Improved determination of femtogram-level organic explosives in multiple matrices using dual-sorbent solid phase extraction and liquid chromatography-high resolution accurate using dual-sorbent

17 Identification and trace quantification of multiple explosives residues, their 18 precursors and transformation products in complex samples remains very 19 challenging. For solid phase extraction (SPE) and liquid chromatography-high 20 resolution accurate mass spectrometry-based methods (LC-HRMS), interferences 21 from co-extracted matrix components can significantly affect recovery during 22 extraction and/or detector signal. The aim of this work was to develop a new, 23 improved and more generalisable extraction approach to trace explosives analysis in 24 a range of matrices using dual-sorbent SPE with LC-HRMS. Recoveries of 44 25 organic explosives from model solutions were optimised and compared for seven 26 different sorbents (Oasis HLB, HyperSep Retain PEP and Isolute ENV+, HyperSep 27 SAX, HyperSep NH2, Strata Alumina-N and Bond Elut CN). On average, Oasis HLB 28 and Isolute ENV+ yielded the best recoveries (>80 %). For three sorbents, mean 29 recoveries remained ≤ 1 %, which made them potentially suitable for matrix removal 30 when used in series with more analyte-selective sorbents. To evaluate matrix effects, 31 a range of aqueous (river- and wastewater), solid (soil), dirty (road sign swabs), oily 32 (oven hood swabs) and biological (dried blood) samples were selected based on 33 complexity and forensic relevance. With the exception of river water, matrix effects 34 were lowest using dual-sorbent SPE, with little/no compromise in recovery. 35 Quantitative method performance assessment is presented for selected 36 explosives, representative of different classes, molecular weights and volatilities, and 37 across three different matrices (i.e. untreated wastewater, cooking oil residues and dried blood). Limits of detection improved by ~10-fold over a single sorbent approach, fg sensitivity in many cases. Finally, application of the method to which could be used to help identify clandestine manufacture or sources of 42 environmental toxicity. This approach offered a versatile solution to sample 43 preparation for robust and highly sensitive detection/quantification of large numbers 44 of explosives residues in a range of complex sample types. 45

and Isolute ENV+ yielded the best recoveries (>80 %). For three sorbents, mean 29 recoveries remained ≤1 %, which made them potentially suitable for matrix removal 30 when used in series with more analyte-selective sorbents. To evaluate matrix effects, 31 a range of aqueous (river-and wastewater), solid (soil), dirty (road sign swabs), oily 32 (oven hood swabs) and biological (dried blood) samples were selected based on 33 complexity and forensic relevance. With the exception of river water, matrix effects 34 were lowest using dual-sorbent SPE, with little/no compromise in recovery. 35 Quantitative method performance assessment is presented for 14 selected 36 explosives, representative of different classes, molecular weights and volatilities, and 37 across three different matrices (i.e. untreated wastewater, cooking oil residues and 38 dried blood). Limits of detection improved by ~10-fold over a single sorbent 39 approach, enabling fg sensitivity in many cases. Finally, application of the method to with a high degree of complexity, thereby potentially posing a significant challenge to 169 method performance. All Nalgene bottles used (500 or 250 mL) were first washed 170 with methanol then water in triplicate. All sampling and pre-treatment procedures for 171 each sample type are given in detail in the supplementary information (SI). 172 The chosen sample types were river water, untreated wastewater, soil and M A N U S C R I P T A C C E P T E D Cheshire, UK) cartridges were supplied by the respective manufacturers (see Table  204 S1 of supplementary information for additional details). Based on the conclusion by 205 were evaluated with a standard mix of explosives (50 or 5 mg L -1 ) in ultrapure water, 212 all using a previously optimised SPE method [16] (see SI for full details). This same 213 method was also used for river water and wastewater. 214 For experiments where combined cartridges were used, two sorbents were 215 connected in series with matrix removal sorbents configured first in the line of

Procedures for SPE of swab and soil extracts 226
The 20 mL extracts from matrix-contaminated swabs and soils were further treated 227 according to the standard procedure used by FEL. Isolute ENV+ cartridges (100 mg, 228 6 mL) were conditioned with 1 mL ethanol:water (50:50 v/v), or 1 mL ethanol:water 229 given in Table S2 of the supplementary information. All data was processed using 283 Thermo Xcalibur v 2.0 software. 284 285

Determination of recovery and MS detection matrix effect 286
Analyte recoveries were expressed as the percentage of the ratio of the measured 287 analyte peak area in the extract by the analyte peak area in the corresponding 288 matrix-matched standard at the theoretical 100 % recovery concentration. Although 289 the majority of these compounds were not ionisable in solution, the apparent pH of 290 the mobile phase was maintained at 7.5 in order to leave flexibility for suspect 291 screening of new acidic or basic compounds in future applications as needed. See SI 292 for specific information about how recoveries from each sample type were 293

determined. 294
To determine the MS detection matrix effect, percentage ion 295 suppression/enhancement was calculated using peak areas in matrix-matched 296 standards prepared in reconstituted soil, swab (of dried blood, cooking oil and dirt 297 residue, separately), river-and influent wastewater extracts (n=3) and comparison to 298 Elut CN and Strata Alumina-N) were initially chosen for recovery assessment (Table  331 1). Overall, very good analyte recoveries were obtained for the 44 explosive residues 332 prepared in model solutions using all three analyte-selective extraction sorbents. 333 While no statistically different recoveries were observed between these three 334 cartridges, only Oasis HLB and Isolute ENV+ were chosen for further investigation. whether there was any advantage to a dual sorbent approach. For river water the 359 lowest matrix effect was generally measured using Oasis HLB alone. For 360 wastewater, matrix effects improved using dual-sorbent SPE, and markedly so using 361 the Hypersep-NH2-Oasis HLB combination (Figure 1(a)). This difference in matrix 362 effects across the two aqueous matrices highlighted the need for a versatile clean-up 363 procedure that could be chosen based on sample type. Based on the assessment of 364 matrix effects across all combinations of sorbents, it was decided to prioritise 365 assessment of recovery from river water using Oasis HLB alone and from 366 wastewater using both Oasis HLB and the Hypersep NH2-Oasis HLB combination 367

Topsoil 379
Recoveries and matrix effects for soil are shown in Figures S1 (c) and Figure S2  were observed for cooking oil residues (Figure 1 (b)). Strong suppression for all 403 analytes but EGDN was observed when a single-sorbent was used. This could lead 404 to potential false negatives and/or inaccurate quantification, both of which are 405 undesirable in high sensitivity forensic analysis. A paired t-test, however, showed 406 matrix effects were significantly lower using a Strata Alumina N -Isolute ENV+ 407 combination (p = 0.010, where p is the probability value that the two observations are 408 not significantly different). For dirt (road sign residue) and blood (Figure 1 (c)), 409 matrix effects in general were lower, with the exception of TNB. Strong signal 410 enhancement was observed for this analyte, regardless of SPE combination. 411 Arguably, suppression would be considered a worse scenario from a qualitative 412

perspective. 413
Recoveries, on average, from cooking oil (Figure 3 (a)) and dirt residues were 414 slightly better using a single Isolute ENV+ sorbent. This benefit was, however, offset 415 by a marked reduction in matrix effects using a Strata Alumina-N -Isolute ENV+ 416 combination. The poorest recoveries overall were observed for blood (Figure 3 (b)) 417 and were significantly lower when Isolute ENV+ was used alone compared to any of 418 the dual-SPE approaches (p < 0.05 in all cases), presumably due to sorbent capacity 419 exceedance and/or competitive sorption of matrix. EGDN, ETN and NG were only 420 detected in blood using a dual-SPE approach and recovery was again best for the 421 Strata Alumina-N -Isolute ENV+ combination. HMTD was not observed in blood M A N U S C R I P T

A C C E P T E D ACCEPTED MANUSCRIPT
19 matrices using a dual sorbent combination. Although it was recovered by Isolute 423 ENV+ alone, its very low recovery of just 12% is unreliable and likely due to the fact 424 that HMTD does not form gas phase ions easily, mainly due to its instability [39]. 425

Method performance assessment 442
Analytical method performance with respect to mass accuracy, linearity, range and 443 limit of detection was evaluated for the subset of 14 analytes using the three most 444 challenging matrices tested (i.e. wastewater, cooking oil residue and dried blood). 445 Results are presented in Table 2