Characterization of EVs released from RCC and immortalized benign epithelial kidney cell lines
EVs were isolated from CCM of ccRCC cells (786-O, 769-P, and Caki1), pRCC cells (ACHN and Caki2), and immortalized benign epithelial kidney cells (HK2 and RPTEC/TERT1). The number of isolated EVs per mL CCM as measured by nanoFCM varied per cell type (Figs. 1a, S2). When corrected for the total cell number at harvest, ACHN released the most EVs (1.4 × 107 EVs/million cells) and Caki1 secreted the least (5.6 × 104 EV/million cells) (Fig. 1a). The particle size-distribution was also measured (Fig. S2). The mean particle size for benign kidney EVs was 44.5 ± 1 nm, for ccRCC EVs was 59.5 ± 9 nm, and for pRCC EVs was 51.6 ± 6 nm. EVs were negatively stained and imaged by TEM, which demonstrated intact membranous vesicles for all cell types recognizable by cup-shape on wide-field and close-up (Fig. 1b, c).
EVs were analyzed for presence of established EV markers (FLOT1, CD81, CD63, and CD9) and the non-EV cellular marker calnexin (Fig. 2a). Calnexin was negative in all EV samples, indicating absence of ER-protein, a frequently detected non-EV contaminant. While relative abundance of EV markers varied across samples, at least three EV-enriched markers were present in all samples, confirming presence of EVs. Flotillin-1 was highly abundant in EVs from HK2, Caki2, RPTEC, 786-O, and ACHN, while it was of lower abundance in 769-P and very low in Caki1. CD81 was highly abundant in HK2, Caki2, 768O, RPTEC, and 769-P, while low in ACHN and very low in Caki1. CD63 was the highest in HK2 and Caki2, followed by RPTEC, 786-O, and ACHN. 769-P and Caki1 had low CD63 abundance. CD9 was only detected in EVs from RPTEC, Caki2, and ACHN EVs.
Spatial proteomics confirms presence of EV protein markers
While western blot is a standard in the EV field, detection of positive EV markers is limited by sensitivity and sample input. To robustly characterize the protein cargo of CCM kidney EVs, a previously developed spatial proteomic approach [29] to examine the differential abundance of proteins was adapted across three sample preparations: EV, cell lysate, and debris (“2 k”) pellet. Following sample lysis and proteolytic digestion, the peptides from six renal cell lines (786-O, 769-P, ACHN, CAKI2, CAKI1, and HK2) were subjected to tandem-mass-tagging isobaric labeling followed by tandem mass spectrometry analysis. A pooled reference channel composed of a mix of cell lysates from each of the individual cell lines was included to link the two TMT-plexes (Fig. S1c). In total, 1,967 proteins were identified with quantification information in the combined search of both TMT-plexes, including 1726 proteins quantified across the two TMT-plexes (Table S1).
The differential abundance of classic EV markers FLOT1, CD81, CD63, and CD9 were also tested by western blot, as well as additional EV markers ALIX and TSG101 in EV fractions relative to cell lysate (Fig. 2b) [22]. Using a threshold cutoff of FC > 1.5 (log2 FC > 0.58), all the EV markers were found to be elevated in EVs from HK2, 786-O, and CAKI2 renal cell lines and four of the six markers to be increased in 769-P, ACHN, and CAKI1 EVs. Whereas CD81 and CD63 had increased abundance in all the EV fractions relative to the cell lysate, ALIX, CD9, TSG101, and FLOT1, while robustly detected, did not always display increased abundance in the EV fractions relative to the cell lysates. This reflects the variable abundance of these proteins in the individual renal cell lines and EV fractions, as well as the potential variable distribution of EV sub-populations [30, 31]. While negative by western blot, CD9 was detected by mass spectrometry in EV samples from CCM of HK2, 786-O, 769-P, ACHN, and CAKI1. CD81, CD63, and Flot1 abundance was the lowest in Caki1, indicating that while the EV proteins were present in the Caki1 sample, they were likely at too low level to be detected by western blot.
In addition to positive EV markers, the differential abundance of several proteins classically considered to be negative in EVs was also evaluated, including CANX (also measured by western blot), HSP90B1, and CYC1 [22]. Using a threshold cutoff of FC > 1.5 (log2 FC > 0.58) between the cell lysate and EV fraction, CANX was found to be the only negative EV marker decreased in all EV samples, following the same expression pattern as shown by western blot, while HSP90B1 was decreased in EVs from five out of the six cell lines.
Finally, to specifically assess the enrichment of EV-specific proteins (i.e., those uniquely present or highly enriched in the EV fraction and absent or depleted in the cell lysate or 2 k debris fraction), proteins with high abundance in EV vs cell lysate and/or in EV vs 2 k debris fraction (log2 FC > 0.58) were considered. Following these criteria, 1279 proteins were annotated that were EV-associated in at least one cell line, 420 proteins in four, and 284 proteins in five cell lines (Table S2). 186 proteins were EV-associated in all six cell line sample sets. Using the bioinformatic tool DAVID [32, 33], gene ontology (GO) analysis of the 186 EV-specific proteins identified in all samples and found enrichment for cell component GO annotations “extracellular exosome,” “extracellular matrix,” “focal adhesion,” and “extracellular vesicle” (FDR < 0.05; Fig. 2c, Table S3). The biological process GO terms “extracellular matrix organization,” “cell adhesion,” and “platelet degranulation” were the most prominent biological process annotations for the EV-associated proteins on our dataset (FDR < 0.05; Table S4).
The quantitation information of the cell lysates from the individual cell lines was leveraged to determine the concordance of protein abundance patterns in the cellular and EV fractions. Focusing on the differential abundance of proteins in RCC cell lines and the HK2 cell line (Tables S5 and S6), concordant trends (displaying either increased abundance or decreased abundance) including 76 proteins in ACHN cells/EVs, 59 proteins in CAKI2 cells/EVs, 47 proteins in 786-O cells/EVs, 70 proteins in 769-P cells/EVs, and 39 proteins in CAKI1 cells/EVs were observed (Table S6). These results provide some evidence of EVs reflecting the molecular characteristics of their respective parental cells [34].
Overall, these results confirm that the SEC approach was successful for enriching EVs, and subsequently EV-associated proteins, while reducing the abundance of several proteins commonly categorized as negative markers of EVs.
Consensus proteomic signature of RCC EV
To better understand the potential biological impact of RCC-derived EVs, the renal cell lines were grouped based on their respective histology: pRCC (ACHN and CAKI2), ccRCC (786-O, 769-P, and CAKI1), and a single benign cell line (HK2) [28]. Using a FC > 1.5 (log2 FC > 0.58) threshold, the differential abundance of EV-associated proteins between each of these groups was examined. Notably, direct pair-wise comparison of the cell line-derived EVs revealed a high degree of heterogeneity in protein abundance, emphasizing the heterogeneity of EV cargo itself, even within a single histology type (Tables S5 and S7).
On comparing pRCC EV cargo (n = 2 cell lines) to benign (n = 1 cell line), 31 proteins were found to be increased and 88 proteins were decreased in abundance in EVs from pRCC cell lines compared to EVs from the HK2 benign cell line (Table S5). Using the bioinformatic tool WebGestalt [35, 36] for pathway analysis revealed disparate signaling-related proteins in each of the pRCC EVs, including NOTCH-regulating proteins increased in abundance in EVs derived from ACHN cells, and MET signaling-regulating proteins increased in abundance in the EVs derived from CAKI2 cells (FDR < 0.05; Table S8). Commonly enriched proteins increased in abundance in pRCC EVs were associated with the proteasome degradation (Table S9). More overlaps were observed in the annotated pathways for the proteins increased in HK2 EVs relative to pRCC EVs, including extracellular matrix organization, focal adhesion, endocytosis, and vesicle trafficking (Tables S8 and S9).
Analysis of the differential abundance of proteins derived from EV released from ccRCC cell lines (n = 3 cell lines) and the benign HK2 cell line revealed that 93 proteins were decreased and 34 were increased in the EVs derived from all three ccRCC cell (Table S5). Cellular processes associated with CCT/TriC protein folding and metabolism were increased in ccRCC EVs relative to HK2-derived EVs, whereas proteins associated with neutrophil degranulation and ERBB2 signaling were decreased in ccRCC EVs (FDR < 0.05; Tables S8 and S9).
Finally, the differential abundance of proteins in EV derived from pRCC cell lines (n = 2 cell lines) and ccRCC cell lines (n = 3 cell lines) was assessed. Overall, there were six differentially abundant EV proteins between ccRCC and pRCC: five proteins were increased in EVs from all pRCC cell lines assessed relative to all ccRCC EVs (EDIL3, GC, HBA1, LTF, OLFML2B), while only one protein, HTRA1, was increased in all ccRCC-derived EV samples relative to pRCC EVs (Table S7). Pathway analysis of the proteins found to be differentially abundant between the pRCC EVs and ccRCC EVs revealed disparate pathways in each of the individual RCC EV comparisons (Tables S10 and S11), albeit with ECM–receptor interaction and focal adhesion being commonly annotated across multiple comparisons.
To specifically identify candidate ccRCC or pRCC EV cargo protein biomarkers, the proteins enriched in EV samples were assessed, regardless of relative abundance. To be included in the analysis, a protein must be detected in the EV preparation in all cell lines of the pathological type (i.e., for ccRCC, 786O, 769P, and CAKI1; for pRCC, ACHN, and CAKI2; for benign, there is a single cell line, EVs from the single cell line HK2). 181 proteins were detected in EVs released from all renal cell lines: the benign epithelial kidney cell lines, both pRCC cell lines, and three ccRCC cell lines. While a large number of proteins were enriched in EV from the immortalized benign epithelial cell line, fewer were specifically enriched in RCC EVs. 34 proteins were enriched in pRCC EVs alone, while 20 were uniquely enriched in EV released from ccRCC cells (Fig. 4a, Table S12). These protein sets represent candidate EV biomarkers to discriminate both RCC from benign renal cells as well as discriminate RCC subtype.
The consensus ccRCC EV mRNA cargo signature
Total RNA from EVs isolated from seven cell lines (immortalized benign renal cells RPTEC and HK2; ccRCC 786O, 769P, and CAKI1; pRCC ACHN and CAKI2) were run in triplicate using the nCounter PanCancer Progression Panel and data were normalized by total RNA content. Following quality control, 1 replicate of HK2 was excluded due to a technical hybridization error and samples CAKI1 and CAKI2 were excluded due to insufficient RNA input. Of the 770 mRNA transcripts assessed, 461 were detected in one or more samples of CCM EVs, and 159 were present in all five cell line CCM EVs (Table S14). Unsupervised hierarchical clustering of the top 100 most differentially abundant mRNA transcripts indicated that benign renal, ccRCC, and pRCC cell line EVs had distinct mRNA cargos (Fig. 3a). Analysis after quality control was limited to a single pRCC sample, not allowing further analyses assessing pRCC-specific mRNA cargo: such research will be the focus of future work.
Differential gene expression comparing EVs released from ccRCC cell lines and benign renal cell lines demonstrated that RELN, CDH2, SERPINE1, STAB1, VCAM1, KCNJ8, SRGNR, EREG, COL6A3, and GDF15 were the 10 most enriched genes in RCC EVs, while HIF1A, KRT19, HOXA7, AGR2, LAD1, EPCAM, LAMA4, MUC1, SCNN1A, and PTRF were the 10 most enriched genes in benign EVs (Fig. 3b, Table S13). To start investigating the possible biologic functions of EVs released from cancer vs benign renal cells, GSEA was utilized to assess the NanoString-defined gene sets related to cancer progression. The NanoString gene set Metastasis Suppressors was significantly enriched in ccRCC EVs (Fig. 3c). Conversely, the two NanoString gene sets Epithelial in EMT spectrum and Basal Lamina were significantly enriched in benign EVs (Fig. 3c).
To identify candidate ccRCC mRNA EV cargo biomarkers, mRNA transcripts found in benign renal cell EVs and ccRCC EVs were directly compared, regardless of abundance. To be scored that the mRNA had to be detected in the majority of replicates per cell line and present/absent in both cell lines per type. 170 mRNA transcripts were present in EVs released from both the benign epithelial renal cell lines and the ccRCC cell lines (Table S14). Ten mRNAs were unique to benign epithelial kidney EVs and absent in EVs released from ccRCC: EPCAM, PRKCZ, PXDN, CXADR, EPS8L1, HOXA7, LAD1, MYO1D, ROCK2, and SLC35A3. Eight mRNAs were found only in ccRCC: CDH2, COL7A1, FGFR2, BMPR1B, HDHD3, ICAM1, KIAA1462, and PFKFB4 (Fig. 4b).