Ratio and Key Clinicodemographic Data for 2023 Active Surveillance Risk Score

Blood was collected from all eligible patients participating in the study consequent to prostate biopsy. Samples of the training cohort were collected as previously described (Van Neste, Leander (2021): Ratio and Key Clinicodemographic Data Table. figshare. Dataset. https://doi.org/10.6084/m9.figshare.16602191.v1). Blood samples of the independent validation set were collected in CPT BD VacutainerTM tubes (Cat. No. 362761, BD Biosciences, San Jose, California). Samples were centrifuged on site according to manufacturer’s instructions to isolate peripheral blood mononuclear cells (PBMCs). Samples were transported and stored at 4˚C for processing within 72 hours of the blood draw. PBMCs from each individual were pooled through a 70µm filter at 4˚C into a single tube. Samples were split into approximately 1/3 and 2/3 aliquots for CD2 and CD14 cell type isolations. Aliquots were centrifuged at 300 xg for 10 minutes at room temperature to produce cell pellets. The supernatant was discarded, and cells were suspended in 225µl for CD2, and 400µl for CD14, 4°C autoMACS running buffer (Cat. No. 130-091-221, Miltenyi Biotech, Bergisch Gladbach, Germany). Specially formulated positive selection MACS Microbeads using anti-CD2 antibodies and anti-CD14 antibodies (Cat. No. 130-091-114 and 130-118-906, respectively, Miltenyi Biotech) were added to the aliquots of PBMCs at a volume of 25µl CD2 beads and 100µl CD14 beads. Beads and cells were incubated together for 15 minutes at 4˚C. After the 15-minute incubation 250µl of 4°C autoMACS running buffer was added to the CD2 sample to bring both samples to a total volume of 500µl. Samples were processed using a positive selection template on the MultiMACS™ Cell24 Separator Plus (Miltenyi Biotech) to isolate CD2 and CD14 cells. Total PSA of patient serum samples was measured using the Cobas e 411 system from Roche. Serum was collected using Greiner Bio-One VACUETTE™ Z Serum Sep Clot Activator Tubes (Greiner Bio-One, Cat# 456073P), which were centrifuged on site to separate the serum from the rest of the blood components. Serum was transferred to the laboratory at 4°C and stored at -80°C until time of analysis. RNA extraction was performed on the KingFisher Flex instrument (Cat. No. 50-152-7925, Fisher Scientific) using the Maxwell® HT simplyRNA Kit (Cat. No. AX7890, Promega). The simplyRNA kit and the KingFisher program were modified to optimize RNA extraction from eluted cells after cell separation. RNA samples were quantified using the Quan-iT RiboGreen RNA Assay (Invitrogen) and the RNA integrity was assessed by electrophoresis using the Tape Station system (Agilent). Samples were required to have a RIN number higher than 7 to proceed with the next steps. A minimum of 300 ng of total RNA was used as template for the library construction. Libraries were constructed using the Universal Plus mRNA-Seq library preparation kit with NuQuant (Tecan). The molar concentration of each individual library was determined by fluorescence using a Qbit fluorometer (Invitrogen) and the corresponding NuQuant standards. Some libraries across the study were also quantified with more traditional methods using fragment size analysis (Bioanalyzer, Invitrogen) and qPCR to correlate the results obtained using NuQuant. Libraries were combined in equimolar proportions generating 10 mM library pools. All pools were pre-run using iSeq 100 (Illumina) instrument to assure all libraries were present in comparable proportions and contributing equivalently to the final sequencing output. Finally, pools were sequenced using a 100 bp paired-end mode in a NovaSeq 6000 sequencer (Illumina). Raw sequencing reads in fastq format were quality-filtered using Seqpurge version 2019_11, also removing adapter sequences. Trimmed reads were mapped to the human genome reference (GRCh38, Ensembl version 100) with STAR version 2.7.2b using a 2-pass alignment with default parameters generating BAM files. Quantification of the reads at gene level was performed from the BAM files using featureCounts version 2.0.1, considering only reads mapped to exons. Quality reports after each step (raw, trimmed, and mapped reads) were generated using FastQC version 0.11.9 and aggregated in a final report using MultiQC version 1.9. After quantification, 24342 transcripts without gene symbol and 2870 mitochondrial and ribosomal genes were excluded using regular expressions, resulting in 33467 transcripts, from which 31918 common gene symbols between both cohorts with nonzero counts in at least one of the cell types (CD2 or CD14) were kept in the training and validation sets, respectively. Read depths were comparable across cell types, but different between cohorts, namely, 30.2±6.9 and 35.3±7.9 million median reads ± standard deviation in the training cohort for CD2 and CD14, respectively, and 60.4±22.0 and 67.4±16.3 million median reads in the validation cohort, respectively. Four samples with cell-type-specific expression profiles that were inconsistent with population estimates via principal component analysis were excluded from downstream analyses. Normalization of raw counts was carried out via trimmed mean of M-values (TMM). As previously described, the log ratio of CD14 over CD2 reads was used to obtain intra-individual normalized counts. Population differences in gene expression profiles between cohorts (<em>i.e.</em>, batch effects) were addressed by simply transferring the gene-wise population means from the training into the validation cohort. Transcriptome variables are log-transformed count ratios of CD14 and CD2 values.