Molecular Cell

Stress granules (SGs) are cytoplasmic biomolecular condensates that assemble when translation initiation stalls, sequestering stalled preinitiation complexes and associated RNA-binding proteins. How individual initiation factors are targeted to SGs and released following stress recovery to reinitiate translation remains poorly understood. Here, combining a NEDD8-conjugate proteome with our previously reported arsenite-induced NEDD8 interactome and curated RNA granule databases, we find that eIF3g and eIF3i are shared, high-confidence NEDDylated SG components. NEDDylation of eIF3g and eIF3i-associated complexes is readily detected at steady state and declines under arsenite stress. Intriguingly, only full-length eIF3g is recruited to SGs. eIF3g lacking the RRM domain strongly inhibits SG formation, whereas the RRM domain alone neither inhibits SG assembly nor localizes to SGs. Blocking the NEDD8 pathway--by NAE inhibition with MLN4924, depletion of NEDD8 pathway components, or expression of the deNEDDylase NEDP1--accelerates the loss of eIF3g and eIF3i protein during stress. Our data indicate that NEDDylation marks a degradation-resistant pool of eIF3g/eIF3i that is competent for SG localization, linking the NEDD8 pathway to initiation-factor proteostasis and condensate partitioning, and potentially making these factors available for translation reinitiation during stress recovery.

The coding potential of mature mRNAs is generally considered fixed once transcription and RNA processing are complete. We previously established RNA dicing as a post-transcriptional process that generates stable, uncapped, translation-competent RNA isoforms. Here, we identify RNA dicing as a transient post-transcriptional program that remodels mature transcripts during macrophage polarization. Long-read transcriptomics reveals widespread, fate-specific dicing that peaks during early cell-state transitions and preferentially occurs between protein domains, preserving downstream coding modules. Fractionated proteomics links these RNA isoforms to truncated protein products, indicating that dicing reshapes proteomic output. Using JAK1 as a mechanistic model, we show that disruption of dicing, impairs macrophage polarization towards pro-inflammatory states. Mechanistically, a diced JH1 kinase module displayed distinct substrate preferences and alters downstream signaling relative to full-length JAK1. These findings establish RNA dicing as an adaptive layer of gene regulation that reprograms transcript architecture, expands protein functional diversity, and helps shape cell-state transitions.

Cells respond to environmental stress by rapidly remodeling translation and assembling stress granules (SGs), which are dynamic ribonucleoprotein condensates that contain untranslated mRNAs, translation initiation factors, and 40S ribosomal subunits. Although the translation initiation factor eIF4B has been implicated in SG biology, the contribution of its highly conserved RNA recognition motif (RRM) to SG assembly has remained unclear. Here, we developed a quantitative live-cell imaging framework that resolves distinct kinetic phases of SG assembly at single-cell resolution and combines these measurements with single-cell analysis of protein synthesis. Using this approach, we show that disruption of the eIF4B RRM delays SG nucleation, slows SG assembly, and reduces the number of SGs formed, while having little effect on mature SG size. Biochemical analyses revealed that the RRM mutant retained high-affinity binding to both RNA and the 40S ribosomal subunit and exhibited only a modest reduction in eIF4A helicase stimulation activity but displayed altered RNA engagement, consistent with impaired RNA-dependent organization of the translation initiation machinery. Coupling SG kinetics with single-cell measurements of protein synthesis further revealed that delayed SG nucleation is associated with reduced translational repression during oxidative stress. Together, our findings identify the conserved eIF4B RRM as a regulator of productive higher-order organization of the translation initiation machinery and establish a quantitative framework for investigating how SG assembly and translational remodeling are coordinated during cellular stress.

Maintenance of genome integrity requires accurate repair of DNA double-strand breaks (DSBs), particularly within transcriptionally active regions. Persistent R-loops at DSBs can impede homologous recombination (HR) repair. While factors that resolve R-loops at DSB sites are known, the mechanisms ensuring timely degradation of nascent RNA to prevent pathological R-loop accumulation remain elusive. Here, we identified a critical role for the RNA-binding protein LSm4 in orchestrating localized RNA decay at DSBs to facilitate repair. We demonstrated that among LSm1-8 subunits, only LSm4 undergoes liquid-liquid phase separation (LLPS) and forms biomolecular condensates (BCs) specifically at DSBs in transcriptionally active chromatin. These damage-induced LSm4 BCs function as hubs that promote nuclear RNA decapping and recruit the 5'[->]3' exonuclease XRN2 to degrade nascent transcripts proximal to DSBs. Accordingly, LSm4-XRN2 axis suppresses R-loop hyperaccumulation, thereby enabling efficient RAD51 filament assembly and intact HR repair. Consequently, loss of LSm4 increases translocations and leads to genomic instability. Collectively, our findings define a new regulatory layer in which LSm4 BCs spatially license RNA degradation, preventing R-loop accumulation at DSB microenvironment to facilitate error-free repair.

CRISPR adaptation requires Cas1-Cas2 to capture prespacers, undergo conformational rearrangement and catalyse integration, but how these steps are coupled remains unclear. Using nine hyperactive Escherichia coli Cas1 substitutions as perturbational probes, we identified a prespacer-coupling module intersecting an interior conformational-coupling module. Single substitutions increased adaptation up to sixfold, whereas combinatorial reassortment generated a genotype with 103-fold greater activity than wild type. Genotype-network analysis and quantitative reconstruction revealed strong background-dependent epistasis: the same substitution could enhance activity in one genotype but impair it in another, and high activity emerged only from compatible combinations spanning both modules. These findings indicate that Cas1-Cas2 activity is constrained by compatibility among changes distributed across the Cas1 dimer, rather than by optimization of individual catalytic or DNA-binding interactions. We propose that coordinated tuning of prespacer engagement and conformational coupling governs Cas1-Cas2 activity during CRISPR adaptation.

Gene fusions involving the FET gene family (FUS, EWSR1, and TAF15) act as drivers of numerous cancer subtypes. The resulting chimeric proteins are widely viewed as aberrant transcriptional regulators that promote malignant transformation through chromatin and enhancer reprogramming. Here, we show that FET fusion oncoproteins also function as regulators of alternative splicing across multiple sarcoma subtypes. Transcriptomic analyses revealed extensive but largely non-overlapping splicing programs driven by the EWSR1::FLI1, EWSR1::WT1, EWSR1::ATF1 and FUS::DDIT3 fusions that nevertheless converged on common oncogenic functions. Fusion-dependent splicing regulation was mechanistically separable from canonical transcriptional activity and was associated with extensive remodeling of cooperative RNA-binding protein (RBP) assemblies on target transcripts. Despite regulating distinct exons, different FET fusions engaged highly similar RBP interaction networks, consistent with a conserved mode of splicing regulation. Transcriptome-wide mapping of RBP occupancy revealed extensive reorganization of local RNA regulatory landscapes following fusion depletion. The requirement of RNA for FET fusion condensate formation, together with the inability of condensation-defective mutants to restore splicing regulation, further supported a role for higher-order assemblies in fusion-dependent alternative splicing (AS) control. Fusion-driven splicing programs stratified Ewing sarcoma patients independently of established clinical covariates, thereby underscoring their clinical relevance. AS of TFDP1 emerged as a common fusion-regulated splicing event required for sarcoma cell fitness and therapeutically actionable using antisense oligonucleotides. Together, our findings establish AS regulation as a conserved function of FET fusion oncoproteins that is mechanistically separable from their canonical transcriptional activity. More broadly, they support a model in which oncogenic fusion proteins can drive malignant phenotypes through large-scale remodeling of RNA regulatory networks.

Epigenetic information is transmitted through covalent DNA modifications that regulate chromatin structure and gene expression. 5-Hydroxymethylcytosine (5hmC) is a key epigenetic intermediate generated during TET-mediated oxidative demethylation of 5-methylcytosine. Whether 5hmC itself is intrinsically genotoxic remains unclear. Here, we combine genome-wide CRISPR loss-of-function screens, isogenic knockouts and mass spectrometry to systematically compare the cellular consequences of exogenous 5-hydroxymethyl-2'-cytidine (5hm-dC) exposure and endogenous oxidative demethylation in mammalian cells. We find that exogenous 5hm-dC causes toxicity, mediated by deamination to 5-hydroxymethyl-2'-deoxyuridine (5hm-dU), followed by excision by the glycosylase SMUG1, which generates base excision repair (BER) intermediates that compromise cell viability. In contrast, toxicity associated with enzymatic oxidative demethylation is primarily driven by TDG-dependent excision of oxidized methylcytosine derivatives. Despite these distinct initiating events, both genotoxins converge on a critical requirement for DNA polymerase {beta} (POL{beta}), indicating that efficient BER completion is essential to mitigate cytotoxic repair intermediates.

Protein homeostasis relies on protein quality control (PQC) pathways that survey the proteome to eliminate aberrant polypeptides. The BAG6 complex is a central PQC factor that recognizes exposed hydrophobic regions, a feature commonly associated with misfolded, mislocalized, and mistranslated proteins. Whether this surveillance machinery also regulates intact, functional proteins as part of physiological proteostasis has remained unclear. Using unbiased quantitative proteomics, we identify the ribosomal protein RPL22L1 as an endogenous BAG6 substrate whose abundance is controlled by continuous proteasomal degradation. This turnover requires the RNF115 E3 ligase activity but not the canonical BAG6 partner RNF126, defining RPL22L1 as a selective RNF115-dependent substrate. Mechanistically, we map a bipartite hydrophobic degron that distinguishes RPL22L1 from its stable paralog RPL22, and show that BAG6-RNF115-mediated degradation is governed by substrate assembly state. Accordingly, RPL22L1 is protected from degradation upon incorporation into the 60S ribosome, where it substitutes for RPL22. When RPL22 is lost, either genetically or through recurrent inactivating mutations in microsatellite-unstable cancers, the vacant ribosomal binding site permits RPL22L1 incorporation, protecting it from BAG6-mediated degradation. These findings establish unassembly-coupled degradation as a mechanism by which BAG6 regulates the abundance of functional protein components, ensuring that they accumulate only when incorporated into their native macromolecular complexes.

Genomic DNA is packaged into chromatin through a fundamental repeating unit known as the nucleosome core particle. Chromatinized DNA is under constant assault from endogenous and exogenous sources of damage, which must be effectively repaired to preserve genome stability. Single-strand breaks (SSBs) with chemically heterogeneous DNA ends are one of the most prevalent forms of genomic DNA damage. These SSBs must be enzymatically processed prior to downstream gap-filling DNA synthesis and/or nick ligation during single-strand break repair (SSBR). Polynucleotide kinase phosphatase (PNKP) is a multifunctional end-processing enzyme that possesses two catalytic activities important for converting non-ligatable SSBs into ligatable SSBs. To date, a mechanistic description for how PNKP processes non-ligatable SSBs in the context of chromatin to initiate SSBR remains undefined. Here, we utilize a combination of biochemical assays and cryogenic electron microscopy (cryo-EM) to define the structural basis of end processing in the nucleosome by PNKP. Cryo-EM structures of PNKP engaged with non-ligatable SSBs at three unique positions within the nucleosome reveal that PNKP locally deforms nucleosomal DNA to reposition the SSBs into the kinase and phosphatase active sites, providing a structural basis for the efficient processing of SSBs throughout the nucleosome. Additional cryo-EM structures reveal the PNKP FHA domain also engages the nucleosome acidic patch during non-ligatable SSB recognition, which accelerates the processing of non-ligatable SSBs in the nucleosome. Together, these findings provide important mechanistic insight into the initial end processing step of chromatin-based SSBR.

Ribosome function depends on interactions with diverse proteins whose activities determine translational efficiency, impose quality control and activate signaling pathways. These many competing activities must in turn be regulated. SERBP1 is an abundant cellular factor that interacts with ribosomes at multiple functionally critical sites on both dormant and active ribosomes. Here, we perform mass spectrometry across sucrose gradients in untreated and stressed cells and find that SERBP1 modulates ribosome interactions with many factors involved in processes including mRNA degradation, translational control, ribosome quality control and ribosome degradation. We then define the role of SERBP1 in protection of ribosomes against selective 40S degradation in the context of mTOR inhibition through its competition with the E3 ligase RNF10 and the atypical kinase RIOK3. Our work reveals SERBP1 as a key player in maintaining ribosome subunit balance and more broadly in the regulation of diverse ribosome activities.

CRISPR-Cas systems are RNA-guided nucleases that enable prokaryotic immunity; however, some loci encode additional associated genes that cooperate with CRISPR effectors to perform diverse biological functions. Here, we uncover a CRISPR-associated kinase (CASK) system that links the recognition of target RNA to protein phosphorylation. We show that the kinase Csx33 phosphorylates Csx34 following activation of Cas13, enabling Csx34 to bind DNA in a sequence-specific manner. Together, Csx33 and Csx34 function as a transcriptional activation module that upregulates cas13 and associated genes, revealing a positive autoregulatory circuit that potentiates the immune response upon detection of foreign RNA. At the molecular level, Csx33 is activated by CCA trinucleotide RNA generated by Cas13-mediated cleavage of tRNA 3' tails, uncovering a novel linear second messenger in bacterial immunity and a previously unrecognized signaling role of collateral RNA fragments. Together, these findings establish CASK systems as a new platform for RNA sensing and for engineering programmable phosphorylation-based signaling systems.

Apoptosis-inducing factor is a mitochondrial flavoprotein that links redox metabolism to mitochondrial homeostasis through its interaction with the disulfide relay protein CHCHD4. Although NADH-dependent AIF dimerization has been proposed as the activated state mediating CHCHD4 engagement, whether it is strictly required for productive AIF-CHCHD4 function remains unclear. Here, combining cellular, biochemical and biophysical approaches, we show that disruption of the AIF dimer interface compromises oxidative phosphorylation, respiratory-chain organization and CHCHD4-dependent mitochondrial homeostasis, yet preserves partial AIF function. Our data reveal that the AIF-CHCHD4 system operates as a conformational dynamic redox module in which distinct AIF oligomeric and redox states sustain CHCHD4 activity with different efficiencies. Mechanistically, dimerization is coupled to NADH-dependent conformational changes that regulate coenzyme binding, charge-transfer complex stabilization and catalytic efficiency. In turn, CHCHD4 binding remodels AIF conformational and redox properties, partially compensating for defects in dimer stabilization or redox coupling. Consistently, a peptide derived from the CHCHD4 N-terminus partially restores redox function in a pathogenic AIF variant defective in dimer stabilization, supporting partner-assisted allosteric regulation as a potential therapeutic strategy.

DNA double-strand breaks (DSBs) are primarily repaired by non-homologous end joining (NHEJ), which tethers and ligates DNA ends within a synaptic complex. PAXX is an NHEJ accessory factor related to XRCC4 and XLF, but its mechanistic role and genetic interaction with XLF are unclear. Using biochemical assays and single-molecule imaging within Xenopus laevis egg extracts and human cell assays, we show that PAXX bridges DNA ends by using its disordered tails to bind opposing Ku molecules. This PAXX tether substantially extends long-range synaptic complex (LRSC) lifetime and modestly stabilizes the short-range synaptic complex (SRSC). LRSC stabilization increases XLF residence and promotes the LRSC-to-SRSC transition. Using domain-swapped chimeras, we demonstrate that PAXXs tails provide synaptic stabilization, whereas XLFs heads interact with XRCC4 to drive the LRSC-to-SRSC transition: an XLF head-PAXX tail chimera recapitulates the functions of both proteins, revealing complementary, sequential roles for PAXX and XLF in NHEJ.

Heat shock factor 1 (HSF1) is the master transcriptional regulator of cellular response to disrupted cytosolic protein homeostasis. Temperature change, oxidation, and other stresses drive the trimerization and activation of HSF1 to induce expression of molecular chaperones, such as heat shock proteins Hsp70 and Hsp90, which sit at the center of cellular proteostatic networks. In turn, the HSPs and co-chaperones regulate HSF1, but mechanistic details of this cycle remain largely unknown. We developed a FRET-based approach to simultaneously monitor HSF1 conformational change and oligomeric state throughout activation and inactivation. By reconstituting Hsp-HSF1 interactions in vitro, we find that monomerization of HSF1 resembles fibril disassembly through coordinated Hsp40-Hsp70 activity. We then used site-specific photocrosslinking to track HSF1 loading into Hsp90 complexes, Hsp90 cycling, and stress-induced shifts in Hsp90-HSF1 interactions. Whereas Hsp90 inhibitors force 'loading state' type Hsp90-HSF1 interactions, heat shock promotes faster Hsp90 cycling. In this reconstituted system, HSF1-Hsp90 interactions are unexpectedly strongly dependent on the co-chaperone HOP, in contrast to canonical Hsp90 clients. Combining cryo-EM structures of Hsp90-HSF1 loading state and maturation state complexes with crosslinking mass spectroscopy and biophysical experiments, we show that Hsp90 holds HSF1 in a pre-activated, extended monomer state that is primed for trimerization. We propose this state both enhances responsivity but also promotes cytoplasmic-nuclear shuttling through exposure of the NLS. Notably, Hsp90-bound HSF1 can trimerize and bind DNA, placing it on-pathway for transcriptional activation. Together, our results unify previously contradictory view on Hsp90's role in HSF1 regulation. The integrated combination of in vitro reconstitution, photocrosslinking, cryoEM and MS is an exciting new paradigm for the study of many dynamic systems including other complex proteostasis components.

DNA helicases preserve genome stability by unwinding DNA during replication, repair, recombination, and transcription, yet their sites of action in vivo remain difficult to define. Here, we describe a sequencing-based strategy to map helicase activity genome-wide by coupling helicases to the single-stranded DNA-specific activation-induced cytidine deaminase (AID). Deamination of cytosines exposed during helicase-mediated DNA unwinding generates strand-specific mutational footprints that can be detected by whole-genome sequencing at near-nucleotide resolution. Using the Saccharomyces cerevisiae RecQ4-family helicase Hrq1, a functional homolog of human RECQL4, we generated the first genome-wide map of Hrq1 activity. Hrq1-dependent deaminations were highly enriched at RNA polymerase III (RNAPIII)-transcribed genes, particularly tRNA genes, where they occurred predominantly on the transcriptional template strand. This localization was reproducible using both overexpressed Hrq1-AID fusions and an inducible dimerization system that recruited AID to endogenously expressed Hrq1, and it was markedly reduced by helicase-inactivating mutation, indicating that active DNA unwinding underlies the observed signal. Hrq1 associated with nearly all tRNA genes irrespective of transcription level, replication orientation, or proximity to transposable elements, yet deletion or overexpression of Hrq1 did not detectably alter pre-tRNA abundance or RNA polymerase III recycling under the conditions tested. Application of the same approach to the PIF1-family helicase Rrm3 recovered its established enrichment at a subset of highly transcribed, head-on tRNA genes, validating the method. Together, these findings establish AID-mediated mutational footprinting as a general approach for mapping helicase activity in vivo and reveal an unexpected, widespread association of the RecQ4-family helicase Hrq1 with RNAPIII-transcribed genes.

Recognition of 5' splice sites by the U1 small nuclear ribonucleoprotein commits pre-mRNAs to splicing, yet splice-site complementarity alone cannot predict productive U1 engagement. Whether dynamic pre-mRNA structure regulates this early spliceosome assembly step remains unclear. Here, we identify a 5'UTR-intron base-pairing interaction positioned near the 5' splice site that acts as an inducible structural gate for U1 engagement. In budding yeast, this element is enriched among introns required for adaptation to nutrient depletion, and in vivo DMS-MaPseq shows that starvation remodels its structure. Structure-guided disruption of pairing impairs adaptation, whereas compensatory mutations restoring pairing without restoring sequence rescue the phenotype, establishing RNA fold as the critical determinant. U1 association decreases when the gate is disrupted and recovers when pairing is restored, and increased Nam8 levels can compensate for gate disruption by stabilizing U1 engagement under stress. Thus, dynamic pre-mRNA folding gates U1 recognition, revealing how transcript architecture converts physiological state into selective splice-site choice.

The integrated stress response (ISR) globally suppresses protein synthesis while selectively permitting translation of a small subset of stress-responsive mRNAs, many of which contain upstream or overlapping open reading frames (uORFs/oORFs). Although translational induction of transcripts such as ATF4 has classically been attributed to delayed re-initiation caused by reduced ternary complex availability, the mechanisms by which uORFs and oORFs allow ISR-selective translation remain incompletely understood. Here, using ribosome profiling during early ISR activation combined with reporter assays, we identify DCP2, encoding a major mRNA decapping enzyme, as a previously unrecognized ISR-induced transcript. We show that translational induction of DCP2 depends on an overlapping ORF whose conserved 3' region, corresponding to a ribosome pausing site, acts as a potent inhibitory element. Both the DCP2 oORF and main ORF increase in translation during stress, indicating that stress relieves repression by this inhibitory element. This reveals a mode of ISR-dependent gene regulation in which inhibition by a nascent peptide or stalling element embedded either in a uORF or an oORF is relieved upon stress to induce translation.

tRNAs are characterized by extensive chemical modifications that influence tRNA fate. N1-methyladenosine at position 58 (m1A58) is a widespread core tRNA modification linked to physiological and pathological processes. However, how m1A58 coordinate tRNA folding and processing to ensure translational efficiency in mammalian cells remains largely unknown. Using acute dTAG-mediated degradation and CRISPR-Cas9 knockout, we identified initiator methionine tRNA (tRNAiMet) as selectively vulnerable to m1A58 loss, lacking the isodecoder buffering observed for most other tRNA isoacceptors. NMR analysis of the tRNAiMet showed that m1A58 stabilizes D/T-loop interactions, consistent with a maturation-competent conformation. In vitro processing assays further demonstrated that m1A58 promotes RNase P-mediated 5'-leader removal and RNase Z-mediated 3'-trailer cleavage, while La/SSB protects accumulated precursors. Disrupting this checkpoint impaired the assembly of the eIF2-containing 43S pre-initiation complex and global protein synthesis, which was substantially rescued by adding m1A58-modified tRNAiMet. Acute TRMT6 degradation elicited temporally coordinated gene-expression responses involving proteostasis, transport and signaling. Together, these findings establish m1A58 as a conformational checkpoint coupling human initiator-tRNA maturation to translation initiation and stress responses.

PARG, the primary enzyme responsible for the reversal of PARP1-mediated poly(ADP-ribosyl)ation, has attracted considerable attention as a therapeutic target in cancer. Yet, the mechanisms underlying PARG inhibitor (PARGi) efficacy remain elusive. Herein, we demonstrate that PARGi prolongs PARP1 residence at damaged chromatin in a manner mechanistically distinct from PARP-inhibitor-induced trapping. That is, PARGi triggers the formation of PAR-driven, FUS-enriched PARP1 nuclear condensates upon DNA damage. Importantly, unrestrained S-phase PARylation during Okazaki fragment maturation also elicited PARP1 condensation in a manner directly reflecting intrinsic PARGi sensitivity, with FEN1 co-inhibition enhancing both PARP1 condensate formation and cytotoxicity. Finally, we discover that dePARylation prevents the rapid nuclear extrusion of PARP1 upon S-phase entry, a phenomenon that is reversible and could undermine PARP1-dependent nuclear processes. Together, our findings reveal that dePARylation precludes the replication-driven condensation of PARP1 and identify Okazaki fragment maturation as a targetable vulnerability that exacerbates condensation and PARGi cytotoxicity.

Nonsense-mediated mRNA decay (NMD) is a translation-dependent mRNA decay pathway triggered by premature termination codons (PTCs). Although NMD is known to eliminate aberrant transcripts, how PTC position influences cell-to-cell heterogeneity in NMD and the resulting protein outputs remains unclear. Here, we used a single-cell NMD analysis system that quantifies cellular variability based on the GFP/mCherry fluorescence ratio. By combining this system with fluorescence-activated cell sorting (FACS), we show that PTC location critically determines not only NMD efficiency and its variability across cells, but also the spectrum of resulting protein products. These include truncated proteins arising from premature termination, full-length proteins generated through translational readthrough, and N-terminally truncated isoforms produced by downstream reinitiation. Our findings reveal that positional and cellular heterogeneity in NMD contribute to proteomic diversity and may underlie the variable phenotypic severity of genetic diseases caused by PTCs. This work establishes a framework for dissecting NMD regulation and its translational consequences.