Prostate cancer is the archetypal lineage-dependent malignancy. Its growth hinges on the androgen receptor (AR), which is why androgen deprivation therapy and second-generation AR pathway inhibitors such as enzalutamide, apalutamide and darolutamide form the backbone of advanced disease treatment. Yet resistance is nearly universal. Tumors frequently shed their AR dependence through a process called lineage plasticity, transdifferentiating into neuroendocrine-like states that no longer respond to AR-targeted drugs. Genetic loss of TP53, PTEN or RB1 can drive this transition, but the translational mechanisms, the events at the level of protein synthesis, that permit such phenotypic flexibility have remained largely unknown.

To find them, the team performed unbiased small RNA sequencing using a multiplex platform capable of quantifying full-length tRNAs, their fragments and selected chemical modifications. They applied this technique to an engineered model in which LNCaP prostate adenocarcinoma cells were pushed into a lineage-plastic state through RB1 knockdown and overexpression of mutant TP53, MYCN, ASCL1, SRRM4, NR0B2, BCL2 and mutant KRAS. Among 49 tRNA isoacceptor families surveyed, only tRNAArg(UCU) was significantly downregulated during the transition, with a log2 fold change of roughly minus 0.34 and a false discovery rate of 2.42 x 10-25 in one analysis. Within the five tRNAArg(UCU) isodecoders, molecules sharing the same anticodon but differing in body sequence, tRNA1Arg(UCU) was the most abundant and the only one substantially decreased. Northern blotting and quantitative PCR confirmed the reduction in both LNCaP and C4-2B cells, and the decline was not accompanied by changes in tRNA fragmentation or chemical modification, indicating a specific loss of the mature molecule.

The correlation with lineage identity was striking. Across nine prostate cancer cell lines, tRNA1Arg(UCU) abundance showed a strong positive correlation with AR activity (Pearson’s r = 0.85, P = 0.0034) and a negative correlation with the neuroendocrine markers ENO2 and SCN3A. In the LTL331 patient-derived xenograft model, which transdifferentiates from adenocarcinoma to neuroendocrine prostate cancer after castration, tRNA1Arg(UCU) levels fell in a stepwise fashion from pre-castration through castration to relapse. The researchers then developed an in situ hybridization assay to visualize the tRNA directly in patient tissue. In specimens from 56 patients in the University of Washington rapid autopsy cohort, AR-positive tumors expressed significantly higher levels of tRNA1Arg(UCU) than neuroendocrine-positive tumors, and tRNA abundance correlated positively with AR activity and negatively with the neuroendocrine marker ELAVL4.

Crucially, the tRNA was not merely a passive marker. Using inducible short hairpin RNAs, the team reduced tRNA1Arg(UCU) levels by 25 to 50 percent in LNCaP and C4-2B cells. The knockdown was specific, leaving other tRNAArg(UCU) isodecoders untouched. The consequence was a coordinated shift in cell identity: AR pathway genes declined at both RNA and protein levels, while neuron-related gene programs rose. Cells depleted of the tRNA became significantly more resistant to enzalutamide, apalutamide and darolutamide, yet more sensitive to alisertib, an Aurora kinase A inhibitor with activity in neuroendocrine prostate cancer. In mice, the team generated animals haploinsufficient for n-Trtct2, the gene encoding tRNA1Arg(UCU), crossed onto the MYC-driven Hi-Myc prostate cancer model. Organoids derived from these mice showed a threefold increase in resistance to AR inhibition, and after surgical castration, haploinsufficient mice developed significantly enlarged prostates, with 25 percent exhibiting high-grade prostatic intraepithelial neoplasia compared with none of the controls.

Most remarkably, the process proved reversible. Re-expressing tRNA1Arg(UCU) in lineage-plastic cells restored AR pathway gene expression, suppressed neuroendocrine markers and resensitized the cells to AR-targeted therapies. Overexpression of other arginine tRNA isodecoders or isoacceptors had no such effect, underscoring the exquisite specificity of this single molecule. In LNCaP-abl cells, a subline that spontaneously evolved androgen insensitivity after prolonged culture in androgen-depleted medium, restoring tRNA1Arg(UCU) enhanced AR signaling and improved enzalutamide sensitivity. In the TRAMP transgenic mouse model, which normally progresses toward neuroendocrine tumors, crossing in a tRNA1Arg(UCU) overexpression allele reduced the incidence of premalignant glands after castration. Together, these experiments demonstrate that lineage dependency in prostate cancer can be toggled by modulating one tRNA species.

The team then asked how tRNA1Arg(UCU) is regulated. Mining the ENCODE database of transcription factor binding, they identified 11 DNA-binding proteins that occupy the TRR-TCT1-1 genomic locus encoding the tRNA. Four of these were downregulated during lineage transition, and functional screening pinpointed two: TARDBP, a DNA- and RNA-binding protein best known for its role in amyotrophic lateral sclerosis, and ZSCAN29, a zinc-finger protein. Silencing either factor reduced tRNA1Arg(UCU) expression, and their occupancy, measured by CUT&RUN profiling, declined selectively at the TRR-TCT1-1 locus during lineage plasticity. The locus carried the highest enrichment of the H3K4me3 histone mark among all six TRR-TCT isodecoder genes, and this mark, along with RNA polymerase III occupancy, was lost specifically at TRR-TCT1-1 during the lineage switch. TARDBP and ZSCAN29 appear to maintain RNA polymerase III recruitment at this chromatin-primed locus, and their expression correlated with tRNA1Arg(UCU) levels in patient specimens. This work provides one of the first demonstrations of isodecoder-specific transcriptional control of tRNA genes in cancer, extending tRNA regulation beyond canonical RNA polymerase III mechanics.

How does a single tRNA change cell fate? The answer lies in codon optimality. tRNA1Arg(UCU) decodes the AGA arginine codon, and the researchers built fluorescent reporters to show that translational capacity at AGA codons dropped roughly fourfold in lineage-plastic cells and was restored by tRNA addback. Polysome RNA sequencing, which separates ribosome-bound messenger RNAs by density to measure translation efficiency genome-wide, revealed that AGA codons were enriched among efficiently translated transcripts in adenocarcinoma cells, lost that enrichment in lineage-plastic cells, and regained it upon tRNA restoration. Among the translationally upregulated targets were five components of the SWI/SNF chromatin remodeling complex, and the team focused on SMARCC2, whose protein levels fell during lineage transition despite unchanged mRNA. A codon-switching experiment confirmed the mechanism: replacing SMARCC2’s AGA codons with synonymous CGC codons rendered the protein insensitive to lineage state. Silencing SMARCC2 in tRNA-restored cells reversed the AR pathway reactivation and restored enzalutamide resistance, establishing SMARCC2 as a key translational mediator through which tRNA1Arg(UCU) sustains lineage fidelity.

The clinical implications are substantial. In the rapid autopsy cohort, patients with the lowest tRNA1Arg(UCU) abundance had significantly shorter time to first bone metastasis, shorter survival after starting androgen deprivation therapy, shorter survival after developing androgen independence, and shorter overall survival overall. In mouse models, tRNA depletion led to a fourfold increase in metastatic burden following intracardiac injection of luciferase-labeled cells. The authors caution that the clinical associations are retrospective and that prospective studies will be needed to validate tRNA isodecoders as biomarkers of AR-targeted therapy resistance. But the therapeutic horizon is already visible: adeno-associated virus delivery of suppressor tRNAs has restored protein function in models of mucopolysaccharidosis type I, and lipid nanoparticle-based tRNA delivery has re-expressed CFTR in cystic fibrosis epithelia. Such platforms could, in principle, be adapted to restore tumor-suppressive tRNAs like tRNA1Arg(UCU), reprogramming lineage dependency and resensitizing tumors to existing therapies. For a disease in which treatment resistance remains the central clinical challenge, the idea that a single RNA adaptor molecule holds a master key to cellular identity is a genuinely paradigm-shifting proposition.

Subject of Research: Regulation of prostate cancer lineage plasticity and therapy resistance by tRNA dosage

Article Title: tRNA dosage regulates lineage dependency and resistance in prostate cancer

Article References: Kim, Y. S., Arora, S., Young, D., Tsou, A., Shiuan, A., Wladyka, C. L., Rudoy, D., Kim, J. Y., Waters, J. A., Schuster, S. L., Coleman, I. M., Kapur, M., Sobczyk, M., Katanski, C. D., Bayat Tork, A. M., Zhang, W., Nelson, P. S., Ha, G., Haffner, M. C., … Hsieh, A. C. (2026). tRNA dosage regulates lineage dependency and resistance in prostate cancer. Nature. https://doi.org/10.1038/s41586-026-11153-8

Image Credits: AI Generated

DOI: 10.1038/s41586-026-11153-8

Keywords: prostate cancer, transfer RNA, lineage plasticity, androgen receptor, therapy resistance, neuroendocrine prostate cancer, translation, SMARCC2, TARDBP, ZSCAN29, RNA polymerase III, codon optimality

The study, conducted by Hidenori Tani of Yokohama University of Pharmacy, takes an unusually careful comparative approach. Rather than comparing mRNA and lncRNA stability data from different experiments, different laboratories or different measurement platforms, Tani assembled half-life measurements derived from the same metabolic labeling and sequencing assays, allowing messenger RNAs and long non-coding RNAs to be evaluated on a genuinely level playing field. This design matters, because the field has been plagued by comparisons in which differences in assay chemistry, cellular context or transcript abundance could masquerade as biological differences in decay regulation. By holding the measurement method constant, the analysis isolates the one variable of interest: whether the sequence of a transcript encodes its own stability.

The findings for messenger RNAs confirm the power of codon-mediated regulation. When Tani re-estimated codon stabilization coefficients, quantitative measures of how strongly each codon is associated with transcript longevity, inside every cross-validation fold to eliminate information leakage between training and test data, the resulting model predicted mRNA half-life with a cross-validated coefficient of determination of 0.174 and a Spearman rank correlation of 0.44. Critically, the codon stabilization coefficient on its own, yielding an R-squared of 0.084, outperformed a six-member set of translation-independent features, which managed only 0.008. That ordering, with codon optimality beating a battery of sequence characteristics such as GC content, transcript length and known destabilizing elements, held consistently across three additional datasets, two distinct measurement technologies and two different species.

The picture for long non-coding RNAs could hardly be more different. In a carefully matched set of 364 lncRNAs measured in the same HeLa cell assay as the messenger RNAs, sequence-based models of half-life performed no better than chance, yielding an R-squared of minus 0.050, a value meaning the predictions were actually worse than simply guessing the average half-life for every transcript. The result was not a quirk of a small or unrepresentative sample. When Tani turned to the largest published lncRNA stability dataset, encompassing 33,285 transcripts, k-mer composition features achieved an R-squared of essentially zero, at 0.0007. Even the cross-validated rank correlation reported in that dataset’s original study, 0.091, is vanishingly small in absolute terms.

Here the study makes a subtle but important statistical argument. A Spearman correlation of 0.091 in a dataset of more than thirty thousand transcripts does clear the permutation null, meaning it is statistically detectable rather than pure noise. But detectability is not the same as explanatory power, and the contrast between the two RNA classes is one of magnitude, not of presence or absence. Messenger RNA half-life is meaningfully sequence-encoded; long non-coding RNA half-life, by every measure Tani applied, is not. The distinction has practical consequences for anyone attempting to model RNA decay: a signal that is real but negligible cannot support the kind of predictive machinery that works for coding transcripts.

The analysis also confronts one of the most fashionable tools in modern genomics: pretrained genomic language models. These deep neural networks, trained on enormous corpora of genomic sequence, have been touted as universal feature extractors capable of discovering biological signals without explicit programming. Tani tested models including HyenaDNA, which can process nearly complete transcripts at single-nucleotide resolution, and DNABERT-2. Even with one model covering 99.6 percent of transcripts in their entirety, the language models left messenger RNA prediction at an R-squared of just 0.042, and lncRNA prediction at values between 0.0002 and 0.0036. The implication is sobering: whatever the language models learned about genomic sequence, they did not uncover a hidden stability code in non-coding transcripts that simpler approaches had missed.

To rule out confounding, Tani stratified transcripts by predicted coding potential, testing whether a subset of lncRNAs with translated open reading frames might behave like messenger RNAs after all. They did not. Nor did differences in transcript length, GC content or sample size explain the gap: when lncRNAs and mRNAs were matched on all three characteristics and analyzed with an identical feature set, the messenger RNAs yielded an R-squared of 0.033 while the lncRNAs yielded 0.0002. Perhaps most convincingly, a signal-injection experiment established that the analytical pipeline could reliably detect effects explaining as little as 0.5 percent of the variance in half-life, a sensitivity threshold comfortably above every single lncRNA result reported in the study. If a sequence-based stability signal existed in these transcripts at even a modest level, the method would have found it.

The biological interpretation is as interesting as the statistical one. The mechanistic chain linking codon optimality to decay runs through translation itself: slow-moving ribosomes on unoptimized codons recruit decay factors, and the DEAD-box helicase Dhh1p in yeast monitors codon optimality directly to couple translation to destruction. Long non-coding RNAs, which are largely untranslated, simply cannot participate in this feedback loop. Their lifespans are instead governed by other forces, including RNA-binding proteins, nuclear retention mechanisms, structural elements and, in many organisms, the nonsense-mediated decay pathway, which can eliminate lncRNAs carrying premature stop codons. A growing literature documents hundreds of short-lived non-coding transcripts in mammalian cells and shows that nonsense-mediated decay restricts lncRNA expression even in RNAi-capable budding yeasts, underscoring that non-coding transcript turnover has its own logic, written in a different biochemical language.

The study’s broader lesson extends beyond RNA biology into the methodology of machine-learning-driven science. By re-estimating features within each cross-validation fold, Tani guarded against data leakage, the insidious practice pattern that has inflated reported performance across computational biology, and the work aligns with recent guidance on reproducibility in machine-learning applications. The bottom-line recommendation is blunt: mRNA-derived decay models should not be transferred to long non-coding RNAs without re-validation. For researchers designing RNA therapeutics, engineering synthetic transcripts or interpreting lncRNA dysregulation in disease, the message is clear. The stability of the non-coding transcriptome cannot be read from codon usage tables, and understanding how these molecules are scheduled for destruction will require looking elsewhere, at the proteins and structures that shepherd them through the cell.

Subject of Research: Codon optimality as a predictor of mRNA half-life and its failure to predict long non-coding RNA stability

Article Title: Codon optimality predicts mRNA half-life but does not transfer to lncRNAs

Article References: Tani, H. (2026). Codon optimality predicts mRNA half-life but does not transfer to lncRNAs. Molecular Genetics and Genomics, 301(1), Article 196. https://doi.org/10.1007/s00438-026-02520-1

Image Credits: AI Generated

DOI: 10.1007/s00438-026-02520-1

Keywords: codon optimality, mRNA stability, long non-coding RNA, RNA decay, RNA half-life, machine learning, cross-validation, genomic language models, nonsense-mediated decay, transcriptome, RNA-binding proteins, gene regulation