Localization of protoporphyrin IX in glioma patients with paired stimulated Raman histology and two-photon 3 excitation fluorescence microscopy

Spatially resolved transcriptomics

Tissue fixation was performed following the ‘Methanol Fixation, H&E Staining & Imaging for Visium Spatial Protocols’ (CG000160 | Rev C), which included heating the slide and immersing it in pre-chilled methanol. In the tissue staining phase, isopropanol was applied to tissue sections followed by a series of air-drying, hema- toxylin application, washing, bluing buffer application, eosin mix addition, and further washing. The slide was then dried on a heating block. Imaging was conducted using the Evos microscope, with the settings following the previously described protocol. Permeabilization and reverse transcription were undertaken without a preceding tissue optimization on Visium Tissue Optimization Slides, as the optimal permeabiliza- tion time for brain tissue had been established at 12 minutes by a previous researcher. The overall library preparation adhered to the ‘Visium Spatial Gene Expression Reagent Kits – User Guide’ (CG000239 | Rev F). During permeabilization, the Visium slide with stained tissue sections was fitted into a slide cassette and exposed to permeabilization enzyme, followed by a wash with 0.1X SSC buffer. For reverse transcription, an RT master mix was dispensed into each well, followed by a 45-minute incubation period in a thermocycler at 53 ° Celsius. In the second strand synthesis stage, each well received an addition of 75 ul 0.08 M KOH, followed by a brief room-temperature incubation. Subsequently, wells were washed with buffer EB and re- ceived the second strand mix, before undergoing a 15-minute incubation at 65 ° Celsius in a thermocycler. The denaturation process involved washing the wells with buffer EB and adding 35 ul 0.08 M KOH in each well, which were then incubated at room temperature. Afterward, Tris 1 M pH 7.0 was pipetted into four tubes of an 8-tube strip, followed by a transfer of samples from each well into these tubes. The tubes were then vortexed, centrifuged, and placed on ice, with the remaining sample stored for subsequent stages. The experiment initiated with the determination of cycle number wherein a qPCR mix was allocated across five wells of a qPCR plate, with a negative control included. The ensuing qPCR and Cq determination followed the standard protocol used for FFPE methods. Notably, uneven Cq values starting from n.5 were rounded up. In the subsequent cDNA amplification phase, an amplification mix was introduced to each sample tube, followed by thermo-cycling for actual PCR using a specified protocol. The cDNA cleanup process involved adding a SPRIselect reagent to each sample tube, followed by a series of incubation, washing, drying, and buffer addition steps. The cleaned-up samples were then transferred to new tubes. Finally, cDNA quality control and quantification were performed using a Tape Station. The total cDNA yield was calculated, factoring in the library concentration and elution volume. The process of fragmentation, end repair, and A-tailing started with using just a quarter of the purified library, with the remaining portion stored at -20 ° Celsius. The selected volume was mixed with buffer EB and fragmentation mix and incubated in a thermal cycler. Double-sided size selection was performed to discard large fragments and retain fragments within the desired size range. This involved the use of SPRIselect reagent, and resulted in a library with reduced total volume and a smaller range of fragment sizes. Adaptor ligation involved mixing adaptor ligation mix with each sample and incubating in a thermocycler. Post-ligation cleanup followed the cleanup steps post-cDNA amplification, with minor adjustments to the quantities of SPRIselect reagent and buffer EB. Sample index PCR was then performed, with an amp mix and dual index TT set A added to each sample, followed by a specific PCR protocol. The total number of cycles was determined based on the cDNA yield. Another round of double-sided size selection was performed, this time with varied substance quantities, to ensure another cleanup stage. The process concluded with a post-library construction quality control, ensuring the success of the library construction. While no exact concentration calculations were necessary, the fragment size in base pairs was of interest. A Fragment Analyzer was used due to its availability and accuracy in fragment size calculation. Sequencing was performed on a NextSeq 550.

Postprocessing and analysis pipeline

The data analysis and quality control for this research was conducted using the 10X Genomics’ space ranger pipeline and the SPATA2 (version 2.0) framework for spatial data analysis. The SPATA2 object was initiated through the ‘SPATA2::initiateSpataObject_10X’ function. This import procedure involved several stages using the Seurat version 4.0 package. Firstly, gene expression normalization was performed by dividing each spot’s values by the estimated total number of transcripts. These normalized values were then multiplied by 10,000 and underwent a natural logarithm transformation to improve interpretability and comparability across genes. Next, a regression model was applied to remove batch effects and scale the data. This model factored in sample batch and the expression percentages of ribosomal and mitochondrial genes, helping to control for potential sources of unwanted variation in the data. For a more detailed understanding of this process, you can refer to the guide provided at this link: https://themilolab.github.io/SPATA2/. This guide provides comprehensive information about the SPATA2 package and its application in spatial transcriptomics analysis.

Postprocessing and imaging analysis

The H&E images along with the PpIX and SRH images were aligned using afine transformation as described recently. For classification of the PpIX patterns we extracted 160x160 sized patches from each barcode spot and predicted the pattern using the pretrained ResNet architecture.