Abu Ali Ibn
Sina
a,
Ting-Yun
Lin
a,
Ramanathan
Vaidyanathan‡
a,
Zhaoran
Wang
a,
Shuvashis
Dey
a,
Jing
Wang
a,
Andreas
Behren
b,
Alain
Wuethrich
a,
Laura G.
Carrascosa
*a and
Matt
Trau
*ac
aCentre for Personalised Nanomedicine, Australian Institute for Bioengineering and Nanotechnology (AIBN), The University of Queensland, Brisbane, QLD 4072, Australia. E-mail: lgcarrascosa@uq.edu.au; m.trau@uq.edu.au
bOlivia Newton-John Cancer Research Institute, La Trobe School of Cancer Medicine, Heidelberg, Victoria, Australia
cSchool of Chemistry and Molecular Biosciences, The University of Queensland, Brisbane, QLD 4072, Australia
First published on 4th June 2020
Extracellular vesicles (EV) play a major role in intercellular communication by transmitting cellular materials (e.g. protein, RNA) among distant cells. Recent evidence suggests that they could also contribute to carrying DNA which could inform on the mutational status of the parent tumour DNA. Thus, the fundamental analysis of evDNA could open a better understanding of tumour metastasis and provide new pathways for noninvasive detection and monitoring of cancer. To explore the potential of evDNA for diagnostics, the isolation of pure evDNA from body fluids free of cfDNA contamination is crucial. Herein, we use a liposome based model system to develop an improved evDNA isolation protocol free from cfDNA contamination and evaluate the methylation dependent physicochemical properties of evDNA to develop a simple test for detecting cancer evDNA. Using a highly sensitive multiplex microelectrode device, we demonstrate that serum-evDNA derived from cancer patients show different solution and surface based properties than normal evDNA due to their different methylation landscape (i.e. methylscape). This microdevice allows simultaneous analysis of multiple samples in a single platform from as low as 500 pg μL−1 of evDNA.
New conceptsExtracellular vesicles (EV) show promise in developing potential liquid biopsy tests for cancer due to their ability to transmit tumor-associated biomolecules such as RNA, proteins, and DNA. Although there is a wealth of research on EV-derived RNA and proteins, only a few works have been done on EV-associated DNA (evDNA), which mostly indicated their existence in EV. Fundamental analysis of evDNA thus could open an entirely new understanding of genetic and epigenetic profiling of parent tumours. In this study, we develop a liposome-based model system to isolate pure evDNA and characterise their methylation dependent physicochemical properties. We find that evDNAs have similar solution and surface-based properties to cellular gDNA, which indicates that they might carry similar methylation levels and patterns like their parent cell gDNA. However, evDNA shows significantly different surface-based properties compared to cfDNA, probably due to their longer size than cfDNA. These findings may shed light on the metastatic mechanism of parent tumour to a distant location via EV. Finally, we show that these physicochemical properties of evDNA can be utilised to develop a simple and multiplex liquid biopsy test for cancer. |
During cancer progression, DNA undergoes epigenetic reprogramming, particularly in their methylation pattern. Normal DNA is usually unmethylated at CpG rich regions (e.g. CpG islands) but heavily methylated across the genome. In contrast, cancer cell DNA features a unique methylation landscape comprising cluster methylation at CpG rich regions and global hypomethylation at the intergenic regions of the genome.20–22 This can significantly influence its global and local physicochemical properties such as flexibility, hydrophobicity, and adsorption towards the metal surface (e.g. gold).20,23–26 Recently, we have discovered that cancer DNA exhibits dispersed orientation in the solution that significantly enhances the adsorption towards the gold surface.27 In contrast, normal DNA displays aggregated orientation that inhibits their adsorption towards gold. Thus, the study of evDNA could be fascinating to understand its methylation dependent physicochemical properties and provide further information about the potential of evDNA as a noninvasive cancer biomarker. However, it is still arguable that the reported evDNA is actually a different entity than cfDNA. This is because DNA tends to stick on any surfaces and hence cfDNA could attach with the lipid envelope of microvesicles and this could result in co-isolation of cfDNA with the EV during the purification protocol.28 This possibility is fueled by the questions regarding unknown biological role of evDNA and related mechanism of EVs-encapsulation. Thus, there is a need to develop a better strategy for isolating evDNA without cfDNA contamination and their characterization for elucidating clinical information.
Herein, we develop a liposome based model system that could provide a framework for the isolation of evDNA free from cf-DNA contamination and perform a fundamental study to understand the physicochemical properties of evDNA from cancer and normal cells. By spiking fragmented DNA to liposome solution, we find that the successful digestion of extra-vesicular DNA (i.e. cfDNA) depends on the vesicle concentration. By optimizing the concentration of liposome and digestion protocol, this model system exhibits high efficiency to digest extravesicular DNA from liposome solution as well as cell derived EVs. Analysis of evDNA reveals that the physicochemical properties of evDNA are similar to that previously observed for cellular gDNA.27 Our data suggest that the evDNA possibly carries a similar methylation pattern to gDNA. Solution based properties of evDNA show that normal evDNA aggregates more in solution probably due to the high level of overall methylation while epigenetically reprogrammed cancer evDNA shows uniform dispersion due to the overall hypomethylation across the genome. Consequently, surface-based properties show that normal evDNA has lower adsorption towards the gold surface in comparison to the cancer evDNA. However, the gold adsorption behaviour of evDNA is found to be similar to cellular gDNA but quite different from the cfDNA. Using a newly designed 16-plex microdevice, this investigation successfully detects cancerous evDNA derived from the serum of melanoma and breast cancer patients. The method is quick and can detect multiple cancerous evDNA from as low as 500 pg μL−1 of a sample. We believe that this study will open a new avenue to better understand the role of EVs in cancer progression and at the same time provide an alternative liquid biopsy test for cancer.
To digest all the contaminating cfDNAs before the EV isolation, we constructed a model system using liposomes which are synthetic vesicles with one or more phospholipid bilayer and finely resemble EV in terms of membrane composition and vesicle size.29,30 A known concentration of DNA was spiked into a known concentration of liposome solution and DNase I enzyme was used to completely digest the spiked DNA present in the liposome solution following the manufacturer's instruction. As shown in the gel picture at Fig. 2A(iv), 100 ng μL−1 of BT474 cell derived DNA was completely digested by DNase I enzyme. However, the same amount of DNA could not be digested completely in the presence of 1 × 1013 particles per mL of liposomes (Fig. 2A(iii)). This indicates that the presence of liposome probably reduced the enzyme activity and decreased digestion efficiency. However, the type of DNA that is likely to contaminate EVs is cell free DNA (cfDNA) rather than gDNA, which typically has an average size of 200 bp. Thus, to mimic the cell free DNA and to facilitate the digestion, we fragmented the BT474 cell derived gDNA up to 200 bp by sonication and then spiked in liposome solution. The results in Fig. 2B(ii) show that with the addition of fragmented DNA (fgDNA), there is no visible band on the gel after digestion, indicating an enhancement of the digestion in the presence of liposome. However, there was also a possibility that the electrophoresis gel had not enough sensitivity to visualize trace amounts of undigested fgDNA. To address this point, we decided to run a polymerase chain reaction (PCR) which targeted a repetitive region that is distributed across the whole genome, so the existence of any trace amount of DNA can be amplified and detected; if no DNA exists after digestion, there will be no amplicon showing on the gel. As shown in Fig. 2C, it is evident from the band for different concentrations of amplified DNA that even after addition of low amounts of fgDNA (i.e.Fig. 2C(iv), 0.3 ng μL−1) in liposome solution; the digestion enzyme could not eliminate the DNA. It appears that the liposomes even provided protection for fragmented DNA and prevented them from complete digestion. Therefore, we sought to investigate whether decreasing the concentration of liposomes in the digestion reaction could facilitate DNAse enzyme to have better access to the fgDNA and ensure the full digestion. Consequently, we performed the digestion on a series of samples entailing several liposome dilutions from the liposome stock (1 × 1013 particles per mL) including 1 in 200, 1 in 500, and 1 in 1000 μL diluted liposomes. Our results in Fig. 2D demonstrate that for the 1 in 1000 μL diluted liposomes (Fig. 2D(vii)), the PCR reaction provided no amplification, indicating that the full digestion of DNA was achieved under these conditions. The corresponding band for each control PCR experiments including (i) DNA + PCR master solution (PCR positive control) (ii) No DNA + PCR master solution(PCR negative control) (iii) DNA + Liposome + PCR master solution (i.e. Positive control) (iv) No DNA + Liposome + PCR master solution, confirms the specificity of the experiments and the absence of any contamination during PCR. This liposome concentration (1/1000 dilution) corresponds to approximately 1 × 1010 particles per mL was therefore selected for the next set of experiments.
To evaluate the performance of the model system in EV samples, we diluted the original stock of BT474 EVs equal to the optimized concentration of liposome (1 × 1010 particles per mL) and performed digestion of 1 ng μL−1 of spiked-in fgDNA and then amplified by PCR as previously did for the liposome system. The absence of any band after the PCR as evident in Fig. 2E(v), suggests that the spiked DNA was fully digested by the enzyme. This outcome indicates that cfDNA from EVs can be completely digested using DNAse I enzyme in optimized conditions where the EVs are partially diluted.
After successful digestion of extravesicular DNA, we sought to extract the EV encapsulated DNA from our isolated EV samples. We lysed the EVs using lysis buffer and extracted the DNA using standard phenol–chloroform extraction procedures. The purity of the DNA was confirmed by measuring 260/280 absorbance ratio using Nanodrop spectrophotometer and the concentration was measured by Qubit Fluorometer. The average length of the DNA in base pairs was also measured by Agilent Bio-analyzer. As shown in Fig. S3 (ESI†), the average size of evDNA is significantly different from the average size of gDNA and cfDNA. While the gDNA was very high molecular weight with an average size above 10 kb, the evDNA size was found 1 to 4 kb long (Fig. S3, ESI†). Moreover, the cfDNA average size is known to be only 50–200 bp. These data support previous reports suggesting evDNAs are fundamentally different in nature. Thus, we speculate that evDNA may exhibit different methylation dependent physicochemical properties in comparison to the gDNA and cfDNA.
To investigate the methylation dependent surface adsorption properties of evDNA, we used electrochemical assays to quantitatively assess the adsorption levels of evDNA. The electrochemical assay involved the direct adsorption of 5 μL of purified evDNA (10 ng μL−1 concentration in SSC5X buffer at neutral pH) onto gold electrodes for 10 min. Subsequently, the adsorption level was quantified by Differential Pulse Voltammetry (DPV) in presence of the [Fe(CN)6]3−/4− redox system (Fig. 1, see methods section for details). Upon adsorption of DNA on gold electrodes, [Fe(CN)6]3−/4− redox system generates a faradaic current signal, which is proportionally lower than the bare electrode signals (i.e., the greater the DNA adsorption is, the larger the relative current signal difference, %ir, with respect to the original baseline). This system has already been reported previously by us and can reliably quantify the adsorption competence of biomolecules (e.g. DNA, RNA, Protein) adsorbed onto the gold surface.31–34 We initially optimized the adsorption time and concentration of evDNA to get a maximum difference between methylated cancer evDNA (i.e. BT474 cell derived evDNA) and fully unmethylated WGA evDNA (prepared by whole genome amplification of BT474 evDNA) in which methylation marks were completely removed by whole genome amplification (see Methods in ESI† for details). As shown in Fig. S4 (ESI†), only 5 μL of 10 ng μL−1 evDNA after 10 minutes of adsorption provided maximum difference between WGA evDNA and BT474 evDNA. An interesting feature of this experiment is that the removal of methylation marks from BT474 evDNA by whole genome amplification significantly dropped the adsorption of this DNA. In our previous study, we have seen that gDNA with different methylation levels had different adsorption properties towards the gold surface. The data for the adsorption of WGA evDNA and BT474 evDNA also support our previous hypothesis and suggests that evDNA methylation has a significant impact on their gold adsorption properties.
Next, we compared the adsorption of different cancer cell lines derived evDNA and gDNA (breast: BT474, MCF7; lung: H1975; melanoma: SKMEL28). As can be seen in Fig. 3A, the relative DPV current for cancer cell derived evDNA (i.e. %ir for BT474 evDNA = 49.5 ± 1.03, MCF7 evDNA = 39.47 ± 0.97, H1975 evDNA = 36.18 ± 1.24, and SKMEL28 evDNA = 34.59 ± 1.67) is similar to cancer cell derived gDNA (i.e. %ir for BT474 = 41.58 ± 0.87, MCF7 = 40.79 ± 1.09, H1975 = 26.15 ± 1.48, and SKMEL28 = 31.27 ± 1.61). This data indicates that the similar adsorption of evDNA and gDNA towards gold electrode is probably due to the presence of similar methylation profiles in cancer evDNA and gDNA. Moreover, these evDNAs and gDNAs were extracted from the same cell lines which further indicates the possibility of carrying a similar cancer specific methylation profile. We also measured the adsorption of normal Human Mammalian Epithelial cell (HMEC) line DNA which provided significant lower current that both cell line and EV derived gDNA. This data supports our previous study where we demonstrated that the cancer Methylscape (cluster methylation with hypomethylated tail) could significantly increase the gold adsorption of gDNA in comparison to the unmethylated or heavily methylated normal DNA.27 Thus, we hypothesize that cancer evDNA probably carries a similar methylation landscape (i.e. Methylscape) like cancer gDNA which increases their adsorption towards the gold surface. To further understand the methylation profile of evDNA and gDNA, we performed methylation analysis. The data in Table S1 (ESI†) shows that evDNA derived from BT474 breast cancer and SKMEL28 melanoma cell shows similar hypomethylation level to BT474 and melanoma gDNA (Methylation level for BT474 evDNA = 26.75%, MCF7 evDNA = 43.59%, SKMEL28 evDNA = 37.84% vs. BT474 gDNA = 43.15%, MCF7 gDNA = 33.56% and SKMEL28 gDNA = 45.23%).
To investigate the solution and surface-based properties of evDNA, we extracted cancer and normal evDNA from the serum of a patient and a healthy individual. To study the solution-based properties, we performed transmission electron microscopy (TEM). As shown in the TEM images in Fig. 3B, normal evDNA derived from serum of a healthy individual tended to aggregate in solution wherein, evDNA derived from the serum of a breast cancer patient provided dispersion in solution. To test the surface adsorption profile of patient serum derived evDNA, we performed atomic force microscopy (AFM). To avoid the drying effect, which is a common limitation of TEM, we performed liquid AFM analysis of normal evDNA and cancer eVDNA using ultra flat gold surface. As shown in Fig. 3C and D, normal evDNA had very less interaction with the gold surface in comparison to the cancer evDNA. Cancer evDNA showed significantly higher adsorption which increased the surface roughness of the gold (Fig. 3D) However, we couldn’t compare the aggregation behavior of cancer vs. normal DNA in AFM probably due to the low concentration and low adsorption of normal evDNA.
The solution and surface-based properties of evDNA resemble our previous observation with gDNA where we have seen that levels and distribution of methylation can significantly influence the solution properties of gDNA.27 Similar to our previous cellular gDNA study,27 the solution-based properties of evDNA showed higher aggregation in normal evDNA. This is attributed to the hydrophobicity of the methyl group. As observed in our methylation analysis, normal evDNA has high level of overall methylation like cellular gDNA (Table S1, ESI†). Thus, we hypothesize that normal evDNA with a high level of overall methylation increased the hydrophobicity of the evDNA and therefore created large aggregates in aqueous solution. In contrast, epigenetically reprogrammed cancer gDNA were found to be uniformly dispersed in solution. This is probably due to the overall hypomethylation across the genome for cancer evDNA (Table S1, ESI†) which reduced the overall hydrophobicity of cancer DNA in solution. We believe that these solution-based properties also influenced the surface-based properties of evDNA. We have previously observed that methyl groups increased the adsorption of gDNA towards the gold surface.27 However, our AFM data shows that normal evDNA with higher methylation levels has lower adsorption towards the gold surface in comparison to the cancer evDNA. This is probably because of the higher aggregation of normal evDNA in solution which blocked the high affinity groups (e.g. methyl groups) of evDNA inside the aggregates and thereby prevented adsorption. In contrast, the cancer evDNA was more dispersed in solution and the high affinity methyl groups are more available for interaction with the gold surface. Moreover, cancer evDNA might comprise of cluster methylated CpG island which could act as an anchor site for evDNA to adsorb towards the gold surface.
To test the surface adsorption profile of patient serum derived evDNA and to develop a diagnostic platform using electrochemistry, we isolated evDNA from the serum of 5 breast cancer patients, 4 melanoma cancer patients and 7 healthy individuals (see ESI† Table S2 for clinical information) using our optimised model isolation procedure. However, the concentration of isolated evDNA from patient serum was very low which limited the ability to compare their adsorption profile using regular macroelectrodes. Thus, we carefully designed a microdevice comprising an array of 16 microelectrodes (see Methods in ESI† for detail fabrication process) which enabled the detection of samples with low concentration and at the same time provided multiplexing capability. The diameter of these microelectrodes is considerably smaller (i.e. 250 μm) than the conventional macroelectrodes (e.g. diameter = 3 mm). As reported in previous studies, these microelectrodes can significantly enhance the detection sensitivity due to the lower surface area of the electrodes and higher signal to noise ratio.35,36 To test the sensitivity of the device, we first performed adsorption profiling of cell line derived gDNA and evDNA using the microchip. As shown in Fig. 4A, only 500 pg μL−1 of gDNA and evDNA from different cancer and normal cell lines provided significant adsorption providing higher relative DPV current (i.e. %ir) which is similar to the results obtained in the case of macroelectrodes.
Next, we isolated both evDNA and cfDNA from the same patient serum samples and performed their adsorption profiling using the microchip. As shown in the Fig. 4B, 500 pg μL−1 of evDNA provided significantly lower relative DPV current than the same concentration of cfDNA (%ir for evDNAP1 = 23.51 ± 0.93, evDNAP2 = 39.72 ± 1.12 vs. cfDNAP1 = 78.83 ± 1.03 and cfDNAP2 = 82.01 ± 0.84). These data also support our previous study where we have seen that cfDNA has higher adsorption in comparison to the gDNA probably due to their shorter length. These data also suggest that the evDNA is a different entity with its own characteristic methylation and adsorption profile. Finally, to evaluate the potential of evDNA as a biomarker for noninvasive cancer detection, we performed experiments with the isolated evDNA derived from the serum of breast and melanoma cancer patients and healthy individuals using the multiplex microchip. The data in Fig. 4C shows that the relative DPV currents (i.e. %ir) for the adsorption of patient serum derived evDNAs are ranging from 20–40% whereas the relative DPV currents for the normal serum derived evDNAs are below 20%. These data suggest that the patient serum derived evDNAs possibly carry the similar Methylscape pattern as overserved in our previous study for gDNAs and may potentially enable liquid biopsy test for cancer. We also believe that the multiplexing capability, high sensitivity and specificity in identifying cancer and normal evDNAs may potentially make this device suitable for clinical application.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d0nh00258e |
‡ Present address: School of Chemistry, University of New South Wales, NSW 2052, Australia. |
This journal is © The Royal Society of Chemistry 2020 |