Serially sectioned FFPE sections (5 μm) of specimens were prepared to generate consecutive sections that were processed for H&E and RNA hybridization. After drying overnight, the slides were baked at 42°C for 3.0 hours and again baked at 60°C for 3.0 hours just prior to deparaffinization. RNA assay slides were deparaffinized and subjected to antigen retrieval (Tris-EDTA, pH9.0; 15 minutes;>99°C), and enzymatic exposure of RNA targets (0.5µg/ml Proteinase K; 15 minutes; 37°C).
Label
n/a
Label protocol
The next morning slides underwent stringent washes and were prepared for morphology marker staining. Slides were treated with blocking buffer for 30 minutes at room temperature in a humidified chamber, then co-incubated with fluorescently labeled visualization markers; PanCK (pan-cytokeratin [CK], (Novus, Clone: AE1+AE3, Cat#: NBP2-34528), and with SYTO 13 (Thermo Fisher, Cat#S7575) for nuclei detection for 1 hour at room temperature in a humidified chamber.
Hybridization protocol
Slides were then hybridized with human whole transcriptome overnight (nanoString, #GMX-RNA-NGS-HuWTA-4) at 37°C.
Scan protocol
Slides were washed twice then loaded into the GeoMx for imaging. Once the samples were imaged, and regions of interest (ROIs) were selected, and any segmentation made. During collection, each capture area (whole ROI or segmented areas) was exposed to UV light that cleaved the linker and released the barcoded oligos for capture by microfluidics. The released barcodes were collected in 96-well plates and were used in the NGS readout library preparation procedure protocol. The resulted libraries were sequenced by the Illumina NextSeq 6000 platform using reversed, 2 x 27 base paired reads. The resulting GeoMx DSP data was then coupled to next generation sequencing (NGS) readout.
Data processing
Nanostring DSP quantification DCC files were analyzed with R version 4.1.3 [42]. R packages GeoMxWorkflows [43], GeomxTools [44], and NanoStringNCTools [45] were used to perform data normalization and quality control. Default values were used unless otherwise indicated. DCC counts and the PKC file (Whole Transcriptome Atlas, https://nanostring.com/wp-content/uploads/Hs_R_NGS_WTA_v1.0.pkc_.zip) were imported and. Ssegments identified as containing primarily ECM were excluded. All 0 counts were imputed to a value of 1. pseudocount of 1 to all counts was added. Candidate segment filter criteria were 1000 minimum reads, 80% trimming, 80% stitching, 80% alignment, 50% sequencing saturation, 10 reads as a minimum negative control count, 1000 reads as a maximum observed in NCT wells, and an area of 5000. Geometric means for negative and endogenous control probes were calculated. One segment was removed due to low negative and endogenous probe geometric means, which also had a sequencing saturation under 30%. A global (across all segments) probe filter criteria was applied of a minimum probe ratio of 0.1 (geometric mean of a given probe’s counts across all segments, divided by geometric mean of all probes for the corresponding gene target across all segments) and if found to be an outlier via Grubb’s test in at least 20% of segments. Additionally, probes were filtered for a given segment if found to be an outlier according to Grubb’s test in that segment. Probe counts were aggregated to gene counts. A limit of quantification (LOQ) per segment was calculated as the product of the negative probes’ geometric mean and the square of negative probes’ standard deviation for each segment, or a value of 2, whichever was larger. Segments with less than 1% of genes above the LOQ were considered for exclusion. Genes were excluded if having a LOQ across segments < 1%, leaving 11,799 genes for downstream analysis. 75th percentile (Q3) normalization was applied to the gene counts, which were then log2-transformed.
