Method Validation for Preparing Serum and Plasma Samples from Human Blood for Downstream Proteomic,Metabolomic,and Circulating Nucleic Acid-Based Applications

Abstract

Background:

Formal method validation for biospecimen processing in the context of accreditation in laboratories and biobanks is lacking. Serum and plasma processing protocols were validated for fitness-for-purpose in terms of key downstream endpoints, and this article demonstrates methodology for biospecimen processing method validation.

Methods:

Serum and plasma preparation from human blood was optimized for centrifugation conditions with respect to microparticle counts. Optimal protocols were validated for methodology and reproducibility in terms of acceptance criteria based on microparticle counts, DNA and hemoglobin concentration, and metabolomic and proteomic profiles. These parameters were also used to evaluate robustness for centrifugation temperature (4°C versus room temperature [RT]), deceleration (low, medium, high) and blood stability (after a 2-hour delay).

Results:

Optimal protocols were 10-min centrifugation for serum and 20-min for plasma at 2000 g, medium brake, RT. Methodology and reproducibility acceptance criteria were met for both protocols except for reproducibility of plasma metabolomics. Overall, neither protocol was robust for centrifugation at 4°C versus RT. RT gave higher microparticles and free DNA yields in serum, and fewer microparticles with less hemolysis in plasma. Overall, both protocols were robust for fast, medium, and low deceleration, with a medium brake considered optimal. Pre-centrifugation stability after a 2-hour delay was seen at both temperatures for hemoglobin concentration and proteomics, but not for microparticle counts.

Conclusions:

We validated serum and plasma collection methods suitable for downstream protein, metabolite, or free nucleic acid-based applications. Temperature and pre-centrifugation delay can influence analytic results, and laboratories and biobanks should systematically record these conditions in the scope of accreditation.

Introduction

Validation of biospecimen processing methods allows for characterization of the method's reproducibility, robustness, and sample stability, and plays an important role in the context of quality assurance, accreditation, and biomarker development. Published validated methods eliminate the need for individuals to establish the conditions, saving time and resources. A fit-for-purpose method reflects not only its feasibility, but also, and most importantly, the reliability of the analytical results. Reliability can be measured in terms of reproducibility of results from a sample, robustness of results from differently processed samples, and sample stability. In addition, biospecimen processing method validation has become a normative requirement for biobank accreditation.^1,2 Here we present the first in a series of biospecimen processing method validations, including urine collection, peripheral blood mononuclear cell isolation, nucleic acid extraction from stools, and nucleic acid and protein extraction from formalin-fixed paraffin-embedded tissue. The purpose of this series is to not only provide methodology references for biospecimen processing, but also to propose methodology for the development of validation plans in the area of biospecimen processing.

Serum and plasma sample preparation is integral to clinical care and biomarker research.³ Samples are used in a broad range of applications including proteomics, metabolomics, and circulating nucleic acid analysis. Biobanks routinely process whole blood for serum or plasma preparation and storage.^4,5 Best Practices⁶ and Standard Operating Procedures^7–9 have been published, however method validation for fit-for-purpose biospecimen processing for downstream applications is lacking.

Serum extraction from peripheral blood requires clotting, while plasma extraction requires anticoagulants. K2EDTA is the most common agent used for proteomic and metabolomic applications although others may be used for specific downstream processing (e.g., citrate anticoagulant for platelet or microparticle-based applications). Serum or plasma is separated by centrifugation from the blood clot or cell mass, respectively.

Microparticles, hemoglobin, and circulating DNA are important endpoints in plasma and serum samples. Microparticles are small (0.1–1.5 μm diameter) membrane-bound vesicles derived from cells such as platelets and endothelial cells, carrying information on the protein composition of the parent cell membrane. However, microparticles with their distinct proteome containing hundreds or thousands of unique proteins,^10,11 many of which are altered in specific pathological conditions,¹² “contaminate” pure plasma, and their uncontrolled production can reduce the specificity of plasma proteomics. Hemoglobin may hamper analyses of clinical chemistry,¹³ metabolomics,¹⁴ and proteomics.^15,16 Circulating nucleic acids are increasingly important as clinically-relevant biomarkers and are relevant in prenatal¹⁷ and oncology¹⁸ diagnostics.

Here we report methodology, including an optimization step, for preparing serum and plasma from human blood. Optimal methodologies were validated in terms of reproducibility and robustness with a fit-for-purpose approach in terms of quality attributes of serum and plasma preparations.

Materials and Methods

Study design

Serum and plasma collection protocols were initially optimized to determine if 2000 g⁸ is optimal for extraction in terms of microparticle and platelet counts. Optimal protocols were then validated according to reproducibility and robustness (for centrifugation temperature and deceleration) in terms of microparticle counts, DNA yield, hemoglobin concentration, and metabolomic output. Robustness for pre-centrifugation delay was validated in terms of microparticle counts, hemoglobin concentration and proteomic output. Peripheral blood samples were collected from healthy volunteers who provided written informed consent (CNER approval #201107/02). The workflow is shown in Tables 1 and 2.

Table 1.

Plan of the Plasma Aliquoting Method Optimization and Validation

	Optimization	Reproducibility	Robustness (Temperature)	Robustness (Deceleration)	Robustness (Pre-centrifugation delay)
Donor
• quantity	two	one	one	one	three
• volume	6x K₂E 10 mL	3x K₂E 10 mL	2x K₂E 10 mL	3x K₂E 10 mL	2x K₂E 10 mL
	2x K₂E 10 mL
Start delay
• time	n.d.	n.d.	n.d.	n.d.	2h
• temperature	n.d.	n.d.	n.d.	n.d.	RT
					2°–8°C
Centrifugation
• speed	Double spin	Single spin	Single spin	Single spin	Single spin
	2000/2000 g	2000 g	2000 g	2000 g	2000 g
	2000/4000 g
	2000/10000 g
	Single spin
	1000 g
	2000 g
	4000 g
• time	(D)10/10 min	20 min	20 min	20 min	20 min
	(S)20 min
• temperature	RT	RT	RT	RT	RT
			4°C
• brake force	(D)medium/hard	medium	medium	soft	medium
	(S)medium			medium
				hard
Transfer and mix	3 mL	4 mL	4 mL	3 mL	3.5–4.0 mL
• aliquoting	3×1.0 mL	4×1.0 mL	4×1.0 mL	3×1.0 mL	3×1.0 mL
Microparticles
• test frequency	triplicate	triplicate	triplicate	triplicate	triplicate
• volume/test	10/100 μL	10 μL	10 μL	10 μL	10 μL
Hemoglobin
• test frequency	n.d.	duplicate	duplicate	duplicate	duplicate
• volume/test		100 μL	100 μL	100 μL	100 μL
Circulating DNA
test frequency	n.d.				n.d.
• extraction		single	single	single
• quantification		duplicate	duplicate	duplicate
• volume/test		1.0 mL	1.0 mL	1.0 mL
Metabolomics
• test frequency	n.d.	triplicate	triplicate	triplicate	n.d
• volume/test		5.0 μL	5.0 μL	5.0 μL
Proteomics
• test frequency	n.d.	n.d.	n.d.	n.d.	single
• volume/test					5.0 μL

D, double; S, single; n.d., not determined.

Table 2.

Plan of the Serum Aliquoting Method Optimization and Validation

	Optimization	Reproducibility	Robustness (temperature)	Robustness (deceleration)	Robustness (pre-centrif. delay)
Donor
• quantity	one	one	one	one	three
• volume	3x CAT 10 mL	3x CAT 10 mL	2x CAT 10 mL	3x CAT 10 mL	2x CAT 10 mL
Clotting
• time	30–45 min	30–45 min	30–45 min	30–45 min	2-h
• temperature	RT	RT	RT	RT	RT
			2–8°C		2–8°C
Centrifugation
• speed	1000 g	2000 g	2000 g	2000 g	2000 g
	2000 g
	4000 g
• time	10 min	10 min	10 min	10 min	10 min
• temperature	RT	RT	RT	RT	RT
• brake force	medium	medium	medium	soft	medium
	hard			medium
				hard
Transfer and mix	3 mL	3 mL	4 mL	3 mL	2 mL
• aliquoting	3×1.0 mL	2×1.0 mL	3×1.0 mL	2×1.0 mL	4×500 μL
		1×600 μL	1×600 μL	1×600 μL
		1×400 μL	1×400 μL	1×400 μL
Microparticles
• test frequency	triplicate	triplicate	triplicate	triplicate	triplicate
• volume/test	100 μL	100 μL	100 μL	100 μL	100 μL
Hemoglobin
• test frequency	n.d.	duplicate	duplicate	duplicate	duplicate
• volume/test		100 μL	100 μL	100 μL	100 μL
Circulating DNA
test frequency	n.d.				n.d.
• extraction		single	single	single
• quantification		duplicate	duplicate	duplicate
• volume/test		1.0 mL	1.0 mL	1.0 mL
Metabolomics
• test frequency	n.d.	triplicate	triplicate	triplicate	n.d
• volume/test		5.0 μL	5.0 μL	5.0 μL
Proteomics
• test frequency	n.d.	n.d.	n.d.	n.d.	single
• volume/test					5.0 μL

n.d., not determined.

Serum and plasma collection protocol for optimization and validation

Three 10-mL blood samples were collected from a single donor. For serum, samples were collected in CAT Vacutainer^® tubes (Becton Dickenson [BD] #367896) and stored upright for 30–45 min at room temperature (RT, 19°–24°C) for clotting. For plasma, samples were collected in K2E Vacutainers^® containing K2EDTA (BD #367525) and centrifuged immediately. Samples were centrifuged in a ThermoScientific SL40R centrifuge (Thermo Fisher) for 10 min, medium brake, at RT. Supernatant was mixed and immediately aliquoted.

For optimization, three centrifugation speeds were evaluated; 1000, 2000, and 4000 g. For plasma, two additional steps were performed. First, a double centrifugation was introduced; supernatant from the initial centrifugation was transferred to 15-mL Falcon tubes and centrifuged at 1000, 2000, or 4000 g (10 min, hard brake). Second, blood samples (4 mL) were divided between two 15-mL centrifuge tubes, and centrifuged at 2000 g (10 min, medium brake), then 2000 g (10 min, hard brake), or at 2000 g (20 min, medium brake).

Optimal serum and plasma were validated for reproducibility and robustness to deceleration (three 10-mL samples from one donor), for robustness to temperature (two 10-mL samples from one donor), and for robustness to pre-centrifugation delay (two 10-mL samples from three donors each).

Microparticle and thrombocyte counting

Serum or plasma samples (10 or 100 μL) were added to 10 mL CASY-ton solution (Roche Applied Science) and counted. Microparticles and thrombocytes were counted using a CASY^® TTC Measuring and Monitoring Device (Roche) equipped with a 45 μm aperture capillary, and CASY^® Measure v1.7 software, using an in-house small-particle setup. For optimization, a 200 μL sample uptake in triplicate was used, plotting counts on a 10-μm scale. For validation, a single 200 μL uptake and a 5-μm scale were used. Events in the range of 0.67–3.00 μm were considered microparticles. Counts were performed in triplicate.

Free nucleic acid extraction and DNA quantification

Free nucleic acids were extracted from serum or plasma (1 mL) using the QIAamp Circulating Nucleic Acid kit (Qiagen). Lysate volume column loading was performed with a QIAvac plus system (Qiagen) equipped with a 20-mL Tube Extender. Carrier RNA was not added to avoid interference with DNA quantification. Free DNA was eluted in 50 μL elution buffer. DNA was quantified using a Quant-it Picogreen dsDNA assay kit (Life Sciences). Fluorescence was read with a SynergyMX spectrophotometer (Biotek) running an in-house DNA quantification protocol with Gen5 software. Quantification was performed in duplicate.

Hemoglobin quantification

Hemoglobin was quantified using a DetectX Hemoglobin detection kit (Arbor Assay) according to the manufacturer's instructions, using the standard supplied. Absorbance was measured with a SynergyMX spectrophotometer with Gen5 software. Quantification was performed in duplicate.

Metabolite extraction

Plasma or serum (5 μL) was mixed with 45 μL methanol/water, placed on a shaking device (5 min, 4°C) then immediately centrifuged at 16,000 g (5 min, 4°C; Eppendorf 5415R). Supernatant (30 μL) was transferred to gas chromatography glass vials and completely dried (-4°C, 40 min) with a refrigerated CentriVap Concentrator (Labconco). To avoid condensation, the Concentrator was heated to RT for 15 min prior to vial removal. Metabolite extraction was performed in triplicate.

Metabolomics analysis: GC-MS

GC-MS measurements were performed on an Agilent 6890 gas chromatograph, with a DB-35MS capillary column. The chromatograph was coupled to an Agilent 5975C mass spectrometer with an electron impact ionization source operating at 70 eV. The spectrometer source was heated to 230°C and the quadrupole to 150°C. Online metabolite derivatization was performed with a Multipurpose Sampler (Gerstel). Dried metabolite extracts were mixed with 15 μL 2% methoxyamine hydrochloride in pyridine (MOX, Thermo Fisher) and incubated for 30 min, 40°C, followed by 15 μL 2,2,2-trifluoro-N-methyl-N-trimethylsilyl-acetamide (MSTFA, Machery Nagel), incubated for 30 min at 40°C. The detector was operated in scan mode and the sampler injected 1 μL of derivatized sample. The injection was set to splitless mode with helium as the carrier gas (1 mL/min flow rate). The gas chromatography oven was maintained at 80°C for 6 min, then increased to 300°C (by 6°C/min), then to 325°C for 4 min (at 10°C/min). GC-MS raw data were calibrated using MetaboliteDetector software¹⁹ with the retention index calibration option, with settings: minimum number of ions, 10; minimum peak height, 5 noise levels; deconvolution width, 5 scans. Samples were analyzed in triplicate. For each replicate, individual metabolite intensity was divided by the sum of all peak intensities from the sample.

Proteomic analysis: LC-MS

Plasma and serum proteins (5 μL) were reduced with tris(2-carboxyethyl)phosphine and alkylated with iodoacetamide before tryptic digestion. Tryptic peptides were labeled with one of the six Tandem Mass Tag reagents (Thermo Fisher) according to the manufacturer's instructions. Labeled samples were pooled and analyzed (LTQ Orbitrap Velos Pro mass spectrometry, Thermo Fisher) with a NanoAcquity HPLC (Waters). For survey scans, the orbitrap resolution was set to 60,000 and the ion population to 5×10⁵ with an m/z window from 400 to 2000. Up to three precursors were selected for fragmentation. For MS-MS, the ion population was set to 7000 in the LTQ and 2×10⁵ in the orbitrap, with resolution of 7500, first mass at m/z=100, and maximum injection time of 750 ms. Normalized collision energies were set to 35% for collision-induced dissociation (CID) and 60% for higher-energy C-trap dissociation (HCD).

Peaklists were generated using the EasyProt software platform,²⁰ and CID and HCD spectra were merged for simultaneous identification and quantification.²¹ Peaklist files were searched against the uniprot_sprot database (2012_06, 13-Jun-2012). TMT-sixplex amino terminus, TMT-sixplex lysine, and carbamidomethylation of cysteines were set as fixed modifications and oxidized methionine as a variable modification. Trypsin was the selected enzyme, with one potential missed cleavage, and the normal cleavage mode was used. All datasets were searched once in the forward and once in the reverse database. Separate searches were used to keep the database size constant. Protein and peptide scores were established to maintain the false positive peptide ratio (i.e., false discovery rate) below 1%, giving a slight FDR overestimation.²² For identification, only proteins matching two different peptide sequences were kept.

Validation criteria

Methodology acceptance criteria were: 1) microparticle counts <2×10⁷/mL for serum and <10⁸/mL for plasma, based on previously published, upper ranges of plasma microparticle concentrations,²³ 2) free DNA yield>10 ng for serum and>5 ng for plasma, corresponding to previously published plasma DNA concentrations in healthy individuals,²⁴ and 3) hemoglobin <0.5 mg/mL, corresponding to the upper reference limit for free hemoglobin in serum.²⁵

Reproducibility acceptance criteria were: CV <20% for mean microparticle counts, DNA, and hemoglobin values, and a nonsignificant difference in concentrations of >95% of detected metabolites between samples from a given donor. These were based on internal experience on the expected reproducibility of different methods.

Robustness was evaluated in terms of temperature (RT versus 2–8°C) and brake speed (soft, medium, vs. hard brake). Acceptance criteria were nonsignificant differences in mean microparticle counts, DNA, and hemoglobin values, and a nonsignificant difference in concentrations of >95% of detected metabolites between different conditions. For evaluation in terms pre-centrifugation delay (immediate vs. 2-h delay including clotting time; at 4°C and RT), acceptance criteria were nonsignificant differences in mean microparticle counts and hemoglobin values, and nonsignificant differences in concentrations of >95% of detected proteins (based on internal experience on expected reproducibility).

Statistical analyses

Mean, SD, and CV% were calculated. Optimization outcomes were compared using paired two-tailed t-tests. For reproducibility and robustness, microparticle counts, DNA, and hemoglobin values, metabolite and protein concentrations were compared between replicates, donors, or protocols using paired two-tailed t-tests or one-way Anova tests. Pairwise multiple comparisons (Holm-Sidak) were performed for robustness (deceleration) if a one-way Anova showed significant difference. For metabolomics, semi-quantitative data were obtained with batch quantification (nontargeted analysis) and statistically tested by Welch's t-test and one-way Anova. Sigma Plot v.12.0 (Systat Software) and R3.0.0 were used with a 5% significance threshold. For proteomics, protein ratio and p values were generated using the isobar package.²⁶

Results

Optimization of serum and plasma collection protocols

Microparticle counts were determined for three serum and eight plasma centrifugation protocols. In serum, differences observed in mean microparticle counts with different centrifugation speeds were non-significant (Table 3). Centrifugation at 2000 g was considered optimal, in order to minimize the probability of red blood cell (RBC) lysis, since recommendations on RBC processing indicate high risk of hemolysis at 5000 g for 5 min.²⁷

Table 3.

Optimization of Serum and Plasma Collection in Terms of Centrifugation Protocol, According to Microparticle Triplicate Counts

	Microparticle counts (counts/mL)
Centrifugation protocol	Count 1	Count 2	Count 3	Mean	SD	Thrombocytes	P value^a
Serum
1000 g	1.95×10⁷	1.90×10⁷	1.63×10⁷	1.83×10⁷	1.72×10⁶	NA	0.069
2000 g	1.43×10⁷	1.17×10⁷	1.38×10⁷	1.33×10⁷	1.38×10⁶	NA
4000 g	1.19×10⁷	1.08×10⁷	1.31×10⁷	1.19×10⁷	1.15×10⁶	NA	0.131
Plasma
1000 g	7.22×10⁷	7.69×10⁷	1.08×10⁸	8.56×10⁷	1.92×10⁷	+++
2000 g	5.93×10⁷	6.00×10⁷	6.19×10⁷	6.04×10⁷	1.33×10⁶	+++
4000 g	1.37×10⁷	1.81×10⁷	1.48×10⁷	1.55×10⁷	2.29×10⁶	++
2000 g/1000 g^b	9.07×10⁶	1.00×10⁷	1.18×10⁷	1.03×10⁷	1.38×10⁶	+	0.059
2000 g/2000 g^b	7.43×10⁶	7.22×10⁶	7.61×10⁶	7.42×10⁶	2.00×10⁵	–
2000 g/4000 g^b	3.02×10⁶	3.12×10⁶	4.05×10⁶	3.39×10⁶	5.68×10⁵	–	0.004
2000 g (2×10 min)^b	1.97×10⁷	1.97×10⁷	2.04×10⁷	2.00×10⁷	4.13×10⁵	–	0.459
2000 g (1×20 min)	1.75×10⁷	1.95×10⁷	2.08×10⁷	1.92×10⁷	1.65×10⁶	–

NA, not applicable; ^aPaired two-tailed t-test comparing 2000 g vs. 1000 g; 2000 g vs. 4000 g; 2×10 min vs. 1×20 min; ^bSecond centrifugation with hard brake, all other centrifugations at medium brake.

Thrombocyte counts in plasma samples with all three single centrifugation speeds were too high for sample analysis (Table 1). This was resolved with a second 10-min centrifugation giving low or no counts. Microparticle counts were not significantly different between 2000 g followed by 2000 g versus 2000 g followed by 1000 g, however a second spin at 4000 g gave significantly lower counts. Two 2000 g centrifugations were considered optimal for balancing adequate microparticle removal while minimizing the probability of RBC lysis.²⁷ The plasma protocol was further refined by comparing the two 10-min centrifugations with a single 20-min centrifugation at 2000 g to allow for automation on a liquid handler. Counts were not significantly different and a single 20-min centrifugation at 2000 g was considered optimal for plasma collection. See Figures 1 and 2 for representative CASY microparticle counts.

FIG. 1.

Representative graph from an optimization CASY microparticle count of a serum sample following centrifugation at 2000 g for 10 min.

FIG. 2.

(A) Representative graph from an optimization CASY microparticle counts of a plasma sample following centrifugation 4000 g for 10 min. (B) Representative graph from an optimization CASY microparticle counts of a plasma sample following centrifugation at 2000 g for 10 min, followed by 2000 g. (C) Representative graph from an optimization CASY microparticle counts of a plasma sample following centrifugation at 2000 g for 2X 10 min. The thrombocyte peak between 2 and 3 μm apparent after a single centrifugation is no longer present after a double or extended single centrifugation protocol.

Validation of serum and plasma collection protocols

Collection protocol methodology

A 10-min 2000 g centrifugation was validated for serum and a 20-min 2000 g centrifugation for plasma, both with medium brake at RT. Methodology acceptance criteria (microparticle counts, DNA and hemoglobin yield) were fulfilled for all samples (Table 4), except for three serum samples with significant hemolysis (hemoglobin concentration >0.50 mg/mL).

Table 4.

Methodology and Reproducibility Validation, According to Acceptance Criteria

	Methodology validation				Reproducibility validation
Evaluation parameter (acceptance criteria)	Sample 1	Sample 2	Sample 3	Mean^a (SD)	CV% (%) NS diff. metabolite conc.	Outcome^b
Serum^c
Microparticle counts/mL (<2×10⁷/mL)	7.12×10⁶	8.1×10⁶	5.97×10⁶	7.04×10⁶ (1.02×10⁶)	CV=14.6%	Accepted^d
DNA yield, ng/mL (>10 ng/mL)	24.9	20.2	25.6	23.6 (2.9)	CV=12.4%	Accepted
Hemoglobin, mg/mL (<0.50 mg/mL)	1.338	1.350	1.294	1.327 (0.029)	CV=2.2%	Accepted
% NS diff. metabolite conc. (>95%)					99.1%	Accepted
Plasma^e
Microparticle counts/mL (<10⁸/mL)	3.29×10⁷	3.29×10⁷	3.01×10⁷	3.20×10⁷ (1.62×10⁶)	CV=5.1%	Accepted
DNA yield, ng/mL (>5 ng/mL)	5.4	5.1	4.3	5 (0.55)	CV=11.1%	Accepted
Hemoglobin, mg/mL (<0.50 mg/mL)	0.106	0.150	0.116	0.124 (0.023)	CV=18.6%	Accepted
% NS diff. metabolite conc (>95%)					90%	Rejected

NS, nonsignificant; ^aMean of three samples from one donor; validation based on microparticles, DNA and hemoglobin yields within acceptance criteria; ^bBased on: 1) CV <20% for each criteria and 2) nonsignificant difference (one-way Anova test) between samples in concentrations of >95% of metabolites; ^c2000 g for 10 min, medium brake, RT; ^dOne reading was above the (<2×10⁷/mL limit); ^e2000 g for 20 min, medium brake, RT.

Reproducibility

Acceptance criteria for all four evaluation parameters (CV <20% for microparticle counts, DNA, and hemoglobin, and nonsignificant difference in concentrations of >95% of metabolites) were met for both protocols except for reproducibility of plasma metabolomics (Table 2). The hemolysis reported in serum samples did not impact reproducibility.

Robustness

Centrifugation temperature

The effect of centrifugation at RT versus 4°C was evaluated. Serum collection was robust for hemoglobin and metabolite content, but centrifugation at RT yielded significantly higher microparticle counts (p=0.028) and circulating DNA yield (p=0.014; Table 5), and yielded 11 out of 297 detected metabolites that were significantly different versus 4°C. For plasma, robustness was confirmed for DNA yield and metabolite content, but centrifugation at RT gave significantly lower microparticle counts (p=0.017) and hemoglobin content (p=0.020), and yielded 6 out of 304 detected metabolites that were significantly different versus at 4°C. Significant hemolysis did not occur with either protocol at either temperature. Overall, neither protocol was considered robust for centrifugation at 4°C versus RT.

Table 5.

Robustness, According to Centrifugation Temperature

Evaluation parameter	4°C^a	Room temperature^a	P value/% NS diff. metabolite conc^b	Outcome^b
Serum^c
Microparticle counts/mL	1.33×10⁷	1.73×10⁷	p=0.028	Rejected
DNA yield, ng/mL	8.9	14.2	p=0.014	Rejected
Hemoglobin, mg/mL	0.108	0.120	p=0.500	Accepted
% NS diff. metabolite conc.			96.3%	Accepted
Plasma^d
Microparticle counts/mL	7.40×10⁷	3.72×10⁷	p=0.017	Rejected
DNA yield, ng/mL	6.0715	5.9185	p=0.500	Accepted
Hemoglobin, mg/mL	0.150	0.118	p=0.020	Rejected
% NS diff. metabolite conc.			98.0%	Accepted

NS, nonsignificant. ^aOne blood sample per temperature from a single donor (two samples in total), with immediate centrifugation; ^bComparison of the two temperatures based on: 1) nonsignificant differences (paired t-test) for microparticles, DNA and hemoglobin yields, and 2) nonsignificant difference (Welch's t-test) between temperatures in concentrations of >95% of metabolites identified; ^c2000 g for 10 min, medium brake, RT; ^d2000 g for 20 min, medium brake, RT.

Centrifugation deceleration

With only two centrifuges simultaneously available and literature reports that microparticle excretion increases with longer pre-centrifugation delays,²⁸ we performed parallel centrifugation under the “extreme” deceleration conditions to limit this potential bias. Thus, samples with medium brake had a longer pre-centrifugation delay than those with the soft and hard brakes.

Serum samples had significantly different microparticle counts and circulating DNA content with the three brake speeds, being highest with medium brake (Table 4). For plasma, different deceleration speeds yielded significantly different microparticle counts, the lowest counts reported with medium brake, and significantly different hemoglobin yields with higher hemolysis at hard brake. The different deceleration conditions did not impact metabolites detected in serum or plasma (7 out of 321 and 3 out of 253 detected metabolites respectively were significantly different). All metabolite results are available upon request from the author.

A pairwise evaluation of rejected parameters did not reveal significant differences between soft and hard brake for either protocol (Table 6). Medium brake was considered optimal relative to microparticle release and hemolysis, and was validated for both protocols.

Table 6.

Robustness, According to Centrifugation Deceleration and Pairwise Comparison

	Centrifuge deceleration speed					Pairwise comparison
Evaluation parameter	Soft brake^a	Medium brake^a	Hard brake^a	P value/% NS diff. metabolite conc²	Outcome^b	Soft brake vs. hard brake
Serum^c
Microparticle counts/mL	1.01×10⁷	1.49×10⁷	9.52×10⁶	p=0.010	Rejected	0.688
DNA yield, ng/mL	14.5	17.0	13.7	p=0.038	Rejected	0.312
Hemoglobin, mg/mL	0.450	0.432	0.441	p=0.489	Accepted	ND
% NS diff. metabolite conc.				97.8%	Accepted	ND
Plasma^d
Microparticle counts/mL	8.41×10⁷	7.66×10⁷	9.44×10⁷	p=0.018	Rejected	0.109
DNA yield, ng/mL	6.8	6.4	6.3	p=0.413	Accepted	ND
Hemoglobin, mg/mL	2.459	2.287	2.616	p=0.030	Rejected	0.085
% NS diff. metabolite conc.				98.8%	Accepted	ND

ND, not done; NS, nonsignificant. ^aOne blood sample per braking speed from a single donor (three samples in total), with immediate centrifugation; ^bComparison of the three speeds based on: 1) nonsignificant differences (one-way Anova) for microparticles, DNA and hemoglobin yields, and 2) nonsignificant difference (one-way Anova) between speeds in concentrations of >95% of metabolites identified; ^c2000 g for 10 min, RT; ^d2000 g for 20 min, RT.

Pre-centrifugation delay

Given the contradictory robustness validation conclusions for centrifugation at RT versus 4°C, with higher circulating DNA concentrations at RT on the one hand in serum samples (fit-for-purpose for free DNA-based applications) but higher microparticle counts on the other hand (poor fit-for-purpose for proteomic-based applications), a stability evaluation was performed. The temperature comparison was repeated with the addition of a 2-h delay prior to centrifugation, instead of 30–45 min for serum and immediate for plasma.

Robustness in terms of microparticle counts was rejected for both serum and plasma protocols (Table 7) with significantly higher counts at RT versus 4°C in all but one donor after the 2-h delay. The hemolysis in serum samples from one donor was higher at room temperature than at 4°C. The hemolysis in plasma samples from one donor was higher at 4°C than at room temperature. No significant differences for hemolysis were observed for the other donors. Proteomics were analyzed to assess the impact of the higher microparticle counts on the proteins detected. Significant differences in protein levels were found for only 4.4% of the detected proteins in plasma samples and 1.6% in serum samples. All proteomic analysis results are available upon request from the author. Despite the statistically significant differences in microparticle counts, the 2 h pre-centrifugation delay did not significantly impact proteomics analyses performed at RT versus 4°C. Pre-centrifugation stability in terms of 2-h delay and temperature was accepted for hemoglobin concentration and proteomics analysis for both protocols.

Table 7.

Robustness, According to Temperature After a 2-Hour Pre-centrifugation Delay

Evaluation parameter	Donor 1^a 4°C	Donor 1^a RT	P value/% sign. diff. protein conc.^b	Donor 2^a 4°C	Donor 2^a RT	P value/% with sign. diff. protein conc.^b	Donor 3^a 4°C	Donor 3^a RT	P value/% sign. diff. protein conc.^b	Outcome^b
Serum^c
Microparticle count/mL	1.53×10⁷ (5.75×10⁵)	1.55×10⁷ (2.51×10⁶)	p=0.927	1.30×10⁷ (1.79×10⁶)	2.68×10⁷ (1.73×10⁴)	p=0.006	–	–		Rejected
	CV 3.7%	CV 16.2%		CV 13.7%	CV 0.1%
Hemoglobin, mg/mL	0.666	1.079	p=0.019	0.096	0.061	p=0.126				Rejected
% Sign. diff. protein conc.			4.8%			0%				Accepted
Plasma^d
Microparticle counts/mL	1.96×10⁷ (2.11×10⁵)	9.15×10⁷ (7.70×10⁵)	p<0.001	2.39×10⁷ (1.80×10⁵)	4.81×10⁷ (2.01×10⁶)	p=0.003	2.39×10⁷ (1.02×10⁶)	3.40×10⁷ (8.51×10⁵)	p<0.001	Rejected
Hemoglobin, mg/mL	0.219	0.002	p=0.042	0.858	0.818	p=0.430	0.937	0.909	p=0.410	Rejected
% Sign. diff. protein conc.			1.5%			4.8%			0%	Accepted

One blood sample per temperature from three donors (6 samples in total); ^bComparison of the two temperatures based on 1) non-significant differences (paired t-test) for microparticles and hemoglobin yields, and 2) significant difference between temperatures in concentrations of <5% of proteins identified; ^c2000 g for 10 min, medium brake, RT and 4°C; ^d2000 g for 20 min, medium brake, RT and 4°C.

Numbers in brackets indicate standard deviations from three independent measures.

Discussion

While microparticles are of interest for identifying biomarkers, in the scope of the serum and plasma processing methods developed here, and in the scope of microparticle-negative protein markers, they present a hindrance and we aimed to obtain the lowest possible microparticle counts with both protocols. The optimal serum collection protocol was established as a 10-min centrifugation at 2000 g using medium brake. Plasma collection required a second 10-min centrifugation to lower thrombocyte counts to acceptable levels. Maintaining the second centrifugation at 2000 g was optimal for balancing low microparticle counts against the higher risk of RBC lysis and consequent presence of hemoglobin associated with a second spin at 4000 g. Refining the protocol to a single 20-min centrifugation allows reduced handling (and an expected lower error rate) and better automation possibilities. Thus, the optimal plasma collection protocol was established as a 20-min centrifugation at 2000 g using medium brake. Acceptance criteria for microparticle counts, circulating DNA yields, and hemoglobin contamination were satisfied for both protocols.

Reproducibility was demonstrated for all criteria with both protocols, with acceptable CV for mean microparticle counts, circulating DNA, and hemoglobin. The metabolic profile showed only 0.9% of detected metabolites in serum with statistically poor reproducibility; however, in plasma, 10% of detected metabolites showed statistically poor reproducibility. The latter case may be due to GC-MS analytic variability. Exposure of whole blood to room temperature may impact the reproducibility of metabolomic analyses between different samples from the same donor.^14,29 A separate validation study is being performed to assess the impact of different pre-centrifugation conditions on plasma GC-MS metabolomics.

Neither protocol proved to be robust in terms of centrifugation temperature, demonstrating the importance of systematically using the same temperature when collecting samples. Serum centrifugation at RT gave significantly higher microparticle counts and circulating DNA than at 4°C, while for plasma it resulted in significantly lower microparticle counts. Given reports of high RBC microparticle shedding over time at 4°C,³⁰ the increased microparticle counts in plasma in RT-centrifuged blood may be due to RBC-derived microparticles. Furthermore, microparticle loss with centrifugation varies between individuals.³¹ For serum, obtaining significantly higher yields of free DNA at RT at the cost of having higher microparticle counts (1.7×10⁷/mL versus 1.3×10⁷/mL) was considered more important, the increased DNA yield offering an advantage in many contexts in which this protocol is used. For plasma, lower microparticle counts are favorable for proteomics analyses, supporting RT centrifugation. Free DNA yields generally correlated with microparticle counts under different centrifugation conditions, likely due to the DNA being of microparticle origin. It was recently suggested that the quantity of residual platelets and microparticles causes differential microRNA concentrations.³²

MetaboliteDetector software identifies compounds using sophisticated compound-matching algorithms based on specific compound libraries. Some peaks could not be associated with a peak identity as the corresponding compounds were not in the in-house compound reference library. Although plasma collection at 4°C gave significantly higher hemolysis than at RT, this was not considered biologically relevant. Microparticle counts in plasma processed immediately were significantly higher at 4°C than at RT, but were significantly lower following a 2-h delay. This is probably due to increased microparticle shedding at RT over prolonged pre-centrifugation delays.³⁰ Overall, centrifugation at RT is the preferred option, giving a higher free DNA yield in serum, and fewer microparticles with less hemolysis in plasma.

While all three deceleration speeds used were robust for plasma and serum collection, significant differences were seen in microparticle counts for soft and hard brakes relative to medium brake in serum. This may be explained by an increased number of particles released in the medium brake sample due to the extended pre-centrifugation delay,³⁰ and highlights the importance of systematically reporting the pre-centrifugation delay and minimizing such variations. Intermediate brake speed is preferred to high.

The pre-analytical robustness of serum and plasma has been studied,^15,33 showing that blood collection and processing variables can have a major impact on specific serum or plasma proteins and peptides.³⁴ Maximum pre-centrifugation delay at room temperature with few changes within protein mass spectrometry spectra is 4 hours^16,34,35 and importantly, room temperature also avoids platelet activation by cold.³⁶ It must be noted that the serum and plasma processing methods studied in this article have limited fitness-for-purpose for peptidomic analyses. Indeed, it has been shown that serum and plasma peptides (e.g., fibrinopeptide A) are trimmed down ex vivo in a time-dependent manner and this decay can partially be prevented by protease inhibitors.³⁷ Blood samples for serum collection should be processed with a minimum delay. Both protocols were stable in terms of proteomic profiles, with significant concentration differences in less than 5% of proteins at 4°C versus RT. Our metabolomics results are in agreement with previously obtained results showing that up to 2 h pre-centrifugation delays at room temperature only influence 2%–11% of detected metabolites in GC-MS metabolomic analyses.^14,38

Plasma and serum stability has been extensively reviewed,³⁹ showing that storage temperatures may influence the stability of specific proteins. Significant protein loss was observed in samples kept at RT for more than 4 h or at 4°C after 24 h. Repeated freeze/thaw cycles can also affect protein stability,³⁹ therefore freeze/thaw cycles should be avoided and must be reported in the Laboratory Information Management System.

We provide formal validation of methods for serum processing for downstream protein, metabolite or free nucleic acid-based applications and for plasma processing for downstream protein or free nucleic acid-based applications (Table 8). Because of the small number of samples used in our study, the power is limited and negative results should be interpreted with caution. The plasma collection protocol is not fit-for-purpose for downstream applications with platelets or microparticles, when citrate anticoagulant and microparticle-enriched plasma are more appropriate. It is not fit-for-purpose for downstream peptidomic applications, when protease inhibitors are more appropriate. Nonetheless, given that preanalytical variables can influence analytical results for specific parameters, laboratories and biobanks operating in the scope of accreditation are advised to systematically track and record preanalytical data using the Standard PREanalytical Code (SPREC).^40,41 Biobanks and clinical laboratories facing accreditation requirements and/or processing biospecimens for clinical biomarker research purposes can implement the serum and plasma preparation methods presented here with reference to this publication.

Table 8.

Sample Processing Method Overview

Parameter	Serum sample	Plasma sample
Primary container type	BD Vacutainer CAT	BD Vacutainer K₂EDTA
Pre-centrifugation delay between collection and processing	30–45 min at room temperature	0–2 h at room temperature
First centrifugation	2000 g; 10 min; 22°C; intermediate brake force	2000 g; 20 min; 22°C; intermediate brake force
Second centrifugation	NA	NA
Post-centrifugation	Collect supernatant, homogenize and aliquot	Collect supernatant, homogenize and aliquot
Long-term storage	−80°C	−80°C

NA, not applicable.

Footnotes

Acknowledgments

We thank Sarah MacKenzie PhD for medical writing assistance.

Author Disclosure Statement

No competing financial interests exist.

References

Bowers

. CAP takes the reins on biorepository accreditation. CAP Today Coll Am Pathol October 2012. http://www.cap.org/apps/cap.portal. Reference resources and publications/CAP Today/2012 Articles. Last accessed 28 June 2014 .

Technical Standard NF S96-900. [Quality in the Biology Resource Centers (BRC) - Management system of a BRC and quality of the biological resources] in French. September 2011.

Latterich

, Abramovitz

, Leyland-Jones

. Proteomics: New technologies and clinical applications. Eur J Cancer, 2008; 44:2737–2741.

Riegman

PHJ

, Morente

, Betsou

, et al. Biobanking for better healthcare. Mol Oncol, 2008; 2:213–222.

Clinical and Laboratory Standards Institute (CLSI). Procedures for the Handling and Processing of Blood Specimens; Approved Guideline, Fourth edition. Ed. Wayne, PA. H18-A4. Vol 30. No. 10. 2010;

Best Practices for Repositories: Collection, Storage, Retrieval and Distribution of Biological Materials for Research. ISBER. Third Edition. Biopreserv Biobank, 2012; 10:79–161.

Canadian Tumor Repository Network. Standard Operating Procedures. https://www.ctrnet.ca/operating-procedures. Accessed November 2013 .

Lehmann

, Roche

, Allory

, et al. [Preanalytical guidelines for clinical proteomics investigation of biological fluids] in French. Ann Biol Clin (Paris), 2009; 67:629–639.

Guerin

, Murray

, McGrath

, et al. Molecular medicine Ireland guidelines for standardized biobanking. Biopreserv Biobank, 2010; 8:3–63.

10.

Østergaard

, Nielsen

, Iversen

, et al. Quantitative proteome profiling of normal human circulating microparticles. J Proteome Res, 2012; 11:2154–2163.

11.

Mathivanan

, Simpson

. ExoCarta: A compendium of exosomal proteins and RNA. Proteomics, 2009; 9:4997–5000.

12.

Smalley

, Ley

. Plasma-derived microparticles for biomarker discovery. Clin Lab, 2008; 54:67–79.

13.

Lippi

, Blanckaert

, Bonini

, et al. Haemolysis: An overview of the leading cause of unsuitable specimens in clinical laboratories. Clin Chem Lab Med, 2008; 46:764–772.

14.

Yin

, Peter

, Franken

, et al. Preanalytical aspects and sample quality assessment in metabolomics studies of human blood. Clin Chem, 2013; 59:833–845.

15.

Timms

, Arslan-Low

, Gentry-Maharaj

, et al. Preanalytic influence of sample handling on SELDI-TOF serum protein profiles. Clin Chem, 2007; 53:645–656.

16.

Hsieh

S-Y

, Chen

R-K

, Pan

Y-H

, et al. Systematical evaluation of the effects of sample collection procedures on low-molecular-weight serum/plasma proteome profiling. Proteomics, 2006; 6:3189–3198.

17.

Wong

, Moturi

, Angkachatchai

, et al. Optimizing blood collection, transport and storage conditions for cell free DNA increases access to prenatal testing. Clin Biochem, 2013; 46:1099–1104.

18.

González-Masiá

, García-Olmo

. Circulating nucleic acids in plasma and serum (CNAPS): Applications in oncology. OncoTargets Ther, 2013; 6:819–832.

19.

Hiller

, Hangebrauk

, Jäger

, et al. MetaboliteDetector: Comprehensive analysis tool for targeted and nontargeted GC/MS based metabolome analysis. Anal Chem, 2009; 81:3429–3439.

20.

Gluck

, Hoogland

, Antinori

, et al. EasyProt—An easy-to-use graphical platform for proteomics data analysis. J Proteomics, 2013; 79:146–160.

21.

Dayon

, Pasquarello

, Hoogland

, et al. Combining low- and high-energy tandem mass spectra for optimized peptide quantification with isobaric tags. J Proteomics, 2010; 73:769–777.

22.

Elias

, Gygi

. Target-decoy search strategy for increased confidence in large-scale protein identifications by mass spectrometry. Nat Methods, 2007; 4:207–214.

23.

Dey-Hazra

, Hertel

, Kirsch

, et al. Detection of circulating microparticles by flow cytometry: Influence of centrifugation, filtration of buffer, and freezing. Vasc Health Risk Manag, 2010; 6:1125–1133.

24.

Szpechcinski

, Dancewicz

, Kopinski

, et al. Real-time PCR quantification of plasma DNA in non-small cell lung cancer patients and healthy controls. Eur J Med Res, 2009; 14 Suppl 4:237–240.

25.

Thomas

. Haemolysis as influence and interference factor. eJIFCC vol 13(4). http://www.ifcc.org/ifccfiles/docs/130401002end.pdf. Last accessed 28 June 2014 .

26.

Breitwieser

, Müller

, Dayon

, et al. General statistical modeling of data from protein relative expression isobaric tags. J Proteome Res, 2011; 10:2758–2766.

27.

Sowemimo-Coker

. Red blood cell hemolysis during processing. Transfus Med Rev, 2002; 16:46–60.

28.

Lacroix

, Judicone

, Poncelet

, et al. Impact of pre-analytical parameters on the measurement of circulating microparticles: Towards standardization of protocol. J Thromb Haemost, 2012; 10:437–446.

29.

Fliniaux

, Gaillard

, Lion

, et al. Influence of common preanalytical variations on the metabolic profile of serum samples in biobanks. J Biomol NMR, 2011; 51:457–465.

30.

Rubin

, Crettaz

, Canellini

, et al. Microparticles in stored red blood cells: An approach using flow cytometry and proteomic tools. Vox Sang, 2008; 95:288–297.

31.

Chandler

. Microparticle counts in platelet-rich and platelet-free plasma, effect of centrifugation and sample-processing protocols. Blood Coagul Fibrinolysis, 2013; 24:125–132.

32.

Cheng

, Yi

, Kim

, et al. Plasma processing conditions substantially influence circulating microRNA biomarker levels. PloS One, 2013; 8:e64795.

33.

Rai

, Vitzthum

. Effects of preanalytical variables on peptide and protein measurements in human serum and plasma: Implications for clinical proteomics. Expert Rev Proteomics, 2006; 3:409–426.

34.

Findeisen

, Sismanidis

, Riedl

, et al. Preanalytical impact of sample handling on proteome profiling experiments with matrix-assisted laser desorption/ionization time-of-flight mass spectrometry. Clin Chem, 2005; 51:2409–2411.

35.

Di Girolamo

, Alessandroni

, Somma

, et al. Pre-analytical operating procedures for serum low molecular weight protein profiling. J Proteomics, 2010; 73:667–677.

36.

Kattlove

, Alexander

, White

. The effect of cold on platelets. II. Platelet function after short-term storage at cold temperatures. Blood, 1972; 40:688–696.

37.

, Liu

, Craft

, et al. Intrinsic peptidase activity causes a sequential multi-step reaction (SMSR) in digestion of human plasma peptides. J Proteome Res, 2008; 7:5112–5118.

38.

KamLage

, Maldonado

, Bethan

, et al. Quality markers addressing preanalytical variations of blood and plasma processing identified by broad and targeted metabolite profiling. Clin Chem, 2014; 60:399–412.

39.

Hubel

, Alptekin

, Skubitz

, et al. State of the art in preservation of fluid biospecimens. Biopreserv Biobank, 2011; 9:237–244.

40.

Lehmann R

. Guadagni

. Moore , et al. Standard preanalytical coding for biospecimens: Review and implementation of the Sample PREanalytical Code (SPREC). Biopreserv Biobank, 2012; 10:366–374.

41.

Betsou

, Lehmann

, Ashton

, et al. Standard preanalytical coding for biospecimens: Defining the sample PREanalytical code. Cancer Epidemiol Biomark Prev, 2010; 19:1004–1011.