Sample and data collection

We analyzed formalin-fixed paraffin-embedded (FFPE) sections from 162 PDAC patients who underwent surgery between April 2008 and March 2017 at the Kobe University Hospital. We retrospectively reviewed the patients’ medical information regarding age, gender, body mass index (BMI), carcinoembryonic antigen (CEA, reference range: < 5 ng/mL), and carbohydrate antigen 19-9 (CA19-9, reference range: < 37 ng/mL), white blood cell count, neutrophil count, total leukocytes count, alcohol consumption, current smoking, diabetes mellitus and administration of antibiotics. The tumor characteristics included tumor size and location, pathological stage (Union for International Cancer Control UICC] 8th classification), histological grade, residual tumor status after surgery, and the history of neoadjuvant and adjuvant chemotherapy. Laboratory data were collected within 1 month before surgery.

The study protocol was reviewed and approved by the ethics committee of the Kobe University Hospital (No.180235). Informed consent was waived due to the retrospective study design and the study information was disclosed on our hospital website, allowing eligible patients to opt out. This study was conducted in accordance with the Declaration of Helsinki. All authors had access to the study data, and reviewed and approved the final manuscript.

DNA extracting and real-time PCR

DNA was extracted from the pancreatic tumoral and adjacent non-tumoral FFPE tissues using the QIAamp DNA FFPE Tissue Kit (Qiagen, Hilden, Germany) following the manufacturer’s protocol. The extracted DNA was quantified using a fluorometer (Invitrogen Qubit 4.0; Thermo Fisher Scientific, Waltham, MA, USA) and stored at −80 ℃ until further analysis.

Quantitative polymerase chain reaction (qPCR) was performed on Applied Biosystems 7500 real-time PCR system (Applied Biosystems Inc, CA, USA) using SYBR green qPCR assay (Applied Biosystems Inc, CA, USA). The V1–2 regions of 16S rRNA gene were amplified using forward primer 27Fmod (5′-AGRGTTTGATYMTGGCTCAG-3′), and reverse primer 338R (5′-TGCTGCCTCCCGTAGGAGT-3′) [12]. The cycling conditions were 1 cycle at 95 ℃ for 10 min to denature DNA, with amplification proceeding for 40 cycles at 95 ℃ for 15 s, 50 ℃ for 20 s, and 72 ℃ for 1 min, followed by a standard denaturation curve protocol.

In situ hybridization (ISH)

Chromogenic RNA in situ hybridization (ISH) targeting 16S rRNA was performed using RNAscope® 2.5 HD Reagent Kit-RED (Advanced Cell Diagnostics, Hayward, CA, USA) and Fast Red according to the manufacturer’s protocol. The probe used was RNAscope® Probe-EB-16S-rRNA (Cat #464461). The chromogenic reaction within the pancreatic tumor indicated a positive result.

Amplicon sequencing and microbiome analysis

The 16S rDNA V1–2 regions were amplified by PCR and sequenced in the MiSeq platform (Illumina) with MiSeq Reagent kit v2 (500 cycles) using the 250 bp paired-end protocol. The QIIME2 (version 2020.11) pipeline was used to perform microbiome analysis [13]. Demultiplexing and quality filtering were performed on the raw sequence data using the q2-demux plug-in, and amplicon sequence variants (ASVs) were counted after denoising by DADA2 [14].

Taxonomy was assigned to ASVs using reference sequences from Silva (138 SSURef NR99 full-length taxonomy). Bacterial contamination was distinguished using R program package “Decontam” with the parameter “method = frequency” by comparing the data from pancreatic tissues with that derived from 15 samples of FFPE pieces without tissues [15]. Alpha diversity analysis with Shannon index was calculated using QIIME2. Beta diversity analysis using weighted-UniFrac Principal Coordinate Analysis (PCoA) and permutation analysis of variance (PERMANOVA) were also performed using Qiime2. The taxonomic types of intratumoral microbiome to distinguish tumor prognosis were analyzed by linear discriminant analysis (LDA) effect size (LEfSe) calculations using Galaxy Version 1.0 [16].

Immunohistochemistry (IHC) and Elastica van Gieson (EVG) staining

FFPE tissues were sectioned at 5 mm thickness and analyzed by IHC and EVG staining. The following antibodies were used: anti-TP53 (Santa Cruz Biotechnology, Dallas, TX, USA, catalog number: sc-47698), anti-CDKN2A/p16 (Roche Diagnostics, Cat #6695221001), anti-SMAD4 (Santa Cruz Biotechnology, Cat #sc-7966), anti-CD4 (Leica Biosystems, Wetzlar, Germany, Cat #CD4-368-L-CE), anti-CD8 (Roche Diagnostics, Cat #5493846001), anti-FOXP3 (Abcam, Cambridge, UK, Cat #ab20034), anti-CD45RO (BioGenex Laboratories, San Ramon, CA, USA,Cat #AM113-5M), anti-CD68 (ProteinTech Illinois, USA, Cat #66231-2-Ig), anti-CD206 (ProteinTech Illinois, USA, Cat #60143-1-Ig) and anti-α-SMA (Santa Cruz Biotechnology, catalog number: sc-53142). EVG staining was performed using an Elastic Stain Kit (Abcam, Cat #ab150667) according to the manufacturer’s protocol.

Evaluation of tumor-infiltrating lymphocytes (TILs), macrophage, and tumor fibrosis

TILs positive for CD4, CD8, FOXP3, CD45RO and macrophage for CD68, CD206 were assessed by immunohistochemical staining on slides with the maximum divided surface of tumors. Each subset of TILs and macrophage were counted at 200 × magnification (counts/mm2) using Image J (Java image processing program inspired by National Institute of Health (NIH), USA). Three fields separated by at least 5 mm each were counted and the mean value was calculated for each case. The cases were classified as high density or low density based on the median value.

To assess fibrosis within pancreatic cancer, tumor stromal collagen and myofibroblasts were evaluated by EVG staining and immunostaining for α-SMA, respectively. The stained sections were digitally scanned and analyzed using Adobe Photoshop CC2019 software (Adobe Inc., San Jose, CA). The red area in EVG-stained sections and the brown area in α-SMA-stained sections were quantified as tumor stromal collagen and αSMA+ myofibroblasts, respectively. Each area was divided by the whole tumor area analyzed and defined as the area proportion of “tumor stromal collagen” and “α-SMA+ fibroblast,” respectively. Additionally, all cases were classified into two groups (high and low) based on the median value of the area proportion. Representative IHC images for TILs, CD68, CD206, α-SMA, and EVG staining image are shown in Supplementary Fig. 1.

Evaluation of driver gene alterations

Alterations of KRAS, TP53, CDKN2A/p16, and SMAD4 genes in the tumor were determined by next-generation sequencing (NGS) analysis, droplet digital PCR (ddPCR) and IHC using DNA extracted from FFPE as reported previously [17]. In brief, KRAS mutations were determined by NGS. TP53 mutations were determined based on a combination of NGS, ddPCR, and IHC. CDKN2A/p16 and SMAD4 mutations were determined using IHC. IHC sections were evaluated by two experienced pathologists (M.K. and T.I.) who were unaware of the clinical data. TP53, CDKN2A/p16, and SMAD4 were evaluated with Kappa values of 0.982 (P < 0.0001), 0.964 (P < 0.0001), 0.942 (P < 0.0001), respectively, and the agreement was high between the pathologists.

Statistical analysis

SPSS (SPSS Inc., Chicago, IL, USA) and GraphPad Prism 8 (GraphPad Software, La Jolla, CA, USA) were used for statistical analyses. The chi-square test or Fisher’s exact test, when applicable, was used to compare frequencies, and the Wilcoxon rank-sum test was used to compare skewed continuous variables. Overall survival (OS) was estimated using the Kaplan–Meier method and compared using a log-rank test. Hazard ratios (HRs) and the corresponding 95% confidence intervals (CIs) were estimated using Cox proportional-hazards models. The multivariate analyses included factors with statistical significance in univariate analysis. All statistical tests were two-tailed, and statistical significance was set at P < 0.05.