Absolute Quantification of Bionanoparticles by Electrospray ... - DOIs

Jan 13, 2017 - This study presents an upgraded electrospray differential mobility analysis (ES-DMA) setup for the absolute quantification of bionanopa...
0 downloads 0 Views 885KB Size
Subscriber access provided by UNIV OF NEWCASTLE

Article

Absolute Quantification of Bio-nanoparticles by Electrospray Differential Mobility Analysis: An application to Lipoprotein Particle Concentration Measurements Noémie Clouet-Foraison, Francois Gaie-Levrel, Loic Coquelin, Géraldine Ebrard, Philippe Gillery, and Vincent Delatour Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.6b02909 • Publication Date (Web): 13 Jan 2017 Downloaded from http://pubs.acs.org on January 17, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Analytical Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 9

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Absolute Quantification of Bio-nanoparticles by Electrospray Differential Mobility Analysis: An application to Lipoprotein Particle Concentration Measurements Noémie Clouet-Foraison1, Francois Gaie-Levrel1*, Loic Coquelin1, Géraldine Ebrard1, Philippe Gillery2, Vincent Delatour1 1

Laboratoire National de Métrologie et d’Essais, LNE, Chemistry and Biology Division, 1 rue Gaston Boissier, 75724 Paris Cedex 15, France 2

University Hospital of Reims, Laboratory of Pediatric Biology and Research, 51092 Reims, France

Lipoproteins, Electrospray Differential Mobility Analysis, ES-DMA, Ion Mobility, non-HDL-P, apoB, Electrospray Efficiency, Enumeration. ABSTRACT: This study presents an upgraded Electrospray Differential Mobility Analysis (ES-DMA) set-up for the absolute quantification of bio-nanoparticle concentrations in biological samples, with a special focus on non-High Density Lipoprotein particle concentrations (non-HDL-P). Metrological characterization of the system’s analytical performances for concentration measurements shows that the mean intermediate precision relative standard deviation is 14% for biological samples, 6% for silica nanoparticles and less than 1% for diameter measurements. This study also demonstrates that the most accurate method for non-HDL-P quantification in native serum samples implies daily calculation of the Electrospray transmission efficiency (E) of the system with the WHO SP3-08 reference material. The establishment of the uncertainty budget reveals that the main contribution to particle concentration measurement uncertainties is the Electrospray transmission efficiency. This data additionally shows that E is not only low (approx. 15-20%) but also highly variable over time and strongly affected by sample composition. This work suggests that absolute enumeration of bio-nanoparticles is achievable with ES-DMA but provided that a special care is taken to quantifying E with a calibrator of nature and matrix highly similar to the samples ones.

INTRODUCTION In the early 1990’s, Dole, Fenn and co-workers developed a pioneering interface to vaporize and aerosolize molecules and large chemical compounds: the Electrospray (ES)1,2. This interface spawned a revolution in chemical analysis and was abundantly optimized and studied ever since for various applications3. Its successful coupling to Mass Spectrometry (ESIMS)4–6 and the numerous applications of this technique rendered ESI-MS an indispensable tool in analytical sciences. Although ESI-MS proved to be a method of choice for absolute quantification of a large range of biomolecules, it still suffers from limitations that make it inappropriate for the quantitative analysis of large intact entities, like nanoparticles. In the past decades, a new coupling of ES was reported for the analysis of (bio)-nanoparticles and large supramolecular entities: Electrospray Scanning Mobility Particle Sizer (ESSMPS)7–10, also known as Electrospray Differential Mobility Analysis (ES-DMA) or Gas Phase Electrophoretic Mobility Molecular Analyzer (GEMMA). SMPS have long been hailed by researchers in the field of aerosol and nanoparticle sciences. These systems rely on the mobility difference of airborne charged particles under an electric field and count particles after their separation according to their electrical mobility diameter11. Recent studies published on ES-DMA demonstrate its added-value and applications for the analysis of (bio)-

nanoparticles and supramolecular assemblies in various fields such as virology12,13, protein analysis9,14–18, (bio)-nanoparticle quantification19–22 or aggregation studies23–25. One promising application of ES-DMA analysis is advanced lipoprotein testing. Lipoproteins are large supramolecular assemblies of lipids and apolipoproteins, ranging from 10 to 60 nm26,27, that are devoted to lipid transport in blood. Some subclasses of these lipoproteins are involved in Cardiovascular Diseases (CVD) and especially in Atherosclerosis28,29. Recent studies demonstrated that the prevalence of CVD related events is correlated to high concentrations of Low-DensityLipoprotein Particles (LDL-P)30–34 and non-High-DensityLipoprotein Particles (non-HDL-P), also referred to as apolipoprotein-B (apoB) containing particles33. Studies published on lipoprotein analysis with ES-DMA highlighted its potential as a novel tool for CVD risk assessment and proved its relevance in clinical trials19,35,36. ES-DMA analysis of a colloidal suspension provides its number size distribution, i.e. the diameter and corresponding count of all particles present in the sample. The potential use of ES-DMA as a quantitative method for absolute enumeration of bio-nanoparticles in biological samples has been significantly studied over the past decade and different approaches have been investigated by various groups. For example, Li et al. suggested an absolute quantification method using regression models based on the non-specific aggregation of particles

1

ACS Paragon Plus Environment

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

generated by an ES source14,37,38. Caulfield et al. suggested a direct equation reflecting the physics of ES-DMA measurements19, and Cole et al.12 and Hutchins et al.20 used gold nanoparticles and protein standards to calibrate ES-DMA. However, to date, very few data are available on the most appropriate and reliable method for accurate particle quantification by ESDMA. In this paper, we present the metrological evaluation of the performances of an upgraded ES-DMA system for concentration measurements of 20 to 60 nm bio-nanoparticles in liquid samples. We additionally present the results of an accuracy study for the measurement of non-HDL-P concentrations in native human serum using different calibration strategies. Finally, we report the uncertainty budget and the major parameters contributing to the final uncertainty associated with ES-DMA particle concentration measurements.

EXPERIMENTAL METHODS Reagents and chemicals LNE RM LIPO 201a (hereafter referred as LNE 201a) is a human frozen serum pool prepared by Solomon Park Research Laboratories according to CLSI C37a Guidelines39. WHO SP3-08 reference reagent is a serum-based material that was value-assigned by immuno-nephelometry, the designated reference method for apoB40. CS-35nm is a colloidal suspension of 35 nm monomodal silica particles. NIST SRM 927e is a Certified Reference Material (CRM, NIST, MA USA) of purified Bovine Serum Albumin (BSA). Its concentration was certified by amino-acid analysis. Additionally, 25 single donations of human frozen serum were collected at Solomon Park according to CLSI C37a Guidelines. They were valueassigned by the Northwest Lipid Metabolism and Diabetes Research Laboratories (NWLMDRL) for apoB concentrations with the immuno-nephelometry designated reference method. Sample preparation All samples were gravimetrically diluted in a 20 mmol/L ammonium acetate buffer (AA in this paper) prepared from a 5 mol/L commercial solution (BioUltra, Sigma-aldrich, France). pH was adjusted at 8.4 using NH4OH 10 mmol/L. For the study on repeatability and intermediate precision, WHO SP3-08 and LNE 201a were both diluted 600-fold (v/v) and CS-35nm was diluted 100-fold (v/v). Three triplicates (n=3) were freshly prepared and analyzed each day, three days a week (p=3) during three weeks (27 assays overall). To evaluate accuracy, WHO SP3-08 and NIST SRM 927e were diluted 1000-fold up to 5000-fold. The 25 patient samples were diluted 5000-fold (v/v). To apply Li et al. methodology, an aqueous solution of 0.063% (v/v) sucrose (#84097, Sigma, USA) was prepared to measure the mean droplet size generated with our ES8. Instrumentation for aerosol measurements (Figure 1) Samples were injected in the ES via a pressuring chamber (MFCS EZ pressure generator 1 bar, Flow rate platform and FLUIWELL-4c device, Fluigent, Villejuif, France). A platinium electrode was immersed in the vial and connected to the earth for electrical grounding. A 25 cm long, 25 µm inner diameter (ID), uncoated silica capillary (TSI Inc. Marseilles, France) was inserted in the vial and connected directly to the ionization chamber entrance of the ES generator (model 3482, TSI Inc.). The pressure applied for sample injection was measured before and after each assay using a calibrated nano2

Page 2 of 9

flowmeter (Flow Unit XS, Fluigent). Pressure was adapted so that the injection flow rate was 100 ± 2 nL/min. High voltage was applied at the capillary tip (2.1 kV, ~320 nA) and a mixture of air (0.600 ± 0.002 L/min) and CO2 (0.100 ± 0.001 L/min) was used as carrier gas. The aerosolized highly charged particles generated by the electrospray were neutralized by a soft X-ray source positioned directly at the exit of the ES chamber, facing the capillary outlet (module integrated in the generator by TSI Inc.). The particles thus exited the neutralization source with a known charge distribution41 and passed through the nano-column of the Differential Mobility Analyzer (DMA, model 3085, TSI Inc.). After successive selection in the DMA according to their electrical mobility diameter, particles were detected and counted by the Water Condensation Particle Counter (WCPC, model 3788, TSI Inc.). The DMA was operated with a 6 L/min sheath flow and the WCPC sampling flow rate was set at 0.6 L/min. The excess aerosol inlet remaining (0.1 L/min) was directed towards an exhaust and filtered. The DMA diameter scanning range was set from 5 to 60.4 nm. The gas flowrates and their stability are critical to obtain a stable Taylor cone and a reproducible aerosol generation10,13. Mass-flow regulators (EL-FLOW - F 201 CV, Bronkhörst, France) were thus added to control air and CO2 flowrates in the ES. These regulators were calibrated by comparison with a mass flow transfer standard MolblocTM, calibrated with a dynamic gravimetric method, thus providing S.I. traceable flowrate measurements. The injection capillary was rinsed in-between each injection with the AA buffer. In order to remove proteins adsorbed at the capillary inner wall surface, an additional 1 min rinse was regularly done with a 10 mmol/L NaOH solution. In addition, injection capillaries were passivated by injecting a diluted serum sample at high flowrate before each analysis.

Data Analysis All data were recorded with the Aerosol Instrument Manager software (AIM V10, TSI Inc.) as raw particle counts per cubic centimeter of gas (#/cm3). Post-analytical data processing was carried out in two steps: (1) processing of the number size distribution to obtain the aerosol phase particle concentration - PNtot - and (2) calculation of the liquid phase particle concentration C. Number size distribution processing (aerosol phase) – PNtot The raw aerosol number size distributions were processed with in-house software that corrected for particle losses due to diffusional processes. Transfer and penetration efficiencies in the DMA and tubings are corrected for by this software as well as the counting efficiency of the WCPC. The number size distributions measurement uncertainties were expressed as the standard deviation calculated on the basis of Monte-Carlo simulations. The equations, physical models and parameters used in this software are thoroughly described in the work of Coquelin et al42,43 and summarized in Supplemental Information. Briefly, all the parameters involved in the equations for SMPS signal inversion and correction (temperature, pressure, gas flows, ionic pathways, penetration factors or DMA column dimensions for example) are listed in the software. The software generates simulated number size distributions based on these parameters’ values and associated uncertainties using the experimental data as model. For each simulation, it creates a random matrix of parameters, choosing randomly a

ACS Paragon Plus Environment

Page 3 of 9

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Figure 1: Scheme of the ES-DMA set-up and data analysis process. The ES-DMA apparatus is constituted of 3 separate modules: (1) an ES nebulization source allowing the generation of an aerosol of highly charged particles neutralized by a soft X-Ray source – (2) a DMA column that gradually selects the particles as function of their electrical mobility and (3) a Condensation Particle Counter (CPC) that counts the selected particles (partially adapted from TSI product sheet). The number size distribution of particles in solution is then reconstructed by AIM V.10 software presenting diameters on the x-axis and corresponding counts (dN, #/cm3) on the y-axis. The particle concentration in the aerosol phase (PN) is calculated as the area of the peak of interest which is used afterwards to calculate the associated particle concentration C in the liquid phase.

value within the uncertainty interval of each parameter. The software then generates a new number size distribution corresponding to a new matrix of parameters. In the end, the mean number size distribution and its associated standard deviation are calculated on all the Monte-Carlo simulations generated. For this study, the mean number size distributions were calculated with 1000 Monte-Carlo simulations. The aerosol phase particle concentrations - PNtot - were determined by peak integration. Repeatability and intermediate precision were calculated for LDL particle concentrations, i.e. particles between 20.2 and 26.9 nm. For accuracy, calculations were done on non-HDL particle concentrations, i.e. between 20.2 and 59.4 nm. As this diameter range includes LDL, IDL and VLDL particles19, the derived non-HDL-P concentration is equivalent to an ApoB concentration and can be directly compared to apoB measurements by immuno-nephelometry.

Evaluation of repeatability and intermediate precision of aerosol measurements Repeatability and intermediate precision of PNtot concentration measurements were evaluated with WHO SP3-08, LNE 201a and CS-35nm materials. Repeatability and intermediate precision of diameter measurements were evaluated on CS35nm only. Data were processed according to the ISO 5725-2 standard. Homogeneity of the variances and mean values were assessed with Cochran and Grubs tests and the nonhomogeneous data-sets were excluded. Repeatability and intermediate precision standard deviations (respectively SDr and SDR) were calculated with the homogeneous data-sets.

In order to account for the variability related to a normal use of ES-DMA over a prolonged period of time, some parameters were deliberately modified during this study. Injection capillaries were changed each week, the system was shut-down for the week-ends and the WCPC wick cartridge was replaced.

Liquid phase particle concentration calculation – C The methodology suggested by Li et al. for absolute protein quantification was tested on NIST SRM 927e standard following the protocol detailed in the article14. The bias between the measured and certified concentrations was calculated. Another methodology involving direct calculation of the liquid phase particle concentration C [mol/L] was also investigated and Equation 1 was developed: 10   !

 1     " with E [%] the transmission efficiency of the ES generator, gElectropsray [L/min] the gas flow rate in the ES chamber, L [L/min] the liquid flow rate in the injection capillary, PNtot [#/cm3] the aerosol phase particle concentration (peak area) and   the Avogadro number [#/mol]. Applying Equation 1 requires the measurement of the ES transmission efficiency E, i.e. the yield of aerosolized particles later measured by the SMPS system, which is a different notion from Ionization efficiency, an extensively studied parameter especially for ESI-MS analysis44,45. In this study, E was determined with five different approaches involving different reference materials (NIST SRM 927e and WHO

3

ACS Paragon Plus Environment

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

SP3-08) and different calibration strategies (linear regression and single point calibration) (Table 1). The first approach investigated was based on the calculation of a mean efficiency E measured over several experiments, E then being considered as a constant in Equation 1. E was cal( culated either as the ratio of the measured ( $%&' ) and certified ( ( ) concentrations (Equation 2 – Methods 1 & 2) or by determining the slope a of an ordinary least square linear regression model between the peak area PNtot and the calibrator’s concentration; *(! ) (Equation 3 – Methods 4 & 5). (

$_,-. 2 "

(  1 " 10/   3     0 Table 1: Sum-up of the five methods used to quantify the ES efficiency, materials and equations. Method Method 1

Calibrator NIST SRM 927e

Method 2

WHO SP3-08

Method 3

WHO SP3-08

Method 4

NIST SRM 927e

Method 5

WHO SP3-08

Method for the determination of E Mean efficiency calculated as the ratio between measured and certified concentrations over several days (Eq.2) Daily calculation of the efficiency as the ratio between measured and certified concentration (Eq. 2) Mean Efficiency calculated with a least square linear regression model over several days ( Eq. 3)

The second approach investigated consisted in using Equation 2 and WHO SP3-08 to calculate daily the generator’s efficiency, E Daily (Method 3). In contrary to Methods 1, 2, 4 and 5, E was then considered as a variable in Equation 1 rather than as a constant. The non-HDL-P concentrations in 25 patient samples were calculated with Equation 1 using the four different mean efficiencies measured previously. Edaily was measured on the same day as the patient samples were analyzed. The bias was assessed for each of the five methods between $1,-. obtained for each sample and the apoB immuno-nephelometry target value. Finally, the uncertainty budget for Cnon-HDL-P was established. Each parameter’s contribution was calculated according to the law of uncertainty propagation described in the Guide for Uncertainty Measurements (GUM). Briefly, the uncertainties associated to particle concentration measurements are determined according to Equation 4: 2( ) 3∑76 689 (56 )  2(56 )

4

With 56 the parameter, 2(56 ) the uncertainty associated to the parameter 56 and (56 ) the sensibility coefficient calculated :; as the partial derivative . The contribution =7< of the pa:7
(7< ) >(;)

5

Page 4 of 9

RESULTS AND DISCUSSION Repeatability and intermediate precision The repeatability variation coefficient (CVr) for aerosol phase particle concentration measurements (PNtot) is 10.2 % for WHO SP3-08, 5.1 % for LNE 201a and 3.6 % for CS35nm (Table 2). These results are consistent with literature: Guha et al. reported a 6 % CVr on PNtot for repeated measurements of 20 nm bacteriophages13, Caulfield et al. reported an intra-assay CVr < 9 % for LDL-P concentration measurements19 and Hutchins et al. determined a CVr < 6 % for repeated measurements of HDL-P concentrations20. An explanation to the differences observed between our results and previous publications could be related to data processing. The models used in our software for Monte Carlo simulations were chosen conservative so that a large variability of each parameter included could be taken into account. As a result, broad experimental conditions were simulated, even though they were not met on the days of the experiments. The uncertainties calculated from these simulations thus tend to be overestimated, which could explain the poorer repeatability we report. On the contrary, intermediate precision on PNtot was improved compared to literature. For biological samples WHO SP3-08 and LNE 201a, intermediate precision CVR is 14% and for CS-35nm, CVR is 6 % (Table 2). In contrast, Guha et al. reported a 20 % CVR on PNtot for bacteriophage measurements13 and Caulfield et al. reported an inter-assay CVR < 20 % for LDL-P concentration measurements19. An important variability source of ES-DMA measurements is the aerosol generation instability. It is the main consequence of a lack of control on flow rates13. As it includes calibrated mass-flow regulators for gases and a liquid flowmeter for sample injection, our set-up ensures improved flow rates stability over days. This was not the case with the former ES system which used rotameters for gas flow rates adjustments and for which no liquid flow control was possible. In addition, sample preparation prior to ES-DMA analysis can greatly impact intermediate precision, especially for biological samples. In this study, analyzed samples consisted of highly diluted raw materials while the results found in literature were obtained on purified materials. The simplicity of our preanalytical steps, together with a tight flow control, may be the explanation for the improved intermediate precision we report. However, Hutchins et al. reported an 11.4 % inter-assay variability for liquid phase particle concentration measurements of purified HDL20. The use of an automated procedure and less concentrated AA buffer at higher pH may explain their better intermediate precision. It can be speculated that these conditions may prevent proteins and HDL particles from adsorbing to the capillary inner wall. The nature of the capillary and the possible coating on its inner surface could also be an explanation to these improved performances. Furthermore, for both repeatability and intermediate precision, CVs vary depending on the material studied. Our hypothesis is that the characteristics of the sample such as matrix complexity, particle’s nature or size, induce varying particle behavior in the injection capillary. For example, non-specific adsorption to the capillary inner wall or progressive clogging by aggregation may occur and affect ES generation stability which will result in a poorer repeatability. Finally, ES-DMA provides extremely robust measurements

4

ACS Paragon Plus Environment

Page 5 of 9

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Table 2: Repeatability and intermediate precision calculated according to ISO 5725-2 standard for particle concentration and mean diameter measurements using the ES-DMA set-up described

Mean value Repeatability SDr Intermediate Precision SDR CVr [%] CVR [%]

Particle Concentration in aerosol phase [#/cm3] WHO SP3-08 LNE 201a CS-35nm Silica 1.2E+09 1.6E+06 3.7E+08 1.3E+08 8.1E+04 1.3E+07 1.8E+08 2.3E+05 2.2E+07 10.2 5.1 3.6 14.5 14.2 6.0

of the particles’ diameter with a 0.6 % CVR for CS-35nm. This is consistent with previous results: Hutchins et al20 reported a 1 % mean CV on diameter measurements for 10 nm-HDL particles and Motzkus et al26 reported a reproducibility standard deviation < 1 nm in an inter-comparison study on the same CS-35nm silica material measured by SMPS systems.

Liquid phase particle concentration calculation - C 14

Li et al. model for absolute quantification Experiments on NIST SRM 927e with Li et al14 methodology for protein absolute quantification were first performed in order to evaluate the accuracy of our ES-DMA set-up. Compared to the mean 65.8 ± 1.6 g/L BSA concentration reported by the authors, we obtained a lower mean concentration of 61.4 ± 10.4 g/L. The concentration certified by amino-acid analysis by IDMS was 67.4 ± 1.4 g/L. The mean bias calculated versus the certified BSA concentration was -8.9 ± 15.4 %. The model suggested by Li et al. is based on non-specific aggregation phenomena occurring during electrospraying. The size of the droplets generated at the capillary tip being directly correlated to its inner diameter, the probability that multiple particles get trapped in a same droplet is higher with a larger capillary that will generate larger droplets. As we used a 25 µm ID capillary, the equations suggested may not be directly applicable to our set-up and may have needed recalculating some of the factors implied for scale-down. In addition, the important results’ variability from one experiment to another resulted in large uncertainties. Finally, even though this method is a very interesting approach, it assumes that particles will aggregate which is very unlikely in the case of lipoproteins. Lipoproteins being large entities, the probability of trapping two of them in one droplet is much lower compared to BSA. This model is thus unsuitable for lipoprotein quantification. Determination of the Electrospray Transmission Efficiency Electrospray transmission efficiency was estimated using the five methods introduced previously (Table 1 – Figure 2). Using Method 1with NIST SRM 927e as calibrator (Figure 2.a), resulted in a mean E of 16.7% (SD = 2.3%) while the same model using WHO SP3-08 (Figure 2.b - Method 2) lead to a mean E of 20.2% (SD = 3.3%). Method 4 with NIST SRM 927e (Figure 2.d) resulted in a mean E of 18.1% (SD = 2.0%) and in a mean 22.1% E (SD = 3.0%) using WHO SP308 (Figure 2.e - Method 5). Finally, the daily calculated efficiencies (Method 3) were plotted on Figure 2.c and resulted in a 24.1% day-to-day CV. In a recent paper, Liu et al. estimated the transmission efficiency of a similar ES generator with soft X-Ray charge reduction. Using fluorescent uranine, they reported a 12.8% mean efficiency46. In this study, we report larger efficiencies 5

Mean diameter [nm] CS-35nm Silica 33.4 0.1 0.2 0.4 0.6

which could be explained by the differences between these two generator set-ups and configurations (capillary’s ID, distance to ES critical orifice or critical orifice diameter). Interestingly, we noted that the mean efficiencies calculated with WHO SP3-08 were 25% higher than with NIST SRM 927e. It can be hypothesized that this is due to differences in calibrators’ conductivities. Aerosolization by electrospraying relies on the formation of a Taylor cone due to the application of a high voltage to a conductive liquid exiting a capillary. Conductivity is thus a key parameter for aerosolization which is why samples must be diluted in high conductivity buffers8,47,48. In the present case, NIST SRM 927e contains 20 mmol/L of NaCl while this concentration is approximately 140 mmol/L in healthy patients’ sera. Therefore, even though calibrators were highly diluted in a 20 mmol/L AA buffer, increasing dilutions lead to decreasing ionic strengths. The fact that E decreased with increasing dilution (data not shown) supports that the lower the conductivity (i.e. high dilution), the lower the efficiency. Using a calibrator which conductivity differs from that of the measured samples could therefore introduce a calibration bias because of differing efficiencies. In addition, the higher efficiencies were obtained using WHO SP3-08, a serum-like material of higher conductivity than NIST SRM 927e. This tends to confirm our hypothesis of a positive correlation between sample conductivity and ES transmission efficiency.

Non-HDL-P liquid phase particle concentration A set of 25 patient samples was analyzed and non-HDL-P concentrations were calculated for each sample using either the mean efficiencies determined previously (Method 1, 2, 4 & 5) or the daily E measured (Method 3). Results are presented on Figure 3 as the mean absolute bias between Cnon-HDL-P and the corresponding immuno-nephelometry reference value. Figure 3 illustrates that using a mean experimental efficiency as intended with Methods 1, 2, 4 & 5, results in highly variable accuracy and poor precision. This highlights the large variability of E from one assay to another (Figure 2). In addition, Figure 3 shows that NIST SRM 927e (Methods 1 & 4), although it provides SI-Traceability to the results, is not a suitable calibrator to measure E because it results in a ~40% overestimation of non-HDL-P concentrations. An explanation to this significant bias could be the difference between the samples and the calibrator in terms of particle nature, matrix or conductivity. On the contrary, the use of WHO SP3-08 as a daily calibrator (Method 3) results in an acceptable +12% bias. The good agreement observed between apoB and Cnon-HDL-P calculated with this method could again be related to the calibrator composition. WHO SP3-08 is an HDL-purified serum containing real lipoproteins, hence this good comparability. However, the material was processed and stabilized by addition of surfac-

ACS Paragon Plus Environment

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 9

Figure 2: Electrospray transmission efficiencies determined with - a) Method 1: Mean efficiency calculated as the ratio between measured and certified concentrations using NIST SRM 927e - b) Method 2: Same as method 1 with WHO SP3-08 - c) Method 3: Daily calculation of the efficiency using WHO SP3-08 – d) Method 3: Mean Efficiency calculated with a least square linear regression model for NIST SRM 927e - e) Method 5: Same as method 4 with WHO SP3-08.

-tants and salts40 which implies a modified matrix that could be the reason for the remaining bias observed. Furthermore, it should be noted that WHO SP3-08 is a reference material intended for calibration of apoB immuno-assays and valueassigned by immuno-nephelometry. This method, designated as the reference, relies on antibody targeting of the apoB in serum and measures the concentrations by light scattering. Therefore, given the different physical principles and measurands these methods involve, the 12% mean difference we observe may not necessarily reflect an accuracy bias. Figure 4 shows the detailed comparison between immunonephelometry apoB and Cnon-HDL-P concentrations calculated using Method 3 for the 25 patient samples. For most samples, good agreement is seen between both concentrations. However, for six of them (CS 011, CS 013, CS 014, CS 016, CS 023 and CS 025) the error bars corresponding to their associated uncertainties do not overlap. This lack of comparability could be attributable to samples’ characteristics depending on serum composition. Further investigation would however be needed to verify this hypothesis and to identify parameter(s) that could explain these discrepancies. Very few other groups studied the accuracy of ES-DMA derived particle concentrations. Hutchins et al. reported very

satisfying correlations between Lowry and calibrated ESDMA for HDL particle concentration measurements20. Their work is however structured in two different parts, first the accuracy assessment and study of the calibration model they developed, then its application to clinical samples. The accuracy assessment was carried out on pure reconstituted HDL particles, i.e. pure proteins in buffer, and calibrated with pure protein materials. Given the observations we report in this study concerning matrix effect (Figure 3), the similar matrices could explain their results. However, the protocol described in their paper for clinical sample analysis involves UC preparation and dialysis. Although this probably was intended to purify HDL particles and normalize sample matrices, this supplementary step may have modified the accuracy due to the fact that the recovery of UC preparation steps is highly variable. The additional possible modifications of the matrix may also have impacted results accuracy. To our knowledge, the results we present on Figure 4 are the first data available on accuracy of ES- DMA derived lipoprotein concentrations in native serum samples. Those underline that choosing an adapted calibration renders possible particle concentration measurements in native serum by ES-DMA with a determined accuracy.

6

ACS Paragon Plus Environment

Page 7 of 9 Method 1

Method 2

Method 3

Method 4

Method 5

apoB immuno-nephelometry

70%

250.0

60%

200.0

Concentration (mg/dL)

50% 40% 30% 20%

Method 3: Daily Efficiency WHO SP3-08

150.0

100.0

50.0

10%

0.0

0% Day 1

Day 2

Day 3

CS 023 CS 006 CS 002 CS 005 CS 007 CS 003 CS 004 CS 011 CS 009 CS 014 CS 022 CS 001 CS 025 CS 021 CS 015 CS 013 CS 012 CS 010 CS 024 CS 017 CS 008 CS 016 CS 018 CS 019 CS 020

Mean Absolute Bias [%]

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Figure 3: Mean absolute bias of the non-HDL-P concentrations using Equation 1 and five different methods to determine the ES efficiency. Bias was estimated against the reference value determined by immuno-nephelometry for 25 patient samples, performed over three days.

Figure 4: Non-HDL-P concentrations calculated by dailycalibrated ES-DMA with WHO SP3-08 (Method 3 – white bars) and corresponding reference immuno-nephelometry concentrations (grey bars) for 25 patient samples. Data are presented in order of decreasing concentration.

Finally, non-HDL-P concentration measurements with our ES-DMA set-up and the associated model for data processing resulted in a mean 11% relative expanded uncertainty for native patient sample analysis. In comparison, Guha et al13 reported a ≈10-20% estimated uncertainty on virus particle concentration measurements and Cole et al12 reported similar results, although more variable.

WHO SP3-08 reference reagent. Results of the comparison between non-HDL-P concentrations measured by this calibrated ES-DMA, and apoB immuno-nephelometry were in good agreement. However, to achieve sufficient accuracy, particular attention should be paid to tight control of liquid and gas flowrates and to carefully estimating the ES transmission efficiency. This key parameter should indeed be measured with an adapted calibrator which nature and matrix composition should be as close as possible to those of the samples of interest. The results of this work additionally suggest that, to date, avoiding external calibration to determine ES transmission efficiency is still not reliable for accurate absolute particle concentration measurements by ES-DMA. In the actual stateof-the-art, ES-DMA cannot pretend to become a primary reference method because it requires calibration. An alternative to reach such a goal would be using internal standards. However, on the contrary to isotope dilution mass spectrometry, ES-DMA only allows discriminating entities according to their electrical mobility diameter. The use of spiked internal standards of close size and nature thus cannot be considered. To conclude, we would not recommend ES-DMA as “gold standard” to perform measurements of particle concentrations that would require sufficiently small measurement uncertainties. However, its wide possible applications and the good performances of an upgraded system including tight flow control and adapted calibration make it a very reliable system for research and routine purposes.

Figure 5: Uncertainty budget for the determination of Cnonfor a native serum sample using daily calibration with WHO SP3-08 (Method 3). HDL-P

An example of the detailed uncertainty budget of Cnon-HDL-P is presented on Figure 5 for a native serum sample. Results highlight that ES transmission efficiency is the major contribution and represents 86.9% of the overall uncertainty. Total peak integration, i.e. variability associated with the determination of PNtot, is the second most important source of uncertainty (9.0%) and is related to the system’s analytical performances and data processing. The liquid flowrate and dilution factor represent much smaller contributions (2.0%) while gas flowrates’ weights in the overall uncertainty are negligible (