Endogenous targets of Zuo1-mediated mRNA surveillance encode proteins with higher predicted disorder

Summary

Zuo1 slows elongating ribosomes in response to nascent peptide misfolding, which in turn triggers co-translational decay of the transcript. The endogenous substrates of this pathway have remained undefined. Using RNA-sequencing and ribosome profiling of unstressed wild-type and zuo1Δ cells, we identified transcripts that accumulate in the absence of Zuo1, representing mRNAs normally downregulated by Zuo1-mediated surveillance. Proteins encoded by these transcripts displayed higher predicted disorder across multiple metrics, were enriched in basic residues, were depleted of proline residues, and were shorter in length. Ribosome profiling revealed that Zuo1 target transcripts exhibited higher ribosome occupancy in the presence of Zuo1, and the ratio of wild-type to zuo1Δ ribosome occupancy positively correlated with predicted disorder. These findings indicate that Zuo1 surveillance acts preferentially on transcripts encoding nascent chains that fail to acquire stable structure.

Graphical abstract

Highlights

• •Transcripts accumulating in zuo1Δ cells encode proteins with higher predicted disorder
• •These Zuo1 target transcripts exhibit higher ribosome occupancy
• •Wild-type to zuo1Δ ribosome occupancy ratio positively correlates with predicted disorder

Molecular biology; Molecular mechanism of gene regulation

Introduction

Ribosome dynamics during messenger RNA (mRNA) translation elongation are a key arbiter of mRNA stability. Many co-translational mRNA decay pathways, including nonstop decay, no-go decay, and codon optimality-mediated decay, monitor ribosome speed and degrade transcripts with stalled or abnormally moving ribosomes.^1,2 Ribosome speed, and thus mRNA stability, is strongly influenced by codon usage, transfer RNA (tRNA) abundance, and amino acid availability.² Beyond these factors, we recently identified a previously unknown regulator of ribosome speed: the folding status of the nascent polypeptide.³

This regulation is mediated by the ribosome-associated chaperone protein Zuotin (Zuo1 in yeast; DNAJC2/MPP11 in humans). Zuo1 is a ribosome-bound HSP40 chaperone that forms the ribosome-associated chaperone complex (RAC) together with the non-canonical HSP70 Ssz1 and interacts with the nascent peptide emerging from the ribosomal exit channel.⁴ Simultaneously, Zuo1 binds to the 18S ribosomal RNA helix 44 (H44), which extends from the ribosomal A-site.⁵ This unique structural arrangement enables Zuo1 to mediate communication between the nascent peptide and the ribosomal A-site, where the decoding rate is determined. When nascent chains misfold, Zuo1 slows elongating ribosomes, which are then recognized by co-translational mRNA decay machineries, resulting in degradation of the transcript.³

The endogenous substrates of this mechanism remain largely unknown. In our recent publication, we showed that global misfolding induced by L-azetidine-2-carboxylic acid (AZC) treatment, heat, or oxidative stress causes ribosome stalling and subsequent downregulation of transcripts harboring stalled ribosomes in a Zuo1-dependent manner.³ However, these stressors each induce different types of protein misfolding, along with many confounding secondary cellular responses, making it difficult to determine what Zuo1-mediated surveillance monitors under normal conditions. Identifying and characterizing the substrates of this pathway in the absence of exogenous stress is essential for understanding its physiological role in regulating the transcriptome and maintaining proteome quality.

Here, using RNA-sequencing and ribosome profiling of unstressed cells, we define the cohort of transcripts that accumulate when Zuo1 is absent and characterize the proteins they encode. We find that these transcripts, which we term Zuo1 targets, encode proteins with higher predicted disorder, are enriched in positively charged residues and depleted of proline, and show elevated ribosome occupancy when Zuo1 is present. Ribosome occupancy also positively correlated with predicted disorder in a Zuo1-dependent manner. These observations are consistent with a surveillance mechanism that reads the conformational state of the emerging nascent chain.

Results

Comparing the transcriptomic profile and ribosome binding in the presence and absence of Zuo1

Using RNA-sequencing and ribosome profiling, we assessed transcriptomic changes and ribosome binding across the transcriptome in the presence and absence of Zuo1. To identify endogenous Zuo1 substrates and avoid confounding effects from stress-induced protein misfolding, we compared wild-type and zuo1Δ cells in an unstressed state. Bulk RNA-sequencing identified mRNAs whose abundance was affected by the absence of Zuo1 (Figure 1A). Increased abundance in zuo1Δ cells indicates downregulation of the given transcript by Zuo1 under normal conditions. Ribosome profiling, which sequences ribosome-protected fragments of mRNA at the time of collection, quantified ribosome binding to these transcripts (Figure 1B). By analyzing these datasets together, we aimed to link ribosome occupancy to mRNA abundance as well as determine whether transcripts regulated by Zuo1 share common characteristics in the proteins they encode.

Figure 1.

Experimental design of the study

From unstressed wild-type and zuo1Δ yeast cells, (A) steady-state mRNA pool was subjected to RNA-sequencing to measure mRNA abundance, and (B) ribosome-protected mRNA fragments were isolated to perform ribosome profiling, examining ribosome binding to transcripts. Created in BioRender. Chang, S. (2026) https://BioRender.com/cd0opro.

Following classification from the Saccharomyces Genome Database (SGD),⁶ all analyses were restricted to verified and uncharacterized open reading frames (n = 5,133), excluding dubious open reading frames, genes within the ribosomal DNA repeat, and subtelomeric Y′ elements to avoid annotation and ribosomal RNA-derived artifacts. This filtering had minimal effect on the composition of the differentially expressed gene sets.

Transcripts accumulating upon Zuo1 deletion encode proteins with higher predicted disorder

Using RNA-sequencing, we identified transcripts whose steady-state abundance differed significantly between wild-type and zuo1Δ cells under unstressed conditions (Figure 2A). Because increased abundance upon Zuo1 deletion indicates that a transcript is normally downregulated by Zuo1, we term the 1,461 transcripts that accumulate in zuo1Δ cells “Zuo1 targets,” and we compared them to the rest of the transcriptome (n = 3,672) or transcripts of reduced abundance in zuo1Δ cells (n = 1,600). Of note, the absence of Zuo1 alone did not induce canonical proteotoxic responses^7,8 (Figure S1), indicating that these gene sets represent physiological Zuo1 substrates rather than stress-induced targets.

Figure 2.

Transcripts accumulating upon Zuo1 deletion encode proteins with increased predicted disorder

(A) Volcano plot of log₂ fold change (zuo1Δ/WT) against adjusted p value. A total of 1,461 upregulated and 1,600 downregulated genes in zuo1Δ cells were identified. Points are colored by the mean AlphaFold pLDDT of the encoded protein. The x axis is clipped for readability; one out-of-range gene (ZUO1) is shown as a triangle at the boundary.

(B) (Left) Density distribution of mean pLDDT for Zuo1 target transcripts (“Upregulated”), transcripts downregulated in zuo1Δ cells (“Downregulated”), and all genes, with dotted lines marking group medians. (Right) Corresponding boxplot comparison showing mean pLDDT values for upregulated and downregulated genes.

(C) Fraction of residues with pLDDT below 50 (“disordered”; left) and above 90 (“structured”; right), for Zuo1 targets and genes downregulated in zuo1Δ cells.

(D) Mean pLDDT plotted against log₂ fold change, with a locally estimated scatterplot smoothing (LOESS) fit and 95% confidence band. Spearman correlation and p values are shown. One out-of-range gene (ZUO1) is shown as a triangle.

(E) Fraction of disordered residues predicted by five alignment-free predictors, comparing Zuo1 targets to the non-upregulated background genes.

(F) The same five predictors comparing Zuo1 targets to transcripts downregulated in zuo1Δ cells.

(B–D) 405 upregulated and 265 downregulated transcripts with no entry in the AlphaFold database were excluded from the analysis. (E and F) Violins show the full distribution, boxes show the interquartile range, and the orange lines indicate the median. Sample sizes given on each image are the number of transcripts carrying a value for the presented disordered metric. (B–F) Statistical comparisons are two-sided Mann-Whitney tests with Cliff’s delta as the effect size.

We first examined predicted local distance difference test (pLDDT) scores, a per-residue structure prediction confidence metric from AlphaFold.⁹ Low pLDDT scores are strongly correlated with protein disorder propensity of a given protein.^10,11 Zuo1 target transcripts encode proteins with significantly lower mean pLDDT scores than transcripts downregulated in zuo1Δ cells, with median values of 77.0 and 85.8, respectively (Figure 2B). Closer examination revealed that the fraction of residues predicted to be disordered, defined as pLDDT below 50, was significantly higher among Zuo1 targets (Figure 2C, left). In contrast, the fraction of well-ordered residues, defined as pLDDT above 90, was significantly lower among Zuo1 targets (Figure 2C, right). We then examined the correlation between predicted disorder and the change in mRNA abundance upon Zuo1 deletion to isolate the Zuo1-dependent component. Across the transcriptome, mean pLDDT correlated negatively with the magnitude of change following Zuo1 deletion (Figure 2D), indicating that transcripts encoding proteins with lower predicted structural order accumulate more strongly when Zuo1 is absent.

Because pLDDT reflects prediction confidence that is influenced by the depth of the multiple sequence alignment available for a given protein, we asked whether this association could be an artifact of the prediction method. We therefore examined five alignment-free disorder predictors that are based on sequences: IUPred (short-disorder configuration),¹² ESpritz in both its X-ray and NMR configurations,¹³ MobiDB-lite,¹⁴ and DisEMBL.¹⁵ All five predictors independently indicated that proteins encoded by Zuo1 target transcripts are more disordered than those encoded by genes that are not upregulated in zuo1Δ cells (Figure 2E), with a larger effect size when contrasted against transcripts downregulated in zuo1Δ cells (Figure 2F). Their agreement with the pLDDT analysis establishes that the association reflects sequence-encoded disorder rather than differences in prediction confidence.

Proteins encoded by Zuo1 target transcripts tend to be more positively charged, depleted of prolines, and shorter in length

The disorder enrichment analysis described earlier raises the question of whether Zuo1 targets differ in amino acid composition in ways not captured by disorder prediction. Compared with transcripts that were not upregulated in zuo1Δ cells, Zuo1 target transcripts encoded proteins enriched in positively charged residues (Figure 3A, left), depleted of proline (Figure 3B, left), and shorter in length (Figure 3C, left). This trend was maintained with a larger effect size when compared with transcripts downregulated in zuo1Δ cells (Figures 3A–3C, right). Consistently, enrichment in positively charged residues correlated positively with transcript stabilization in zuo1Δ cells (Figure 3D), while proline frequency (Figure 3E) and protein length (Figure 3F) correlated negatively.

Figure 3.

Characteristics shared by proteins encoded by Zuo1 target transcripts

(A–C) Distributions in Zuo1 targets compared to genes (left) not upregulated and (right) downregulated in zuo1Δ cells, shown as violins with interquartile boxes and median lines. Statistical comparisons used a two-sided Mann-Whitney U test with Benjamini-Hochberg correction. (A) Positively charged residue fraction (lysine and arginine), (B) proline frequency, and (C) protein length; 6 mitochondrial genes were eliminated due to lack of length annotation.

(D–F) Log₂ fold change of gene expression in zuo1Δ versus wild-type cells plotted against (D) positively charged fraction, (E) proline frequency, and (F) protein length. Gray points show individual genes, orange points show decile medians ±standard error of the mean, and the red line is an ordinary least-squares linear fit. Spearman’s rho and Pearson’s r, with associated p values, are reported in each image. The y axis is clipped to the central 99% of values.

Enrichment in positively charged residues also correlated modestly with predicted disorder by IUPred (Figure S2A), suggesting that basic residue enrichment may be detected by Zuo1-mediated surveillance through its association with structural disorder. In contrast, proline frequency showed a positive correlation with predicted disorder (Figure S2B), and shorter protein length did not correlate with disorder prediction metrics (Figure S2C).

Ribosome occupancy is increased on Zuo1 target transcripts

We next asked whether Zuo1 target transcripts exhibit elevated ribosome occupancy, as predicted by a model in which Zuo1 slows elongating ribosomes on these transcripts. Ribosome density per transcript was calculated by normalizing ribosome-protected fragment (RPF) counts by steady-state mRNA level. Over 95% of Zuo1 target transcripts exhibited higher ribosome occupancy in wild-type compared with zuo1Δ cells (Figure 4A, left). This trend was specific to Zuo1 target transcripts, as higher ribosome occupancy in wild-type compared with zuo1Δ cells was not observed across the genome (Figure 4A, right). This specificity argues against a global effect of Zuo1 deletion on translation and supports a model in which Zuo1 selectively modifies ribosome occupancy on a defined subset of target transcripts.

Figure 4.

Ribosome occupancy correlates with gene expression changes and predicted disorder

(A) Ribosome occupancy (RPF/mRNA) in wild-type and zuo1Δ cells on (left) Zuo1 target transcripts and (right) a genome-wide reference. While over 95% of Zuo1 target genes show higher ribosome occupancy in wild-type cells (Wilcoxon signed-rank test, p = 9 × e⁻²²⁹), only 50% of genes in the genome-wide reference show higher ribosome occupancy in wild-type cells, with no statistical significance.

(B) Ratio of wild-type to zuo1Δ ribosome occupancy, across transcripts assigned to IUPred quintiles.

(C) Ratio of wild-type to zuo1Δ ribosome occupancy, across transcripts assigned to pLDDT quintiles.

(B and C) (Top) Per-transcript paired difference in ribosome occupancy by disorder quintile, with a reference line at zero. Violins show the full distribution of paired differences within each quintile; each violin is annotated with the fraction of transcripts above zero, and the number of genes per quintile is labeled in the x axis. (Bottom) The Hodges-Lehmann estimate of the paired difference within each quintile, with its 95% confidence interval. The Jonckheere-Terpstra trend p value and the covariate (protein length and expression level)-adjusted slope from the continuous regression are presented together.

Ribosome occupancy positively correlates with predicted disorder

We examined whether ribosome density per transcript in wild-type versus zuo1Δ cells correlates with the protein folding predictors examined in earlier analyses. The ratio of wild-type to zuo1Δ ribosome occupancy for each gene showed a significant positive correlation with IUPred short-form disorder (Figure 4B), indicating that transcripts encoding more disordered proteins exhibited higher ribosome binding in a Zuo1-dependent manner. This trend is not an artifact of transcript abundance or protein length, as adjustment for wild-type expression level and reading frame length left the disorder slope essentially unchanged at 0.146 (Figure 4B, bottom).

The same relationship was observed using mean pLDDT as the disorder measure (Figure 4C), where the ratio of wild-type to zuo1Δ ribosome occupancy showed a negative correlation with pLDDT scores, as lower pLDDT indicates higher disorder. Of note, while IUPred prediction covers 5,110 of the 5,133 transcripts, the pLDDT analysis was restricted to the 3,964 transcripts with an AlphaFold model, potentially under-representing short and poorly characterized reading frames.

Although the trend was shallow, it was consistent in direction and monotonic across increasing disorder for both prediction metrics. The direction of the relationship and its dependence on predicted disorder are consistent with a model in which Zuo1 slows ribosomes on transcripts encoding poorly structured nascent chains (Figure 5).

Figure 5.

Model for Zuo1-mediated mRNA surveillance in response to co-translational protein misfolding

(Top) Under normal conditions, Zuo1 (yellow) associates with ribosomes without affecting the elongation rate (green signal). (Bottom) When the nascent chain has difficulty in folding (red, tangled), Zuo1 causes ribosome slowdown/stalling (yellow/red signals), leading to ribosome pileup on the transcript and triggering degradation of the transcript by the co-translational mRNA decay machinery. Created in BioRender. Chang, S. (2026) https://BioRender.com/3dbt63s.

Discussion

Our results provide the first characterization of the endogenous substrates monitored by Zuo1-mediated surveillance. Transcripts that accumulate in the absence of Zuo1 encode proteins with higher predicted disorder, and this association is robust to the choice of prediction method. Ribosome profiling revealed that these transcripts exhibit higher ribosome occupancy in the presence of Zuo1, and the ratio of wild-type and zuo1Δ ribosome occupancy scales with predicted disorder of the encoded protein. Together, these observations support a model in which Zuo1 reads the conformational state of the nascent chain as it emerges and slows the ribosome when stable structure fails to form.

The identity of the Zuo1 substrate cohort refines the model we proposed previously. An intrinsically disordered protein does not misfold in the conventional sense, because it never adopts a stable fold.¹⁶ What the surveillance mechanism appears to detect is therefore the absence of compact structure at the exit channel rather than the presence of a specific misfolded species. This interpretation is consistent with the reporter constructs used in our recent study, in which shuffled open reading frames encoding sequences with no capacity to fold were efficient triggers of Zuo1-dependent slowing and decay.³ In that study, differently shuffled HIS3 ORFs produced decay intermediates of different sizes, indicating that Zuo1-dependent ribosome pausing depends on how the nascent chain is folding rather than on the ribosome’s precise position within the mRNA. The endogenous cohort identified here exhibits the same property.

That transcripts encoding intrinsically disordered proteins are constitutively subject to this surveillance raises the question of how such mRNAs are expressed at all. Disordered proteins are abundant, functional, and in many cases, essential.¹⁶ One possibility is that the surveillance imposes a tolerated cost on this class of transcripts, setting their steady-state abundance lower than their transcription rate would otherwise dictate. Another is that specific features, such as co-translational binding partners or ribosome-associated factors acting downstream of Zuo1, exempt particular disordered transcripts from decay. Distinguishing these possibilities will require measurement of decay rates on individual targets rather than steady-state abundance.

The enrichment of positively charged residues among Zuo1 targets is consistent with a known determinant of elongation rate. Positively charged amino acids slow ribosomes through electrostatic interaction with the negatively charged exit tunnel surface.^17,18 In the present dataset, however, the charge enrichment is not separable from the predicted disorder metrics, since basic residues are themselves modestly disorder-promoting. Enrichment of positively charged residues and shorter length was also observed in transcripts downregulated by the protein misfolding-inducing drug AZC and rescued by Zuo1 deletion in our previous work.³ Transcripts containing heat and oxidative stress-induced ribosome stalling were also enriched in basic residues approximately 30 amino acids upstream of stall sites,³ where the nascent chain emerges from the exit channel. Whether Zuo1 responds to nascent chain charge as such or to the conformational consequences of a highly charged sequence cannot be resolved from composition analysis alone.

The proline result points in the opposite direction and is informative for that reason. Zuo1 targets are depleted rather than enriched in proline, distinguishing this surveillance from proline-induced elongation stalling, which is resolved by eIF5A.^19,20 Zuo1-dependent slowing therefore appears to act on a substrate class defined by the conformational state of the emerging chain rather than by the intrinsic difficulty of peptide bond formation, and the two mechanisms are unlikely to converge on the same transcripts.

Understanding the endogenous substrates of Zuo1-mediated surveillance has implications beyond yeast. Our findings indicate that transcripts encoding proteins that fail to acquire stable structure are inherently susceptible to this quality control mechanism. Characterizing the features that trigger the pathway provides a framework for predicting, and potentially manipulating, which transcripts are subject to it, with applications in both basic research and therapeutic mRNA design.

Limitations of the study

RNA-sequencing measures steady-state abundance, not stability, and zuo1Δ cells carry pleiotropic defects. Direct decay rate measurements on a panel of Zuo1 targets, such as transcription shutoff experiments to measure half-lives of individual transcripts, would be needed to establish that accumulation reflects stabilization rather than altered transcription. Additionally, the ribosome occupancy trend, while robust to gene set selection, choice of disorder measure, and adjustment for expression level and open reading frame length, is shallow. Positional analysis of ribosome density within individual target open reading frames, and examination of cells deficient in co-translational decay factors, would further strengthen the causal link between Zuo1-dependent ribosome slowing and transcript degradation.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Jeff Coller (jmcoller@jhmi.edu).

Materials availability

Yeast strains in this study are available from the lead contact upon request. This study did not generate new unique reagents.

Data and code availability

• •RNA-seq and ribosome profiling data are publicly accessible at NCBI’s Gene Expression Omnibus (GEO: GSE345693 and GSE345702).
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

The authors thank the Johns Hopkins University School of Medicine Genetics Resources Core Facility and Single Cell and Transcriptomics Core for technical support. This work was supported by Bloomberg Philanthropies (J.C.), Cystic Fibrosis Foundation Path to a Cure Postdoctoral Research Fellowship Award 006606F223 (S.C.), and the National Institutes of Health: R35GM144114 (J.C.).

Author contributions

Conceptualization, methodology, resources, data curation, software, formal analysis, writing, and funding acquisition, S.C. and J.C.; investigation, S.C.; project administration and supervision, J.C.

Declaration of interests

The authors declare no competing interests.

Declaration of generative AI and AI-assisted technologies in the manuscript preparation process

ChatGPT-4 (OpenAI) and Claude (Anthropic) were used to assist with the writing process for this manuscript. The authors reviewed and edited the content produced by the generative AI and take full responsibility for the content of the publication.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, peptides, and recombinant proteins
RNase I	EpiCentre	N6901K
SUPERase∗In RNase inhibitor	Invitrogen	AM2694
T4 Rnl2(tr) K227Q ligase	NEB	M0351
5′-deadenylase	EpiCentre	DA11101K
RecJ exonuclease	EpiCentre	RJ411250
CircLigase II	EpiCentre	CL9021K
Protoscript II RT	NEB	M0368
RNaseOUT	ThermoFisher	10777019
DNase I	ThermoFisher	18068015
TRIzol	Invitrogen	15596026
KAPA SYBR® FAST	Roche	08203407001/KM4107
Critical commercial assays
Ultra II RNA library prep kit	NEB	E7770
Unique Dual Index	NEB	E6440, E6442
RNA ScreenTape Analysis	Agilent	5067–5576, 5577, 5578
D1000 High Sensitivity ScreenTape Analysis	Agilent	5067–5584,5585
QIAquick Gel Extraction kit	Qiagen	28704
QIAseq FastSelect –rRNA Yeast Kit	Qiagen	334215
Qubit RNA High Sensitivity Assay	ThermoFisher	Q32852
Oligo Clean & Concentrator kit	Zymo Research	D4060
Deposited data
Raw data for RNA sequencing	This paper	GEO: GSE345693
Raw data for Ribosome profiling	This paper	GEO: GSE345702
Experimental models: Organisms/strains
ura3, leu2, his3, met15	ResGen Collection	yJC151
ura3, leu2, his3, met15, zuo1:kanMX6	Horizon MATa KO Collection	yJC4067
Software and algorithms
FastQC v0.12.1	Andrews et al.21	RRID:SCR_014583
MultiQC v1.33	Ewels et al.22	RRID:SCR_014982
Fastp v1.3.0	Chen et al.23	RRID:SCR_ 016962
Salmon v1.11.4	Patro et al.24	RRID:SCR_017036
DEseq2	Love et al.25	RRID:SCR_015687
apeglm v1.32.0	Zhu et al.26	RRID:SCR_026951
Cutadapt v4.9	Martin27	RRID:SCR_011841
Bowtie 2 v2.4.4	Langmead and Salzberg28	RRID:SCR_016368
Samtools v1.23.1	Li et al.29	RRID:SCR_002105
featureCounts v2.1.1	Liao et al.30	RRID:SCR_012919
AlphaFold Protein Structure Database	Jumper et al.9	RRID:SCR_025454
IUPred (short-order configuration)	Mészáros et al.12	RRID:SCR_014632
Espritz (X-ray and NMR configurations)	Walsh et al.13	RRID:SCR_025732
MobiDB-lite	Necci et al.14	RRID:SCR_014542
DisEMBL	Linding et al.15	RRID:SCR_003070
HSF1base	Kovács et al.31	RRID:SCR_015530
R v4.3.2	R Core Team	RRID:SCR_001905
Python v4.3.2	Python Software Foundation	RRID:SCR_008394
Other
Thick-wall polycarbonate tube	Beckman Coulter	343778
TLA120.2 rotor	Beckman Coulter	TLA120.2
4–15% Criterion TGX gel	Biorad	5671084
Spex 6775 freezing grinding mill	Spex sample prep	N/A
Qubit Fluorometer 2.0	Invitrogen	RRID:SCR_020553
Illumina NovaSeq X Plus	Illumina	RRID:SCR_024568

Experimental model and study participant details

Yeast strains, genotyping and culture

Both the wild-type and zuo1Δ yeast strains used in this study are based on BY4741. zuo1Δ strain was validated by PCR amplification of the genomic DNA using various combinations of primers internal to or flanking the Zuo1 coding sequence. For samples used in RNA-sequencing and ribosome profiling, cells were grown in Yeast extract Peptone Dextrose (YPD) media at 24°C starting from OD₆₀₀ of 0.1, and collected at mid-log phase at around OD₆₀₀ 0.4.

Method details

RNA isolation

Upon collection, yeast cultures were filtered using a 0.45 μm Whatman membrane (GE Life Sciences, USA) using a vacuum filter apparatus. After filtration, Cells were immediately scraped from the filter with a straight-edge metal spatula and plunged into 50 mL liquid nitrogen. Next, samples were ground using a Spex 6775 freezing grinding mill (Spex sample prep, USA) in the presence of 10 mL lysis buffer (20 mM Tris (pH 7.4), 150 mM NaCl, 5 mM MgCl₂, 1 mM DTT, 100 μg/mL cycloheximide, 1% Triton X-100, 25 U/mL DNase I). The RNA concentration of each cell lysate was determined using Qubit RNA HS Assay Kit (ThermoFisher Q32852) with the Qubit Fluorometer 2.0 (Invitrogen, USA).

RNA-sequencing library prep

After confirming the quality of RNA via ScreenTape Analysis (Agilent 5067–5576, 5577, 5578), 1 μg total RNA went through ribosomal RNA (rRNA) depletion using QIAseq FastSelect rRNA Yeast kit (Qiagen 334215) according to the manufacturer’s instructions. mRNA libraries were prepared using Ultra II RNA library prep kit (NEB E7770L) and Unique Dual Index (NEB E6440S, E6442S). Libraries were quantified with DNA High Sensitivity ScreenTape Analysis (Agilent 5067–5584, 5585) and the pooled library was sequenced on a NovaSeq X Plus in Johns Hopkins Genetic Resources Core Facility.

Ribosome profiling library prep

RNA was diluted to 300 μL with polysome buffer (20 mM Tris (pH 7.4), 150 mM NaCl, 5 mM MgCl₂, 1 mM DTT, 100 μg/mL cycloheximide), digested using 4 U RNase I (EpiCentre, USA N6901K) per 10 μg RNA for 10 min at room temperature and quenched with SUPERase∗In RNase inhibitor (Invitrogen, USA AM2694). To pellet the ribosomes, samples were transferred to a 13 mm × 51 mm polycarbonate ultracentrifuge tube containing 0.7 mL sucrose cushion (1 M sucrose in polysome buffer) and centrifuged in Beckman TLA-120.2 rotor at 55,000 rpm for 1.5 h. Translucent ribosome pellets were resuspended in 300 μL polysome buffer followed by RNA purification using TRIzol reagent (Invitrogen, USA 15596026), according to the manufacturer’s instructions. Samples were size-separated by running on 15% polyacrylamide TBE-Urea gels for 60 min at 200 V followed by gel-purification of the approximately 20–40 nucleotide (nt) length ribosome protected fragments (RPFs) using 17–34 nt length oligonucleotides as size markers. RPFs were treated with T4 PNK (NEB M0201S) to heal the 2′–3′ cyclic phosphate that results from RNase I cleavage. After end-healing reaction, pre-adenylated adapter oligonucleotides (35 nt length) were ligated to RPFs using T4 Rnl2(tr) K227Q ligase (NEB M0351S). Excess adapter oligos in the reaction were depleted by adding yeast 5′-deadenylase (EpiCentre, USA DA11101K) and RecJ exonuclease (EpiCentre, USA RJ411250) enzymes directly to the ligase reaction and incubated at 30°C for 45 min. Samples were purified using Oligo Clean & Concentrator kit (Zymo Research D4060) followed by ribosomal RNA (rRNA) depletion using QIAseq FastSelect rRNA Yeast kit (Qiagen 334215) according to the manufacturer’s instructions. Ligated RNA samples were reverse transcribed using Protoscript II RT (NEB M0368). Reverse transcription products were separated from the unextended primer products by polyacrylamide TBE-Urea gel electrophoresis followed by gel purification of the full-length cDNAs (80–120 nt length). Size-selected cDNAs were circularized using CircLigase II (EpiCentre, USA CL9021K). 10–15 rounds of PCR enrichment of the libraries were performed with the proper forward and reverse primers. Enriched libraries were used for cluster generation and sequencing. Paired-end (150 bp) sequencing was carried out for five cDNA libraries (five biological replicates of each sample) on a NovaSeq X Plus in Johns Hopkins Genetic Resources Core Facility.

Analysis of RNA sequencing data

Raw FASTQ files (five biological replicates each of wild-type and zuo1Δ) were assessed with FastQC v0.12.1²¹ and aggregated with MultiQC v1.33.²² Reads were trimmed with fastp v1.3.0²³ using --detect_adapter_for_pe --qualified_quality_phred 20 --length_required 36, with adapter sequences detected automatically. Trimmed reads were quantified with Salmon v1.11.4²⁴ using -l A --validateMappings --gcBias --seqBias. The Salmon index was decoy-aware and built with k = 31 from the Saccharomyces cerevisiae R64-1-1 cDNA reference (Ensembl Fungi, assembly GCA_000146045.2, strain S288C) concatenated with the corresponding genomic sequence as decoys, comprising 6,571 targets and 17 decoy sequences. Transcript-level estimates were imported with tximport v1.38.2³² and summarized to gene level, with average transcript length offsets retained and no rounding applied. Differential expression was performed with DESeq2 v1.50.2²⁵ using the design ∼ condition, with wild-type as the reference level. Genes with fewer than 10 total counts across all samples were removed prior to testing. Log₂ fold change estimates were shrunk with apeglm v1.32.0.²⁶ Significance was assessed at an adjusted p value below 0.05 using the Benjamini-Hochberg procedure.³³ No samples were excluded.

Open reading frames were classified using the metadata pulled from the Saccharomyces Genome Database (SGD).⁶ Analyses were restricted to Verified and Uncharacterized open reading frames. Genes were excluded if they fell within the chromosome XII ribosomal DNA repeat (chrXII:450,000–491,000, or annotated as TAR1, matching YLR154W-[A-Z], or beginning with RDN), or if they corresponded to subtelomeric Y′ elements (annotated as Y_prime_element, described as a Y′ or telomeric Y′ element, or beginning with YRF1, together with YFL064C and YFL066C). This yielded a universe of 5,133 genes from 5,167 expressed genes. Zuo1 target transcripts were defined as those significantly increased in zuo1Δ, yielding 1,461 transcripts; 1,600 were significantly decreased and 2,072 did not change.

Analysis of ribosome profiling data

Raw FASTQ files (five biological replicates each of wild-type and zuo1Δ) were assessed with FastQC v0.12.1.²¹ Adapter and quality trimming was performed with cutadapt²⁷ using -a CACTCGGGCACCAAGGACAGATCGGAAGAGCACACGTCT -q 20 -m 25 -M 45 --trim-n. No 5′ trimming was applied. Reads mapping to noncoding RNA were removed by alignment to a Bowtie2²⁸ index built from the ribosomal RNA, transfer RNA, small nuclear RNA, and small nucleolar RNA features of the Ensembl annotation. Remaining reads were filtered to an aligned length of 20–35 nucleotides, calculated as the sum of CIGAR match operations.²⁹ Filtered reads were aligned to the S. cerevisiae genome with Bowtie2 v2.5.5.²⁸ Alignments were sorted with samtools v1.23.1,²⁹ and reads mapping within the chromosome XII ribosomal DNA repeat (chrXII:450,000–491,000) were removed. Ribosome-protected fragments were assigned to features with featureCounts v2.1.1³⁰ against the Ensembl annotation, counting over the full coding sequence without exclusion of terminal codons. No samples were excluded.

Protein structure and disorder prediction

Mean per-residue pLDDT values were obtained from the AlphaFold Protein Structure Database.⁹ Five alignment-free disorder predictors were used: IUPred (short-disorder configuration),¹² ESpritz (X-ray and NMR configurations),¹³ MobiDB-lite,¹⁴ and DisEMBL.¹⁵ Positive charge fraction was calculated as the combined frequency of lysine and arginine residues, and proline frequency as the fraction of proline residues.

Ribosome occupancy

Ribosome occupancy was calculated as the ratio of RPF counts to mRNA counts. Counts from each assay were normalized separately using the DESeq2²⁵ median-of-ratios method, and condition means were taken across the five normalized replicates for each genotype. Transcripts were retained if at least three replicates carried 10 or more counts in both assays. Paired comparisons between genotypes were made on the per-transcript log₂ ratio of wild-type to zuo1Δ ribosome occupancy. Location was estimated by the Hodges-Lehmann estimator,³⁴ with confidence intervals obtained by inverting the Wilcoxon signed-rank test, and monotone trend across disorder quintiles was assessed by the Jonckheere-Terpstra test.^35,36

Quantification and statistical analysis

Statistical analysis

Group comparisons used two-sided Mann-Whitney tests, with Cliff’s delta³⁷ as a non-parametric effect size and 95% confidence intervals obtained by bootstrap resampling. Correlations between protein features and log₂ fold change are reported as Spearman rank correlations across all transcripts. Trend fits are generalized additive models with cubic spline smoothers, fitted with mgcv. Propensity matching used nearest-neighbor matching on standardized covariates, with balance assessed after matching. Joint models were logistic regressions predicting Zuo1 target status from standardized protein features, wild-type expression level, and a standardized disorder term, with variance inflation factors reported. Analyses were performed in R and Python. Statistical significance for all figures is indicated as follows: n.s. p > 0.05 | ∗p ≤ 0.05 | ∗∗p ≤ 0.01 | ∗∗∗p ≤ 0.001 | ∗∗∗∗p ≤ 0.0001.