RINT1 Regulates SUMOylation and the DNA Damage Response to Preserve Cellular Homeostasis in Pancreatic Cancer

Despite intensive research in basic and translational medicine, pancreatic ductal adenocarcinoma (PDAC) still has a poor prognosis. The 5-year overall survival rate of patients diagnosed with PDAC is currently estimated at less than 9% (1). The poor prognosis is due to manifold reasons, for example, (i) insufficient availability of early detection methods, (ii) high number of unresectable PDAC (around 80%), (iii) dramatic risk of early metastases in up to 90% of the cases, and (iv) frequent chemotherapy resistance to standard therapeutics (1–3). Conventional chemotherapeutic treatment as commonly used in PDAC care, such as FOLFIRINOX [5-fluorouracil (5-FU), folinic acid, irinotecan, and oxaliplatin) or gemcitabine are often limitedly effective and enriched in side effects (4). In addition, alternative treatment strategies based on mutational pattern of patients with PDAC or immunotherapy have also failed to achieve promising breakthroughs in the context of PDAC so far (2, 3, 5). Therefore, it is paramount to gain a better understanding of the cellular behavior of PDAC, which may help to identify novel therapeutic targets.

RAD50-interacting protein 1 (RINT1) has been shown to play a major role in cell cycle progression, maintenance of chromosome segregation and telomere length, and endoplasmic reticulum (ER)–Golgi trafficking (6–9). The first reported functions of RINT1 were the control of radiation-induced G₂–M checkpoint progression through the interaction with DNA repair protein RAD50 (6). However, the interaction with RAD50 alone appears not sufficient to compensate the accumulating DNA damage as shown in the fission yeast RINT1 homolog Drp1 (7). Later, RINT1 was also reported to form a vesicle-tethering complex together with ZW10 and NAG at the ER to maintain ER–Golgi trafficking of coat protein complex I (COPI) vesicles (10, 11). RINT1 was also proven to be important for the development of the central nervous system. Specifically, it was shown to preserve genomic stability and ER–Golgi homeostasis and seems to block the induction of neurodegeneration (12, 13).

Over the last decade, RINT1 was increasingly reported to participate in the development/progression of diverse diseases mostly driven by rare mutations perturbing RINT1 function. Such sporadic mutations were found to be associated with predisposition to cancer development as shown for breast, colorectal cancer, and Lynch syndrome–related cancer (14, 15), albeit a follow-up study could not confirm a role of RINT1 in breast cancer (16). Moreover, RINT1 mutations were found to be associated with other disease such as recurrent acute liver failure (17). Initially, RINT1 was reported to act as tumor suppressor gene due to abundant tumor development upon heterozygote deletion of Rint1 in mice (8). Later, functional genomics screens identified RINT1 as an oncogene in glioblastoma (18). Furthermore, RINT1 overexpression favors development of colorectal cancer (15).

Here, we report that low RINT1 protein expression correlates with prolonged survival in patients with PDAC. RINT1 abrogation negatively impacts growth of human PDAC cell lines as organoid cultures in vitro and limits tumor progression in subcutaneous and orthotopic xenotransplantation models in vivo. Mechanistically, PDAC homeostasis is affected by accumulation of DNA double-strand breaks (DSB) and Golgi fragmentation, resulting in cell cycle arrest, activation of ER stress, and cell death. Therefore, tumor-specific RINT1 inhibition in human PDAC in a clinical setting might turn out to be a novel therapeutic approach to improve clinical outcome.

PANC-1 (ATCC CRL-1469) and MIA PaCa-2 (ATCC CRL-1420) cell lines were purchased from ATCC. PA-TU-8988S (PancTu-2, DSMZ, ACC 204) were purchased from DSMZ. HPDE-E6E7 were kindly provided by A. Trauzold (University of Kiel, Kiel, Germany). Patient-derived organoids (PDO) were provided by the biobank of the University of Ulm (Ulm, Germany) and obtained between 2018 and 2020. PDOs were cultivated on a growth factor reduced (GFR) Matrigel (Corning)-coated plate in organoid culture medium containing WNT-3/RSPOI conditioned medium and 5% GFR Matrigel (19). Low-passage PANC-1, MIA PaCa-2, and PA-TU-8988S were propagated in standard culture medium containing DMEM (Life Technologies) high glucose supplemented with 10% FBS (PanBiotech), 1% penicillin-streptomycin (P/S, Sigma-Aldrich). HPDE-E6E7 (short: HPDE) were cultivated in Gibco Keratinocyte-SFM (1×) 1:1 medium and RPMI1640 Medium (ratio 1:1) containing 1% GlutaMAX, 0.025% bovine pituitary extract, 2.5 μg/L EGF, 10% FBS, and 1% P/S (20). Lenti-X (Takara Clontech) cells were cultivated with standard culture medium on 0.1% gelatin-coated cell culture plates. Mycoplasma tests (Lonza) were routinely performed. Cells were used for maximal 15 passages.

Detailed procedure is described in the Supplementary Materials and Methods. After 4 days of selection, knockdown was confirmed by either qRT-PCR or Western blot analysis, followed by outperforming the experiments.

Virus-containing supernatants with either scramble (EZ-scramble) or RINT1-shRNA (EZ-shRINT1) were produced as described above. Viral transduction was performed as described previously (21). Cells were propagated under selection for 14 days. Expression of RINT1 was measured by qRT-PCR and Western blot analysis upon 3 and 4 days of treatment with doxycycline (1 μg/mL, Sigma-Aldrich), respectively. To obtain inducible cell lines with substantial knockdown (>95% downregulation of RINT1 by RT-PCR), monoclonal cell lines were generated by limiting dilution. Therefore, 50 cells were seeded in a 96-well plate. Single-cell colonies were propagated, and RINT1 expression levels were confirmed by qRT-PCR and Western blot analysis.

Guide RNAs (gRNAs) targeting RINT1 were predicted using CCTOP (22). The top overall candidate (targeting sequence: AGG AGA GGC GCC GAT CTC GC) and the candidate with the top efficacy score (targeting sequence: AAT GGA GCC GAG GAC TCG CG) were chosen. The nontargeting sequence was ACG GAG GCT AAG CGT CGC AA. HiFi Cas9 Nuclease V3 (IDT) and the gRNAs (IDT) were mixed in a ratio of 80 pmol/240 pmol and incubated for 20 minutes at room temperature. The complex was transfected into 500,000 PANC-1 cells using the Amaxa SE Cell Line 4D-Nucleofector X Kit (Lonza), pulse code DN-100 using the Amaxa 4D nucleofector (Lonza). At 4 days after Cas9/sgRNA transfection, the cells were collected for Western blot analysis, cell viability assay, and colony-forming assay.

Cells were seeded in 8-well chamber slide and cultivated in standard medium with 1 μg/mL doxycycline or vehicle (PBS). Twenty-four hours later, cells were infected with Premo Autophagy Tandem Sensor RFP-GFP-LC3B-Kit (Thermo Fisher Scientific) according to manufacturer's guidelines and further cultivated in standard medium with 1 μg/mL doxycycline or vehicle. After 3 days, cells were incubated in medium supplemented with a cocktail of doxycycline 200 nmol/L bafilomycin A1 (Cell Signaling Technology) and Hank's Balanced Salt Solution (HBSS; Gibco, Life Technologies), for 3 hours at 37°C. Afterward, cells were fixed with 4% PFA, followed by mounting with DAPI containing ProLong Diamond Antifade Mountant (Life Technologies). Images were taken using Zeiss Axio Imager Z1 fluorescence microscope (Carl Zeiss). RFP, GFP, and RFP-GFP double-positive vesicle were quantified from three independent experiments and at ≥ 25 cells per condition using Fiji (ImageJ, NIH).

Eight-week-old female Hsd:Athymic Nude-Foxn1^nu mice were purchased from Envigo. Animals were kept in a conventional health status–controlled animal facility. Animal experiments were conducted under ethical and animal protection regulations of the German Animal Protection Law and previously approved by the governmental review board of the state of Baden-Württemberg (TVA-1458).

Data were extracted from cBioportal (23). PDAC mutation frequency and copy-number alterations (CNA) were compared with other cancer entities including all The Cancer Genome Atlas (TCGA) PanCancer Atlas Studies (32 studies, 10,967 patients). Distribution and number of mutations were extracted by compiling data from the PDAC studies (total 739 patients) from the International Cancer Genome Consortium (ICGC, n = 99; ref. 24), the Queensland Centre of Medical Genomics (QCMG, n = 456; ref. 25), TCGA (n = 184; ref. 26), and the pancreatic cancer study from the UT Southwestern Medical Center (UTSW, Dallas, TX; n = 109; ref. 27). mRNA level of RINT1 was correlated to CNA using the pancreatic adenocarcinoma study from TCGA PanCancer Atlas (n = 184). Analyses were performed with publicly available datasets from the February 20, 2020. Datasets from TCGA PanCancer Atlas (n = 184) were further clustered into classical (n = 85) and basal-like (n = 65) PDAC subtypes according to Moffitt and colleagues (28) by the “Cancer Genome Atlas Research Network” (29) and correlated with the corresponding RINT1 mRNA expression levels.

Prediction of functional effects of extracted mutations in RINT1 gene was performed with RINT1 sequence extracted from UniProt (ID: Q6NUQ1) using Polymorphism Phenotyping v2 (PolyPhen-2; ref. 30) and Align-GVGD (31). Mutations were defined as likely pathogenic, when Align-GVGD score was > C0 and PolyPhen-2 result “probably damaging.”

Gene expression of RINT1 from tumor tissue and paired normal tissue was extracted from Gene Expression Profiling Interactive Analysis (GEPIA) and was performed with datasets available on May 26, 2020 (32). RINT1 mRNA expression of established cell lines were extracted from the Expression Atlas of EMBL-EBI using the 675 Genentech pancreatic carcinoma (n = 9) and pancreatic adenocarcinoma (n = 16) datasets on October 9, 2020.

The previously published ULM cohort included 122 patients with resected PDAC (33, 34). PDACs were classified with the International Union Against Cancer (UICC) according the 7th Edition of the tumor–node–metastasis Classification of Malignant Tumors. In few cases, metastatic PDAC has also been resected (35). All 110 tumor samples analyzed in this study were treatment naïve. To complete 40 patients received adjuvant chemotherapy after primary resection. In detail: 24 patients received gemcitabine monotherapy and 14 received a combination of gemcitabine with either capecitabine (2 patients), erlotinib (8 patients), or cetuximab (4 patients). 5-FU was administered to 1 patient. One patient received a not further specified study drug. RINT1 protein expression was determined by IHC in 110 samples. Twelve patients were excluded because of absence of tumor in formalin-fixed paraffin-embedded tumor specimens. RINT1 protein expression level within the tumor compartment was evaluated by a board-certified pathologist at the University of Ulm (Ulm, Germany). RINT1 expression level was scored from 0 (negative) to 5 (strong). RINT1 expression level were correlated to clinical characteristics of patients.

Mass spectrometry (MS) analysis of whole proteome MS analysis was performed as followed. Database search was performed using MaxQuant Ver. 1.6.3.4 (www.maxquant.org; ref. 36). Employing the build-in Andromeda search engine (37), MS-MS spectra were correlated with the UniProt human reference proteome set (www.uniprot.org) for peptide identification (version November 26, 2018, 73,940 protein entries). Carbamidomethylated cysteine was considered as a fixed modification along with oxidation (M), and acetylated protein N-termini as variable modifications. FDRs were set on both, peptide and protein level, to 0.01. For label-free quantification (LFQ) values were generated by Andromeda. LFQ intensities were imported into Perseus Software 1.6.2.3 (www.maxquant.org; ref. 36). Only protein groups with at least 3 of 6 valid values in at least one condition [shRINT1-transfected cells (two constructs with and without doxycycline in triplicate)] were accepted. Missing values were imputed using standard settings. Mean values were calculated for each condition and a ratio between conditions was deemed significant with a fold change of more than 2 and a P value (t test) smaller than 0.05.

The MS-MS spectra of the interactome analysis were searched against the SwissProt Database (07/2013, Taxonomy homo sapiens, 20,341 protein entries) using Mascot v 2.4.1 (Matrix Science Inc). Fragment ion mass tolerance was set to 0.5 Da and precursor mass tolerance to 100 ppm. Carbamidomethylation was set as fixed modification and deamidation of asparagine and glutamine, oxidation of methionine, phosphorylation of serine, threonine and tyrosine and GlyGly of lysine were set as variable modifications. Scaffold (version Scaffold_4.0.7, Proteome Software Inc.) was used to validate MS-MS based peptide and protein identifications. Peptide identifications were accepted if they could be established at greater than 90.0% probability by the Peptide Prophet algorithm (38) with Scaffold delta-mass correction. Protein identifications were accepted if they could be established at greater than 90.0% probability. Protein probabilities were assigned by the Protein Prophet algorithm (39). Proteins that contained similar peptides and could not be differentiated on the basis of MS-MS analysis alone were grouped to satisfy the principles of parsimony. Proteins identified in EGFP-RINT1 sample, but not in GFP-EV were characterized as putative RINT1 interaction partners.

Here, identified putative interaction partners of the interactome analysis were correlated with differentially regulated proteins (DRP) from the whole proteome analysis. Protein–Protein interaction network was generated from the overlap using STRING database. Here, minimum required interaction score was set on medium confidence (0.400) and k-means clustering on 4.

Statistical analysis was performed using GraphPad Prism. Student t test or two-way ANOVA with Bonferroni posttest were used for growth curve, viability and cytotoxicity assay, Annexin V, proliferation assay by EdU, histology, cytology, subcutaneous experiments, qRT-PCR, and Western blot quantification. Significance for qRT-PCR data was determined using Mann–Whitney U test. Significance for survival differences of orthoptic xenografts was calculated using log-rank (Mantel–Cox) test. Significance for survival differences of human patients with PDAC from tissue microarray was calculated using log-rank (Mantel–Cox) test and log-rank test for trend. Significance for the patient cohort characteristics sex, tumor size, nodal and systemic metastasis, stage, grade, and surgical margins was determined using either χ² or Fisher exact test. All tests were considered to be statistically significant when P ≤ 0.05.

For a detailed description of further materials and methods, see Supplementary Materials and Methods section, including primer sequences (Supplementary Table S1) and antibodies used in the study (Supplementary Table S2).

Rare mutations in RINT1 were described as predisposing to or favoring tumorigenesis in a variety of cancers including breast, colorectal cancer, and Lynch syndrome–related cancer (14, 15). First, we queried genomic databases (TCGA PanCancer Atlas, and ICGC, QCMG, UTSW) for somatic RINT1 mutation in various tumor entities. RINT1 was mutated in human sporadic PDAC with an overall mutation rate of 1.8% (Fig. 1A). Five missense and one nonsense mutations distributed across the RINT1 coding sequence were identified (Fig. 1B). Among them, the D48Y and R609H missense mutations are likely pathogenic as demonstrated by alignment analysis using Align-GVGD and PolyPhen-2 classifiers (Supplementary Table S3). Accordingly, these mutations may have the potential to interfere with RINT1 function by impairing its interaction with ZW10 and RAD50. Because RINT1 functions as an oncogene in glioblastoma with high frequency of copy-number gains (18), we extended our analysis to RINT1 CNAs in PDAC. Indeed, 37.5% of all PDAC cases (n = 184) displayed a CNA of RINT1, albeit none showed a significant modification of mRNA levels (Fig. 1C–E). There was also no correlation between previously reported classical and basal-like PDACs and RINT1 expression (Supplementary Fig. S1A). We then assessed RINT1 protein expression in an independent cohort of resected PDACs (33, 34) by IHC using RINT1-specific antibody (Fig. 1F). Here, we observed different RINT1-positive staining patterns (from weak to intense, score 1 to 5), whereas normal pancreatic tissue appeared to be mainly weak positive (Supplementary Fig. S1B), leading to a significant correlation between RINT1 expression and median overall survival (mOS; P = 0.0339). Interestingly, patients with a weak RINT1 staining pattern (score 1), however, not negative (score 0), showed the best overall survival with a mOS of 865.0 days (Fig. 1F and G). In contrast, the RINT1-negative PDAC subgroup (score 0; Fig. 1F) showed the worst mOS (338.0 days) with significantly more poorly differentiated (grade 3/4) tumors (Fig. 1G and H). Surprisingly, increasing RINT1 protein expression correlates with a more advanced (stage III/IV) of PDACs accompanied with an higher frequency of metastasis (Fig. 1G and I; Supplementary Table S4), and goes along with a steady decrease in overall survival. Stratification of our cohort according to GATA6 expression levels into the PDAC subtypes (classical and basal-like) did not correlate with RINT1 protein expression scores as suggested by published transcriptome datasets (Supplementary Fig. S1A and S1C). Collectively, these results show that RINT1 expression affects patient outcome and outline its potential oncogenic role in pancreatic cancer tumorigenesis.

To elucidate the effects of its depletion on cellular behavior, RINT1 was knocked down in three PDAC cell lines (PANC-1, MIA PaCa-2, and PancTu-2) expressing aberrant levels of RINT1 (Supplementary Fig. S2A) using short hairpin (shRNA; Fig. 2A). RINT1 depletion resulted in substantially reduced cell growth and impaired colony-forming capacity (Fig. 2B–D; Supplementary Fig. S2B–S2H), whereas the extent directly correlates with RINT1 protein knockdown (Supplementary Fig. S2B–S2E). A CRISPR/Cas9 RINT1 knockout approach confirmed shRNA-mediated growth defects (Fig. 2E–H). Of note, the human pancreatic duct epithelial cell line (HPDE), harboring no pancreatic oncogenic driver mutation but already a defective Tp53 pathway (20), showed similar RINT1-associated growth defects (Supplementary Fig. S2I–S2K). On the basis of these findings, we assessed whether the impaired cell growth is associated with increased cell death. To investigate the effect of RINT1 on cell viability, stable MIA PaCa-2 Tet-On shRNA cell lines were generated (Fig. 2I; Supplementary Fig. S2L; ref. 21). After doxycycline induction, stable MIA PaCa-2 clones expressing shRINT1 (EZ-shRINT1 #2 and #12) displayed a significantly increased number of dead cells (Fig. 2J; Supplementary Fig. S2M). PDAC cells can undergo programmed cell death via apoptosis and necroptosis (40). We further investigated the mechanism of cell death (Fig. 2K). First, chemical blockage of apoptosis with the pan-caspase inhibitor Z-VAD-FMK resulted in a pronounced reduction of dead cells in both RINT1-depleted MIA PaCa-2 clones (Fig. 2L). The induction of cell death via apoptosis in EZ-shRINT1 cells was validated by the observation of elevated levels of early and late apoptotic cells, and increased PARP cleavage, both of which could be reversed through caspase inhibition (Fig. 2M–O; Supplementary Fig. S2N and S2O). We next prevented necroptosis with necrostatin-1, a RIPK1 inhibitor. Here, a significant decreased number of dead cells was observed in one EZ-shRINT1 clone (Fig. 2L), indicating that cell lethality is mainly but not exclusively driven by apoptosis. Interestingly, inhibition of both apoptosis and necroptosis did not completely restore cell viability (Supplementary Fig. S2P), arguing that cell death caused by RINT1 depletion is at least partially triggered by an independent mechanism. This might result from (i) a disruption of cell cycle progression and an accumulation of DNA damage and/or (ii) an impaired ER–Golgi homeostasis (Fig. 2P).

RINT1 was previously described to control radiation-induced G₂–M checkpoint progression (6). Having demonstrated the role of RINT1 in cell growth and viability, we sought to investigate its effects on cell cycle progression (Fig. 3A). Indeed, depletion of RINT1 in PDAC cells was associated with cell cycle arrest in G₂–M-phase (Fig. 3B and C). To test whether RINT1-depleted cells remain in G₂-phase or proceed toward mitosis, we investigated the key mitosis marker phospho-S10-Histone H3 (H3P; Fig. 3D). Upon RINT1 knockdown, we observed a decreased number of H3P-positive cells as well as a significantly increased size of nucleus (Fig. 3E–G), demonstrating that deficiency in RINT1 is associated with cell cycle arrest in G₂-phase. G₂ cell cycle arrest typically occurs in response to DNA damage (Fig. 3H). Indeed, RINT1 depletion led to a significant increase of DSB levels as determined by the amount of phospho-S139-Histone H2AX-positive (H2AX p-S139) foci (Fig. 3I and J; Supplementary Fig. S3A and S3B). Of note, the amount of DNA single-strand breaks as determined by RPA-positive foci in MIA PaCa-2 cells remained unchanged upon RINT1 deficiency (Supplementary Fig. S3C and S3D). Upregulated H2AX p-S139 protein levels during immunoblotting and accumulation of WEE1 kinase transcripts, a critical G₂–M checkpoint, after RINT1 loss further substantiated these findings (Fig. 3K and L; ref. 41). RINT1-depleted cells also displayed a higher number of 53BP1-positive foci (Fig. 3J). Although we observed an increase in H2AX p-S139 and 53BP1 double-positive foci in doxycycline-induced shRINT1 cells, their number was surprisingly reduced compared with the total amount of 53BP1-positive foci (Fig. 3I), suggesting a defect in 53BP1 recruitment to DSB sites. Microarray-based profiling followed by gene set enrichment analysis confirmed these effects and identified major biological processes such as G₂–M cell cycle transition and DNA repair as strongly enriched in shRINT1 MIA PaCa-2 cells (Fig. 3M). Altogether, these data suggest that RINT1 deficiency leads to an accumulation of DSBs, associated with a G₂-phase cell cycle arrest, in PDAC cell lines.

Several studies reported RINT1 as crucial for retrograde transport of COPI vesicles between Golgi and ER (Fig. 4A; refs. 8, 10, 42). Because disruption of this complex was associated with Golgi dispersion, we analyzed its structure after RINT1 downregulation in PDAC cell lines. Depletion of RINT1 in PANC-1 and MIA PaCa-2 cells resulted in significantly increased Golgi fragmentation (Fig. 4B and C), indicating a compromised vesicular transport between ER and Golgi, in line with published data (10). During brain development, RINT1 inactivation has been shown to impair Golgi homeostasis, which in turn, activates ER stress (12). Activation of ER stress is mirrored by upregulation of different unfolded protein response (UPR) pathways (Fig. 4D). Indeed, shRINT1 cells showed an upregulation of several UPR genes [i.e., spliced XBP1, ATF4, DDIT3 (CHOP), and HSPA5] in both PANC-1 and MIA PaCa-2 PDAC cell lines (Fig. 4E–G; Supplementary Fig. S4A), suggesting ER stress induction in the absence of RINT1. Both Golgi fragmentation after the loss of COPI vesicles and upregulation of spliced-XBP1 after ER stress activation have been reported to activate autophagy (43, 44), whereas RINT1 inactivation has been shown to inhibit autophagy (12). Here, RINT1 depletion led to significantly increased protein levels of LC3BII in MIA PaCa-2 cells (Fig. 4H and I), indicative of a proper autophagosome assembly in RINT1-defective PDAC cells. However, no degradation of SQSTM1 (p62) was observed (Fig. 4H). Therefore, it might be possible that RINT1 depletion causes a defect in fusion between lysozyme and autophagosome at later stages of autophagy (Fig. 4J). To test this hypothesis and assess consequences on autophagic flux, MIA PaCa-2 clones were transduced with an RFP-GFP-LC3B tandem sensor, as a readout for autophagosome/autolysosome formation (Fig. 4J). An increased amount of RFP/GFP double-positive autophagosomes was observed in shRINT1-expressing MIA PaCa-2 cell lines, showing that the loss of RINT1 triggers early autophagy initiation (Fig. 4K and L; Supplementary Fig. S4B). Of note, a similar accumulation of autophagosomes was detected in scramble control cells treated with the autophagy inhibitor bafilomycin A1. Moreover, forced activation of autophagy by starvation led to a significant reduction of GFP in the acidic autolysosome environment, in both scramble and RINT1-depleted cells, revealing that late autophagy (formation of autolysosomes) could still be induced in these cells (Fig. 4K and L). Taken together, these data indicate that RINT1 is crucial for maintaining ER–Golgi homeostasis. However, RINT1 is not involved in the assembly of autophagosomes and autolysosomes, in line with previous findings (42). The late autophagy inhibition might result from an accumulation of cellular stress, including DNA damage, rather than from the activation of cell death.

To decipher the event cascade triggered by RINT1 downregulation, we performed time-resolved microarray-based transcriptome profiling in MIA PaCa-2 cells (Fig. 5A). Interestingly, upon RINT1 depletion, the number of differentially expressed genes (DEG) steadily increased over time. At early time points, 24 and 48 hours after shRINT1 induction, 75 (49 downregulated and 26 upregulated transcripts) and 139 (69 downregulated and 70 upregulated transcripts) DEGs were respectively detected upon RINT1 loss (Fig. 5B). Ninety-six hours after induction of shRINT1, a total of 852 DEGs were identified, including 466 downregulated and 386 upregulated transcripts (Fig. 5B), with minimal overlap to scramble DEGs at the respective time points (Fig. 5B and C). Pathway enrichment analysis revealed that RINT1 depletion induces initially (24 hours) alterations in DNA damage response (DDR) [gene ontology (GO) “regulation of response to DNA damage stimulus”] and cell cycle progression (Reactome “M-phase”) followed by impaired COPI-mediated vesicular transport (Reactome “Golgi-to-ER retrograde transport”) and autophagy (GO “autophagy”; 48 hours), and finally leads to SUMOylation defects (Reactome “SUMOylation of DNA replication proteins”) 96 hours after shRNA induction (Fig. 5D). Importantly, while early alterations in DDR, cell cycle progression, and vesicular transport remain similarly deregulated throughout the experiment (Fig. 5D–F; Supplementary Fig. S5A and SB), corroborating the RINT1-deficient cell phenotype observed above (Fig. 2P). DEGs involved in apoptosis emerge only at the late time point (96 hours) upon RINT1 loss (Fig. 5F), in accordance with RINT1 protein loss starting around 72 hours after doxycycline induction (Fig. 5G and H). This suggests that onset of cell death directly correlates with the absolute RINT1 protein levels, albeit mRNA alterations occur far earlier, indicating a rather prolonged stability of RINT1 protein (Fig. 5F–H). To probe this hypothesis, protein synthesis was blocked in PDAC cells in a cycloheximide chase assay for up to 72 hours, indeed, confirming high RINT1 protein stability (Supplementary Fig. S5C). Taken together, these data dissect, in a time-resolved manner, the events following RINT1 deficiency in human PDAC and pinpoint SUMOylation of DNA replication proteins as critical downstream event.

To corroborate the observed transcriptional modulations with respective changes in whole proteomes, we performed quantitative LC/MS-MS mass spectrometry in RINT1-depleted MIA PaCa-2 cells (Fig. 6A). Out of the 3,613 quantified proteins, 385 proteins were differentially regulated (DRP) upon RINT1 depletion (208 increased and 177 decreased in amount; log₂(fold change) < −1.0 and > 1.0 and P < 0.05). To identify a putative RINT1 protein–protein interaction network, we additionally performed an interactome analysis using immunoprecipitation (IP)-coupled LC/MS-MS with GFP-empty vector and EGFP-RINT1 293T cells (Fig. 6A). We identified 173 unique proteins including RINT1. In total, 24 of these putative interaction partners of RINT1 were also found to be differentially expressed in the proteomics analysis (Fig. 6B). Interestingly, RAD50 and DNA-dependent protein kinase catalytic subunits (DNA-PKcs), two major players in homologous recombination (HR) and non-homologous end joining (NHEJ), emerged as putative RINT1 interaction partners.

To explore the interactions and connections of the identified 24 proteins with RINT1, we constructed a STRING protein–protein association network. Using k-means clustering, we identified four protein clusters, three being directly connected to each other and coordinating (i) Golgi-to-ER retrograde transport and G₂–M transition, (ii) DNA DSB repair, (iii) SUMOylation of DNA replication proteins and the nucleocytoplasmic transport (Fig. 6C). Of note, these processes were similarly identified during time-resolved transcriptomics (Fig. 5D). In addition, the fourth cluster includes proteins involved in (iv) actin cytoskeleton organization and did not show any connection to the other groups. Nevertheless, this interaction network revealed RINT1 as a pivot protein linking these different protein clusters (Fig. 6C). Intriguingly, enrichment analysis using predefined protein sets identified a variety of SUMOylation processes and terms indicating protein transport such as protein localization, protein import, or nucleocytoplasmic transport enriched in the STRING protein interaction network (Fig. 6D).

According to the STRING protein–protein interaction network, we found many key players involved in DSB repair (e.g., RAD50 and DNA-PKcs), nucleocytoplasmic transport (e.g., NUP153 and NUP50), and SUMOylation (e.g., SUMO2 and UBC9) to be differentially expressed (Fig. 6E). Downregulation of the small ubiquitin-like modifier 2 (SUMO2) and the E2-conjugation enzyme UBC9 substantiate the existence of a SUMOylation defect in RINT1-deficient cells (Fig. 6E). Thus, we matched putative SUMO2 target proteins identified by Hendriks and colleagues (45) with the differentially expressed proteins from our proteome and interactome analyses (Supplementary Fig. S6A). Among the overlapping proteins, we found several putative RINT1 interaction partners (Fig. 6C) to be SUMO targets such as NUP153 and RNF40 (Supplementary Fig. S6A and B). NUP153 was shown to be essential for the translocation of 53BP1 to DSB repair foci in association with NUP50 (46, 47). Strikingly, both NUP153 and NUP50 were significantly downregulated in RINT1-depleted cells (Fig. 6E), which is supportive to the above mentioned potential translocation defect of 53BP1 in RINT1-deficient PDAC cells (Fig. 3I and J). Interestingly, RNF40 together with RNF20 were shown to be critical for DSB repair as well (48). While both were identified as RINT1 interactors in our IP-MS (Fig. 6B and C), only RNF40, a putative SUMO target protein (45), was downregulated after RINT1 depletion in MIA PaCa-2 cells (Fig. 6E and F). The absence of RNF40 upon RINT1 knockdown prevented IP of endogenous RNF40 for SUMOylation analysis. Accordingly, IP of either GFP-tagged SUMO1 or SUMO2 revealed pronounced SUMOylation of RNF40 (Fig. 6G). Thus, RINT1 may modulate RNF40 protein stability by mediating its SUMOylation. Moreover, we identified differentially expressed RINT1 interaction partners that are dynamically SUMOylated upon a heatshock-mediated stress response (Supplementary Fig. S6A and B). Among them, DNA-PKcs (NHEJ) undergoes deSUMOylation while RAD50 (HR) is SUMOylated after stress. GFP-tagged SUMO1 and SUMO2 IP validated DNA-PKcs as SUMO1 and SUMO2 target protein before stress response (Fig. 6G). RINT1, not identified by Hendricks and colleagues as putative SUMO2 target proteins, was exclusively SUMO1-modified (Supplementary Fig. S6C), proving the validity of our system. Overall, these results indicate that the absence of RINT1 entails a dysregulation of SUMOylation of DSB repair proteins, thus impairing nucleocytoplasmic transport and the DDR.

Having demonstrated that low RINT1 levels are associated with improved survival in patients with PDAC (Fig. 1F–I) and that its depletion decreases viability of PDAC cell lines (Fig. 2B–H), we sought to confirm the role of RINT1 in relevant in vivo models (Supplementary Fig. S7A). Therefore, we performed xenograft experiments using a monoclonal MIA PaCa-2 cell line showing in vitro a substantial RINT1 depletion upon doxycycline treatment (Supplementary Fig. S2L). In contrast to their EZ-scramble counterparts, EZ-shRINT1 #2 tumor growth was dramatically reduced upon doxycycline treatment (Supplementary Fig. S7B and S7C). Indeed, all EZ-shRINT1 xenografts responded to doxycycline-induced RINT1 depletion as shown by the complete disappearance of 3 of 6 tumors (Supplementary Fig. S7D). The remaining tumors compromised a cell population with pronounced reduction in their proliferative capacity; however, without increased cell death (Supplementary Fig. S7E–S7H), indicating an escaper population tolerating RINT1-mediated cell death. To challenge these findings in a more physiologic PDAC model, also allowing the assessment of animal survival, we performed orthotopic transplantation of respective cell lines into the pancreas. After tumor engraftment, as assessed either by MRI or bioluminescence imaging, RINT1 knockdown was induced upon doxycycline treatment (Fig. 7A). Strikingly, depletion of RINT1 led to a spectrum of phenotypes, ranging from strong tumor growth reduction to almost complete tumor regression, over a prolonged period of time, which significantly improved overall survival (Fig. 7B–F; Supplementary Fig. S7I and S7J). Of note, two mice within the doxycycline-treated group transplanted with the MIA PaCa-2 EZ-shRINT1 sacrificed after a maximal observation time of 60 days, bearded no macroscopic PDAC tumors. However, histologic analysis of resected pancreata confirmed a proper tumor cell engraftment. Altogether, these data show that the downregulation of RINT1 affects PDAC cell viability, restricts tumor progression, and prolongs overall survival.

Finally, we aimed to lift these tumor cell line–based data to a more clinically relevant human system, namely, primary human PDAC-derived organoid model (19, 49, 50). Specifically, we validated our findings in a set of 4 PDOs. The PDOs were randomly selected regardless their treatment status at time of biopsy, and were isolated from either the primary tumor or liver metastasis biopsies (Supplementary Fig. S8A). In line with our above-described observation, shRNA-mediated downregulation of RINT1 in PDOs led to significantly decreased viability of the organoid culture (Fig. 7G and H). IHC confirmed RINT1 depletion, which went along with increased proliferative defects and accumulation of DSB levels, as indicated by Ki-67 and H2AX p-S139 levels, respectively (Fig. 7I; Supplementary Fig. S8B).

RINT1 operates context-dependent either as a tumor suppressor or oncogene (8, 15, 18). Our study reveals for the first time RINT1 expression levels are inversely correlated with overall survival in a human PDAC cohort, regardless of their adjuvant treatment regimen, implicating that an increased RINT1 gene dosage may drive malignancy. Interestingly, we revealed a subpopulation of RINT1 completely negative patients with PDAC with dramatically reduced patient survival, as compared with all other groups, suggesting a tight corridor of inverse correlation, turning into the opposite at a certain RINT1 expression threshold. Therefore, RINT1-negative tumors may have acquired growth advantages, for example, from inaccurate mitosis, leading to a dedifferentiated phenotype and, thus, poor clinical outcome. Nevertheless, RINT1 expression is overall not associated with a classical or basal-like PDAC subtypes. Still, due to the small number of patients this has to be taken with caution. Of note, our xenograft assays similarly suggest the existence of a RINT1-deficient, still actively cycling tumor cell population, rendering these to a potentially more aggressive cell type through accumulation of mitotic defects, as described for the “unstable” PDAC subtype (51). Indeed, RINT1 downregulation causes increased mitotic defects, with accumulation in sub-G₁-phase in HeLa cells (8) and leads to increased genomic instability followed by apoptosis in brain cells (12). In PDAC, RINT1 knockdown caused accumulation of DSBs followed by (i) failed mitotic entry, (ii) WEE1-mediated G₂ cell cycle arrest, and (iii) finally apoptosis and necroptosis. This inability to properly repair damaged DNA in the absence of RINT1, positions the latter at the DDR center.

Previously it has been shown that RINT1 is indispensable for retrograde transport of COPI vesicles, rendering RINT1 as a hub of Golgi apparatus homeostasis and ER stress–dependent autophagy. Disruption of the cytoplasmic tethering complex (8, 10, 42) causes Golgi fragmentation, which results in a defective retrograde ER–Golgi membrane trafficking. In line, RINT1-deficient, cells displayed massive Golgi fragmentation and pronounced collateral ER stress. Consecutively, UPR induction activates autophagy cascade and autophagosome synthesis, favored by the dysfunction in COPI-mediated vesicular transport. The accumulation of autophagosomes might directly result from the release of UVRAG after ER tethering complex disruption in the absence of RINT1. Nevertheless, it cannot be excluded that high levels of DNA damage could have triggered UPR pathway activation (52) and autophagy initiation (43, 53, 54). We further report that RINT1-depleted cells incompletely execute the autophagy program, as demonstrated by the absence of fused autophagosomes. Lysosomal degradation requires a functional coatomer-mediated vesicle transport (43). In our shRINT1 cells, we postulate that abrogation of vesicle traffic machinery might have caused the observed late autophagy inhibition. Moreover, in certain acute stress conditions, an excess of autophagosome production can reduce lysosomal activity, which finally results in ineffective autophagy (55).

Autophagy is intricately linked to programmed cell death in its role in protecting cells from apoptosis (56) in line with our observations. Of note, the accumulation of nonfused autophagosomes causes cell toxicity independent of apoptosis and necroptosis (55), which may explain why inhibition of both regulated cell death pathways did not fully restore viability of RINT1-depleted cells. In conclusion, RINT1 deletion results in severe cellular stress, caused by defects in membrane trafficking mediated by Golgi dispersion as well as massive accumulation of DNA damage, identified as the initiating events triggered by RINT1 loss, ultimately leading to a fatal program launch.

These functional data cumulate in our time-resolved transcriptome data together with comparative MS-based proteome and interactome analyses. Here, we identify a protein network with RINT1 as a pivotal regulator of (i) Golgi-to-ER retrograde transport and G₂–M transition, (ii) DNA DSB repair, (iii) SUMOylation of DNA replication proteins and the nucleocytoplasmic transport via its ability to modulate its interaction partner abundance. Specifically, RINT1 interacts with key mediators of the DDR such as RAD50 and DNA-PKcs, which are impaired in RINT1-deficient PDAC, thus, offering a potential explanation of increased DNA damage. Upon DNA damage, a plethora of proteins need to be conveyed from the cytoplasm to the nucleus to signal and orchestrate an adequate DDR (54). SUMOylation and ubiquitylation processes function in concert to promote DSB repair. SUMO modifications were specifically shown to be required for both HR and NHEJ-mediated DSB response. In addition, they are of similar importance at the nuclear pore complex to regulate protein stability and nucleocytoplasmic transport (57). Interestingly, several direct RINT1-interacting proteins were previously described as dynamic SUMO targets (45). Among them, the nuclear pore complex protein NUP153 plays a central role in SUMOylation initiation and deSUMOylation processes. Of note, NUP153 is required for the SUMO1-dependent nuclear import of 53BP1 (46) to DSB sites through the NUP153-NUP50 complex (47). Alongside, the lack of 53BP1 localization at the vicinity of DSB may concur to improper processing of DNA lesions. Here, RINT1 might regulate the NUP153-NUP50 interface concentration, affect 53BP1 SUMOylation and guide its translocation to DSB sites, an event axis perturbed upon RINT1 loss. Moreover, NUP153 reduction might act on other key DDR proteins such as RAD50 or DNA-PKcs (45, 58). Similarly, the key HR player BRCA1 (59), is UBC9-dependent SUMOylated in response to genotoxic stress to form an active triad with BRCA2 and BRCC3. Indeed, BRCC3 is also downregulated upon RINT1 loss. Vice versa, UBC9 downregulation itself may disrupt SUMOylation. In sum, high levels of genotoxic stress upon RINT1 loss in PDAC appear as a direct consequence of both alterations in HR and NHEJ activity due to dysregulation of SUMOylation of proteins involved in nucleocytoplasmic transport and DNA damage response. Indeed, SUMO pathway inhibition was just recently shown to induce G₂–M-phase arrest and apoptosis (60), perfectly mirroring our RINT1-depletion phenotype in PDAC cells. Still, therapeutic interference with the SUMO machinery requires caution (61). Similarly, RINT1 inactivation in neural progenitors and postmitotic neurons led to severe developmental neurologic defects and neurodegeneration (12, 62). Therefore, ubiquitous RINT1 or SUMO inhibition could lead to potential neurodegenerative defects. Nevertheless, many DDR players such as DNA-PKcs and RAD50, both deregulated after RINT1 loss, trigger an immune response after its activation (63). Thus, and because of launching cell death programs, tumor-specific targeting of RINT1 using RNAi strategy and nanoparticles as carriers (64) might be an elegant approach to target pancreatic cancer.

L. Perkhofer reports grants from DFG (PE 3337/1-1) during the conduct of the study. K.M.J. Sparrer reports grants from BMBF during the conduct of the study. P.-O. Frappart reports grants from University Medical Center of the Johannes Gutenberg University Mainz outside the submitted work. No disclosures were reported by the other authors.

F. Arnold: Data curation, formal analysis, visualization, methodology, writing–original draft. J. Gout: Data curation, supervision, methodology. H. Wiese: Data curation, writing–original draft. S.E. Weissinger: Data curation. E. Roger: Data curation. L. Perkhofer: Data curation, writing–review and editing. K. Walter: Data curation. J. Scheible: Formal analysis, investigation. C. Prelli Bozzo: Investigation. A. Lechel: Data curation, formal analysis. T.J. Ettrich: Resources. N. Azoitei: Writing–review and editing. L. Hao: Data curation. A. Fürstberger: Data curation, software, formal analysis. E.K. Kaminska: Resources, data curation, formal analysis. K.M.J. Sparrer: Investigation. V. Rasche: Data curation. S. Wiese: Data curation, funding acquisition. H.A. Kestler: Software, funding acquisition. P. Möller: Data curation. T. Seufferlein: Conceptualization, funding acquisition, writing–review and editing. P.-O. Frappart: Resources, supervision, equally contributed to this work together with A. Kleger. A. Kleger: Conceptualization, supervision, project administration, writing–review and editing.

The authors are deeply grateful to Kuhn Elektro-Technik GmbH for supporting their research to fight pancreatic cancer. They also thank Ralf Köhntop, Eleni Zimmer, Sandra Widmann, and Claudia Längle for technical assistance. The authors thank the animal facility platform Tierforschungszentrum, Ulm University (Ulm, Germany), and its members for animal care. The authors acknowledge Servier Medical Art, licensed under a Creative Common Attribution 3.0 Generic License (https://smart.servier.com/) for providing graphics. Main funding is provided by a German Cancer Aid grant to A. Kleger (111879), the INDIMED-Verbund PancChip, and Else-Kröner-Fresenius Excellence funding to A. Kleger. Additional funding comes from the Deutsche Forschungsgemeinschaft (DFG) “Sachbeihilfe” K.L. 2544/1-1 and 1-2, “Heisenberg-Programm” KL 2544/6-1, as well as the Baden-Württemberg-Foundation ExPO-Chip to A. Kleger. The DFG (K.L. 2544/7-1 and 5-1), the INDIMED-Verbund PancChip, and the Else-Kröner-Fresenius Excellence funding to A. Kleger. L. Perkhofer received supportive funds from the Bausteinprogramm of Ulm University and the DFG (PE 3337/1-1). F. Arnold is a fellow of HEIST RTG DFG GRK 2254/1. H. Wiese and parts of proteomics method development were funded by the DFG (German Research Foundation) - SFB 1074. P.-O. Frappart was funded by the DFG: FR 2704/1-1 “Emmy Noether program.”

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.