked gene loss occurred around their origin; however, this is difficult to reconstruct because they differ so much from the other archaeplastid lineages, and the relationships between these lineages are unclear. Here we describe the novel eukaryotic phylum Rhodelphidia and, using phylogenomics, demonstrate that it is a closely related sister to red algae. However, the characteristics of the two Rhodelphis species described here are nearly opposite to those that define red algae: they are non-photosynthetic, flagellate predators with gene-rich genomes, along with a relic genome-lacking primary plastid that probably participates in haem synthesis. Overall, these findings alter our views of the origins of Rhodophyta, and Archaeplastida evolution as a whole, as they indicate that mixotrophic feeding—that is, a combination of predation and phototrophy—persisted well into the evolution of the group.
Evolution of the Rhodelphis heme biosynthetic pathway.
Maximum likelihood phylogenetic trees were generated for all Rhodelphis heme
biosynthetic proteins including: aminolevulinate synthase (ALAS), delta aminolevulinic acid dehydratase (HemB), porphobilinogen deaminase (HemC), uroporphyrinogen-III synthase (HemD), uroporphyrinogen-III decarboxylase (HemE), coproporphyrinogen-III oxidase (HemF), protoporphyrinogen-IX oxidase (HemY), ferrochelatase (HemH). Only the first step in heme synthesis, catalyzed by ALAS, is most similar to mitochondrial homologs, and is predicted to be mitochondrial. All others, with the exception of HemH, are plastid-type proteins. The atypical HemH protein is found in some, but not all red algae, and is closely related to HemH from myxobacteria. Untrimmed and trimmed alignments used to generate the trees are also included.
Heme_Data_DRYAD.zip
Evolution of the Rhodelphis plastid FeS cluster biosynthesis pathway.
Maximum likelihood phylogenetic trees were generated for all Rhodelphis plastid iron-sulfur biosynthetic proteins including: SufB, SufC, SufD, SufE, SufS, ferredoxin, ferredoxin reductase, and NiFU. Untrimmed and trimmed alignments used to generate the trees are also included.
FeScluster_Data_DRYAD.zip
Rhodelphis retain relic plastids including subunits of conserved plastid protein import complexes.
Maximum likelihood phylogenetic trees were generated for Rhodelphis plastid Cpn60, along with conserved subunits of plastid protein import complexes. The presence of plastid import complexes is strong evidence for the retention of plastids in Rhodelphis cells. Untrimmed and trimmed alignments used to generate the trees are also included.
Tic_Toc_Cpn_Data_DRYAD.zip
Unaltered microscopic photographs of Rhodelphis cells.
Raw photographs of Rhodelphis limneticus and R. marinus that correspond to photographs in Figure 1, along with Extended Data Figures 1 and 2 are presented.
Raw_Rhodelphis_photos.zip
Individual and concatenated protein alignments for 151/253 dataset
This folder contains trimmed individual protein alignments that were used to construct the 151/253 concatenated supermatrix using SCaFoS. The resulting concatenated alignment - also included - was used to generate Figure 2a-b, along with Extended Data Figure 4.
151-253_alignments.zip
Individual and concatenated protein alignments for 153/253 dataset
This folder contains trimmed individual protein alignments tthat were used to construct the 153/253 concatenated supermatrix using SCaFoS. The resulting concatenated alignment - also included - was used to generate Extended Data Figure 3.
153-253_alignments.zip
Coalescence recovers Rhodelphis as sister to red algae based on individual gene trees
Individual bootstrapped gene trees were generated with RAxML v8.1.6 and used to generate a species tree with ASTRAL-III under default parameters and 100 bootstrap replicates. Species trees were made from either all 253 single-gene trees from the 151/253 dataset, or b, the 50 trees with the highest relative tree certainty. The sister relationship of Rhodelphis and red algae is recovered with both datasets, and is in agreement with concatenated phylogenomic analyses. The resulting coalescent trees for each dataset, along with the gene trees used to generate the species tree, are included.
Astral.zip
Concatenation of alignments from the 50 single-gene datasets with highest relative tree certainty scores recovers Rhodelphis as sister to red algae.
A maximum likelihood phylogenetic tree was generated with IQ-TREE based on a concatenated alignment of the 50 single gene trees with the highest relative tree certainty (151 taxa, 21,886 sites). The sister relationship of Rhodelphis and red algae still receives full statistical support with a highly reduced, phylogenetically well-supported dataset. This dataset represents Extended Data Figure 6.
Top50_RTC.zip
Maximum likelihood phylogenomic analyses of concatenated datasets demonstrate that Rhodelphis are sister to red algae
Maximum likelihood phylogenomic analyses were carried out for two very similar datasets: 153/253 (153 taxa, 253 proteins, 56,312 sites) and 151/253 (151 taxa, 253 proteins, 56,530 sites) using IQ-TREE under the LG+C60+F+G4 model. In each case, Rhodelphis were shown to be sister to red algae; in the 153/253 dataset, picozoa were shown to be sister to Rhodelphis + red algae. Unlike Bayesian analyses, maximum likelihood did not recover the monophyly of Archaeplastida, instead grouping Cryptista with green algae/plants and glaucophytes, though with only modest statistical support. The 153/253 tree corresponds to Extended Data Figure 3b, and the 151/253 tree corresponds to Extended Data Figure 4b.
IQ-TREE_concat.zip
Bayesian phylogenomic analyses of concatenated protein datasets recover Rhodelphis as sister to red algae
Bayesian phylogenomic analyses were carried out for two very similar datasets: 153/253 (153 taxa, 253 proteins, 56,312 sites) and 151/253 (151 taxa, 253 proteins, 56,530 sites) using PhyloBayes under the CAT+GTR+G4 framework. In each case, Rhodelphis were shown to be sister to red algae; in the 153/253 dataset, picozoa were shown to be sister to Rhodelphis + red algae. Unlike maximum likelihood analyses, Bayesian analyses recovered the monophyly of Archaeplastida.
Phylobayes_concat.zip
Rhodelphis limneticus genome assemblies and predicted protein sequenecs
This folder contains R. limneticus genomic scaffolds from mixed culture (including the prey kinetoplastid Parabodo caudatus and bacteria) and manually isolated cells amplified by whole genome amplification (WGA). In each case, the assembly is based on a combination of Illumina HiSeq X and Oxford Nanopore minION reads that were assembled using SPAdes. Contaminant contigs were identified and removed using a combination of megablast and Autometa. For the WGA dataset, proteins were predicted using braker2; protein sequences are provided with soft repeat masking and no repeat masking.
Rlimneticus_genome.zip
Rhodelphis limneticus filtered transcriptome assemblies and protein sequences
RNA-seq was performed on R. limneticus from mixed culture (including the prey kinetoplastid, Parabodo caudatus and bacteria) and from manually isolated single cells after amplification using the Smart-seq2 protocol. Paired-end reads were generated with the Illumina MiSeq platform and assembled with Trinity. In each case, contamination from prey organisms was subtracted using prey-only transcriptomes. Trinity contigs and Transdecoder protein sequences collapsed at 95% sequence identity with CD-HIT are included.
Rlimneticus_transcriptome.zip
Rhodelphis marinus filtered transcriptome assembly and predicted proteins
RNA-seq was performed on R. marinus from mixed culture (including the prey kinetoplastid, Procryptobia sorokini and bacteria). Strand-specific paired-end reads were generated with the Illumina HiSeq 2500 platform and assembled with Trinity. In each case, contamination from prey organisms was subtracted using prey-only transcriptomes. Transdecoder protein sequences collapsed at 95% sequence identity with CD-HIT are included, along with Transdecoded mRNA sequences corresponding to the protein predictions.
Rmarinus.zip
Fast site removal analyses demonstrate the robustness of the sister relationship of Rhodelphis and red algae
The fastest evolving sites in the 151/253 and 153/253 concatenated datasets were estimated with AgentSmith and removed progressively in blocks of 3,000 amino acids. Maximum likelihood phylogenies were generated under the PROT+CAT+LG+F model with 100 rapid bootstrap replicates in RAxML 8.1.6. Here we include the individual phylogenies and alignments generated by AgentSmith. The support for various relationships is depicted in Figure 2c and Extended Data Figure 3c.
FastSite.zip