A common germline 3′ UTR variant controls DDIT4 translation efficiency and stress responses in cancer cells

Summary

Systematic strategies for identifying and classifying variants that directly modulate ribosome engagement are lacking. Here, we develop and extensively validate a pipeline to identify translation-associated single-nucleotide polymorphisms (tranSNPs) by integrating analysis of total and polysome-associated mRNAs. Common variants showing allele-specific enrichment in polysomes, without corresponding changes in mRNA levels, were uncovered. We focused on rs1053639, a frequent germline SNP located in the 3′ UTR of DNA damage-inducible transcript 4 (DDIT4), a key negative regulator of mTORC1. Using edited cell models, we showed that this variant controls DDIT4 protein output. Mechanistically, the two rs1053639 alleles differ in their interaction with the RNA-binding protein RBMX, leading to altered mRNA subcellular localization and translation. Functionally, allele-specific regulation of DDIT4 translation impacts stress-dependent mTORC1 repression and autophagy, resulting in divergent competitive fitness in co-culture or xenograft assays and response to therapeutic perturbations. Analysis of cancer patient datasets revealed an association between rs1053639 genotype and clinical outcome in colorectal cancer.

Graphical abstract

Highlights

• •HCT116 translatome revealed allele-specific mRNA translation for 40 SNPs
• •DDIT4 rs1053639 (T/A) alleles differed in mRNA localization and protein expression
• •RBMX modulates DDIT4 mRNA localization and translation in an allele-specific manner
• •rs1053639 AA homozygous cells showed weaker mTOR inhibition and higher proliferation

Cancer; Genomics; Molecular biology

Introduction

Many studies in the last decade have demonstrated the importance of translational regulation in shaping cell phenotype.^1,2,3,4 Instead, relatively few studies focused on the identification and characterization of genetic variants impacting directly on mRNA translation, although allele-specific expression finely modulating epigenetic and transcriptional regulation has been revealed as critical for many pathophysiological processes.^5,6,7,8 To investigate instances of allele-specific mRNA translation efficiency (TE), we recently developed a method that focuses on single-nucleotide polymorphisms (SNP) and integrates information from SNP genotyping with RNA sequencing (RNA-seq) data performed on both total and polysome-bound mRNAs.⁹ Briefly, heterozygous SNPs mapped by genotyping assays are followed in RNA-seq data by calculating allelic fractions (AFs) for both total and polysomal RNA. Coverage threshold, consistency among biological replicates, and the differences in AF between paired total and polysomal RNA samples are used to nominate candidate translation-associated SNPs (tranSNPs). The method assumes that the translatome, defined as the ensemble of transcripts that co-sediment with polysomes on sucrose gradients, represents a proxy for the proteome.¹⁰ Also, it assumes that if two alleles expressed by a cell can be distinguished due to heterozygous SNPs or other types of variants, a difference in their relative abundance within the polysomes would signal an instance of allele-specific difference in TE. Hence, the name tranSNPs is attributed to the variants emerging from our pipeline.⁹ Besides polysomal profiling, Ribo-seq is another approach typically used to characterize the translatome. The method can reach a single-nucleotide positional resolution of the ribosomes along a transcript. The polysome-seq variant protocol could be potentially applied to our scope, as 80S monosomes, whose actual translational status is still somewhat debated,¹¹ are removed before processing the samples by RNase treatment to retrieve ribosome-protected fragments.¹² However, we decided to use polysomal profiling to increase the possibility of obtaining robust coverage also in 5′- and 3′-UTR regions of mRNAs. Our earlier study identified 147 tranSNPs, corresponding to about 4% of analyzable SNPs in a dataset obtained in the MCF7 cancer cell line. Of those, 39 were located in the 3′ UTR⁹ and could impact, directly or through structural features, the binding of trans-factors that, in turn, can affect TE or mRNA stability.

We have now extended our analysis using data produced in the HCT116 p53 wild-type cell line cultured in control conditions or treated with the MDM2 inhibitor nutlin.¹³ While tranSNPs could potentially impact multiple cellular phenotypes and outcomes, we decided to continue to focus on a non-genotoxic p53-activating treatment in cancer cells due to profound changes in mRNA translation in cancer⁴ and the growing evidence linking the tumor suppressor p53 to various features of mRNA translation control. For example, p53 was shown to modulate eIF4EBP1, mTORC1, and endoplasmic reticulum (ER) stress,^14,15,16 and, through the modulation of ribosomal proteins (RBPs) and microRNAs, to impact the fate of specific sets of mRNAs.^17,18,19 We identified 40 candidate tranSNPs, corresponding to about 7% of analyzable SNPs. Among these tranSNP candidates, four were prioritized for validation experiments considering minor allele frequency (MAF), the relevance of the associated genes in cancer, the position along the gene sequence, and the presence of predictions for overlapping trans-factor binding sites. One candidate, rs1053639 T/A in the 3′ UTR of the DNA damage-inducible transcript 4 (DDIT4) (REDD1 and RTP801) gene, caught our particular interest. DDIT4 is a negative modulator of mTORC1, is an established target of p53, RFX7, and ATF4, and acts as a tumor suppressor gene.^20,21 We generated HCT116 cell line trios for rs1053639 exploiting a CRISPR-Cas9-mediated knockin strategy²² starting from the heterozygous (AT) parental cells. Homozygous TT and AA cell clones were characterized, which revealed differences in DDIT4 protein expression both in basal and induced conditions, alterations in the relative expression of mTOR markers, and changes in mRNA subcellular localization and translatability. Mechanistically, we implicated the RNA-binding protein RBMX in a distinct fate of DDIT4 mRNA for rs1053639 alleles. We also established higher relative proliferation and fitness for A-homozygous cells in both in vitro and xenograft experiments in zebrafish embryos and explored the correlation between rs1053639 alleles and DDIT4 protein expression in primary tumor samples and the prognostic value of this SNP using The Cancer Genome Atlas (TCGA) data.

Results

TranSNPs can be identified in mock- and nutlin-treated HCT116 cells

Single-end RNA-seq data obtained from HCT116 cells under either mock or 10 μM nutlin treatment for 16 h were processed starting from raw files by trimming, aligning, and deduplication steps (see STAR Methods). SNPs in exonic and UTR regions were extracted from dbSNP. A list of heterozygous SNPs was compiled based on AF values between 0.2 and 0.8 in at least one of the four replicates in one of the two fractions sequenced (total and polysomal RNA), setting a minimal threshold of 20 reads for AF calculations (Figure S1). In addition, only the SNPs with AF measures available for all four biological replicates, the two fractions, and across the two treatments were considered to calculate changes in AF between the fractions (N = 582, Table S1). For each SNP, we exploited the concordance among differential AF measures across pairs of polysomal and total RNA biological replicates, performing paired statistics (Figure S1). This approach identified 40 putative tranSNPs (Figure 1A), of which 21 are located in 3′ UTRs (Table S2). Overall, we extended the methodology previously developed⁹ and observed that in HCT116, about 11% of genes with analyzable heterozygous SNPs show a difference in AF between paired total and polysome-bound mRNAs, suggesting allele-specific post-transcriptional and translational control.

Figure 1.

The tranSNPs in the DDIT4, DDX17, and PERP genes are validated by orthogonal approaches

(A) Heatmap view of average differences in allelic fractions (AF) in the comparison between polysome-associated and total RNA for the 40 tranSNPs identified from the RNA-seq data from either mock or nutlin-treatment conditions (Tables S1 and S2 present the AF value for each biological replicate; accession number for the RNA-seq data: GSE95024). AF was computed as the number of reads for the alternative allele over the sum of reads of the alternative and reference alleles.

(B) Distribution across coding and regulatory regions of the transcripts for all analyzable SNPs (left) and for the identified tranSNPs (right). The result of chi-square contingency test is shown.

(C) Plot presenting the absolute delta AF (dAF) and the minor allele frequency (MAF) for the identified tranSNPs located in 3′ UTRs. MAF values were derived from the 1000 Genomes Project phase 3, all individuals. The SNP rsID and associated gene names are shown.

(D) Plots of the AF for the rs1053639 (in DDIT4), rs763121 (in DDX17), and rs6588 (in PERP) tranSNPs obtained through RNA-seq of cytoplasmic total (TOT) and polysome-bound mRNAs (POL). Individual replicates and the AF mean are plotted. ∗p value < 0.05, two-tailed, paired t test.

(E) AF values of the selected tranSNPs calculated from Sanger sequencing electropherograms, comprising also the subpolysomal-RNP to 80S (SUB) fractions. Individual replicates and the AF mean are plotted. ∗p value < 0.05, two-tailed, paired t test.

(F) Results of the dual-luciferase assays presenting the relative activity (mean, standard deviations, and individual biological replicates) for the reference (REF_) and alternative (ALT_) alleles for each of the three selected tranSNPs cloned separately into the pGL4.13 vector immediately downstream of the Firefly luciferase stop codon. Data were normalized using Renilla luciferase. ∗∗p value < 0.01; ∗∗∗p value < 0.001; ∗∗∗∗p value < 0.0001; two-tailed, unpaired t test. See also Figure S1.

Candidate tranSNPs for downstream analyses and validation were chosen considering several SNP features (Figures 1B and 1C; Tables S1 and S2) and information about the associated genes. Higher priority was given to SNPs with high magnitude of the AF imbalance, high MAF, and positioned in 3′ UTR of genes whose expression is altered in cancer or whose function is associated with cancer-relevant features, especially p53-related. We selected a total of four SNPs for validation (Figures 1D–1F and S1). First, we sought to confirm differential allele-specific expression found in polysomes versus total RNA. Sucrose gradient density sedimentation of cytoplasmic lysates was performed,¹⁰ and fractions were collected and analyzed by UV absorbance. RNA was recovered from the combined polysomal fractions and fractions corresponding to ribosomal subunits and 80S monosomes. Total RNA was also extracted. Amplicons were generated using primers specific for sequences surrounding the chosen SNPs (Table S4), Sanger sequenced, and the electropherograms were analyzed, expressing the results as AF (alternative allele signal over the sum of alternative + reference allele signals), as done for the RNA-seq data. For three out of the four chosen SNPs, an imbalance consistent in direction with what was measured in the RNA-seq dataset was found (Figures 1E, S1D, and S1F).

We also used an Ampliseq-based targeted re-sequencing approach to measure AFs from total and polysome-associated mRNAs, obtaining results consistent with the RNA-seq and Sanger sequencing (Figure S1G). Next, we asked whether the candidate tranSNPs could exhibit allele-specific differences in activity when extracted from their native context. To this aim, we cloned ∼500 nt fragments of the 3′ UTR sequences harboring those SNPs downstream of a luciferase plasmid reporter, using HCT116 genomic DNA. Given the heterozygous state of the cell line, we retrieved clones for both alleles for each SNP. Dual-luciferase assays were then performed in HCT116 cells, with results further supporting the observation that the SNPs and their immediate surroundings are sufficient to show allele-specific differences in reporter activity (Figures 1F and S1).

Editing of HCT116 cells to obtain the three genotypes of rs1053639 tranSNP

It is worth noting that while the heterozygous state and the comparison of AF among two RNA compartments (total versus polysomal) are necessary to enable reliable identification of tranSNPs in our approach, we expect that the homozygous state for alternative and reference alleles would reveal the highest level of phenotypic impact of this category of genetic variants. Hence, our next step was constructing potentially isogenic cell clones by CRISPR-Cas9-mediated editing. We exploited the CRISPR-Cas9-mediated knockin delivering the Cas9 and the guide RNA (gRNA) as a ribonucleoparticle (RNP) to reduce its off-target potential²² (Figure S2A). Based on the collective RNA-seq, validation data, editing amenability, and the relevance of the gene in cancer, including its functional links with p53,^20,23,24 the T/A rs1053639 in DDIT4 was a promising tranSNP.²⁵ Homology-directed repair (HDR) efficiency, as measured by the ICE Synthego analysis web tool²⁶ and two sgRNA guides, ranged from 5% to 13% for the bulk populations. ∼150 clones were isolated and characterized by Sanger sequencing. We obtained at least four clones homozygous for the rs1053639 T allele and four homozygous for the A allele. However, three clones (#3 TT, #5, and #6 AA) contained the same GGT 3 nt deletion, starting three nucleotides after the SNP site. We also obtained two clones bearing either a 9-nucleotide deletion (#7) or a 29-nucleotide deletion (#8) encompassing the SNP position (Figure S2B, Table S6, and STAR Methods). We considered at least two clones among the rs1053639 TT (#2, #3, #12, and #13) and AA (#5, #6, #7, and #10), along with the unedited AT clone (#1), to overcome the potential influence of clonal selection and adaptation and the above-mentioned editing limitations (Figures 2A, 2B, and S2C). We then used these cell models to study the effect of the SNP on various phenotypic assays. No evidence of editing was found at CRISPOR-predicted²⁷ off-target sites (Figures S2D and S2E).

Figure 2.

rs1053639 tranSNP impacts DDIT4 protein expression in HCT116-derived cells with the three genotypes

HCT116 cells were edited by CRISPR-Cas9 RNPs and donor DNA to obtain cell trios with the three genotypes for the rs1053639 tranSNP located in the proximal portion of the DDIT4 3′ UTR.

(A) Western blot analysis of DDIT4 protein expression in HCT116 cells (A/T heterozygous) and DDIT4 rs1053639 homozygous clones (TT and AA). β-Tubulin was used as a reference.

(B) Densitometry analysis of the blots from several biological replicates shows a significant difference in relative DDIT4 protein abundance with the different tranSNP genotypes. Mean ± SD and individual biological replicates are presented. ∗∗∗p value < 0.001; ordinary one-way ANOVA.

(C) Western blot analysis of DDIT4 protein in HCT116 TT and AA homozygous clones for the rs1053639 treated with 100 μg/mL cycloheximide (CHX) at the indicated time points. β-Tubulin was used as a reference.

(D) Relative DDIT4 mRNA expression as measured by RT-qPCR and normalized to GAPDH and YWHAZ. Mean ± SD and individual biological replicates are presented. The Student’s t test showed no significant differences in DDIT4 mRNA expression in the different clones. ns, not significant.

(E) Estimation of DDIT4 mRNA stability following inhibition of transcription by actinomycin D treatment (10 μg/mL). The curve shows residual relative DDIT4 mRNA levels (mean and standard deviation from at least 3 biological replicates) at various treatment time points, measured by RT-qPCR decay in HCT116 clones harboring TT or AA genotypes. The estimated half-lives of the DDIT4 mRNA in TT and AA homozygous clones are indicated.

(F) HCT116 clones were exposed to 100 nM thapsigargin (Tg) or vehicle for 4 h, and the induction of DDIT4 mRNA was measured by RT-qPCR with clones of the three rs1053639 genotypes. Average, standard deviation, and individual replicates are presented. ns, not significant, Student’s t test.

(G) Western blot and densitometry analysis showing relative DDIT4 protein induction upon thapsigargin (Tg, 100 nM) treatment of the three rs1053639 genotypes. β-Tubulin served as a control. Mean ± SD and individual replicates are shown. ∗∗∗p value < 0.001, unpaired t test.

(H and I) Cytoplasmic lysates from HCT116 rs1053639 homozygous (TT and AA) clones treated with thapsigargin (100 nM) or vehicle for 4 h were subjected to fractionation through sucrose density gradients (15%–50%). Polysome profiles are shown in Figure S3C. The distribution of DDIT4 mRNA (mock or thapsigargin conditions) across the ten gradient fractions was analyzed in both genotypes by RT-qPCR.²⁸ The relative mRNA percentage in each fraction was plotted. Three independent biological experiments were performed. Presented are the average proportions and the standard deviations of the three replicates. Statistically significant p values at the unpaired t test are indicated by an asterisk. The p values are shown below the images. The shading pattern in each image refers to the sucrose-gradient fractionations (Figure S2F). RNP = light fraction corresponding to free RNA and ribonucleoprotein complexes; SUB = ribosomal subunits; 80S = monosomes. Fractions corresponding to the light (L-POL) and heavy polysomes (H-POL) are also highlighted. See also Figures S2 and S3.

Higher DDIT4 protein levels in rs1053639 TT homozygous clones

First, we tested the impact of the tranSNP status on DDIT4 steady-state protein levels (Figures 2A, 2B, and S2C). TT clones exhibited significantly higher relative protein expression than AA clones, with the AT cells intermediate. This result was consistent with the change in AF from RNA-seq or Sanger sequencing that showed lower A allele representation in the polysomal fraction. The difference in protein levels was not dependent on changes in protein half-life (Figure 2C), nor on mRNA levels and turnover (Figures 2D and 2E). An antisense DDIT4-AS1 transcript corresponding to the 3′ UTR region was reported to affect the stability of the DDIT4 transcript.^29,30 However, DDIT4-AS1 was equally expressed across TT and AA clones, and its expression was significantly lower than that of the DDIT4 transcript (Figure S3A). Also, we did not detect significant allele-specific differences in the expression of DDIT4 mRNA when using primers spanning the two possible exon-exon junctions, indicating that alternative splicing cannot explain the observed phenotype (Figure S3B). Next, we challenged the cells with thapsigargin (100 nM for 4 h),^31,32 an ER stressor³³ that is a strong inducer of DDIT4 expression (Figure 2F). As expected, the treatment resulted in a marked induction of DDIT4 protein levels (Figure 2G). Notably, the difference in expression associated with the SNP genotype was maintained, and the lower expression of DDIT4 in AA clones was even more apparent. DDIT4 is also a recognized transcriptional target of p53, directly or through the transcription factor RFX7.^20,24 Hence, we examined whether the rs1053639 alleles differed in DDIT4 expression, also in response to p53 activation. Results showed that while the MDM2 inhibitor nutlin led to comparable induction of p53 protein across the HCT116 DDIT4 clones, the associated induction in DDIT4 protein was higher for the TT homozygous clones while the AA clones did not show significant DDIT4 induction (Figures S4A–S4C). Since we observed higher polysomal association and higher DDIT4 protein levels from the rs1053639 T over the A allele, we examined at higher resolution the relative DDIT4 mRNA distribution across the sucrose gradient in mock conditions and upon thapsigargin treatment (100 nM for 4 h). For this experiment, RNA was recovered from ten individual fractions along the polysome profile (fractions 3 to 12; Figure S3C), and the DDIT4 mRNA distribution, along with CDKN1A and RPS26 controls, was measured by qPCR. Results were plotted as relative abundance, as previously described.²⁸ In the mock condition, no significant differences in apparent translation efficiency for the DDIT4 mRNA were observed between TT and AA homozygous cells (Figure 2H). No differences were also apparent for the CDKN1A and RPS26 mRNAs, except for a slight increase in the proportion of CDKN1A in subpolysomal fraction 5 in TT cells (Figures S3D and S3F). When the same analysis was performed after thapsigargin treatment, a shift was visible for the DDIT4 mRNA in TT cells, with significant higher proportion in polysomal fraction 10 and corresponding reduction in subpolysomal and light polysomal fractions, significant for fractions 6 and 8 (Figure 2I). Thapsigargin treatment led to a general reduction in the polysome association for the three mRNAs, which was more pronounced for the RPS26 mRNA (Figures S3E and S3G). A higher proportion of RPS26 mRNA was observed in the subpolysomal fractions 5 and 6 in AA cells. Collectively, these results are consistent with the higher DDIT4 protein expression in TT cells, particularly in response to thapsigargin treatment.

Luciferase reporter assays show higher TE for the T allele

We confirmed that the TT and AA homozygous HCT116 clones exhibited equivalent activity when 3′-UTR portions with the rs1053639 alleles were cloned downstream of a luciferase reporter and transfected into HCT116 edited clones (Figure 3A). This result confirmed that the DDIT4 3′-UTR sequences, down to 109 nt in length surrounding the rs1053639 position, are sufficient to show allele-specific differences. The fact that those differences were equivalent in TT and AA edited cells also argues against clonal adaptation. Interestingly, a significant difference in the reporter activity was also observed in A549 and RPE-1 cells (Figures 3B and 3C). Finally, considering that the dual-luciferase assays were normalized to firefly mRNA levels, the allele-specific differences in reporter activity can be ascribed to post-transcriptional regulation.

Figure 3.

DDIT4 rs1053639 allelic differences in gene reporter activity across different cell lines

(A–C) Dual-luciferase assays showing relative activity (average, standard deviation, and individual replicates) for the reference (Ref) and alternative (Alt) alleles. Two fragments of the DDIT4 3′ UTR comprising 100 (labeled as short) and ∼500 nt (labeled as long) centered on the rs1053639 position were cloned separately into the pGL4.13 vectors immediately downstream of the firefly luciferase stop codon. These constructs were transiently transfected in the indicated HCT116 edited clones (A), or the wild-type A549 cells (B) or the wild-type RPE-1 cells. (C) Data were normalized to Renilla luciferase activity and relative to firefly mRNA levels, which were measured by RT-qPCR. Mean ± SD and individual biological replicates are presented. _∗p value < 0.05, ∗∗∗∗p value < 0.0001, two-tailed, unpaired t test.

The rs1053639 tranSNP genotype impacts DDIT4 mRNA fate through the RBMX protein

We developed predictions for allele-specific binding starting from a list of 53 RNA-binding proteins (RBPs) with consensus binding motifs,³⁴ using the TESS software³⁵ (Table S3). The predictions and allele-specific binding scores indicated that ACO1, YBX1, and particularly RBMX were RBPs that might differentially bind to the T/A alleles. Recent PAR-CLIP data in HEK293T cells³⁶ highlighted that RBMX binds in the 3′-UTR of DDIT4 mRNA; therefore, we decided to focus on this target RBP. RBMX protein is mainly nuclear, can interact with structural features of mRNAs, and was reported to modulate mRNA splicing, processing, and degradation.^37,38,39,40 Hence, we performed RNA electromobility shift assays using the full-length recombinant RBMX protein and RNA probes mimicking the secondary structure of the T and A alleles of the endogenous DDIT4 mRNA. For the RNA probe design, we took advantage of three learning-based methods (EternaFold, CONTRAfold, and Ufold), two thermodynamic-based methods (ViennaRNA and RNAstructure), and one hybrid method (MXFold2).^41,42,43,44 Although the global structure of the 3′ UTR was not significantly impacted, significant differences in the local secondary structure were observed after changing the base at the SNP site. RNA electrophoretic mobility shift assay (EMSA) revealed that both RNA probes are bound by RBMX in vitro, with a possible preference for the T allele (Figure 4A). We further explored the publicly available web service AlphaFold 3 to generate accurate structure predictions for protein-RNA interactions. Interestingly, AlphaFold 3 provided several models confirming that the T allele RNA EMSA probe is located in a more favorable spatial conformation than the A allele probe for RBMX binding (Figure S5A). It also predicts that the C-terminal low-complexity domain of RBMX binds to DDIT4 mRNA.

Figure 4.

The rs1053639 tranSNP genotype impacts mRNA fate through the RBMX protein

(A) Structural RNA probes were designed based on RNA folding prediction tools on the endogenous DDIT4 3′ UTR sequence. RNA EMSA was performed using the full-length recombinant RBMX protein and the above-mentioned RNA probes. In the “0” lane, the RNA probe was incubated with the binding buffer and the protein elution buffer. RNA-protein complexes were resolved on a native 6% gel. Right, quantification of the shifted bands in the RNA electromobility shift assay (REMSA) experiments. Error bars plot the standard deviations among replicates. The apparent Kds obtained using non-linear regression fit are shown.

(B) The endogenous RBMX protein was immunoprecipitated from a whole-cell lysate in TT and AA clones. Results are expressed as fold enrichment of the target DDIT4 mRNA relative to IgG. Three independent biological replicates were performed. Mean ± SD are shown. ∗∗p value < 0.01, unpaired t test.

(C) Cell fractionation was performed in HCT116 cells, and protein levels were evaluated. Lamin A/C (70–75 kDa) was used as a marker of the nuclear fraction, and β-tubulin (55 kDa) was used as a marker of the cytoplasmic fraction. The DDIT4 protein was found exclusively in the cytoplasm. RBMX is mostly a nuclear protein, as previously reported. WCL, whole-cell lysate; CYTO, cytoplasm.

(D) RNA levels were evaluated in TT and AA cells upon cell fractionation, followed by RT-qPCR. The distribution of DDIT4 mRNA is significantly different among TT and AA cells, with DDIT4 mRNA from TT cells being considerably more enriched in the cytoplasm. Data represent the mean ± SD of at least four independent experiments. The relative expression of the gene in the cytoplasm is calculated as the fold change of the Ct in the cytoplasm compared to the Ct in the nucleus, with nuclear and cytoplasmic RNAs eluted in the same volume of water. ∗∗∗∗p value < 0.0001. Unpaired t test.

(E) DDIT4 mRNA levels in control (siNT) and in RBMX knocked down (siRBMX) TT and AA cells were evaluated upon cell fractionation and RT-qPCR. The cytoplasmic localization of DDIT4 mRNAs was slightly reduced by RBMX depletion. Data represent the mean ± SD of three independent experiments. ∗p value = 0.0114; ∗∗p value = 0.0031; unpaired t test.

(F) DDIT4 protein levels in control (siNT) and RBMX-depleted (siRBMX) TT and AA cells were evaluated upon cell fractionation, followed by western blot. Representative blot of DDIT4 protein levels across TT and AA clones (top). Densitometry analysis (mean ± SD) of the blots from several biological replicates shows that DDIT4 protein levels in TT clones decrease to the levels of the AA clones upon RBMX knockdown (bottom). ∗∗p value < 0.01, unpaired t test.

(G) Cytoplasmic RBMX proteins associate with ribosomal subunits and polysomes. Sucrose-gradient fractionation of cytoplasmic lysates of HCT116 cells in control (top) and thapsigargin treatment (Tg) (100 nM for 4 h) (bottom). Proteins were extracted from each fraction and used for western blotting. RBMX was immunodetected together with the ribosomal protein RPS6.

(H) Confocal images of smiFISH-IF of a TT and an AA cell clone after control (siNT), RBMX silencing for 72 h, or thapsigargin treatment (100 nM for 4 h). DDIT4 mRNA is shown in yellow. DAPI staining is shown in blue. Scale bars, 10 μm.

(I) At least three fields (15–20 cells per field) for each coverslip and each condition were imaged, and cytoplasmic spots were quantified using Cell-Profiler v.4.0.7 (Broad Institute, Inc.) and segmented as objects with a typical diameter range of 10–30 pixels. The plot presents the number of DDIT4 mRNA spots per cell and the median of the distribution. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ordinary one-way ANOVA. ns, not significant. See also Figure S5.

We then performed RNA immunoprecipitation (RIP) assays to evaluate this interaction in a cellular context. The antibody used proved to be effective for the RIP protocol (Figure S5B). We first performed RIP in parental HCT116 cells, followed by Sanger sequencing of the input and IP samples. However, we were unable to appreciate quantitative differences in RBMX binding across the two alleles with this assay (Figure S5C). We next used both the rs1053639 TT and AA clones to perform RIP-qPCR. The qPCR analysis revealed that RBMX binds to DDIT4 mRNA slightly less than the positive control CSDE1, and we confirmed the preferential binding to the rs1053639 T allele, particularly in the cytoplasm (Figures 4B, S5D, and S5E). Given the nuclear localization of RBMX and its binding to DDIT4 mRNA, we investigated the relative subcellular localization of DDIT4 mRNA alleles. Specifically, nuclear and cytoplasmic fractions were obtained and verified at both protein (Figure 4C) and RNA levels (Figure S5G). Results indicate that the DDIT4 mRNA exhibits higher cytoplasmic localization in the TT clones, whereas it is significantly more nuclear in the AA clones (Figure 4D). To shed light on the contribution of RBMX to this phenotype, we confirmed that our edited cells expressed it at comparable levels and used small interfering RNAs (siRNAs) to deplete its expression by ∼50% (Figure S5F). Nuclear and cytoplasmic fractions upon RBMX silencing were verified at the RNA level (Figure S5H). RBMX silencing led to a slight increase in nuclear DDIT4 mRNAs in TT clones (Figure 4E) and, importantly, to a significant decrease in DDIT4 protein levels (Figure 4F). No significant changes were observed in AA cells. Notably, although predominantly a nuclear protein in HCT116 cells (Figure 4C), cytoplasmic RBMX seems to associate with the polysomes (Figures 4G and S5I) similarly to the RBP RPS6. Single-molecule fluorescence in situ hybridization (FISH) showed a punctuated distribution of the DDIT4 mRNA, particularly in the perinuclear region of the cytoplasm (Figure 4H). The number of granules per cell was higher in the AA compared with the TT clones (Figure 4I), while the spot mean intensity and area were not significantly different across clones (Figures S5J and S5K). Interestingly, the depletion of RBMX led to a significant increase in granularity in TT cells, which reached levels comparable to those observed in the AA ones (Figures 4H, 4I, and S5L). Conversely, thapsigargin treatment increased the number of spots per cell in the AA clones but not in the TT clones. Overall, the number of DDIT4 mRNA spots per cell appears to inversely correlate with DDIT4 protein levels, suggesting that DDIT4 mRNA is accumulated in conditions where it is poorly translated. Hence, we suggest that RBMX contributes to the allele-specific regulation of DDIT4 protein levels through multiple mechanisms, including export, cytoplasmic localization, and translatability of the DDIT4 mRNA.

The rs1053639 tranSNP genotype impacts on mTOR pathway markers

Since DDIT4 is an established inhibitor of mTORC1,^20,45,46 we next sought to investigate whether the effect of rs1053639 on DDIT4 protein levels would impact features of the mTOR pathway as well. In particular, we analyzed the levels of EIF4EBP1 and S6K phosphorylation both in basal culture conditions and in response to thapsigargin treatment. We observed unexpectedly higher levels of relative phospho-eIF4EBP1 and phospho-S6K for TT clones in basal conditions, suggesting that our clones may have acquired compensatory mechanisms to balance their metabolism (Figures 5A–5D).^47,48 Instead, in response to thapsigargin treatment, TT clones were associated with a significant reduction in those mTORC1-dependent phosphorylations, which were not changed in the AA clones. While the AA clone #6 showed a distinct eIF4EBP1 phosphorylation pattern, analysis on additional AA clones revealed that this feature was not attributable to the AA genotype but only to this specific clone. Analysis of the protein levels of the ER stress response markers ATF4, a transcription factor regulating DDIT4 expression^49,50 (Figure 5E), and of the transcript levels of CHOP, spliced XBP1, and SESN2 (Figure 5F) did not reveal differences among HCT116 clones with the three rs1053639 genotypes in response to thapsigargin.

Figure 5.

DDIT4 rs1053639 genotype-specific effect on the mTOR pathway

HCT116 parental cells and clones differing for the rs1053639 homozygous genotypes were treated with either thapsigargin 100 nM (Tg) or vehicle control for 4 h.

(A and C) Cell lysates were analyzed by immunoblotting for the phosphorylation state and levels of the mTORC1 target eIF4EBP1 and S6K proteins.

(B and D) Densitometry analysis of all blots reveals that significant inhibition of the phosphorylation of both eIF4EBP1 and S6K proteins is apparent only in the rs1053639 TT homozygous clones. Mean ± SD and individual replicates are shown. ∗p value < 0.05; ns, not significant; unpaired t test.

(E and F) ER stress response markers are activated in HCT116 cells in response to thapsigargin treatment independently of the rs1053639 genotype. (E) Western blot analysis of ATF4 protein induction upon thapsigargin (Tg) treatment in HCT116 cells (A/T heterozygous) and DDIT4 rs1053639 homozygous clones (TT and AA). β-Tubulin was used as a reference. (F) Relative CHOP, SESN2, and spliced XBP1/unspliced XBP1 mRNA expression as measured by RT-qPCR and normalized to GAPDH and YWHAZ. Mean ± SD and individual replicates are shown. ns = not significant, unpaired t test. See also Figure S4.

We concluded that the rs1053639 SNP status does not broadly impact the activation of the ER-stress response upstream of DDIT4. Consistent with the effect of thapsigargin treatment, a more pronounced reduction in phospho-eIF4EBP1 was observed for TT homozygous cells, also in response to nutlin treatment (Figures S4A and S4D).

Next, we investigated whether the differences in the mTOR pathway led to changes in global translation, proliferation, or autophagy of the HCT116 edited clones. The global translation was monitored either in mock culture conditions or in response to thapsigargin treatment by the measurement of newly synthesized peptides using either a commercially available global protein synthesis kit (Figure 6A) or using a quantitative analysis of the TE measured as the relative abundance of polysomes compared to the 80S from polysomal profiling (Figure S6A) or a puromycilation assay (Figures S6B and S6C).

Figure 6.

The DDIT4 rs1053639 genotype does not significantly impact global translation and cell proliferation in response to thapsigargin

(A) Global translation was measured using a Click-it reaction assay after incubation of the cells with modified methionine for 30 min. TT and AA HCT116 clones in mock conditions or upon thapsigargin treatment (100 nM for 4 h) were tested. Mean ± SD and individual replicates are shown. ns = not significant, ∗∗p < 0.01; ∗∗∗∗p < 0.0001, unpaired t test.

(B) The proliferation of HCT116 clones was followed every 24 h by the Operetta HCS system for 72 h. Relative cell number curves were plotted as the mean of several replicates ± standard deviation (SD). ns, not significant; ordinary two-way ANOVA.

(C) (Left) Western blot analysis of autophagy markers LC3-II and p62 (SQSTM1) upon chloroquine treatment (25 μM indicated with + or 75 μM indicated with ++) or vehicle for 6 h. GAPDH was used as a loading control. (Right) Densitometric analysis of LC3-II and p62 protein levels normalized to GAPDH levels. Data from at least three biological replicates are shown. Mean ± SD are shown. ∗ p value <0.05, ∗∗p value < 0.01, ns, not significant; ordinary one-way ANOVA. See also Figure S6.

In all cases, there were no significant differences in global translation between TT and AA rs1053639 clones, although the latter showed a trend toward a stronger reduction under perturbed conditions. The relative proliferation of the clones was similar in the mock condition (Figure 6B). TT cells exhibited higher autophagic flux when treated with chloroquine, as indicated by the relative levels of LC3-II and p62 proteins (Figure 6C). We concluded that the rs1053639 tranSNP genotype did not show an overt difference in global translation or cell proliferation but did show a difference in autophagy that is consistent with both mTORC1-dependent and -independent DDIT4 functions.

The rs1053639 tranSNP AA genotype shows a competitive growth advantage in prolonged culture conditions and in zebrafish embryos

Given the contribution of DDIT4 in the negative modulation of a critical and intricate metabolic hub such as mTOR,⁵¹ we tried to tease out possible differences in cellular phenotypes due to the rs1053639 genotypes through co-culture experiments where cells could experience fluctuations in nutrient availability in the media. To visualize and count the contribution of each genotype to the total co-culture population, derivative GFP- or RFP-positive clones were obtained with high clonal purity. Cells were seeded in 6-well plates, and relative cell number was measured either at Operetta high-content screening (HCS) or by fluorescence-activated cell sorting (FACS), examining the population every four days, up to fifteen days. Results were expressed as relative fitness, as recently described⁵² (Figure 7A). Interestingly, while the rs1053639 TT cells showed higher fitness at early times, AA cells progressively gained a proliferative advantage. This result was not influenced by the presence or the type of fluorescent marker (Figure S7A). Notably, the depletion of RBMX both in TT and AA cells three days prior to the co-culturing reversed the proliferative advantage phenotype of the AA cells (Figures 7B and S7B). To further explore the significant difference related to rs1053639, cells were sorted seven days after co-culture, and the two rs1053639 genotypes were examined separately. No significant difference in cell cycle distribution was observed (Figure 7C). Finally, the co-culture experiment was repeated in the presence of the Q-VD-OPh pan-caspase inhibitor.⁵³ The progressive advantage of the AA genotype was confirmed, suggesting that apoptosis was not involved in this phenotype (Figure 7D). Doxorubicin (3 μM) was included as a control to induce apoptosis^54,55 (Figures S7C and S7D).

Figure 7.

DDIT4 rs1053639 impacts cellular dynamics and proliferation under stress and competitive conditions in vitro and in zebrafish larvae xenotransplants

(A) DDIT4 clones harboring the TT genotype were transduced with lentiviral particles to express EGFP constitutively. After transduction, cells were treated with 5 μg/mL blasticidin for at least 15 days to reach at least 95% of GFP+ cell populations. DDIT4 clones harboring the AA genotype (not fluorescent) were mixed ∼1:1 with EGFP-expressing TT genotype cells and co-cultured for 15 days. The relative growth was assessed by flow cytometry, analyzing 10K cells. Cellular fitness of the GFP+ clones relative to GFP- competitor clones was calculated as follows: pGFP+(t) = NGFP+(t)/(NGFP+(t) + N GFP−(t)), W = ln ((pGFP+(ti)/pGFP−(ti))/(pGFP+(t0)/pGFP−(t0)), adapting a described method.⁵² The error bars represent the mean ± SD of at least eight independent experiments (four TT GFP+, four AA GFP+). ∗p value < 0.05; ∗∗p value < 0.01; ordinary two-way ANOVA.

(B) Relative fitness of rs1053639 AA homozygous cells (GFP labeled) in co-culture experiments with the TT homozygous counterpart. Both TT and AA-GFP cells were transfected to deplete RBMX (siRBMX) (or control, siNT) 3 days before starting the co-culture and silenced again at days 3 and 7 during the co-culture experiment. The relative fitness of GFP-positive cells is shown. Mean ± SD are shown. ∗p value < 0.05; ∗∗p value < 0.01; ∗∗∗p value < 0.001; ordinary two-way ANOVA.

(C) Cell cycle analysis by flow cytometry (FACS) for single clones before co-culture (left) and sorted clones seven days after co-culture (right).

(D) TT and AA clones were co-cultured as described in (A) up to 7 days, in the presence of the pan-caspase inhibitor Q-VD-OPh (10 μM) or vehicle control, and analyzed by FACS at days 0, 3, and 7. As a control, co-cultured cells were treated with doxorubicin (3 μM) ± Q-VD-OPh (10 μM) for 3 days. The relative fitness of GFP-positive cells across time points and conditions is shown. Mean ± SD are shown. ∗p value < 0.05; two-tailed unpaired t test.

(E) DDIT4 rs1053639 homozygous clones, transduced with lentiviral particles to constitutively express either EGFP or MiRFP, were xenotransplanted into zebrafish larvae two days post-fertilization. Tumor growth was monitored over a three-day period, with a minimum of 30 larvae per genotype condition. Representative confocal microscopy images were captured on days 1 and 3 post-transplantation, with TT clones in green and AA in purple. DDIT4 clones expressing GFP (TT clones) were mixed ∼1:1 with RFP-expressing clones (AA genotype). They were initially co-cultured in vitro for four days, followed by xenotransplantation into ∼30 zebrafish larvae for an additional three days of in vivo co-growth before confocal imaging. Scale bars, 200 μm.

(F) Bar graph (mean ± SD) illustrating the percentage of tumor growth relative to day 1 of transplantation, with each data point representing an individual larva. AA clones harbor significantly higher fitness. ∗∗p value < 0.01, unpaired t test.

(G) TT and AA clones were separately xenotransplanted into zebrafish larvae, which were followed for three days and imaged by confocal microscopy. Scale bars, 200 μm.

(H) Quantitative analysis of tumor growth indicates a significantly faster growth rate for AA than TT cells in vivo. Mean ± SD are shown. Each data dot represents one larva. ∗∗p value < 0.01, paired t test. See also Figure S7.

The proliferation experiments in co-culture suggested to us that the rs1053639 tranSNP genotype could impact cell growth in challenging environments where nutrient concentrations vary and different cell types can compete. To explore this hypothesis and also gain some information on the potential aggressiveness of our edited cancer cell lines, we developed xenograft experiments in Casper zebrafish embryos,⁵⁶ following a published protocol adjusted for our experiments.⁵⁷ An equal number of cells of the two different genotypes labeled either with GFP or RFP, as described earlier, were injected into the swim bladder of two-day post-fertilization embryos. Cells were then imaged the following day and again 48 h later. Results, plotted as a percentage of tumor cells’ growth comparing 24 to 72 h post-injection, showed that, when cells of the two genotypes were pre-co-cultured in vitro for three days prior to injection in the embryos, the growth advantage of the AA genotype was again apparent (Figures 7E, 7F, and S7E).

Furthermore, we evaluated the proliferative potential of the TT and AA cells separately in the in vivo setting. Notably, the AA cells grew more compared to the TT ones as well (Figures 7G, 7H, and S7F). Collectively, co-culture experiments indicate that rs1053639 AA cells display a competitive advantage over TT cells both in vivo and in vitro, although the underlying mechanisms remain to be clarified, and it does not seem to fit under a classical cell competition response.⁵⁸

TT and AA clones exhibit different drug susceptibility and can have prognostic significance in cancer patients

The phenotypic differences related to the rs1053639 genotype, a relatively common SNP in the human population, led us to explore whether this tranSNP status can impact clinically relevant variables. We explored whether the SNP genotype could determine allele-specific susceptibility to drugs. We found that TT clones are significantly more resistant to niclosamide, nutlin, and thapsigargin than the AA clones (Figures 8A–8C).

Figure 8.

TT and AA clones exhibit different drug susceptibility and can have prognostic significance in cancer patients

(A–C) HCT116 clones were treated with a wide range of concentrations of (A) niclosamide, (B) nutlin, (C) thapsigargin, or vehicle for 48 h and imaged using the ImageXpress Micro Confocal system (molecular devices). Live cells were counted based on the standard Hoechst and propidium iodide staining and normalized to the vehicle control. Bar graphs indicating the IC₅₀ (half maximal inhibitory concentration) values from at least three independent experiments (mean ± SD and individual points; ∗p value < 0.05; ∗∗p value < 0.01; unpaired t test), as well as the corresponding dose-response curves, were plotted.

(D) Kaplan-Meier analysis of disease-free interval (DFI) according to rs1053639 genotype in TCGA patients. Survival curves are shown for colon adenocarcinoma (COAD).

Based on the results obtained with the three genotypes, particularly the reduced control of the mTOR pathway of the AA homozygous cells, we decided to focus on a recessive model, comparing AA homozygous patients with AT plus TT to interrogate clinical variables in cancer patients. Using TCGA data, we observed a worse prognosis for AA patients in colorectal cancers (Figure 8D). Further studies are needed to establish the value of rs1053639 status as a prognostic marker.

Discussion

Variation in allelic imbalance in polysomal RNA as a tool to identify functionally relevant SNPs

Our approach to identifying tranSNPs is based on computing and comparing relative changes in allele counts across mRNA fractions, building upon the methodology previously introduced,⁹ and the underlying assumption is that the interaction of an mRNA with polysomes is directly proportional to the amount of protein being produced.^59,60,61 By fully leveraging biological replicates, our enhanced method eliminates the need for a predefined threshold to detect allele-specific expression. Instead, it evaluates differential AFs between RNA-seq data from the two distinct cellular compartments using pairwise statistics, thereby improving the robustness of the analysis. Despite the limitations imposed by low coverage, which introduced noise in AF estimates and constrained our ability to detect and characterize SNPs with significant allelic imbalances, our findings support the presence of signal in over 11% of analyzable genes. Each of these genes harbors one or multiple imbalanced SNPs, collectively accounting for approximately 7% of all analyzable SNPs, aligning with previous reports.^8,9

Mechanism of rs1053639 impact on DDIT4 protein expression

For rs1053639 A/T we observed a significant reduction in the alternative allele (A) in polysome-associated mRNAs in the mock condition. Using edited HCT116 clones homozygous for the alternative and reference alleles, respectively, we observed a marked difference in DDIT4 protein expression associated with the different SNP status. In our analysis pipeline, we also used predictions of trans-factors binding to mRNAs. In this respect, rs1053639 is predicted to affect an RBMX binding motif. Given that RBMX is reported as mainly nuclear,³⁷ an observation confirmed also for HCT116 cells (Figure 4C), we measured DDIT4 mRNA localization by comparing the TT and AA homozygous clones. In general, RNA-binding proteins binding 3′-UTR sequences can contribute to subcellular mRNA localization and mRNA fate.⁶² Results indicated higher cytoplasmic localization for the T allele. The transient silencing of RBMX slightly increased the nuclear localization of the T allele and significantly increased the number of perinuclear granules of DDIT4 mRNA. This effect was associated with a significant reduction of DDIT4 protein expression with the TT genotype, making it comparable to the levels of the AA clones. In cells of the AA genotype, the number of DDIT4 mRNA perinuclear granules was lower than in TT cells in the mock condition. Interestingly, RBMX depletion did not significantly change the nuclear localization of the DDIT4 mRNA A allele, the number of perinuclear granules per cell, or DDIT4 protein levels. The observed increase of perinuclear DDIT4 mRNA granules in the AA clones, irrespective of RBMX expression, and in the TT clones only upon RBMX depletion, suggests an impact of the tranSNP, together with this trans factor, on mRNA export and translation. Indeed, mRNA granules near the nuclear membrane have been considered intermediate sites during mRNA export.^63,64,65 Moreover, recent reports highlighted the relevance of mRNA localization in cytoplasmic compartments for controlling its protein production.⁶²

Collectively, our validation experiments confirmed the observation of allelic imbalance in polysomes and provided evidence that rs1053639 can impact DDIT4 protein expression through a mechanism that comprises mRNA localization, export, modification, and TE.

Implications of rs1053639 allelic status on DDIT4-mediated functions

DDIT4 is an established important negative modulator of the mTOR pathway, acting at the level of mTORC1. Consistently, its gene expression is modulated in response to ER stress, with ATF4 as an important transcription factor. It is also inducible by hypoxia and is considered a p53 target gene, although recent data proposed that the contribution of p53 is indirect, via the transcription factor RFX7.²⁰ In general, mTOR inhibition is growing in relevance in p53-dependent tumor suppression.⁶⁶ The negative modulation of mTORC1 by DDIT4 is rather established, although reports differ on the exact mechanism^46,51 and its relative relevance in different cell types.⁶⁷ DDIT4, as the extended name testifies, is considered a player in the DNA damage response, although also, in this regard, mechanistic details are missing. Overall, DDIT4 function in cancer is considered to be tumor suppressive. For example, a relatively recent screen indicated DDIT4 in a short list of p53 pathway targets with strong tumor-suppressive potential. In agreement with the centrality of the mTOR pathway for general cell metabolism, DDIT4 has emerged as an important gene in contexts other than cancer, such as liver disease, neurodegeneration, pulmonary injury, and retinopathy.^68,69,70,71 In fact, post-transcriptional regulation of DDIT4 expression has been the subject of different studies that can bear relevance for understanding the impact of rs1053639. For example, two studies pointed to the important role of an antisense transcript (DDIT4-AS1) corresponding to a large portion of DDIT4 3′ UTR and of the 3′ coding sequence, although with different conclusions on the role of the antisense transcript on the sense one.^29,30 rs1053639 is included in the antisense. We checked if the allelic status could somehow impact the amount of spliced DDIT4-AS1. In HCT116, the antisense was expressed at a very low level, and there were no appreciable differences among the edited clones (Figure 2), indicating that it does not contribute to the difference in DDIT4 protein expression in our context.

In HCT116 cells, our experiments demonstrate that rs1053639 SNP status significantly impacts DDIT4 protein levels. The lower expression of DDIT4 in AA homozygous clones is even more pronounced when the gene transcription is induced by ER stress or by nutlin. Hence, the tranSNP we identified is predicted to impact p53-dependent and, more in general, stress-dependent modulation of DDIT4 with potentially appreciable consequences on the negative modulation of mTOR, impacting translation control, proliferation, autophagy, and cell survival. Given the complex regulation of the mTOR pathway,⁷² the impact of DDIT4 is expected to be at least in part influenced by tissue type. In HCT116 cells, higher DDIT4 protein levels in the TT homozygous clones are associated with stronger repression of mTORC1 markers. Despite this, TT and AA clones showed similar proliferation in culture plates. However, in vitro co-culture experiments showed a higher proliferation of AA cells. The two separated populations did not show evidence of cell cycle checkpoint activation or apoptosis induction. Silencing of RBMX, which negatively impacts DDIT4 protein levels in TT but not AA homozygous cells, led to a reversal of that proliferation advantage. The results of the xenograft transplantation experiments in zebrafish embryos were also consistent with higher proliferation of HCT116 cells homozygous for the rs1053639 A allele.

In general terms, DDIT4 is considered a tumor suppressor gene. However, gene expression data from TCGA indicate that in several cancer types, DDIT4 expression is higher in normal tissues than in cancer (Table S8). Consistent with these conflicting data, high DDIT4 expression has been associated both with better and worse prognosis. Although exploratory, we found consistent prognostic predictions when TCGA patients’ cohorts were stratified by DDIT4 mRNA levels or by rs1053639 genotype in colon adenocarcinoma (COAD). We also established that the differences in DDIT4 protein levels associated with the rs1053639 genotype can be assessed by an IHC protocol (Figure S8).

In summary, we have developed a robust pipeline to catalog genetic variants associated with allele-specific changes in mRNA translation, leveraging biallelic mRNAs and polysomal profiling as a proxy for TE. Our approach provides a framework for identifying genetic drivers of inter-individual variability in mRNA translation, shedding light on mechanisms that can influence cancer progression.

Limitations of the study

The limited coverage of our RNA-seq dataset resulted in a relatively small number of heterozygous SNPs available for analysis. This was further constrained by our conservative approach, which requires the heterozygous state of an mRNA variant, a minimum of twenty reads, and consistency among four biological replicates of total and polysomal RNA sequencing, which are then compared to nominate a tranSNP candidate. This strategy reduces the number of genetic variants that can be analyzed with a chosen cell line but enhances the likelihood of detecting true signals. Validation experiments using orthogonal approaches and reporter assays were limited to four selected candidates of which three showed consistent results across methods. While the heterozygous state is a requirement to start our analysis, the phenotypic impact of a variant is expected to be most pronounced in homozygous states. Therefore, confirming the causal role of a tranSNP on cell phenotypes requires the development of genomically edited, isogenic cell clones differing for SNP status, a time-intensive process that, in our study, was performed only in HCT116 cells.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Alberto Inga (alberto.inga@unitn.it).

Materials availability

Plasmid and edited cell lines generated in this study are available from the lead contact.

Data and code availability

• •The previously published RNA-seq dataset of HCT116 total and polysomal RNA re-analyzed in this study is publicly available in the Gene Expression Omnibus (GEO) under accession no. GSE95024. The targeted resequencing data generated in this study have been deposited in the NCBI BioProject database under accession no. PRJNA1438951. TCGA SNP 6.0 Birdseed genotype calls used in this study are available through the Genomic Data Commons (GDC) under controlled access and require authorization through dbGaP. TCGA clinical information was obtained from Liu et al. (2018),⁷³ Table S1. Uncropped western blot images are provided in the supplemental information and have also been deposited in Mendeley Data (https://doi.org/10.17632/9j4wjn3xxb.1). Accession numbers, DOIs, and links to the relevant datasets and resources are provided in the key resources table.
• •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

We thank Dr. Ainan Geng and Prof. Hashim M. Al-Hashimi, Department of Biochemistry and Molecular Biophysics, Columbia University, NYC, USA, for the DDIT4 3′ UTR RNA structural predictions. We thank Dr. Mounira Chalabi, Ribosome, Translation and Cancer Team Université Claude Bernard Lyon 1, France, for technical support to M.H.H. We thank Alice Casata, Irene Cagol, Massimo Andreis, Micol Bottamedi, Federica Sontacchi, and Christian Cestaro for technical support during their bachelor’s internships. We thank Dr. Ezequiel Mariano Rivero and Prof. Fatima Gebauer, Gene Regulation, Stem Cells, and Cancer Program, Centre for Genomic Regulation (CRG), Barcelona, Spain, for support with the RIP protocol. We thank Dr. Gabriella Viero, Institute of Biophysics, CNR Unit at Trento, Italy, for support with sucrose gradient fractionation and data analysis. This work was supported by Fondazione AIRC under the grant IG #25849, “Mining common genetic variants impacting on allele-specific translation and cancer risk,” to A.I.; the European Union under NextGenerationEU, PNRR M4C2 INV 1.1, PRIN 2022 PNRR Prot. n. P2022A9J9L to L.L.F. L.A. was supported by an AIRC short-term fellowship. M.M. and F.L. were supported by AIRC IG 2021 no. 25704. L.A. and M.H.H. were also supported by Short-term scientific mission fellowships by the “Translation Control in Cancer” European Network, Translacore Cost Initiative (CA21154). This work was also partially supported by the initiative “Dipartimenti di Eccellenza 2023–2027 (Legge 232/2016)” funded by the MUR, and by FESR 2023,“Sostegno alle Infrastrutture di Ricerca.” Department CIBIO Core Facilities are supported by the European Regional Development Fund (ERDF) 2014–2020 and 2021–2027.

Author contributions

M.H.H. and L.A. performed most of the experiments, prepared figures, and contributed to the first draft of the manuscript; T.V., S.V., and D.D. performed the RNA-sequencing analysis and developed the pipeline to identify tranSNPs; F.L. performed with M.H.H. and L.A. the experiments in zebrafish and prepared the first draft of the relative methods and results; A. Mattivi and V.V. supported the DDIT4 gene-editing approaches; G.P.G. performed with L.A. the smiFISH experiments; A. Matte supported the design and execution of the cell-sorting experiments; P.G. and M.P. supported all HTS-based experiments; C.V., V.D.S., and R.B. designed the targeted resequencing and performed RNA-sequencing experiments; V.M. supervised M.H.H. in planning and performing translation assays; E.C. contributed to the design and the evaluation of the results of smiFISH experiments; S.F. and G.B. performed IHC experiments to establish their suitability to distinguish DDIT4 protein levels by rs1053639 genotype; S.Z. supervised L.A. in designing and performing in vitro RBMX-RNA binding and GLORI at DDIT4 3′ UTR; M.M. supervised and contributed to the design and interpretation of the experiments in zebrafish; L.L.F. supervised the editing approaches, contributed to the design, visualization, and interpretation of the co-culture experiments, contributed to drafting and revising the manuscript; A.R. supervised the study, in particular RNA-seq data analyses, preparation of tables, drafting of the manuscript; A.I. supervised the study, secured the main findings, and drafted the manuscript.

Declaration of interests

The Authors declare no competing interests.

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

During the preparation of this work, the authors used Grammarly and ChatGPT to assist with grammar checks and improvements in writing clarity. The authors reviewed and edited all generated text and take full responsibility for the final content of the manuscript.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
ATF4	SantaCruz	sc-390063; RRID: AB_2810998
beta-tubulin	SantaCruz	sc-5274: RRID: AB_2288090
DDIT4	Proteintech	67059-1-Ig; RRID: AB_2882369
DDIT4 IHC	Proteintech	10638-1-AP; RRID: AB_2245711
GAPDH	SantaCruz	sc-32233; RRID: AB_627679
Lamin-A/C	Abcam	ab40567; RRID: AB_775967
LC3	Sigma	L8918; RRID: AB_1079382
p62	Cell Signaling	5114; RRID: AB_10624872
p-4EBP1 Thr 37/46	Cell Signaling	236B4; RRID: AB_560835
p53 (DO-1)	SantaCruz	sc-126; RRID: AB_628082
p-RPS6 Ser 240/244	Genetex	GTX133942; RRID: AB_2887145
p-S6K Thr 421/Ser 424	Genetex	GTX635621; RRID: AB_2888544
Puromycin	Millipore	MABE343: RRID: AB_2566826
RBMX	Cell Signaling	D7C2V Cat#14794; RRID: AB_2798614
RPL21	SantaCruz	sc-393663
RPS3	SantaCruz	sc-376008; RRID: AB_10991105
tot-4EBP1	Cell Signaling	9452; RRID: AB_331692
tot-RPS6	Genetex	GTX113542; RRID: AB_2037923
tot-S6K	Genetex	GTX107562; RRID: AB_1951779
anti-Normal Rabbit IgG antibody	Cell Signaling Technology	Cat #2729; RRID: AB_1031062
Bacterial and virus strains
E. coli strain	Laboratory stock	DH5alpha
E. coli strain	Laboratory stock	XL1blue
Chemicals, peptides, and recombinant proteins
TRIzol	Thermo Fisher Scientific (Invitrogen)	Cat# 15596026
Nutlin	Cayman chemicals	Cat#10004372
Thapsigargin	Prodotti Gianni	Cat# ab120286
Actinomycin D	Thermo Fisher Scientific (Gibco)	Cat# 11805017
Cycloheximide	Cayman Chemicals	Cat# 14126
Xba I	New England Biolabs	Cat# R0145S
EcoR I	New England Biolabs	Cat# R0101S
T4 DNA ligase	New England Biolabs	Cat# M0202S
AmpliTaq Gold 360 Master Mix	Thermo Fisher Scientific	Cat# 4398833
Phusion High-Fidelity DNA Polymerase	Thermo Fisher Scientific	Cat# F530L
cOmplete™, Mini Protease Inhibitor Cocktail	Merck (Roche)	Cat# 11836153001
Protein A Magnetic Beads	Thermo Fisher Scientific (Dynabeads, Invitrogen)	Cat# 88845
full-length recombinant RBMX protein	Origene	Cat# TP710403
Critical commercial assays
Lipofectamine LTX with Plus reagent	ThermoFisher Scientific	Cat# 15338100
FuGene4K Transfection reagent	Promega	Cat# E5911
SE Cell Line 4D-Nucleofector™ X Kit S	Lonza	Cat# V4XC-1032
INTERFERin	Polyplus	Cat# 101000028
NU7441 (KU57788)	DBA Italia (MedChemExpress)	Cat# HY-11006
Agilent RNA 6000 Nano kit	Agilent	Cat#5067-1511
RevertAid RT Kit	Thermo Fisher Scientific	Cat# K1691
Dual Luciferase Reporter assay system	Promega	Cat# E1910
Pierce™ BCA Protein Assay Kits	Thermo Fisher Scientific (Pierce)	Cat# 23227
Click-iT™ Plus OPP Alexa Fluor™ 488 Protein Synthesis Assay Kit	Thermo Fisher Scientific (Invitrogen)	Cat# C10456
qPCRBIO SyGreen Mix	PCR Biosystems (Resnova)	Cat# PB20/11-20
Deposited data
RNA-seq of HCT116 (total and polysomal RNA)	Rizzotto et al.13	GSE95024
Targeted resequencing	This study	BioProject PRJNA1438951
TCGA SNP 6.0 array calls	Genomic Data Commons	https://gdc.cancer.gov
TCGA clinical information	Liu et al., 2018	https://doi.org/10.1016/j.cell.2018.02.052 (Table S1)
RBPDB	http://rbpdb.ccbr.utoronto.ca/	https://doi.org/10.1093/nar/gkq1069
Uncropped Western Blot Images	Mendeley Data	https://doi.org/10.17632/9j4wjn3xxb.1
Experimental models: Cell lines
HCT116	ICLC, certified by BMR genomics, Padua, Italy	N/A
A549	ICLC	N/A
Oligonucleotides
DDIT4 Short Cloning_F	Eurofins Genomics	aattcGACGACTGAACTTTTGGGG TGGAGACTAGAGGCAGGAGCTG AGGGACTGATTCCWGTGGTTGG AAAACTGAGGCAGCCACCTAAG GTGGAGGTGGGGGAATAGTGT TTCca
DDIT4 Short Cloning_R	Eurofins Genomics	tatgGAAACACTATTCCCCCACCT CCACCTTAGGTGGCTGCCTCAG TTTTCCAACCACWGGAATCAGT CCCTCAGCTCCTGCCTCTAGTC TCCACCCCAAAAGTTCAGTCG TCg
DDIT4 Long Cloning_F	Eurofins Genomics	ACTTCAACCTGAGGGGGCCGACA
DDIT4 Long Cloning_R	Eurofins Genomics	AATCTAGAAGGTCAGTAGTGATGCTCCG
DDX17 Cloning_F	Eurofins Genomics	AATCTAGAATTCAGGGATGCCC TACTCTCTCG
DDX17 Cloning_R	Eurofins Genomics	AATCTAGAGAGCACTGATCCCAAGCA
PERP Cloning_F	Eurofins Genomics	AATGAATTCCATGGTCCAAACCTGTTGCC
PERP Cloning_R	Eurofins Genomics	ATAATCTAGATTCTCCTGTG GCATGTTTTATATC
RELA Cloning_F	Eurofins Genomics	AATCTAGAATTCGGGA GCTGGGGAAACTCAAA
RELA Cloning_R	Eurofins Genomics	AATCTAGAAGTCCCAGACCAAACCCC
UBE2I Cloning_F	Eurofins Genomics	AATCTAGAATTCGC AGCGACCTTGTGGCATCGTCAA
UBE2I Cloning_R	Eurofins Genomics	GGGTCTGAGGCTCAAAAGCA
18S_F	Eurofins Genomics	GTAACCCGTTGAACCCCATT
18S_R	Eurofins Genomics	CCATCCAATCGGTAGTAGCG
7SL_F	Eurofins Genomics	ATCGGGTGTCCGCACTAAGTT
7SL_R	Eurofins Genomics	CAGCACGGGAGTTTTGACCT
B2M_F	Eurofins Genomics	AGGCTATCCAGCGTACTCCA
B2M_R	Eurofins Genomics	ATGGATGAAACCCAGACACA
CDKN1A_F	Eurofins Genomics	CTGGAGACTCTCAGGGTCGAAA
CDKN1A_R	Eurofins Genomics	GATTAGGGCTTCCTCTTGGAGAA
CHOP_F	Eurofins Genomics	AGAACCAGGAAACGGAAACAGA
CHOP_R	Eurofins Genomics	TCTCCTTCATGCGCTGCTTT
CSDE1_F	Eurofins Genomics	CACGGGAATGGATGGGTCTG
CSDE1_R	Eurofins Genomics	ACGGCCTCTTTACCCAGAGA
DDIT4 CDS_F	Eurofins Genomics	CTAGCTGCGGCTTCTACGC
DDIT4 CDS_R	Eurofins Genomics	CCAAAGGCTAGGCATGGTGA
DDIT4 3UTR_F	Eurofins Genomics	ACTTCAACCTGAGGGGGCCGACA
DDIT4 3UTR_R	Eurofins Genomics	CTAAGCCTTTGTTTCATGCT
DDIT4 RIP_F	Eurofins Genomics	TTGGAAAACTGAGGCAGCCA
DDIT4 RIP_R	Eurofins Genomics	CTAAGCCTTTGTTTCATGCT
DDIT4 Tide_F	Eurofins Genomics	CCAAGATCCAGGGGCTGTTT
DDIT4 Tide_R	Eurofins Genomics	GTGCCTCTTCCTTCACCCTG
DDIT4-AS1_F	Eurofins Genomics	CTACAACAGGTCATAACAAAAAT
DDIT4-AS1_R	Eurofins Genomics	ATGAAACAAAGGCTTAGGG
GAPDH_F	Eurofins Genomics	CCACTCCTCCACCTTTGAC
GAPDH_R	Eurofins Genomics	ACCCTGTTGCTGTAGCCA
NEAT1_F	Eurofins Genomics	GGGGCGGATCGGTGTTGCTT
NEAT1_R	Eurofins Genomics	CCCGGTTCCATCTGCTCGCC
RPS26_F	Eurofins Genomics	CCTAAGGATTCTCCCGGTGT
RPS26_R	Eurofins Genomics	GCACGACCATTGTTCCTTCT
SESN2_F	Eurofins Genomics	CTCCTCCTTCGTGTTTGGCT
SESN2_R	Eurofins Genomics	CTCAAAGCCCCCAGAGTTGT
Total XBP1_F	Eurofins Genomics	TGGCCGGGTCTGCTGAGTCCG
Total XBP1_R	Eurofins Genomics	ATCCATGGGGAGATGTTCTGG
Unspliced XBP1_F	Eurofins Genomics	CAGCACTCAGACTACGTGCA
Unspliced XBP1_R	Eurofins Genomics	ATCCATGGGGAGATGTTCTGG
Spliced XBP1_F	Eurofins Genomics	CTGAGTCCGAATCAGGTGCAG
Spliced XBP1_R	Eurofins Genomics	ATCCATGGGGAGATGTTCTGG
U6 snRNA_F	Eurofins Genomics	CTCGCTTCGGCAGCACATATACT
U6 snRNA_R	Eurofins Genomics	ACGCTTCACGAATTTGCGTGTC
YWHAZ_F	Eurofins Genomics	CAACACATCCTATCAGACTGGG
YWHAZ_R	Eurofins Genomics	AATGTATCAAGTTCAGCAATGGC
DDIT4_F	Eurofins Genomics	TCGTCGGCAGCGTCAG ATGTGTATAAGAGAC AGTCCGAGTCATCAAGAAGAAGCTGTA
DDIT4_R	Eurofins Genomics	GTCTCGTGGGCTCG GAGATGTGTATAAGAGACAGTTGTTT CATGCTACAGTACTGAGGG
DDX17_F	Eurofins Genomics	TCGTCGGCAGCGTCA GATGTGTATAAGAGACAGTTATAAGTG ACAGACATGGAGGCAG
DDX17_R	Eurofins Genomics	GTCTCGTGGGCTCGGA GATGTGTATAAGAGACAGCAGCTCTT GAAAGCATCTGTTTGAG
PERP_F	Eurofins Genomics	TCGTCGGCAGCGTCAGAT GTGTATAAGAGACAGTAGACT GGATTGAAAGATGGACTG
PERP_R	Eurofins Genomics	GTCTCGTGGGCTCG GAGATGTGTATAAGAGACA GAGCAAGTCATTTATTCTCCTGAGG
RELA_F	Eurofins Genomics	TCGTCGGCAGCGTCAGA TGTGTATAAGAGACAGTTTGAGCCCACA AAGCCTTATCAAG
RELA_R	Eurofins Genomics	GTCTCGTGGGCTCGGAGA TGTGTATAAGAGACAGCTTGAGAAGAGGG AGAGCAAGGAAG
siRBMX 13.1	IDT	AUCAAGAGGAUAUAGCGAUAGAGAT
siRBMX 13.2	IDT	AUGUUCUCAGUACGAAAAACCUGAA
siRBMX #3	IDT	CAAAUCAAGAGGAUAUAGCGAUAUU
DDIT4 FISH probe 1	Eurofins genomics	GAAGCCGCAGCTAGCGC GGTCAGCGAGAACTGTTACACTCGGACCT CGTCGACATGCATT
DDIT4 FISH probe 2	Eurofins genomics	CCAGGTAAGCCGTGTCTT CCTCCGGCCCGATTACACTCGGACCTCGT CGACATGCATT
DDIT4 FISH probe 3	Eurofins genomics	AGGGCGTTTGCTGATGAA CTCAGAGTGCCGTTACACTCGGACCTCGTC GACATGCATT
FAM-labelled fluorescent RNA probe	Metabion	GAGGGACUGAUUCCWGUGGUUGGAA-3’FAM
DDIT4 Alt-R® CRISPR-Cas9 crRNA	IDT	GTGGTTGGAAAACTGAGGCA
DDIT4 Ultramer® DNA Donor Oligonucleotide	IDT	TTTGGGGTGGAGACTAGA GGCAGGAGCTGAGGGACTGATTCCWGTG GTTGGAAAACTGAGGCAGCCACCTAAG GTGGAGGTGGGG
Software and algorithms
ASEQ	https://bitbucket.org/CibioBCG/aseq/	https://doi.org/10.1186/s12920-015-0084-2
Ensembl rest API	https://rest.ensembl.org/	https://doi.org/10.1093/bioinformatics/btu613
SHAPEIT	https://mathgen.stats.ox.ac.uk/genetics_software/shapeit/shapeit.html	https://doi.org/10.1016/j.ajhg.2013.09.002
IMPUTE2	http://mathgen.stats.ox.ac.uk/impute/impute_v2.html	https://doi.org/10.1371/journal.pgen.1000529
Survival R package	https://cran.r-project.org/web/packages/survival/index.html	https://doi.org/10.1007/978-1-4757-3294-8
TESS	https://www.cbil.upenn.edu/tess	https://doi.org/10.1002/0471250953.bi0206s21
VariantAnnotation R package	https://bioconductor.org/packages/release/bioc/html/VariantAnnotation.html	https://doi.org/10.1093/bioinformatics/btu168
Trimmomatic v0.39	Bolger et al.74	https://doi.org/10.1093/bioinformatics/btu170
CutAdapt 4.1	Martin75	https://doi.org/10.14806/ej.17.1.200
STAR (≥v2.7.8a)	Dobin et al.76	https://doi.org/10.1093/bioinformatics/bts635
SAMtools (≥v1.0)	Li et al. 2009	https://doi.org/10.1093/bioinformatics/btp352
Picard MarkDuplicates (v2.17.8)	Broad Institute	https://broadinstitute.github.io/picard/
GATK (≥v3.8)	Broad Institute	https://gatk.broadinstitute.org/
BamUtil (clipOverlap, v1.0.15)	University of Michigan	https://genome.sph.umich.edu/wiki/BamUtil:clipOverlap
PacBAM (v1.6.0)	Valentini et al.77	https://doi.org/10.1186/s12864-019-5899-3
R (v4.x)	R Foundation for Statistical Computing	https://www.r-project.org/
limma (R package)	Ritchie et al., 2015	https://doi.org/10.1093/nar/gkv007
RBPDB	Cook et al.34	https://doi.org/10.1093/nar/gkq1069
CRISPOR	Concordet and Haeussler27	https://crispor.gi.ucsc.edu
Fiji	https://doi.org/10.1038/nmeth.2019	https://imagej.net/software/fiji/downloads
TIDE	Brinkman and van Steensel25	https://tide.nki.nl
Synthego ICE	https://doi.org/10.1089/crispr.2021.0113	https://ice.editco.bio/#/
Image Lab	Bio-Rad Laboratories, Inc	https://www.bio-rad.com/it-it/product/image-lab-software?ID=KRE6P5E8Z#fragment-6
Leica Application Suite X (LAS X)	Leica Microsystems	https://www.leica-microsystems.com/products/microscope-software/p/leica-las-x-ls/

Experimental model and study participant details

Cell lines and cell culturing

The human cell line HCT116 was purchased from the Biological Resource Center ICLC Cell bank. Cells were cultured in standard Roswell Park Memorial Institute (RPMI) 1640 medium (GIBCO) supplemented with 10% fetal bovine serum (FBS), 100 units/ml penicillin, 100 mg/ml streptomycin antibiotic mix (GIBCO), and 2 mM Glutamine (GIBCO). Cells were periodically checked to control for mycoplasma contamination. The cells were maintained in a humidified atmosphere at 37°C and 5% CO₂.

Method details

Dataset and data preprocessing

The RNA-sequencing data of RNA extracted from polled polysomal and total cytoplasmic fractions of HCT116 cells cultured either in mock condition or treated with the MDM2 inhibitor Nutlin was produced in our earlier study.¹³ The raw data is deposited in GEO, under the GSE95024 accession number. Single-end RNA-seq data was processed to obtain from row FASTQ files properly processed BAM files. The trimming step, essential to remove sequence adapters, was performed with Trimmomatic.⁷⁴ The data were aligned with STAR⁷⁶ to the reference genome GRCh38 to obtain a BAM file for each sample. To improve the characterization of SNP positions, a step of deduplication to remove PCR artifacts was performed by employing PICARD module MarkDuplicates (broadinstitute.github.io/picard/). Then, GATK⁷⁸ was employed to split, realign, and recalibrate the data in order to improve the quality of the alignment. The pileup of SNPs positions was performed with ASEQ⁷⁹ using parameters mode=0, mbq=20, and mrq=20 in order to obtain allelic fraction (AF) calculations. Only SNPs in exonic and UTR regions extracted from dbSNP were considered. We emphasize that, as expected, the use of single-end RNA-seq data led to high duplication rates (on average 70%), limiting the number of SNPs we could characterize and analyze. Nevertheless, this step, along with subsequent quality control processes, was essential to ensuring high-quality AF estimations.

Differential ASE analysis

The allelic fraction at SNPs’ positions was used to compute differential allele-specific expression (ASE) events between total and polysomal RNA fractions. In order to capture those events, we started by identifying all SNPs that were heterozygous in at least one biological replicate of the two fractions. More specifically, to define the analyzable SNPs, for the data obtained both in mock and Nutlin-treatment conditions, we searched for SNPs having a value of AF between 0.2 and 0.8 (i.e., heterozygous genotype) in at least one replicate of one of the two fractions (polysomal or total) and that were confidently observable, i.e., with a depth of coverage of at least 20x of the SNP site in the each fraction, condition, and biological replicate. A total of 1060 SNP/condition pairs were identified. Of these, only those with AF measures available across all four biological replicates and both fractions were then considered for differential ASE analysis (N=582, Table S1).

To assess the presence of a differential ASE event at a SNP position we extended the methodology we developed.⁹ In particular, for each SNP we exploited the concordance of differential AF measures across pairs of polysomal and total RNA biological replicates by performing paired t-test statistics. Although none of the 41 SNP/condition pairs with nominal p-value < 0.05 retained significance after FDR correction, likely due to limited read coverage and increased noise in allelic fraction estimation, we observed that 11.8% of the analyzed genes harbored one or more SNPs in linkage disequilibrium, showing consistent allelic imbalance between total and polysomal RNA at the nominal significance level. This global enrichment was used to prioritize candidate variants for downstream analyses. Accordingly, SNPs with nominal p-values < 0.05 were defined as putative tranSNPs and retained for experimental validation as hypothesis-generating candidates rather than statistically definitive associations.

Expression analysis of DDIT4 in cancer and normal tissue samples

To test whether the gene containing the rs1053639 tranSNP exhibits differential expression in tumor versus normal tissue samples, we compared transcript expression values (normalized RPKM values) of the DDIT4 gene exploiting primary tumor and normal tissue samples from the TCGA COAD dataset. In detail, we downloaded the gene counts for all genes from the recount 3 database,⁸⁰ and RPKM values were computed using the function getRPKM from the recount 2 R package. Then, protein-coding transcripts were selected, and their RPKM values were normalized across samples using the R limma quantile normalization function. Then, focusing on the DDIT4 gene, the distributions of log2(RPKM+1) values across tumor and normal samples were compared using Wilcoxon’s test statistics, and p-values were finally corrected for multiple hypotheses using the FDR method.

Survival analysis using TCGA data

Genotype and survival data for TCGA patients were collected. Specifically, genotype information for rs1053639 tranSNPs was retrieved from,⁸¹ while survival data across different endpoints (Overall Survival, Disease-Specific Survival, Disease-Free Interval, and Progression-Free Interval) was retrieved from.⁷³ A recessive model (AA+AB vs. BB) was built, and for each endpoint, a Kaplan-Meier survival curve was calculated along with summary statistics obtained from a multivariate Cox proportional hazards regression model, where age at diagnosis was included as a covariate. For each endpoint, the analysis was performed across 23 different cancer types, and p-values were corrected for multiple hypothesis testing, retaining only associations with a p-value < 0.05 and an FDR < 0.25. Survival analyses were performed using the survival R package.

Identification of SNPs disrupting putative RBP consensus motifs

To investigate the RNA protein binding landscape at the rs1053639 tranSNPs locus, we retrieved 53 human RNA binding consensus motifs from the RBPDB database³⁴ and created a FASTA file containing a list of 60bp DNA sequences containing the reference and the alternative bases. Then, TESS software³⁵ was run to compute RBP consensus motifs scores. A set of log-likelihood-ratio-based scores were computed: the log-odds ratio of the match for the alternative and reference bases (altScore and refScore), and the maximum possible log-odds ratio for a match for a given RBP consensus motif (maxScore). The scores were filtered by keeping only the forward strand analysis. Results were retained only if the score for the reference or the alternative allele was at least 3 and greater than 0.5 when normalized by the maxScore.

Polysome profiling, RNA extraction, and Sanger sequencing

HCT116 cells were grown in 150 mm dishes to ∼80% confluence followed by 100 nM Thapsigargin or vehicle treatment for 4 hours. Cells were then incubated for 10 min with 50 μg /ml cycloheximide at 37°C to disrupt the protein synthesis process and halt translation elongation, thereby immobilizing the ribosomes on the mRNAs and preventing ribosome disassembly. Cells were rinsed with ice-cold PBS containing 50 μg /ml cycloheximide and then gently scraped using 600 μl of ice-cold lysis buffer (20 mM Tris-HCl pH 7.5, 100 mM KCl, 5 mM MgCl₂, 0.5 % Nonidet P-40, 0.2 U/μl RNasin® Ribonuclease Inhibitor [Promega], cOmplete™, Mini Protease Inhibitor Cocktail 1X [Roche], 100 μg /ml cycloheximide) and collected in an Eppendorf tube. The cells were then incubated on ice for 10 min and centrifuged for 10 min at 12,000 x g at 4°C to remove the nuclei. The cytoplasmic lysate was then separated by ultracentrifugation in a Beckman SW41 rotor at 40.000 rpm for 1 hour and 40 minutes at 4°C after being loaded onto a 15-50% linear sucrose gradient dissolved in salt buffer (100 mM NaCl, 20 mM Tris-HCl pH7.5, and 5 mM MgCl₂). The sucrose gradients were fractionated, and the absorbance was measured at 254 nm by the Teledyne Isco model 160 gradient analyzer equipped with a UA-6 UV/VIS detector. Based on density separation, polysomes were isolated from the fractions, and RNA was harvested from these polysomes with the TRIzol® reagent (Invitrogen™) following the manufacturer’s protocol. The RNA was then retro-transcribed, and the sequences containing the candidate TranSNPs were amplified by PCR using AmpliTaq Gold 360 Master Mix (ThermoFisher Scientific) and specific primers (Table S4). The PCR products were purified by the spinNAker GEL&PCR DNA purification kit (Euroclone) and subjected to Sanger sequencing by Mix2Seq Kit (Eurofins Genomics) following the manufacturer’s protocol. Total RNA was also harvested in parallel for Sanger sequencing. Analysis of the electropherograms was performed by Fiji software. Quantification was performed by assessing the area under the peaks of the tranSNP nucleotides as a metric to calculate allelic fraction. The normalization process considered the relative efficiency of the four nucleotides’ incorporation during sequencing, using the information of local surroundings. The allelic fraction for targeted genes was determined for at least three independent experiments. Student’s t-test was used to calculate the p-value.

Targeted RNA sequencing

Targeted RNA-seq libraries were prepared using custom multiplex primers, each containing an ∼25-nucleotide (25-nt) gene-specific sequence and either a 33-nt Illumina adaptor (5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAG-3′) at the 5′ end of the forward primer or a 34-nt Illumina adaptor (5′-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAG-3′) at the 5′ end of the reverse primer. Targeted RNA-seq libraries were prepared from DNase-treated RNA. cDNA was synthesized using a dTVN primer and purified with SPRIselect beads to remove excess primers. Fifty nanograms (50 ng) of cDNA was amplified (first PCR) using gene-specific forward and reverse primers attached to Illumina-specific adapters for cluster generation and sequencing (5′-Illumina Nextera adaptor + sequence-specific forward and reverse primer sequence-3′, see Table S4). A five-cycle second PCR was performed using Phusion High-Fidelity DNA Polymerase (ThermoFisher Scientific, Waltham, MA) to incorporate Illumina Nextera primers for sample indexing. After purification with Beckman Ampure XP beads, quantification with Qubit, and quality control using the PerkinElmer LabChip GX, the indexed libraries were pooled, quantified by qPCR using the Roche KAPA Library Quantification Kit (for Illumina platforms), and sequenced on the Illumina MiSeq platform (Illumina, San Diego, CA) using paired-end 250 bp reads with a NANO flowcell and v2 chemistry reagents. Quality control of raw reads was performed using FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Adapter trimming and quality filtering were then conducting with Trimmomatic v0.39⁷⁴ and CutAdapt 4.1⁷⁵ to remove Nextera adapters and primer sequences. The resulting FASTQ files were then aligned to the GRCh38 reference genome using STAR v2.7.10b,⁷⁶ and alignment files were further processed with clipOverlap tool implemented into BamUtil software v1.0.15 (https://genome.sph.umich.edu/wiki/BamUtil:clipOverlap) to trim overlaps between paired reads. Pileup files and variant statistics, including coverage supporting the reference allele, coverage supporting the alternative allele, and the variant allelic fraction, were generated using PacBAM.⁷⁷

Cloning of 3’UTR sequences, transfection, and luciferase assays

The 3’UTR regions surrounding the candidate TranSNPs were amplified from the HCT116 genomic DNA, using primers containing XbaI and EcoRI restriction sites and resulting in ∼500 bp fragments (Table S4). The resulting PCR fragments were cloned immediately downstream of a modified Firefly Luciferase in pGL4.13 [Luc2/SV40] vector (Promega), where a linker sequence adding an EcoR I site (F: 5’ cAagaattcatatgcccgggt 3’, R: 5’ ctagacccgggcatatgaatt 3’) was inserted at the original Xba I site to facilitate the insertion of fragments in a desired orientation. Correct cloning was confirmed by colony PCR and Sanger sequencing. HCT116 cells were seeded into 96-well plates and transiently transfected the empty pGL4.13 modified plasmid or those containing the allele-specific 3’UTRs fragments along with the Renilla control luciferase (pRL-SV40) in a 3:1 ratio (60 ng plus 20 ng) using Lipofectamine™ LTX and Plus™ Reagent (ThermoFisher Scientific) or FuGene4K (Promega). After 48 hours, Firefly and Renilla luciferase activities were quantified using a Dual-Luciferase™ Reporter (DLR™) Assay System (Promega). The transfected cells were lysed with 50 μL of 1X Passive Lysis Buffer for 15 minutes. The lysates were resuspended, and 15 μL of the resuspension was transferred to a white 384-well plate. Subsequently, 15 μL of LAR (Luciferase assay reagent) was added, and the luminescence was read at the Varioskan LUX Microplate reader. Following this, 15 μL of the Renilla substrate was added, and the luminescence was measured again to quantify Renilla luciferase expression. RNA (TRIzol Reagent, Invitrogen) was extracted from the lysates, and the Firefly Luciferase mRNA was quantified by qPCR for additional normalization.

CRISPR/Cas9-mediated knock-in

Edited HCT116 clones at the SNP sites were generated exploiting a CRISPR-Cas9 mediated knock-in with the Cas9 delivered as a ribonucleoparticle (RNP), as previously described.²² We designed a single gRNA (Table S5) in a position as close as possible to the tranSNP site to cut both alleles and maximize the efficiency of Homology-directed Repair (HDR) (Figure S2A). The same protocol with few modifications (not providing the donor DNA as a template for repair and not inhibiting DNA-PKcs) was used beforehand to test the cutting efficiency of the guide RNA (gRNA). For both protocols, the gRNA was prepared by mixing 1:1 Alt-R® CRISPR-Cas9 crRNA (IDT™) and Alt-R® CRISPR-Cas9 tracrRNA (IDT™). The mixture was heated at 95°C for 5 minutes and let cool down at room temperature for 5 minutes for optimal annealing. 3 μl of gRNA were further incubated with 120 pmol of recombinant Cas9 for 20 minutes at room temperature to generate the ribonucleoparticle (RNP). In parallel, previously seeded HCT116 wild-type cells were harvested by trypsinization, and 2.5 x 10⁵ cells per condition were electroporated in the 4D-Nucleofector™ System (Lonza) using the SE Cell Line 4D-Nucleofector™ X Kit S (Lonza) as described in.²² The electroporation mixture combined 20 μl of cell suspension, 1.2 μl of Alt-R® Cas9 Electroporation Enhancer (IDT™), 5 μl of RNP complex, and PBS to a final volume of 30 μl. In addition to these, 1.2 μl of Ultramer® DNA Oligonucleotide (100 μM, IDT™) were added in the case of knock-in generation to provide a template donor DNA for HDR. Electroporated cells were recovered from the strip and seeded in a 12-well plate format either untreated or, in the case of knock-in generation, treated with 1 μM of NU7441 DNA-PKcs inhibitor for 48 hours to favor HDR occurrence. The media, both with and without the drug, was changed after 24 hours. The cells were kept in culture up to full recovery and expanded until 2.5 x 10⁵ cells could be harvested for DNA extraction and Sanger sequencing. DNA was extracted by adding 20 μl of the QuickExtract™ DNA Extraction Solution (Lucigen) to the cell pellet. After vortexing, the sample was incubated at 65°C for 15 minutes and 98°C for 10 minutes. 1 μl of this DNA was used for the PCR amplification using the GoTaq® Green Master Mix (Promega) in 25 μl of reaction. PCR primers (Table S4) were specifically designed to produce 300-700 bp long amplicons and to have the SNP of interest asymmetrically located 100-200 bp from one or the other end of the amplicon, to have optimal sequencing interpretation. The PCR product was then purified using the MinElute™ PCR Purification Kit (QIAGEN). 75 ng of purified PCR products were sequenced using the Mix2Seq Kit and Sanger sequencing service provided by Eurofins. The TIDE software²⁵ was employed to identify and quantify insertions and deletions (indels) in the bulk edited population. The rs1053639 sgRNA presented a cutting efficiency of around 80%. The Synthego ICE software (Synthego Performance Analysis, ICE Analysis. 2019. v3.0) was employed to assess the knock-in efficiency in the edited bulk population. A limiting dilution assay was performed on the knock-in edited bulk population by seeding 1 cell/well in 96-well plate format to eventually obtain clones. The media was changed once a week for 3-4 weeks, then potential clones were expanded to 24-well plates, and at the same time, pellets were harvested for DNA extraction and sequencing. The efficiency of clone production was about 1 in 3 seeded cells; 85 clones were screened by Sanger sequencing. Three CRISPOR-predicted²⁷ off-target sites were also inspected by Sanger sequencing in the bulk populations and in selected clones (Figure S2).

cDNA synthesis and qRT-PCR

RNA was extracted by using the TRIzol® reagent (Invitrogen™) or the NucleoSpin ® RNA kit (Machery-Nagel) based on the downstream applications. RNA purity and quantity were assessed by NanoDrop™ Spectrophotometer (Thermo Fisher Scientific), followed by random primer-mediated reverse transcription using the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific) with 1000 ng of RNA as input. The cDNA was diluted 1:8 and 12.5ng -total RNA equivalent- in 10μl reaction were used in the following RT-qPCR analysis with the qPCRBIO SyGreen Mix (PCR Biosystems). All reactions were performed in 384-well plates using the QuantStudio™ 5 Real-Time PCR System (Applied Biosystems®) instrument. Primers were designed using Primer Blast (NCBI) and are listed in Table S4. All kits were used following the manufacturer’s instructions. All qPCR reactions were performed in triplicate, and Cq values were averaged. The fold change was calculated using dCt and normalized to at least two gene references.

RNA half-life measurement

3 x 10⁵ HCT116 cells from edited clones were seeded in a 6-well plate format. After 24 hours, cells from the first timepoint (t0) well were collected by trypsinization. The cell pellet was resuspended in 500 ul of TRIzol® reagent (Invitrogen™) and frozen at -80°C. The media was aspirated from the remaining wells and replaced with 10 ug/ml Actinomycin D in 2 ml of fresh media to achieve RNA polymerase II inhibition. After 20 minutes, 40 minutes, 1 hour, 2 hours and 4 hours from the treatment, cells were pelleted from the corresponding timepoint wells and resuspended in TRIzol®. Once all the pellets were recovered and frozen, they were thawed at room temperature for 15 minutes and RNA was extracted following the TRIzol® manufacturer’s instructions. RNA was quantified using the NanoDrop™ Spectrophotometer (Thermo Scientific™) and 1000 ng of timepoint t0 RNA from each HCT116 clone was retro-transcribed using the RevertAid First Strand cDNA Synthesis Kit (Thermo Scientific™) with random hexamers. For timepoints > t0, RNA was retro-transcribed volumetrically to the reference t0 for each clone. The cDNAs were diluted 1:8 and 2 μl in 10 μl of reaction and were used in the following RT-qPCR analysis with the qPCRBIO SyGreen Mix (PCR Biosystems) and the QuantStudio™ 5 Real-Time PCR System (Applied Biosystems®) instrument. Primers are listed in Table S4. GAPDH and 18S rRNA levels were also evaluated as stable mRNAs. The mRNA abundance was calculated relative to the start of transcription inhibition (t0) for each HCT116 clone (2^∧-dCt). The mRNA decay rate was calculated by a nonlinear regression curve fitting using the one-phase decay function (GraphPad Prism version 10).

Western blot analysis

1.6 x 10⁵ HCT116 cells from edited clones were seeded in a 6-well plate format. When cells reached 70 - 80 % confluence, they were treated with 100 nM Thapsigargin or vehicle for 4 hours. Cells were washed with PBS, and total cellular protein was isolated using a RIPA buffer containing protease inhibitor cocktail (cOmplete™, Mini Protease Inhibitor Cocktail, Roche). Lysates prepared for analysis of phosphorylated proteins were supplemented by a phosphatase inhibitor cocktail (Phosphatase Inhibitor Cocktail I, MedChemExpress). After vortexing and incubation at 4°C for 20 min, cell lysates were centrifuged at 15000 rpm at 4°C for 15 min, with supernatant collected and stored at -80°C. Prior to western blotting, the lysates were denatured by boiling at 95°C for 10 min in 1x Laemmli Sample Buffer (Bio-Rad). The protein concentration in each sample was determined using Pierce™ BCA Protein Assay Kits (Thermo Fisher Scientific). Equal amounts of protein from each lysate were then subjected to SDS-PAGE and electrophoretically transferred to a nitrocellulose membrane (Cytiva Amersham™ Reinforced Nitrocellulose Blotting Membrane). Each membrane was blocked with 5% dry milk in phosphate-buffer saline and 0.1% Tween-20 (PBST) for 1 h and then incubated with the indicated primary antibodies overnight at 4°C. The membranes were rinsed and incubated with peroxidase-conjugated secondary antibodies for 1 hour at RT. Immunoblotted membranes were developed with ECL™ Select western blot detection reagents (Amersham Biosciences, UK) and imaged on the ChemiDoc™ XRS imaging system (Bio-Rad). Image processing and densitometric quantification of the bands were performed using the Image Lab software. Protein expression was normalized against housekeeping protein, which was an internal control of protein loading. Relative protein expression was then calculated by dividing the normalized intensity of the condition sample by the control. Antibodies used for western blot analysis are listed in Table S7. The corresponding anti-mouse and anti-rabbit secondary antibodies were used at 1:10000 dilutions.

Protein half-life by cycloheximide chase assay

6 x 10⁵ HCT116 cells with the three genotypes were seeded in a 6-well plate format. After 24 hours, cells from the first timepoint (t0) well were collected by trypsinization. The cell pellet was resuspended in 50 μl of RIPA lysis buffer prepared as in the Western Blot section and supplemented with 1X protease inhibitors (Roche). In the meantime, the media was aspirated from the remaining wells and replaced with 100 ug/ml of cycloheximide (Cayman Chemical) in 2 ml of fresh media to prevent further protein synthesis. After 10 minutes, 20 minutes, 30 minutes, and 1 hour from the treatment, cells were pelleted from the wells of the corresponding timepoints and resuspended in RIPA buffer 1X as well. Once all the pellets were harvested, proteins were extracted by centrifugation at 15000 rpm for 20 minutes at 4°C, quantified by the Bicinchoninic Acid Assay (Thermo Scientific™ Pierce™ BCA Protein Assay Kit), and Western Blot was performed as described earlier. 35 μg of timepoint t0 protein extract from each HCT116 clone was used in Western Blot. For timepoints > t0, the protein amounts were loaded volumetrically to the reference t0, for each clone. DDIT4 and β-Tubulin antibodies were used as primary antibodies. Mouse secondary antibody was used for both the primary antibodies. β-Tubulin was found to be a stable protein and thus was used as a loading control even in this context. Upon normalization of each DDIT4 band intensity to that of the corresponding β-Tubulin, relative protein abundance over time was calculated to the start of translation inhibition (t0) for each HCT116 clone.

Global translation by Click-iT® metabolic labelling or puromycin incorporation assay

Global translation was measured by using the Click-iT™ Plus OPP Alexa Fluor™ 488 Protein Synthesis Assay Kit according to the manufacturer’s instructions. The fluorescence was measured by the ImageXpress High-Content imaging system (Molecular Devices). For the puromycin incorporation assay, 6 x 10⁵ HCT116 cells with the three genotypes were seeded in a 6-well plate format. When the cells reached 60% confluence, they were treated with vehicle or 100nM Thapsigargin for 4 or 16 hours. 1μM puromycin was added for 30 min at 37°C prior to pellet collection. Cycloheximide 100 μg/ml was used as a positive control. The cells were lysed, and the puromycin incorporation was assayed by western blot as described above. Stain-free intensity was measured as the reference.

Cell proliferation assay

HCT116 DDIT4 edited clones were seeded at 2.5 x 10³ per well onto 96-well plates and incubated at 37°C in standard conditions. The cell number was followed by the high content fluorescent microscope Operetta (PerkinElmer) in digital phase contrast every 24h for 96 h.

RNA immunoprecipitation-qPCR (RIP-qPCR)

8 x 10⁶ HCT116 cells from edited clones were seeded in 150 mm cell culture dishes (Sarstedt) and harvested after 48 hours by scraping. Cells were lysed in 800 μl of mild lysis buffer (20 mM HEPES pH 7.9, 150 mM NaCl, 1% Triton-X 100, 10% glycerol, 1 mM MgCl₂, 1 mM EGTA, 100 U/mL RNasin (Promega), 1X protease inhibitors (Roche)). After 30 minutes of incubation on ice, the lysate was obtained by centrifugation at 15000 rpm, 4°C for 10 minutes. The protein amount was quantified by the Bicinchoninic Acid Assay (Thermo Scientific™ Pierce™ BCA Protein Assay Kit). Meantime, 20 μl of Protein A Dynabeads (Invitrogen™) per condition were washed twice with 500 ul of IPP500 buffer (20 mM HEPES pH 7.4, 500 mM NaCl, 1.5 mM MgCl₂, 0.5 mM DTT, 0.05% NP40) and incubated with either 0.22 μg of anti-RBMX/hnRNP G antibody (#14794, 1:40, Cell Signaling Technology®) or 0.22 μg of anti-Normal Rabbit IgG antibody (#2729, Cell Signaling Technology®) in 100 μl of IPP500 buffer for 1h 30 at room temperature in a rotating wheel. After antibody conjugation, beads were further washed once with 900 μl of IPP500 buffer and twice with 900 μl of IPP150 1x buffer (20 mM HEPES pH 7.4, 150 mM NaCl, 1.5 mM MgCl2, 0.5 mM DTT, 0.05% NP-40) and once with PBS. Nonspecific binding was blocked by incubating the beads with 300 μl of 10 mg/ml Bovine Serum Albumin (BSA) for 30 minutes at room temperature in a rotating wheel. Beads were further washed three times with 800 μl of IPP150 1X buffer and incubated with 1 mg of the protein lysate in IPP150 1X buffer and 1x protease inhibitors in a final volume of 400 μl O/N at 4°C in a rotating wheel. For nuclear and cytoplasmic RIP (Figure S5) the lysates prepared in fractionation buffer (see Cell Fractionation section below) were quantified and incubated similarly in IPP150 1X buffer. Before the incubation with the beads, 10% of lysate in IPP150 1X buffer was isolated as protein input for Western Blot analysis, and 1% was isolated as RNA input for qPCR analysis. After O/N incubation, the flow-through (unbound lysate) was collected for Western Blot analysis, the beads were washed four times with 800 μl of IPP150 1X buffer, and 10% of the last wash was also collected for Western Blot analysis upon elution in LB 1X (Laemmli Sample Buffer, Bio-Rad) heating at 98°C for 10 minutes. Conversely, elution both for RIP samples (beads) and RNA inputs was performed by incubation with 5 μl of >600 U/ml proteinase K solution (Qiagen) and 150 ul of Proteinase K buffer (20 mM Tris-HCl pH 7.4, 10 mM MgCl₂, 0.2% Tween-20) at 55°C for 30 minutes in a lab heat block shaking at 700 rpm. The output RNA is extracted with TRIzol® reagent (Invitrogen) according to the manufacturer’s instructions. The total amount of eluted RNA was retrotranscribed using the RevertAid First Strand cDNA Synthesis Kit (Thermo Scientific) and 1 μl of cDNA in 10 μl of reaction was used in the following RT-qPCR analysis with the qPCRBIO SyGreen Mix (PCR Biosystems) and the QuantStudio 5 Real-Time PCR System (Applied Biosystems®) instrument. Primers are listed in Table S4.

RNA electromobility shift assay (REMSA)

The interaction between DDIT4 mRNA and RBMX protein was assessed in vitro by REMSA. The full-length recombinant RBMX protein was purchased from Origene (TP710403). FAM-labelled fluorescent RNA probes (GAGGGACUGAUUCCU/AGUGGUUGGAA-3’FAM) for the T and A alleles were purchased from Metabion. The binding reaction was set up in 10 μl in the following binding buffer (10 mM HEPES, 50 mM KCl, 1 mM EDTA, 0.05% NP-40, 10% Glycerol, 0.5 mM DTT, pH 11.5). 10 U of RNasin (1 μl) were also added to each reaction and incubated for 30 minutes at RT in the dark. Upon incubation, 1 μl of 6X Orange Dye was added to each reaction to facilitate the visualization of the samples. The binding complex was resolved by running this reaction on a 6% TBE polyacrylamide native (non-denaturing) gel at 80V for 1 hour in TBE 1X at RT. The gel was pre-run in cold TBE 0.5X at 100V for 30 minutes and imaged using the SYBR filter in the ChemiDoc™ XRS imaging system (Bio-Rad).

Cell fractionation

Cell fractionation was performed using a 0.1% Igepal buffer in PBS 1X as in the REAP fractionation protocol,⁸² coupled with an RNA extraction protocol⁸³ to improve the extraction of the RNA from the nuclei. Briefly, HCT116 TT and AA clones were seeded in 10-cm dishes. Once confluent, cells were detached by trypsinization, washed with PBS, and lysed in 900 μl of ice-cold 0.1% Igepal buffer in PBS 1X buffer for 2 minutes; 300 μl of this lysate were removed as whole cell lysate (WCL). The lysate was then fractionated by centrifugation at 12000 rpm for 40s, at RT. 300 μl of the supernatant were isolated as a cytoplasmic fraction (CYTO). The pellet was washed with 1 ml of 0.1% Igepal buffer and centrifuged at 12000 rpm for 40s, at RT to obtain the nuclear fraction. 50 μl out of 300 μl were isolated from each fraction for Western Blot analyses upon further lysis by RIPA buffer. The remaining 250 μl were used for TRIzol-based RNA extraction. To implement RNA extraction, the CYTO and NUCLEI fractions were resuspended in 1 ml of TRIzol. Lower volumes of TRIzol did not successfully extract RNA in our conditions. RNA was incubated for 10 minutes in TRIzol at RT, then 10 μl of EDTA 0.5 M were added before incubation at 65°C for 5 min, 600 rpm (or until the pellet was completely dissolved). Subsequently, chloroform was added to follow the conventional TRIzol protocol.

RBMX silencing

HCT116 TT and AA cells were transfected with a single or a mixture of two out of three siRNAs targeting RBMX (IDT) at a final concentration of 40 nM for 48 hours (Table S4). Control cells were transfected with a non-targeting siRNA (IDT, 51-01-14-04) in the same conditions. INTERFERin (Polyplus) was used as a transfection reagent.

Xenograft of HCT116 cells in zebrafish larvae

Animal rearing: Zebrafish (Danio rerio) strains were raised and maintained in the Model Organism Facility (MOF) at the Department CIBIO, University of Trento under standard conditions.⁸⁴ The transgenic zebrafish line Casper⁵⁶ was used to acquire live images in transparent larvae. All zebrafish studies were performed according to European and Italian law, D.Lgs. 26/2014, authorization11/2023-PR to MM. Larvae preparation: Embryos were initially maintained at 28°C at a maximum density of 50 embryos per Petri dish in fish water supplemented with 0.0002% methylene blue (Sigma, Burlington, MA, USA) as an antifungal agent. After 24 h, embryos were placed in fish water without methylene blue. Xenotransplantation of human cancer cell lines: the protocol of xenotransplantation procedure was adapted from.^57,85 GFP or RFP labeled HCT116 cells (edited AA, TT, and AT clones) were harvested from a 100 mm Petri dish on the day of xenotransplantation and resuspended in 70 μl of PBS. 0.5 μL of cell suspension containing 200-400 cells was transplanted in the swim bladder of 2 dpf zebrafish larvae anesthetized in 0.16 mg/ml PBS/Tricaine (MS-222) under a Nikon SMZ800 stereomicroscope, using a micromanipulator (Marzhouser Wetzkar MM 33 links) and the manual microinjector (CellTram® 4r Air Eppendorf). Transplanted embryos were maintained at 33°C in fish water. Larvae were selected for good transplantation profile using a Leica MZ 10F fluorescence stereomicroscope. Live-imaging and image analysis: each larval xenograft was imaged one-day post-transplantation (D1, i.e., 3 dpf) and 3 days post-transplantation (D3, i.e., 5 dpf) using Leica TCS SP8 laser scanning confocal system (CD7) (Objectives: 10x/1). Images were acquired in z-stack mode with a z-step interval of 7 μm to a range of 170-230 μm depending on the tumor size. The 488/633 nm lasers were used with appropriate filters to acquire green and red fluorescent cancer cells. Image analysis was performed using Fiji software. To study tumor growth of cancer cells in vivo, the maximum projection of every image was created, and a threshold was applied for each larva on day 1 (D1) and day 3 (D3) to generate a mask of the tumor area. The area on D3 was normalized to the area of D1 to assess the percentage of tumor growth for each larva.

Single-molecule inexpensive RNA fluorescence in situ hybridization

The smiFISH protocol was adapted from.⁸⁶ Cells were grown on 22 × 22 mm glass coverslips (Prestige) previously stripped with 1 M HCl, sterilized, and stored in 100% ethanol. Cells were fixed in 4% paraformaldehyde (PFA) (Electron Microscopy Science) for 20 min at room temperature and then washed two times with 1× PBS. Permeabilization was performed in 0.5% Triton X-100, 1× PBS for 5 min at room temperature. Once permeabilized, samples were incubated at room temperature in a solution containing 1× SSC and 15% formamide, for 40 min and placed cell-side down on a 50-μl drop of hybridization buffer containing the probe [1× SSC, 15% formamide, bovine serum albumin (BSA) (4.5 mg/ml), 10.6% dextran sulfate, 2 mM Vanadyl Ribonucleoside Complex (VRC), yeast tRNA (0.4 mg/ml), 1 μl of probe mix] in an air-tight hybridization chamber protected from light overnight at 37°C. The probe mix preparation is described in the next paragraph. The following day, coverslips were washed twice with prewarmed 1× SSC solution containing 15% formamide for 30 min at 37°C. After that, coverslips were rinsed twice with 1× PBS, cells were stained with DAPI (1 ng/μl) (4′,6-diamidino-2-phenylindole) in 1× PBS, for 12 min at room temperature and washed with 1× PBS. Samples were mounted on precleaned glass slides using ProLong Diamond Antifade Mountant (Thermo Fisher Scientific). smiFISH probe preparation: probe preparation was performed as described by.⁸⁶ For DDIT4 detection, 3 primary probes targeting the DDIT4 sequence were used and prepared as a 20 μM equimolar mixture in milliQ water. Sequences of the probes are shown in Table S4. Primary probes were mixed with 50 pmol of fluorescently labeled secondary probes in 1× NEB3 solution [100 mM NaCl, 50 mM tris- HCl (pH 8), 10 mM MgCl2] in a final volume of 10 μl. Primary probes were annealed with secondary probes in a PCR thermocycler (85°C for 3 min, 65°C for 3 min, 25°C for 5 min. As described in the previous paragraph, the probe mix (1 μl) was used for the smiFISH hybridization buffer preparation. smiFISH combined with IF: the smiFISH technique was performed first, using the protocol described in the smiFISH paragraph, and followed by the immunofluorescence protocol as described hereafter. After the overnight probe hybridization, cells were washed twice in prewarmed 1× SSC, 15% formamide solution. After two rinses in 1× PBS, cells were incubated in blocking solution (1× PBS, 2% BSA, 0.1% Triton X-100) for 3 hours and incubated for 1 hour at room temperature with the primary antibody diluted in blocking solution. For RBMX detection, we used the antibody in Table S7 at 1:200 dilution. Cells were then washed five times for 6 min with 1× PBS, 0.1% Triton X-100, and incubated with the appropriate fluorescent secondary antibody for 1 hour [goat anti-rabbit Alexa Fluor 488 (AF488), from Thermo Fisher Scientific, catalog no. A-11034, 1:1500 dilution; donkey anti-mouse AF647, from Thermo Fisher Scientific, catalog no. A-31571, 1:1000 dilution]. After incubation, cells were washed five times for 6 min with 1× PBS, 0.1% Triton X-100, and stained with DAPI (1 ng/μl) in 1× PBS for 12 min. After two washes with 1× PBS, coverslips were mounted on glass slides using ProLong Diamond Antifade Reagent (Thermo Fisher Scientific). smiFISH/IF image acquisition by confocal microscopy: images were acquired using a confocal laser scanning Leica TCS SP8 inverted microscope (Leica Camera AG) with a plan apochromatic 63×/1.40 oil immersion objective and 2× zoom. Normal photomultipliers (PMTs) were used to detect DDIT4 smiFISH signal, RBMX, b-tubulin, and DAPI signals. Optimal acquisition parameters, including nanometer range of excitation, gain, offset, and laser power, were optimized for the visualization of each fluorophore at the beginning of the acquisition and maintained throughout each experiment. Z stack images were acquired in each experiment with a 0.3-μm step and 2048 × 2048 pixel size resolution using Leica Application Suite X (LAS X) imaging software.

Confocal image analysis

Images were analyzed using CellProfiler 4.0.7 (Broad Institute, Inc.). Briefly, cell nuclei were identified as the primary object using the DAPI signal. For each object, the cellular region was defined by propagating from the nucleus, applying a low segmentation threshold to the smoothed smiFISH signal. The cytoplasm was defined as a tertiary object by subtracting the nuclear masks from the cellular regions. DDIT4 smiFISH signal spots were segmented as objects with a typical diameter range of 10 to 30 pixels. Subsequently, objects were localized based on their position relative to the nuclear or cytoplasmic regions. Since the majority of the spots exhibited a perinuclear localization, the nuclear regions were reduced by 5% to prevent misleading classification. Morphological features and intensity values for both DDIT4 smiFISH and RBMX signals were calculated at the object level.

Immunohistochemistry

HCT116 rs1053639 TT and AA homozygous cells were collected by centrifugation and paraffin embedded. Sections from the paraffin were then cut and incubated overnight at 37°C. IHC staining was performed using Bond III IHC autostainer for full Automated Immunohistochemistry (Leica Biosystems). Antigens were unmasked with Bond Epitope Retrieval Solution 1 Leica AR9961 and incubated with the following antibody (diluted with Bond Primary Antibody Diluent, AR9352 Leica): DDIT4/REDD1 at dilution 1:400 (Proteintech 10638-1-AP). Samples were stained with BOND IHC Polymer Detection Kit (DS9800) and counterstained using Hematoxylin solution (Leica).

Quantification and statistical analysis

Statistical analysis

Results were plotted and statistically analyzed using GraphPad Prism, version 10. The statistical tests applied, the number of replicates, the mean, standard deviations, and the level of significance are provided in each figure legend.

Supplemental information

Data S1. Uncropped western blot images, related to Figures 2, 4, 5, 6, S1, S4, and S5

The specific figure panels are indicated in the file.

Table S2. List of the tranSNPs identified

rsID, gene name, location, nominal p values, AF data, and experimental conditions are reported, related to Figure 1.

Table S4. Sequences of primers used for cloning, qPCR, Sanger sequencing, and preparation of targeted libraries

Sequences of siRNAs targeting RBMX. Sequences of DDIT4 FISH probes.