Antibody production relies on the tRNA inosine wobble modification to meet biased codon demand

Abstract

Antibodies are produced at high rates to provide immunoprotection, which puts pressure on the B cell translational machinery. Here, we identified a pattern of codon usage conserved across antibody genes. One feature thereof is the hyperutilization of codons that lack genome-encoded Watson-Crick transfer RNAs (tRNAs), instead relying on the posttranscriptional tRNA modification inosine (I34), which expands the decoding capacity of specific tRNAs through wobbling. Antibody-secreting cells had increased I34 levels and were more reliant on I34 for protein production than naïve B cells. Furthermore, antibody I34-dependent codon usage may influence B cell passage through regulatory checkpoints. Our work elucidates the interface between the tRNA pool and protein production in the immune system and has implications for the design and selection of antibodies for vaccines and therapeutics.

Fig. 1.. Biased codon usage patterns across immunoglobulin heavy chain constant regions.

(A) Frequency of codon usage in the human IGHM gene. Mean±SD. (B) Quartile distribution of SCUO scores, representing total codon usage bias across gene of the human IGHM gene compared to all genes from human genome CCDS dataset. (C) Mean of codon P(Usage) scores for human IGHM. On the x-axis are the codons ranked by the combined mean scores for all five human IgH isotypes using a linear correlation model. (D) Mean of codon P(Usage) scores for all other human IgH isotypes. The x-axis is ordered as in (C) and were analyzed using a linear correlation model. (E) As (C) and (D) but showing scores for each human TCR constant region chain with the x-axis ordered as in (C), analyzed using a linear correlation model. (F) Hierarchical clustering analysis of codon P(Usage) scores for human and mouse immunoglobulin IgH genes. Each row represents one gene isoform (as delimited by IMGT), each column represents a codon. (G and H) tRNA gene copy number in human genome for the amino acids corresponding to hyperutilized codons bracketed in (F). Codons and tRNAs are matched by Watson–Crick base pairing. Codons lacking genome-encoded Watson–Crick complementary tRNAs are highlighted in red. Bars color-coded by the average codon probability score for all IgH isotypes across humans. Source data for this figure can be found in data S1 and S2.

Fig. 2.. The tRNA modification profiles of antibody-secreting plasma cells and naive B cells differ.

(A) Total ion chromatogram of LC-MS/MS analysis of modified small RNA nucleosides by multiple reaction monitoring (MRM) from WEHI231 cells. (B and C) Normalized signal intensities for modifications from WEHI231 small RNA. Each dot represents the average of three replicates run concurrently. N=3 independent experiments. (B) Modifications found throughout tRNAs. (C) Modifications that occur exclusively on tRNA position 34 (wobble position). (D) Principal component analysis (PCA) of mass spectrometry-based quantification of 25 tRNA modifications in various cell types from one set of samples run concurrently. Each dot represents the average of three replicates run concurrently. (E) As in (D), PCA of mass spectrometry-based quantification of the five tRNA modifications detected that occur exclusively on position 34 of tRNAs. Each dot represents the average of three replicates run concurrently. (F) Log of fold-change values for position 34 tRNA modifications in murine PC lines (J558, MPC11; red), murine splenic B cells (dark blue), and human HEK293T cells (gray), relative to the murine NBC-like line WEHI231 (blue line). Each dot represents the average of three replicates run concurrently. Mean±SEM. (G) Schematic of tRNA deamination to introduce inosine at the wobble position (I34). (H) Hierarchical clustering analysis of codon probability scores for I34-dependent codons in human and mouse IgH genes. Each row represents one gene isoform (as delimited by IMGT), each column represents a codon. (I and J) Quartile distribution of combined I34-dependent codon probability scores for various gene families in humans (gray) and mice (brown). (I) Immunoglobulin heavy chain constant regions (IgH). (J) Immunoglobulin light chain constant regions (IgL), immunoglobulin superfamily proteins (IGSF), T cell receptors (TCR), and histones. In panels B through F, unfilled red circles and triangles denote antibody-secreting B cell lines, and filled blue diamonds and squares denote non-antibody-secreting B cell lines. Source data for mass spectrometry can be found in data S3 and source data for I34-dependent codon probability scores can be found in data S4.

Fig. 3.. The tRNA pool of plasma cell lines is enriched for inosine at the wobble position, compared to naive B cell lines.

(A) Heatmap of significantly differentially expressed tRNA genes between murine plasma cell (PC) lines (red) and murine naive B cell (NBC)-like lines (blue), showing normalized expression values (rlog normalization, using DESeq2) that are centered and scaled by row. I34-modifiable tRNAs are highlighted in orange. Each column represents an individual sample. N=1 independent experiment. (B) Fraction of significantly upregulated tRNAs in NBC lines vs upregulated in PC lines that are I34-modifiable. (C) Volcano plot of differentially expressed anticodons between PC lines and NBC-like lines. I34-modified anticodons are labeled and highlighted in orange. (D) Fraction of nucleotide mismatch at the wobble position of I34-modifiable tRNA species from tRNA sequencing. (E and F) Correlation between mean P(Usage) codon probabilities for murine IgH sequences and the differentially expressed anticodons between murine PC lines and NBC-like lines. (E) Anticodon-codon pairs are restricted to Watson–Crick (non-wobble) base pairing interactions, where applicable. (F) Anticodon-codon pairs are restricted to wobble, where applicable. I34-preferred codons are labeled. Raw tRNA sequencing data available from Dryad (51).

Fig. 4.. I34 and I34-dependent codons play key roles at multiple stages of B cell development.

(A) Representational density plots of log scores for normalized GFP (GFP/mCerulean), by cell. The red label denotes an antibody-secreting B cell line whereas the blue label denotes a non-antibody-secreting B cell line. (B) Mean of normalized GFP by cell type. N=5. (C) Difference in log normalized GFP between GFP_WT and GFP_I34mod by cell type. Each dot represents the average of technical duplicates. N=5 independent experiments. Mean±SD. Linear mixed model with Tukey post-hoc correction. * P<0.05, ** P<0.01, *** P<0.001. Source data for (A to C) can be found in data S5. (D) Representative flow cytometry plots of blood cells from (Top) WT and (Bottom) Adat2-flox Mb1-Cre mice. (E) Quantification of percentage of B220⁺ cells in the blood. Each dot represents a single mouse. N=1. Mean±SEM. Welch’s two-sample t test. Source data for (D and E) can be found in table S1. (F) Percentage of peripheral B cells expressing full-length CLK19 variant from Igh^CLK19/WT or Igh^I34/WT knock-in mice, by single-cell RNA sequencing. (G) Quantification of (F), with mean±SEM. Welch’s two-sample t test. Source data for (F and G) can be found in table S2. (H) Boxplot of I34-dependency P(Usage) scores in variable region antibody repertoires from eight human donors, in various B cell compartments. (I and J) As (H), but showing individual median donor scores in (I) naive B cells versus long-lived BMPC and (J) naive B cells versus memory B cell repertoires. Linear mixed model with Tukey post-hoc correction. * P<0.05, ** P<0.01, *** P<0.001. Source data for (H to J) can be found in data S6.