Evidence for strong purifying selection of human 47S ribosomal RNA genes

By analyzing over 3,000 human genomes, this study demonstrates that strong purifying selection suppresses deleterious mutations in conserved regions of the multicopy 47S rRNA genes, explaining why disease-causing variants in these highly expressed genes have historically remained undetected in association studies.

The Gist

The Mystery of the "Silent" Ribosomes

Imagine your body is a massive, bustling factory. The workers in this factory are called ribosomes, and their job is to build proteins, which are the bricks and mortar of your body. To do their job, these workers need a set of blueprints. These blueprints are made of rRNA genes.

Here is the weird part: Your factory has hundreds of copies of these blueprints (about 300 to 400 per cell). Usually, when you have hundreds of copies of something, if one copy gets a typo, it doesn't matter much because the other hundreds are perfect. It's like having 400 copies of a recipe for chocolate cake; if one page has a smudge, you just use the others.

Because of this, scientists have long been puzzled: Why don't we see diseases caused by typos in these rRNA genes? We know that typos in other parts of our genetic code cause diseases, but for these ribosome blueprints, it seems like nothing goes wrong.

The Detective Work

The researchers in this paper decided to play detective. They looked at the DNA of over 3,000 people from the famous "1,000 Genomes Project." They didn't just look for the big, obvious typos; they looked for the tiny, rare ones that might only exist in a few of the hundreds of copies in a single person.

They found something surprising: There are actually thousands of typos (variants) in these genes. In fact, they found over 14,000 different variations across the people they studied.

The "Natural Selection" Filter

So, if there are so many typos, why aren't people getting sick?

The answer is Purifying Selection. Think of this as a strict quality control manager in your factory.

1. The "Safe" Zones (Spacers): Some parts of the blueprint are like the margins of the page or the decorative borders. If you scribble in the margins, the recipe still works. The researchers found that typos in these "safe" areas were common and often copied around to many of the 400 blueprint copies.
2. The "Critical" Zones (The Core): Other parts of the blueprint are the actual instructions for mixing the ingredients. If you change a word here, the cake burns. The researchers found that in these critical zones, typos were extremely rare.

The Big Discovery: The researchers realized that if a "bad" typo happens in a critical zone, the body's quality control system immediately tries to get rid of it. It doesn't wait for the typo to spread to all 400 copies. It suppresses it so aggressively that the typo stays at a very low level (maybe just 1 or 2 copies out of 400).

The "Needle in a Haystack" Problem

This explains the mystery!

• The Old Way: When doctors look for genetic diseases, they usually look for typos that are present in many copies of the gene. They assume that if a typo is rare (only in 1 or 2 copies), it's probably harmless noise.
• The New Reality: This paper shows that the "harmful" typos are being kept at a low count on purpose by evolution. They are the "needles in the haystack." Because current medical tests ignore these low-count typos, we have been missing the connection between rRNA errors and disease.

The "Poisoned" Factory Analogy

The authors suggest a scary but fascinating idea: Ribosomes are so abundant and work so hard that even a few "poisoned" ones can ruin the whole batch.

Imagine if 5% of the workers in your factory started building the wrong product. Even if 95% are perfect, the 5% that are broken might clog the machines or produce toxic waste that hurts the whole factory.

The study suggests that if a person has a "bad" rRNA variant that manages to sneak past the quality control and reach a slightly higher number of copies, it could cause disease. But because these variants are usually kept very low by nature, we haven't been able to spot them until now.

Summary in One Sentence

This paper proves that our bodies have a super-strong "immune system" for our genetic blueprints that ruthlessly hunts down and suppresses bad errors in the most important parts of the ribosome, which is why we haven't found rRNA-related diseases yet—we just haven't been looking closely enough at the tiny, hidden errors.

Technical Summary

1. Problem Statement

The human genome contains hundreds of copies of the 47S ribosomal RNA (rRNA) gene family, which are among the most highly expressed genes. Despite their critical role in protein synthesis and the known fact that defects in mitochondrial rRNA or ribosomal protein dosage cause severe diseases, no disease-causing sequence variants have been identified in human 47S rRNA genes.

This absence of disease association is paradoxical. The authors hypothesize that this gap in knowledge stems from two main factors:

1. Experimental limitations: The high-copy nature and chromosomal isolation of rDNA make analyzing sequence variants difficult.
2. Methodological bias: Standard disease-mapping algorithms typically ignore low-frequency variants (those present in only a small percentage of sequencing reads), assuming they are noise or artifacts. The authors propose that deleterious rRNA variants might exist at very low copy numbers but still cause significant phenotypic consequences.

2. Methodology

To test the hypothesis that strong purifying selection suppresses deleterious rRNA variants, the authors employed an evolutionary genomic approach using high-coverage data.

• Data Source: They analyzed high-coverage (30x) whole-genome sequencing (WGS) data from 3,190 individuals in the 1,000 Genomes Project.
• Mapping & Variant Calling:
  • Reads were mapped to a prototype human rDNA sequence (GenBank: U13369.1) using Bowtie2.
  • A custom pipeline (adapted from yeast rDNA analysis) was used to calculate Intragenomic Variant Frequencies (iVFs), defined as the fraction of reads containing a variant at a specific position.
  • Filtering:
    • Variants in homopolymers (>9 nucleotides) were excluded to avoid sequencing errors.
    • Variants likely originating from pseudogenes were filtered out. These were identified as variants shared across >5% of samples but consistently showing very low iVFs (<0.07) without copy-number amplification.
  • Benchmarking: The pipeline was validated using synthetic data (NEAT toolkit), confirming reliable detection of variants with iVFs $\ge$ 1%.
• Analytical Framework:
  • Heritability Analysis: Hypergeometric tests were used to determine if shared variants were population-specific (indicating heritability) or random somatic mutations.
  • Consensus Prototype: A new consensus 47S prototype was constructed based on the major alleles across the dataset to correct for reference bias.
  • Variation Scoring: A "variation score" was calculated for different gene regions by multiplying the fraction of variant nucleotides by the median iVF, capturing both mutational incidence and amplification.
  • Regional Classification: The 28S rRNA was divided into three classes: Conserved Nucleotide Elements (CNEs) (highly conserved across eukaryotes), Expansion Segments (evolutionarily young, variable), and the Rest of 28S.

3. Key Contributions

• Comprehensive Variant Catalog: Identified 14,878 unique 47S rDNA variants across >3,000 individuals, covering >61% of the 47S sequence.
• Reference Correction: Generated a new, robust consensus 47S prototype sequence that eliminates reference-driven inflation of variant frequencies.
• Methodological Adaptation: Successfully adapted a pipeline to detect and quantify low-frequency variants (iVF < 1%) in a high-copy gene family, a task previously hindered by technical noise.
• Evolutionary Insight: Provided the first evidence that purifying selection in human rDNA acts not just on the presence of mutations, but specifically on their copy number amplification.

4. Key Results

• Stratified Variant Distribution:
  • Variants were unevenly distributed. Spacer regions (5' ETS, ITS1/2, 3' ETS) and Expansion Segments showed high variation and frequent amplification (high iVFs).
  • Core rRNA sequences (18S, 5.8S, and parts of 28S) showed significantly lower variation.
• Strong Purifying Selection in CNEs:
  • Conserved Nucleotide Elements (CNEs) in the 28S rRNA (identical across >90% of eukaryotes) exhibited the strongest selection.
  • In CNEs, the fraction of variant nucleotides was extremely low (median 0.11%), and median iVFs were significantly lower than in expansion segments (3.18%).
  • Indels in CNEs were significantly shorter, and variants were rarely shared between individuals, suggesting they are rapidly purged.
• Low-Copy Deleterious Effects:
  • The data indicates that deleterious variants are suppressed even when present at very low frequencies (often <1% of copies).
  • The lack of high-copy deleterious variants suggests that if a harmful mutation spreads to even a small fraction of the ~300-400 rRNA copies, it likely confers a fitness cost severe enough to be selected against.
• Heritability:
  • A significant portion of shared variants (46.9% of minor variants and 96% of medium-frequency variants) showed enrichment in specific populations, confirming they are heritable germline events rather than sequencing artifacts or somatic mutations.

5. Significance

• Explaining the "Missing" Disease Link: The study provides a mechanistic explanation for why rRNA-associated diseases have been elusive. Current association studies typically filter out low-frequency variants (low iVF). However, this paper demonstrates that deleterious rRNA variants are kept at low copy numbers by strong purifying selection. Therefore, they are invisible to standard GWAS but likely pathogenic.
• Ribosome Poisoning Hypothesis: The findings support the "poisoning" model, where even a small fraction of defective ribosomes (due to low-copy rRNA variants) can disrupt cellular physiology, likely because translation involves multiple ribosomes on a single transcript.
• Clinical Implications: The authors argue that future disease-mapping studies must specifically account for low-copy number variants in multi-copy gene families like rDNA. Ignoring these variants may lead to false negatives in identifying genetic causes of ribosomopathies and other diseases.
• Evolutionary Dynamics: The results highlight that concerted evolution (gene conversion, unequal exchange) is highly active in humans, but its ability to amplify variants is strictly constrained by selection in functionally critical regions of the ribosome.