Fabric Phase Sorptive Extraction—A Metabolomic Preprocessing Approach for Ionizing Radiation Exposure Assessment

Abstract

The modern application of mass spectrometry-based metabolomics to the field of radiation assessment and biodosimetry has allowed for the development of prompt biomarker screenings for radiation exposure. Our previous work on radiation assessment, in easily accessible biofluids (such as urine, blood, saliva), has revealed unique metabolic perturbations in response to radiation quality, dose, and dose rate. Nevertheless, the employment of swift injury assessment in the case of a radiological disaster still remains a challenge as current sample processing can be time consuming and cause sample degradation. To address these concerns, we report a metabolomics workflow using a mass spectrometry-compatible fabric phase sorptive extraction (FPSE) technique. FPSE employs a matrix coated with sol–gel poly(caprolactone-b-dimethylsiloxane-b-caprolactone) that binds both polar and nonpolar metabolites in whole blood, eliminating serum processing steps. We confirm that the FPSE preparation technique combined with liquid chromatography-mass spectrometry can distinguish radiation exposure markers such as taurine, carnitine, arachidonic acid, α-linolenic acid, and oleic acid found 24 h after 8 Gy irradiation. We also note the effect of different membrane fibers on both metabolite extraction efficiency and the temporal stabilization of metabolites in whole blood at room temperature. These findings suggest that the FPSE approach could work in future technology to triage irradiated individuals accurately, via biomarker screening, by providing a novel method to stabilize biofluids between collection and sample analysis.

Graphical Abstract

INTRODUCTION

Exposure to ionizing radiation (IR) can cause acute or long-term deleterious effects dependent on the rate of exposure, dose, radiation quality, type of exposure (whole or partial body), and internal vs external exposure. Active research has been dedicated to the development of radiation biomarkers to better estimate absorbed radiation dose in certain accidental situations and to guide risk stratification of exposed populations. Mass spectrometry-based metabolomics has become a reliable platform tied to the discovery of radiation biomarkers in mice, rats, nonhuman primates (NHPs), and humans and has provided valuable insight into the metabolic perturbations induced by different IR sources,^1,2 doses, and dose rates.^3–6 Several researchers have established metabolic markers found in easily collected biofluids,^6–9 such as blood,^4,10–15 urine,^1,13,16–20 feces,^10,21–23 and saliva,^24–26 which may help with evaluating injury and treatment assessment during a nuclear or radiological incident.

Following a radiological incident, individual samples will need to be collected, packaged, and shipped to collaborating laboratories to confirm potential exposures to radiation. Currently, methods to assess metabolomics biomarkers require several pretreatment steps such as protein precipitation, centrifugation, solvent extraction, sonication, and preconcentration as well as post-treatments including solvent evaporation and sample reconstitution. These time-consuming steps can introduce sample handling errors into the analysis and require at least 100–200 μL of whole blood. The reality remains that during a radiological emergency the demand to process samples from exposed responders and individuals will reach into the thousands. Therefore, it is crucial to develop a deployable extraction technique that can minimize pre/post-treatment steps, can utilize small volumes of whole blood from a fingerprick, and is compatible with the current high-throughput analytical tools. Here we examine a fabric phase sorptive extraction (FPSE) technique that not only allows for processing and stabilization of whole blood while eliminating the need for serum separation but also is compatible with the downstream UPLC-TOFMS analysis.

Fabric phase sorptive extraction (FPSE),²⁷ uniquely addresses most of the aforementioned shortcomings of conventional techniques. FPSE has been optimized and popularized in the fields of environmental pollutant analysis,^28–34 food safety,^35,36 and biological and pharmaceutical analysis^34,37–43 as it offers a high-throughput approach easily paired with liquid chromatography-mass spectrometry (LC-MS). FPSE offers a number of advantages over classical sample preparation techniques including: (a) simple, fast, and minimized solvent usage; (b) direct introduction into the sampling container; (c) use of any organic solvent for back-extraction; (d) minimal number of steps in sample preparation workflow, (e) no solvent evaporation or sample reconstitution needed due to small solvent volume, and (f) maximizes extraction from whole blood.^42,43

The FPSE technique can utilize a variety of polymeric sol–gel microextraction sorbents that can be distributed on multiple substrates such as cellulose, polyester, or fiberglass fibers. Different sol–gel sorbents, expectedly, enhance the extraction of different molecules. Prior studies using polydimethyldiphenylsiloxane (PDMDPS), polytetrahydrofuran (PTHF), polyethylene glycol-block-polypropylene glycol-block-polyethylene glycol (PEG-b-PPG-b-PEG) triblock, and polyethylene glycol (PEG)²⁹ have focused on sorbent/substrate pairs tailored toward the polarity of a target molecule, but for radiation metabolomics an assay of multiple biomarkers spanning polarities will be needed for proper assessment. For this study a biocompatible medium-polarity sorbent coating was created using the poly(caprolactone-block-dimethylsiloxane-block-caprolactone) (PCL-b-PDMS-b-PCL) triblock co-polymer.⁴⁴ This study, to our knowledge, is the first to apply FPSE sol–gel PCL-b-PDMS-b-PCL-coated cellulose, polyester, or fiberglass membranes and to compare their extraction efficiency, temporal stability, and ability to preserve the integrity of specific IR metabolomic signatures.

MATERIALS AND METHODS

Chemicals and Materials

All solvents (acetonitrile, methanol, water) were UPLC-MS grade (Fisher Scientific Inc., Hanover Park, IL), and chemicals (debrisoquine sulfate, 4-nitrobenzoic acid, chlorpropamide, l-carnitine, linolenic acid, oleic acid, and arachidonic acid (MilliporeSigma, St. Louis, MO) were the highest purity available. Centrifugation of sol solutions to obtain particle-free solutions was carried out in an Eppendorf Centrifuge model 5415 R (Eppendorf North America Inc. USA). A Fisher Scientific Digital Vortex Mixer (Fisher Scientific, USA) was used for thoroughly mixing different solutions. A 2510 BRANSON Ultrasonic Cleaner (Branson Inc., USA) was used to trap gas molecules from the sol solutions. A Barnstead NANOPure Diamond (model D11911) deionized water system (Dubuque, IA) was used to obtain ultrapure deionized water (18.2 MΩ) for sol–gel synthesis. Methyltrimethoxysilane (MTMS) and trifluoroacetic acid (TFA) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Sodium hydroxide and hydrochloric acid were purchased from Thermo Fisher Scientific (Milwaukee, WI, USA). PCL-B-PDMS-B-PCL triblock copolymer was purchased from Gelest, Inc. ((Morris-ville, PA, USA). The FPSE media was laser cut (Versa Laser VLS2–30 from Universal Laser, AZ) into small circular disks at 5 mm in diameter.

Preparation of Sol–Gel Poly(caprolactone-block-dimethylsiloxane-block caprolactone)-Coated Fabric Phase Sorptive Extraction Media

FPSE membrane preparation was performed as previously reported.^43,45,46 The sol solution for creating sol–gel poly-(caprolactone-block-dimethylsiloxane-block-caprolactone) coating on a cellulose, polyester, or fiberglass fabric substrate was prepared by sequentially adding poly(caprolactone-block-dimethylsiloxane-block-caprolactone) (PCL-B-PDMS-B-PCL) polymer, tetrahydrofuran, methyltrimethoxysilane, trifluoroacetic acid, and water into the reaction container a molar ratio of 0.2:1:4:0.48:3, respectively. The mixture was thoroughly vortexed after adding each of the individual ingredients into the sol solution. Subsequently, the sol solution was vortexed for 3 min and centrifuged for 5 min. The clear, supernatant solution was then carefully transferred into a clean 60 mL amber-colored glass reaction bottle. The clean and preconditioned fabric was gently submerged into the sol–gel solution. The respective fabric was kept in the sol solution for 6 h. During the residence time into the sol solution a sol–gel network with randomly incorporated PCL-b-PDMS-b-PCL polymer continues to grow under acidic hydrolysis and polycondensation. In addition to the expanding sol–gel network, the sol–gel network also gets chemically bonded to the cellulose substrate via polycondensation. Upon completion of the coating period, the sol solution was expelled from the reaction bottle and the coated fabric was dried and aged in a homemade conditioning device built inside a gas chromatography oven with continuous helium gas flow at 50 °C for 24 h. Before using the FPSE, the sol–gel PCL-B-PDMS-B-PCL-coated FPSE media was rinsed sequentially with methylene chloride and methanol followed by drying at 50 °C under an inert atmosphere for 1 h. At this condition, the sorbent loading of sol–gel PCL-B-PDMS-B-PCL was calculated as 6.14 mg/cm².

Animal Experiment

Eight-week-old male C57BL/6J mice (Jackson Laboratories, Bar Harbor, ME) were maintained in a 12 h dark and 12h light cycle at 22 °C in 30–70% humidity and provided a certified rodent diet along with filtered water ad libitum. Mice (n = 6 mice per study group) were exposed to 0 (sham) or 8 Gy whole-body X irradiation using an XRAD-320 X-ray irradiator (Precision Xray, North Branford, CT). We used 320 kV and 12.5 mA with a 1.5 mm Al/0.25 mm Cu/0.75 mm Sn low-pass filter placed in front of the source for the 8 Gy irradiation with a mean dose rate of 86.30 cGy/min. Immediately after irradiation, mice were returned to their home cages and monitored regularly. Blood samples were collected at necropsy by cardiac puncture at 24 h (hs) after exposure. Serum was recovered from a paired sample of whole blood and separated using BD Vacutainer SST tubes. All animal procedures were performed according to protocols approved by the Institutional Animal Care and Use Committees (IACUC) at Georgetown University. For our research we followed the Guide for the Care and Use of Laboratory Animals, prepared by the Institute of Laboratory Animal Resources, National Research Council, and U.S. National Academy of Sciences.

FPSE Experiment

One FPSE membrane was placed in a siliconized 1.5 mL tube and conditioned with 500 μL of 1:1 methanol:acetonitrile. After a 5 min incubation, the membrane was carefully removed and placed in a 96-well plate and dried under vacuum. A 50 μL amount of whole blood or serum was then added to the membrane and incubated at room temperature for 24 h. The membrane was then dried under vacuum and moved to a new siliconized tube before addition of 100 μL 1:1 methanol:acetonitrile. After incubation for 10 min, the membrane disk was then removed from the eluent and the tube was centrifuged at 17 200g for 5 min at room temperature. A 20 μL amount of the supernatant was recovered and diluted with 2% acetonitrile in HPLC-grade water containing the internal standards debrisoquine sulfate (2 μM), 4-nitrobenzoic acid (30 μM), and chlorpropamide (5 μM). For temporal stability studies, serum was separated and then pooled prior to aliquoting 50 μL onto the FPSE membrane. After the extraction, samples were then subjected to LC-MS as described below. As a comparison to traditional serum sample preprocessing, serum samples were also prepared by diluting the sample in 1:20 water:acetonitrile, followed by centrifugation at 17200g for 5 min at 4 °C. The supernatant was recovered and subjected to LC-MS analysis. Prior to analysis the total protein concentration of each sample (2 μL) was assessed with a micro BCA protein assay kit (Thermo Fisher Scientific, Grand Island, NY).

Metabolomic Analysis

Metabolomic analysis was performed by injecting a 2 μL aliquot of each sample into a reverse-phase 50 × 2.1 mm H-class UPLC Acquity 1.7-μM BEH C18 column coupled to a Xevo G2 QTOF (Waters). The mobile phase consisted of water and 0.1% formic acid (solvent A), 100% acetonitrile (solvent B), and isopropanol:acetonitrile (90:10 v:v) with 10 mM ammonium formate (solvent C). The gradient for this analysis switched from 98% aqueous solvent A to 40% solvent A and 60% solvent B after 4 min and to 98% solvent B at 8 min for 2 min and to 75% solvent C for 3 min and back to 98% solvent A for the last 1 min of the 13 min gradient at a flow rate of 0.5 mL/min. The Xevo G2 QTOF mass spectrometer was operated in positive (ESI+) and negative (ESI−) electrospray ionization (ESI) modes over a mass range from 50 to 1200 Da in two channels, MS and MS^E, with a low collision energy of 10 eV for the precursor ions and a collision energy ramp of 10–50 eV for the productions. The lock spray consisted of leucine–enkephalin (556.2771 [M + H]⁺ and 554.2615 [M – H]⁻), and the data was acquired in centroid mode. Quality control (QC) samples consisting of a pooled sample were run every 10 samples for instrument quality assessment and retention time drift. Putatively identified ions were validated by UPLC-QTOF tandem (MS/MS) mass spectrometry as previously described.⁴⁷ Pure standards (dissolved in 1:1 water:acetonitrile) were fragmented with a 5–50 eV ramped collision energy. Mass spectrometry data have been deposited to the Center for Computational Mass Spectrometry Database and can be accessed through the following links : ftp://massive.ucsd.edu/MSV000083720and ftp://massive.ucsd.edu/MSV000083719.

Metabolomics Data Processing

Progenesis QI software (NonLinear Dynamics, Newcastle UK) was used to assign peaks for each ion using the raw chromatograms. Peaks were aligned in Progenies based on the internal standards debrisoquine sulfate, 4-nitrobenzoic acid, and chlopropamide. The software MetaboLyzer⁴⁸ was used to analyze the data and identify statistically significant ions. MetaboLyzer allowed for extraction of the ions with nonzero abundance values, which were detected in at least 70% of samples in each study group, called complete-presence ions. Data were then log transformed; outliers were removed via 1.5 interquartile range (IQR), analyzed for statistical significance via the nonparametric Mann–Whitney U statistical hypothesis test (p value < 0.05), and FDR adjusted (<0.1). Statistical significance testing for ions with nonzero abundance values in at least 70% of the samples in only one group (partial-presence ions) were analyzed as categorical variables for presence status (i.e., nonzero abundance) via Fisher’s exact test (p value < 0.05). The log-transformed data for statistically significant complete-presence ions were then utilized for an unsupervised principal component analysis (PCA) via singular value decomposition for the purpose of data visualization. MetaboAnalyst⁴⁹ was used to generate a multigroup PCA. In MetaboAnalyst the aligned data was log transformed, and outliers were removed via 1.5 IQR. Statistically significant ions were putatively identified via MetaboLyzer, which utilizes the Human Metabolome Database (HMDB),⁵⁰ LipidMaps,⁵¹ and the Kyoto Encyclopedia of Genes and Genomes (KEGG) database,^52,53 and confirmed using the METLIN^54,55 database with Progenesis QI. The m/z values were used to putatively assign IDs to the ions by neutral mass elucidation, which was accomplished by considering the possible adducts, H⁺, Na⁺, or NH₄⁺ in the ESI+ mode and H⁻ and Cl⁻ in the ESI− mode. The masses were then compared with the exact mass of small molecules in the databases, from which putative metabolites were identified with a mass error of 20 ppm (ppm) or less. KEGG-annotated pathways associated with these putative metabolites were also identified. To extract structural information on the putative identities of the metabolite, we explored MS^E data via Progenesis QI. Putatively identified ions were validated using retention times and fragmentation patterns from the experimental samples compared to pure standards and further verified using the METLIN database. Receiver operating characteristic (ROC) curves for each validated metabolite assessed, graphical representation, and P values after identification were all determined with Prism 6 (GraphPad Software Inc., La Jolla, CA). The 4-biomarker ROC curve and area under the curve (AUC) values were generated using multivariate exploratory ROC analysis with MetaboAnalyst based on VIP scores (PLSDA).

RESULTS

FPSE Membrane Substrate Effects on IR metabolome Coverage

As a proof of concept, we compared the efficiency of sol–gel PCL-b-PDMS-b-PCL-coated cellulose, polyester, or fiberglass membranes (Figure 1) at extracting metabolites perturbed by radiation. We examined a small quantity (50 μL) of serum or whole blood using each membrane. Additionally, the FPSE technique was compared to a traditional serum metabolomics method. Here, samples were incubated on the membrane for 24 h to mimic possible shipping transit times between sample collection and sample analysis. Sample extracts were analyzed using UPLC-TOFMS followed by statistically based data reduction (Mann–Whitney U, FDR adjusted, p < 0.05) and putative metabolite identification using the software MetaboLyzer.⁴⁸

Figure 1.

Chemical structure of sol–gel PCL-b-PDMS-b-PCL triblock polymer and exploded visual representation of polymeric coating porous fabric substrate. One membrane is made of fabric substrate interior and coated on the top and bottom with sol–gel polymer and measures 5 mm in diameter.

The sol–gel FPSE method captured the altered metabolomic signature of serum samples from mice exposed to 8 Gy vs serum from sham mice with variations in coverage based on substrate type (Figure 2). Cellulose and polyester membranes recovered more statistically significant putatively identified metabolites ($\overline{\text{x}}=203$ and 178, respectively) when assessing sham versus irradiated samples compared to the traditional extraction method ($\overline{\text{x}}=164$) (Figure 1A). Using a fiberglass base, however, resulted in fewer significant features ($\overline{\text{x}}=80$). A key strength of sol–gel FPSE is its ability to eliminate blood preprocessing and simplify the sample preparation protocol; thus, we evaluated paired whole blood and serum samples. As expected, chromatograms from whole blood samples had more unique peaks than those from serum (Supplemental Figure S1). The number of putative metabolites identified in whole blood were similar to those found in serum samples and were mainly by FPSE substrate-based modifications (Figure 2A). Multivariate statistics, using PCA, further highlights the patterned differences found in whole blood samples when using a different membrane substrate. Though each membrane adequately separates the 8 Gy samples from the sham samples, clusters separate farthest based on substrate material (Figure 2B). Samples extracted using a cellulose or polyester substrate cluster closer than those extracted using a fiberglass substrate. A detailed examination of putatively identified metabolites significantly dysregulated by radiation, revealed that the cellulose substrate provided the best metabolite coverage among the three tested sol–gel FPSE substrates. The cellulose substrate recovered metabolites classified as sugars (2.04%), nucleic acids (6.12%), amino acids (17.35%), general polar molecules (24.49%), and lipids (50%) (Figure 2C). When compared, coverage using a polyester substrate was limited to amino acids (15%), general polar molecules (21.67%), and lipids (63.33%), and when using a fiberglass substrate, it was further limited to only amino acids (15.38%) and lipids (84.62%) (Figure 2C). Putatively identified metabolites were also mapped to their corresponding KEGG pathways (Figure 2C, Supplementary Table 1). The pathways with the most assignments were α-linolenic acid metabolism, taurine and hypotaurine metabolism, biosynthesis of unsaturated fatty acids, linoleic acid metabolism, 2-oxocarboxylic acid metabolism, tricarboxylic acid (TCA) cycle, cysteine and methionine metabolism, and tyrosine metabolism. These pathways are reflective of changes involving energy metabolism, inflammatory signaling, and antioxidative upregulation, which have previously been characterized in the serum of mice, rats, NHPs, and humans after irradiation.⁶

Figure 2.

Descriptive analysis of membranes. (A) Total statistically significant spectral features (p < 0.05, FC ≥ 2) found using cellulose (blue), polyester (aqua), or fiberglass (green) fabric-based membrane with FPSE or with traditional extraction (purple), comparing serum (top) or whole blood (bottom) from sham versus 8 Gy irradiated mice. Mean and SD plotted. Sham: n = 6; 8 Gy: n = 6; repeated n = 2. (B) Unsupervised principal component analysis (PCA) (MetaboAnalyst) of metabolites profiles from whole blood, extracted after 0 or 8 Gy exposure, using a cellulose, polyester, or fiberglass FPSE membrane. (C) Alluvial diagram (RAWgraphics) displays coverage of metabolites which are first grouped by simple classes (lipid, polar, amino acid, nucleic acid, or sugar) and then by the associated KEGG pathways. Pathway matching was performed by Metaboanalyst. Stroke weight represents the number of significant metabolites (p < 0.05, FC ≥ 2) associated with each class and pathway. Tabular data in Supplementary Table 1. Pie graphs below illustrate breakdown of metabolites found in each class (lipids, polar molecules, amino acids, nucleic acids, and sugars) per FPSE membrane used.

Table 1 details some of the highly dysregulated metabolites associated with these pathways and the extraction method with which they were identified. Specifically, a notable increase in pro-inflammatory lipids such as arachidonic acid and linoleic acid were observed paired with a decrease in anti-inflammatory lipids such as α-linolenic acid, docosahexaenoic acid, and palmitoylethanolamide. These lipids are associated with an IR lipid mediator response previously documented in the serum from mice and NHPs.^4,14 Additionally, the antioxidant capacity of the whole blood was altered with IR, as supported by an increase in glutathione disulfide.⁵⁶ Other metabolites documented as biomarkers of radiation injury in mice,⁶ such as carnitine, taurine, valine, hypoxanthine, cystathionine, oleic acid, and phosphatidylcholines, were also recognized to be significantly dysregulated. Together, the data confirms that sol–gel FPSE is comparable to traditional protocols for accurate detection of IR-induced injury markers and can capture these metabolites from a small quantity of whole blood. This is taken with the caveat that the chemical diversity of the captured metabolome is dependent partially on the membrane substrate material, pore size, and other intrinsic physio-chemical properties.

Table 1.

Metabolites Associated with IR-Injury Found with FPSE^a

m/z	compound	p value	mode	substrate	trend
118.00	valine	0.0467	negative	CELL	↓
126.02	taurine	0.0007	positive	CELL/PE	↓
223.07	l-cystathionine	0.0149	positive	CELL/PE	↓
137.05	hypoxanthine	0.0217	positive	CELL/PE	↑
612.83	glutathione disulfide	0.0423	negative	CELL/PE	↑
298.26	palmitoylethanolamide	0.0002	negative	CELL/PE/FG	↓
277.21	α-linolenic acidb	0.0018	negative	CELL/PE/FG	↓
279.23	linolenic acid	0.0021	negative	CELL/PE/FG	↑
281.25	oleic acidb	0.0027	negative	CELL/PE/FG	↓
162.11	l-carnitineb	0.0239	positive	CELL/PE/FG	↓
303.24	arachidonic acidb	0.0372	negative	CELL/PE/FG	↑
327.24	docosahexaenoic acid	<0.0001	negative	CELL/PE/FG	↓
818.60	PC (40:6)	<0.0001	positive	CELL/PE/FG	↓
794.60	PC (40:4)	<0.0001	positive	CELL/PE/FG	↓

^aIdentification are putative unless otherwise noted.

^bMetabolites validated with chemical standards.

Cell = cellulose, PE = polyester, FG = fiberglass.

Validation of FPSE Cellulose Membrane for IR Injury-Assessment Technology

On the basis of its reliable performance and coverage we further explored the role of the sol–gel FPSE cellulose membrane for future IR exposure-assessment technology. The perturbations by 8 Gy irradiation were clearly captured by the cellulose membrane (Figure 3). The separation between the whole blood samples from sham and 8 Gy irradiated mice is evident in the PCA and recapitulated in the heatmap, which displays the increase (red) and decrease (green) in the levels of the statistically significant metabolites (Mann–Whitney U, FDR adjusted, p < 0.05) (Figure 3A). It should be noted that a majority of the displayed significant metabolites are down-regulated in response to 8 Gy. Previously established radiation biomarkers were easily identifiable after FPSE with the cellulose membrane (Figure 3B), including carnitine, α-linolenic acid, and oleic acid, which were decreased after 8 Gy, and arachidonic acid, which increased after 8 Gy. These metabolites were also validated using MS/MS with respective standards (Supplementary Figure S2, Supplementary Table 2). Receiver operating characteristic (ROC) curve analysis was carried out to determine the predictive ability of the aforementioned biomarkers. Generally, a biomarker with an AUC score of >0.8 is considered a robust marker of IR. The AUC for the individual metabolites was 0.86 for carnitine, 0.78 for α-linolenic acid, 0.86 for arachidonic acid, and 0.94 for oleic acid (Figure 3C). The combined panel of all four metabolites produced an AUC of 0.86 (95% CI = 0.5–0.1) and a diagnostic predictive accuracy of 72% (Supplementary Figure S3).

Figure 3.

Overall metabolomic signature of whole blood from mice exposed to 8 Gy of X-ray in comparison to that of control mice as captured by FPSE. (A) Principal component analysis (PCA) plot demonstrates a distinct separation between the metabolomic profiles of irradiated mice (blue triangles) and sham mice (red circles). Heat map in this panel shows the changes in the abundance of individual ions postirradiation. Top half of the figure shows ions with a significant decrease in their blood abundance postirradiation, while bottom half of the heat map depicts those ions with increasing levels after irradiation. (B) Box and whisker plots of individual ions extracted from whole blood using FPSE showing significant response to ionizing radiation. Carnitine, α-linolenic acid, and oleic have decreasing levels after IR exposure, while arachidonic acid increases exposure. Nonparametric Mann–Whitney U statistical hypothesis test used. (C) Individual receiver operator character (ROC) curve analysis (Metaboanalyst) for carnitine, arachidonic acid, α-linolenic acid, and oleic acid with area under the curve (AUC) score and 95% confidence intervals (CI).

FPSE Membrane Temporal Study

Future use of the FPSE method for patient-based blood biodosimetry may require a lag between sample collection and analysis, as samples need to be shipped between sites. Here we perform an initial study of temporal fluctuation of the blood metabolomics when incubated on the sol–gel FPSE cellulose, polyester, or fiberglass membrane at room temperature. Serum samples were first collected from mice and pooled. The pooled samples were divided and incubated on each substrate membrane for 1, 3, 24, 72, or 120 h at room temperature, extracted using the FPSE method, and analyzed using the UPLC-TOFMS protocol described previously. An example total ion current (TIC) chromatogram is shown for serum analyzed after incubation on a cellulose membrane at each respective time point (Figure 4A). The low baseline and well-defined chromatographic regions for each compound class highlights a strength of the FPSE method. The TICs generated after FPSE at room temperature were more defined compared to those from serum samples prepared using a traditional protocol at 4 °C (Supplementary Figure S1), and strong peak signals are retained up to 120 h of incubation. The highlighted time point (24 h) showed the greatest general peak intensity and feature retention when compared to other examined time points. On the basis of this data we deemed 3–24 h of incubation optimum to ensure proper serum adsorption onto the membrane. We also monitored the stability of creatinine (Figure 4B), carnitine (Figure 4C), arachidonic acid (Figure 4D), and α-linolenic acid (Figure 4E) based on normalized peak intensity over time and by substrate type. All three substrates stabilized creatinine, a constant marker found in serum, up to 120 h with minimal loss in intensity (<30%) (Supplementary Table 3). The cellulose membrane also stabilized the IR markers carnitine, arachidonic acid, and α-linolenic acid, keeping the signal intensity loss to <20%, with no apparent bias toward metabolite class. The polyester membrane appeared to retain the lipophilic metabolites, arachidonic acid and α-linolenic acid, better than the polar metabolite carnitine which dropped to 52% peak intensity after 72 h. The fiberglass membrane displayed a uniform drop in peak intensity after 24 h, especially for the polar metabolite carnitine which fell below the level of detection after 72 h. This data indicates that the sol–gel PCL-b-PDMS-b-PCL-coated membranes can maintain the integrity of specific radiation metabolites found altered in blood samples, at room temperature, and with variation based on substrate composition.

Figure 4.

(A) Example chromatogram of extracts from serum analyzed after FPSE using a cellulose membrane. Serum was incubated on the membrane for 1, 3, 24, 72, or 120 h at room temperature. Highlighted time point (24 h) showed greatest peak intensity and feature retention when compared to other examined time points. Relative intensity of (B) creatinine, (C) carnitine, (D) arachidonic acid, and (E) α-linolenic acid in whole blood after incubation for 1, 3, 24, 72, or 120 h at room temperature stays relatively stable. Serum was incubated with either a cellulose (open circle), polyester (open square), or fiberglass (open triangle) membrane. Mean and SD plotted. Biological replicates n = 3 per experimental group. ANOVA–Multiple comparison statistics Supplementary Table 2.

DISCUSSION

In this study we investigated three different sol–gel PCL-b-PDMS-b-PCL-coated FPSE membranes and their ability to capture the metabolomic signature of serum and whole blood from mice exposed to 8 Gy of IR as an introductory test of their practicality in operational scenarios of a radiological incident and medical countermeasures. The study’s novelty resides in the application of FPSE to enrich for metabolites in whole blood at room temperature while conserving the metabolomics signature of the samples. FPSE was first developed for pharmaceutical and toxicological application and was previously applied to targeted analyses of exogenous molecules⁵⁷ including anticancer and inflammatory bowel disease treatment drugs from whole blood, serum, and urine.^42,43 Here we successfully applied FPSE to an untargeted analysis of whole blood and serum and performed an extraction of endogenous metabolites altered in response to radiation.

To test FPSE substrates for the future application to IR-exposure assessment, we focused on the analysis of whole blood following 8 Gy or sham exposures. Whole blood is a complex matrix and as result of its complexity is often used in the form of plasma or serum in most traditional sample preparation techniques. Whole blood separation to serum seems to promote the loss of metabolomic features as made evident by analyte loss when comparing serum to plasma using mass spectrometry,^58,59 and whole blood to serum using NMR.⁶⁰ FPSE is designed to retain analytes through sorbent sol–gel affinity, which allows for utilization of whole blood analysis without preprocessing.^42,43 Additionally, the PCL-b-PDMS-b-PCL sorbent coating is known to reduce platelet binding and decrease adsorption of clotting promoters such as thrombin and fibrinogen.⁵⁷ Here we report that the sol–gel FPSE membranes had an affinity for a wide range of metabolites when incubated with either serum or whole blood, specifically with use of the cellulose membrane, and that metabolite recovery was comparable to our previously documented liquid–liquid metabolomic approach. Most importantly, we confirmed that the sol–gel FPSE membranes could capture the same key IR-specific metabolites using whole blood as compared to serum by verifying metabolites and pathways associated with the biological impacts of radiation.

Cellulose, polyester, and fiberglass had previously been used as substrates with FPSE due to their readily available sol–gel active functional groups,^45,57 but their differing substrate effects had not been detailed with the use of the PCL-b-PDMS-b-PCL sorbent. It is known that certain triblock copolymers such as PCL-b-PDMS-b-PCL have beneficial blood contacting properties, thought to derive from their patterned surface features.⁴⁴ Our comparative analysis of PCL-b-PDMS-b-PCL-coated cellulose, polyester, and fiberglass membranes revealed that the extracted metabolite coverage in both whole blood and serum could also be dictated by the utilized fabric substrate. A number of other factors, such as sample volume, extraction time, back-extraction solvent type, and volume, can impact extraction performance, and we endeavored to keep these factors consistent throughout our experimentation.⁵⁷ As our method relies on capillary action, with no additional forces (such as stirring), it seemed that the substrate hydrophobicity affected sample recovery altering sample permeation. We noted in Figure 1 that as substrate hydrophobicity and material stiffness increased (cellulose < polyester < fiberglass), metabolite recovery and class coverage decreased. Though not proven, the cellulose and polyester membranes generally appeared to retain adhesion to the surface material in a manner greater than the biofluid cohesive forces, making them more wettable to the biofluid than fiberglass.

Of the three substrates examined, the sol–gel PCL-b-PDMS-b-PCL cellulose membrane provided the best metabolite recovery and class coverage. Further exploration proved that the sol–gel PCL-b-PDMS-b-PCL cellulose membrane generated sufficient group separation when comparing whole blood samples from 8 Gy IR versus sham mice. From these samples we were able to accurately identify four dysregulated metabolites using MS/MS and build a sample panel for radiation injury. The biomarkers carnitine, α-linolenic acid, oleic acid, and arachidonic acid serve as good signatures of radiation exposure, and the prediction accuracy for assessment is likely to increase with the addition of more validated biomarkers. A key limitation of other established IR-assessment assays, such as the dicentric chromosome assay,^61–63 cytokinesis-block micronucleus assay,^64,65 and γ-H2AX foci assay,^66–68 is the critical need for blood culture time. Radiation biodosimetry using FPSE was conducted on a generally qualitative level, but in order to construct a signature capable of determining dose range exposure, we eventually will need to compile quantitative biomarker data by which treatment decisions can be made. For now, the FPSE sol–gel PCL-b-PDMS-b-PCL cellulose membrane is a promising platform for future targeted and quantitative analysis of known radiation biomarkers that can be incorporated with other biodosimetry assays to offer accurate radiation exposure assessment.

In the case of a high demand for radiation-exposure analysis, one limitation of current metabolomic and biodosimetry procedures for blood analysis is that they require sample refrigeration. Blood metabolite signals can deteriorate and change over time through oxidation and enzymatic activity, a process intimately connected to temperature. Ideally, our technology would allow for blood collection and shipping at ambient temperature, eliminating the need for maintaining a cold chain transportation and logistics. A trend of substrate-based stability was apparent and seemed to be tied to surface energy (e.g., hydrophobicity). We observed strong signal retention of key metabolites and consistent internal standard recovery when using the cellulose substrate (least hydro-phobic) which outperformed the polyester and fiberglass membrane. Recent work analyzing the metabolomic signature of dried blood spots (DBS) on cellulose paper noted changes to the signal intensity of metabolites that ranged from 20% to 60% depending on the class.⁶⁹ The FPSE sol–gel PCL-b-PDMS-b-PCL cellulose membrane maintained the integrity of IR-induced metabolomic profiles of the samples up to 120 h with changes to the signal intensities only ranging from 20% to 30%.

In conclusion, PCL-b-PDMS-b-PCL-coated FPSE membranes were evaluated as a simpler sample-processing technique for radiation metabolomics. We confirmed that sol–gel PCL-b-PDMS-b-PCL-coated FPSE membranes can be reliably used to capture the IR-exposure metabolomic signature of whole blood comparably to serum. Comparable results using whole blood with FPSE eliminates the need for serum separation and protein precipitation required for traditional LC-MS-based metabolomics. We observed differences in metabolomic recovery and coverage that were coupled to the use of a cellulose, polyester, or fiberglass sol–gel PCL-b-PDMS-b-PCL-coated FPSE membranes and identified cellulose as the membrane most suited for radiation metabolomics. Our metabolite stability measurements indicated that the sol–gel PCL-b-PDMS-b-PCL-coated FPSE cellulose membrane could also stabilize whole blood at room temperature with minimal signal change for up to 120 h. Thus, the FPSE sol–gel PCL-b-PDMS-b-PCL cellulose membrane is an attractive option for high-throughput biodosimetry in a mass radiation exposure scenario of a large radiological or nuclear event.