Glioblastoma-derived extracellular vesicle subpopulations following 5-aminolevulinic acid treatment bear diagnostic implications

Abstract

Liquid biopsy is a minimally invasive alternative to surgical biopsy, encompassing different analytes including extracellular vesicles (EVs), circulating tumour cells (CTCs), circulating tumour DNA (ctDNA), proteins, and metabolites. EVs are released by virtually all cells, but at a higher rate by faster cycling, malignant cells. They encapsulate cargo native to the originating cell and can thus provide a window into the tumour landscape. EVs are often analysed in bulk which hinders the analysis of rare, tumour-specific EV subpopulations from the large host EV background. Here, we fractionated EV subpopulations in vitro and in vivo and characterized their phenotype and generic cargo. We used 5-aminolevulinic acid (5-ALA) to induce release of endogenously fluorescent tumour-specific EVs (EV_PpIX ). Analysis of five different subpopulations (EV_PpIX , EV_CD63 , EV_CD9 , EV_EGFR , EV_CFDA ) from glioblastoma (GBM) cell lines revealed unique transcriptome profiles, with the EV_PpIX transcriptome demonstrating closer alignment to tumorigenic processes over the other subpopulations. Similarly, isolation of tumour-specific EVs from GBM patient plasma showed enrichment in GBM-associated genes, when compared to bulk EVs from plasma. We propose that fractionation of EV populations facilitates detection and isolation of tumour-specific EVs for disease monitoring.

Keywords: 5-ALA; 5-aminolevulinic acid; EV; IFC; PpIX; extracellular vesicles; glioblastoma; imaging flow cytometry; liquid biopsy; nano-FACS; nanoparticle fluorescence activated sorting; protoporphyrin IX.

FIGURE 1

Effects of 5‐aminolevulinic acid (5‐ALA) dosing on EV subpopulations. GBM cell lines were dosed with 5‐ALA and the released EVs were analyzed pre‐5‐ALA and post‐5‐ALA dosing. Biological replicates (n = 2) and technical replicates (n = 3) per each biological replicate (total n = 6) were utilized for each analysis. The total number (A) and size distribution of EVs (B) was determined using NTA. Comparison of EV concentration between pre‐ and post‐5‐ALA showed p‐values of 0.37, 0.99, and 0.87 for Gli36^EGFRvIII, Gli36^EGFRwt, and U87, respectively. EVs from 5‐ALA dosed GBM cells pre‐ and post‐5‐ALA dosing were stained with CFDA‐SE and phycoerythrin‐conjugated antibodies (CD63, CD9, EGFR). All EV subpopulations including EV_PpIX were studied using IFC. Average positive events are described by EVs/ml of conditioned media. Relative fold changes of each EV subpopulation pre‐ and post‐5‐ALA dosing were determined and calculated for significance (C; Gli36^EGFRwt EV_CD63: p = 0.03; U87 EV_CD63: p = 0.002; U87 EV_CD9: p = 0.0008). Relative subfractions of EV subpopulations following 5‐ALA dosing (as quantified by IFC) represented as a fraction of total EVs (gray, measured by NTA) were analyzed for abundance (D): EV_CFDA (purple), EV_CD63 (dark blue), EV_CD9 (yellow), EV_EGFR (green), and EV_PpIX (red). The mean cell fluorescence of GBM cells pre‐ and post‐5‐ALA dosing (E), the mean EV_PpIX fluorescence (F) and the relative fraction of EV_PpIX from the NTA‐derived total (G), were plotted to compare relative fluorescence and abundance. Correlative analysis of mean cell PpIX fluorescence and mean EV_PpIX fluorescence (H), mean cell PpIX fluorescence against relative fraction of EV_PpIX (I), and mean EV size against mean EV PpIX fluorescence (J) was performed.

FIGURE 2

Quantification of cell line‐derived EV subpopulation cargo. EVs were fluorescently labeled and sorted using nanoFACs. A workflow for EV sorting and downstream library preparation was developed for low‐input, bulk RNA sequencing (A). The relative RNA landscape of EV subpopulations as defined by long and small RNA biotypes was mapped according to detection (B). Gene expression (transcripts per million; TPM) across EV subpopulations and cell lines were assessed in order of phylogenetic relation (C).

FIGURE 3

Transcriptomic commonalities across cell line‐derived EV subpopulations. Comparative analysis was performed to determine the average number of common genes among subpopulations (A) and within subpopulations across cell lines (B). All genes expressed in each subpopulation were analyzed via KEGG orthological analysis and all significant pathways associated with cancer were identified (C). Significance was designated at ‐log₁₀(p value) ≥ 2.

FIGURE 4

RNA sequencing of matched EV_PpIX, plasma total EVs, and tumor tissue. A workflow for biological sample collection and the downstream extraction, library preparation, and low‐input, bulk RNA sequencing was developed for application in the patient cohort (A). A patient cohort (GBM; n = 8) and an age‐matched healthy cohort (HC; n = 8) was built for sample collection. Both cohorts are composed of an even sex distribution. Patient cohort inclusion parameters consisted of a clinical glioblastoma diagnosis and 5‐ALA fluorescence guided surgery (B). The RNA landscape of sorted EV_PpIX and total EVs derived from plasma was examined and plotted relative to total long RNAs and small RNAs (C). Average patient sample gene expression of the genes detected in the EV_PpIX transcriptome was calculated and mapped in order of phylogenetic relationship for each biological sample type (D).

FIGURE 5

Differential expression of patient and healthy samples. Differential expression analysis of patient EV_PpIX [case] against total EVs from healthy plasma [control] (A), patient EV_PpIX [case] against total EVs from patient plasma [control] (B), patient EV_PpIX [case] against patient tumor tissue [control] (C), total EVs from patient plasma [case] against total EVs from healthy plasma [control], and total EVs from patient plasma [case] against patient tumor tissue [control] was conducted to examine significant gene expression. For each of the differential expression analyses, control groups were established in order of priority (healthy plasma, tumor tissue, and patient plasma) and have been depicted as the left condition above each plot. Significance cutoffs were defined at |log₂(Fold Change)| ≥ 2.00 and ‐log₁₀(p‐value) ≥ 1.30, with significantly upregulated genes expressed in blue and significantly downregulated genes expressed in red.

FIGURE 6

EV_PpIX origin identification and PCR validation of notable genes. Gene lists with organ‐specific, elevated expression were obtained via the Human Protein Atlas (Sjöstedt et al., , The Human Protein Atlas) and compared against the cumulative gene lists for each biological type: patient tumor, patient EV_PpIX, patient plasma, and healthy plasma (A). Percentage alignment was calculated from the [number of genes aligned to the elevated expression list] out of the [total number genes detected] in each biotype. Significance was determined at a cutoff of p ≤ 0.05. Expression of genes common across all patient EV_PpIX was determined (B) and the resulting list was analyzed using gene ontology (C). The gene common to both patient EV_PpIX and the notable genes in cell line subpopulations was considered for assay validation. GREM1 expression levels (TPM) were compiled for comparative use (D). GAPDH expression was extracted for use as a control. The validation study was performed on 100,000 sorted patient EV_PpIX and a cohort of patient and healthy total plasma EVs (n = 3, respectively). ddPCR for the GREM1 assay (E) was performed multiplexed with GAPDH (F) and expression levels were compared across biotypes.