GC high

Deep research

Generated by just research-trait; source: research/traits/genomics/gc_high-deep-research-falcon.md

# Curation report: microbial genomic low-GC phenotype

## Executive summary

The trait identifier must be quoted exactly as **METPO:1000432**. Despite its upstream label, **“GC high,”** the authoritative synonym and numerical definition describe the opposite phenotype: **whole-genome GC content ≤ approximately 42.65% (`GC_<=42.65`)**. For `data/traits/genomics/gc_high.yaml`, the numeric bin should control interpretation, while the misleading label should be retained only as provenance.

This is a **genome-composition class**, not a physiological activity. The strongest causal route supported by experiments is:

> spontaneous cytosine deamination and guanine oxidation → GC-to-AT/TA substitutions; loss of the corresponding repair capacity amplifies these substitutions → long-term decline in genome-wide GC.

The broader literature supports a systems-level model in which the composition of DNA replication and repair (DRR) machinery, phylogenetic history, mutation bias, recombination-associated GC-biased gene conversion, drift, and selection jointly determine genomic GC. Direct environmental adaptation to low GC is not established as a universal mechanism.

## 1. Trait scope and current understanding

### 1.1 Operational definition

**Recommended curation definition:** “A microbial genome-composition phenotype in which G+C bases constitute no more than approximately 42.65% of the complete or representative genome sequence.”

The phenotype should be calculated as:

\[
GC\% = 100\times\frac{G+C}{A+T+G+C}
\]

preferably over a complete, high-quality whole-genome assembly. The bin is somewhat stricter than the broad “low-GC” grouping used in recent comparative work, where most low-mode genomes occur below 45%. In 11,083 representative bacterial genomes, GC ranged from about 16% to 77% and was bimodal, with most genomes below 45% or above 60%; more than 60% of variance was explained at phylum level, with Blomberg’s K=1.47 and Pagel’s λ=0.998. Thus, 42.65% is an ontology-specific discretization, not a universal biological breakpoint (teng2023genomiclegaciesof pages 2-5).

### 1.2 Boundaries and nearby traits

The trait is **not equivalent to**:

- **GC3:** GC fraction at third codon positions. GC3 is especially responsive to synonymous substitution and codon usage and can differ substantially from whole-genome GC. Recombination studies often analyze GC3 rather than genomic GC (lassalle2015gccontentevolutionin pages 11-14).
- **Coding-sequence GC, noncoding GC, or local GC windows:** these can identify islands, horizontally transferred regions, or strand effects but do not alone establish the whole-genome bin. Teng et al. explicitly separated whole-genome GC, coding GC, noncoding GC, amino-acid-contributed GC, and codon-contributed GC (teng2023genomiclegaciesof pages 2-5).
- **GC skew:** strand asymmetry such as `(G−C)/(G+C)`; this concerns replication/transcription asymmetry rather than total composition. Whole bacterial genomes can be compositionally homogeneous while retaining strand-specific biases (lind2008wholegenomemutationalbiases pages 1-1).
- **An AT-biased mutation spectrum:** mutation bias is an upstream process. A currently high-GC genome may have AT-biased new mutations because equilibrium composition changes over long evolutionary periods (teng2023genomiclegaciesof pages 8-10, lassalle2015gccontentevolutionin pages 11-14).
- **Genome reduction:** reduced genomes are frequently AT-rich, especially in endosymbionts, but genome size and GC percentage are separate traits. Neither implies the other universally.
- **Low-GC Gram-positive bacteria:** an historical taxonomic description, not a mechanistic or phylogenetically exclusive class. Low-GC clades occur in multiple bacterial groups (teng2023genomiclegaciesof pages 2-5).

### 1.3 Expert synthesis

Recent authoritative analysis favors **indirect evolution through replication/repair systems and historical contingency**, rather than a single adaptive advantage of low GC. A phylogenetically informed model based on 217 DRR-related KEGG orthologs explained 88% of observed GC variance (multiple correlation 0.94); however, this is predictive comparative evidence, not proof that each correlated gene causes the phenotype (teng2023genomiclegaciesof pages 2-5). Figure 3 of that study shows the model fit and opposing associations of DnaE2 and MutS2, as well as pathway-level correlations involving BER, MMR, replication, recombination, and translesion synthesis (teng2023genomiclegaciesof media 5c6ce460).

## 2. Candidate graph nodes

### Trait and measurement nodes

- **Low whole-genome GC content:** **METPO:1000432**
- Parent trait: **METPO:1000127**
- Whole-genome GC percentage — label-only assay/measurement node
- GC3, coding-sequence GC, noncoding-sequence GC, local GC, and GC skew — label-only boundary/measurement nodes

### Genes, proteins, and complexes

- **MutM/Fpg DNA glycosylase**, **MutY adenine glycosylase** — repair oxidized guanine-associated lesions
- **Ung** and **Mug** uracil-DNA glycosylases — remove uracil arising from cytosine deamination
- **Vsr endonuclease** — very-short-patch repair of G:T mismatches
- **MutS/MutL mismatch-repair system** — canonical MMR; direction of compositional effect is taxon dependent
- **MutS2** — MutS homologue; do not conflate automatically with canonical MutS-directed MMR
- **NucS/EndoMS** — noncanonical mismatch-repair endonuclease in certain archaea and actinobacteria; potentially relevant but presently not supported as a universal low-GC determinant
- **DnaE/Pol III α**, **PolC**, **DnaE2**, **DinB/Pol IV**, and **Pol V** — replicative or error-prone/translesion polymerases
- **RecA/RuvC and homologous-recombination machinery** — candidates connecting recombination and gene conversion

Showing the first 60 of 200 lines of findings; the linked file also carries the run's front matter and the prompt it was given — read the full report.