A novel protein RASON encoded by a lncRNA controls oncogenic RAS signaling in KRAS mutant cancers

Mutations of the RAS oncogene are found in around 30% of all human cancers yet direct targeting of RAS is still considered clinically impractical except for the KRASG12C mutant. Here we report that RAS-ON (RASON), a novel protein encoded by the long intergenic non-protein coding RNA 00673 (LINC00673), is a positive regulator of oncogenic RAS signaling. RASON is aberrantly overexpressed in pancreatic ductal adenocarcinoma (PDAC) patients, and it promotes proliferation of human PDAC cell lines in vitro and tumor growth in vivo. CRISPR/Cas9-mediated knockout of Rason in mouse embryonic fibroblasts inhibits KRAS-mediated tumor transformation. Genetic deletion of Rason abolishes oncogenic KRAS-driven pancreatic and lung cancer tumorigenesis in LSL-KrasG12D; Trp53R172H/+ mice. Mechanistically, RASON directly binds to KRASG12D/V and inhibits both intrinsic and GTPase activating protein (GAP)-mediated GTP hydrolysis, thus sustaining KRASG12D/V in the GTP-bound hyperactive state. Therapeutically, deprivation of RASON sensitizes KRAS mutant pancreatic cancer cells and patient-derived organoids to EGFR inhibitors. Our findings identify RASON as a critical regulator of oncogenic KRAS signaling and a promising therapeutic target for KRAS mutant cancers.


INTRODUCTION
The human RAS gene family includes KRAS, HRAS and NRAS, all of which encode 21 kDa small GTPase proteins that are activated on the inner cell membrane and transduce extracellular growth stimuli signals to intracellular effector signaling cascades. 1 RAS proteins are binary molecular switches that cycle between an 'ON' state (GTP bound) and an 'OFF' state (GDP bound). The subtle balance between these two states is mainly controlled by guanine nucleotide exchange factors (GEFs), which convert RAS-GDP to RAS-GTP; and GTPase-activating proteins (GAPs), which convert RAS-GTP back to RAS-GDP through hydrolysis. 2 Activating RAS mutations are found in around 30% of all human cancers. 3,4 Among the three RAS genes, mutation frequency of KRAS in human cancer is considerably higher than that of HRAS and NRAS. For example, KRAS G12C mutation is detected in around 12%-15% of non-small cell lung cancer (NSCLC) patients; whereas more than 90% of pancreatic ductal adenocarcinoma (PDAC) patients bear KRAS mutations, especially the KRAS G12D and KRAS G12V subtypes. 5 These mutations are thought to impair both the intrinsic GTPase activity of KRAS-GTP and the rate of GAP-mediated GTP hydrolysis in KRAS-GTP, 6 thus leading to constitutively active oncogenic KRAS signaling.
Despite the critical role of RAS signaling in human cancers, direct targeting of the RAS protein has remained a great challenge for the past forty years. Alternative therapies against the upstream growth factor receptors such as EGFR have been applied in NSCLC and colon cancer, but those patients with RAS mutations were excluded. 7 Moreover, secondary resistance to EGFR inhibitors in patients with wild-type RAS is often driven by acquired RAS mutations. 8 On the other hand, targeting RAS downstream effectors, including RAF and MEK, showed limited benefits in RAS-mutant cancers. 9 Very recently, small-molecular compounds that could directly target KRAS G12C mutant through covalent binding have been approved for the treatment of NSCLC. [10][11][12][13][14] However, strategies for targeting other KRAS mutations such as KRAS G12D and KRAS G12V , which are the prevalent mutation subtypes in PDAC, are desperately needed. Here we report the discovery of a novel protein, termed RAS-ON (RASON), that plays a critical role in oncogenic RAS signaling and represents a vulnerable therapeutic target for mutant KRAS-driven cancers.

RESULTS
LINC00673 encodes a novel protein RASON Advances in genome-wide translatome lead to the identification of hundreds of functional peptides or proteins from noncoding regions of the genome, including long intergenic noncoding RNAs (lincRNAs), 5′ un-translational region (5′-UTR) and pri-micro RNAs. [15][16][17] We screened small open reading frame (sORF) in lincRNAs that could encode novel proteins in PDAC by performing ribosome-sequencing in the normal pancreatic ductal cell line HPNE and three PDAC cell lines BxPC-3, PANC-1 and AsPC-1. Certain Ribosome Protected Fragments (RPFs) were mapped to transcripts annotated as non-coding regions including 5′-UTR, 3′-UTR (of known protein-coding genes) and lincRNAs (Supplementary information, Fig. S1a). A total of 14,504 sORFs were identified with read counts ≥1, with an average length of 100 bp (Supplementary information, Fig. S1b). To identify active translation, the ribosome release score (RRS) and modified ORF score for each lincORF were calculated (Fig. 1a). Forty actively translated lincORFs were retained and subjected to differential analysis. Compared with HPNE cells, eight lincRNAs were found differentially translated in all three PDAC cell lines and LINC00673 ranked top of these candidates (Fig. 1b). Inspection of the LINC00673 transcript revealed an sORF in exon 1 that may encode a 108 amino acid protein (referred to as RASON hereafter), with differential RPFs between HPNE and PDAC cells (Fig. 1c). We also searched several published ribosome-sequencing datasets, RPFs of RASON were detected as well (Supplementary information, Fig. S1c). Using cohorts from TCGA and GTEx databases, we found that LINC00673 RNA was highly expressed in PDAC and several other cancer types (Supplementary information, Fig. S1d). We also enrolled twenty-four paired samples (tumor (T), adjacent normal tissue (N), and metastatic lymph node (M)) from eight PDAC patients and performed RNA-seq. A total of 834 differentially expressed lincRNAs were identified between N vs T, 959 between N vs M, and 13 between T vs M (Supplementary information, Fig. S1e). Among the eight differentially translated lincRNAs previously identified in PDAC cell lines (Fig. 1b), five were differentially expressed in N vs T and N vs M; LINC00673 also ranked top in these differentially expressed lincRNAs (Supplementary information, Fig. S1f).
To confirm the translation of RASON in cells, we generated a CRISPR/Cas9-induced FLAG tag insertion inside LINC00673 genomic DNA in HEK293T cells, right after the RASON sORF ( Fig. 1d; Supplementary information, Fig. S2a). Immunoblotting (IB) with anti-FLAG antibody confirmed the translation of a~14 KD protein (Fig. 1e). RASON is highly conserved among primates, whereas mouse RASON (78 aa) has a shorter C-terminus but shares a fair homology with human RASON on the N-terminus (Supplementary information, Fig. S2b), we thus generated a rabbit monoclonal antibody against RASON (1-14 aa, Supplementary information, Fig. S2c) and performed IB in 293T cells overexpressing LINC00673, RASON-ORF, RASON-ORF with mutated ATG start codon, or shRNA against LINC00673 (sh-LINC00673). RASON expression was elevated after transfection of either LINC00673 or RASON-ORF but not RASON-ORF-ATG mutant and was inhibited by sh-LINC00673 (Fig. 1f). RASON-specific peptide sequences were also identified by mass spectrometry (MS) analysis in LINC00673transfected 293T cells (Supplementary information, Fig. S2d), and endogenously in the AsPC-1 PDAC cell line (Supplementary information, Fig. S2e). Collectively, these data demonstrated the translation of RASON protein in human cells.
RASON is aberrantly overexpressed in PDAC and correlates with poor prognosis LINC00673 was identified as an oncogenic lncRNA in multiple human malignancies including PDAC. [18][19][20][21][22] In an independent PDAC patient cohort including 105 PDAC cancerous samples and paired adjacent normal pancreatic samples, we confirmed that LINC00673 is highly expressed in cancerous tissues and its expression positively correlates with tumor grade (Supplementary information, Fig. S2f, g). Consistent with this, RASON protein expression is also significantly higher in cancerous tissues (Fig. 1g, the same patient cohort with 47 additional PDAC tumor samples) and positively correlates with both tumor grade (Fig. 1h) and poor overall survival (Fig. 1i). RASON expression levels were determined by immunohistochemistry (IHC) staining using the anti-RASON antibody, the specificity of which was confirmed by the addition of a blocking peptide (Fig. 1j). In PDAC cell lines, RASON expression is high in AsPC-1 (KRAS G12D ) and PANC-1 (KRAS G12D ) cells, and relatively low in BxPC-3 (KRAS WT ) and Capan-2 (KRAS G12V ) cells (Fig. 1k). The aberrant overexpression of RASON in human PDAC was also confirmed by IB in paired cancerous/ adjacent normal tissue (>2 cm away) from 14 randomly selected PDAC patients (Fig. 1l). Taken together, these results suggested that RASON is an oncogenic protein and may serve as a prognostic marker for PDAC. To elucidate the molecular mechanisms underlying the protumor function of RASON, we performed RNA-seq in AsPC-1 and PANC-1 RASON-KO cells. Both DEG and KEGG analysis implied the involvement of RAS downstream signaling including the MAPK (RAF-MEK-ERK) and PI3K-AKT pathways (Supplementary information, Fig. S6d). Immunoblotting experiments showed that both pathways were inhibited by RASON KO in AsPC-1 and PANC-1 cells, but enhanced by RASON OE in BxPC-3 and Capan-2 cells (Fig. 2d). While total KRAS level was not affected (Fig. 2d), active KRAS-GTP level was downregulated in AsPC-1 and PANC-1 cells by RASON KO, and upregulated in BxPC-3 and Capan-2 cells by RASON OE (Fig. 2e), as measured by a pull-down assay using the RAS binding domain of the RAS effector protein RAF1 (GST-Raf1-RBD). Moreover, KRAS KD by shRNA in BxPC-3 and Capan-2 cells abolished RASON OE-induced activation of MAPK and PI3K-AKT pathways, further demonstrating that RASON exerts its biological function through RAS signaling (Fig. 2f). Notably, the downregulation of RAS signaling in AsPC-1 cells by RASON KO could be readily rescued by the overexpression of mouse Rason (Supplementary information, Fig. S6e), suggesting an evolutionarily conserved mechanism.
RAS signaling is critical for the proliferation and migration of normal cells. 23 To investigate whether RASON modulates RAS signaling in vivo, we generated Rason homologues KO mice by deletion of the entire sORF (Supplementary information, Fig. S7a). Rason −/− mice are viable; however, mouse embryonic fibroblasts (MEFs) generated from prenatal Rason −/− mice and pancreas tissues isolated from adult Rason −/− mice both exhibited significant downregulation of RAS signaling, as evidenced by reduced phosphorylation of RAF and AKT, as well as reduced level of KRAS-GTP ( Fig. 2g; Supplementary information, Fig. S7b, c). Together, the above results suggested that RASON is critical for tumor maintenance through activating RAS signaling.

RASON is essential for KRAS-driven tumorigenesis
To explore whether RASON is also involved in KRAS-driven tumorigenesis, we used the RAS-less MEF cell lines 24 stably transfected with KRAS G12D or KRAS G12V and tested the effect of Rason KO on KRAS-induced transformation of MEF cells. We generated two Rason KO clones for both KRAS G12D and KRAS G12V MEF cell lines using CRISPR/Cas9 (Fig. 3a). Rason KO almost completely diminished the tumorigenesis potential of KRAS G12D or KRAS G12V MEF cells in 3D anchorage-independent growth assays (Fig. 3b). In vivo subcutaneous tumor formation assays in nude mice also validated the critical role of RASON in KRAS G12D or KRAS G12V -induced transformation of MEF cells, as Rason KO almost Survival(%) B x P C -3 C a p a n -2 P a n c 0 2 . 0 3 C F P A C -1 P A N C -  Fig. S9). The KC model is a well-established model for studying early stages of PDAC development, i.e., the formation of pancreatic intraepithelial neoplasia (PanIN). Compared with KC mice, KCR mice exhibited a significant reduction in the formation of PanIN lesions of all grades (IA, IB, II, III) at both 6 and 9 months of age (Fig. 4b, c), suggesting RASON is involved in the early stage of KRAS-driven pancreatic tumor initiation. Next, to assess the role of RASON in PDAC formation in the classical LSL-Kras G12D ; Trp53 R172H/+ ; Pdx1 Cre model (KPC mice), we introduced Rason KO into the KPC background by CRISPR/Cas9-mediated mutation of the start codon in Rason ORF ( Fig. 4d; Supplementary information, S10; natural breeding was not feasible because of the proximity of Rason and p53 gene allele on the same chromosome) and generated LSL-Kras G12D ; Trp53 R172H/+ ; Pdx1 Cre ; Rason mut/mut mice (KPCR mice). KPC mice develop highstage PDAC in 3-6 months. Compared to KPC mice, KPCR mice exhibited a significant reduction in both tumor incidence (6/15 in KPC vs 2/10 in KPCR, 3 months) and tumor size (Fig. 4e, f). Notably, as a byproduct of KPCR mice, Rason mut/mut mice (with mutations in the start codon of Rason ORF) showed a significant reduction in KRAS signaling in primary MEF cells (Supplementary information, Fig. S10b-d), similar to that observed in Rason −/− mice (Fig. 2h). These data showed that RASON is required for KRAS-driven pancreatic tumorigenesis in mice.
Another genetic mouse model that is often used for investigating KRAS-driven tumorigenesis is the LSL-Kras G12D ; Trp53 R172H/+ (KP mice) lung cancer model. In this model, simultaneous KO of Rason and activation of Kras/p53 alleles in the lung was achieved by intratracheal administration of a lentivirus construct containing Cre, Cas9, and Rason sgRNA (Fig. 5a). Rason KO markedly inhibited lung tumor formation in KP mice, as monitored by micro-CT measurement of lung tumor incidence and size ( Fig. 5b-d), and significantly prolonged mouse survival (Fig. 5e), suggesting a common mechanism underlying the protumor function of RASON in KRAS mutant cancers. Altogether, the above results demonstrated an essential role for RASON in KRAS-driven tumorigenesis in vitro and in vivo. We next tested the binding between purified RASON protein and different KRAS mutants using Surface Plasma Resonance (SPR). RASON binds to KRAS G12D and KRAS G12V with remarkably high affinity (K d = 5.9E-8 M and 1.0E−7 M, respectively, Fig. 6b; Supplementary information, Fig. S14a). Notably, the K d value for KRAS-RASON binding is within the same order of magnitude as that between KRAS and RAS-GEFs or RAS-GAPs, 26,27 suggesting that RASON might function as GEF or GAP to interact with KRAS. To further confirm this binding, we performed Nuclear Magnetic Resonance (NMR) titration experiments. 15 N-labeled KRAS G12D was mixed with unlabeled human RASON protein at molar ratios ranging from 1:0 to 1:0.6, and a dose-dependent decrease of peak intensities were observed in the 2D 1 H-15 N heteronuclear single quantum coherence (HSQC) spectra of KRAS in the presence of RASON (Fig. 6c), suggesting a strong interaction between human RASON and KRAS G12D . Taken together, SPR and NMR data strongly support the direct binding between KRAS and RASON.

RASON inhibits GAP-induced GTP hydrolysis and maintains KRAS in the active state
In the classical KRAS regulation model, RAS-GEFs activate KRAS by promoting GDP-GTP exchange, while RAS-GAPs deactivate KRAS by catalyzing GTP hydrolysis; mutations in the G12 position partially inhibit GAP-mediated GTP hydrolysis, thus keeping KRAS in the active state. Since RASON promotes oncogenic KRAS signaling, we first tested the possibility of RASON being a new GEF for KRAS by performing in vitro nucleotide exchange assays. Surprisingly, however, RASON had no effect on the rate of either intrinsic or GEF-induced GDP-GTP exchange for KRAS G12D or KRAS G12V , suggesting that it is not a RAS-GEF (Supplementary information, Fig. S14b). To test the other side of the coin, i.e., RASON being an inhibitor of GAP activity, we carried out in vitro GTP hydrolysis experiments. As a result of the G12 mutations, KRAS G12D and KRAS G12V proteins exhibited a very slow rate of intrinsic GTP hydrolysis but still retained certain GAP-induced extrinsic GTP hydrolysis activity. However, the presence of RASON almost Fig. 1 RASON is overexpressed in human pancreatic cancer. a Scatter plot showing the ribosome release score (RSS) and the ORF score from main ORF (mORF, blue) and lincORF (purple) (x-axis, Log 10 RSS; y-axis, ORF score). The dashed lines represent the 95% percentile of RSS and ORF scores from mORF, which are set as the threshold to identify actively translated lincORF. Along each axis, all points are summarized using an overlaid density plot. b Venn diagram of differentially translated lincRNAs in three PDAC cell lines BxPC-3, PANC-1, and AsPC-1 compared to HPNE normal cells by Ribo-seq. c Upper, differential RPFs of RASON sORF between HPNE and PDAC cells; lower, diagram of the sORF encoding RASON. d Scheme of CRISPR/Cas9-induced FLAG-tag insertion after RASON sORF inside LINC00673 genomic DNA in 293T cells.  We next tested if RASON could also compete with GAP for KRAS binding and inhibit GAP-induced GTP hydrolysis in cancer cells. 28,29 IP experiments in HEK293T cells showed that increasing levels of RASON overexpression dose-dependently inhibited KRAS-NF1 binding (Supplementary information, Fig. S15a), suggesting that RASON can compete with NF1 (as the representative GAP) in cells. A prerequisite for this hypothesis is that RASON is expressed in cells at a sufficiently high level to bind KRAS protein and to compete against RASGAPs. To evaluate the abundance of RASON in cells, we used MS to quantify the respective protein levels of RASON, KRAS and NF1 in AsPC-1 and PANC-1 cells. Purified RASON, KRAS and NF1 proteins were first used to develop MS methods and generate a standard curve for each protein (Supplementary information, Fig. S16). Intracellular levels of each protein were then determined by detection and quantification of the respective signature peptides according to the standard curve. Cellular levels of RASON protein were found to be 10-20 folds higher than that of KRAS and 3-5 folds higher than NF1 (Fig. 6e), suggesting that RASON is expressed at a sufficiently high level to bind to KRAS and competes with NF1 in PDAC cells. Moreover, overexpression of NF1 in BxPC-3 and Capan-2 RASON-OE cells abolished the activation of KRAS signaling induced by RASON overexpression, further supporting our hypothesis (Fig. 6f) Fig. S17). These results are consistent with our hypothesis that RASON inhibits NF1-induced KRAS deactivation and maintains KRAS in a hyperactive state. Furthermore, RASON could also compete with another RAS-GAP, RASA2 (Supplementary information, Fig. S15b, c) for binding KRAS, implying a common biochemical mechanism. Taken together, these data demonstrated that RASON is a high-affinity KRAS binding protein that prevents both intrinsic and GAP-induced GTP hydrolysis, thus sustaining the hyperactive KRAS phenotype; a graphic model of KRAS regulation by RASON is shown in Fig. 6h.
RASON is a potential therapeutic target for PDAC Given the critical role of RASON in promoting oncogenic KRAS signaling, we investigated whether this could be utilized for that cetuximab itself had no effect on AsPC-1 tumor growth, whereas AAV9 delivery of RASON shRNA significantly inhibited xenograft tumor growth. Combination therapy of cetuximab and AAV9-delivered RASON shRNA maximally diminished tumor growth in nude mice (Fig. 7a, b; Supplementary information,  Fig. S18b). IB and IHC analysis confirmed target engagement (Supplementary information, Fig. S18c, d), demonstrating that RASON inhibition sensitized KRAS mutant cancer cells to cetuximab through impairment of KRAS activation.
Lastly, we tested the effect of combination therapy with cetuximab and RASON shRNA in PDAC patient-derived organoid (PDO) models (Fig. 7c). Two PDOs established in our lab, namely PDO-001 (KRAS G12D ) and PDO-068 (KRAS G12V ), were first stably transfected with tetracycline-inducible RASON shRNA constructs. PDOs were then treated with cetuximab, RASON shRNA, or both. In the combination group, RASON KD was activated three days before cetuximab treatment (Fig. 7d). The growth of organoids was monitored by measuring organoid area on pictures taken at various time points or by measuring the number of viable cells using the Cell Titer-Glo assay at the end of the experiment. Therapeutic responses similar to that in PDAC xenograft experiments were observed in both KRAS G12D and KRAS G12V PDOs (Fig. 7e-h). Collectively, these data indicate that RASON could be a potential therapeutic target for KRAS mutant PDAC patients.

DISCUSSION
The mechanism of RASON function as we revealed here complements the classical model of RAS regulation (Fig. 6h). Conventional wisdom that KRAS activating mutations (G12C, G12D, G12V, G12S, etc.) lose their GTP hydrolysis capability and cannot cycle back to the GDP-bound state has already been challenged by recent evidence. 31 In fact, these mutants still retain various degrees of both intrinsic and extrinsic GTP hydrolysis capability and could indeed cycle back to their inactive state, 31 implicating the existence of an additional layer of fine-tuning for RAS regulation. To our knowledge, RASON is the first identified positive regulator of KRAS that is neither GEF nor GAP but directly binds to KRAS and stabilizes it in the hyperactive state. We speculate that RASON does this by competing with RAS-GAPs like NF1 for KRAS binding, thus abolishing the GTP hydrolysis capacity induced by RAS-GAPs. Whether a feedback mechanism exists between RASON and KRAS warrants further investigation. Moreover, our work also puts LINC00673 out of the class of lincRNAs and into the class of more conventional genes. Although divergent functions have been reported for LINC00673 in human cancers, 32 our data support an oncogenic role of it in PDAC and lung cancer through regulation of KRAS.
Direct targeting of KRAS has been considered 'mission impossible' for many years until the recent discovery of smallmolecule covalent inhibitors of KRAS G12C . 12,33 KRAS-driven cancers are also resistant to therapies targeting RTKs such as EGFR. Besides, drugs that target KRAS effectors, including sorafenib (RAF inhibitor), rigosertib (RAS mimetic that prevent PI3K interaction) did not achieve desired clinical benefits in patients with KRAS mutation. Due to the critical role of RASON in KRAS-induced tumorigenesis and maintenance, as well as the aberrant overexpression of RASON in many cancers (our unpublished data), targeting RASON, either alone or in combination with other KRAS pathway inhibitors, may provide a novel approach for treating KRAS mutant cancers.
In conclusion, we have discovered a novel protein RASON that is essential for KRAS signaling. Encoded by LINC00673, RASON directly binds to KRAS and stabilizes KRAS in its GTP-bound hyperactive state and is therefore required for KRAS-driven tumorigenesis and tumor maintenance. The therapeutic potential of our findings was confirmed by the sensitization of pancreatic cancer to EGFR inhibitors upon RASON depletion in PDO models. KRAS−RASON interaction, therefore, represents a vulnerable target for the 'undruggable' KRAS oncogenic signaling in human cancers. Cells were maintained at 37°C in a humidified atmosphere with 5% CO 2 . Cells were cultured in either RPMI 1640 or Dulbecco's modified Eagle's medium (DMEM) according to ATCC instructions with 10% fetal bovine serum (FBS). RAS-less MEF cell lines stably overexpressing KRAS G12D or KRAS G12V were obtained from NIH's RAS Initiative and cultured as indicated in the instructions online (https:// www.cancer.gov/research/key-initiatives/ras/ras-central/blog/2017/raslessmefs-drug-screens). The in-house primary KRAS G12D MEF cell line was established from E13.5 LSL-Kras G12D genetic mouse and cultured in DMEM with 10% FBS and 1% GlutaMAX (Thermo Fisher, MA, USA, 35050061). Two additional cell lines were used to produce condition medium for PDAC organoid culture. L Wnt-3A (CRL-2647™) was purchased from ATCC for Wnt3a conditioned medium production and cultured in advanced DMEM/ F12 supplemented with 10% FBS. HA-R-Spondin1-Fc 293T Cells (R&D, Minneapolis, USA, 3710-001-01) for producing RSPO-1 conditioned medium was cultured in DMEM supplemented with 10% FBS. All cell lines were regularly tested for mycoplasma contamination.

Ribosome sequencing and data processing
The ribosomal profiling technique was processed as reported previously, 34 with a few modifications as described below. Cells or tissues were treated with harringtonine (2 µg/mL, Abcam, Waltham, USA, ab141941) for 2 min Fig. 4 RASON is required for KRAS-driven pancreatic tumorigenesis. a-c Effect of Rason KO on pancreatic intraepithelial neoplasia (PanIN) formation in LSL-Kras G12D ; Pdx1 Cre mice (KC mice). a Scheme of establishing LSL-Kras G12D ; Pdx1 Cre ; Rason −/− mice (KCR mice). Rason −/− mice was crossed with KC mice to generate KCR mice. b Rason KO inhibited PanIN formation in KC mice. Both KC and KCR mice were grown to 6 or 9 months old and the percentage of mice with different PanIN grades were shown for each group (n = 30 mice). c Histopathologic lesions observed in the pancreas of mice from three different groups (Rason − /(R), LSL-Kras G12D ; Pdx1 Cre ; Rason −/− (KCR) and LSL-Kras G12D ; Pdx1 Cre (KC)) at the age of 6 and 9 months. Representative images (left, 9 months) and quantification (right, 6, and 9 months) of HE (top), Alcian blue (middle), and Sirius red (bottom) staining are shown (n = 30 mice). d-f Effect of Rason KO on PDAC tumor formation in LSL-Kras G12D ; Trp53 R172H/+ ; Pdx1 Cre mice (KPC mice). Scheme of establishing LSL-Kras G12D ;Trp53 R172H/+ ; Pdx1 Cre ; Rason mut/mut mice (KPCR mice) (d). Two mutations in Rason ORF (M1X and M15X) were introduced by CRISPR/Cas9 into the KPC mice to generate KPCR mice. Rason KO inhibited PDAC formation in KPC mice (e). Mice in all groups (KPC, n = 15; KPCR, n = 10; R (Rason mut/mut ), n = 10) were sacrificed at 3 months of age. Representative images of mouse pancreas from each group are shown; red circles indicate pancreatic tumor. Histopathologic analysis of mouse pancreas from each group (f). Top, HE staining; bottom, Alcian blue staining; left, representative images (bars, 100 µm); right, quantification. Data shown are means ± SD. P values were calculated by paired Student's t-test (b, c, f). **P < 0.01. and cycloheximide (100 µg/mL, Sigma, MO, USA, 66-81-9) for 3 min. Samples were then dissolved in lysis buffer. Lysates were centrifuged at 20,000× g for 10 min at 4°C, then the supernatant was incubated with 7.5 µL RNase I (NEB, Ipswich, USA, M0307) and 5 µL DNase I (NEB, Ipswich, USA, M0303) in 300 µL total volume for 45 min at room temperature followed by the addition of 10 µL SUPERase·In RNase inhibitor (Ambion, TX, USA, AM2696) to stop nuclease digestion. Size exclusion columns (Illustra MicroSpin S-400 HR Columns, GE Healthcare, 27-5140-01) were equilibrated with 3 mL of polysome buffer by gravity flow and then centrifuged at 600× g for 4 min at room temperature. 100 μL of digested RPFs was then added and centrifuged at 600× g for 2 min. 10 μL of 10% (wt/vol) SDS was then added and RPFs with size greater than 17 nt were isolated according to the RNA Clean and Concentrator-25 kit (Zymo Research, Irvine, USA, R1017). rRNA was removed using the method reported previously. 35 In brief, short (50-80 bases) antisense DNA probes complementary to rRNA sequences were added to solution containing RPFs; then RNase H (NEB, Ipswich, USA, H0110) and DNase I was added to digest rRNA and residual DNA probes. Finally, RPFs were further purified using magnet beads (Vazyme, Nanjing, China, N412). After obtaining RPFs, Ribo-seq libraries were constructed using NEB Next® Multiple Small RNA Library Prep Set for Illumina® (E7300S, E7300L). In brief, adapters were added to both ends of RPFs, followed by reverse transcription and PCR amplification. The 140-160 bp size PCR products were enriched to generate a cDNA library and sequenced using Illumina HiSeq TM X10 by Gene Denovo Biotechnology Co. (Guangzhou, China).

CRISPR-mediated mouse Rason KO in Ras-less MEF cells
The sgRNA sequences targeting mouse Rason (CGCGGATTGTGTGTTTGCGT and GGGTTGGGGTTCCCCTAACG) were cloned into the eSpCas9-2A-GFP (PX458) plasmid by clone EZ. Plasmid amplification was carried out using DH5α competent cells. Ras-less MEF cells were seeded in 100 mm dishes and transfected at approximately 85% confluency. In brief, 30 μg of sgRNA expression plasmids were transfected using Lipofectamine 3000 according to the manufacturer's protocol. Three days later, GFP-positive cells were sorted into 96-well plates by flow cytometry, single cell clones were picked and validated after 2 weeks.

CRISPR-mediated NF1 KO in Ras-less MEF cell lines
The sgRNA sequences targeting mouse NF1 (CCAGGACATCTCCAAGGATG) was cloned into pLentiCRISPR v2. KRAS G12D MEF cells were seeded in 100 mm dishes and transfected at approximately 85% confluency, followed by puromycin selection.

Antibody generation and immunoblotting
A rabbit monoclonal antibody against the N-terminal 1-14 residues of human RASON protein was obtained by inoculating rabbits with synthesized peptides. The antibody was purified using affinity chromatograph columns. Cell or tissue lysates were separated by 7.5%-17%

LC-MS/MS analysis
Total proteins were collected, separated by 17% SDS Gel, stained with Commassie Blue, and the bands at corresponding molecular weight were excised and subjected to trypsin digestion. The resulting peptides were analyzed by QExactive mass spectrometer coupled to a nano-LC (AdvanceLC, Michrom Inc.) The acquired spectra were analyzed with the SEQUEST HT algorithm.

Immunoprecipitation
HEK293T cells were transfected with different plasmids for 72 h. Cells were lysed in ice-cold lysis buffer (0.3% CHAPS, 10 mM β-glycerol phosphate, 10 mM pyrophosphate, 40 mM HEPES (pH 7.4), 2.5 mM MgCl 2 and EDTAfree protease inhibitor). The soluble fractions from cell lysates were immunoprecipitated with primary antibodies against RASON, KRAS, FLAG, NF1, Rho, cdc42, or ARHGAP10 by incubation overnight at 4°C. Protein A or G beads were then added and incubated for 2 h at room temperate. Immunoprecipitants were washed five times with PBST and subjected to immunoblotting.

RAS activity assay
GTP-bound RAS (active Ras) was measured using the C-Raf RAS-bindingdomain (RBD) pull-down and detection kit (8821, Cell Signaling Technology) as instructed by the manufacturer. In brief, for the negative control, 500 µL cell lysates were treated with 10 µL 0.5 M EDTA, pH 8.0 (for a final concentration of 10 mM) followed by adding 5 µL of 100 mM GDP (for the final concentration of 1 mM). The mixture was incubated at 30°C for 15 min with constant agitation and the reaction was stopped with 32 µL of 1 M MgCl 2 (for final concentration of 60 mM). All samples (but not including the Input one) were mixed with GST-Raf-RBD and glutathione resin and incubated at 4°C for 30 min. Bounded proteins were eluted and analyzed by immunoblot.

Immunofluorescence
Cultured cells were fixed with 4% formaldehyde for 10−15 min and then blocked with 3% BSA and 0.1% Triton X-100 in PBS for 20 min at room temperature. Immunostaining was performed using the following primary antibodies: anti-RASON (This study), anti-KRAS (Abcam, Waltham, USA, Ab180772). Nuclei were counterstained with DAPI. Images were taken with an Olympus FV1000 confocal microscope (Olympus, Japan) or Leica Inverted Confocal SP8 (Leica, Germany).
well in 96-well plates in triplicate for 36-48 h. Cells were then exposed to 50 μM EdU for 2 h at 37°C and washed with PBS three times. Next, cells were fixed with 4% formaldehyde for 30 min at room temperature and washed with PBS three times, followed by treatment with 0.5% Triton X-100 for 10 min at room temperature. After three PBS washes, the cells were added with 100 μL of Apollo® reaction cocktail for 30 min and washed with PBS three times. Then 100 μL Hoechst 33342 (5 mg/mL) was used to stain the cells for 30 min followed by three PBS washes. Representative images were taken with an Olympus fluorescence microscope.

Cell proliferation assay
The cell counting kit-8 assay (CCK8, Dojodo, Tabaru, Japan, CK04) was performed to measure cell proliferation. Five hundred cells were seeded into each well of 96-well plates and allowed to grow or be treated with drugs for different amounts of time as indicated by each experiment. Cell numbers were determined following the manufacturer's protocol and absorbance was measured at 450 nm on a microplate reader (Thermo Fisher, MA, USA).

Colony formation assay
For colony formation assays, 300 cells per well were seeded into 6-well plates and incubated for 2 weeks with complete medium. The colonies were fixed with 4% paraformaldehyde and stained with 0.1% crystal violet and counted under a microscope.

Alcian blue and sirius red staining
Alcian blue and Sirius red staining of mouse pancreas tissues were carried out according to published protocols. 36 The expression of mucin was stained with Alcian blue reagents (IHC WORLD, Woodstock, USA, IW3000) according to manufacturer's instructions. Briefly, pancreas sections were deparaffinized, hydrated, and immersed in Alcian blue solution for 30 min at room temperature. After washing with distilled water, the slides were counterstained in nuclear fast red or eosin. Strongly acidic mucosubstances will be stained blue. The presence of collagen was stained with Sirius red (IHC WORLD, Woodstock, USA, IW3012). In brief, slides were deparaffinized and hydrated and nuclei were stained with Weigert's hematoxylin for 10 min. Following 10 min of washing in running tap water, the slides were stained in picro-Sirius Red solution containing 0.1% Sirius Red in saturated aqueous solution of picric acid for 1 h. Collagen-rich tissue is stained red on a pale-yellow background.

Protein expression and purification
For eukaryotic expression, RASON ORF was cloned into the pcDNA3.1 vector and transfected into HEK293T cells using Sinofection Transfection Reagent (Sino biological Inc. Beijing, China, STF02), and cultured with SMM 293-TI medium (Sino biological Inc. Beijing, China, M293TI-1). Shake flask culture condition: 37°C, 5% CO 2 , 175 rpm table speed. One week after transfection, the medium was collected and filtered through 0.45 µm filters. The purification process contained the following steps: Step 1, the Ni 2+ packed column was washed with ddH 2 O and equilibrated with binding buffer (50 mM Tris-HCl, 150 mM NaCl, pH 8.0) at a flow rate of 0.75 mL/min. Step 2, samples were then loaded onto the Ni 2+ packed column at a flow rate of 0.5 mL/min. Step 3, elution buffer was used to elute unbound proteins and molecules (washed twice with 50 mM imidazole and 100 mM imidazole, respectively).
Step 5, RASON protein was verified by immunoblotting.
Step 6, RASON protein was washed with wash buffer (10 mM Tris-HCl, pH 7.6, 150 mM NaCl) at a flow rate of 0.75 mL/min and kept in −80°C.
For prokaryotic expression, human RASON protein fused with a C-terminal TEV protease cleavage site and Tamavidin 2 protein was incorporated into the pET-30a vector between NdeI and XhoI endonuclease cleavage sites with a 6× His tag at the C-terminus. The protein was overexpressed in BL21(DE3) bacteria as incubated at 37°C and induced with 1 mM IPTG. After 4 h of expression, the cells were collected by centrifugation at 4°C and 4000 rpm for 10 min, and then suspended in the buffer of 50 mM Tris-HCl, pH 8.0, 300 mM NaCl, and 2 mM DTT and lysed by sonication in an ice-water bath. The inclusion bodies were washed twice using the lysis buffer and then dissolved in 8 M urea and loaded onto a nickel column. After removing urea step wisely, TEV protease was added to cleave the recombinant protein at 4°C overnight on the column. The eluted RASON was further purified by Superdex 75 10/300 GL gel-filtration column (GE Healthcare) and the protein concentration was measured by NanoPhotometer N50.

NMR experiment
All NMR experiments were carried out at 298 K on a Bruker AVANCE NEO 800 MHz spectrometer equipped with a 5 mm z-gradient 1 H& 19 F/ 13 C/ 15 N TCI cryogenic probe. The 2D 1 H-15 N HSQC spectra of 15 N-labeled KRAS G12D (20 μM) in the absence and presence of unlabeled RASON were obtained to detect the interaction. All NMR spectra were processed and plotted by topspin 4.1.1.

GTP hydrolysis assay
The malachite green assay was used to measure intrinsic and extrinsic GTP hydrolysis rates of KRAS G12D and KRAS G12V by detecting the production of inorganic phosphates. 39  For extrinsic hydrolysis, the experiment was carried out similarly as described for the intrinsic hydrolysis assay using the same plate set-up and plate reader. In brief, 7.4 µL GTP-loaded KRAS protein in reaction buffer was added to each well. Then 0.6 µL extrinsic mixture including ingredients of intrinsic mixture plus NF1 (RAS-GAP) was added to corresponding wells followed by incubation for specified durations at 37°C. For extrinsic hydrolysis of KRAS, the duration of incubation was set as follows: 0 s, 30 s, 60 s, 120 s, 300 s, 600 s, 1800 s, 3600 s, respectively. Finally, 2 µL EDTA was added to stop reaction. The concentrations of each component in the final mixture were the same as the intrinsic hydrolysis assay except for NF1 (1.0 µM). To eliminate contaminating background phosphate signal, we included a control group with all ingredients except KRAS for every corresponding experiment.

Nucleotide exchange assay
Nucleotide exchange assay for KRAS G12D and KRAS G12V proteins were carried out according to published protocols. 14

SPR
A Biacore S200 instrument (GE Healthcare, Chicago, USA) was used to detect protein-protein interactions using a direct binding assay format. Prior to activation, the CM5 chip surface was preconditioned using two 50 μL injections each of 10 mM HCl, 50 mM NaOH, 0.1% SDS, and 0.085% H 3 PO 4 at a flow rate of 100 μL/min. RASON was immobilized on the sensor surface using standard amine coupling. This was achieved by activating the sensor surface using 7 min injections of a mixture of 11.5 mg/mL N-hydroxysuccinimide with 75 mg/mL 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride. Protein immobilization was accomplished using a 10 min injection of protein (30 μg/mL) in 10 mM NaAc, pH 5.0 buffer. Remaining reactive esters were blocked using a 7 min injection of 1 M ethanolamine, pH 8.5, at a flow rate of 10 μL/min. Reference flow cells were prepared without the protein. Denaturation was achieved by a 30 s injection of 10 mM HCl at a flow rate of 30 μL/min. All binding measurements were performed in 0.01 M HEPES, pH 7.4, 0.15 M NaCl, 3 mM EDTA, 0.005% surfactant P20, at a flow rate of 30 μL/min. KRAS G12D or KRAS G12V was injected over the immobilized RASON protein and reference surface with at least 30 s association and dissociation time. Surface regeneration was achieved by dissociation for a time period allowing the response to return to baseline. Control injections of different concentrations of KRAS G12D and KRAS G12V were used to allow monitoring of the functionality of the protein surface. SPR equilibrium binding data, consisting of Req values from several concentration series, were analyzed by fitting a simple 1:1 binding to yield Rmax and K d values using Biacore S200 Evaluation Software.

Quantification of KRAS, NF1 and RASON protein levels in cells
LC-MS was used to quantify the protein levels of KRAS, NF1, and RASON in AsPC-1 and PANC-1 pancreatic cancer cell lines. Purified KRAS, NF1, and RASON proteins were first serial diluted, digested, and analyzed by NanoLC/ Triple TOF 5600 (AB SCIEX, Framingham, USA) to generate a standard curve for each protein. Total proteins from cell lysates were separated by 15% SDS-PAGE, stained with Commassie Blue, and the bands at corresponding molecular weight were excised and subjected to trypsin digestion. The resulting peptides were analyzed by LC-MS. The acquired data were analyzed with the Peakviewer 2.0 and Proteinpilot software. Intracellular levels of each protein were then determined by detection and quantification of the respective signature peptides according to the standard curve.

Mice and animal housing
Athymic (Ncr nu/nu) mice at 6-8 weeks of age were purchased from GemPharmatech Co, Ltd (Nanjing, China). All mice were housed in a SPF facility (5 mice per cage) under a 12-h light-dark cycle with free access to food and water. All experiments using animals were conducted under the Institutional Animal Care and Use Committee (IACUC)-approved protocols at Nanjing University (IACUC-2101008) or Sun Yat-sen University (2021878) in accordance with NIH and institutional guidelines.
Subcutaneous tumor formation of MEF cells. Ras-less MEF cells (1.5-2 × 10 6 ) with different genetic manipulations were injected subcutaneously into the dorsal flanks of 6-week-old male athymic nude mice, with each mouse bearing one control tumor and one Rason-KO tumor. Mice were euthanized when the larger tumor reached 1000 mm 3 . The tumors were then excised and measured with a caliper in two dimensions.

Micro-CT scanning
Micro-CT scans were performed using Hiscan XM Micro CT (Suzhou Hiscan Information Technology, Suzhou, China). The X-Ray tube settings were 60 kV and 133 uA and images were acquired at 50 µm resolution. A 0.5°r otation step through a 360°angular range with 50 ms exposure per step was used. The images were reconstructed with Hiscan Reconstruct software (Version 3.0) and analyzed with Hiscan Analyzer software (Version 3.0).

Human pancreatic tumor organoid culture and drug treatment
Human PDAC PDOs were established according to the protocol established by the Tuveson group. 43 In brief, freshly collected tumor tissues from PDAC patients undergoing surgery were minced and digested with 5 mg/mL collagenase II (Thermo Fisher, MA, USA, 17101015) and 10 μg/mL DNase1