Recovery of 400 Chemicals with Three Extraction Methods for Low Volumes of Human Plasma Quantified by Instrumental Analysis and In Vitro Bioassays

Human biomonitoring studies are important for understanding adverse health outcomes caused by exposure to chemicals. Complex mixtures of chemicals detected in blood − the blood exposome − may serve as proxies for systemic exposure. Ideally, several analytical methods are combined with in vitro bioassays to capture chemical mixtures as diverse as possible. How many and which (bio)analyses can be performed is limited by the sample volume and compatibility of extraction and (bio)analytical methods. We compared the extraction efficacy of three extraction methods using pooled human plasma spiked with >400 organic chemicals. Passive equilibrium sampling (PES) with polydimethylsiloxane (PDMS) followed by solid phase extraction (PES + SPE), SPE alone (SPE), and solvent precipitation (SolvPrec) were compared for chemical recovery in LC-HRMS and GC-HRMS as well as effect recovery in four mammalian cell lines (AhR-CALUX, SH-SY5Y, AREc32, PPARγ-BLA). The mean chemical recoveries were 38% for PES + SPE, 27% for SPE, and 61% for SolvPrec. PES + SPE enhanced the mean chemical recovery compared to SPE, especially for neutral hydrophobic chemicals. PES + SPE and SolvPrec had effect recoveries of 100–200% in all four cell lines, outperforming SPE, which had 30–100% effect recovery. Although SolvPrec has the best chemical recoveries, it does not remove matrix like inorganics or lipids, which might pose problems for some (bio)analytical methods. PES + SPE is the most promising method for sample preparation in human biomonitoring as it combines good recoveries with cleanup, enrichment, and potential for high throughput.


Table of content
Tables: Table S1-1: Temperature gradient and split of the TDU unit Table S1-2: Temperature gradient of the GC oven Table S1-3: Internal standard mix used for GC-HRMS Table S1-4: Solvent gradients for LC-HRMS analysis in positive and negative mode Table S1-5: Internal standard mix used for LC-HRMS Table S1-6: Index on prediction quality and standard deviation for concentration addition Table S1-7: Index on prediction quality and standard deviation for independent action Table S1-8: Effect recoveries (ER) per method and bioassay/cell line calculated with equation (3) from the effect concentrations in Table S2-6.

Text S2. Lipid removal
Lipids may can cause problems in both, bioassays and instrumental analysis by causing unspecific effects, 1 binding chemicals and therefore reducing the bioavailability in bioassays, 2 or influencing ionization efficiency in the mass spectrometer. 3,4 n this study we evaluated the Phree® phospholipid removal plates 5 in combination with SPE.
The total sample volume was 300 µL with the following sample types: A) 300 µL PBS (blank) B) 300 µL pooled plasma (unspiked) C) 64 µL of a 500 ng/mL compound mix which was blown-down, then adding 300 µL of pooled plasma (spiked).All samples were prepared in glass vials.300 µL of sample were added to the 96-well Phree® phospholipid removal plate.900 µL ACN with 1% FA were added and samples were shaken at 300 rpm using an orbital plate shaker for 2 minutes.The plate was placed on top of a glass coated 96-well plate and elution took place by centrifugation at 500 g for 10 minutes.ACN was blown-down and 900 µL of water were added.Then, SPE was performed without adding 4% FA as described in the Solid-Phase-Extraction paragraph.

Text S3. PDMS-plasma partition constants
Independent experiments were performed to derive the PDMS-plasma partitioning constants (KPDMS/plasma.)Samples were prepared from 64 µL of a 500 ng/mL spike mix, which was blown down in a nitrogen stream and resolubilized in 300 µL of pooled human plasma.PES was added as 10 mg, 50 mg, 100 mg, and 150 mg pieces in triplicates.The samples were equilibrated in batch one for three days and batch two for eight days at 7.5°C and 400 rpm on an orbital shaker.Afterwards the PDMS was removed, washed in MilliQ water and tapped dry.The PDMS was extracted twice using ethyl acetate in the volumes of 1.05 mL (150 mg PDMS), 0.7 mL (100 mg PDMS), 0.35 mL (50 mg PDMS), and 0.07 mL (10 mg PDMS).The extract was evaporated to dryness under nitrogen, transferred into 2 mL autosampler vials with 200 µL glass micro inserts and reconstituted in 40 µL of MeOH.Samples were measured by GC-HRMS and LC-HRMS.
For the calculation of the KPDMS/plasma a mass balance equation for partitioning between PDMS and the plasma was set up, assuming negligible loss due to evaporation or binding to compartments besides plasma or PDMS.Since 64 µL of 500 ng/mL spike mix were used and 25% of extract were injected into the LC and GC systems, the total mass with 100% chemical recovery equals 8 ng.The respective concentrations in PDMS (CPDMS,i) and plasma (Cplasma,i) per chemical i were calculated by equation S(1) and the concentration remaining in plasma with equation S(2).The factor 1.025 is the correction for density (g/mL) of human plasma.
C plasma,i ng g = 8 ng×(1-Recoveryi) V plasma (mL)×1.025(g mL ) The KPDMS/plasma of chemical i can be described as the concentration ratio (equation S(3)). 6DMS/plasma,i = The final PDMS partition constant KPDMS/plasma,i were calculated from the mean of all experiments.
The experimental KPDMS/plasma were compared with theoretical predictions from a simplified mass balance model (equation S(4)) adapted from Baumer et al., 7 where the partitioning constant between PDMS and lipids, KPDMS/lipid, was approximately constant over a wide hydrophobicity range and amounted to 10, 8 the partition constant between lipids and serum proteins, Klipid/protein, was approximately constant over a wide hydrophobicity range and predicted as 20. 7The mass fraction of lipid, mflipid was measured as in Baumer et al. 7 and amounted to 0.00446, that of proteins mfprotein was 0.05647.The remainder was assumed to be non-binding and assigned to water, mflwater, and was calculated as mfwater = 1mflipidmfprotein = 0.93907.

Text S4. GC-HRMS analysis
Samples were prepared by adding 10 µL of extract and 5 µL of 250 ng/mL internal standard mix to 10 µL of MeOH.Pure MeOH was used as solvent blanks.GC-HRMS analysis was performed on Thermo Trace 1310 GC system and Q Exactive TM Orbitrap TM .A J&W DB-5ms Ultra Inert GC Column, 30 m x 0.25 mm, 0.25 µm film thickness was used for separation.With helium as carrier gas.For injection, we used a Gerstel multi-purpose autosampler with thermal desorption unit (TDU) and a Gerstel cold injection system (CIS).The injection volume was 2 µL into thermodesorption tubes equipped with glass inserts.The TDU was heated according to Table S1-1.The CIS was initially cooled to -20°C and heated up quickly to 300°C using a rate of 720°C/min in splitless mode at 75 kPa for desorption.The CIS transfer temperature was 320°C.This was the injection of the sample into the GC system which had a subsequent oven temperature gradient as shown in Table S1-2 at a constant flow of 1.2 mL/min.
The use of a TDU in combination with CIS is applied to reduce the injection of matrix by initially desorbing the sample into the CIS leaving behind non-desorbing components of the sample in the glass tubes (which can be considered as single use liners) and is the standard injection system used for this instrument.
During measurements the ion source temperature was 200°C and transfer line temperature was 250°C.Electron ionization was used at an emission current of 50 µA and an electron energy of 70 eV.
The instrument was operated in full scan MS1 mode with a m/z range from 60 -810 and a nominal resolving power of 60,000 referenced to m/z 200.The internal standard mix consisted of the compounds listed in Table S1-3 and was at 1 µg/mL.Internal standards were assigned by nearest retention time or to the respective analyte without isotope label.The internal standard mix consisted of the compounds listed in Table S1-3 and was at 1 µg/mL.Internal standards were assigned by nearest retention time or to the respective analyte without isotope label.
Table S1- A total of 500 mg PDMS was prepared per sample by cutting 12 pieces of the size of 1 mm thickness.Total volume was 900 µL with the following sample types: A) 900 µL PBS ("blank"), B) 300 µL pooled plasma + 600 µL PBS ("unspiked plasma"), C) 64 µL of a 500 ng/mL spike mix which was blown down in a nitrogen stream, then adding 300 µL of pooled plasma and 600 µL PBS ("spiked plasma").All samples were prepared in glass vials.
To investigate the necessary time for equilibration in human plasma, the extraction was performed with a chemical spike mixture with a lower number of components (identified by the respective column in Supporting Information S2, Table S2-1) in triplicates for one, three, and six days.After preparation, the samples were equilibrated on an orbital shaker with 400 rpm for the respective duration at 7.5°C.After this time, the PDMS was removed, washed in MilliQ water and tapped dry on lint-free tissue.The PDMS was extracted for one day with 3.5 mL of ethyl acetate on a horizontal shaker at room temperature.This step was repeated once.The extract was evaporated to dryness under nitrogen, transferred to 2 mL autosampler vials with 200 µL glass micro inserts and reconstituted in 40 µL of MeOH.The supernatant was prepared for SPE by adding 300 µL of 4% of formic acid.SPE was performed with 96-well SPE plates with HLB sorbent using a negative pressure unit with a suitable manifold.The SPE plate was conditioned with 1 mL of ethyl acetate, 1 mL of methanol, and 1 mL of water.The sample was transferred, extracted, and the sorbent washed with 0.5 mL of 5% MeOH in water.The plates were left under vacuum for 30 minutes and subsequently centrifuged at 1,500 × g for 30 minutes.The plates were left over night in a desiccator at room temperature to achieve full dryness.The plates were eluted using 1 mL of MeOH, which was collected in glass-coated 96-well plates.The eluate was evaporated to near dryness under nitrogen, transferred to 2 mL autosampler vials with 200 µL glass micro inserts, evaporated to dryness and reconstituted in 40 µL of MeOH.
As depicted in Figure S1-2, PDMS alone had a mean ± standard deviation recovery of 22 ± 23% (one day), 25 ± 25% (three days), and 25 ± 25% (six days) with n = 351.Details regarding respective recoveries can be accessed from Supporting Information S2, Table S2-  The rank means were significantly different with p < 0.0001 for paired Friedman test with recoveries after one day equilibration time being significantly lower (p < 0.0001) than three days and six days equilibration time.There was no significant difference in rank sums of recoveries after three days and six days equilibration (p = 0.2791).For SPE after PDMS the mean ± standard deviation recoveries were 26 ± 22% (one day), 27 ± 24% (three days), and 23 ± 22% (six days) with n = 285.Individual chemical's recover can be accessed in the Supporting Information S2, Table S2-3.The rank means of all distributions were significantly different using paired Friedman test with p < 0.0001.All multiple comparisons were significantly different with p < 0.0001 with three days having the highest recoveries.

Text S7. PDMS-plasma partition constants
The KPDMS/plasma were converted from the recoveries with equation S(3) for all neutral chemicals (Table S2-7).The mean KPDMS/plasma,i per chemical i of chemicals with logKow > 2 did not show a strong dependence of hydrophobicity expressed as logKow (Figure S1-3, Table S2-7, mean per compound in Table S2-8).The slope of the linear regression over nine orders of magnitude was merely 0.12, and the KPDMS/plasma varied unsystematically over three orders of magnitude with a mean logKPDMS/plasma of -0.154 (standard deviation 0.80, 95% confidence interval -0.238 to -0.070, KPDMS/plasma 0.70) but there appears to be an upward trend at logKow ≤ 2. These KPDMS/plasma were approximately a factor of 10 to 100 lower than the KPDMS/blood measured in full blood of humans 9 and turtles, 9 despite the higher protein and lipid content of full blood. 10The simple mass balance model (equation S( 4)) provided a maximum estimate but overestimated the measured KPDMS/plasma (Figure S1-4) for many chemicals.The model and the data showed an apparent trend that partition constants were increasing from 0 < logKow < 2 and independent of hydrophobicity at logKow > 2. For all directly fitted assay data, the standard errors were calculated using the included functions in GraphPad Prism assuming symmetric confidence intervals.For the mixture calculations, the error in the predicted effect concentration for effect level F of concentration addition, ECCA,F, was calculated using equation S (5).
For independent action, first the errors of effects at level F per chemical i were calculated using equation S (6), which were then used to calculate the error of predicted mixture effects via equation S (7).
The deviation of effect was calculated using equation S( 8) at the effect level F of interest.For AREc32 a linear evaluation was selected and hence the deviation was calculated according to equations S(9) and S (10).

S15
Text S9.Index on Prediction Quality The comparisons between predicted and measured mixture effects of the binary mixture of unspiked samples and reference chemical were expressed as index on prediction quality (IPQ, equation 7) and standard deviation (SD).IPQs are summarized for concentration addition in Table S1-6 and for independent action in Table S1-7.

5 :
Internal standard mix used for LC-HRMS Mono-Time-dependence of recovery with PES 2.

Figure S1- 4 :
Figure S1-4: PDMS-plasma partition constants logKPDMS/plasma of the neutral chemicals (n = 352) as a function of the octanol-water partition constant logKow.Data from Table S2-8.red dashed line: mean of all logKPDMS/plasma, blue line prediction by the mass-balance model (MBM) described by equation S(4) with Klipid/PDMS of 10.
Figure S1-5: Concentration-response curves of method blanks in all cell lines.Effect and inhibitory concentration values (ECF and ICF) listed in Table S2-6.

Figure S1- 7 :
Figure S1-7: Concentration-response curves of the spike mix in all cell lines.The vertical dotted line indicates the inhibitory concentration at 10% cytotoxicity IC10).Effect and inhibitory concentration values (ECF or ECIR1.5 for AREc32 and IC10) are listed in Table S2-6.

Figure S1- 9 :
Figure S1-9: Concentration-response curves of unspiked plasma samples in PPARγ-BLA.The vertical dotted line indicates the inhibitory concentrations at 10% cytotoxicity Effect and inhibitory concentration values (ECF and IC10) are listed in Table S2-6.

Figure S1- 10 :
Figure S1-10: Concentration-response curves of unspiked plasma samples in AhR-CALUX.The vertical dotted line indicates the inhibitory concentration at 10% cytotoxicity Effect and inhibitory concentration values (ECF and IC10) are listed in Table S2-6.

Table S1 -
1: Temperature gradient and split of the TDU unit

Table S1
LC-HRMS was performed on Thermo Ultimate 3000 LC system with an electrospray ion source and a Q Exactive Plus quadrupole-Orbitrap instrument.The column was an ACQUITY UPLC® BEH C18, 100 x 2 mm, 1.7 µm particle size (Waters).Eluents were 1 mM ammonium formate with 0.1% FA in water and methanol for positive mode, and for negative mode 2 mM ammonium bicarbonate in water and 95:5 MeOH:water.90:10 water:MeOH was used for solvent blanks.Samples were prepared by adding 10 µL of extract and 10 µL of 100 ng/mL internal standard mix to 180 µL of water.The injection volume was 50 µL.The heated ESI source and the transfer capillary were both operated at 300°C, the spray voltage was 3.8 kV (positive mode) or 3.5 kV (negative mode), the sheath gas flow rate was 45 a.u. and the auxiliary gas flow rate 1 a.u.