We analyzed RNF31 polymorphisms in exon 10 of RNF31 in lung cancer patients who underwent surgery in our department from 2003 to 2019. Details of the clinical courses of patients with the RNF31 Q622H polymorphism are described in the Results section.

The primers used to amplify RNF31 exon 10 were as follows: forward 5′-CTGGGCTGGGTGCCTTTTCCTGTCAGG-3′ and reverse 5′-GAGTAATTCTTGGACCAGGTATCG-3′ (10). The PCR products were purified using the ExoSAP-IT Kit (Thermo Fisher Scientific) and sequenced using the BigDye sequencing system (Applied Biosystems).

Immunohistochemistry was performed on four-micrometer-thick formalin-fixed [10% neutral buffered formalin for 24 to 48 h at room temperature (RT)], paraffin-embedded (FFPE) tissue sections as previously described (13,14). For RNF31 immunostaining (Fig. S1), deparaffinized and rehydrated sections were treated with 0.3% hydrogen peroxide (H₂O₂) in methanol for 30 min at RT to block endogenous peroxidase activity. Antigen retrieval was performed by autoclaving (5 min, citrate buffer pH 6.0). The sections were incubated with anti-RNF31/HOIP antibody (ab187976; Abcam; 1:50 dilution) overnight at 4°C, followed by incubation with a Histofine Simple Stain MAX PO (Nichirei), for 45 min at RT. The peroxidase reaction was carried out using 0.02% 3,3′-diaminobenzidine tetrahydrochloride and 0.01% H₂O₂ in 0.05 M Tris-HCl (pH 7.4). The immunoreaction was visualized with diaminobenzidine (DAB) and briefly counterstained with hematoxylin. Negative control tissue sections were stained as described above, except that the primary antibody was omitted. For p65 immunostaining, antigen retrieval was performed by autoclaving (10 min, citrate buffer pH 6.0). The sections were incubated with anti-NF-κB p65 antibody (D14E12, Cell Signaling Technology; 1:400 dilution), overnight at 4°C, followed by immersing the sections in 3% solution of H₂O₂ for 10 min at RT to block endogenous peroxidase activity. Next, addition of a Histofine Simple Stain MAX PO (Nichirei) was carried out for 30 min at RT. The immunoreaction was visualized with DAB and briefly counterstained with hematoxylin for 30 to 60 sec. Nuclear staining of NF-κB p65 was rated positive.

DNA from fresh frozen lung cancer tissue and adjacent normal lung tissue were extracted using a DNeasy Tissue Kit (QIAGEN). DNA from peripheral blood mononuclear cells was extracted using a QIAamp DNA Blood Mini Kit (QIAGEN) and that from ABC-DLBCL FFPE tissue was extracted from four micro-dissected slides using a GeneRead DNA FFPE Kit (QIAGEN), all in accordance with the manufacturer's instructions.

QIAseq DNA QuantiMIZE Kit (QIAGEN) was used to qualify and quantify amplifiable FFPE DNA prior to the library preparation. Ten nanograms (fresh frozen samples) or 100 ng (FFPE samples) genomic DNA were subjected to genetic analysis using the Human Comprehensive Cancer QIAseq DNA Panel (QIAGEN). For each library, DNA concentrations and fragment sizes were measured using the Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific) and the Bioanalyzer High Sensitivity DNA Kit (Agilent), respectively. Paired-end sequencing was performed on a NextSeq 500 platform (Illumina) for 151×2 cycles. The average number of read fragments was 22,444,402 (range 17,939,522-30,178,462). Read fragments were aligned to the hg19 assembly of the human genome (Genome Reference Consortium Human Build 37, GRCh37) and variants were called using the GeneGlobe v.2 smCounter (QIAGEN). The average coverage depth and the percentage of target coverage at ≥20× were 3,573.19 (range 3,025.23-4,366.68) and 98.68% (range, 98.18%-99.03%), respectively. VCFtools (v. 0.1.17) was used to filter out all variants other than ‘Pass’ variants with following command: ‘vcf-annotate-hard-filter’. Variants were annotated using the ANNOVAR (http://annovar.openbioinformatics.org/en/latest/) pipeline. Affectable mutation candidates were those with: 1) amino acid changes or variants on splice site; 2) low minor allele frequency in Japanese and east Asian populations (<0.01, HGVD2, DBexome20161214; https://www.hgvd.genome.med.kyoto-u.ac.jp, gnomAD exome EAS, v. 2.0.1); and 3) variant allele frequency >0.05. In total, 15 variants were detected in at least one sample. Furthermore, five variants with COSMIC v.70 (15) (cancer.sanger.ac.uk) registration were selected.

RNF31-knockout (KO) 293T cells were cultured in DMEM containing 10% fetal bovine serum, 100 IU/ml penicillin G, and 100 µg/ml streptomycin at 37°C under 5% CO₂. Transfection experiments were performed using polyethyleneimine (PEI MAX; Polysciences). Briefly, plasmid DNA and PEI [1:2 ratio of total DNA (µg): PEI (µg)] were mixed in PBS, and incubated for 15 min at RT. Then the DNA/PEI mixture was added to cells. For the luciferase assay, a pGL4.32 [luc2P/NF-κB-RE/Hygro] vector and a pRL-TK Renilla Luciferase control reporter vector (Promega) were co-transfected into RNF31-KO 293T cells with a FLAG-RNF31 expression vector (wild-type or mutant), RBCK1-myc, and HA-SHARPIN. At 24 h after transfection, the cells were lysed, and the luciferase activity was measured using a GloMax 20/20 luminometer (Promega) using the Dual-Luciferase Reporter Assay System (Promega). The enzyme activity of Renilla luciferase was used to normalize the firefly luciferase enzyme activity.

A gRNA cloning vector and a pCAG-hCas9 vector were obtained from Addgene. The nucleotide sequence 5′-TCAACCCTCAGGAAGCTCAGC-3′ in exon 2 of human RNF31 was selected as the target. These plasmids and a puromycin-resistant vector (pXS-Puro) were co-transfected into 293T cells (ATCC), and puromycin-resistant cell clones were selected by limiting dilution. Genome editing of RNF31 was screened by a BtsCI digestion assay (New England BioLabs), and mutations were confirmed by sequencing. RNF31 protein deficiency was confirmed by immunoblotting (Fig. S2).

FLAG-RNF31, RBCK1-myc, and HA-SHARPIN plasmids were co-transfected into RNF31-KO 293T cells. At 24 h after transfection, the cells were lysed with 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100, 2 mM PMSF, and complete protease inhibitor cocktail (Sigma). The Bradford protein assay was performed to determine the protein concentration. For immunoprecipitation, 800 µg of the cell lysates was incubated with 1 µg of anti-FLAG antibody (F7425; Sigma-Aldrich) or normal rabbit IgG (PM035; MBL) for 1 h at 4°C, and centrifuged at 20,000 g for 10 min at 4°C. The supernatants were incubated with Protein G agarose beads (30 µl of the 50% slurry; GE Healthcare) for 1 h at 4°C with gentle rotation. Then, beads were washed three times with 1 ml of lysis solution, centrifuged at 1,500 g for 4 min at 4°C. The samples were heated at 95°C for 5 min with SDS-PAGE sample buffer, and separated by SDS-PAGE and transferred to PVDF membranes. For SDS-PAGE, 2/15% gradient gel (Cosmobio) or 7.5% gel was used and transferred to PVDF membranes. After blocking the membrane in Tris-buffered saline containing 0.1% Tween-20 (TBS-T) with 5% skim-milk for 2 h at RT, the membrane was incubated with the appropriate primary antibodies diluted in TBS-T containing 5% skim-milk at 4°C overnight. Then, the membranes were incubated with anti-mouse IgG horseradish peroxidase-conjugated secondary antibody (NA931V; 1:10,000; Cytiva) or anti-rat IgG horseradish peroxidase-conjugated secondary antibody (NA935; 1:10,000; Cytiva) diluted in TBS-T containing 5% skim-milk for 2 h at RT when using non-HRP-conjugated primary antibodies. For detection, SuperSignal West Pico PLUS (34577; Thermo Fisher Scientific) or Luminata Forte (WBLUF0100; Millipore) was used.

The human cDNA open reading frame of RNF31 (16,17) was amplified by reverse transcription PCR. Mutants of this cDNA was prepared by the QuikChange method, and all nucleotide sequences were verified. The cDNAs were ligated to the appropriate epitope sequences and cloned into the pcDNA3.1 vector (Invitrogen).

The following antibodies were used for immunoblotting analyses: DYKDDDDK (1E6, 015-22391; HRP-conjugate; 1:20,000; Wako), tubulin (CLT9002; 1:3,000; Cedarlane), β-actin (sc-47778, 1:250; Santa Cruz Biotechnology), Myc (HRP-Conjugate, M192-7; 1:20,000; MBL), HA (HRP-Conjugate, M180-7; 1:20,000; MBL), HA (11867423001; 1:1,000; Roche), and linear ubiquitin (clone LUB9, MABS451; 1:1,000; Millipore), RNF31/HOIP (ab125189; 1:1,000; Abcam), RBCK1 (sc-49718, 1:250; Santa Cruz Biotechnology), SHARPIN (14626-1-AP; 1:3,000; Proteintech).

Data are shown as means ± SEM from at least three experiments performed in triplicate. One-way ANOVA followed by a post hoc Tukey HSD test or Student's t-test was performed using KaleidaGraph software (Synergy Software, PA, USA). For all tests, a P-value of less than 0.05 was considered statistically significant.

To identify the RNF31 Q584 and Q622 polymorphisms, we sequenced exon 10 of RNF31 in 481 patients with lung adenocarcinoma, 152 with squamous cell carcinoma, 16 with large cell neuroendocrine carcinoma, 2 with adenosquamous carcinoma, and 22 with small cell lung cancer (Table SI). Although the reported prevalence of Q584H and Q622L polymorphisms in healthy individuals is 0.95% (GO Exome Sequencing Project and 1000 Genome Project), no patient harbored these polymorphisms in our cohort. However, two patients had a Q622H polymorphism (Fig. 1A), which was not identified in public databases (SNPnexus: http://www.snp-nexus.org/v4/). More interestingly, one of these patients also had a history of ABC-DLBCL.

The first case was a 71-year-old Japanese patient who had undergone resection of lung adenocarcinoma, stage IB. The patient received postoperative oral chemotherapy for 2 years and had no recurrence at 5 years (Fig. 1B). The second case was a 78-year-old Japanese patient with a history of ABC-DLBCL (Fig. 1B). The patient underwent six cycles of R-CHOP therapy with radiation. One year later, the patient was diagnosed with lung cancer and underwent surgical resection. Pathological diagnosis was lung adenocarcinoma, stage IA. Two years later, the patient developed multiple lymph node metastases. The patient was administered vinorelbine but further developed malignant pleuritis. The patient then had pemetrexed as second-line, docetaxel as third-line, vinorelbine rechallenge as fourth-line, and tegafur/gimeracil/oteracil (TS-1) as fifth-line treatment. However, the disease progressed, and the patient eventually died of respiratory failure.

The previously reported RNF31 Q584H/Q622L polymorphisms strengthen RNF31-RBCK1 binding, which subsequently increase M1-Ub formation and NF-κB activation (10). We therefore analyzed whether the RNF31 Q622H polymorphism had similar effects. Immunohistochemistry showed that both patients strongly expressed RNF31 in tumor tissues compared with the adjacent lung tissue (Fig. 2). NF-κB activation, evaluated by nuclear localization of p65, was only observed in the ABC-DLBCL specimen, and not in the lung cancer or recurrent lymph node specimens.

Next, we examined co-mutations in the lung cancer and ABC-DLBCL specimens by cancer panel analysis. Fifteen variants were selected as candidate mutations. All samples harbored an MSH2 mutation (Fig. 3, Table SII) with similar variant allele frequencies (0.59, 0.51, and 0.55 for lung cancer, ABC-DLBCL, and normal lung, respectively). We also detected ARID1A, KMT2A, BRCA2, NFKBIA, and EP300 mutations common to all samples. The lung cancer sample was negative for EGFR, KRAS, BRAF, and NRAS mutations, but we detected mutations in TP53, MET, and MTOR that were not detected in the ABC-DLBCL lesion. In contrast, the ABC-DLBCL lesion harbored multiple mutations, including B2M, FBXW7, CCND3, PIK3R1, PRDM1, and TNFAIP3 mutations.

We next investigated the effect of RNF31 Q622H on NF-κB activation, M1-Ub chain formation, and LUBAC complex formation. We compared the NF-κB activation induced by co-expression of RNF31 Q622H with RBCK1 and SHARPIN, in comparison to RNF31 Q584H and Q622L. When wild-type RNF31 and each mutant were overexpressed in 293T cells (Fig. 4A, lower panel), elevated NF-κB activation was detected in samples with RNF31 Q584H and Q622L mutants, in contrast to RNF31 Q622H, which showed no difference with wild-type RNF31 (Fig. 4A, upper panel). We then compared the formation of linear ubiquitin chains. In contrast to NF-κB activation, the level of M1-Ub formation was similar amongst all RNF31 mutants, with no significant increase in RNF31 Q584H and Q622L mutants (Fig. 4B). We also compared the binding ability of RNF31 mutants with RBCK1 and SHARPIN by co-immunoprecipitation but found no significant difference between RNF31 wild-type and RNF31 mutants (Fig. 4C).

We further analyzed the effect of RNF31 Q622H on RNF31 and RBCK1 binding using the data from deposited crystal structures of RNF31 (Fig. 4) (18,19). We hypothesized that RNF31 Q622H polymorphism could do one of the following: (1) directly affect the RNF31-RBCK1 interface; or (2) indirectly affect RNF31-RBCK1 binding by changing high-layer structures, i.e., RNF31 Q622L polymorphism, which was proposed to affect the electrostatic interaction between RNF31 Q622 and E618 residues (10). In the crystal structures of human RNF31-RBCK1 and murine RNF31-RBCK1-SHARPIN complexes, the Q622 residue (or Q616 in mouse) is located in the ubiquitin-associated (UBA) domain of RNF31, which interacts with the ubiquitin-like (UBL) domain of RBCK1, leading to LUBAC formation (Fig. 4D and E). However, the side chain of the Q622 residue (Q616 in mouse) pointed towards the outside of the molecule and not the binding cleft between RNF31 and RBCK1. Therefore, the change in the Q622 residue might not directly affect RNF31- RBCK1 binding. Next, to compare the conformation of Q622 in the human and murine RNF31 structures, the Cα atoms of the C-terminal helix of the human RNF31 UBA (residues 610–624) in RNF31-RBCK1 was superposed onto the equivalent region of murine RNF31-RBCK1-SHARPIN with a root-mean-square deviation value of 0.857 (15 residues in total) (Fig. 4F). Superposition of the two structures showed that the human RNF31 Q622 formed an electrostatic interaction with E618, whereas the side chain of murine RNF31 Q616 and E612 pointed towards different directions (Fig. 4F). This suggests that the electrostatic interaction between murine E612 and Q616 (human E618 and Q622), which was previously proposed to affect the binding affinity between RNF31 and RBCK1 (10), is not a crucial interaction in the conformation of RNF31, at least in mice. Because this area is a highly conserved region of RNF31 amongst species, we speculated that the previously suggested human RNF31 E618 and Q622 electrostatic interaction is not crucial in humans either.