Irreversible alteration of extracellular vesicle and cell-free messenger RNA profiles in human plasma associated with blood processing and storage

The discovery and utility of clinically relevant circulating biomarkers depend on standardized methods that minimize preanalytical errors. Despite growing interest in studying extracellular vesicles (EVs) and cell-free messenger RNA (cf-mRNA) as potential biomarkers, how blood processing and freeze/thaw impacts the profiles of these analytes in plasma was not thoroughly understood. We utilized flow cytometric analysis to examine the effect of differential centrifugation and a freeze/thaw cycle on EV profiles. Utilizing flow cytometry postacquisition analysis software (FCMpass) to calibrate light scattering and fluorescence, we revealed how differential centrifugation and post-freeze/thaw processing removes and retains EV subpopulations. Additionally, cf-mRNA levels measured by RT-qPCR profiles from a panel of housekeeping, platelet, and tissue-specific genes were preferentially affected by differential centrifugation and post-freeze/thaw processing. Critically, freezing plasma containing residual platelets yielded irreversible ex vivo generation of EV subpopulations and cf-mRNA transcripts, which were not removable by additional processing after freeze/thaw. Our findings suggest the importance of minimizing confounding variation attributed to plasma processing and platelet contamination.


Materials and methods
Blood sample collection and processing. All experimental protocols were reviewed and approved by the Oregon Health & Science University Institutional Review Board. All methods were carried out in accordance with relevant guidelines and regulations. Blood samples from healthy individuals were obtained from the Cancer Early Detection Advanced Research center (CEDAR) at Oregon Health and Science University. All samples were collected under institutional review board (IRB) approved protocols with informed consent from all participants for research use. Whole blood was collected from healthy individuals with 8.5 ml in ACD-A (BD Vacutainer, Becton Dickinson, cat. 364606), 10 ml in K2EDTA tubes (BD Vacutainer, Becton Dickinson, cat. 36643), 10 ml in heparin (BD Vacutainer, Becton Dickinson, cat. 367874), or 3 ml in sodium citrate tubes (BD Vacutainer, Becton Dickinson, cat. 369714) via antecubital vein puncture using a 21G butterfly needle (BD Vacutainer, Becton Dickinson, cat. 367281). Tubes were transported vertically at room temperature before processing. Within 1 h of blood withdrawal, 10 ml of whole blood was centrifuged at 1000 × g for 10 min at room temperature with the highest acceleration and deceleration setting at '9' using Eppendorf 5810-R centrifuge with S-4-104 Rotor. Plasma was collected until 10 mm above the buffy coat and was labelled as S1. To obtain double spun plasma, S1 plasma was centrifuged in Eppendorf 5424R centrifuge at 15,000 × g for 10 min at room temperature. The resulting supernatant of platelet-depleted plasma was collected and labelled as S2. S1 and S2 plasma samples did not undergo a freeze/thaw cycle. Plasma samples that were frozen at − 80 °C and thawed at room temperature were labelled as S1FR and S2FR respectively. For post-thaw processing, S1FR was centrifuged in Eppendorf 5424R centrifuge at 15,000 × g for 10 min at room temperature. The resulting supernatant was carefully transferred and designated as S1FRS2. All samples were collected and processed using a uniform protocol at Oregon Health and Science University. The overall schematic diagram of plasma processing steps was described in Supplementary  Fig. S1.
Platelet counting using improved Neubauer haemocytometer. The platelet count was measured by the improved Neubauer haemocytometer (VWR Scientific Products, Piscataway, NJ) by two independent, experienced researchers. The total number of platelets were counted from the central 1 × 1 mm area consisting of 25 groups of 16 squares separated by closely ruled triple lines.

Flow cytometry set-up for light scatter and fluorescence calibration. Beckton-Dickinson FAC-
SAria Fusion equipped with 488 nm (60 mW), 561 nm (100 mW), and 640 nm (100 mW) lasers was used. For optimal configuration of submicron size detection, 0.1 µm size filter was applied to the sheath fluidic system to reduce sheath fluid noise. The sample flow rate was set at 1, which was measured by mass discharge 30 and determined to be 45 μl/min. Timed collections were recorded for 60 s. Data collection was set using the SSC trigger threshold value of 200 using scatter wavelength at 488 nm. In order to calibrate light scattering, 152, 203, 303, 401, 510, and 600 nm polystyrene NIST-traceable beads (ThermoFisher Scientific, cat. 3150A, 3200A, 3300A, 3400A, 3500A, and 3600A) were serially diluted in 0.1 μm filtered D-PBS without calcium and magnesium (ThermoFisher Scientific, cat. 14190250). A minimum of 5000 events were recorded for 60 s. Particle diameter and scatter relationship was established utilizing FCMpass software (v3.09, http:// nanop ass. ccr. cancer. gov) [32][33][34] . Median SSC-H intensity in arbitrary units was converted to standardized unit in EV diameter. To approximate EV diameter size, the average of effective refractive index (RI) data based upon published measurements were used. Detailed instructions for light scattering calibration based on a core-shell structure to model EVs were followed (Shell RI = 1.4800, Core RI: 1.3800, and shell thickness: 5 nm) 32  Characterization of platelets using flow cytometry. For counting platelets in differentially processed plasma, blood samples from three healthy individuals were obtained in 10 ml K2EDTA tubes (BD Vacutainer, Becton Dickinson, catalog number: 36643). Plasma was processed using single spin at 1000 × g (S1). To obtain double spun plasma, S1 plasma was centrifuged in Eppendorf 5424R centrifuge at 15,000 × g for 10 min at room temperature (S2). To investigate the CD9, CD63, and CD41 expression, 5 µl of S1 and S2 plasma was incubated with 5 µl of antibody mix prepared after established dilution series. CD9 Alexa Fluor 647 ( 38 . The concentration of platelets (platelets per µl) was calculated using the following formula: A = ((B × C)/(D × E)) × F, where A is the concentration of platelets (platelets per µl), B is the gated events for platelet population, C is the volume of CountBright beads, D is the gated events for Count-Bright beads, E is the volume of platelet samples (µl), and F is the CountBright bead concentration (beads per µl) from the manufacturer.
RT-qPCR profiling of cell free mRNA. For characterizing the effect of freeze thaw on cell free mRNA expressions, RNA was extracted using plasma processed with S1, S2, S1FR, S2FR, and S1FRS2 conditions. Cell free mRNA was isolated by using plasma/serum circulating and exosomal RNA Statistical analysis. Significance for platelet count measurements using the improved Neubauer haemocytometer and flow cytometry across differential centrifugation were both assessed using the Wilcoxon test. To determine the impact of overall preanalytical factors, statistical analysis was performed on CD9 + , CD63 + , or CD41 + EVs on specific gated populations across differential centrifugation and freeze/thaw processing. The significance of individual preanalytical factor comparisons were determined using Tukey's multiple comparison test. P values < 0.05 were considered statistically significant (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001). Analyses were conducted using R package.  [32][33][34] . For light scatter, each bead sample was analyzed at the same acquisition setting until > 5000 bead events were recorded. The histogram of each sized bead population revealed distinct side scattering in arbitrary units, where a progressive increase in SSC-H with increasing NIST bead diameter (152-600 nm) was observed (Fig. 1A). The median light scatter statistic of each bead size gated population was inputted into flow cytometry postacquisition analysis software (FCMpass) to calibrate light scattering [32][33][34] . The collection half-angle of our system, which is important to quantify the amount of light reaching a detector in absolute units, was determined to be 45.3° using FCMpass software. Utilizing the side scattering collection angle, recorded side scattering value in arbitrary units was standardized to predicted scattering cross-section using Mie theory [32][33][34] . The linear regression between our observed light scattering power in arbitrary units and predicted scattering cross-section resulted in R-squared value = 0.9991 (Fig. 1B). The acquired scattering intensity of standard beads (red dots) was plotted on modelled data (black line) for polystyrene beads, which revealed the model fitted actual data accurately (Fig. 1C). After scatter-diameter relationship for EVs is extrapolated using FCMpass software, the measured scatter signal for polystyrene beads corresponding to vesicle diameter is revealed (Fig. 1C). Approximate diameters of EVs was calculated using the average of effective EV refractive index (Shell RI = 1.4800, Core RI: 1.38000, and shell thickness: 5 nm) [32][33][34] . Next, we performed fluorescence calibration using forward scatter as the trigger threshold to gate micron-sized MESF beads. MESF beads for each fluorophore was analyzed until > 5000 bead events were recorded. While FSC-A vs. SSC-A revealed a single microsphere population, four fluorescent microspheres were observed with varying fluorescent intensity in arbitrary units ( Fig. 2A Flow cytometry reveals distinct vesicle populations differentially affected by blood processing condition. After establishing light scatter and fluorescence calibration, we investigated the impact of differential centrifugation on plasma EVs using the flow cytometer. Whole blood was collected in EDTA, and plasma was differentially processed using single centrifugation at 1000 × g for 10 min (S1) and double centrifugation (S2: 15,000 × g secondary spin for 10 min after the initial single spin S1) (Fig. 3A, Supplementary Fig. S1). Complete counts of residual platelets in plasma were measured using a haemocytometer and flow cytometry with platelet marker CD41. Using the haemocytometer, single spun plasma S1 contained an average platelet concentration of 313 ± 74 thousand/µl while secondary spin resulted in the removal of more than 99.99% of residual platelets in S2 (Fig. 3B). Using flow cytometer, single spun plasma S1 contained an average CD41 + platelet concentration of 127 ± 32 thousand/µl while secondary spin resulted in the removal of more than 99.90% of residual platelets in S2, consistent with the reduction measured by the haemocytometer (Supplementary Fig. S2A). For EV analysis, plasma was stained with anti-CD9, anti-CD63, and anti-CD41 fluorescent antibodies and measured by flow cytometry. The fluorescently positive gated data revealed that there are distinct populations in EV diameter distribution ranging between 150 and 3000 nm (Fig. 3C). It is noted that the subset of EVs within 150-1000 nm range at around 500 nm is an artifact of Mie scattering calibration from our calculated flow cytometer collection angle and geometry. Specifically, this corresponds to a plateau in the scatter-diameter curve from ~ 400-480 nm using predicted EV light scattering from the estimated average EV refractive index employed in our model (Fig. 1C). Welsh et al. reported a similar observation, suggesting that a plateau from the scatter-diameter curve resulted in an artifact between 400 and 480 nm accordingly 32 . Therefore, we gated EVs into two populations: 150-1000 nm which may be comprised of small and medium EVs, and 1000-3000 nm comprised of large EVs and platelets 30 . Notably, we observed much less 150-1000 nm CD41 + EVs for both S1 and S2 compared to 150-1000 nm CD9 + and CD63 + EVs. Flow cytometer assay controls included unstained samples, isotype controls, serial dilution of stained plasma, and antibody with buffer alone (Supplementary Fig. S3). Serial dilution of stained plasma showed the linear detection of EVs while the median fluorescence intensity remained constant, suggesting that EVs were detected and counted as single particles via flow cytometry. Plasma condition at S2 resulted in a clear reduction in 1000-3000 nm populations compared to S1 while the 150-1000 nm EVs remained similar for plasma from both processing conditions (Fig. 3C).
Distinct subsets of cf-mRNA influenced by differential centrifugation and post-thaw processing. Since platelets and EVs contain mRNA, we sought to determine if blood centrifugation and post-thaw processing affected cf-mRNA levels. We analyzed cf-mRNA profiles in single and second spin plasma freshly (S1 and S2), after freezing at − 80 °C (S1FR and S2FR), and for samples subjected to a second spin following S1FR processing (S1FRS2). We selected a panel of housekeeping, platelet and tissue-specific genes for multiplex RT-qPCR measurements (Fig. 6A, B). Hierarchical clustering analysis of relative gene expression between post-thaw processed samples revealed three distinct clusters (Fig. 6A). Overall, these clusters were either dependent (nontissue specific) or independent (tissue specific) of post-thaw processing conditions, wherein non-tissue specific genes segregated into two clusters. The first cluster included genes (HBG1 and SMC4 for example), which could be removed by post-thaw processing and therefore were likely related to large EVs or platelets (Fig. 6A, B). The second cluster, including platelet genes and house-keeping genes such as PF4 and B2M, was partially removed by post-thaw processing and therefore was likely associated with ex vivo generated small and medium EVs which Figure 3. Effect of differential centrifugation on EVs using flow cytometry. (A) Schematic diagram of differentially processed plasma using single spin (S1: 1000 × g centrifugation) and double spin (S2: 15,000 × g secondary spin after the initial single spin S1). (B) Platelet concentration in differentially processed plasma from three healthy individuals (n = 3) was measured in independent technical replicates using a haemocytometer. The error bar represented standard deviations for the indicated blood processing conditions. P-value was calculated using Wilcoxon test (*P < 0.05 www.nature.com/scientificreports/ remained after post-thaw processing (Fig. 6A, B). Importantly, our results revealed that tissue specific gene signatures (such as genes expressed in liver tissue; including APOE and ALB) were retained regardless of spinning and post-thaw processing conditions (Fig. 6A, B), suggesting they are present in non-platelet small or medium EVs. The relationship of cf-mRNA transcripts with EV subpopulations requires further investigation and is the subject of future studies. Overall, as genes from different biological roles are uniquely affected by preanalytical differences, the selection of novel cfRNA biomarkers should consider the effects of preanalytical variability.

Discussion
Circulating EVs and cf-RNA are promising biomarkers for disease diagnosis and prognosis [41][42][43][44] . However, significant variability in standardizing blood processing across published methods has led to a lack of reproducibility between studies 2, 16,17,45 . In this study, we utilized multiparametric flow cytometry and cf-mRNA profiling to characterize preanalytical influences on EV and cf-mRNA subpopulations in plasma. We observed two distinct subpopulations by flow cytometry which are differentially impacted by centrifugation and post-thaw processing. Interestingly, we observed small and medium CD9 + EVs and CD41 + EVs were irreversibly generated via freezing single spun plasma while CD63 + EVs remained similar. Importantly, these ex vivo generated EVs could not be removed by additional centrifugation after freeze/thaw, and thereby can significantly affect downstream analyses. As a first in cf-mRNA studies, we also found groups of genes significantly, partially, or unaffected by post-thaw processing in plasma.
Since different types of EV purification methods (ultracentrifugation, density gradients, size-filtration, etc.) affect the yield and purity of EVs 46,47 , we chose to fluorescently label EVs directly in plasma using EV tetraspaninspecific antibodies with proper assay controls according to recent MIFlowCyt-EV reporting framework 48 . Previous study highlighted effects of centrifugation on pre-isolated EVs from platelets and erythrocytes by examining recovery of EVs through differential centrifugation 49 . However, their freeze-thaw cycle was performed on purified EVs, leading to no significant change across different temperatures of single freeze-thaw cycle. By examining EVs directly in plasma, we observed irreversible ex vivo generation of EVs and cf-mRNA subpopulations altered by differential centrifugation and post-thaw processing.
The choice of anticoagulant used during blood collection is another important preanalytical factor which influences downstream analyses and the extent of ex vivo EV release 15,50,51 . Other studies also have tested different anticoagulant tubes focusing on platelet microparticles or large EVs 28,50,52 . György et al. showed the highest CD42 + /Annexin V + platelet microparticle counts using flow cytometry were found in heparin compared to citrate differentially processed plasma (S1, S2) and respective freeze thaw processes (S1FR, S2FR). (B) Box plot of CD9 + , CD63 + , and CD41 + of gated events from 1000-3000 nm (red) and 150-1000 nm (green) for S1, S1FR, S2, and S2FR using R. CD9 + , CD63 + , CD41 + events were converted to concentrations using calibrated flow rate in a given acquisition time. EV concentration defined as the number of EVs per μl was determined by number of EVs detected in a given sample volume multiplied by the dilution factor. The sample volume was calculated by the product of measured flow rate and acquisition time. Statistical significance were obtained from three healthy volunteers for each freeze thaw processing condition using Tukey's multiple comparisons (ns = not significant, P > 0.05; *P < 0.05, ***P < 0.001, ****P < 0.0001). www.nature.com/scientificreports/ and ACD tubes 50 . Lacroix et al., also examined Annexin V + platelet microparticles using flow cytometry and found the most Annexin V + platelet microparticles in decreasing order for EDTA, heparin, and citrate tubes 39 . After establishing light scattering calibration using flow cytometry, we have found the impact of anticoagulant tubes differed based on the markers being tested (CD9, CD63, or CD41) and size ranges (150-1000 nm EVs vs 1000-3000 nm EVs). Our study revealed that the combined effect of both freezing and anticoagulation with the highest fold change in 1000-3000 nm CD41 + EVs was found in heparin compared to EDTA, ACD, and sodium citrate tubes. However, EDTA generated the highest fold change in 1000-3000 nm CD9 + and CD63 + EVs. For 150-1000 nm EVs, EDTA generated the highest fold change for CD9 + and CD41 + EVs while the relative levels of CD63 + EVs were similar across all anticoagulants. Similar to Jayachandran et al. and Lacroix et al., we also found that counts of 1000-3000 nm large EVs were similar both before and after freezing double spun plasma 28,39 . Notably, while EDTA is commonly used for RNA analysis 53,54 , EDTA also caused the largest change in 150-1000 nm CD9 + and CD41 + EVs as well as 1000-3000 nm CD9 + and CD63 + EVs when freezing single spun plasma. Although we found the relative effect of anticoagulant types can be minimized by freezing samples as secondary spun plasma, it is important to test and choose anticoagulant types to fit specific research questions and downstream analysis. Although previous studies highlighted the preanalytical influences on microparticle generation associated with platelet activations 40,49,55 , the effect of blood processing on EV subpopulations using flow cytometry with light scattering and fluorescence standardized calibration is lacking. The enumeration of microparticles in previous studies mostly utilized flow cytometry that was validated to discriminate between 0.5 and 0.9 μm Megamix beads 39,49 . Since considerable efforts have been directed to establish a standardized methodology for EV measurements by flow cytometry [30][31][32][33][34] , we applied this standardized approach to investigate enumeration of EVs influenced by preanalytical factors. Utilizing the FCMpass software developed by Welsh et al. [32][33][34] , we observed differential centrifugation results in distinct EV subpopulations within the diameter range between 150 and 3000 nm. In addition, our freeze thaw analysis further revealed ex vivo generated CD9 + EVs, adding to the body of literature on platelet-associated blood processing artefacts 2,16 . We addressed the importance of preanalytical influences on EVs using a standardized approach, which will improve the reproducibility with respect to effective EV diameter and given fluorochrome molecule standards across literatures. Figure 5. Effect of post-thaw processing on EVs using flow cytometry. (A) Schematic diagram of differentially processed plasma (S1, S2), respective freeze thaw samples (S1FR, S2FR), and secondary spin after post-freeze/ thaw plasma S1FR (S1FRS2). (B) Representative flow cytometry dot plot of EV diameter (nm) versus fluorescent intensity in Quantum MESF units for CD9 + EVs, CD63 + , and CD41 + EVs in S1FR, S2FR, and S1FRS2 conditions using FlowJo. Quantum Alexa Fluor 647 MESF is used for Alexa Fluor 647 conjugated CD9 stained plasma, Quantum Alexa Fluor 488 MESF is used for Alexa Fluor 488 conjugated CD63 stained plasma, and Quantum PE MESF is used for PE conjugated CD41 stained plasma. Events were gated from 150 to 1000 nm (green box) and from 1000 to 3000 nm (red box). (C) Box plot of CD9 + , CD63 + , CD41 + EV concentration from 1000-3000 nm (red) and 150-1000 nm (green) for S1FR, S2FR, and S1FRS2 using R. Statistical significance were obtained from three healthy volunteers for each freeze thaw processing condition using Tukey's multiple comparisons (ns = not significant, P > 0.05; *P < 0.05, ***P < 0.001, ****P < 0.0001). www.nature.com/scientificreports/ Comprehensive assessment of light scattering sensitivity on multiple different flow cytometers was performed by Van der pol et al. 30 . For small particle detection, only a few flow cytometers detected more than three different sized reference beads using both side scatter (SSC) and forward scatter (FSC). Similarly, our instrument could not detect more than three sized reference beads by FSC. Instead, we utilized FCMpass software to calibrate SSC using the effective refractive index of EVs (Shell RI = 1.4800, Core RI: 1.3800, and shell thickness: 5 nm) 32 . Since the true refractive index of different EV subpopulations is currently unknown, the average EV refractive indices based on core-shell theory has been implemented 32,33 . Although EVs in the ~ 1000 nm diameter range may overlap with small platelets 56 , precise refractive indices which considers platelet granule content and shapes is currently unknown. Specific studies, which definitively parse EVs from small platelets and understanding refractive indices of EV subpopulations, are needed to better define EV physical characteristics and compositions.
How blood processing influences circulating microRNA has been previously shown 16,17 , and yet the impact on cf-mRNA is poorly understood. Cheng et al. provided preanalytical influences on miRNA expression due to differing residual platelet amount 17 . Conversely, our study investigated the impact of blood processing conditions through differential centrifugation, respective freezing condition, and post-thaw processing on cf-mRNA. We revealed cf-mRNA groups whose extent of preanalytical variability differed based on the degree of residual platelets in plasma. In particular, non-tissue specific genes were further classified as either partially or fully removed by post-freeze/thaw processing. Intriguingly, tissue-specific cf-mRNA were less prone to blood processing conditions, revealing them as potentially more robust biomarkers, or differentially associated with smaller vesicle subpopulations retained through centrifugation.
In conclusion, our study provides an assessment of the preanalytical effect of differential centrifugation and freeze/thaw cycles on plasma EVs and cf-mRNA. Employing multiparametric flow cytometry, our work provides insights into how preanalytical factors influence EV subpopulations and ex vivo release of EVs in association with residual platelets. Notably, these artifacts appear to be irreversible for CD9 + and CD41 + small and medium EVs and mRNA transcripts of genes present in platelets. However, an increasing number of other marker-specific studies, including non-platelet specific markers, will help to address the origin of the ex vivo EV generation observed in our study. Our results indicate distinct subpopulations of EVs and cf-mRNA are not removable by additional spinning after freeze/thaw. Therefore, consideration should be taken when analyzing EVs and cf-mRNA from banked plasma and designing robust EV and cf-mRNA based liquid biopsy tests. Figure 6. Effect of freeze thaw and post-thaw processing on cf-mRNAs using qRT-PCR. (A) Hierarchical clustering analysis of relative levels (in ΔCt) of 16 custom selected genes using RT-qPCR. Ct difference (ΔCt) between S1 and individual processing conditions are indicated from lowest (blue) to highest (red) using R. Nontissue specific genes that are fully or partially removed, and tissue-specific genes which are retained in S1FRS2 with respect to S1 are shown. (B) Box plot of the median expression levels (in Cts) for representative non-tissue specific genes which are fully removed (i.e. HBG1 and SMC4) or partially removed (i.e. PF4, and B2M), and tissue-specific genes (i.e. APOE, ALB) which are retained in S1FRS2 with respect to S1 are shown using R. Higher raw Ct indicates lower levels of cfRNA transcripts.