Readthrough Extensions in the Cytoplasmic DNA-Sensing Pathway

When a ribosome reaches the period at the end of a gene — the stop codon — it occasionally ignores it and keeps translating into the trailer region of the mRNA, producing a protein with an extra peptide tail. This "stop codon readthrough" is not rare noise: comparative genomics has identified hundreds of genes across insects and mammals where the tail sequence is preserved by evolution, implying it does something useful, yet for the large majority of these extensions the actual job is unknown Evidence of abundant stop codon readthrough in Drosophila and other metazoa; Evolutionary Dynamics of Abundant Stop Codon Readthrough. Meanwhile, cells run a surveillance system that detects DNA loose in the cytoplasm — a sign of viral infection or a broken nucleus — and converts that detection into interferon signaling, inflammation, or cell death. The one place where these two stories could collide has never been examined: nobody has asked whether the proteins that sense cytoplasmic DNA carry readthrough extensions, or whether activating the DNA-sensing pathway changes how often ribosomes read through stop codons.

The collision point is concrete rather than speculative, because the decision-making hardware of cytoplasmic DNA sensing sits exactly where readthrough extensions would be appended. Apaf-1 — an evolutionarily conserved cytoplasmic DNA sensor — chooses between driving apoptosis and driving NF-κB inflammation through competition at its C-terminal WD40 repeat domain, where double-stranded DNA and cytochrome c vie for the same binding interface, with dsDNA binding recruiting RIP2 to switch the cell toward inflammation Apaf-1 is an evolutionarily conserved DNA sensor that switches the cell fate between apoptosis and inflammation. A peptide tail added to that C-terminus by readthrough would land directly on this life-or-death switch. The precedent that a tail can rewire a signaling protein is established: the readthrough-extended human vitamin D receptor (VDRx) dimerizes differently from the canonical receptor and needs roughly 100-fold more ligand to mount the same transcriptional response Stop codon readthrough generates a C-terminally extended variant of the human vitamin D receptor with reduced calcitriol response.

Core Claim

We claim that programmed stop-codon readthrough is an uncharacterized regulatory layer of cytoplasmic DNA sensing, in two directions. First, at least some components of the pathway — Apaf-1, its partner RIP2, or regulators of the cGAS–STING axis such as ARF1 — produce C-terminally extended isoforms whose interaction partners and downstream transcriptional output differ from the annotated canonical proteins, with the strongest mechanism-anchored prediction at Apaf-1, where an extension would sit on the WD40 interface that arbitrates apoptosis versus inflammation. Second, readthrough efficiency of these loci is itself condition-dependent: cytoplasmic DNA challenge shifts it, making readthrough a tunable post-transcriptional dial on the innate response rather than a fixed rate.

This is deliberately distinct from the closest existing work. Loughran et al. established the full discovery-to-function pipeline for readthrough — computational candidate selection, dual-luciferase quantification, Co-IP of extended isoforms — but applied it to a nuclear hormone receptor and never to innate immunity Stop codon readthrough generates a C-terminally extended variant of the human vitamin D receptor with reduced calcitriol response; Evidence of efficient stop codon readthrough in four mammalian genes. Ruan et al. and Lang et al. characterized Apaf-1 and the ARF1–COPI–STING recycling loop entirely as canonical, single-isoform proteins, with no analysis of alternative termination or extension Apaf-1 is an evolutionarily conserved DNA sensor that switches the cell fate between apoptosis and inflammation; Distinct pathogenic mutations in ARF1 allow dissection of its dual role in cGAS-STING signalling. The proposal occupies the empty intersection: their methods, our pathway, and a regulatory question neither asked.

Research Program

The program has three linked aims, moving from discovery to mechanism to regulation.

The first aim is a targeted computational screen of every annotated component of the cytoplasmic DNA-sensing machinery — the sensors (cGAS, Apaf-1), the signaling hub (STING, TBK1, IRF3, RIP2), and the trafficking and degradation regulators (ARF1, COPI components, p62/SQSTM1) — for readthrough signatures downstream of their annotated stop codons. We will run PhyloCSF, the codon-substitution model that distinguishes protein-coding from non-coding sequence across multi-species alignments, using the fast PhyloCSF++ implementation, which reproduces the original tool's scores with a Pearson correlation of 1.0 on raw scores while running 1.3–2.4× faster depending on mode PhyloCSF++: A fast and user-friendly implementation of PhyloCSF with annotation tools. We will prioritize loci carrying the UGA-CUAG tetranucleotide that drives high readthrough across vertebrates, and confirm conserved stem-loop structures downstream of candidate stops with RNAz, as in prior screens Evidence of efficient stop codon readthrough in four mammalian genes. Because PhyloCSF is blind to evolutionarily young, lineage-specific extensions — a limitation the whole-genome PhyloCSF study makes explicit Discovery of high-confidence human protein-coding genes and exons by whole-genome PhyloCSF helps elucidate 118 GWAS loci — we will not require phylogenetic signal for candidacy: any pathway locus with the motif, a predicted stem-loop, or ribosome-footprint density downstream of the stop in the GWIPs-viz repository (the evidence route used to support endogenous VDR readthrough Stop codon readthrough generates a C-terminally extended variant of the human vitamin D receptor with reduced calcitriol response) enters the experimental list even without alignment depth.

The second aim quantifies readthrough of every candidate in mammalian cells using the bicistronic dual-luciferase reporter, in which the stop codon and its flanking sequence are cloned between Renilla and Firefly luciferases so that Firefly light reports translation past the stop. Readthrough efficiency is computed against a 100%-readthrough control carrying a sense codon:

$$\text{RTE} = \frac{(\text{Firefly}/\text{Renilla})_{\text{construct}}}{(\text{Firefly}/\text{Renilla})_{\text{sense-codon control}}}$$

The critical new element is that we measure RTE not only in resting HEK-293T cells, where the reference values come from, but also under pathway activation: transfection of the HSV60 double-stranded DNA ligand used in the Apaf-1 binding work, exogenous 2′3′-cGAMP, and interferon treatment. A shift in RTE under these conditions would establish the feedback arm — cytoplasmic DNA sensing regulating the ribosomal decoding of its own genes — which no dual-luciferase study has examined; prior work quantified readthrough and interferon induction with the same instrument but never in the same experiment Evidence of efficient stop codon readthrough in four mammalian genes; Distinct pathogenic mutations in ARF1 allow dissection of its dual role in cGAS-STING signalling. We will repeat key measurements in A549 and HeLa cells, the lines in which Apaf-1-dependent, cGAS-independent inflammatory activation was demonstrated Apaf-1 is an evolutionarily conserved DNA sensor that switches the cell fate between apoptosis and inflammation, since transformed kidney cells are a poor proxy for innate effector cells.

The third aim tests mechanism with co-immunoprecipitation — pulldown of one tagged protein to recover its physical binding partners — comparing extended and canonical isoforms. For Apaf-1, the decisive Co-IP panel asks whether an extended isoform still binds cytochrome c, double-stranded DNA, and RIP2, and whether the ratios shift; direct affinities for the DNA interaction will be measured by surface plasmon resonance against the canonical benchmark of $K_d = 1.71 \times 10^{-7}\,\text{M}$ for human Apaf-1 binding HSV60 dsDNA Apaf-1 is an evolutionarily conserved DNA sensor that switches the cell fate between apoptosis and inflammation. Unbiased Co-IP coupled to mass spectrometry will catch unanticipated partners of the extension, the strategy that resolved the STING–TBK1–IRF3 interactions subverted by iridoviral VP088 SGIV envelope protein VP088 facilitated virus replication via interacting with other viral proteins and promoting p62-dependent autophagic degradation of TBK1. Downstream consequences are read out with NF-κB and ISRE/IFNB1 luciferase reporters, the standard transcriptional readouts of this pathway Distinct pathogenic mutations in ARF1 allow dissection of its dual role in cGAS-STING signalling, alongside apoptosis-versus-inflammation outcome assays. Endogenous production of any extended isoform will be verified with extension-specific antibodies and Western blot, the approach that confirmed the ~55 kDa VDRx band Stop codon readthrough generates a C-terminally extended variant of the human vitamin D receptor with reduced calcitriol response.

Baselines

The naive reference point is the implicit assumption of the entire DNA-sensing literature as indexed here: that the annotated open reading frame is the whole protein. Both Ruan et al. on Apaf-1 Apaf-1 is an evolutionarily conserved DNA sensor that switches the cell fate between apoptosis and inflammation and Lang et al. on the ARF1–COPI–STING loop Distinct pathogenic mutations in ARF1 allow dissection of its dual role in cGAS-STING signalling operated under this assumption, and it remains the field's working model. Beating this baseline means producing any reproducible extension isoform of a pathway component with a measurable effect on partner binding or interferon/inflammatory transcription; if no such isoform exists, the do-nothing model stands and our screen falsifies the claim.

The strongest competitor is the Loughran et al. discovery pipeline, because it is the only published workflow that spans candidate selection, ratiometric quantification, and isoform-specific interaction mapping, and it holds the field's reference numbers: approximately 31% readthrough for OPRL1 in HEK-293T cells and a drop to roughly 5% when the 5′ stop-proximal sequences were deleted Evidence of efficient stop codon readthrough in four mammalian genes, and 6.7% readthrough for VDR, the highest among tested human UGA-CUAG genes, with the extended VDRx requiring approximately 100-fold higher calcitriol concentrations for comparable transactivation Stop codon readthrough generates a C-terminally extended variant of the human vitamin D receptor with reduced calcitriol response. We will run their pipeline as the baseline on our gene set with the same alignment data, the same PhyloCSF scoring via PhyloCSF++ (correlation 1.0 with the original PhyloCSF++: A fast and user-friendly implementation of PhyloCSF with annotation tools), the same HEK-293T reporter format, and the same motif criteria, so that any difference in outcome is attributable to the pathway being queried, not to method drift. We beat this baseline not by out-detecting readthrough but by doing what the pipeline never did: showing that an extension changes innate immune transcription and partner choice at a pathway signaling hub. Our pre-registered threshold for a functional claim is a reproducible ≥2-fold change in reporter output or a partner-binding change resolvable by Co-IP, applied uniformly; no reported result found on Lacuna for ISRE-luciferase fold-induction values, so we cannot anchor the reporter threshold to a published interferon-induction figure and must set it ourselves.

The key ablation isolates whether the extension peptide, rather than the stop-proximal mRNA context, carries the effect. Following the mutagenesis logic that deleting the conserved CUAG tetranucleotide drastically reduced readthrough for all four confirmed mammalian genes Evidence of efficient stop codon readthrough in four mammalian genes, we compare three constructs per locus: the wild-type stop context, a CUAG-deleted version that collapses readthrough toward baseline, and a sense-codon version forcing 100% extension production. A functional phenotype present in the forced-readthrough construct but absent in the CUAG-deleted one demonstrates that the tail itself, not mRNA structure or translation kinetics at the stop, is the active ingredient.

Why not just…?

Why not just do the screen in yeast or a cell-free lysate? This was directly tested and failed. Chan et al. cloned 147-nucleotide stop-codon neighborhoods of metazoan candidates into the same bicistronic reporter in S. cerevisiae and found that metazoan architectures collapse there: the human candidate SACM1L showed only 0.9% readthrough in wildtype yeast, versus the ~31% measured for a mammalian candidate in HEK-293T cells, indicating that metazoan readthrough depends on trans-factors and mRNA structures absent in fungi Heterologous Stop Codon Readthrough of Metazoan Readthrough Candidates in Yeast. Cell-free systems are worse still: the questions here involve ER-to-Golgi trafficking of STING and apoptosome assembly, which only exist in intact metazoan cells Distinct pathogenic mutations in ARF1 allow dissection of its dual role in cGAS-STING signalling. Mammalian culture is not a preference; it is a requirement established by negative result.

Why not just check for alternative splicing or RNA editing instead of readthrough? This is the right first control, and it has already been run at genome scale: Jungreis et al. used 2.3 billion modENCODE RNA-seq reads to confirm the absence of splice junctions and A-to-I editing events across their readthrough candidates' stop codons, establishing that the unspliced, unedited termination codon persists in the mRNA pool Evidence of abundant stop codon readthrough in Drosophila and other metazoa. We will still perform RT-PCR junction checks on any new candidate in mammalian cells — that is cheap and closes the door — but the burden of proof is on the splicing explanation, not on readthrough, given the prior genome-wide refutation and the reporter-based evidence that mutating the UGA itself abolishes the extended product.

Why not just overexpress a tagged extension construct and skip measuring endogenous readthrough? This is a real threat and we do not argue around it: overexpression would give a clean mechanistic answer about what the tail can do, at a fraction of the effort, and Aim 3 includes such constructs. But it cannot answer whether the tail is produced at meaningful levels physiologically or whether its production is regulated by pathway activation — the second half of the core claim. Overexpression also risks artifacts, since supraphysiological stoichiometry of a competitor at the Apaf-1 WD40 interface could force phenotypes that never occur in vivo. We therefore treat overexpression as mechanism, not evidence: the claim stands or falls on detecting the endogenous extended isoform by extension-specific antibody (the approach that detected endogenous-relevant VDRx as a discrete ~55 kDa band Stop codon readthrough generates a C-terminally extended variant of the human vitamin D receptor with reduced calcitriol response) and on reporter-measured RTE shifting with DNA challenge. If endogenous protein is undetectable despite measurable reporter-level readthrough, we will report the mechanism as latent rather than active, and that would materially weaken the proposal's significance.

Decisive Experiments and Checks

Three outcomes decide the claim. First, the screen and dual-luciferase measurements must identify at least one pathway component with readthrough at or above the level at which VDR readthrough (6.7%) proved functional Stop codon readthrough generates a C-terminally extended variant of the human vitamin D receptor with reduced calcitriol response; anything far below the ~0.9–2.9% range, where metazoan candidates behaved as borderline noise in yeast Heterologous Stop Codon Readthrough of Metazoan Readthrough Candidates in Yeast, would be treated as non-functional. Second, the Co-IP panel must show a partner-choice difference between extended and canonical isoforms — for Apaf-1, a shift in the cytochrome-c versus dsDNA versus RIP2 binding balance, with SPR affinities compared against the published $K_d$ of $1.71 \times 10^{-7}$ M Apaf-1 is an evolutionarily conserved DNA sensor that switches the cell fate between apoptosis and inflammation. Third, the activation-dependent RTE measurements must show a reproducible shift under cytoplasmic DNA challenge; the evolutionarily grounded prior is that such regulation is plausible because these extensions are under strong purifying selection and evolve as gain-and-loss events within existing genes Evolutionary Dynamics of Abundant Stop Codon Readthrough, but no direct measurement exists — no reported result found on Lacuna for condition-dependent readthrough efficiency in any mammalian gene. If the Apaf-1 extension is confirmed and functional, the decisive in vivo test is an HSV-1 challenge in mice bearing extension-disrupting versus wild-type alleles, benchmarked against the reported survival split of 20% in Apaf-1 heterozygous knockouts versus 80% in wild-type Apaf-1 is an evolutionarily conserved DNA sensor that switches the cell fate between apoptosis and inflammation: a survival or cytokine shift caused by removing a peptide tail would establish readthrough as a genetically encoded tuner of antiviral immunity, not a curiosity of reporter constructs.