Towards establishing extracellular vesicle-associated RNAs as biomarkers for HER2+ breast cancer

Bioreactor culture

To prevent bovine EVs present in foetal calf serum (FCS) from contaminating the cancer-specific EVs, we cultured BT-474 cells from ATCC (ATCC® HTB-20™) (seeded at 4.5 × 10⁸ cells/mL) in 15 mL Advanced DMEM/F-12 medium (Gibco, ThermoFisher Scientific, Waltham, USA) supplemented with 2% CDM-HD serum replacement (FiberCell Systems, New Market, USA) in the lower cell chamber of a CELLine AD 1000 bioreactor flask (Argos, Elgin, USA). The same media (150 mL) was used in the upper media chamber but supplemented with 2% FCS (Figure 1A). The dialysis membrane that separates the cell and media compartments allows FCS-specific nutrients <10 kDa but not EVs to pass through and nourish the cells. Every three to four days, the 15 mL of conditioned medium from the cell chamber was harvested for EV isolation, and the media from the upper chamber was replaced.

Figure 1. Purification and characterisation of BT-474 EVs.

(A) experimental procedure employed for extracellular vesicle (EV) production, isolation, and purification; (B) transmission electron microscopy image of a small EV; (C,D) hydrodynamic diameter distribution profiles of isolated large and small EVs measured by nanoparticle tracking analysis (NTA) wherein red vertical lines and blue numbers denote standard deviation and particle diameters at specific peaks, respectively; (E) EV concentration (empty squares) determined by NTA, and protein levels (filled squares) determined by BCA assay of fractions acquired during separation on a qEV Original size exclusion chromatography (SEC) column; and (F) immunoblot with antibodies specific for HER2, EpCAM, α-tubulin and TSG101 proteins. Tetraspanin TSG101 is a loading control expected to be present in both EVs and cells. The α-tubulin should be present only in the cell lysates but not in purified EVs. MDA-MB-231 cell lysate serves as the negative control for HER2 and EpCAM proteins. Representative images/data from three independent experiments were shown in B–F.

EV isolation and purification

EVs were isolated using differential centrifugation and size exclusion chromatography (SEC) as outlined in Figure 1. Conditioned medium (15 mL) was first centrifuged at 2,000 x g for 10 min to remove large debris, 10,000 x g for 30 min to isolate large EVs, and 100,000 x g for 70 min to isolate small EVs (Figure 1A). The resulting small EV suspension (in 500 µL PBS) was loaded onto a 35 nm qEV original size exclusion column (Izon, Christchurch, New Zealand), and fractions 7 through 24 were collected using an automated fraction collector (500 µL per fraction). BCA protein quantitation assay (Cat # 23225, Pierce, ThermoFisher Scientific, Waltham, USA) and Nanosight NS300 nanoparticle tracking analysis (NTA; Malvern Panalytical, Malvern, UK) were performed to quantitate protein and particle concentrations in each fraction, respectively. EV concentrations and size distributions were quantified on NTA by recording three 30 seconds videos under low flow conditions, with large EVs diluted at 1:100 in PBS and small EVs diluted at 1:500 in PBS. Small EV-rich fractions (7–11) were pooled, quantified again using NTA and BCA, and concentrated by ultracentrifugation (Avanti, Beckman Coulter, Brea, USA) at 100,000 x g for 70 min.

EV visualisation by transmission electron microscopy (TEM)

Negative staining TEM of pooled EV fractions was conducted by adsorption onto Formvar-coated copper grids (Electron Microscopy Sciences, Hatfield, USA) for 2 min, then treated with 2% uranyl acetate for 2 min. Grids were then visualised on a Tecnai G2 Spirit TWIN (FEI, Hillsboro, OR, USA) transmission electron microscope at 120 kV accelerating voltage and images were captured using a Morada digital camera (SIS GmbH, Munster, Germany).

Protein analysis by western blotting

This procedure was carried out as described previously²⁶. Breast cancer cell lines were grown to log-phase, washed twice with ice-cold PBS, and lysed in an sodium dodecyl sulphate (SDS) lysis buffer [60 mM Tris-HCl (pH 6.8 at 25°C), 2% (w/v) SDS, 10% glycerol]. Proteins (25 μg) were separated by SDS-polyacrylamide gel electrophoresis (PAGE) and transferred to PVDF membranes. Membranes were subsequently immunoblotted with antibodies recognising human HER2 (mouse monoclonal, anti-Neu, Santa Cruz, Cat # sc-33684, RRID:AB_627996), human EpCAM (rabbit monoclonal, AbCAM, Cat # ab223582, RRID:AB_2762366), human alpha-Tubulin (mouse monoclonal, Sigma-Aldrich Cat# T6074, RRID:AB_477582) and human TSG101 (rabbit polyclonal, AbCAM, Cat # ab30871, RRID:AB_2208084) and corresponding secondary antibodies. Bound antibodies were visualized using Pierce™ ECL Western Blotting Substrate (ThermoFisher Scientific, Waltham, USA) and the chemiluminescence was measured using a BioRad ChemiDoc MP imaging system (Bio-Rad Laboratories, Inc., Hercules, USA).

RNA quantitation by qRT-PCR

Technical triplicates of Trizol-purified RNA from each experimental condition were reverse transcribed into cDNA using qScript Flex cDNA kit (Cat # 95049, Quantabio, Beverly, USA) primed with equal molar ratio of oligo-dT and random primers according to the manufacturer’s instructions. Quantitative RT-PCR was carried out using SYBR Green MasterMix (Life Technologies, Carlsbad, USA) and gene-specific primers previously validated in the literature (Table 1). These included protein-coding mRNAs EpCAM²¹, BIRC5²², YBX1²³, GAPDH, and HPRT1, and lncRNAs ZFAS1¹⁷, HOTAIR¹⁹, and AGAP2-AS1²⁰. Three independent experiments were performed with duplicate PCR reactions per sample. RT-qPCR data were presented as cycle threshold (CT) values. Expression values were normalized relative to GAPDH mRNA expression. Statistical analysis was performed using multiple T-test.

Bioinformatic meta-analyses

For this meta-analysis, the “RSEM expected count (DESeq2 standardized)” dataset was downloaded on 31st March 2020 from the TCGA_GTEx_TARGET cohort included in the UCSC Xena portal (https://xenabrowser.net/datapages/) and was manually annotated. This procedure has resulted in a dataset called “Figure 2B and C_meta_analysis_rawdata.xlsx” deposited in the DRYAD Digital Repository and used for all subsequent analyses. All data manipulations, plotting and statistical analyses were carried out in R computing environment (v 3.5.3) running in R Studio (v 1.1.414) on a Windows 10 x64 machine. The ggplot2 package (v 3.3.0) was used to render Figures 2B and 2C. Magnitude of the gene expression difference between non-tumour breast tissues and breast tumours (Hedges g effect size) was calculated using the cohen.d function included in the effsize R package (v 0.8.0). The R script containing the code for all the above computations and visualisations is available in the DRYAD Digital Repository.

Figure 2. Bioinformatics meta-analysis of BT-474 extracellular vesicle (EV)-associated RNAs in tumour and non-tumour tissue.

(A) Mean mRNA abundance (Ct value) of five protein-coding genes (EpCAM, BIRC5, YBX1, GAPDH, HPRT1) and three long non-coding RNAs (ZFAS1, HOTAIR, AGAP2-AS1) in BT-474 cells and their EVs. Each data point represents an average Ct value obtained in a PCR experiment using technical duplicates of an independently prepared sample. Three independent experiments were performed. Error bars denote standard errors of the mean. (B) Comparison of RNA expression of the gene panel studied in (A) between human tumours and their respective non-tumour tissues deposited in TCGA and GTEx portals. Data were manually classified into 20 different organ categories (y-axis) including 8,867 samples across 28 different cancer types and 6,874 samples across 24 non-tumour tissue types. Colour and area of the circles represent median RNA abundance; darker and larger circles indicate higher RNA expression. (C) Distribution of RNA expression of studied genes in breast tumours and breast non-cancer tissues. Open diamonds denote means of each population. Hedges g effect sizes indicate a number of standard deviations that separates the tumour and non-tumour groups. Hedges g > 0.8 demonstrates large effect size, i.e., difference between the means clearly stands out from the “noise” within the groups.

An earlier version of this article can be found on bioRxiv (doi: https://doi.org/10.1101/2020.09.27.309252).

EV production and isolation

The CELLine AD 1000 bioreactor increased the cell density and EV production due to the unique growth surfaces and fluid interactions^25,27. In addition, the common issue of contaminating bovine EVs^28,29 was avoided by using the serum replacement CDM-HD, which is chemically defined, protein free, and animal component free. From three independent experiments, we obtained an average of 1.9 ± 0.3 × 10¹¹ large EVs of a mean diameter 150 ± 3 nm and 8.5 ± 0.7 × 10¹¹ small EVs of a mean diameter 127 ± 5 nm. Negative-stained transmission electron microscope imaging showed the expected round EV morphology, and NTA size distributions resemble those seen from EVs produced in conventional culture flasks (Figure 1B–D). Low levels of contaminating proteins were observed in fractions 11–24 due to 2% CDM-HD serum replacement instead of the standard 5–10% FCS (Figure 1E). This allowed the accurate quantification of EV-associated protein markers without the concern of contaminating cellular proteins and demonstrated that the small EVs obtained using ultracentrifugation are suitable for RNA analysis.

EV molecular characterization

Both the BT-474 cell lysates and BT-474 EVs of all sizes and purities isolated contained TSG101, EpCAM, and HER2 proteins (Figure 1F). Consistent with the literature, the triple-negative MDA-MB-231 breast cancer cell line did not express detectable levels of HER2 and EpCAM³⁰. TSG101 is a regulator of the endosomal sorting and trafficking process and is expected to be present in both cells and EVs³¹. EpCAM is a cell adhesion glycoprotein that has been used extensively as a liquid biopsy marker for several epithelial cancers³², whilst HER2 plays an important role in breast cancer subtyping. Interestingly, HER2-positive EVs appear to increase tumour proliferation and resistance to trastuzumab therapy³³.

Quantification of the abundance of several EV-associated RNAs, including protein-coding mRNAs EpCAM, BIRC5, YBX1, GAPDH, and HPRT, as well as lncRNAs ZFAS1, HOTAIR, and AGAP2-AS1, was then performed using RT-qPCR from small EVs purified by ultracentrifugation. Despite well-documented differential expression in breast cancer, EpCAM mRNA was not found to be associated with the BT-474 EVs, while BT-474 small EVs were clearly associated with established breast cancer-specific RNAs, including mRNA BIRC5 and lncRNA HOTAIR (Figure 2A). Apart from EpCAM, no significant difference (unpaired T-test) was found between cells and EVs in the RNA analysed (Figure S1).

Differential expression of selected RNAs in cancer and normal tissues

We then explored the expression of the identical set of RNAs in 15,741 tumour and non-tumour tissue samples included in The Cancer Genome Atlas (TCGA) and Genotype Tissue Expression (GTEx) databases, respectively. Tumour and non-tumour tissues in all 20 tissues analysed expressed similar levels of YBX1, GAPDH, HPRT1, ZFAS1, and AGAP2-AS1 RNAs. The result indicates a limited use of these RNAs for differentiating tumour and non-tumour EVs. This result is consistent with the canonical “housekeeping” role of HPRT1 and GAPDH and suggests potential use of ZFAS1 and AGAP2-AS1 as housekeeping genes for analyses of lncRNAs in samples including tumour and non-tumour tissues, as well as cultured cells. Of the six candidate biomarkers investigated in this study, only BIRC5²², EpCAM²¹ and lncRNA HOTAIR¹⁹ were found to be differentially expressed in a wide range of cancer types including breast cancer (Figure 2B and 2C).

Underlying data

DRYAD: Towards establishing extracellular vesicle-associated RNAs as biomarkers for HER2+ breast cancer. https://doi.org/10.5061/dryad.jdfn2z393⁴⁰.

This project contains the following underlying data:

Data are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).