Tumor cells distribution (intra-assay precision and accuracy) analyses.

In order to verify the random distribution of CTC on the five spots located in the large compartment of the Rarecells^® Block, cultured A549 cells were made fluorescent and precisely counted as described in section 6. An aliquot of 100 μL of the diluted suspension containing 50 cells was added to 5 mL of healthy donor's blood and processed by filtration (Fig 1A).

The number of tumor cells found on each spot (each corresponding to the filtration of 1 mL of blood) was recorded.

For each experiment, we calculated the average number of tumor cells found if considering randomly only 1 spot, only 2 spots, only 3 spot and only 4 spots. We analyzed all possible single spots or randomly grouped spots (any 1, any 2, any 3, any 4 spots) and calculated corresponding mean values and their standard deviations. Intra-assay precision (percent coefficient of variation, %CV) and accuracy (%Error) were calculated (formula given in section 6).

Counting of residual leukocytes.

In order to evaluate the number of remaining White Blood Cells (WBC), one mL of blood from four different healthy donors was filtered and stained with Hematoxylin and Eosin. Leucocytes were then carefully counted under microscope.

Counting of residual leukocytes.

We assessed the number of remaining leukocytes after live tumor cells enrichment using 2 x 5 mL of blood obtained from 2 distinct healthy donors in 2 separate tests, and using a TC20™ Automated Cell Counter (Bio Rad, USA) after cell staining with Trypan Blue.

Elimination of residual leukocytes after live tumor cells enrichment by ISET^®.

We tested the feasibility of eliminating the residual leucocytes after live tumor cells enrichment by the ISET^® system. To this aim, the live tumor cells samples obtained with the ISET^® system were treated with the manual column-free method EasySep^TM and the ‘Human CD45 Depletion kit’ (StemCell Technologies, Canada) that contains a combination of monoclonal antibodies against the leucocyte marker CD45 [36]. This kit is originally intended for use with whole blood after Ficoll-type density centrifugation, a step that is known to lead to highly significant tumor cells loss [37, 38]. In order to adapt this kit to the protocol for isolation of live tumor cells by ISET^®, the amount of antibody and nanoparticles was reduced by 4-fold while the other manufacturer’s recommendations were followed. In brief, CD45 positive cells were specifically labeled with dextran-coated magnetic nanoparticles using bispecific Tetrameric Antibody Complexes (TAC). Tumor cells were then recovered after two successive CD45 depletions using the EasySep^TM Purple Magnet.

Live tumor cells' number, size and viability after their enrichment by ISET^®.

Cells were prepared and carefully counted as described in section 6. A fraction of the cell suspension (typically 100 μL) was mixed with 1 mL of blood from a healthy donor, or 1 mL of PBS, diluted tenfold with the Rarecells^® Live Cells Buffer and kept at room temperature for 5 minutes, or 3 minutes when cells were treated without blood, under gentle constant stirring. Filtration was then performed with the new protocol as described above. For comparison of cell counts, sizes and viability on the TC20™ Automated Cell Counter (Bio Rad, USA), 10 μL of either the unfiltered cells or the enriched fraction recovered after filtration were diluted 1:2 with Trypan Blue and loaded into a counting slide (Dual chamber for cell counter, BioRad, USA). Three or more counts were performed to obtain consistent counting values.

Live tumor cells antigen expression assessment before and after their enrichment by ISET^®.

We assessed the potential impact of this new ISET^® filtration protocol on cell surface antigens' expression by fluorescent labeling of the Epithelial Cell Adhesion Marker (EpCAM) on MCF-7 cultured cells before and after live cell enrichment by ISET^®. MCF-7 cells in culture were first incubated with 5 μg/mL of Hoechst 33342 (Sigma-Aldrich, USA) for 2 hours before being rinsed twice with PBS, detached from the flask by incubation with trypsin (0.125%) and resuspended in 5 mL of culture media as described in section 6. A fraction of the cell suspension (typically 100 μL) was mixed with 1 mL of blood from a healthy donor, diluted 1:10 with the Rarecells^® Live Cells Buffer and kept at room temperature for 5 minutes under gentle constant stirring. The enriched fraction recovered after live cells filtration was centrifuged for 2 minutes at 150 g. The supernatant was carefully eliminated leaving 100 μL of suspension. Fluorescent labeling of EpCAM was performed on both the unfiltered cells and the enriched fraction recovered after ISET^® filtration. 20 μL of Human FcR binding inhibitor (eBioscience, USA) were added to each 100 μL suspension and kept in the dark at 4°C for 20 minutes. The Alexa Fluor 488 conjugated anti-EpCAM antibody clone VU1D9 (Cell Signaling Technology, USA) was diluted 1 to 30 in PBS and 20 μL of that dilution were then added to each sample followed by a 30 minutes incubation in the dark at 4°C. After washing with 1 mL of PBS, the cells were centrifuged at 300 g for 4 minutes and transferred to a microchamber (Ibidi, Germany). Microscopy evaluation was performed with an epi-fluorescence microscope (Zeiss, Germany). The DAPI filter (emission at 465 nm) and the FITC filter (emission at 525 nm) were used for the detection of the Hoechst and the EpCAM signals, respectively. Images were processed with Image J software. To perform quantitative analysis of EpCAM fluorescence, we analyzed two consecutive optical fields and measured the cell mean fluorescence, the cell area and the integrated density (fluorescence per area) for each cell (50 MCF-7 cells before and 50 MCF-7 cells after live cell enrichment). We then calculated the corrected total cell fluorescence (CTCF).

CTCF = Integrated_density–(Area_of_selected_cell x mean_background_fluorescence). Furthermore we applied cutoffs at 91000 units and 200000 units of CTCF based on the calculations of the first and third quartiles of each data set.

Live tumor cells growth after their isolation by ISET^®.

We performed in vitro culture experiments of live tumor cells enriched from blood by ISET^®. In two separate tests with 3 replicates, 10⁴ A549 cells were added to 1 mL of healthy donor’s blood before treatment with the Rarecells^® Live Cells Buffer as described above and in section 6. Samples were incubated for 5 min before filtration as described above.

The enriched cells were recovered and centrifuged for 2 minutes at 150 g. After removal of the supernatant, each pellet was carefully resuspended in 100 μL of complete media, deposited onto a round glass coverslip of 1.2 cm in diameter (MGF-slides, Microscopic Glass Factory) and placed in a 24-wells culture plate (Multiwell^TM 24, Becton Dickinson, USA). Cells were left to attach for 2 hours in a culture incubator. Coverslips were then washed with 300 μL of PBS and the cells were either collected for immediate microscopy observation (D0 time points) or supplemented with fresh media and kept in culture for 48 hours (D2 time points) or up to 5 days (D5 time points) before observation.

Confocal microscopy analysis of cytoskeleton markers before enrichment and after enrichment by ISET^® and culture for 72 hours.

To assess the potential impact of our live tumor cells isolation protocol on cytoskeleton structures, we used cytoskeleton markers and immunofluorescence staining of actin microfilaments and microtubules together with confocal microscopy. Live A549 and H2228 tumor cells were seeded onto round glass coverslips and placed in a 24-wells culture plate before and after live cell enrichment from blood by ISET^®, as described above. Cells were kept in culture for 72h before fixation by incubating 15 minutes with a 4% paraformaldehyde solution. Cells were then washed twice with PBS and permeabilized 15 minutes in PBS containing 0,2% Triton X-100 before washing again twice with PBS. Antigens were then blocked by incubating with PBS containing 10% FCS for 1h at room temperature. Primary mouse antibody targeting human actelyl-α-Tubulin (6-11-B1 clone, Thermofisher Scientific, USA) was diluted 1:200 in PBS containing 10% FCS and incubated 1h at room temperature. Cells were then washed twice with PBS and incubated 30 minutes in the dark with a 1:200 dilution of the secondary goat anti mouse antibody conjugated to a Dylight 488 fluorophore (Thermofisher Scientific, USA). Without washing, samples were then supplemented with Alexa Fluor 568 conjugated Phalloidin (Thermofisher Scientific, USA) for F-actin labeling at 20 units/mL and kept in the dark for another 30 minutes at room temperature. Cells were then washed three times in PBS and incubated 15 minutes at room temperature and in the dark with PBS containing 8 μM Hoechst for nuclear staining. After rinsing twice with PBS, coverslips were finally mounted on glass slides using Ultramount permanent mounting medium (Dako, USA). Microscopy evaluation was performed with an SP5 confocal microscope (Leica, Germany) and all images were acquired with identical parameter settings. Images were processed with Image J software and the quantitative analysis of both actin and tubulin fluorescence was performed on 30 cells for each sample evaluated by calculating the corrected total cell fluorescence (CTCF) of each signal as described above.

NGS analyses of single tumor cells isolated by ISET^®.

Live or fixed tumor cells enriched from blood by ISET^®, collected without sticking them to the filter as described in section 5, and transferred in suspension to a cell culture plate could be gently picked up manually and individually with a micropipette (Gilson P2) under an inverted Olympus CK30-F2000 microscope at the x10 objective and transferred to an individual 0.2 mL PCR tube.

Single cells’ proteins were lysed so as to release the DNA using our in house protocol (in 15 μL, 100 mM TrisHCl pH 8 and 400 μg/mL of proteinase K incubated for 2 hours at 60°C followed by proteinase K inactivation for 5 min at 94°C). The single-cell DNA was then pre-amplified by combining our lysis protocol and by adapting it to single-cell whole genome amplification (WGA) commercial kits: MALBAC (Yikon Genomics, China); Picoplex (Rubicon Genomics and New England BioLabs, USA); GenomePlex (Sigma-Aldrich, USA). Amplified DNA was then purified using the DNA Clean & Concentrator™-5 kit (Zymo Research, Germany) and quantified by Quant-iT™ PicoGreen^® dsDNA Assay Kit (Thermo Fisher Scientific, USA). The WGA products' quality was assessed by PCR as described elsewhere [40].

For NGS single-cell analysis using the Illumina approach, 250 ng of the amplified DNA was used to prepare libraries for high-throughput sequencing with the Nextera Exome Enrichment Kit (Illumina, USA) following the supplier’s protocol. A fraction of the library DNA from the pre-capture step was used for whole genome analysis. The libraries were sequenced on a HiSeq2000 (Illumina) in 2 x 100 bp paired-end runs. Data were mapped against the human genome and the number of reads per chromosome was calculated.

In order to perform single-cell targeted theranostic mutations' NGS analysis, we used the ThermoFisher approach, 10 ng of purified WGA DNA was used to generate 207 amplicons libraries including 50 oncogenes and tumor suppressor genes through the use of the optimized primer pool “Hotspot cancer panel v2” and Ion Torrent™ Library Kit V2 (ThermoFisher, USA), according to manufacturer’s recommendations. Briefly, Ion adapters and barcodes were added using the “bioanalyzer protocol” (for living single cells and bulk extracted DNA) or the “FFPE protocol” (for fixed single cells), respectively for 17 cycles and 20 cycles of pre-amplication. The quality of purified libraries was verified with the Bioanalyzer (Agilent, USA) and quantified by qPCR using the Ion Library TaqMan™ Quantitation Kit. Libraries were templated using the Ion One-Touch system and sequenced on Ion Torrent™ 318 chips using the Ion Personal Genome Machine™ (ThermoFisher, USA).

Data analysis was performed using the Ion Torrent™ suite (ThermoFisher, USA) and Excel (Microsoft, USA) softwares. Sequencing artifacts in homopolymer regions were filtered out. In theory, when analyzing diploid single cells, only three allele frequencies should be possible: 0% (wild-type), 50% (heterozygous mutant/wild type) and 100% (homozygous mutant). However, WGA and sequencing methods introduce variability and quantitative amplification biases. In consequence, COSMIC hotspot mutations were included in our analysis only if present in at least one single cell WGA sample with a quality score over 100 and 15% allele frequency while using 10X depth amplicon coverage. Venn diagrams were drawn using a public online tool (bioinformatics.psb.ugent.be/webtools/Venn/).

NGS analysis of tumor cells isolated by ISET^® and cultured for 72 hours.

In order to investigate the feasibility of studying possible tumor cells' genetic changes induced by their isolation from blood with our method and culture for 72 hours, we performed NGS analysis on tumor cells before and after isolation by ISET^® and culture. Approximately 10⁴ live A549 and 10⁴ HCT116 live tumor cells were seeded in a 24-wells culture plate before and after their enrichment from blood by ISET^®, as described above. Cells were kept in culture for 72h before their collection for DNA extraction by incubating them for 2 minutes with trypsin (0.125%) at 37°C. Trypsin reaction was stopped by addition of 2 mL of complete media. Cells were further diluted with PBS, centrifuged at 300 g for 4 minutes, and carefully resuspended in 200 μL of PBS for DNA extraction.

We extracted bulk DNA from our samples using the QIAamp DNA Blood Mini Kit (Qiagen, Germany) following the manufacturer's instructions. We then used 10 ng of extracted bulk DNA from unfiltered control tumor cells and from tumor cells cultured for 72 hours after ISET^® filtration using the same ThermoFisher protocols with the ‘Hotspot Cancer Panel V2’ and Ion Torrent™ Library Kit V2 (ThermoFisher, USA) used for single tumor cells.

Data from experiments on tumor cells' populations were further processed with Sophia DDM^® software (Sophia Genetics, Switzerland) to calculate variant fraction and coverage. Only variants with more than 5% of allele frequency were reported.

B1. Sensitivity and precision (repeatability and reproducibility) of the ISET^® system for isolation of fixed tumor cells from blood.

We assessed the analytical sensitivity, sensitivity threshold (or Lower Limit of Detection (LLOD), i.e. the smallest amount of cancer cells which can be detected), repeatability and reproducibility of the ISET^® standard system for isolation of fixed tumor cells (Fig 1A). Because of the particular clinical relevance of detecting tumor cells in blood when they are very rare, we tested the LLOD and the sensitivity at the LLOD of the ISET^® system by performing analyses with very low numbers of fluorescent tumor cells, counted one by one by micropipetting and spiked in blood from healthy donors. Typical "spiking tests" experiments were performed by adding individually micropipetted A549, MCF-7, HeLa or MMTV-PyMT fluorescent tumor cells to one to ten mL of blood and processing the blood by ISET^® filtration (Fig 2A). We then analyzed the filter and detected the recovered spiked cells by looking for their fluorescent signal. We only counted fluorescent signals proven to be fluorescent cells (see Methods, section 6). This in vitro assay was 100% specific (n = 4 tests with 0 spiked cells in 1 mL of blood).

To assess reproducibility by two different operators using different reagent lots and instruments, we performed 12-fold replicate tests with two A549 cells spiked in one mL of blood. We found a total number of 21 and 20 out of 24 spiked cells per operator, respectively (Table 1), with an average recovery success rate of 88% and 83%

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. In vitro assay of the repeatability and reproducibility of ISET^® sensitivity tests for isolation of fixed cells.

https://doi.org/10.1371/journal.pone.0169427.t001

Given the structure of the ISET^® cartridge (which separates 10 mL of blood in 10 samples of one mL of whole blood i.e. 10 samples of 10 mL of diluted blood) and since the LLOD of one tumor cell per mL of blood had been found by our team and by independent teams [12, 25, 27–29], we thought that the LLOD and physical limit of the ISET^® system down to one tumor cell per 10 mL of blood could potentially be achieved and should be tested. Consequently we spiked by individual cell counting in 6-fold replicate tests two A549 cells in one mL, five mL and ten mL of blood. We found a total of 10 of 12, 10 of 12 and 12 of 12 spiked cells, respectively, with an average recovery success rate of 83%, 83%, and 100% respectively (Table 2). Similarly, in 10-fold replicate tests with 1 MCF-7, we recovered 9 out 10 spiked MCF-7 cells. In 3-fold replicate tests with 1 and 3 HeLa cells, we recovered 3 out 3 spiked HeLa cells, and 9 out of 9 spiked HeLa cells, respectively (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. In vitro assay of ISET^® sensitivity threshold for isolation of fixed A549 cells.

https://doi.org/10.1371/journal.pone.0169427.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. In vitro assay of ISET^® sensitivity using various types of cancer cells.

https://doi.org/10.1371/journal.pone.0169427.t003

Using the same protocol and workflow, we also spiked the mouse MMTV-PyMT cells which isolation had been studied on 8 micron-pore and 5 micron-pore filters (Fig 3B). In 4-fold replicate tests with two MMTV-PyMT cells, we found a total number of 6 out of 8 spiked cells (Table 3).

Our sensitivity tests with the ISET^® System were found to be reproducible and the LLOD was determined as one cancer cell per 10 mL of blood (Tables 1–3). Sensitivity at the closest concentration to the LLOD was 100%. Importantly, similar average recovery success rates were found with various volumes of blood processed (1 mL, 5 mL or 10 mL of blood) (Table 2) and tumor cells having different sizes (median sizes ranging from 12 to 22 μm) (Tables 1–3, A in S1 File and Fig 3).

B2. Linearity and accuracy of the ISET^® System for isolation of fixed tumor cells from blood.

We next assessed the ISET^® System's linearity (i.e. predictability of sample recovery with known dilution factors within an assay range or overall sensitivity) and accuracy (i.e. trueness as compared to a reference) by using fluorescent A549 cells carefully counted by dilution (see Methods, section 6) and processed with the ISET^® fixed cells protocol (Fig 1A and Methods, section 3). Recovery is calculated when cells are spiked into the relevant matrix, blood in our case, at a minimum of five concentration levels covering the linear range of the assay (according to the FDA definition, Docket No. 2004D-0163).

We have assessed linearity by spiking 30, 100 and 300 A549 cells, all counted by dilution, in one mL of blood in 9, 13 and 2 replicates, respectively. We have also plotted the number of observed fluorescent A549 cells on ISET^® filters after their counting by micropipetting versus the expected number of A549 (Tables 1 and 2 and Fig 4A). Using a linear regression analysis, we found a slope of 0.9991 and correlation coefficient (R²) of 0.96 (Fig 4A and Table B in S1 File). After addition of a number of 2, 30, 100 and 300 cells in 1 mL of blood, we recovered an average number of 1.7 cells per mL (85%, n = 30 replicates), 28 cells per mL (94%, n = 9), 85 cells per mL (85%, n = 13) and 334 cells per mL (111%, n = 2), respectively (Fig 4B and Table B in S1 File). We also obtained similar cell recovery rates by spiking dilutions of 50 and 100 HeLa cells in 1 mL of blood (respectively, 91% n = 14 and 104% n = 9) and 50 MCF-7 cells in 5 mL of blood (105%, n = 5) (S1A Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. In vitro assay of ISET^® linearity.

(A) Number of tumor cells (A549) detected on ISET^® filters plotted against expected number of tumor cells. Cells were spiked into whole blood after counting by dilution (30 to 300 cells) or counting by individual cells micromanipulation (2 cells). 0 cells (n = 4 replicates), 2 cells (n = 30), 30 cells (n = 9), 100 cells (n = 13) or 300 A549 cells (n = 2) were added to 1 mL of blood. Error bars (when visible) indicate standard error. (B) Mean % of tumor cells (A549) detected on ISET^® filters plotted against expected number of tumor cells spiked into whole blood. (C) Number of tumor cells (A549) observed on ISET^® filters plotted against expected number of tumor cells after extrapolation to 10 mL of blood (log scale). In addition to the test with 2 cells spiked in 10 mL (n = 6 replicates), to facilitate comparison with different volumes of blood, tests in 1 mL or 5 mL were plotted as their equivalent in 10 mL (See Methods). Error bars (when visible) indicate standard error. (D) Number of tumor cells (HeLa) detected on ISET^® filters plotted against expected number of tumor cells. Cells were added into the whole blood after their counting by dilution (50 to 100 cells) or after their counting by individual cells micromanipulation (1 or 3 cells). 1 cell (n = 3 replicates), 3 cells (n = 3), 50 cells (n = 14) or 100 cells (n = 9) were added to 1 mL of blood (See Methods). Error bars (when visible) indicate standard error.

https://doi.org/10.1371/journal.pone.0169427.g004

Since the ISET^® cartridge partitions 10 mL of blood in 10 times 1 mL volumes, cell recovery in 10 mL of blood has the same efficiency as cell recovery in 1 mL of blood (also demonstrated by our tests (Table 2)). We have thus plotted the equivalent tests in 10 mL of blood (i.e. 2 cells in 1 mL equivalent to 20 cells in 10 mL) and calculated linearity over 6 concentrations of A549 tumor cells ranging from 2 to 3000 CTC in 10 mL equivalent of blood. In this analysis, our average recovery rate (overall sensitivity) was 99.91% (Fig 4C). We observed similar linearity (overall sensitivity) with HeLa cells over 4 tumor cell concentrations (Fig 4D), with an average recovery rate of 100% ± 1%.

Finally, we calculated the percent coefficient of variation (% CV), a measure of precision, for each condition tested (cell line, volume of blood, number of spiked cells) (Table B in S1 File). Average %CV was 22% in experiments with spiked cells counted by cell dilution (n = 6 conditions, 52 tests). In experiments with micromanipulated spiked cells, average %CV was also 22% (n = 9 conditions, 62 tests) (Table B in S1 File). As expected, in these tests with an extremely low number of spiked cells, which assess the detection limit, the %CV values are usually higher because of the binary nature of the test results (0 or 1 cell) which has an impact on the readout. Accuracy (measured as recovery error) was below 25% for all conditions (Table B in S1 File). Under the rigorous conditions of our spiking tests and readings (see Methods), tumor cells' recovery obtained using spiked cells counted by dilution was consistent with that obtained using spiked cells counted by micropipetting.

It is worth mentioning that cell spiking with the individually micropipetted cells protocol requires technical skills in single cells' micromanipulation and is a particularly demanding test. One single micropipetted cell added to ten mL of blood (containing on average 50 to 100 million leukocytes and 50 billion erythrocytes) can be lost if it is damaged by the trypsin used to detach cells from the petri dish. In this case, the spiked rare cells can be lost independently from the capability of the ISET^® system to capture them from blood.

To verify if the cells are evenly distributed over 5 spots, we used the methodology described by Krebs et al. [12] and assessed intra-assay precision and accuracy (Fig 5). We examined the number of tumor cells found on each spot after spiking of 50 A549 cells counted by dilution into 5 mL of blood and filtration on 5 spots (each corresponding to the filtration of 1 mL of blood). To study the impact of tumor cells' distribution on 5 spots, we diluted tumor cells and obtained 43 to 59 cells per 5 mL in three different experiments (on average 8.6 to 11.8 A549 per mL). For each experiment, tumor cells' counts were calculated from all possible combinations of spots (i.e. any 1, any 2, any 3, any 4). The average tumor cells' number (± Standard Error) with 95% confidence intervals (CI) was calculated for each experiment and each combination of spots as described by Krebs et al. [12]. In an experiment with on average 8.6 tumor cells per mL, analyzing four spots provided an accurate estimation of the true mean (mean ± SE = 8.6 ± 0.13 tumor cells per mL, 95% CI: 8.35–8.85). When analyzing 3 spots or 2 spots, the estimation was less accurate, respectively 8.6 ± 0.14 (CI: 8.33–8.87) and 8.6 ± 0.21 (CI: 8.19–9.01) tumor cells per mL. Similarly, in an experiment with on average 11.8 tumor cells per mL, analyzing four spots provided an accurate estimation of the true mean (11.8 ± 0.31, 95% CI: 11.41–12.19). When analyzing 3 spots or 2 spots, the estimation was less accurate, respectively 11.8 ± 0.22 (CI: 11.37–12.23) and 11.8 ± 0.33 (CI: 11.16–12.44) (Fig 5).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Assessment of ISET^® intra-assay accuracy and precision.

50 A549 cells were spiked into 5 mL of blood (n = 3 experiments, 43 to 49 cells per 5 mL). The number of tumor cells found on each spot after ISET^® filtration (each corresponding to the filtration of 1 mL of blood) was recorded. Experiments were done on 5 spots but for intra-assay precision and accuracy only the assessment of the number of cells found on single spots or randomly grouped spots (any 1, any 2, any 3, any 4 spots) is relevant. The cell counting on the combination of all the 5 spots was used as reference. Results show that cell counting on four spots exhibited a representative mean tumor cells value per mL of blood. (A) Bar chart with the mean tumor cell number per spot and corresponding standard error of the mean (error bars) depending on the number of spots analyzed. Error bars are calculated using the standard deviation in different combinations of any 4 spots, any 3 spots, any 2 spots or any 1 spot, respectively. If only one spot is considered, standard deviation is higher than when counting 4 spots, showing that counting on four spots gives a reliable mean tumor cells' value per mL of blood. (B) Table indicating the number of tumor cells found on each spot for each of the five experiments, the 95% confidence interval (CI), the precision (%CV) and the accuracy (%error) depending on the number of spots analyzed (1 to 4) as compared to the analysis on five spots.

https://doi.org/10.1371/journal.pone.0169427.g005

Altogether, ISET^® intra-assay average precision was 4% (range 2.9% to 5.8%) and average accuracy was 2.7% (range 2.5% to 3.1%) when tumor cells were enumerated on 4 spots. Enumerating tumor cells on 3, 2 and 1 spot(s) leads to a decreased average precision (respectively, 5.2%, 7.7% and 13.4%) and a decreased average accuracy (respectively, 4%, 6% and 10.8%) (Fig 5). We thus found that the ISET^® System's intra-assay precision and accuracy related to individual spot’s cell counting is consistent when 50 cells are filtered on a total of 5 spots and cells are enumerated on at least 4 spots. We have also performed further analyses showing that intra-assay precision and accuracy is maintained with an extended cell range of 29 to 70 cells in 5 mL of blood (6 to 14 cells per mL) (S2 Fig). These results show that, if the number of tumor cells in 5 mL of blood is comprised between 29 and 70 (between 58 and 140 in 10 mL of blood), counting the tumor cells on 4 spots gives the same number of tumor cells per mL of blood as counting the tumor cells in 5 or 10 mL of blood.

Overall, we found that the ISET^® filtration's accuracy, precision and linearity are very high. Furthermore, the average recovery rate from two up to 3000 tumor cells per 10 mL (or equivalent) of blood is close to 100%.

B3. Sensitivity of the ISET^® system for isolation of fixed tumor cells from blood after collection of plasma.

Since detection of rare circulating tumor cells logically prefers methods using a non-negligible volume of blood (8 to 10 mL), we checked the feasibility of isolating plasma and tumor cells from the same blood sample. This protocol aims to perform multiple tests using the same sample, in order to make full use of the precious patient’s blood sample (Fig 1B). We assessed the in vitro sensitivity of this protocol by spiking 10 fluorescent A549 cells, counted one by one by micropipetting, in 5 mL of healthy donor's blood. In 4-fold replicate tests, we found 9 cells twice, 10 cells once and 7 cells once (Table 4) with a recovery success rate of 88%. The morphology of the spiked cells was intact as we found by performing MGG cytological staining (S1D Fig). The very high recovery success rate demonstrates that the LLOD and overall sensitivity for dual collection of plasma and tumor cells using the ISET^® system are similar to that of the standard ISET^® protocol.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. In vitro assay of ISET^® sensitivity with dual collection of tumor cells and plasma.

https://doi.org/10.1371/journal.pone.0169427.t004

B4. Assessment of the number of contaminant leucocytes on filters after isolation of fixed tumor cells.

The ISET^® system successfully eliminates all red blood cells and the majority of white blood cells, however some of them remain. By using cytopathological staining, we counted the number of residual leucocytes on the filters obtained by filtrating the blood of 4 different healthy donors (see Methods, section 3). With the standard ISET^® workflow for isolation of fixed cells, the number of leucocytes on the filter is variable, ranging from 1208 to 3188 cells per mL (1779 on average). This corresponds to an approximate enrichment factor of 4 logs, in agreement with the data reported by other teams [12].

C1 Sensitivity and reproducibility of the ISET^® system for enrichment of live tumor cells.

Isolating by filtration live cells with no antibody-related bias and no or minimum cell loss is a very challenging goal as live tumor cells are flexible and can more easily slip through the pores. We have concentrated our efforts on developing a new filtration buffer and a variant ISET^® protocol allowing enrichment of live tumor cells and their recovery in suspension for molecular analysis, in particular transcripts' analyses, and cell culture (Fig 1C). We have then assessed the in vitro sensitivity of this new workflow by spiking various numbers of cultured live A549 and LNCaP fluorescent tumor cells, collected one by one by micropipetting, in 1 mL of healthy donor's blood and performing the new protocol. The enriched cells have been transferred to an Ibidi microchamber and identified under microscope due to their fluorescent signal (Fig 2B), as described in the Methods, section 6.

In 6-fold replicate tests using 1, 3, 5 and 10 spiked live fluorescent A549 cells in 1 mL of blood, we found 6/6, 16/18, 24/30 and 50/60 cells (Table 5), with an overall recovery success rate of 100%, 89%, 80% and 83%, respectively. Spiking tests were repeated with 5-fold replicate tests using 1 and 5 spiked live fluorescent LNCaP cells in 1 mL of blood. We found a total number of 4/5 and 21/25 cells with an overall recovery success rate of 80 and 84%, respectively (Table 5). Furthermore, using a Trypan blue stain, we found that the viability of the enriched live tumor cells was maintained. The recovery rate and the analytical sensitivity we obtained with live A549 and LNCaP cells are very high. Sensitivity at the closest concentration to the LLOD tested (i.e. 1 tumor cell in 1 mL of blood) was 100% and similar to that obtained with the standard ISET^® protocol for isolation of fixed cells (Tables 1–3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. In vitro assay of ISET^® sensitivity for enrichment of live tumor cells.

https://doi.org/10.1371/journal.pone.0169427.t005

C2 Assessment of cell size and viability before and after enrichment of live tumor cells with the ISET^® system.

The precise size of at least 200 live A549 and MMTV-PyMT cells has been carefully calculated before and after filtration through 8-micron pores (Table C in S1 File and Fig 6). The median diameter remained essentially the same before and after filtration of live cells from both cell lines (13.6 μm ± 2.6 and 13.9 μm ± 2.6 for A549 cells and 10.0 μm ± 2.2 and 9.9 μm ± 2.2 for MMTV-PyMT cells, n = 3 experiments, respectively). Results demonstrate that filtration does not induce a selection in the cell size, including for the smallest cells (MMTV-PyMT) when they are kept alive (Fig 6). By using Trypan Blue exclusion, only the size of living cells has been taken into account. We have also calculated the rate of viability of live cells before and after filtration and found 99% ± 2% and 92.7% ± 2.4% for A549 cells and 97% ± 1.3 and 85% ± 3.4% for MMTV-PyMT cells, respectively (Fig 6 and Table C in S1 File).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Cell sizes and viability before and after live cell enrichment.

Cell sizes (median diameter) (A) or viability (B) according to each cancer cell line tested before or after live cell enrichment (n = 3 experiments). Error bars indicate standard error. Cells from human and mouse tumor cell lines were incubated 5 min with the Rarecells^® Live Cells Buffer with blood and recovered on standard (8 micron-pore) ISET^® filters. MMTV = MMTV-PyMT. Cell size and viability were analyzed using the TC20™ Automated Cell Counter (Bio Rad) and Trypan Blue stain.

https://doi.org/10.1371/journal.pone.0169427.g006

C3 Assessment of cell surface marker expression before and after enrichment of live tumor cells with the ISET^® system.

As changes in protein expression could potentially occur during filtration, we also tested whether filtration of live tumor cells with our protocol induces a change in antigen expression by using EpCAM antibody immunofluorescence and labeling live MCF-7 cells before and after filtration. The result of our analysis is shown in Fig 7. We did not find any difference in EpCAM expression induced by filtration under the tested filtration conditions. In fact, quantitative analysis of fluorescence using ImageJ (Fig 7D) revealed that the median corrected total cell fluorescence (CTCF) calculated on 50 MCF-7 cells was very similar before (median of 114665 ± 14363 with CTCF ranging from 43754 to 494919) and after (median of 124021 ± 12697 with CTCF ranging from 65799 to 594996 units) live cell enrichment. Furthermore, when applying cutoffs of CTCF to distinguish cells with low EpCAM expression from those with medium and high EpCAM expression we found very similar distributions of cells across all three categories: 28% (low), 50% (medium), 22% (high) before live cell enrichment and 24% (low), 54% (medium), 22% (high) after live cell enrichment, respectively (Fig 7C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Comparison of EpCAM fluorescence on labeled cells before and after live cell enrichment by ISET^®.

Representative images of MCF-7 cells before live cell enrichment showing EpCAM fluorescence alone (A1) and merged Hoechst and EpCAM fluorescence with bright field image (A2) are each presented with a scale bar of 50 microns. Representative images of MCF-7 cells after live cell enrichment showing EpCAM fluorescence alone (B1) and merged Hoechst and EpCAM fluorescence with bright field image (B2) are each presented with a scale bar of 50 microns. (C) Comparison of cell distribution across three levels of EpCAM expression for 50 MCF-7 cells before and 50 MCF-7 cells after live cell enrichment. The low EpCAM group comprises cells with corrected total cell fluorescence (CTCF) below 91000 units, the medium EpCAM category regroups cells with CTFC between 91000 and 200000 units and the high EpCAM group contains cells with CTCF above 200000 units. (D) Comparison of median corrected total cell fluorescence (CTCF) calculated on 50 MCF-7 cells before and 50 MCF-7 cells after live cell enrichment. Error bars indicate standard error.

https://doi.org/10.1371/journal.pone.0169427.g007

C4 In vitro culture of live tumor cells enriched from blood with the ISET^® system.

We tested if cells from cell lines could survive and proliferate in vitro after their enrichment from blood using the ISET^® System. We spiked A549 cells in blood in duplicate tests and processed the samples with the live cell enrichment protocol as described in the Methods section. We then compared imaging results at day zero (D0) with those obtained after 2 (D2) and 5 days (D5) of in vitro culture of enriched A549 cells isolated from blood (Fig 8). In both replicate tests, tumor cells observed at D0 were scattered as individual cells across the surface of the round coverslips (1.2 cm in diameter) used for their culture with an overall similar yet heterogeneous cell population density (Fig 8A). Tumor cells observed at D2 were mainly present as small clusters with a higher overall cell population density and some mitotic cells, clearly indicating cell proliferation (Fig 8B). Tumor cells observed at D5 were present exclusively as large clusters or colonies of proliferating cells with higher density, nearly covering the whole surface of the coverslips (Fig 8C). Interestingly, very few residual leucocytes were found at D0 after the first washing step and none of them were observed at D2 and D5, suggesting that stringent in vitro culture conditions can potentially select pure tumor cells populations. These results show that tumor cells' growth in culture is possible after live tumor cells enrichment by the ISET^® System.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. In vitro culture of A549 tumor cells after enrichment from blood using ISET^®.

Representative bright field images of A549 cells after live cell enrichment from blood and in vitro culture were obtained at 3 distinct time points. Scale bar of 50 microns. (A1, A2) Images at day zero (D0 = 2h of culture) showing A549 cells after live cell enrichment from blood obtained using the 10X (A1) and the 40X objective (A2). At 40X (A2) a single normal leucocyte is pointed by a single black arrow and a representative tumor cell is pointed by two arrows. (B1, B2) Images at day 2 (D2 = 2 days of culture) showing A549 tumor cells growing after live cell enrichment from blood obtained using the 10X (B1) and the 40X objective (B2). (C1, C2) Images at day 5 (D5 = 5 days of culture) showing A549 tumor cells growing after live cell enrichment from blood obtained using the 10X (C1) and the 40X objective (C2).

https://doi.org/10.1371/journal.pone.0169427.g008

C5 Assessment of cytoskeleton markers before and after enrichment of live tumor cells with the ISET^® system and short term in vitro culture.

In order to determine the potential impact of filtration on the cytoskeleton of live tumor cells, we performed immunofluorescence staining of actin microfilaments and microtubules of A549 and H2228 cells. We compare the profile of cells before enrichment to the one after their enrichment with the ISET^® System and short term in vitro culture (for 72 hours). For both cell types, key features of cytoskeleton dynamics involved in crucial cellular processes such as proliferation were observed before and after live tumor cells isolation using ISET^®. Results are shown in Fig 9.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. Cytoskeleton analysis of A549 and H2228 cells before and after live cell enrichment by ISET^® and in vitro culture.

Representative images of A549 and H2228 cells showing merged Hoechst (blue), actin (red) and tubulin [41] fluorescence were taken after 72h of in vitro culture using the 63X objective and are each presented with a scale bar of 20 microns. Images of A549 cells before filtration show examples of cells in G1/S phase (A1) and during mitosis (A2) for comparison with images showing A549 cells after filtration in G1/S phase (B1) and during mitosis (B2). Images of H2228 cells before filtration show examples of cells in G1/S phase (C1) and during mitosis (C2) for comparison with images showing H2228 cells after filtration in G1/S phase (D1) and during mitosis (D2). Comparisons of median corrected total cell fluorescence (CTCF) of actin (E) and of tubulin (F) calculated on 30 A549 and 30 H2228 cells before and 30 A549 and 30 H2228 cells after live cell enrichment. Error bars indicate standard error.

https://doi.org/10.1371/journal.pone.0169427.g009

Microscopy analysis revealed very similar actin and tubulin structures when comparing untreated A549 cells (Fig 9A) to A549 cells enriched live using the ISET^® system (Fig 9B) and kept 72h in culture. Importantly, no significant difference was observed in spindle morphogenesis during mitosis when comparing A549 cells before (Fig 9A2) and after live cell enrichment with ISET^® (Fig 9B2). In addition, all A549 cells observed at G1/S phase (Fig 9A1 and 9B1) displayed polygonal global shape, uneven distribution and unaligned orientations of F-actin filaments as well as frequent weakening of the peripheral actin cortex colocalized with microtubule-enriched plasma membrane extensions. These observations are consistent with previous literature on A549 cells [42] and correlate with known invasive properties [43] and stiffness' plasticity [44] of A549 cells which are prone to EMT transformation. Similarly, enrichment followed by 72 hours of in vitro culture had no visible impact on the cytoskeleton of H2228 cells (Fig 9D) as compared to untreated H2228 cells (Fig 9C). At G1/S phase, scattered microtubules and long straight actin filaments were observed with thicker actin bundles present near attachment sites of the plasma membrane for H2228 cells before (Fig 9C1) and after live cell enrichment (Fig 9D1). Enriched H2228 cells entering mitosis (Fig 9D2) displayed mitotic rounding, chromatin condensation and spindle morphogenesis with no apparent alteration compared to untreated controls (Fig 9C2).

Furthermore, global quantitative analysis of actin fluorescence (Fig 9E) performed on 30 randomly picked cells revealed very similar results of median CTCF before (median of 214166 ± 9416 for A549 and 234552 ± 19036 for H2228) and after live cell enrichment and culture (median of 209915 ± 7405 for A549 and 236599 ± 21468 for H2228). Using a Student's t-test, the differences of actin CTCF before and after live cell enrichment were found to be not significant (p value of Student test = 0.997 for A549 and 0.912 for H228).

Similarly, global quantitative analysis of tubulin fluorescence (Fig 9F) performed on 30 randomly picked cells also revealed very similar results of median CTCF before (median of 144482 ± 8824 for A549 and 188390 ± 12406 for H2228) and after live cell enrichment and culture (median of 151118 ± 6732 for A549 and 174672 ± 16327 for H2228). Using a Student test, the differences of tubulin CTCF before and after live cell enrichment were found to be not significant (p value of Student test = 0.296 for A549 and 0.916 for H2228).

Of note, median CTCF was higher for H2228 as compared to A549, which could be due to the larger size of H2228 cells or a higher expression level of tubulin and actin.

Altogether these results indicate that tumor cells' morphology and cellular functions related to the cytoskeleton distribution of F-actin and acetyl-α-tubulin can be either conserved or restored after 72h of in vitro culture following live tumor cells isolation from blood using the ISET^® system and protocol.

C6 Estimation of the remaining leukocytes after live tumor cells filtration and their depletion using CD45 magnetic beads.

The ISET^® System workflow for live cells successfully eliminates all red blood cells and the majority of white blood cells. The remaining white blood cells (WBC) were counted in 2 tests as described in the Methods section 5. Results showed from 640 to 2485 cells per mL of blood (1563 on average), thus similar to the amount of WBC remaining using the fixed cell protocol, and an approximate enrichment factor of 4 logs.

We next tested the feasibility of eliminating the remaining leucocytes after live cells enrichment with the ISET^® System using immuno-magnetic beads coated with CD45 antibodies (Fig 1C). We spiked 10 A549 fluorescent cells, counting them one by one by micropipetting, in 1 mL of blood and processed the blood as described in the Methods section 6. In 9 replicate experiments, we found 6 A549 once, 5 A549 once, 4 A549 three times and 3 A549 four times with an average recovery rate of 40%, as shown in Table 6. As compared to the much higher recovery rate obtained after enrichment of live cells (80 to 100%, Table 5), the additional CD45-depletion step causes a loss of 50 to 60% of the enriched cells. However, no remaining leucocytes were identified with this protocol and the morphology of the spiked cells remained intact.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. In vitro assay of ISET^® sensitivity for enrichment of live tumor cells followed by CD45-immunomagnetic mediated leukocytes depletion.

https://doi.org/10.1371/journal.pone.0169427.t006

D1 Comparative NGS analysis of A549 and HCT116 cells before and after live tumor cell enrichment and 72h in vitro culture.

In order to further validate the potential clinical interest of our live tumor cells filtration workflow for theranostic applications, we used Next Generation Sequencing (NGS) to study DNA mutations in A549 and HCT116 cells before and after live tumor cells isolation followed by their short term (72h) in vitro culture. We used the hotspot cancer panel v2 on the Ion Torrent™ platform as described in the Methods section 7. This amplicon panel can detect 6893 possible COSMIC mutations over 50 oncogenes and tumor suppressor genes [45].

Median coverage ranged from 1423X to 2073X among the different samples (S3 Fig). Median sequencing depth across all variants detected was 2957X and 2357X for A549 DNA extracted before and after live enrichment, respectively and 3103X and 3306X for HCT116 DNA extracted before and after live enrichment, respectively (Table D in S1 File). 17 SNP variants were found in A549 DNA including 11 COSMIC mutations classified as most likely pathogenic (Fig 10 and Table D in S1 File). Allele frequencies of all variants were concordant in the sample ‘before filtration’ as compared to the sample ‘after filtration’ (Fig 10A and 10C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 10. Ion Torrent^TM molecular characterization of A549 and HCT116 cells before and after live cell enrichment from blood using ISET^®.

Comparison of variants' mutant allele frequencies in bulk extracted DNA of approximately 10⁴ A549 (A) and 10⁴ HCT116 (B) cells at 72h of in vitro culture, before and after live cell enrichment. Correlation of variant allele frequencies in bulk extracted DNA of approximately 10⁴ A549 (C) and 10⁴ HCT116 (D) cells at 72h of in vitro culture, before and after live cell enrichment.

https://doi.org/10.1371/journal.pone.0169427.g010

Similarly, 25 variants were found in HCT116 cells including 23 SNPs and 2 INDELs. 16 of these mutations are classified as most likely pathogenic (Table D in S1 File). Allele frequencies of all variants were concordant in the sample ‘before filtration’ as compared to the sample ‘after filtration’ (Fig 10B and 10D). Of note, 4 of these heterozygous mutations were found in the EGFR pathway (PIK3CA H1047R, SMO V404M, ABL1 Y257C and KRAS G13D) in agreement with previous literature [46]. Of note, 11 variants were common in A549 and HCT116.

Interestingly, variants were also identified with confidence at a similar low frequency (6 to 10%) in both samples ‘before filtration’ and ‘after filtration’. (STK1 Q37L and FGFR3 P402P in A549, and FGFR3 P402P in HCT116, Fig 10). This reflects a known phenomenon of genetic and phenotypic heterogeneity of cells lines [47–49] that could be identified in our cell populations.

In A549 DNAs, 3 regions had low coverage (under the 500X threshold) (S3 Fig). Two of these regions were in the CDKN2A gene (amplicon coverage 0) which is depleted in A549 [50]. The last low coverage region was the same in all samples (A549 and HCT116); it corresponds to a variant in CSF1R 3'UTR identified in all the samples, albeit with a coverage under the 500X threshold (Table D in S1 File).

These results provide a proof of principle that the tumor cells’ genetic profile can remain stable after their isolation from blood using our protocol and their short-term culture.

D2 Single-cell genetic analysis after live cell enrichment.

Since studying the whole genome of individual CTC when they are very rare is important to explore the CTC heterogeneity at the beginning of tumor invasion, we then assessed if our protocol allows single circulating tumor cells' capture and their NGS analysis.

NGS exome libraries were first prepared using the Illumina technology with purified whole-genome amplified DNA from fresh single cells and fixed (not microdissected) single cells isolated by the ISET^® System, and from unamplified human gDNA as control. Respectively, we obtained 52, 41, and 44 million reads as well as a genome coverage of 76%, 70%, and 95% with average sequencing depth varying from 20 to 27X. To further assess the quality of our libraries, we next prepared whole genome and exome libraries with whole-genome amplified DNA from 4 fresh single cells and from 4 fixed (not microdissected) cells. The sequencing and mapping results show that we recovered about 70% of the exomes from the analyzed single cells.

Since performing NGS analysis of druggable mutations in very rare single CTC may be clinically relevant, we studied single tumor cells enriched from blood by the ISET^® System using the hotspot cancer panel v2 on the Ion Torrent platform as described in the methods section 7. Overall, we analyzed three live single leucocytes, three live single HCT116 cells, six live single A549 cells, and 3 fixed single A549 cells, isolated by the ISET^® System, WGA-amplified bulk DNA from HCT116 and A549 cells, as well as bulk genomic DNA from HCT116, A549 cells and the donor’s leukocytes.

We first evaluated standard sequencing quality control parameters of the DNA obtained from the WGA of our single cells (S1 File and S4 Fig). Then, we studied the presence of hotspot mutations registered in the COSMIC database in the WGA DNA from single cells isolated by the ISET^® System. We identified the presence of heterozygous mutations in these 4 hotspots in all HCT116 single cells, WGA-amplified HCT116 DNA and non-amplified HCT116 DNA but not in single leukocytes (Fig 11A). Additionally, when pooling data from three single HCT116 cells, the frequency of mutant allele was consistent with an heterozygous mutation, respectively 56% for KRAS G13D, 50% for PIK3CA H1047R, 45% for SMO V404M, and 51% for ABL1 Y257C (Fig 11A). Furthermore, the coverage for these amplicons ranged from 11 to 2000 (Table E in S1 File). Of note, in HCT116, additional mutations in the APC, FGFR3 and STK11 genes have been described in the literature [46] but are not included in the mutations that can be detected by the hotspot cancer panel v2. A549 cells are known to harbor KRAS G12S homozygous mutation [51] and this mutation was found in all analyzed live A549 single cells (n = 6) as well as fixed A549 single cell (n = 3) (S5 Fig). Furthermore, for all A549 samples, the two amplicons for the CDKN2A (p16) gene failed to be detected (Fig 10A and Table D in S1 File), which is consistent with the loss of this locus in this cell line [50]. Interestingly, when analyzing the presence of nonsense COSMIC mutation in leucocytes, we found the presence of heterozygous KIT M541L mutation in all single leucocytes and the matching bulk DNA (Fig 11A and Table E in S1 File). However, although this mutation has been suggested to be associated with leukemia, it is actually a frequent polymorphism among the Caucasian population [52] consistent with our finding.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 11. Ion Torrent^TM molecular characterization of single HCT116 and leukocytes enriched from blood using ISET^®.

(A) Non-sense variants mutant allele frequency from Catalogue Of Somatic Mutation In Cancer (COSMIC) database in whole genome amplified DNA from single HCT116 tumor cells (H1 to H3), whole genome amplified DNA single leukocytes (L1 to L3), whole genome amplified bulk HCT116 DNA (WH), unamplified bulk HCT116 DNA (BH) and unamplified bulk extracted DNA from the blood donor (BL). (B) Venn diagram showing the concordance of COSMIC non-sense variants determination in whole genome amplified DNA from single HCT116 tumor cells (H1 to H3) as compared to unamplified bulk HCT116 DNA (bulk HCT116) and whole genome amplified bulk HCT116 DNA (WGA). # indicates SMARCB1 R201* and ## indicate CTNNB1 S45P, NOTCH1 L1574P and RB1 E137*. (C) Venn diagram showing the concordance of COSMIC non-sense variants determination in whole genome amplified DNA from single leukocytes (L1 to L3) as compared to unamplified bulk extracted DNA from the blood donor (bulk).

https://doi.org/10.1371/journal.pone.0169427.g011

Finally, we plotted all non-sense COSMIC mutations found in the HCT116 samples and leucocytes and studied personal mutations (i.e. mutations which can only be identified in a single sample). Personal mutations could reflect either heterogeneity (which is not expected among leukocytes) or technical artifacts (error introduced by polymerase during WGA or sequencing) (Fig 11B and 11C). While 0, 1 and 3 personal mutations were identified in the three live single HCT116, personal mutations were not found in single leucocytes (n = 3), nor in WGA-amplified bulk DNA from HCT116, nor in bulk DNA from HCT116 and the donor’s leukocytes. The personal mutations found in HCT116 could represent variants, which are present in less than 5% of the cells (the sensitivity of mutation detection in bulk DNA) or technical artifacts.

Overall, our data suggest that NGS-mediated whole genome analyses, exome analysis and multi-genes panel analyses of tumor cells isolated by the ISET^® System are feasible. Of note, we found consistent results with live and fixed tumor cells isolated by the ISET^® System, but fixed cells required 3 additional cycles of pre-amplification at the NGS library preparation step (S4A Fig). These data are susceptible to provide reliable and informative results for a deeper understanding of CTC heterogeneity including at the critical early steps of tumor invasion.