Candidate Homologs of REST-Associated Chromatin Regulatory Components in Choanoflagellates

Technical Summary: Candidate Homologs of REST-Associated Chromatin Regulatory Components in Choanoflagellates

Problem Statement
REST/NRSF is a well-characterized animal transcriptional repressor that silences neuronal gene programs by binding RE1/NRSE DNA motifs via C2H2 zinc fingers and recruiting a chromatin co-repressor complex (including CoREST/RCOR, SIN3, HDACs, and LSD1). To understand the evolutionary origins of this regulatory logic, it is necessary to determine which components of this pathway predate the animal lineage. Choanoflagellates, as the closest living relatives of animals, serve as the critical outgroup for this inquiry. The study addresses two distinct questions: (1) Do choanoflagellates encode a direct ortholog of REST/NRSF? and (2) Do they encode candidate homologs of the broader protein families (DNA binders, co-repressors, chromatin modifiers) that constitute the animal REST-associated repression machinery?

Methodology
The study employed a layered computational survey across 23 choanoflagellate protein datasets, comprising three complete-genome proteomes (Monosiga brevicollis, Salpingoeca rosetta, Choanoeca flexa) and twenty transcriptome-derived sets, totaling 702,196 predicted proteins.

1. HMM-Based Screening: A curated panel of Hidden Markov Models (HMMs) was assembled from Pfam/InterPro to detect domains associated with REST function. These included:

   • DNA-binding: C2H2 zinc fingers, bHLH, homeobox, POU, Myb/SANT, forkhead, MADS, GATA, HMG box, T-box.
   • Repressor Architecture: BTB/POZ, KRAB, SCAN.
   • Co-repressor Complex: ELM2, SIN3 PAH, histone deacetylase (HDAC), LSD1-type amine oxidase, SWIRM.
   • Chromatin Readers/Modifiers: PHD, bromodomain, chromodomain, BAH, SET, SIR2.
     The panel was searched against all protein sets using HMMER with gathering thresholds.

2. Candidate Prioritization: 5,123 candidate proteins were identified. They were ranked using a scoring formula that weighted the presence of specific domains (e.g., C2H2, BTB/POZ, corepressor domains) and the annotation of the best Swiss-Prot hit (e.g., "REST," "HDAC," "transcriptional repressor"). Candidates were stratified into Low, Moderate, High, and Priority tiers.

3. Sequence Verification: Candidates were verified via DIAMOND searches against Swiss-Prot (E-value < 1e-5) and a secondary BLASTP check on the top 50 priority candidates. HH-suite was noted as a future extension but not utilized due to database constraints.

4. Structural Preparation: AlphaFold-ready FASTA files were generated for 40 priority candidates. Reference structures for their top Swiss-Prot homologs were retrieved from the AlphaFold DB (yielding 12 unique UniProt reference structures) to facilitate future structural comparisons.

5. Motif Analysis: A promoter scan (up to 1 kb upstream of genes) was conducted on the three full-genome assemblies for the canonical RE1/NRSE motif (TTCAGCACCACGGACAGCGCC), allowing up to five mismatches. An analytical GC-matched null model was computed to assess enrichment, calculating Poisson tail probabilities and Benjamini-Hochberg FDR values.

Key Results

• Domain Prevalence: The survey identified 8,299 domain hits across 5,123 candidates. Families such as C2H2 zinc fingers, BTB/POZ, HDACs, SIN3 PAH, and various chromatin readers were detected in all 23 samples, indicating a broad regulatory toolkit in choanoflagellates.
• Absence of REST/NRSF Orthologs: No candidate protein returned a convincing best-hit annotation to NRSF/REST. The study found no evidence of a direct REST/NRSF ortholog in the surveyed datasets.
• Presence of Co-Repressor Candidates: The strongest signal was the identification of candidate homologs for the machinery recruited by REST, rather than the repressor itself.
  • HDAC/SIN3: Histone deacetylase-like and SIN3-like PAH-domain proteins dominated the top priority tiers.
  • RCOR/CoREST: Five candidates contained ELM2 and/or Myb-SANT domains with best Swiss-Prot hits to RCOR1/2/3 or "REST corepressor" proteins. These are treated as provisional homologs, as ELM2/SANT domains are widespread and orthology requires further phylogenetic validation.
  • YY1-like Repressors: Ten C2H2 zinc-finger candidates matched human YY1-like transcriptional repressors, suggesting the presence of sequence-specific repressors distinct from REST.
• Motif Enrichment: 717 loosely matching RE1-like sequences were found. The GC-matched null model indicated enrichment at relaxed thresholds (4 and 5 mismatches) in S. rosetta and C. flexa. However, the strictest threshold (3 mismatches) yielded only marginal evidence (e.g., 2 hits in S. rosetta with FDR = 0.057), suggesting these are candidate sites rather than confirmed regulatory elements.
• Structural Data: 12 unique reference structures were prepared for future comparison against predicted choanoflagellate candidate structures.

Significance and Claims
The paper concludes with a conservative interpretation: choanoflagellates encode candidate homologs of several protein families that participate in animal REST-associated repression (specifically HDACs, SIN3-like proteins, provisional RCOR/CoREST-like proteins, and C2H2 repressors). However, the study explicitly does not claim that these components function together as an integrated REST-like pathway in choanoflagellates.

The primary significance lies in distinguishing between the presence of individual components and the existence of a functional pathway. While the "parts list" for REST-associated repression appears to be largely pre-animal, the study finds no evidence for a direct REST/NRSF ortholog or a confirmed integrated complex in choanoflagellates. The identification of provisional RCOR/CoREST-like candidates is highlighted as the most direct link to the animal REST complex, warranting future phylogenetic and structural investigation. The motif results are presented as exploratory leads for future regulatory analysis rather than proof of functional binding.

Limitations Noted by the Author

• Sequence similarity does not equate to biochemical function.
• Structural comparisons were prepared but not executed (no local AlphaFold predictions were generated).
• Motif analysis used a simple GC-matched null model, lacking controls for phylogenetic conservation or intergenic background.
• Transcriptome-derived sets may contain artifacts or isoforms that complicate copy-number interpretation.
• RCOR/CoREST orthology remains provisional pending reciprocal similarity and phylogenetic reconstruction.