The lncRNA lincNMR regulates nucleotide metabolism via a YBX1 - RRM2 axis in cancer

Long intergenic non-coding RNA-Nucleotide Metabolism Regulator (lincNMR) is a long non-coding RNA (lncRNA) which is induced in hepatocellular carcinoma. Its depletion invokes a proliferation defect, triggers senescence and inhibits colony formation in liver, but also breast and lung cancer cells. Triple-label SILAC proteomics profiles reveal a deregulation of key cell cycle regulators in lincNMR-depleted cells like the key dNTP synthesizing enzymes RRM2, TYMS and TK1, implicating lincNMR in regulating nucleotide metabolism. LincNMR silencing decreases dNTP levels, while exogenous dNTPs rescues the proliferation defect induced by lincNMR depletion. In vivo RNA Antisense Purification (RAP-MS) identifies YBX1 as a direct interaction partner of lincNMR which regulates RRM2, TYMS and TK1 expression and binds to their promoter regions. In a Chick Chorioallantoic Membrane (CAM) in vivo model, lincNMR-depleted tumors are significantly smaller. In summary, we discover a lincRNA, lincNMR, which regulates tumor cell proliferation through a YBX1-RRM2-TYMS-TK1 axis governing nucleotide metabolism.

A comprehensive landscape of the transcriptome has revealed that although over a 75% of human genome is transcribed, only about 2% of it encodes for proteins 1 . In recent years, over 10,000 long non-coding RNAs (lncRNAs) have been identified, and their number is ever increasing. Although lncRNAs were initially viewed as background noise from junk DNA, it is now evident that they play an important role in various biological processes, and their deregulation has been linked to various cancers including liver cancer 2,3 . LncRNAs act as drivers of carcinogenesis by regulating one or multiple hallmarks of cancer [4][5][6] . Several well-studied examples include lncRNAs regulating viability [7][8][9] , proliferation [10][11][12] , growth suppression 13,14 , migration [15][16][17] , angiogenesis 18,19 , and cellular immortality [20][21][22] . Detailed epigenetic, genomic, and transcriptional analyses have revealed that lncRNAs have cancer type-specific deregulation patterns 23 and could be attractive molecules for therapeutic applications 24 . LncRNAs are heterogeneous molecules and play diverse functional roles by interacting with DNA, RNA, or proteins, such as recruitment of chromatin modifying complexes 25 or transcription factors 26 , controlling mRNA stability 27 , and acting as competing endogenous RNAs 28 .
Liver cancer is the fourth leading cause of cancer-related mortalities worldwide in 2018, and is predicted to be sixth most commonly diagnosed cancer 29 . Of all liver cancers diagnosed, primary liver cancer called hepatocellular carcinoma (HCC) constitutes 75-85% of the cases with limited treatment options in advanced stages warranting further investigations. Notably, lncRNAs like HULC, TERC, and HOTAIR have also been implicated in hepatocarcinogenesis 28,30,31 .
In this study, we investigate lncRNAs induced in liver cancer patient samples derived from high-throughput RNA sequencing data and identify the lncRNA lincNMR (long intergenic noncoding RNA-nucleotide metabolism regulator) upregulated in HCC and impacting tumor growth in vivo. Its in-depth molecular characterization unravels its role in controlling nucleotide metabolism via interacting with YBX1 and regulating RRM2, TK1, and TYMS.

Results
lincNMR is upregulated in hepatocellular carcinoma. To identify long noncoding RNAs (lncRNAs) deregulated in hepatocellular carcinoma (HCC), lncRNA expression was analyzed genome-wide based on the TCGA RNA sequencing dataset of liver cancer patients (tumor = 200 samples, normal = 50 samples). Out of 12,727 annotated lncRNAs in the TANRIC liver cancer dataset 32 , 217 lncRNAs were found to be significantly (P < 0.05) induced by at least fivefold (median tumor/normal). In total, 49 lncRNAs were selected based on their genomic location, repeat content, pseudogene content, and association with clinical properties, and their expression was validated in nine liver cancer cell lines (Supplementary Data 1). Nine expressed lncRNAs were further selected based on expression levels, coding potential, and novelty for RNAi-based phenotypic analysis using siPOOLs 33 . Among these candidates, the uncharacterized transcript RP6-65G23.3 yielded the strongest decrease of cell viability as determined by ATP content measurement in liver cancer cells, comparable with the effect of PLK1 34 or HULC 35 knockdown (Fig. 1a). For reasons described below, we named this lncRNA lincNMR (long intergenic noncoding RNA-nucleotide metabolism regulator).
Basic characterization of lincNMR. LincNMR is a lncRNA transcribed from a bidirectional promoter in a head-to-head orientation on chromosome 14. Since the lincNMR transcript had never been studied, we defined its gene boundaries using rapid amplification of cDNA ends (RACE). 5′RACE identified a transcription start site (TSS) upstream of the current GENCODE annotation ( Supplementary Fig. 1a). This finding was supported by RNA-Pol II Chip and switchgear TSS datasets ( Supplementary  Fig. 1b) corroborating the extended transcript identified in our 5′-RACE. 3′-RACE confirmed the previously annotated 3′-end of lincNMR, but also identified a second, less abundant isoform of lincNMR, including an additional internal exon ( Supplementary  Fig. 1c). Both isoforms have been deposited into Genbank with the accession numbers MK652436 and MK652437, respectively. Our 5′-and 3′-RACE results were supported by ENCODE/Cold Spring Harbor long RNA-Seq tracks from the ENCODE consortium 36 (Supplementary Fig. 1a, c). Next, we analyzed the coding potential of lincNMR using scores from phyloCSF 37 ( Supplementary Fig. 1d) and the Coding Potential Calculator 38 ( Supplementary Fig. 1e). Both algorithms classified lincNMR as a noncoding transcript. We determined the copy number of lincNMR per cell with at least two to seven copies. Since the subcellular localization maybe linked to the biological function of a noncoding RNA 39,40 , we performed subcellular fractionation with fraction-specific controls NEAT1 (chromatin fraction), RNU-1 (nucleoplasmic fraction) and DANCR (cytoplasmic fraction). LincNMR predominantly localized with 60-70% to the cytoplasm, but also showed considerable abundance in the nucleoplasm ( Supplementary Fig. 1f).
LincNMR depletion affects cell proliferation and induces senescence. To elucidate the cellular function of lincNMR, we depleted lincNMR using two independent siPOOLs for additional specificity and to exclude any off-target effects observed with single siRNAs 33 Fig. 2c) cancer cell lines. Since lincNMR knockdown decreased cell viability in liver cancer cells (Fig. 1a), cell proliferation was determined by performing BrdU incorporation assays. LincNMR silencing with two independent siPOOLs resulted in 30-80% decrease in cell proliferation in four liver cancer cell lines (HLE, HLF, SNU-387, and FLC-4) (Fig. 1b). Depletion of lincNMR also impaired cell proliferation in three breast (MCF-7, KPL-1, and T47D) ( Supplementary Fig. 2d) and three lung (A549, NCI-H460, and NCI-H1299) cancer cell lines ( Supplementary Fig. 2e). The overexpression of lincNMR rescued the proliferation defect caused by lincNMR depletion attesting to its specificity (Fig. 1c). Furthermore, a cell cycle analysis using flow cytometry confirmed an increase of cells in the G0/G1 phase of the cell cycle after depletion of lincNMR in multiple cell lines (Figs. 1d and Supplementary S2f, g).
The arrest of cells in the G0 / G1 phase prompted us to evaluate the induction of senescence. Depletion of lincNMR triggered senescence in three liver cancer cells with two independent siPOOLs as evident by β-GAL-positive blue cells in SA-β-GAL assay (Fig. 1e, f). The induction of senescence was supported by the induction of the pro-inflammatory cytokines IL-1a and IL-1b, which are bona fide markers of the senescenceassociated secretary phenotype (SASP) as well as the senescenceassociated proteins EDN and IGFBP7 (Fig. 1g). The induction of senescence was largely independent of the expression of p53 or pRB ( Supplementary Fig. 2h) and also did not have gross consistent effects on p53 or pRB signaling ( Supplementary  Fig. 2i). LincNMR knockdown neither had a consistent impact on apoptosis ( Supplementary Fig. 3a) nor did its overexpression affect proliferation in cancer cells ( Supplementary Fig. 3b). Data shown are normalized to si-Neg Ctrl siPOOL transfected with empty vector pcDNA3.1 (n = 5). Significance was calculated by paired, two-tailed t test with *P < 0.05. d Silencing of lincNMR with 10 nM siPOOLs induces cell cycle arrest in the G0/G1 phase shown by flow cytometry 72 h post transfection in HLE cells (n = 3). Analysis was performed using FlowJo v10 software. e Representative microscopic images showing increased β-Gal activity indicating senescence in HLE cells 96 h post lincNMR knockdown with 10 nM siPOOLs (n = 3). The scale bar represents 100 µm. f Bar graph representing percent β-Gal-positive cells in multiple liver cancer cell lines after lincNMR depletion (n = 3). g Induction of senescence-associated secretary phenotype (SASP) markers IL-1a,IL-1b, EDN, and IGFBP7 determined by RT-qPCR at 72 h after lincNMR knockdown in HLE cells with 10 nM siPOOLs (n = 3). a-d, f, g Data represent mean, and error bars represent SEM. a, b, d, f, g Significance was calculated by unpaired, two-tailed t test with *P < 0.05; **P < 0.01; ***P < 0.001. a-c, g Data shown were normalized to negative control siPOOL. NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17007-9 ARTICLE NATURE COMMUNICATIONS | (2020) 11:3214 | https://doi.org/10.1038/s41467-020-17007-9 | www.nature.com/naturecommunications entities, lincNMR was also significantly induced between tumor and normal tissues across multiple cancer types, including breast invasive carcinoma, lung adenocarcinoma, lung squamous cell carcinoma, bladder urothelial carcinoma, and cervical squamous cell carcinoma (Fig. 2a). In addition, lincNMR was also expressed in a broad panel of 73 cell lines derived from different tumor entities or normal tissues ( Supplementary Fig. 3c).
lincNMR depletion leads to decreased tumor growth in vivo. To test the impact of lincNMR knockdown on tumor growth in vivo, we turned to the chicken chorioallantoic membrane (CAM) model for in vivo xenograft experiments following the ethical responsibility aiming to replace, reduce or refine (3R) the use of animal models for research purposes. Knockdown efficiency ( Supplementary Fig. 3d) and the presence of the growthinhibitory phenotype (Fig. 2b) were confirmed for the duration of the CAM assay until 168 h (day 7). After seeding lincNMRsilenced or control-treated cells onto the CAM, tumors were harvested, measured, and weighed. Tumors derived from lincNMR-depleted cells were significantly smaller in size than tumors derived from the control group (Fig. 2c). The tumor . d Quantification of tumors harvested from CAM assay: decreased tumor weight in lincNMR-depleted tumors compared to the control group. a, b, d Significance was calculated by unpaired, twotailed t test with *P < 0.05; **P < 0.01; ***P < 0.001.
weight was also significantly reduced in lincNMR-depleted tumors in comparison with the control group (Fig. 2d).
lincNMR binds to YBX1 protein, which regulates cell survival.
In order to gain insight into the molecular function and protein interaction partners of lincNMR, we performed in vivo RNA antisense purification (RAP-MS) 41,42 . After cross linking RNA and protein in vivo, we used biotinylated DNA oligos complementary to the lincNMR sequence to pull down lincNMR and its associated protein-binding partners from HLE cell lysates ( Supplementary Fig. 4a). Pull down efficiency was confirmed by RT-qPCR with three different lincNMR amplicons, whereas PPIA and GAPDH mRNAs were used as negative controls ( Supplementary Fig. 4b). Mass spectrometry analysis of proteins cross linked to and pulled down with lincNMR compared with non-cross linked lysates identified 701 proteins in total (Fig. 3a), with 48 proteins or isoforms enriched at least twofold and with an adjusted P-value < 0.001 (Supplementary Data 2). To further select relevant interaction partners linked to HCC, we analyzed their correlation to survival in HCC (Fig. 3b). In total, 22 candidates were significantly associated with the overall survival of liver cancer patients (Supplementary Data 2). YBX1 emerged as a potential binding partner with strongest and significant correlation to survival in liver cancer patients ( Fig. 3b; Supplementary Fig. 4c). In addition, we searched for predicted RBP binding sites in the target genes using RBPmap 43 and found three YBX1 sites and two SRSF3 sites (P < 0.001) in the lincNMR transcript. Mutation of the first of the three predicted consensus YBX1 sites in lincNMR ( Supplementary Fig. 4d) abrogated the proliferation rescue of wild-type lincNMR (Fig. 3c). UV-RIP followed by RT-qPCR validated the interaction between lincNMR RNA and YBX1 protein. After immunoprecipitation of Flag-HA-tagged YBX1 ( Supplementary Fig. 4e), lincNMR was enriched in the YBX1 pulldown compared with the negative control SRSF4 (Fig. 3d).
In addition, biotinylated lincNMR RNA pulled down endogenous YBX1 protein in an in vitro RNA-affinity purification ( Supplementary Fig. 4f).  Fig. 3 lincNMR directly binds to and controls YBX1. a Volcano plot depicting total candidates (n = 701) identified by lincNMR in vivo RNA antisense purification followed by mass spectrometry (RAP-MS). In total, 48 candidates highlighted in red were selected for fold enrichment (FC > 2) and adjusted Pvalue of enrichment (P < 0.001). XL: cross link. Significance was calculated by moderated t test. b Interaction candidates identified by in vivo RAP-MS selected for fold enrichment in cross linked over non-cross linked samples (FC > 2), adjusted P -value of enrichment (P < 0.001), and sorted for log-rank P -value in liver cancer patient survival (based on TCGA data). c While the overexpression of wild-type (WT) lincNMR partially rescues the decreased cell proliferation caused by lincNMR knockdown, a lincNMR mutant with a mutation in the first (A) of three (A, B, C) putative YBX1-binding sites is not capable of rescuing cell proliferation in HLE cells at 72 h post lincNMR depletion (assayed by BrdU incorporation, n = 4). d Determination of RNA pull down efficiency in UV-RIP by RT-qPCR validating the interaction between YBX1 and lincNMR in HLE cells (n = 4). Data shown are normalized to SRSF4 as a negative control gene and control vector. e Silencing YBX1 with 10 nM siPOOL imparts a proliferative disadvantage to HLE cells determined 72 h post transfection (n = 4). Data shown are normalized to negative control siPOOL. f Dual-luciferase assay for the transactivational activity of YBX1, which unravels the inhibition of YBX1 by lincNMR silencing. Data shown are the normalized ratios of YBX1-dependent Firefly luciferase activity divided by Renilla luciferase used for standardization after depletion of lincNMR with 10 nM siPOOLs in HLE and FLC-4 cells (n = 4). g Overexpression of YBX1 partially rescues the proliferation defect induced by lincNMR depletion in HLE cells assayed by BrdU cell proliferation assay (72 h depletion of lincNMR and 66 h overexpression of YBX1). Data shown are normalized to empty vector transfected with negative control siPOOL (n = 5). c-g Data represents mean and error bars represent SEM. Significance was calculated by unpaired, two-tailed t test with *P < 0.05; **P < 0.01; ***P < 0.001. NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17007-9 ARTICLE NATURE COMMUNICATIONS | (2020) 11:3214 | https://doi.org/10.1038/s41467-020-17007-9 | www.nature.com/naturecommunications In addition to the significant association of high YBX1 expression with survival in hepatocellular carcinoma patients ( Supplementary Fig. 4c), knockdown of YBX1 impaired cell proliferation by 50% (Fig. 3e) resembling the effect of lincNMR knockdown. LincNMR levels were decreased by 49% after YBX1 depletion ( Supplementary Fig. 4g). Vice versa, YBX1 protein levels were decreased by~20% when lincNMR was silenced by two independent siPOOLs ( Supplementary Fig. 4h). We also found a significant positive correlation between lincNMR and YBX1 mRNA in HCC patient samples ( Supplementary Fig. 4i).
The direct interaction of lincNMR and YBX1 protein prompted us to assess the activity of YBX1 upon lincNMR depletion using luciferase assays for YBX1 transactivational activity. Knockdown of lincNMR significantly decreased the YBX1 activity in two independent liver cancer cell lines (Fig. 3f). In turn, overexpression of YBX1 partially rescued the proliferation deficit caused by lincNMR knockdown (Fig. 3g). Together, these data show that lincNMR interacts with and regulates YBX1, YBX1 mimics the impact of lincNMR on cell proliferation, and the regulation of lincNMR by YBX1 generates a feedforward loop leading to the correlation of expression in liver cancer.
lincNMR and YBX1 share target genes in nucleotide metabolism. To investigate the impact of lincNMR on the cellular proteome, we employed a triple-label stable isotope labeling by amino acids in cell culture (SILAC) approach (schematic Supplementary Fig. 5a). SILAC ratios were used for analysis comparing siPOOLs targeting lincNMR to negative control siPOOL (si-lincNMR-A/si-Neg Ctrl = M/L and si-lincNMR-B/si-Neg Ctrl = H/L). A correlation analysis served as a quality control on the complete dataset: a significant correlation was observed across three biological replicates and in between both siPOOLs targeting lincNMR with an average correlation coefficient of R = 0.74 ( Supplementary Fig. 5b). Notably, the correlation between replicates of the same siPOOL was slightly higher (range 0.69 -0.87) than between two different siPOOLs (range 0.43-0.54), which may point towards sequence-mediated off-target effects, but still indicates the high specificity and correlation (P < 10 −100 ) of the complex siPOOLs of 30 siRNAs compared with an average correlation of only 0.07 between individual siRNAs or lower complexity pools 44 .
For 242 candidates (P-value <0.001) deregulated by both lincNMR-targeting siPOOLs in the same direction, a Panther Overrepresentation Analysis with FDR correction was performed using Reactome and PANTHER pathway datasets. This revealed a significant enrichment of key terms like "G1 / S transition" and "Cell Cycle Checkpoints" as well as "De novo pyrimidine deoxyribonucleotide biosynthesis" matching the phenotype observed after lincNMR depletion ( Supplementary Fig. 5c). Among the strongly downregulated proteins upon lincNMR , and TYMS (si-lincNMR-A = −43%, si-lincNMR-B = −57%) and other key enzymes in nucleotide metabolism pathways (Fig. 4a, b), which were also part of the enriched gene ontology terms (Supplementary Fig. 5c). This decrease of RRM2, TK1, and TYMS proteins after lincNMR depletion was validated by western blotting (Fig. 4c) in good accordance with the independent triple-label SILAC-MS approach (Fig. 4d). In addition, a decrease of RRM2, TK1, and TYMS mRNAs was also observed after lincNMR depletion with both siPOOLs (Supplementary Fig. 5d). Since we identified lincNMR as a regulator of the transcription factor YBX1, we tested whether also YBX1 depletion would affect RRM2, TK1, and TYMS expression levels. Indeed, loss of YBX1 induced a significant decrease of RRM2, TK1, and TYMS protein levels (Fig. 4e, f) establishing the lincNMR-YBX1-RRM2 / TK1 / TYMS axis. Depletion of RRM2, TK1, and TYMS with siPOOLs significantly impaired cell proliferation mimicking the phenotype observed after lincNMR depletion (Fig. 5a) with a knockdown efficiency in the range of 95-99% ( Supplementary Fig. 5e). Similarly, the RRM2 inhibitor triapine 45 induced an arrest of cell cycle progression in the G1 phase similar to lincNMR knockdown (Fig. 5b).
Knockdown of YBX1, TK1, or TYMS also significantly induced cellular senescence (Fig. 5c, d), further phenocopying the lincNMR knockdown. Knockdown of lincNMR, YBX1, TK1, and TYMS also significantly inhibited the colony-formation capacity of liver cancer cells (Fig. 5e, f). After the knockdown of RRM2, no cells were left precluding quantification of senescence or colony formation.
Next, we found RRM2, TK1, and TYMS to be strongly and significantly induced by about ten-, six-and fourfold, respectively, in liver cancer patient datasets from TCGA ( Supplementary  Fig. 6a-c). Furthermore, all three target genes, RRM2, TK1, and TYMS were significantly associated with poor survival in HCC patients ( Supplementary Fig. 7a-c). The lincNMR expression level significantly positively correlated with RRM2, TK1, and TYMS mRNA levels in HCC patient samples ( Supplementary Fig. 7d-f), further corroborating the strong link between lincNMR, these three regulators of nucleotide metabolism, and liver cancer.
Also, the expression of YBX1 mRNA significantly positively correlated with RRM2, TK1, and TYMS mRNA expression in 374 human hepatocellular carcinoma samples (TCGA, Fig. 6a-c). The YBX1 protein interacted with the promoter regions of the RRM2, TK1, and TYMS genes as revealed by chromatin immunoprecipitation (ChIP, Fig. 6d). Lastly, we performed luciferase assays with the promoter regions of RRM2, TK1, and TYMS fused to firefly luciferase. Knockdown of lincNMR or YBX1 significantly decreased the luciferase activity for all three promoters (Fig. 6e), further corroborating the regulatory interaction of lincNMR and YBX1 with RRM2, TK1, and TYMS.
Depletion of lincNMR leads to reduced dNTP levels. Since dNTP-synthesizing enzymes were downregulated by the knockdown of lincNMR, we further investigated whether the levels of dNTPs were accordingly affected. All four dNTPs, dATP, dCTP, dGTP, and dTTP, were significantly downregulated after lincNMR depletion with two independent siPOOLs in two independent cell lines (Fig. 7a, b). The knockdown of YBX1 recapitulated this phenotype also leading to decreased dNTP levels in two cell lines (Fig. 7c, d).
dNTPs rescue the lincNMR proliferation phenotype. Importantly, supplying exogenous dNTP pools by bathing the cells rescued the impact of lincNMR knockdown on cell proliferation by preventing this phenotype (Fig. 7e, f). This effect was dosedependent and statistically significant, illustrating the essential role of nucleotide metabolism in the pro-proliferative function of lincNMR.
In summary, their regulation, their association with survival, and their correlation of expression links lincNMR, YBX1, RRM2, TK1, and TYMS to liver cancer and to each other, respectively. LincNMR affects cell viability, proliferation, senescence, colony formation, and tumor growth in vivo. The interactor YBX1 and its targets RRM2, TK1, and TYMS mimic the phenotypes of lincNMR. At the molecular level, these data suggest a model (Fig. 7g) in which the lncRNA lincNMR binds to YBX1, increases its activity resulting in the upregulation of the enzymes RRM2, TK1, and TYMS, which mediate an increase in nucleotide metabolism. In turn, lincNMR depletion causes a decrease of dNTPs leading to cellular senescence.

Discussion
In this study, we identified a lncRNA -lincNMR, a first lncRNA to regulate nucleotide metabolism in cancer cells. Silencing of lincNMR leads to impaired cell proliferation and colony formation, induction of a G0 / G1 phase cell cycle arrest, deregulation of nucleotide metabolism, and eventual induction of senescence in multiple cancer cell lines. LincNMR overexpression rescues the proliferation defect pointing toward the specificity of the knockdown. Cellular senescence is defined by an irreversible cell growth arrest. While cells undergoing senescence do not replicate, they remain metabolically active and undergo alterations in cell metabolism pathways, including nucleotide, glucose, mitochondrial, and lipid metabolism 46 .
Nucleotide pools are essential for a multitude of biological processes, and their synthesis is carefully regulated during cell proliferation. dNTP pools are synthesized de novo or via the salvage pathway 47 . The rate-limiting step in the synthesis of dNTP pools is the reduction of ribonucleoside di-or tri-phosphates (NDPs / NTPs) to deoxyribonucleotide di-or tri-phosphates (dNDPs / dNTPs) by ribonucleotide reductase (RNR) 48,49 .  Deregulation of nucleotide metabolism has been reported to play a pathogenic role in various diseases, including cancer 50,51 .
In our triple-label SILAC-MS, knockdown of lincNMR leads to strong downregulation of key enzymes essential for dNTP biosynthesis -RRM2 (ribonucleotide reductase subunit 2), TK1 (thymidine kinase), and TYMS (thymidylate synthetase) among other cell cycle regulators and proliferation markers. These are also induced in HCC, associated with poor survival, and correlated with lincNMR expression, and their knockdown phenocopies the effect of lincNMR on cell proliferation, senescence, and colony formation. f Quantification of colony formation after knockdown represented relative to the negative control siPOOL (n = 3). #: Following RRM2 knockdown, no colonies were growing (0) and hence, no statistical significance could be calculated. a, b, d, f Data represent mean, and error bars represent SEM. Significance was calculated by unpaired, two-tailed t test with *P < 0.05; **P < 0.01; ***P < 0.001.
Consequently, all four dNTPs are depleted upon lincNMR knockdown, consolidating its role in the nucleotide metabolism axis. Our findings are in line with previous studies, finding levels of all four dNTPs significantly decreased during OIS 50-52 due to a suppression of RRM2 as a driver and not as an effect of cell cycle exit 51  Significance was calculated by unpaired, two-tailed t test. b YBX1 mRNA expression significantly and positively correlates with TK1 mRNA expression in n = 374 hepatocellular carcinoma patient samples (TCGA LIHC, indicated: log2 expression). Significance was calculated by unpaired, two-tailed t test. c YBX1 mRNA expression significantly and positively correlates with TYMS mRNA expression in n = 374 hepatocellular carcinoma patient samples (TCGA LIHC, indicated: log2 expression). Significance was calculated by unpaired, two-tailed t test. d The binding of YBX1 to the promoters of RRM2, TK1, and TYMS was determined by chromatin immunoprecipitation (ChIP) in HLE cells. Bar graph represents qPCR data after YBX1 ChIP compared with IgG as a negative control (n = 3). Data represent mean, and error bars represent SEM. Significance was calculated by unpaired, two-tailed t test where *P < 0.05; **P < 0.01; ***P < 0.001. e Luciferase assays show decreased transactivation of RRM2, TK1, and TYMS promoter regions after lincNMR or YBX1 silencing with 10 nM siPOOLs in HLE cells (n = 4). Data shown are ratios of firefly luciferase divided by Renilla luciferase used for standardization. Data represent mean, and error bars represent SEM. Significance was calculated by unpaired, two-tailed t test where *P < 0.05; **P < 0.01; ***P < 0.001.   4). Data shown represent the results of the cell proliferation assay normalized to negative control siPOOL with the respective dNTP concentration. g Proposed model of lincNMR molecular mechanism. a-d, f Data shown are normalized to negative control siPOOL. Data represent mean, and error bars represent SEM. Significance was calculated by unpaired, two-tailed t test with *P < 0.05; **P < 0.01; ***P < 0.001. that senescence induced by RRM2 depletion is independent of p53 and pRB 51 .
Importantly, bathing cells in exogenous dNTP pools dosedependently rescues the proliferation phenotype caused by lincNMR knockdown in two liver cancer cell lines, which is in good accordance with previous studies stating that increasing dNTP levels by RRM2 overexpression or exogenous nucleoside supply overcomes aberrant DNA replication, DNA damage, and senescence induced by oncogenic RAS or BRAF 51,52 . Ectopic coexpression of TYMS and RRM2 also suppresses OIS in normal human fibroblasts 52,53 . TYMS and RRM2 are suppressed in c-MYC-depleted melanoma cells undergoing OIS, and this senescent effect is rescued by overexpression of TYMS and RRM2 or by addition of deoxyribonucleosides 54 .
To gain an insight into the molecular mechanism, we identified YBX1 as a direct binding partner of lincNMR in vivo and in vitro. YBX1 is a transcription factor and also binds to lncRNAs linked to cancer [55][56][57][58] . YBX1 drives tumorigenicity and invasiveness of melanoma cells and its expression represents a negative prognostic factor in primary melanoma patients 59 . YBX1 expression correlates with poor outcomes in breast cancer patients 60 . Accordingly, we find high YBX1 expression correlated with poor overall survival in liver cancer, as well. LincNMR and YBX1 mRNA expression significantly correlate in liver cancer patient samples as well as with RRM2, TK1, and TYMS mRNA expression. Importantly, YBX1 depletion mimics the lincNMR depletion phenotype in decreasing cell proliferation, inducing senescence, diminishing colony formation, affecting RRM2, TK1, and TYMS levels, decreasing dNTP levels and YBX1 also partially rescues the growth-inhibitory effect of lincNMR knockdown. YBX1 binds to the promoter regions of RRM2, TK1, and TYMS. LincNMR controls the transactivational activity of YBX1-hence, we report a role for the lincNMR-YBX1 axis in regulating nucleotide metabolism in liver cancer cells. Nonetheless, future studies will unravel whether lincNMR primarily acts via recruiting, regulating, or activating YBX1, and these data also do not exclude that lincNMR may have additional functions and relevant interactors.
While the findings that lincNMR RNA interacts with YBX1 protein, that it shares the same target genes RRM2, TK1, and TYMS, that lincNMR, YBX1, RRM2, TK1, and TYMS share the same loss-of-function phenotype regarding cell proliferation, senescence, and colony formation, that YBX1 expression or exogenous dNTPs rescue the proliferation defect caused by lincNMR depletion, and that lincNMR and YBX1 depletion reduces the transactivation of RRM2, TK1, and TYMS promoter fragments in luciferase assays, arguing in favor of a direct effect of the lincNMR-YBX1 axis, future research may also investigate secondary effects by cell cycle disturbance or by an impact on YBX1 regulation.
The copy number with a conservatively approximated minimum range of two to seven copies per cell on average classifies lincNMR as a moderately expressed lncRNA while other functional lncRNAs show much lower copy numbers 61 . On the one hand side, this may already be sufficient to recruit YBX1 to specific loci in the genomeon the other hand, this is likely an underestimation since the comparison is necessarily done to plasmid DNA, so the efficiency of RNA isolation and reverse transcription is assumed to be quantitatively complete. Furthermore, the copy number of lincNMR could vary between different cell states (e.g., cell cycle, heterogeneity) and hence reach higher numbers for activity.
Depletion of lincNMR not only reduces proliferation in liver cancer but also impairs proliferation in multiple breast and lung cancer cell lines. In addition, it is overexpressed in multiple cancer entities like liver, lung, breast, bladder, and cervical cancer making it a likely broadly relevant oncogenic lncRNA. Xenograft experiments using lincNMR-depleted cells in the CAM model reveal the impact of lincNMR on tumor size in vivo potentially implicating it as a therapeutic target in the future.
Notably, targeting nucleotide metabolism via ribonucleotide reductase inhibitors has been identified as a promising therapeutic strategy in multiple cancer types. Ribonucleotide reductase (RNR) inhibitors show promise in the clinic for treating multiple cancer types with some even serving as a first-line cancer treatment. Gemcitabine is the first nucleoside analog clinically approved, and it continues to be a frontline therapy against pancreatic, bladder, and lung cancer 62 . Clofarabine is the second approved drug targeting refractory pediatric leukemia 63,64 . A combination therapy with gemcitabine, clofarabine, and carboplatin significantly improves progression-free survival of patients with platinum-sensitive recurrent ovarian cancer 65 . Another RNR inhibitor, hydroxyurea (HU), is used in treatment of AML, CML, and glioblastomas [66][67][68] .
Overall, our study identified a tumor-promoting lincRNA -lincNMRand unveils its mechanism along a YBX1-RRM2 / TK1 / TYMS axis in regulating nucleotide metabolism and governing the cancer cell fate between proliferation and senescence.

Methods
Cell culture. Liver cancer cell lines (HLE, HLF, FLC-4, and SNU-387) used in this study were kindly provided by Dr. Kai Breuhahn (Institute of Pathology, University Heidelberg, Heidelberg Germany). Lung cancer cell lines (A549, NCI-H1299, and NCI-H460) and breast cancer cell lines (MCF-7, KPL-1, and T47D) were purchased from ATCC. Liver cancer cells were cultured in RPMI-1640 (Sigma-Aldrich, R8758) with 10% FBS. Lung and breast cancer cell lines were cultured in DMEM (Sigma-Aldrich, D5671) supplemented with 10% FBS. All cell lines were cultured in a cell culture incubator at 37°C and 5% CO 2 without addition of any antibiotics. Cell lines were periodically tested at a 3-month interval for mycoplasma contamination using a PCR-based detection kit (PromoCell, PK-CA91-1048). All cell lines used in this study were verified using cell authentication services from Multiplexion, Heidelberg, Germany 69 .
siPOOL and plasmid transfections. siPOOLs were obtained as complex pools of 30 siRNAs targeting the same gene from siTOOLs Biotech GmbH, Martinsried, Germany, to minimize off-target effects 33 . siPOOLs were reverse transfected using Lipofectamine RNAiMAX transfection reagent (Life Technologies, 133778150) with RPMI-1640 medium. Plasmids were transfected using a forward transfection protocol by using TurboFect transfection reagent (Fischer Scientific, R0531) and Opti-MEM (Gibco, 31985054). Transfections were performed according to the manufacturer's recommendations. Sequences of all siPOOLs used in this study are provided in Supplementary Data 3.
The double knockdown of TP53 and pRb was performed as described 70 . siPOOL sequences are listed in Supplementary Data 3.
RNA isolation, reverse transcription, and quantitative PCR. Cells were lysed in Trizol (Sigma, T9424-200ML), RNA isolation and DNase I (Roche, 4716728001) digestion were performed as per the manufacturer's instructions. RNA isolation for in vivo RAP-MS and UV -RIP qPCR experiments was performed as per the manufacturer's protocol using the miRNeasy Mini kit (Qiagen, 217004), and DNase digestion was performed using the TURBO DNAfree kit (Life Technologies, AM1907).
In total, 1 µg of RNA was reverse transcribed using random hexamer primers with RevertAid reverse transcriptase (ThermoFisher, EP0442). cDNA was diluted 1:40 with DNase-and RNase-free water, and 4 µl was used in a qPCR reaction. RT-qPCR was performed using Power SYBR Green PCR Master Mix (ThermoFisher, 4367659) in an Applied Biosystems StepOne Plus thermal cycler with holding stage of 95°C for 10 min followed by 40 cycles of 95°C for 15 s and 60°C for 30 s. Normalization was performed with PPIA, GAPDH, and / or SRSF4 as internal reference controls as indicated. Data were analyzed using StepOne Software v2.3. Primer sequences used in this study are provided in Supplementary Data 4.
Protein isolation, protein quantification, and western blot. Cells were briefly washed with 1× PBS and lysed in 200 µl RIPA buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% Triton X-100, 0.5% Na-Deoxycholate, 0.1% SDS, 1% DTT) supplemented with 1× Protease (Roche, 4693132001) and phosphatase inhibitors (Sigma-Aldrich, 4906837001) on ice for 30 min with intermittent mixing. The lysate was centrifuged at 17,000 g for 10 min at 4°C for clearing. The supernatant was transferred to new tubes and flash-frozen until further analysis. Protein quantification was performed using the Pierce BCA Protein Assay Kit (Thermo Fischer, 23224) as per the manufacturer's instruction. Samples were boiled for 10 min at 95°C in 4× Laemmli Buffer (0.25 M Tris pH 6.8, 20% glycerol, 10% SDS, 355 mM 2-mercaptoethanol, 0.002% bromophenol blue) before loading onto 10% SDS-PAGE gels. Proteins were transferred to nitrocellulose membranes using transfer buffer (0.25 M Tris-base, 1.92 M Glycine, 1% SDS with 20% methanol) at 120 V for 90 min. The membrane was blocked with 5% milk in 1× TBS-T (247 mM Tris, 1.37 M NaCl, 26.8 mM KCl, 0.05% Tween-20) for 30 min. Blots were incubated with the primary antibodies in 5% milk in 1× TBT-T overnight at 4°C with indicated dilutions. After incubation with primary antibody, blots were washed five times with 1× TBS-T for 5 min at room temperature on a shaker. Next, blots were incubated for 1 h at room temperature with respective anti-mouse or anti-rabbit HRP-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories, 115-035-003 and 111-035-144) were used at a dilution of 1:2500. After incubation with secondary antibody, blots were washed again for five times with 1× TBS-T for 5 min at room temperature on a shaker. Membranes were developed using Supersignal Pico (Fisher Scientific, 34580). Images were acquired on the Intas ChemoCam Imager, and signal quantification was performed using LabImage 1D software. A list of antibodies and dilutions used is provided in Supplementary Data 5.
Cell viability. CellTiter-Glo Luminescent Cell Viability Assay (Promega kit, G7572) was performed 72 h after knockdown with the respective siPOOLs. At the time point for the assay, growth medium was removed from the cells using multichannel pipette and 60 µL of 1:4 CellTiter-Glo reagent: 1× PBS was added to the cells. Plate was incubated for 15 min at room temperature in the dark using an orbital shaker. After the incubation, chemiluminescence was measure using luminometer (Fluoroskan Ascent FL, Thermo Scientific). Data obtained were normalized to siPOOL-negative control.
Rapid amplification of cDNA ends (RACE). The total RNA (treated with DNase I) from HLE cells was used for first-strand cDNA synthesis. The SMARTer RACE cDNA Amplification Kit (Clontech, 634923) was used to perform 5′-and 3′-RACE analysis according to the manufacturer's instructions. The gene-specific primers used for RACE are listed in Supplementary Data 6. Subcellular fractionation. Subcellular fractionation was performed in HLE cells to separate chromatin-associated, nucleoplasmic and cytoplasmic fractions as previously described 71 . Fraction-specific controls were used to assess the quality of fractions obtained (chromatin fraction: NEAT1, MALAT1; nucleoplasmic fraction: RNU-1; cytoplasmic fraction: DANCR). Primer sequences are provided in Supplementary Data 4.
Cell proliferation. BrdU incorporation efficiency of cells was measured at 72 h post knockdown with respective siPOOLs using the Cell Proliferation Assay kit (Roche, 11669915001) as per the manufacturer's instructions.
Cell cycle analysis. Seventy-two hours post treatment with siPOOLs, the cells were trypsinized and fixed in 70% ethanol overnight at −20°C. Fixed cells were pelleted and washed with 1× PBS. After washing, cells were resuspended in 1× PBS containing 100 µg/ml RNase A (Sigma, 10109169001) and incubated at 37°C for 30 min. Post RNase treatment, the cells were stained with 100 µg/ml propidium iodide (Sigma-Aldrich, P4170). In total, 10,000 cells were acquired on BD FACS-Canto II Flow Cytometer, and data analysis was performed using FlowJo v10 software.
Triapine (Selleckchem, S7470), the RRM2 inhibitor 45 , was dissolved in DMSO, and HLE cells were treated with respective concentrations. Cell cycle analysis was performed as described, and data were analyzed using FlowJo software.
Apoptosis Caspase-Glo assay. The assay was performed using the Caspase-Glo assay kit from Promega (G8091) in a 96-well format. Cells were transfected with 10 nM of respective siPOOLs and incubated at 37°C with 5% CO 2 for 72 h. The supernatant was discarded and replaced with diluted (1:2 in 1× PBS) Caspase-Glo solution. The plate was incubated for 60 min at RT in the dark. Luminescence was measured with a FLUOstar Omega microplate reader (BMG Labtech).
Senescence-associated β-Gal assay. The Cells were reverse transfected with the respective siPOOLs on a six-well plate, and SA-β-Gal activity was detected 96 h post transfection. Cells were washed with 1× PBS and fixed at room temperature with 0.5% glutaraldehyde for 20 min. Cells were washed twice with 1× PBS supplemented with 1 mM MgCl 2 (pH 6.0) for 10 min on a rocker. In all, 2 ml X-Gal staining solution (1× PBS containing 1 mM MgCl 2 , 41 mg of potassium hexacyanoferrate (III), 52.5 mg of potassium hexacyanoferrate (II) trihydrate, 1 mg/ml X-Gal (5-bromo-4-chloro-3-indolyl-beta-D-galacto-pyranoside), pH 6.0) was added, and the dishes were sealed with parafilm and incubated overnight at 37°C. Next day, the cells were washed three times with distilled water, and microscopy pictures were taken using a 10x objective of a Zeiss Cell Observer microscope. For analysis, 100 cells were counted, and percent senescent cells per condition are depicted in the bar graph.
Chick chorioallantoic membrane (CAM) assay. Pathogen-free fertilized eggs were purchased from Valo Biomedia GmbH and incubated in an incubation oven with 60% humidity at 37°C. On day 8, lincNMR was knocked down with 10 nM siPOOLs using a forward transfection protocol described before in HLE cells. On day 9, eggs were windowed with an electric drilling tool. On the same day, in parallel, 1 × 10 6 transfected HLE cells were resuspended in 10 µL growth media, mixed with 10 µL Matrigel (Corning, 354262) and incubated for 10 min at 37°C and 5% CO 2 . The cell-matrigel mix was seeded onto the CAM, and the window was sealed for five days and incubated at 37°C with 60% humidity. During these days, eggs were observed, and dead or injured embryos were excluded from further experiments. On day 6, after seeding the cells on the CAM, chicken embryos were euthanized by a quick decapitation, and tumors were harvested from the CAM. Harvested tumors were processed, cleaned, and collected in ice-cold 1× PBS. Images of tumors from HLE cells transfected with negative control siPOOL, and two independent siPOOLs targeting lincNMR were taken. Tumors were weighed on a microscale.
Colony formation assay. Colony formation assays were performed as described in ref. 72 . In brief, HLE cells were reverse transfected with 10 nM of respective siPOOLs. Twenty-four hours later, 500 cells were reseeded into six-well plates to allow formation of colonies for 14 days. After the incubation, cells were fixed with 6% glutaraldehyde and stained with 1% crystal violet solution. After subsequent washes, plates were allowed to dry at room temperature, and colonies were counted.
In vivo RNA antisense purification (in vivo RAP-MS). Biotinylated DNA oligos complementary to lincNMR sequence were ordered from IDT, sequences are available in Supplementary Data 7. One billion HLE cells were used per pulldown per condition per biological replicate. In vivo RAP-MS was performed as per the protocol described previously 41 .
Generation of plasmids and mutagenesis. Gateway entry vectors were obtained from the DKFZ plasmids and clone repository. LincNMR-001 was amplified using the primers listed in Supplementary Data 10. Gateway LR reaction was performed with 50-150 ng of entry vector using LR Clonase II (Thermo, 11791020) as per the manufacturer's instructions into the gateway destination vector pFRT-Flag/HA. Mach1 cells were used for transformation. Mini-Prep was performed using the NucleoSpin ® Plasmid kit (Macherey & Nagel, 740588.250). Midi-Prep was done using the PureLink™ HiPure Plasmid kit (Invitrogen, K210004). Services from Eurofins genomics/GATC were used for sequencing with CMV.for CGCAAATGG GCGGTAGGCGTG and BGH.rev TAGAAGGCACAGTCGAGG primers. Finally, cells were transfected with respective plasmid and empty vector pFRT-Flag-HA-ΔCmR-ΔccdB as a control plasmid. Overexpression was confirmed by western blot using anti-Flag-M2 or anti-HA antibody (Supplementary Data 5).
Mutagenesis of the predicted high-confidence YBX1-binding sites (RBPmap) in the lincNMR transcript was performed using Phusion DNA polymerase as per supplier's instructions (NEB, M0530S). Primers used for performing PCR are listed in Supplementary Data 10.
UV cross linking RNA immunoprecipitation (UV-RIP) assay. Cell Seeding, Transfection, UV Cross linking, harvesting: on day 1, 4 × 10 6 HLE cells were seeded onto a 15-cm dish. On day 2, 2 µg of the respective constructs (Empty vector plasmid pFRT-Flag-HA-ΔCmR-ΔccdB or pFRT-Flag/HA-YBX1) were transfected using TurboFect transfection reagent with a forward transfection protocol. On day 4, cells were UV cross linked at a wavelength of 254 nm using 0.8 J/cm 2 (instrument setting: 8000 × 100 µJ/cm 2 ) and then lysed in high-strength cell lysis buffer (10 mM Tris-HCl pH 7.5, 500 mM lithium chloride, 0.5% dodecyl maltoside, 0.2% SDS, 0.1% sodium deoxycholate) supplemented with inhibitors cocktail containing SUPERase In (Thermo Scientific, AM2696), protease and phosphatase inhibitors. Lysate was passed through a syringe to break up the pellet, and DNase digestion was performed as per the manufacturer's instruction using the Turbo DNA-free Kit (Ambion, AM1907). Protein quantification was performed using the BCA reagent, and overexpression of YBX1 was confirmed by probing with anti-HA antibody on a western blot.
Pulldown: 1.5 µg of cell lysate was used to perform UV-RIP using anti-Flag magnetic beads, and IP was confirmed by probing for anti-HA using western blot. For UV-RIP, 150 µl of anti-Flag magnetic beads were used per pulldown. Beads were prewashed 5× using 1× TBS and resuspended in supplemented cell lysis buffer. In total, 5% lysate was removed for the Input fraction, and 150 µl of prewashed beads were added to the cell lysate and incubated for 1 h at 4°C with rotation. After the incubation, beads were magnetically separated. 50 µl of flowthrough was saved to confirm the depletion of YBX1, and the remaining flowthrough was discarded. Beads were washed 2× with low salt-wash buffer (0.1% SDS, 0.5% sodium deoxycholate, 0.5% NP-40, 0.01 M NaCl, 0.002 M KCl, 0.001 M Na 2 HPO 4 , 0.0001 M KH 2 PO 4 ) and then with high salt-wash buffer (0.1% SDS, 0.5% sodium deoxycholate, 0.5% NP-40, 0.05 M NaCl, 0.010 M KCl, 0.005 M Na 2 HPO 4 , 0.0005 M KH 2 PO 4 ) with rotation for 5 min per wash. Beads were magnetically separated and resuspended in 500 µl of RNase-free pure water. Beads were then separated into 20% (for protein extraction) and 80% (RNA isolation) to confirm the IP and lincNMR pulldown, respectively. Antibodies and magnetic beads used are listed in Supplementary Data 5.
Elution of RNA: RNA was eluted from the beads by reversal of UV cross linking in high salt buffer combined with Proteinase K digestion (160 µl of Proteinase K buffer (125 mM Tris-HCl pH 7.8, 62.5 mM NaCl, 12.5 mM EDTA) and 40 µl of Proteinase K) by incubating at 37°C with shaking for 30 min at 1000 rpm. Next, after addition of 500 µl of Trizol to the tube, the tube was vortexed for 10 s and stored at −20°C until ready for RNA isolation. RNA isolation, reverse transcription, and qPCR were performed as described above.
Elution of proteins: Captured, washed beads were boiled with 1× SDS loading buffer at 95°C for 10 min, and western blot was performed with input, supernatant, and flow-through samples to confirm the YBX1 pulldown.
In vitro lincNMR RNA-affinity purification. In vitro lincNMR RNA-affinity purification was performed as described previously 73 . The lincNMR sequence (1100 nt) was cloned into a pcDNA3.1 vector, which was further linearized using EcoRV and MluI digestion. As a negative control, the lncRNA HULC (560 nt) was used for comparison, which had been previously cloned 73 . The MEGAscript T7 Transcription Kit (Ambion, AMB13345) was used according to the manufacturer's instructions for in vitro transcription. In vitro transcribed biotinylated lncRNAs were used to pull down the interacting proteins from HLE cell lysate using streptavidin sepharose beads (GE Healthcare, 17-5113-01). After subsequent washing steps, proteins were eluted using wash buffer (20 mM HEPES pH 7.9, 300 mM KCl, 10 mM MgCl 2 , 0.01% NP-40, 1 mM DTT) supplemented with 50 μg/ml RNase A. Eluates were acetone precipitated and the pellet was washed twice with 80% ethanol. The washed pellet was dissolved in 2× SDS sample by heating at 95°C for 5 min. Western blot was performed to identify the RNA interacting proteins.
Cell culture, transfections, lysate preparation, protein quantification, LC-MS/ MS: After confirmation of label incorporation, isotopically labeled HLE cells were grown to 80% confluency. Cells were reverse transfected with the respective siPOOLs targeting lincNMR or the control siPOOL. Lysates were harvested 48 h post transfection. Cells were lysed in 200 µl RIPA buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% Triton X-100, 0.5% Na-deoxycholate, 0.1% SDS, 1% DTT) supplemented with protease and phosphatase inhibitors. In addition, 0.1% benzonase (Merck, Darmstadt, Germany) was added to digest nucleic acids. After incubation on ice for 1 h, cell lysates were cleared by centrifuging at 15,000 g, for 30 min at 4°C. The 2D Quant Kit (GE Healthcare, 806483356) was used to measure protein concentrations. YBX1 ChIP assay. Chromatin immunoprecipitation (ChIP) assay was performed as described before 74 to identify YBX1 interacting promoter regions. HLE cell lysates were used for the ChIP experiment using YBX1 antibody (Abcam, ab12148) and isotype control rabbit IgG antibody (Abcam, ab171870). ChIP pull-down DNA was analyzed using the qPCR and primers listed in Supplementary Data 8.
Dual-luciferase reporter assay. HLE and FLC-4 cells were reverse transfected with 10 nM of control siPOOL and siPOOLs targeting lincNMR on day 1. Control pRL-TK/pRL-SV40 reporter constructs and Y-Box-TATA-Luc were co-transfected using a forward transfection protocol on day 2. Cell culture medium was changed on day 3. Luciferase reporter assay was performed using a Dual-Luciferase Reporter Assay System kit (Promega, E1960) on day 4 as per the manufacturer's instruction, and luminescence was measured on Spectra Max M5e (Molecular Devices).
For the characterization of the YBX1 impact on RRM2, TK1, and TYMS promoter regions, genomic DNA from HLE cells was PCR-amplified using Phusion polymerase (NEB, M0530). The first round of PCR was performed using primers without restriction sites. After gel elution of PCR purified products, a nested PCR was performed using primers with restriction sites. Gel purified PCR products and empty pcDNA3.1 vector were digested using Xhol / NotI and ligation was performed using T4 DNA ligase (Thermo, EL0011). Primers used for creating the constructs are listed in Supplementary Data 9.
dNTP quantification. Cellular dNTPs were extracted from cells (HLE and FLC-4) transfected with the respective siPOOLs 72 h post transfection as per a previously published protocol 75 . The dried dNTPs were resuspended to proper volumes of water and added to the HIV-1 RT-mediated single dNTP incorporation reactions. The percent of primer extension were converted to the incorporated dNTP amounts, and the determined dNTP amounts were normalized by 1 million cells for comparison. Further, data were normalized to negative control siPOOLs. dNTP bathing and rescue assay. In total, 1250 cells were reverse transfected with the respective siPOOLs in a clear bottom 96-well plate. The desired concentration of extracellular dNTPs was added to the cells 24 h later. A cell proliferation assay was performed at 96 h post siPOOL transfection to assess the proliferation of the cells.
RNA copy number detection. 1 µg of RNA was used for reverse transcription with MaximaRT (Thermo Fischer EP0751) from respective cell lines (HLE, HLF, SNU-387 and FLC-4). 1:10 diluted cDNA was used for qPCR. Primers used for qPCR are listed in Supplementary Data 4. Plasmids containing lincNMR-001 sequence were linearized using BstBI digestion and gel purified. Serial dilutions were made in the range of 0.02 pg -5 pg. The copy number was calculated using the online tool https://cels.uri.edu/gsc/cndna.html.
Statistics and reproducibility. All experiments were performed at least in three biological replicates, and information about statistical tests used is detailed in the respective figure legends. Exact P-values are provided in the Source Data file. Numbers of replicates always refer to independent biological replicates.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Sequencing data from 5′-and 3′-RACE experiment supporting both lincNMR isoforms have been deposited at Genbank with accession numbers MK652436 and MK652437. Source data for figures shown in this study are available upon request if not available in Supplementary Data and in the attached source data files. The source data underlying Fig. 4c