Resurrecting Full-length Ancestral Schizorhodopsins and Heliorhodopsins with Structure-guided, Indel-aware Sequence Reconstruction

This study successfully reconstructs and experimentally validates full-length ancestral schizorhodopsins and heliorhodopsins using a novel structure-guided, indel-aware sequence reconstruction method that overcomes alignment ambiguities to produce stable, retinal-binding proteins, thereby enabling direct investigation of how extra-membrane architectures evolve alongside the conserved transmembrane core.

The Gist

Imagine you are a detective trying to solve a mystery that happened millions of years ago. Your suspects? Ancient proteins called microbial rhodopsins. These are tiny, light-sensing machines found in bacteria and archaea that act like solar panels, pumps, or sensors.

The problem is, these proteins have changed so much over time that it's hard to guess what their original, "ancestor" forms looked like. It's like trying to rebuild a 1920s Model T Ford just by looking at a pile of modern sports cars, a few rusted parts, and a blurry photo.

This paper is about a team of scientists who successfully rebuilt two of these ancient proteins: Anc-SzR (an ancestor of Schizorhodopsins) and Anc-HeR (an ancestor of Heliorhodopsins). They didn't just guess; they used a high-tech, step-by-step method to resurrect them in a lab, and the results were stunning.

Here is how they did it, explained with some everyday analogies:

1. The Problem: The "Missing Pieces" Puzzle

Scientists have found thousands of these modern proteins. To figure out what the ancestors looked like, they usually try to line them up like a game of "Telephone" to see how they changed.

However, there was a major glitch. These proteins have a rigid, 7-part core (like a sturdy 7-story building), but the "floors" and "balconies" sticking out (the extra parts) are messy and full of holes.

• The Old Way: Previous methods would just chop off the messy parts and only try to rebuild the sturdy 7-story core. It was like trying to understand a house by only studying the foundation and ignoring the roof, windows, and front door.
• The Glitch: When they tried to guess the missing "balconies," the computer kept adding too many extra bricks, making the ancestors look like giant, bloated monsters with long, floppy tails that didn't make sense.

2. The Solution: The "Indel-Aware" Detective Work

The authors developed a new method called ConsistASR. Think of this as a super-smart editor that knows exactly where the "missing pieces" (gaps) should be.

• The Analogy: Imagine you are trying to reconstruct a torn family photo album.
  • Old Method: You guess where the tears are and just paste the photos together, often stretching the paper until it rips or looks weird.
  • New Method (Indel-Aware): The detective looks at the pattern of the tears. They realize, "Ah, this family always lost a page here, and this one always gained a page there." They use that logic to cut out the extra paper and fill in the gaps perfectly.
• The Result: Instead of bloated monsters, they got compact, realistic ancestors that looked just like the modern versions, but with the right "flavor" of the past.

3. The "Crystal Ball" Check (AlphaFold)

How do you know your reconstruction is right if you can't see the past? The scientists used AlphaFold, a super-computer AI that predicts what a protein looks like in 3D.

• The Analogy: It's like using a weather forecast to check if your reconstructed house makes sense. If you built a house with a roof made of jelly, the weather forecast (AlphaFold) would scream, "That won't hold up!"
• The Check: The AI looked at their reconstructed ancestors and said, "Yes! These look stable. They have the right 7-story core, and the messy balconies actually form nice little structures (like tiny stairs or loops) instead of just floppy strings."

4. The Big Reveal: Bringing Them Back to Life

This is the coolest part. Most of these "ancestral reconstruction" studies stay on the computer screen. These scientists actually built the proteins in a lab.

• The Process: They took the digital code they created, put it into bacteria (like E. coli), and let the bacteria act as tiny factories.
• The Result: The bacteria started making the ancient proteins. When they looked at the bacteria, they turned purple/red.
• Why? Because these proteins are "rhodopsins," which means they love to grab a molecule called retinal (the same stuff in your eyes that lets you see). The fact that the ancient proteins grabbed the retinal and turned color proved they were real, functional, and folded correctly.

5. Why Does This Matter?

This paper is a breakthrough for three reasons:

1. We Don't Have to Chop: It proves we don't need to throw away the "messy" parts of proteins to study them. We can reconstruct the whole thing, including the parts that stick out.
2. Evolution in Action: They found that the "balconies" (the extra parts) changed in specific ways. One group of ancestors developed a tiny spiral staircase, while another group developed a long bridge. This helps us understand how nature invents new tools.
3. Reliability: They created a new "trust score" system. It combines the math (how likely is this sequence?), the structure (does the AI think it's stable?), and the family tree (does it fit the history?). This helps scientists know which ancient proteins are safe to build in the lab.

The Bottom Line

The scientists successfully time-traveled. They took a jumbled mess of modern data, used a clever new method to fix the missing pieces, checked their work with a super-AI, and then built the ancient proteins in a petri dish. The proteins worked perfectly, proving that we can now "resurrect" complex ancient machines and study how they evolved, one tiny piece at a time.

Technical Summary

1. Problem Statement

Ancestral Sequence Reconstruction (ASR) is a powerful tool for inferring the sequences of extinct proteins to study evolutionary history. However, applying ASR to seven-transmembrane (7TM) proteins (like microbial rhodopsins) faces significant challenges:

• Alignment Ambiguity: 7TM proteins have conserved transmembrane (TM) cores but highly variable extra-membrane (EM) loops and termini. Standard multiple sequence alignments (MSAs) often struggle with these gap-rich EM regions.
• Indel Uncertainty: Traditional ASR workflows often treat insertions and deletions (indels) implicitly or ignore them, leading to artificially overextended ancestral sequences. These "inflated" ancestors often contain long, low-confidence tails that do not fold correctly, making experimental resurrection difficult.
• Experimental Tractability: Most studies focus on "trimmed" TM cores, discarding EM regions. This prevents the experimental testing of how full-length architecture, including lineage-specific secondary structures in EM regions, evolves.

The authors aimed to reconstruct full-length ancestral schizorhodopsins (SzR) and heliorhodopsins (HeR)—two families with distinct topologies and oligomeric states—to test if a robust, indel-aware pipeline could produce experimentally viable ancestors.

2. Methodology: The ConsistASR Pipeline

The authors developed a novel workflow called ConsistASR (Consistency-aware ASR) that integrates evolutionary modeling with structural validation.

• Sequence Curation & Alignment:
  • Compiled a dataset of 228 SzR, HeR, and outgroup sequences.
  • Generated MSAs using two strategies: PSI/TM-Coffee (structure-aware, incorporating transmembrane constraints) and MAFFT L-INS-i (high-accuracy standard).
• Phylogenetic Inference:
  • Used IQ-TREE with Q.pfam+R profile models (membrane-aware) to infer maximum-likelihood trees.
  • Selected the best-fit substitution model using the Bayesian Information Criterion (BIC).
• Indel-Aware Refinement (The Core Innovation):
  • Two-Stage Process: Instead of ignoring gaps, the pipeline explicitly models them.
    1. Amino Acid ASR: Infer ancestral amino acid states in IQ-TREE.
    2. Binary Gap ASR: Recode the alignment as a binary matrix (1=residue, 0=gap) and infer ancestral gap states using RAxML on the fixed topology.
  • Node-Specific Masking: Merge the amino acid states with the inferred binary gap states. Sites inferred as gaps for a specific node are masked (removed), while residue sites are retained. This prevents the "inflation" of ancestral sequences.
• Structural Validation:
  • Predicted structures using AlphaFold3 (monomeric and oligomeric).
  • Evaluated confidence using pLDDT (per-residue confidence) and mapped it against Posterior Probabilities (PP) from ASR.
  • Defined a Composite Reliability Score: $Score = 100 \times b \times PP \times pLDDT$, where $b$ is the bootstrap support. This integrates evolutionary, statistical, and structural confidence.
• Experimental Resurrection:
  • Synthesized and expressed the indel-corrected Anc-SzR and Anc-HeR in E. coli with minimal tags (His6).
  • Added all-trans retinal to test holoprotein formation.
  • Characterized spectral properties via UV-Vis absorption spectroscopy.

3. Key Contributions

1. Full-Length Reconstruction: Successfully reconstructed compact, full-length ancestors for SzR and HeR, including complex EM secondary structures (e.g., $\beta$-strands and short helices), rather than just the TM core.
2. Indel-Aware Refinement: Demonstrated that explicit binary gap modeling is critical. Without it, ancestors were 400+ residues (inflated) with low confidence; with it, they contracted to ~200–260 residues with high structural confidence.
3. Integration of Metrics: Proposed a composite metric combining Bootstrap Support (BS), Posterior Probability (PP), and AlphaFold pLDDT to identify "conditionally reliable" ancestors (e.g., Anc-HeR has high structural confidence but alignment-sensitive topology).
4. Experimental Validation: Proved that full-length ASR sequences, without stabilizing mutations or truncations, can fold into stable, colored, retinal-binding holoproteins in E. coli.

4. Key Results

• Impact of Indel Correction:
  • Uncorrected: Anc-SzR and Anc-HeR were ~400 residues long with mean pLDDT of ~60–70 (low confidence) and inflated EM tails.
  • Corrected: Lengths contracted to 206 (Anc-SzR) and 259 (Anc-HeR) residues, closely matching extant proteins (202 and 256 residues, respectively). Mean pLDDT rose to >93, and mean PP increased to >80%.
• Structural Plausibility:
  • AlphaFold models of corrected ancestors formed compact 7TM folds with high confidence.
  • Lineage-Specific Motifs: The pipeline correctly recovered lineage-specific EM structures: Anc-HeR retained long $\beta$-strands (TM1-TM2) and a short helix (TM2-TM3), while Anc-SzR retained short $\beta$-strands (TM2-TM3) and a loop (TM1-TM2).
• Oligomeric State Prediction:
  • AlphaFold-Multimer predicted Anc-SzR as a trimer and Anc-HeR as a dimer, matching the known oligomeric states of extant proteins.
• Experimental Success:
  • Both Anc-SzR and Anc-HeR expressed in E. coli as stable, pigmented proteins.
  • Spectral Properties: Anc-SzR showed $\lambda_{max}$ at 549 nm; Anc-HeR at 543 nm. These values are consistent with their respective clades, confirming functional retinal binding.
• Reliability Analysis:
  • Anc-SzR: Highly robust across alignment methods (high BS, PP, and pLDDT).
  • Anc-HeR: High structural confidence (PP/pLDDT) but alignment-sensitive topology (BS dropped to 0% with MAFFT L-INS-i). This highlights the need for composite scoring.
  • Anc-SH (Common Ancestor): Topologically stable but with lower residue-level certainty (diffuse PP), typical of deep nodes.

5. Significance

• Paradigm Shift in 7TM ASR: This work challenges the convention of "trimming" 7TM proteins. It demonstrates that full-length architecture, including complex EM loops, can be reconstructed with high confidence and is essential for understanding evolutionary mechanisms like oligomerization and topology shifts.
• Methodological Framework: The ConsistASR pipeline provides a reproducible, scriptable framework for future ASR studies of membrane proteins, explicitly handling indels and validating results with structural AI.
• Evolutionary Insights: The study reveals that EM secondary structures (like $\beta$-strands) are evolutionary characters that can be gained or lost along specific lineages (e.g., the HeR lineage gained long $\beta$-strands), providing a new layer of evolutionary data beyond the TM core.
• Biological Feasibility: The successful resurrection of functional, full-length ancestors proves that computational predictions of ancient membrane proteins can directly translate to wet-lab experiments, opening doors to studying the evolution of GPCRs and other complex membrane systems.