Re-evaluating the eukaryotic Tree of Life with independent phylogenomic data

By reconstructing the eukaryotic Tree of Life using a largely independent, marker-rich dataset that minimizes shared biases with previous studies, this research confirms most major supergroups while proposing revised phylogenetic positions for specific lineages, thereby supporting the hypothesis that extant eukaryotic diversity is organized into a small number of high-rank supergroups.

The Gist

Imagine the history of life on Earth as a massive, ancient family tree. For over a century, scientists have been trying to draw this tree, specifically for eukaryotes—the complex life forms that include everything from humans and mushrooms to tiny algae and amoebas.

For the last 20 years, scientists have been using a specific set of "family heirlooms" (a standard list of about 300 proteins) to figure out who is related to whom. They've built a pretty good map, grouping these organisms into big "supergroups." But, there's a problem: everyone has been using the exact same set of heirlooms.

If those heirlooms were slightly damaged or biased, the whole family tree could be wrong. It's like trying to solve a mystery where every detective is reading from the same biased witness statement.

The New Approach: A Fresh Set of Clues

In this new study, the researchers decided to stop using the old, overused heirlooms. Instead, they went to a completely different library to find a brand new set of 277 proteins (called BUSCO markers) that no one had really used for this big picture before.

Think of it this way:

• The Old Way: Imagine trying to identify a suspect by only looking at their shoes. Everyone agrees on the shoe style, but maybe the shoes are misleading.
• The New Way: The researchers decided to look at the suspect's fingerprints, DNA, and voice patterns instead. They used a dataset that shares almost no overlap with the old one.

What They Found

When they built the family tree with these new clues, they found that the big picture was mostly correct, but some specific branches were surprisingly different.

1. The "Glissogyra" Family (The Sticky Swimmers)
They discovered a strong new family unit called Glissogyra. It's a marriage between two groups of single-celled organisms (Ancyromonadida and Malawimonadida) that were previously thought to be distant cousins.

• The Analogy: It's like realizing that two neighbors who look very different actually share a great-grandparent you didn't know about. They even share a unique "family trait" (a specific type of DNA tool) that proves they belong together.

2. The "Telonemia" Switch
For years, scientists thought a group called Telonemia was related to the massive "SAR" supergroup (a huge cluster of algae and parasites).

• The New Twist: The new data says, "Actually, no!" Telonemia is much more closely related to Haptophytes (a type of algae).
• The Analogy: Imagine you thought your neighbor was related to the family down the street because they both drove red trucks. But when you looked at their birth certificates (the new data), you realized they were actually related to the family across town who both love gardening. The red trucks were a coincidence; the gardening was the real link.

3. The "Long Branch" Trap
Some groups of organisms evolve very fast, making their DNA look very different from everyone else. In the old trees, these fast-evolving groups kept getting glued together incorrectly, like two magnets that just happen to stick because they are both "weird."

• The Fix: By removing the "noisy" parts of the data and using better math, the researchers showed that these fast groups (like Discoba and Metamonada) aren't actually sisters. They are just two different families that happen to have messy, fast-changing DNA.

Why This Matters

The most important takeaway isn't just the new names or the shuffling of branches. It's that the big picture holds up.

Even with a completely different set of data, the major supergroups (like plants, animals, and fungi) still look the same. This gives us confidence that the "Tree of Life" we've been building is real and not just a trick of the data we've been using.

However, the study also shows that the deep, ancient roots of the tree are still a bit foggy. It's like looking at a mountain range in the fog: you can clearly see the big peaks (the major groups), but the small valleys between them are still hard to map.

The Bottom Line

This paper is a "second opinion" from a different doctor. The first doctor (the old studies) gave a diagnosis that was mostly right. This new doctor (the new study) used a different set of tests and confirmed the main diagnosis but corrected a few specific details about who is related to whom.

It proves that to truly understand our biological family tree, we need to keep asking questions with fresh eyes and new data, rather than just repeating the same old answers.

Technical Summary

1. Problem Statement

Understanding the deep phylogenetic relationships among eukaryotic lineages is critical for reconstructing the evolution of the Last Eukaryotic Common Ancestor (LECA) and mapping the emergence of key phenotypic traits. However, the current consensus on the Eukaryotic Tree of Life (eToL) relies heavily on a limited set of "legacy" protein marker datasets (e.g., those derived from Strassert et al. and Tice et al.).

• Systematic Bias: These legacy datasets share a high degree of marker overlap (~66–87%) and are enriched in ribosomal proteins. Ribosomal proteins are known to have distinct amino acid compositions (rich in Arginine and Lysine) due to structural constraints, potentially introducing compositional bias and systematic errors (such as Long Branch Attraction) into phylogenetic reconstructions.
• Uncertainty: Despite the identification of major "supergroups" (e.g., SAR, Amorphea, Diaphoretickes), deep relationships remain unstable. It is unclear whether current topologies reflect true evolutionary history or are artifacts of the specific markers and models used.

2. Methodology

The authors constructed a largely independent, marker-rich phylogenomic dataset to test the robustness of existing eToL hypotheses.

• Dataset Construction (BUSCO-based):

  • Source: Instead of legacy markers, the authors mined the BUSCO (Benchmarking Universal Single-Copy Orthologs) Eukaryota odb9 collection (303 universal genes).
  • Curation: They identified 405 orthologous markers, manually curated them to remove paralogs, contaminants, and sequences with non-vertical evolutionary histories, and retained 277 high-quality markers.
  • Independence: This new dataset shares <25% overlap with previous major datasets (Strassert21 and Tice21), with 77% of markers being unique to this study.
  • Compositional Balance: The new dataset contains significantly fewer ribosomal proteins (2.1% vs. ~19% in legacy sets), resulting in an amino acid composition that closely mirrors the whole proteome average, thereby minimizing compositional bias.

• Taxon Sampling:

  • Assembled a comprehensive dataset of 741 eukaryotic proteomes from EukProt v2/v3 and other sources.
  • Applied rigorous quality control (removing prokaryotic contamination and low-coverage proteomes), resulting in a final set of 651 high-quality taxa.
  • Created subsamples (264 taxa for ML robustness tests; 61 taxa for Bayesian Inference) to manage computational costs.

• Phylogenetic Inference:

  • Models: Used the ELM+C60+G4 substitution model (specifically designed for pan-eukaryotic data) and the CAT+GTR+G4 model (Bayesian Inference) to account for site-heterogeneity and reduce Long Branch Attraction (LBA).
  • Robustness Analyses:
    1. Site Filtering: Incremental removal of the fastest-evolving sites (10–90%) to test for mutational saturation.
    2. Marker Subsampling: Random resampling of 20% and 60% of markers to test signal consistency.
    3. Dataset Comparison: Direct comparison of trees built from the new "Leroy-specific" (LS) markers vs. the "Strassert-Tice" (ST) legacy markers.

3. Key Results

A. Validation of Major Supergroups

The new dataset strongly supports the monophyly of most established high-rank eukaryotic groups, including SAR (Stramenopiles, Alveolata, Rhizaria), Obazoa, CRuMs, Podiata, Opimoda, and Diaphoretickes. This confirms that the "few-supergroup" structure of the eToL is robust to independent marker sets.

B. Novel and Revised Topologies

• Glissogyra (New Supergroup): The study provides robust support for a monophyletic group uniting Ancyromonadida and Malawimonadida at the base of the Opimoda. This clade is proposed as a new supergroup named Glissogyra. Both groups share a unique DNA polymerase (rdxPolA), suggesting a synapomorphy.
• Telonemia and Haptophyta: Contrary to the widely accepted "TSAR" hypothesis (Telonemia + SAR), the new data consistently places Telonemia as the sister group to Haptophyta with high support. This relationship holds even after removing fast-evolving sites.
• Excavata and LBA Artifacts: The monophyly of "Excavata" (specifically the grouping of Metamonada and Discoba) collapses when fast-evolving sites are removed or when using the CAT+GTR+G4 model (less prone to LBA). The authors conclude that the grouping of these long-branch lineages is likely an artifact of Long Branch Attraction. Instead, Discoba appears as the sister to Diaphoretickes within the "Diphoda+" clade.
• Amoebozoa Position: The position of Amoebozoa remains sensitive to data sampling. While the full dataset and site-filtered data support Amoebozoa as sister to Obazoa (Amorphea), marker subsampling sometimes supports Amoebozoa as sister to CRuMs.

C. Compositional Bias Effects

• Picozoa: The placement of Picozoa was found to be sensitive to compositional bias. In analyses including Telonemia, Picozoa was grouped with Telonemia/Haptophyta due to similar amino acid compositions. When Telonemia was removed, Picozoa correctly grouped within Archaeplastida.
• Haptista: The monophyly of Haptista (Haptophyta + Centroplasthelida) was recovered in legacy datasets but broken in the new LS dataset, suggesting convergent evolution of feeding structures rather than shared ancestry.

4. Significance and Contributions

• Independence from Legacy Bias: This study demonstrates that the core structure of the eToL is not an artifact of the specific ribosomal-protein-heavy datasets used for the past two decades. The independent BUSCO-derived dataset validates the major supergroups while resolving deep nodes differently.
• Refining Deep Nodes: The study challenges the "TSAR" hypothesis and the monophyly of "Excavata," proposing a split between Opimoda+ (including Metamonada) and Diphoda+ (including Discoba and Diaphoretickes).
• Taxonomic Proposal: The formal description of Glissogyra (Ancyromonadida + Malawimonadida) as a new basal opimodan supergroup adds a critical piece to the early eukaryotic radiation puzzle.
• Methodological Insight: The results highlight that deep eukaryotic relationships are highly sensitive to marker choice and compositional heterogeneity. The study underscores the necessity of using independent, compositionally balanced datasets and sophisticated models (like CAT+GTR) to distinguish historical signal from systematic artifacts.

Conclusion

Leroy et al. provide a robust, independent reconstruction of the eukaryotic Tree of Life. While confirming the broad validity of the supergroup paradigm, the study significantly revises the placement of key lineages (Telonemia, Glissogyra, Excavata), suggesting that previous topologies were influenced by systematic biases in legacy datasets. The findings support a model of eukaryotic diversity dominated by a small number of high-rank supergroups, with a major early split separating Opimoda from Diphoda.