Metabolism and Gas Phase Reactions of Peroxide Explosives Using Atmospheric Pressure Ionization Mass Spectrometry

The toxicity or pharmacodynamics of many of the nitrated explosives have been well documented. Trinitrotoluene (TNT) is known to cause liver toxicity while nitrate esters (nitroglycerine) are known vasodilators. One class of explosive that has been on the rise due to the ease of manufacturing from household products are the peroxides. Of particular interest are the cyclic peroxides used for many home-made explosives (HME): triacetone triperoxide (TATP) and hexamethylene diamine triperoxide (HMTD). Very little is known about the toxicity or potentially beneficial effects of these compounds. This may be primarily due to the difficulty in detecting or working with these materials, particularly when they are extracted from living tissues. The use of liquid chromatography (LC) mass spectrometry (MS) is ideally suited to handle this type of sample, provided that the proper detection limits can be achieved. Additionally, this technique provides a very sensitive detection with gentle ionization for more definitive confirmation of the chemical in question over many other techniques historically chosen. In our efforts to reduce the limits of quantification for TATP and HMTD, several remarkable discoveries were made. Most importantly, acetonitrile, one of the most commonly used LC/MS solvents used throughout many industries has shown direct inhibition of ionization. The proposed mechanism of this suppression is by the formation of neutral aggregates of the nitrile moiety with various, common functional groups. Peroxides are one of the most intensely affected moieties. Also, TATP and methyl ethyl ketone peroxide (MEKP) have been shown to react with one or more alcohols under atmospheric pressure ionization (API) conditions to produce new species which may be exploited to improve limits of detection. Caution must be used while working with these products since the conditions can directly affect the signal intensity and multiple related analytes can all provide this common product. Lastly, HMTD has been found to react with both primary and secondary amines and alcohols in the gas phase to produce unique products related to the nature of the amine or alcohol. This research has allowed limits of detection to improve by 20 to 50 times our original analysis limits. Toxicity of HMTD and TATP were primarily in question. However, with the volatility associated with TATP, it seems prudent that this should be the first compound studied since the exposure to this chemical entity is highly probable for any scientist or animal (bomb-sniffing canines) working with it. Simple in vitro analysis using canine liver microsomes (DLM) and lung microsomes (DLgM) in the presence of NADPH (electron donor) were performed to determine the rate, product and nature of the metabolism. Since most of the Phase I metabolism associated with the cytochrome P450 (CYP) enzymes requires molecular oxygen, the incubations are performed in open containers. The exceptional volatility of solid TATP extended to solutions of the compound as well, thus preventing this technique for experimentation. To overcome this issue, oxygen gas was bubbled through the buffered solution used as the matrix for the in vitro studies prior to sealing the containers for the duration of the experiment. Based on this work, several discoveries have been made. The metabolism that does occur appears to be NADPH-dependent, which limits the types of enzymes with may be responsible. The affinity for the non-specific metabolism is very high, with a Km value of 2.21 μM (±14.8%) with a Vmax of 1.13 nmol/min/mg protein (±3.27%). This also indicates that the enzyme responsible for the metabolism is saturated at relatively low concentrations. Work with recombinant isoforms of specific CYP enzymes (rCYP) has shown that only rCYP2B11 has any effect (of the 5 major liver 5CYP’s commercially available) and that this metabolism is enhanced by the presence of cytochrome b5. The metabolism of CYP2B11 does not seem to account for all of the total metabolism of TATP. Only one metabolite has been identified, the mono-oxidation of a single primary methyl carbon (TATP-OH), for TATP. Monitoring the relative amount of this metabolite has been performed. After the rate of metabolism of TATP begins to level, the TATP-OH begins to drop, without detection of a second metabolite. Attempts to trap a second metabolite with semicarbazide (for aldehydes and ketones) or glutathione (for soft electrophiles) did not provide any conclusive products related to the metabolism of these species. With the successful synthesis of TATP-OH, we were able to directly incubate this metabolite. Although it was metabolized more rapidly that TATP, we were unable to detect any metabolites. It was also shown to degrade to acetone in oxidized aqueous buffer, but this did not appear to be related to the metabolism. TATP metabolism was not affected by the presence of TATP-OH or additional undetected metabolite(s). TATP-OH is metabolized only by rCYP2B11, providing evidence that TATP and TATP-OH competitively compete for the same enzyme and TATP dominates this competition. Of particular note is that very little metabolism was observed with the lung microsomes compared to liver. This may have the consequence of significant systemic exposure to those coming into contact with this material.

In our efforts to reduce the limits of quantification for TATP and HMTD, several remarkable discoveries were made. Most importantly, acetonitrile, one of the most commonly used LC/MS solvents used throughout many industries has shown direct inhibition of ionization. The proposed mechanism of this suppression is by the formation of neutral aggregates of the nitrile moiety with various, common functional groups. Peroxides are one of the most intensely affected moieties. Also, TATP and methyl ethyl ketone peroxide (MEKP) have been shown to react with one or more alcohols under atmospheric pressure ionization (API) conditions to produce new species which may be exploited to improve limits of detection. Caution must be used while working with these products since the conditions can directly affect the signal intensity and multiple related analytes can all provide this common product. Lastly, HMTD has been found to react with both primary and secondary amines and alcohols in the gas phase to produce unique products related to the nature of the amine or alcohol. This research has allowed limits of detection to improve by 20 to 50 times our original analysis limits.
Toxicity of HMTD and TATP were primarily in question. However, with the volatility associated with TATP, it seems prudent that this should be the first compound studied since the exposure to this chemical entity is highly probable for any scientist or animal (bomb-sniffing canines) working with it. Simple in vitro analysis To overcome this issue, oxygen gas was bubbled through the buffered solution used as the matrix for the in vitro studies prior to sealing the containers for the duration of the experiment. Based on this work, several discoveries have been made. The metabolism that does occur appears to be NADPH-dependent, which limits the types of enzymes with may be responsible. The affinity for the non-specific metabolism is very high, with a Km value of 2.21 μM (±14.8%) with a Vmax of 1.13 nmol/min/mg protein (±3.27%). This also indicates that the enzyme responsible for the metabolism is saturated at relatively low concentrations. Work with recombinant isoforms of specific CYP enzymes (rCYP) has shown that only rCYP2B11 has any effect (of the 5 major liver 5CYP's commercially available) and that this metabolism is enhanced by the presence of cytochrome b5. The metabolism of CYP2B11 does not seem to account for all of the total metabolism of TATP.
Only one metabolite has been identified, the mono-oxidation of a single primary methyl carbon (TATP-OH), for TATP. Monitoring the relative amount of this metabolite has been performed. After the rate of metabolism of TATP begins to level, the TATP-OH begins to drop, without detection of a second metabolite. Attempts to trap a second metabolite with semicarbazide (for aldehydes and ketones) or glutathione (for soft electrophiles) did not provide any conclusive products related to the metabolism of these species. With the successful synthesis of TATP-OH, we were able to directly incubate this metabolite. Although it was metabolized more rapidly that TATP, we were unable to detect any metabolites. It was also shown to degrade to acetone in oxidized aqueous buffer, but this did not appear to be related to the metabolism. TATP metabolism was not affected by the presence of TATP-OH or additional undetected metabolite(s). TATP-OH is metabolized only by rCYP2B11, providing evidence that TATP and TATP-OH competitively compete for the same enzyme and TATP dominates this competition. Of particular note is that very little metabolism was observed with the lung microsomes compared to liver. This may have the consequence of significant systemic exposure to those coming into contact with this material. xiii                              [1][2][3] or from reaction (either gas or liquid phase) of the analyte with the matrix, [4] these effects compromise the ability to detect the analyte. [1][2][3] Furthermore, these problems are frequently very difficult to recognize; [5] ion suppression may easily be misinterpreted by the absence of an unknown component that significantly enhances ionization. [6] Efforts to minimize these effects have been extensive for electrospray ionization (ESI) and atmospheric pressure chemical ionization (APCI), although APCI typically experiences fewer of these issues. [2,7] Usually co-eluting background interference from matrix components are the most significant problem and frequently can be addressed by changing the liquid chromatography (LC) conditions to separate undetected suppressors from the analyte. [5] Solvent suppression, either from aqueous mobile phase modifiers, pH adjustment, or organic solvent selection is typically identified in the initial analysis for the compound of interest.

LIST OF TABLES
Addition of some degree of organic solvent is known to improve ionization in atmospheric pressure ionization (API) sources by improving the volatility of the solution. The organic modification can disrupt surface tension of droplets and generally assist in the droplet evaporation process. [8][9][10] The two most abundantly used organic solvents in reverse-phase liquid chromatography (RPLC) are methanol (MeOH) and acetonitrile (ACN). [2,8,11,12]

Chemicals and Reagents
Caution: The sensitive organic peroxides mentioned below are powerful explosives.
Take all necessary precautions when working with these compounds. MeOH. An eighth vial was added to include a 2/48/50 solution ratio and 10 μL of standard was added to each vial (neglecting the 1% addition of MeOH). These samples were re-infused and used to develop the FIA system.

Flow Injection Analysis (FIA)
An adequate number of scans (>10) across the peak and minimal mixing with the flow were required for this system. Previous work with the cyclic peroxide, HMTD showed very little response in ESI, so analysis was performed in APCI. Since MeOH showed reactivity toward HMTD in the APCI source, ACN had been chosen as the solvent for subsequent APCI analyses. [4] With this work still being in an area of active investigation in our lab, HMTD was examined in the same fashion as described for TATP above. With APCI or ESI, the HMTD signal was significantly more intense when no ACN was present.
Irrespective of the ion source, HMTD showed as much as 47% signal suppression with as little as 2% ACN present in the solvent (Figure 1-1 and Figure 1-2). With the solubility of HMTD being much greater in ACN, this precludes any notion that the compounds analyzed were simply more soluble in MeOH compared to ACN.
Hexamine, the starting material for HMTD synthesis, was also analyzed by this method with no suppression by ACN observed (Figure 1-2).    To investigate the generality of the ACN ionization suppression effect other ketones were examined. Significant ACN ion suppression was observed for acetone, cyclohexanone, cyclopentanone, and diphenyl isophthalate (Figure 1-2 Table S1-1), which shows a consistent decrease in ion response as ACN concentration is increased, is mirrored over the entire calibration curve dynamic range from 0 to 10% ACN concentration. it may explain the reason TATP was affected (since it is known to sublime). [23] However, this idea falls short when considering HMTD has such a low vapor pressure that it cannot be accurately measured and must be estimated. [24] To explain the source of ion suppression observed for some analytes with ACN, the theory applied in APCI was considered. For the volatilized analyte to be ionized, it must have a higher proton affinity than the reagent molecules. [7] Literature values for the proton affinity (PA) and gas phase basicity (GPB) data for some of the solvents and analytes used are readily available online (Appendix 1: ACN present for that analysis). It may be that the dimer or the ammonium adduct of ACN scavenged the positive charge, reducing the formation of analyte ions. However, this does not explain the reason the ammonium adduct was reduced proportionally to the proton adduct. With the understanding that solvated molecules will increase the proton affinity for the analyte, [12] it may be that the analyte-solvent cluster for these compounds increased the proton affinity for ACN and therefore did not form analyte With PA/GPB failing to fully explain the suppression phenomenon, determining the mechanism of ACN ion suppression was attempted by substituting ACN with pivalonitrile (TMACN), cyanamide or bromoacetonitrile (BrACN). These nitriles were tested against cyclohexanone to determine if they would behave similarly to ACN with regards to ion suppression. Since TMACN and BrACN were immiscible in water, the aqueous portion was replaced with MeOH (also tested against ACN). The electron donating properties of cyanamide were expected to exacerbate the ion suppression, while the electron-withdrawing Br on ACN was expected to improve analyte signal. Both cyanamide and BrACN performed as expected as can be seen in  Since the majority of molecules in this study contained carbonyl or peroxide groups, 1,

Conclusions
With little success we attempted to correlate the ACN suppression effect to ion size and shape, functionality, volatility, gas phase energy and solvation. This has been rigorously tested in multiple mass spectrometers with different ionization sources.
Currently accepted mechanisms for ion formation fail to fully explain the phenomenon. Although the mechanism is still unclear, we have tentatively proposed a polarity aggregation model involving nitriles and carbonyls, peroxides or other polar molecules that may inhibit ionization. An important objective to this work is alerting the LC/MS community to the significant ion suppression that may be caused by the presence of ACN. Chemical analysis/trace detection of peroxides, ketones, and related compounds would be particularly impacted fields. With attention being on rapid detection and analysis, traditional analytical tools, infrared [4], Raman [4,5], and xray [5] have been applied. However, screening usually employs ion mobility mass spectrometry (IMS) [6][7][8]. Spectroscopy, which offers no possibility of separation from interferences, has reported limits of detection (LOD) ranging between 1[9] and 5 ppm [10] in standoff mode. IMS, which has some ability to separate interferences, has a reported LOD of 23.3 ng for TATP and 0.2 ng for HMTD. [8] For unequivocal identification and quantification some type of separation is essential prior to detection.

Figure 2-1. Structures of peroxides analyzed.
Volatile organic compounds have traditionally relied on separation by gas chromatography (GC) coupled to either a mass spectrometer (MS) [11][12][13][14][15] or electron capture detector (ECD). [13,16]. In one of the earliest reports of TATP detection in a criminal case study, both GC/MS electron impact (EI) and chemicals ionization (CI) techniques were used. [17] Since that time, the number of GC/MS applications for TATP and HMTD have grown exponentially; today it is one of the prominent techniques for their detection. The reported LODs for TATP in a condensed phase range between 0.05 and 2 ng, [18] depending on the mode of ionization and type of mass spectrometer used; even lower LODs (<0.1 ng) are recorded for headspace analysis. [19] Low nanograms levels were reported by DART™-time-of-flight-MS [20] for HMTD analysis. The major drawback using GC is the potential for thermal degradation of explosives in the inlet or ion source. For this reason, liquid chromatography mass spectrometry (LC/MS) is becoming a predominant technique for unequivocal structural elucidation and quantification of most organic molecules.
The benefits over GC include room temperature sample introduction, availability of soft ionization techniques, and high resolution accurate mass capability. [21,22] Selected LC methods with monitored ions and LODs are presented in Table 2-1. 29] exact mass MS shows this fragment contains four carbons, which cannot readily be explained from the structure of TATP (Figure 2-1). Our work investigates the origin of that fragment and addresses chromatographic and mass spectrometric parameters (e.g. solvents, [35] temperatures, gas flows and voltage differentials) that can affect ion production.

Chemicals and Reagents
Caution: The sensitive organic peroxides mentioned below are powerful explosives.
Take all necessary precautions when working with these compounds.

TATP Analysis
The mass spectrometry gas flows and temperature were originally optimized using a and 95% aqueous 10 mM NH4OAc (channel B) for introduction onto a Thermo Syncronis C18 column (2.1 x 50 mm, 5 µm). Initial conditions were held for 1.5 minute before a linear ramp to 35% A/65% B over 1.5 minutes followed immediately by a linear ramp to 95%A/5% B over the next minute. This concentration was held for 2 minutes before a 30 second transition to initial conditions with a hold of 1.5 minutes.
As an internal standard (IS), d18-TATP at 10 μg/mL (41.7 μM) in ACN was added 1:1 to aqueous TATP samples with a final concentration of 5000 ng/mL (20.8 uM). XIC were integrated using the Genesis peak detection algorithm in Thermo Xcalibur Quan Browser. Linear dynamic range comparing concentration to peak area response ratio, relative to the IS, extended from 25 ng/mL (112.6 nM) to 20000 ng/mL (90.1 μM) using 10 points and 1/x weighting of the calibration curve. Identical procedures were followed for the calibration curve of DADP (discussed later). Stability determination for TATP did not use an IS and calibration was determined by peak area response vs.
concentration (external calibration). Linear range and curve conditions were the same as above. All dilutions were made in 50/50 ACN/water. Stability was determined by comparing quality control (QC) samples made on day 1 to freshly prepared standards made on the day of stability determination.

TATP Volatility
Volatility of TATP was interrogated by 2 methods. This was repeated 3 times for each vessel before sample solutions were injected onto the HPLC/MS system described above (without IS). Samples were injected in duplicate; average TATP vapor concentration is reported.

MEKP Analysis
With the MEKP's lacking any true "standard", stability determination or quantitative analysis of any specific one of these was not possible. Purification of these compounds, particularly the cyclic trimer (MEKP C3), was attempted using a

Isotope Incorporation Studies
To examine the origin of certain products/fragments observed in the LC/MS experiments, isotope incorporation studies were performed as follows.
Hydrogen/deuterium exchanged (HDX) began with concentrated (20 µg/mL), 0.5 mL samples of TATP and d18-TATP prepared in deuterium labeled methanol/water (CD3OD/D2O) and unlabeled (MeOH/H2O) solvents, respectively. An ammonium source was provided by the addition of 5 µL of 500 mM NH4OAC. Solutions were individually infused at 20 µL/min. Two samples containing an estimated 30 μg/mL of total MEKP and 20 μg/mL of both TATP and d18-TATP were produced from highly concentrated standards prepared in MeOH. These samples were briefly placed under a light stream of N2 gas to evaporate the solvents but prevent significant evaporation of TATP. One sample was reconstituted in 100 μL of MeOH and 20 μL of water before infusion onto the optimized APCI-MS conditions for in-source fragment production (discussed later). Once this sample was successfully observed, the second sample was reconstituted in 100 μL of Me 18 OH and 20 μL of water and infused. Conditions were held for 1 minute followed by a linear ramp to 90% PrOH/20% B over 8 minutes. This was held for 1 minute before ramping to initial conditions over 30 seconds and holding for 2 minutes.

Alcohol Incorporation and Infusion Experiments
Infusion of TATP into the APCI source was performed by two methods. To generally optimize MS voltages, TATP (20 µg/mL in 90/10 MeOH/10 mM NH4OAc) was directly infused onto the ACPI source at 20 μL/min. We termed this "direct infusion." To assess the effects of temperature and gas flow, TATP (20 µg/mL in 90/10 MeOH/10 mM NH4OAc) was infused at 20 µL/min into a 230 μL/min flow of 95% MeOH/5% 10 mM NH4OAc (total flow was 250 µL/min, the environment of TATP eluting from a C18 column). This we termed "MP infusion." An additional MP infusion study was performed but using the TATP gradient program (described above

Results and Discussion
Stability and Volatility of Analytes Many unexpected challenges and unusual findings were encountered during the development of LC/MS analysis methods for the peroxide explosives TATP, DADP, HMTD, TMDDD and MEKP. Samples can be prepared and stored in ACN without issue as long as MeOH is used as the organic mobile phase modifier for reverse phase LC/MS and the compound is not eluting in the void volume. The peroxides described in this work are generally well-retained and free of can, which is washed away in the sample plug. However, if ACN is present in the source using either ESI or APCI, the signal will be significantly reduced or almost completely eradicated. [35] Individually synthesized products of TATP, HMTD and TMDDD were treated separately for stability analysis. While attempts were made to purify HMTD, it always contained a small amount of TMDDD and vice versa. Solutions of each peroxide were prepared and stored in acetonitrile. Autosampler stability (for the 10000 and 5000 ng/mL samples), is presented in the Online Resource for the first reanalysis (day 7) since concentrations were changing rapidly once the container seal was compromised.
Samples were run at N=2 (RSD <5%). Calibration curve values were within ±15% of nominal concentrations with R 2 values > 0.98 for 1/x or 1/x 2 weighted curves. TATP showed ~9% recovery after 7 days storage in the refrigerated autosampler (in a 96well plate, Appendix 2: Table S2-1). This loss of TATP is attributed to its high volatility and the fact that the 96-well mat was no longer sealed; subsequent data supports this conclusion.
The peak shape of HMTD on a C18 column was strongly dependent on the organic concentration in the sample plug. As the organic concentration increased, the peak fronting became severe, with optimal peak shape occurring at low organic content. By placing the samples in 50/50 ACN/water, peak shape and limit of detection for HMTD was compromised to keep sample processing consistent with the previously developed method for TATP analysis. Despite this compromise, HMTD curve and QC data was within acceptable criteria of ±15% accuracy. As ACN evaporated from the 96-well plate, concentration of HMTD and TMDDD increased significantly. Evidence of ACN evaporation was observed by the greatly improved peak shape of the HMTD sample after sitting in the autosampler for 7 days (Appendix2 : Table S2-1, Figure S2-1). In fact, due to the change in concentration of HMTD or TATP over a relatively short period of time, separate curves using the same vial or well-plate position could not be used to bookend analyses greater than 3 or 4 hours apart.
TATP concentration was relatively unchanged over 60 days under conditions where the storage vessel was airtight. HMTD degraded ~15% in 40 days at room temperature and about 7% in the refrigerator or freezer in 60 days (Appendix2: with the molecular formula C4H3D6O2 + . We speculated that the source of the nondeuterated methyl group was derived from the addition of the solvent methanol into TATP, as we had observed for HMTD [38] and proposed by Rondeau, et al for dialkyl mono-peroxides. [14] In order to determine the extent of methanol dependence for the m/z 89 signal, TATP was infused post-column into the normal LC gradient used for TATP analysis. Initially, m/z 240 increased with increasing methanol (as would be expected with increased organic modifier), but that signal quickly leveled off and began to diminish while m/z 89 continued to increase. To clearly illustrate the effect of methanol on the m/z 89 fragment, the ratio of m/z 89 to m/z 240 was plotted against time and compared to the methanol concentration of the gradient (Figure 2-3). There was a significant increase in m/z 89 when increasing from 5% to 95% methanol in the gradient. This accounted for some of the previously unexplained variations between the levels of m/z 89 and m/z 240 we had experienced.     Table 2-2), respectively. At a vaporizer temperature of 250 ºC, the signal for all ions seemed to be optimized, with 3 being the most intense ion followed closely by 1. As gas flow from the auxiliary/sheath gas was increased, 1 began to dominate over 3 with little change in either parent or 4 (which was only marginally detected). However, when the LC flow was removed and the same d18-TATP solution was directly infused at 20 uL/min into the high gas flow at 250 ºC, product 4 became the most intense ion with nearly the same intensity as the parent. As gas flow was pushed even higher, parent and product 4 were nearly all that could be seen, with no 3 present at all. As gas flow was dropped to minimal values (sheath 8 AU and aux 5 AU), product 1 dominated the spectrum, and 3 became slightly more intense than 4, which had dropped significantly. This suggests that at higher gas flows, ions are pushed into the MS more rapidly with either less time in the corona region or less time exposed to the vaporizer temperature to react with the solvent to form alcohol-incorporated products. Additionally, when infused into the mobile phase (vs. direct infusion) which contains a significantly higher population of MeOH ions/molecules, far more of the alcohol incorporated products are observed (Figure 2-6). To assure that the incorporation of the alcohol is completed, a study with the infusion of TATP, d18-TATP and MEKP in Me 18 OH was performed. Complete incorporation of both the carbon and the oxygen from MeOH into each of the proposed products was observed (Appendix2: Figure S2-2). For this analysis, source conditions that favored the formation of the alcohol incorporated products were used. Due to the cost of the 18 O solvents, we were unable to attempt this at mobile phase levels; therefore, doubly incorporated products at m/z 93.0693and 96.0881 were not as significant as shown in

Scheme 2-1
Scheme 2-2 shows the proposed mechanism for the addition of the second alcohol from the product of Scheme 2-1. With the abundance of this fragment (depending on conditions used) it may transform by a more concerted mechanism than proposed. It should be noted that as the alcohol chain length increased, the addition of two alcohols seemed to became more significant that the addition of one alcohol. Also, formation of the tert-BuOH product for either 1 or 2 additions of alcohol was nearly non-existent. This supports the proposed mechanisms since steric interactions would prevent this reaction.

Scheme 2-2
Cyclic peroxides appear to have several analytical nuances that separate them from their linear counterparts. The structure of TATP has, of course, been confirmed by Xray diffraction and other spectroscopic techniques with subsequent DFT calculations to corroborate this data. [39] , [40] , [5] However, there has been no definitive identification of the cyclic MEKP species. MEKP product 9 (m/z 88.0519) in Table   2-2 was only produced by the presumed cyclic MEKP C3 trimer. Because TATP also forms this analog (product 6, and is most likely cyclic in nature.

Attempts to Enhance Signal Intensity
Although the sodium adduct of TATP has been used by Desorption ESI (DESI) [26], Extractive ESI (EESI) [24] and LC-ESI [23] to produce abundant ions at m/z 245.0996, our attempts at adding controlled amounts of very low concentrations of sodium to the mobile phase for quantitative analysis ended with plugged electrospray capillaries.
However, even using sodium we have been unable to approach the level of quantification provided by APCI for TATP (currently, 1 ng on column for m/z 240.1442 and 200 pg on column for m/z 89.0597). TATP (and all the peroxides associated with MEKP) have historically been observed only as ammonium or sodium adducts in our lab. While in ESI, the TATP sodium adduct is rather intense, the addition of lithium and potassium did not produce a significant signal compared to either sodium or ammonium, suggesting the size of these ions is optimal for gas phase adduction.
As mentioned above, variation in gas flow affected the intensity and abundance of each ion associated with TATP.

Conclusions
Two cautionary notes come from this research. First, while researchers have long been aware that TATP has a rather high vapor pressure for a solid, the fact that it readily volatilizes from solution has not been fully appreciated. Second, while lowlevels of TATP may be quantified by LC/MS using the molecular fragment m/z 89, it must be recognized that this fragment has conditions. It represents two different species which are both dependent on the MeOH concentration, mobile phase modifiers, temperature, gas flow, and flow rate (among the parameters tested). Since the two most common LC/MS solvents for reverse phase chromatography are methanol and acetonitrile, the analyst is faced with a dilemma. If acetonitrile is used as the mobile phase, ionization is suppressed below reasonable levels of analytical detection. [35] If methanol is used, the compound will invariable react with the species in the gas phase. The gas-phase alcohol attack of peroxides (TATP, DADP, MEKP, and HMTD) at the α-carbon is apparently a general phenomenon. [14] , [38] This phenomenon can be exploited to lower the limits of detection for these compounds.
However, understanding the origins of a particular fragment is very important, and all variables must be considered prior to using these ions for quantification. Proper separation must be achieved to prevent unwanted materials (many small compounds may have a mass associated with C4H9O2 + ) from providing a false positive response.

Introduction
Hexamethylene triperoxide diamine (HMTD) is a sensitive peroxide explosive that is relatively easy to synthesize from hexamethylenetetramine (hexamine), hydrogen peroxide and catalytic levels of citric acid. Although it has never found use as a military explosives due to poor thermal stability and high sensitivity to impact, friction and electrostatic charge, it has become more commonly used by terrorist. [2] [3] [4] [5] Our efforts to successfully prevent the use or production of HMTD by terrorists require fundamental understanding of mechanistic principles associated with its formation and decomposition.
First synthesized in 1885 by Legler, [1] the structure was not proposed until 1967 by Urbanski [6] and not confirmed until 1985 by Schaefer et al. [7] using X-ray crystallography. Its structure is unusual in that there is a planar 3-fold coordination about the two bridgehead nitrogen atoms rather than a pyramidal structure. [7] Ring strain in HMTD may account for the stability and sensitivity issues mentioned above.
Despite a plethora of information on HMTD, a mechanism for formation of this compound has only recently been tentatively proposed. [8] Development of an analytical method for HMTD was investigated to identify potential, non-volatile decomposition products by liquid chromatography interfaced During solution optimization it is important to take future chromatography conditions into consideration. Though previous separation work reported for HMTD used methanol and water, [9] we preferred to perform initial testing using the aprotic organic solvent, acetonitrile. When HMTD was later infused into the mass spectrometer in a methanol/water solution, the spectrum suggested a gas-phase chemical reaction occurred between a methylene carbon of HMTD and the alcohol. The purpose of this work is to help describe the behavior of HMTD in the gas phase. This information may aid present efforts to elucidate formation and destruction mechanisms of this molecule. Additionally, the ability of HMTD to react with alcohols under chemical ionization conditions may prove useful to other fields of research.

Chemicals and Reagents
Water HMTD was produced in house by standard methods reported in previous work. 10 Methanol labelled with 18  Advantage PFP column (100 x 2.1 mm, 5 μm). In order to gain some retention of hexamine, neutral pH conditions were preferable, but this caused broadening of the HMTD peak shape. To remedy this problem, 3 different mobile phase solvents were used to provide both pH and solvent strength gradients. Initially, 95% solvent A (10 mM ammonium acetate, pH 6.8) and 5% solvent C (acetonitrile) were held for 3 minutes following injection to retain hexamine. The system was then rapidly ramped to 85% solvent B (0.1% acetic acid), 5% solvent A and 10 % solvent C over the next 3 minutes. Organic levels increased slowly for 9 minutes to 35% C, 60% B and 5% A, then rapidly for 3 minutes to 90% C and 5% of both A and B. This was held for 2 minutes before returning to initial conditions and re-equilibrated for 5 minutes prior to the next injection. Data collection and analysis was performed with Thermo Xcalibur software version 2.2, SP 1.48.

Results and Discussion
Infusion acetic acid (pH ~ 3.2) or 10 mM ammonium acetate (pH ~ 6.8). Due to the lack of retention or reasonable peak shape, even under highly aqueous conditions, it was decided to switch to methanol (MeOH) as the organic phase. This was consistent with literature methods described by Crowson [9] who also used APCI positive ion mode conditions, but with an isocratic method using 5% MeOH. Although retention was improved by this alteration, peak shape was inconsistent and unacceptably broad.
Additionally, a new peak at m/z 207 was observed in the spectrum obtained for HMTD (also reported by Crowson). [9] To assure that this was not an impurity from MeOH, additional infusion experiments were performed using various solutions. When 100 %

MeOH was infused, presence of impurities was ruled out. Infusion of HMTD in 100%
MeOH provided ion signals of both m/z 209 and 207 in roughly equal abundances.
Using an aqueous solution with 10% MeOH showed only a small amount of m/z 207 (roughly 10% relative abundance, Figure 3-2). Initially, the assignment of m/z 207 was thought to result from protonated HMTD losing two hydrogen atoms (H2 gas) in the gas phase, as previously reported. [9] [11] [12] Crowson [9] , using a nominal mass, quadrupole instrument, attributed the mass at m/z 207 as a fragment of HMTD, but did not specifically designate the fragment structure. In 2004, Xu [11] using a ThermoFinnigan (San Jose, CA, USA) TSQ7000 triple quadrupole mass spectrometer  (Figure 3-3).
Cotte-Rodrıguez, [13] used a nominal mass Thermo Electron LTQ ion trap instrument affixed with a variation of the Desorption Electrospray Ionization (DESI) source called DAPCI, which provides APCI-like results. This direct analysis technique combined with alkali metal (sodium or potassium) doped solvents was able to detect the sodium adduct of the stable HMTD-methanol product [M+CH3OH+Na] + at m/z 263.. Since the fragmentation pathway was inconsistent with a normal solvent adduct, they proposed a mechanism in which one peroxide bond of HMTD reacts with MeOH by a homolytic mechanism consistent with peroxide reactions, forming a methyl ether with the loss of water and formaldehyde. [13] The Cotte-Rodrıguez mechanism [13] involves cleavage of the methanol oxygen and subsequent loss of that oxygen as water. If this mechanism is correct, then the oxygen from the alcohol would be lost as water rather than being incorporated into the HMTD molecule upon gas phase ionization. Therefore, we performed the experiment by initiation as opposed to only three intermediates by a homolytic reaction mechanism of the peroxide. [13] Compared to ESI, the protonated molecule of HMTD was produced in far greater abundance using APCI, but it was still detected by ESI. To determine whether the

Conclusions
Despite the considerable body of work performed on HMTD over the years, it still possesses many secrets. The work presented here shows that a gas phase chemical reaction occurs with HMTD in the presence of alcohols to produce a hemiaminal ether under APCI conditions. We are hoping that this unusual behavior may be exploited to provide insight into the formation and degradation mechanism(s) of HMTD, neat, in solution and in gas phase. The idea that the methylene groups of HMTD may be more reactive than the peroxides is an interesting prospect when considering the behavior of this molecule. An added benefit of this study is that it provides a method for quick characterization of various alcohols in solution; a property of HMTD that may find use in other fields of science, possibly as a probe substrate. Research efforts into HMTD mechanisms are ongoing in our lab.

Figure 3-8. HMTD infused with alcohol mixtures of methanol, ethanol, isopropanol and A) n-butanol, B) isobutanol and C) tert-butanol. 2-butanol was not included due to large impurities found in this alcohol, but the trend for HMTD adducts was similar
to N-butanol and isobutanol.

Chapter 4 : Using Gas Phase Reactions of Hexamethylene Triperoxide Diamine (HMTD) to Improve Detection in Mass Spectrometry
Status
Combined with its ease of formation, the native sensitivity of this peroxide has facilitated use in illicit explosive devices.

Solvent Incorporation Studies
To examine the origin of certain products/fragments observed in the LC/MS experiments, HMTD or TMDDD (10 μg/mL) were prepared in solutions containing 1 mM of various amine compounds (see Table 4

Results and Discussion
Early efforts to identify all species related to HMTD in the APCI source led to the frequently encountered m/z 224.0877, associated with C6H14N3O6 + . When chromatographically separated, two peaks with this same m/z 224 were observed. The first peak eluted early with a major signal of m/z 224 and a minor signal of m/z 207.0611 [HMTD-2H+H] + . We believed this compound to be TMDDD (Figure 4-1), with m/z 224 being the ammonium adduct. The second peak eluted at the same retention time as HMTD and exhibited all other masses associated with HMTD ionization. Although this second peak showing m/z 224 was produced in varying degrees from one analysis to another, we believed this to be TMDDD formed from HMTD in the gas phase under APCI conditions. Marr and Groves had reported that, in the gas phase, a small amount of HMTD is converted to the dialdehyde product (TMDDD). [10] When an authentic sample of TMDDD was prepared and analyzed, it eluted with the same retention time (tR) and peak shape as the early eluting m/z 224, indicating that our HMTD standard was contaminated with a small amount of

TMDDD. Quantification of the TMDDD and HMTD samples showed that HMTD
was contaminated with about 1% TMDDD and TMDDD contained about 1.5% of HMTD (Appendix 4: TMDDD was found to have a significantly better signal in ESI than APCI and a very high affinity for ammonium or amine adducts. In fact, using our standard mobile phase of ammonium acetate/methanol, the minor contamination of TMDDD in HMTD (~1% depending on the batch) produced a signal for the ammoniated TMDDD in ESI that was nearly comparable to the HMTD signal in APCI. Krawczyk discovered that TMDDD also has a high affinity for metal ions in the ESI source and suggested that purposeful oxidation of HMTD be exploited to improve detection levels. This process involved off-line oxidation of the HMTD samples prior to analysis by ESI in the presence of metal ions. [11] In contrast, the APCI source produced a significantly better signal for HMTD than ESI with variable amounts of conversion from HMTD to TMDDD. We then began to consider ways of intentionally increasing this in-source conversion to improve detection limits with the addition of an adducting agent. Since the production of TMDDD from HMTD required heat in the presence of O2, several experimental parameters were examined in both APCI and heated ESI (HESI) to cause in-source conversion. Attempts to convert HMTD to TMDDD using the HESI source did not succeed even at temperatures exceeding 300°C. Experiments using APCI show that TMDDD formation increases with increasing temperatures, up to the point (~350°C) where both HMTD and TMDDD begin to decompose.   (Figure 4-3). It should be noted that each source produced both the low and high m/z ions to some degree, but the lower m/z, e.g. 238 or 252, clearly dominated in ESI, and the higher m/z, in APCI (Figure 4-3). To clarify these results, the same HMTD sample was injected onto the LC-MS system with post-column addition of the same amines using both ESI and APCI (Figure 4-4). In ESI, the intense signal at m/z 252.1190 eluting early with no detectable m/z 254.1347 was the TMDDD contaminant in our HMTD sample. The second peak being HMTD had almost no m/z 252, but a reasonable signal for m/z 254. Results using APCI showed very low intensity for the TMDDD m/z 252 signal (no m/z 254 at all) and a very intense signal of m/z 254 for HMTD with a small amount of m/z 252 present (from TMDDD formed in-source). These results confirm our observations above that TMDDD contaminates HMTD samples. Even HMTD, chromatographically separated from TMDDD, forms TMDDD in the APCI source. Furthermore, it was observed that HMTD produced the strongest signal in APCI, while TMDDD was best observed by ESI. Additionally, both HMTD and TMDDD form products or adducts with amines.
At this point, it was not certain whether the amine adduct could be used to improve the APCI signal for TMDDD.  Attempts to trap and dissociate TMDDD-amine adduct peaks in ESI provided no fragmentation at all, just depletion of the parent ion. Krawczyk also reported that attempting to fragment metal adducts of TMDDD produced no observable fragments. [11] When larger amines (see list in Table 4-1) were studied, where the protonated amine had a m/z greater than the 50 Da cut-off of the Orbitrap, we observed only the fragment corresponding to the protonated amine. This suggests that the amine sequestered all charge and explains the reason no fragments were observed during CID of the TMDDD adduct. It appears that the affinity of TMDDD for amines is related to the basicity of the amine, with the more basic amine producing a larger adduct signal. Post-column addition of organic amines indicated that the formation of organic amine adducts with TMDDD are favored over ammonium. Unfortunately, the signal observed for the amine was equivalent in intensity to the ammonium adduct, but not significantly better. Unlike TMDDD, trapping and dissociation of the HMTD amine adducts in APCI produced multiple stable fragments (see Figure 4-5, Table 4-2). CID spectra and proposed fragments for many of the amines tested can be viewed in the Appendix 4: Figures S41-46. Interpretation of the spectra for each of the amines showed that although each exhibited a fragment of 209.0768 (product 4, Table 4-2) suggesting adduct formation, they all lost water (product 2, Table 4-2) and H2O2 (product 3, Table 4-2). Notably, in all spectra was the formation of the fragment m/z 197.0768 (product 5, loss of exactly 12.000 Da from HMTD) and a fragment corresponding to the parent amine increasing by exactly 1 carbon (product 10, Table 4-2). This suggested that the amine performed a nucleophilic attack on one of the methylene groups of the HMTD molecule, similar to the reaction with alcohols. [1] To confirm this result d12-HMTD was synthesized and full hydrogen/deuterium exchange (HDX) studies were performed using isopropylamine. Results are shown in Table 4

Scheme 4-1 (arrow colors correspond to formed structure color)
To improve detection of HMTD, post-column addition of organic amines (with or without neutralization) was examined with increasing temperature in the APCI source.
This attempt at in-source conversion of HMTD to TMDDD was successful, but the signal for the TMDDD-amine adduct was not greater than HMTD-methanol product (m/z 207.0975). In another attempt to improve the HMTD signal, we attempted to attach a charged quaternary amine, producing a permanently charged ion. Two organic quaternary amines, one with a pendent primary amine ((2-aminoethyl) trimethylammonium) and the other with a pendent primary alcohol (choline), were infused in a MeOH/H2O solution of HMTD. No reaction or reaction product (including multiply charged products) was observed in either ESI or APCI. We speculate that the quaternary amine pulled the electron density from the alcohol or amine tail which created a species that was not a strong enough electrophile to attack the methylene group. Additionally, the electron rich oxygens and nitrogen surrounding the methyl groups can draw the quaternary group toward molecule thus preventing the nucleophilic group from proper approach for reactivity. Similar experiments using amines were attempted with other peroxides including TATP and MEKP. No reaction products were detected in either ESI or APCI for these compounds. [7] Using the ammonium acetate/methanol mobile phase, chromatographic data frequently  Optimizing gas flow was performed by injecting the same 10 µg/mL sample onto the PFP column at 250°C vaporizer temperature under various flow conditions (Table S-3, Online Resource). Although a higher sheath than auxiliary gas provided more intense signal, it was also associated with the most variability. The best stability with the most intense signal for the [M+MeOH2-H2O2] + product was achieved when the sheath and auxiliary gasses were both set to 15 AU. Additionally, HMTD peak shape is strongly affected by the amount of organic in the sample plug. More than 20% ACN in the sample plug can produce sever fronting of the peak, reducing our ability to detect low levels of HMTD (see Figure 4-7). Since HMTD is not volatile and we are using a deuterated IS, samples can be evaporated to dryness and reconstituted in 90% water/10% ACN. Although concerns about solubility did arise, the assay linear range spanned from 10 ng/mL (48nM) to 20000 ng/mL (96 μM) demonstrating that this was not an issue. This also allows for the concentration of higher volume samples to push detection limits even lower. Using optimized conditions for HMTD, we have detected HMTD as low as 100 pg on column with a robust analysis of 300 pg on column.

Conclusions
Although TATP and HMTD have been observed to undergo a gas-phase reaction with alcohols[7] [1], no corresponding reaction has been observed for TMDDD. While enhancing detection of HMTD we discovered that TMDDD is frequently a contaminant in purified HMTD and can also be formed in the gas phase within the APCI source. TMDDD [M+H] + (m/z 207.0612) produces a much better signal in ESI than APCI, and has tremendous affinity for ammonium or organic amine ions. HMTD and cyclic peroxides have exhibited significantly better ionization by APCI, while linear peroxides (MEKP) and TMDDD respond best to ESI. [7] We speculate that the open nature of TMDDD (a 10-membered ring) may allow this to behave more like a linear peroxide. Our attempts to create TMDDD in-situ were a success, in that HMTD is oxidized in a temperature-dependent fashion within the APCI source. Unfortunately, we were unable to exploit the high TMDDD affinity for amines to improve HMTD detection limits. Off-line conversion of HMTD to TMDDD as suggested by nmol/min/mg of protein based on substrate depletion. Only one metabolite, hydroxy-TATP (TATP-OH), was identified. Canine CYP2B11 was the only enzyme specifically determined to catalyze metabolism, but the degree to which it metabolized TATP was insufficient to account for observed DLM metabolism, suggesting more than one enzyme may be functioning. This metabolite disappears over extended incubation times, but no other metabolites were detected. Trapping of either hard or soft electrophilic products was unsuccessful. Similar work performed on MEKP indicated significant metabolism of the hydroperoxides and rapid oxidation of reduced glutathione (GSH). This suggests that TATP does not metabolically form any ringopened or hydroperoxide product(s). The hydroxy metabolite was synthesized and tested for stability in DLM. At 10 μM concentration, TATP-OH metabolism progressed 3x more rapidly than TATP with no metabolites found or trapped.
Chemical degradation of the metabolite, with acetone appearing to be trapped as a byproduct, proceeded in oxygenated pH 7.4 buffer much slower than did its metabolism. The metabolite (TATP-OH) was stable in all organic solvents tested. Data suggests that TATP and TATP-OH may be competitively competing for the same enzyme with TATP dominating this competition. The formation of a second metabolite(s) that is either undetectable by MS/UV or covalently bound to a protein or polymer in the incubation reaction is very probable. The fate of this second metabolite(s) does not appear to affect TATP metabolism, but is currently still a mystery.

Introduction:
Every drug that is currently on the market has been thoroughly tested in multiple pre-clinical species to predict human exposure, metabolism and toxicity.
While it is a common and necessary process to determine the toxicity of many poisons, toxins and toxicants by animal testing [1], alternative in vitro methods are constantly being developed or improved to reduce the need for this practice. Over several decades, a significant body of work correlating in vitro testing to in vivo results has been published. [5][6] These models are intended to predict what a chemical will do in humans based on the in vitro/in vivo correlation (IVIVC) with preclinical species. Some of these techniques can also be applied to the determination of human or animal hazards arising from environmental and/or occupational exposure to compounds with unknown biological effects. Determination of toxicity from continual exposure to new or different chemical entities is of great concern to the general public. This area of research has implications for both acute, short-term problems as well as long-term, genetic or epigenetic effects. [7] [8] For employees or people living in areas where environmental exposure to explosives may be unavoidable, knowledge of potential toxicity is essential. While some of the more common, older explosives such as trinitrotoluene (TNT) have been fully investigated for metabolism and subsequently found to have toxic metabolites [9][10], many of the newer or peroxide-based explosives have never been tested for toxicity. The peroxide-based explosives are easy to produce, sensitive to initiation, and relatively unstable; therefore, they are not used or produced by the military.
However, with the ease of production and power of these explosives, they have become very appealing to those wishing to inflict damage and destruction. [11][12] [13] Therefore, research into the formation and safe destruction of these compounds as well as applications for their trace detection must continue. Currently, canines are being trained to detect trace levels of peroxide explosives [14], (triacetone triperoxide, TATP and hexamethylene triperoxide diamine, HMTD) to mitigate terrorist risk in airports, train stations, etc. With that in mind, there may exist some significant risk of toxicity to both humans and canines from exposure to these compounds. No information on the metabolism or potential toxicity of these easy to produce homemade explosives (HME) currently exists.
Hydrogen peroxide is a reactive oxygen species (ROS) that is endogenously produced through many sources including, mitochondrial respiration [15], superoxide dismutase activity [16], and metabolism by P450 [17] or other oxidase enzymes. [18] While ROS are generally thought to be responsible for cellular damage, H2O2 is necessary for the redox regulation of many physiological processes. [15] The catabolism of this by catalase and enzymes like glutathione peroxidase are well studied. [19][20] [21] There is also a fair amount of work done on the metabolic fate of hydroperoxides [22] [23], which, in the presence of cytochrome P-450 (CYP) and NADPH, react to provide an organic aldehyde and hydrogen gas or a ketone and free alkane (see Figure 5-1). [22]  respectively. This comes at the expense of two reduced glutathione (GSH) molecules being oxidized into GSSG. [29] [30] Whereas organic hydroperoxides are generally far too reactive to be used as forms of medicine [31], cyclic peroxides are used as anti-parasitic drugs, like artemisinin or its analogs. [32] [33][34] [35] One theory on cyclic peroxide drug mechanism of action is that they are activated to carbon-centered radicals in the erythrocyte by iron(II) that has been freed by the actions of the parasite. [33][32] [35] An alternative to the carbon-centered radical formation premise is a compelling theory involving cyclic peroxide oxidation of flavin cofactors, that disrupt the homeostasis of ROS removal by GSH and NADPH. With the parasites having no natural protection from ROS of their own, they rely on the host for protection. With the host now overburdened by ROS, the parasite is destroyed. [36][37] However, involvement of selenium-containing enzymes may also contribute to this process. Specific mechanisms notwithstanding, literature data would suggest that cyclic peroxides may be stable in the body and available for systemic circulation. When (and if) there is interaction with ferrous iron or some other agent that may initiate radical formation or two-electron reduction, significant toxicity or mutagenicity may occur, depending on where these molecules eventually reside. It should be noted that TATP was shown to be stable in the presence of iron(II) when solvated in tetrahydrofuran, but not in ethanol. [38] While HMTD is not volatile [39] and is most likely detected by the scent of its degradation products, [40] TATP is quite volatile as an intact molecule and is known to sublime. [41] This would make inhalation the most likely route of exposure. Furthermore, with sensitive explosives, using gloves is generally not an acceptable practice as the static associated with nitrile or latex can cause them to initiate. With these compounds being rather lipophilic (log Po/w: TATP = 3.21 and HMTD = 1.99), the risk of exposure due to absorption through the skin is rather high. Investigation of the metabolism of TATP and HMTD may determine if measures should be instituted to mitigate exposure for both animals and humans working with these compounds.
We have previously established that TATP vapor in a closed vessel exists at a concentration of about 375 μg/L [42]. With an average dog lung capacity of about 40 mL/Kg [43], a 30 Kg dog (~65 lbs), has a lung capacity of 1.2 L. A full breath would lead to an exposure of 450 μg. For humans, with a vital lung capacity of 4 to 5 L, [44] exposures in a closed room over a short time could lead to very large doses. As a forensic consideration, if TATP and HMTD are not extensively metabolized and are stable in the body, individuals producing large quantities of these materials for nefarious reasons may be identified by the analysis of small amounts of blood.
The analysis of TATP and HMTD by reverse phase liquid chromatographymass spectrometry (LC-MS) is the most amenable means of separation and detection for aqueous-based samples of these molecules and their potential metabolites.
Development of assays for these compounds have presented significant analytical challenges. For instance, LC-MS analysis of peroxides cannot have acetonitrile in the mobile phase solvent due to severe, direct, gas-phase ion suppression by the solvent. [45] While methanol is a better solvent for ionization, both HMTD [46] and TATP [42] react with alcohols in the gas phase depending on the conditions used.
Since concentration of TATP solutions cannot be performed due to the volatility of TATP, it is fortunate that the chromatographic peak shape is relatively unaffected by high levels of strong solvent content in the injection plug. Also fortunate is that HMTD is not affected by solvent evaporation since its peak shape and sensitivity are tremendously altered by small changes to the organic content in the sample plug. [47] The fully deuterated TATP and HMTD molecules have been synthesized for use as an

TATP, TATP-OH and MEKP Synthesis:
Triacetone triperoxide (TATP) was synthesized following the literature methods with the exception that hydrochloric rather than sulfuric acid was used. [48]   Methyl ethyl ketone peroxides (MEKP) were synthesized by a modified literature method. [49] In a test tube containing a micro stir bar, hydrogen peroxide (50-wt%, 1.4 mL) was mixed with methyl ethyl ketone (0.82 mL, 9.49 mmol). The solution was chilled in an ice bath and concentrated H2SO4 (0.5 mL, 9.38 mmol) was added slowly so that the temperature did not exceed 20 °C. Stirring continued for 15-18 hours before the solution was extracted with pentane, washed with saturated ammonium sulfate (3x3 mL), deionized water (3x3 mL) and dried with sodium sulfate. The product was stored as a solution in pentane and was pipetted into tared vials for immediate dilution with MeOH to desired concentrations. Purification of individual MEKP was attempted, but to date has been unsuccessful. With no one specific MEKP structure or concentration, crude MEKP product was weighted and diluted to 10 mg/mL in MeOH. LC-MS detection of the observed products, primarily ammonium adducts of the proposed linear dihydroperoxide (DHP) dimer, trimer, tetramer and pentamer, were used to estimate an average molecular weight based on their relative intensities (346 g/mol, Supporting Information, Appendix 5). This sample was then diluted to the appropriate concentration to prepare incubations at ~30 μM.

Instrumentation
Using a Thermo Electron LTQ Orbitrap XL or Exactive mass spectrometer equipped with an APCI interface, ions were generated and introduced into the ion transfer tube set 275 ºC. All work was performed using positive ion mode. Tune conditions for APCI were as follows: discharge current, 5 µA; N2 sheath gas, 40  MeOH (channel A) and 95% aqueous 10 mM NH4OAc (channel B) for introduction onto a Thermo Syncronis C18 column (2.1 x 50 mm, 5 µm). Initial conditions were held for 1.5 minute before a linear ramp to 35% A/65% B over 1.5 minutes followed immediately by a linear ramp to 95%A/5% B over the next minute. This concentration was held for 2 minutes before a 30 second transition to initial conditions with a hold of 1.5 minutes. As an internal standard (IS), d18-TATP at 10 μg/mL (41.7 μM) in ACN was added 1:1 to aqueous TATP samples with a final concentration of 5000 ng/mL (20.8 μM). XIC were integrated using the Genesis peak detection algorithm in Thermo Xcalibur Quan Browser. Linear dynamic range comparing concentration to peak area response ratio, relative to the IS, extended from 25 ng/mL (112.6 nM) to 20000 ng/mL (90.1 μM) using 10 points and 1/x weighting of the calibration curve. Quality control (QC) samples were 75, 1500 and 15000 ng/mL. Stability determination for TATP stored in ACN has been previously determined. [42] All necessary dilutions were made in 50/50 ACN/water. An example calibration curve and QC data can be seen in the Supporting Information, Appendix 5. The same analytical procedure for TATP was used to quantify the synthesized TATP-OH. The LLOQ was 50 ng/mL, but linearity extended to only 5000 ng/mL.

Results and Discussion:
Initial interrogation of TATP metabolism was limited by poor detection limits for the analysis. With a lower limit of quantification (LLOQ) of only 500 ng/mL, we had to perform preliminary work with a substrate (TATP) incubation concentration of 100 μM in 1 mg/mL [protein] of dog liver microsome (DLM), knowing that this was probably above enzyme saturation. The obtained information was still useful as preliminary data on the Phase I metabolism. Only one metabolite, TATP-OH, was detected and identified as the hydroxylation of one of the primary methyl groups of TATP ( Figure 5-2). A significant amount of the TATP remained intact. This product formation was NADPH-dependent and confirmed by incubation of the fully deuterated TATP. To perform any type of enzyme kinetics, incubations would require detection well below 1 μM (222 ng/mL). With that level being diluted in half with ACN/IS addition and our inability to concentrate the samples by evaporation, significant efforts to lower the detection limit were required. The target LLOQ was 10 ng/mL, approximately 10x less than the required 111 ng/mL needed for 1 μM incubations.
Achieving this level was possible by adjusting the MS conditions and monitoring m/z 89.0597, the gas phase reaction product of TATP with MeOH. [42] However, to assure that related metabolites could also be detected we chose to look at the intact TATP ammonium adduct at m/z 240.1442 which could now be detected with an LLOQ of 25 ng/mL.
With adequate assay conditions, incubations were then performed at 5,10,20,30,40,50,75 and 100 μM, initiating the analysis by the addition of NADPH. Due to significant daily variability in the data, closer analysis of the certain aspects of the procedure were investigated. It was considered that the TATP might be insoluble at higher buffer concentrations; therefore, 100 mM potassium phosphate was reduced to 10 mM. This had little effect on the results, but tests were continued with 10 mM concentration. Initiation of the reaction with NADPH was associated with approximately 5 to 15% decrease from the initial TATP concentrations. Additionally, detectable levels of the TATP-OH were present in the time zero samples, suggesting very rapid metabolism (approximately 20 seconds to add substrate, mix and sample).
Varying protein concentration from 1 mg/mL to 0.5, 0.2 and 0.1 mg/mL did not account for this initial drop, but it was decided that 0.5 mg/mL provided data that was more manageable with regards to sampling times. Time zero was then specifically changed to 0.5 minutes. Speculations that NADPH or the magnesium ions in the MgCl2 could react with the peroxide were investigated. Performing 10 μM TATP DLM incubations at MgCl2 concentrations of 0, 2 and 5 mM and incubating TATP under three different conditions (in only MgCl2; only NADPH; and MgCl2 with NADPH) showed that neither magnesium nor NADPH were directly associated with TATP depletion. It was, however, observed that the rate of metabolism was higher with MgCl2 present and that there was no difference between 5 and 2 mM MgCl2. It might be possible that the addition of cold NADPH to the reaction causes the TATP to precipitate. It was also speculated that the TATP might bind tightly in a specific enzyme pocket and addition of the NADPH reducing equivalent caused rapid metabolism that appeared as a drop in initial concentration. This latter speculation was dismissed since the subsequent metabolic rate was not consistent with this behavior and since the TATP-OH metabolite was not detected at levels corresponding to 5-15% production (as was confirmed later following TATP-OH synthesis). To date, this issue has never been resolved but was overcome by initiating reactions with the addition of TATP to the incubations already containing NADPH.
With these issues controlled, day to day variability was still unacceptable.
Evaporation in the headspace of the tubes was the prime suspect. On several different days, incubation of two closed, aqueous TATP samples for 1 hour were performed.
One sample remained closed the full hour and one was sampled every 15 minutes.
Fortunately, there was no detectable substrate degradation, but significant sample loss (frequently > 3% depending on concentration) was observed due to the opening of the tubes for sampling. Attempting to perform separate incubations for each time point in individual tubes provided data with even more inconsistency. With many variables to affect specific evaporation at any given time, it was decided that every incubation would have an identical, parallel incubation performed in buffer alone. The concentration loss at each time point from these buffer-only incubations was added to the TATP concentrations from the metabolic incubation to account for non-metabolic TATP loss due to evaporation. Data for a single incubation trial at 50 μM TATP in DLM is shown in Figure 5  Considering that further oxidation of TATP-OH to the corresponding aldehyde may occur, SC was added to the reaction when TATP-OH formation appeared to plateau.
No TATP-specific reaction products, including acetone, were detected.  CYP2B11, this only accounts for about 5-6% of the 40% metabolized. With cytochrome b5 not available in the rCYP's for canines, human CYP b5 was added at rCYP2B11 at 5x the P450 concentration (as recommended by the vendor) and reanalyzed. The resulting data showed little change from rCYP 2B11 alone at 5 minutes. However, when incubations were carried out to 10 minutes, 30% of the TATP was metabolized in the rCYP2B11 with b5 (compared with little change to rCYP2B11 alone, Figure 5-6). This would support the Lineweaver-Burk kinetic plot suggesting that multiple enzymes may participate in the oxidative metabolism of TATP.
Several experimental conditions were designed in an attempt to identify the actors in TATP metabolism. Although flavin-containing monooxygenase (FMO) enzymes are unlikely to participate in TATP oxidation, their involvement was considered, as cyclic peroxide interaction with flavoenzymes are suspected with antiparasitics. [36] There was with no evident change to the metabolism when performing the DLM incubation at a temperature (45 °C) known to deactivate FMOs. With the high affinity and low capacity for metabolism, it is possible that the metabolism is perpetrated by a CYP that constitutes a minor portion of the total P450 content in DLM. An attempt was made to deactivate all P450 enzymes by pretreatment with 1-mM ABT, although it has been reported that 1-ABT fails to inhibit all P450. [51] Pretreatment of the microsomes with 1mM 1-ABT did not change the metabolism of TATP.  It was determined that identifying the fate of the TATP molecule would be a more valuable use of resources than trying to determine specific enzymes. To evaluate the possibility that the TATP ring may open to produce a free hydroperoxide, we began to investigate the effects of MEKP metabolism. Incubations of ~30 μM MEKP in DLM were performed with and without NADPH, and in the presence of both GSH and SC. The metabolism of each polymer showed an NADPH-dependent formation of methyl ketone products suggesting that ethane is exclusively lost (not methane).
Addition of SC showed many related products corresponding to this metabolism as well as a large signal for trapped MEK (m/z 130.0975). It may be that either MEK is a direct metabolic product or that metabolism initiates the degradation into MEK. These peaks were not detected with SC added to MEKP crude standard and only to a minor degree without NADPH addition. Glutathione however, was highly reactive with MEKP with or without the presence of enzymes or NADPH. In each sample, all MEKP products eventually disappeared and the oxidized glutathione (GSSG) was extensively formed (Scheme 5-1). Retrospective analysis of TATP incubations with GSH or SC show no sign of the oxidized GSSG or of SC-trapped acetone. This lack of analogous reactivity with TATP suggests that only linear hydroperoxides participate in these reactions and TATP ring-opening does not occur under the conditions tested. Attempts to chemically synthesize and purify the metabolite of TATP provided a small quantity of compound with the same exact mass, chromatographic retention time and product ion spectrum as the TATP-OH metabolite. Both this compound and the metabolite react with methanol in the gas phase to produce a large product at m/z 105.0546, the oxidized product corresponding to m/z 89.0597. [42] This was enough supporting evidence to conclude that the synthesis was successful. Initially, we tested this compound for volatility and found that TATP-OH was not volatile in an aqueous environment, but under certain conditions it was not stable. Upon incubation at 37 °C in oxygenated buffer for one hour, TATP-OH was mostly degraded. To see if this compound was stable in our storage solution (ACN) and other matrices, we incubated it in several solutions-MeOH, ACN, water, buffer and buffer with SC-in closed containers for 60 minutes. The results ( Fortunately, it did appear to be more stable in pure water and very stable in organic environments. When SC was added to the oxygenated buffer, at 60 minutes, only 70% of the TATP-OH was gone but an appreciable increase in the amount of SC-trapped acetone (m/z 116.0818) was observed (Table 5-2). While the SC-trapped acetone did significantly increase with TATP-OH incubated for 60 minutes, some m/z 116.0818 was present in all buffer samples with SC added (but not in SC samples placed in water only). This could be the traces of acetone remaining from the synthesis of TATP or TATP-OH, but signal was detected in the oxygenated phosphate buffer with SC added. This cast some doubt on the SC trapping experiments.  It is likely that the residual acetone was metabolized away, but no other SC-trapped products based on known acetone metabolism (acetol and methyglyoxal) [52] [53] were detected. Following an additional 30 minutes of incubation, it appears as though the signal for m/z 122.1195 increased by about 25%, suggesting that acetone may be produced (see Table 5-3). However, this may be due to previously observed chemical degradation and not metabolism. The results for TATP and SC were inconclusive since the m/z 116.0818 was significant in all SC-containing samples tested. Incubation of TATP-OH was then performed in DLM at a concentration of 10 μM and compared to a similar incubation of TATP. Figure     of total loss appears much closer to the TATP-OH degradation rate than the metabolic rate, suggesting TATP is metabolized and TATP-OH is chemically degraded.
However, with the metabolism of TATP-OH being so rapid, it is unlikely that no metabolism of this metabolite occurs.
To investigate the possibility of any time-dependent inhibition of either TATP or TATP-OH resulting from a second metabolite, the following experiment was performed. An incubation of 10 μM TATP-OH (5 μL ACN organic) and a blank of 5 μL of ACN were incubated in DLM. After 20 minutes, TATP (10 μM) was added to each incubation, and samples were analyzed at 0, 2, 4, 6, 10 and 15 minutes. The rate of TATP loss proceeded identically for both incubations. Also, both pre-treated and non-pre-treated samples produced very similar final concentrations of TATP-OH in 15 minutes. Since there was probably TATP-OH remaining in the pre-incubated samples, the rate of TATP-OH formation in these samples appeared to be nearly half that of the non-pretreated samples ( Figure 5-9). This may suggest that the Vmax for TATP-OH metabolism is constant with and without TATP present. Data indicates that neither TATP-OH nor its undiscovered metabolite(s) interfere with the metabolism of TATP.
Addition of SC after 20 minutes of TATP incubation in DLM showed no SC-trapped products, including the SC-trapped acetone. While data suggests that TATP-OH degrades into acetone in solution, there appears to be no evidence that it is metabolized into acetone. The loss of the TATP-OH metabolite could be due to complete degradation of the molecule into products smaller than acetone. It could also be due to some non-specific binding of a yet undiscovered, second metabolite to a protein or other material within the system.
With TATP-OH metabolism being considerably more rapid than that of TATP, the question becomes, why is the formation and buildup of TATP-OH observed?
Incubations of TATP, TATP-OH and a mixture of TATP plus TATP-OH (all performed with 5 μM concentrations) in DLM provided very interesting results ( Figure 5-10). In both the TATP and TATP+TATP-OH incubations, the rate of TATP depletion was nearly identical. TATP-OH alone showed an initial metabolic depletion rate approximately 1.5 times faster than that of TATP. Mixing TATP with TATP-OH showed an initial formation rate of TATP-OH (exceeding the 5 μM initial concentration) to be nearly identical to the initial formation rate of TATP-OH in the TATP only incubation. This formation of TATP-OH in the mixture plateaued quickly (between 4 and 6 minutes). This data indicates that TATP-OH and TATP are probably metabolized by the same enzyme (if not the same enzymatic pocket), otherwise, buildup of TATP-OH would not be observed due to its faster metabolism that TATP. Reversible competitive inhibition is the most likely mechanism of TATP-OH inhibition as the Km is clearly increased in the presence of TATP and is also very likely that the Vmax is unaffected. This study would also support that as TATP is metabolized into TATP-OH, the metabolite does not greatly affect TATP metabolism until concentrations significantly exceed that of TATP. isoforms showed about 20% reduction of TATP-OH, similar to the chemical degradation observed in buffer only. Since this was the only enzyme found to metabolize TATP-OH and it was very similar to the DLM, it may be fair to say that only CYP2B11 metabolizes TATP-OH and that TATP dominates a competition for this enzyme. While reversible competitive inhibition seems likely, determining the specific mechanism may not be easily performed since the initial substrate is also the metabolic inhibitor of its only metabolite.

Conclusions:
TATP metabolism was characterized in canine liver microsomes. Only one hydroxylated metabolite was detected. Although the clearance was high, the low capacity of metabolism suggests that large exposure to TATP vapor could lead to significant systemic exposure. This is could be further evidenced by the lack of lung microsomal activity, since inhalation is the most likely route of exposure. With the assumption that absorption would not be much of a barrier, TATP may be sequestered in cells (and toxic) if its clearance does not progress by other means (cytosolic or Phase II metabolism). While TATP may be metabolized by more than one enzyme in the microsomes, only CYP2B11 (in conjunction with cytochrome b5) was identified.
Its only detected metabolite, TATP-OH, is metabolized by the same enzyme (CYP2B11 only) and TATP appears to dominate a reversible competitive inhibition of the TATP-OH metabolism. By comparing the TATP and TATP-OH levels following a 10 μM TATP incubation, their concentrations combine to account for all TATP initially added until approximately 15 minutes when the sum of these concentrations both begin to disappear. The data suggests that another metabolite is formed that is either not detected or non-specifically binds with some protein or other material related to the system. Whatever the fate of this second, unknown metabolite, it does not appear to affect the metabolism of TATP or TATP-OH, but its lack of detection is suggestive of high reactivity.

Supplemental Information:
Supporting Information  Figure S2-1. The same HMTD standard curve sample (10000 ng/mL) injected on day 1 and day 7 for autosampler stability.