Human Specimens. We obtained 40 fresh SGC tumor tissues from patients undergoing surgical resection at Yokohama City University Hospital or Yokohama City University Medical Center (Yokohama, Japan) and stored them in culture medium on ice until further use (< 12 h). The pathological diagnosis for each case was confirmed by independent pathologists after sample collection. We obtained written informed consent from all patients prior to surgery. The current study was approved by the ethical committees of the Institutional Review Boards of the Yokohama City University (approval IDs: A171130010) and was conducted in accordance with the Declaration of Helsinki.
Patient-derived xenografts (PDX). PDX were established by subcutaneous implantation of fresh minced tumors into NOD Cg-Prkdcscid Il2rgtm1Wjl (NSG) mice as previously described23. All mice were dissected to visually examine for metastases in lung, liver, and abdominal cavity. All mice were maintained and handled in accordance with the procedures approved by the Institutional Animal Care Use Committee at Yokohama City University, School of Medicine (Yokohama, Japan).
Organoid culture. Organoid cultures from patient specimens and PDX were performed as previously described23. The tissue, cut into 2–4 mm pieces, was enzymatically digested with Liberase TM Research Grade (Sigma Aldrich, St. Louis, MO, USA) and Hyaluronidase (Sigma Aldrich) for 30–60 min at 37°C. Processed tissue was passed through a 70 µm cell strainer (Corning Incorporated, Corning, NY, USA) to eliminate macroscopic pieces. The isolated cells were suspended in complete media (AdDMEM/F12 medium supplemented with HEPES [Invitrogen, Carlsbad, CA, USA], Glutamax [Invitrogen], penicillin/streptomycin [Invitrogen], B27 [Invitrogen], 1 mg/m Primocin [InvivoGen, San Diego, CA, USA], 1 mM N-acetyl‐l‐cysteine [Sigma‐Aldrich], 500 ng/mL Wnt3a [R&D systems, Emeryville, CA, USA], 0.1 µg/mL R‐spondin1 [Peprotech, Rocky Hill, NJ, USA], 0.1 µg/mL Noggin [Peprotech], 50 ng/mL epidermal growth factor [Sigma‐Aldrich], 100 ng/mL fibroblast growth factor 10 [Peprotech], 10 mM Nicotinamide [Sigma‐Aldrich], 0.1 µM A83‐01 [Sigma‐Aldrich] and 1 µM Dexamethasone [Sigma‐Aldrich]) and seeded on growth-factor-reduced (GFR) Matrigel (Corning Incorporated) coated plate, pre-prepared as a lower layer. After incubation for 16–24 hours, the media was removed and the organoids formed on the lower layer were covered with additional GFR Matrigel as an upper layer. The complete media was added after the formation of solid coating, and the media was changed every 2–3 days.
For passaging, the Matrigel containing organoids was collected from the plate, and digested with TrypLE Express Enzyme (Thermo Fisher Scientific, Waltham, MA, USA). Isolated organoids were suspended in DMEM/F12 media and physically crushed into smaller cell clumps by pipetting. Cells were centrifuged, re-suspended in complete media with 10 µM ROCK inhibitor (Y-27632, Sigma), and then embedded in GRF Matrigel as described above. Organoids were passaged at a 1:2 to 1:1.5 dilution ratio every 2–3 weeks. To prepare frozen stocks, organoids were isolated and suspended in CELLBANKER 1 (TAKARA-BIO, Kusatsu, Shiga, Japan) and stored in − 80°C freezer or liquid nitrogen. Stocks have been successfully recovered for up to at least 6 months after freezing. STR analysis was performed at BEX. CO., LTD. (Tokyo, Japan) to authenticate the identity of organoids and corresponding patient tissue. To check the contamination of mouse cells, we performed PCR of animal species-specific mitochondrial DNA sequences42 using the primers listed in Supplementary Table S3. Organoids were routinely tested for Mycoplasma using e-Myco VALiD Mycoplasma PCR Detection Kit (iNtRON Biotechnology, Seoul, Korea).
Orthotopic transplantation. For orthotopic transplantation of SGC organoids, organoids were injected into the submandibular gland and subcutaneously in NSG mice as previously described23. The submandibular gland was injected with 0.5 × 105 to 1 × 106 cells suspended in a mixture of DMEM/F12 media and Matrigel, and the subcutaneous area was similarly injected with 1 to 2 × 106 cells. Orthotopic xenograft volumes were measured weekly. Xenografts were harvested when the tumor diameter reached > 1 cm or 6 months after implantation and fixed for 24 h in 10% formalin.
IHC analysis. Fresh PDXs and orthotopic xenografts were fixed in 10% formalin for 24 h and then embedded in paraffin following standard histological procedures. Organoids were isolated by digesting Matrigel using dispase (Sigma) for 30 min at 37°C and embedded into a gel using iPGell (GenoStaff, Tokyo, Japan) according to the manufacturer’s protocol. Next, the organoids were fixed in 10% formalin for 24 h and paraffin-embedded. Haematoxylin–eosin staining and IHC were performed using standard protocols on 5-µm-thick paraffin sections. The following antibodies were used for IHC: human-Androgen Receptor (AR441, Dako, Carpinteria, CA, USA) 1:500, pan keratin AE1/AE3/PCK26 (Roche, Basel, Switzerland) 1:1, HER2 (4B5, Roche) 1:1, alpha-smooth muscle actin (S131, Leica Biosystems, Buffalo Grove, IL, USA) 1:1, p63 (4A4, Biocare medical, Concord, CA, USA) 1:200, S-100 (Roche) 1:1000, and GCDFP15 (D6, Biocare medical) 1:400. Images were acquired using an OLYMPUS BX41 microscope.
DNA/RNA extraction. Organoids were extracted from Matrigel using TrypLe. Total RNA was extracted from organoids using TRIzol (Thermo Fisher), followed by isolation and precipitation in chloroform and 70% ethanol, and then purified via column-based separation using the RNeasy Mini Kit (QIAGEN, Valencia, CA). DNA was extracted from organoids using DNA mini kit (QIAGEN) according to the manufacturer’s protocol. PDX tissues harvested from mice, as well as tissue fragments of primary salivary gland tumor, were physically homogenized using a plastic homogenizer vessel. RNA and DNA extraction from these homogenized tissues were processed similarly as described above.
RNA-sEq. RNA sequencing was performed at the Laboratory of Systems Genomics, Department of Computational Biology and Medical Sciences, at the University of Tokyo (Chiba, Japan). RNA quality and quantity were measured with an Agilent Bioanalyzer 2100. Libraries for sequencing were constructed using TruSeq Stranded mRNA (Illumina, San Diego, CA, USA)) according to the manufacturer’s protocol, followed by sequencing on an Illumina NovaSeq6000 platform to generate 70 million paired-end reads of 150 bases. The RNA-seq data are available at the DNA Data Bank of Japan Sequence Read Archive (DRA) under the accession number DRA011243.
Gene expression analysis. RNA-seq reads were quality checked and adapter trimmed using fastp (v0.20.1)43. Since RNA-seq reads derived from PDX tumors and PDX-derived organoids both essentially contain mouse reads, we distinguished the trimmed reads into those of humans (GRCh38/hg38) or mice (GRCm38/mm10) using xenome (v1.0.0)44. Only human reads were used for subsequent processing. Mouse reads and indistinguishable reads were discarded. To ensure consistency in process sampling, samples that do not contain intrinsic mouse reads, such as primary organoids, were processed in the same manner as described above. The human reads were aligned to human genome reference sequence (GRCh38/hg38) using STAR (v2.7.5c)45 and counted for each gene using featureCounts (v2.0.1)46. For a heatmap, hierarchical clustering analysis with complete linkage and Euclidean distance, and correlation analysis, the raw read counts per gene with at least an average of 5 counts were TMM normalized using edgeR (v3.30.3)47 and log2-transformed. The heatmap and clustering analysis were visualized with the top 2000 variable genes using R package “pheatmap.” The Pearson’s correlation coefficients were for calculated for all genes. For a principal component analysis (PCA), we combined our samples with RNA-seq datasets of multiple salivary gland cancers downloaded from public databases. SRP067524 (including 42 samples of ACC and 5 samples of normal salivary gland), SRP067827 (including 3 samples of acinic cell carcinoma), SRP096726 (including 16 samples of SDC), and SRP109264 (including 40 samples of MEC) were downloaded from the NCBI Sequence Read Archive. The raw read counts per gene for all samples were calculated as described above and were normalized for library size by converting to CPM (counts per million) using edgeR47. The R package “sva” (v3.36.0)48 was applied to adjust for batch effects, along with information of histological subtype of each sample. PCA was performed using the “prcomp” function in R.
Variant calling. SNP discovery and filtering from RNA-seq data were performed using HaplotypeCaller under standard parameters according to GATK24 (v4.1.8) Best Practices (https://github.com/gatk-workflows/gatk3-4-rnaseq-germline-snps-indels). Additionally, SNPs with a depth < 25 and an allele frequency < 0.2 were excluded. The functional effects of the mutations were predicted using SnpEff (v5.0)49, and SNPs with "high" or "moderate" functional importance were retained. To visualize representative genes that are mutated in salivary gland tumors in COSMIC25, the vcf format data was converted to maf format data using ANNOVAR50 and annovarToMAF under standard parameters, and the "waterfall" function of R package GenVisR (v1.20.0) 51 was applied.
Detection of fusion genes. Candidate fusion genes were explored from RNA-seq data using STAR-Fusion (v1.6.0)45, FusionCatcher (v1.20)52, and a combination of kallisto (v0.46.2)53 and pizzly (v0.37.3)54. The detected candidate fusion genes were cross-referenced to ChimerDB4.055, and those reported in salivary gland carcinoma were extracted and validated by RT-PCR. RT-PCR was performed as previously described50 using PrimeScript 1st strand cDNA Synthesis Kit (TAKARA-BIO), and RT-PCR products were subjected to Sanger sequencing at Macrogen Japan (Tokyo, Japan). All primers used are listed in Supplementary Table S3.
Drug sensitivity test. Drug sensitivity test was performed as previously described23 using CellTiter-Glo 3D (Promega). All available organoids were evaluated for sensitivity to cisplatin (3.12–800 µM range), docetaxel (0.24 nM to 4.0 µM range), and fluorouracil (0.39 µM to 6.4 mM range). Additionally, organoids derived from SDC were evaluated for sensitivity to trastuzumab (0.3 µM to 5 mM range), an anti-HER2 antibody agent. A253 cell line, as a control group, was evaluated for sensitivity to cisplatin, docetaxel, fluorouracil, and trastuzumab. Each test was repeated at least thrice. The dose-response curve was obtained as the output, and the IC50 was calculated using the four parameter logistic equation and the R package “drc” (v3.0.1)56.
Statistical analysis. The association between the establishment rate of each experimental model and the clinical information of the patients was tested using Fisher's exact ratio test and Student’s t-test in the open-source R Statistical Computing software (http://www.r-project.org/). Statistical significance was set at P < 0.05.