Abstract

   Acetaminophen (APAP), a commonly used over-the-counter analgesic,
   accounts for approximately fifty percent of the cases of acute liver
   failure (ALF) in the United States due to overdose, with over half of
   those unintentional. Current clinical approaches for assessing APAP
   overdose rely on identifying the precise time of overdose and
   quantitating acetaminophen alanine aminotransferase (ALT) levels in
   peripheral blood. Novel specific and sensitive biomarkers may provide
   additional information regarding patient status post overdose. Previous
   non-clinical metabolomics studies identified potential urinary
   biomarkers of APAP-induced hepatotoxicity and metabolites involved
   pathways of tricarboxylic acid cycle, ketone metabolism, and tryptophan
   metabolism. In this study, biomarkers identified in the previous
   non-clinical study were evaluated in urine samples collected from
   healthy subjects (N = 6, median age 14.08 years) and overdose patients
   (N = 13, median age 13.91 years) as part of an IRB-approved multicenter
   study of APAP toxicity in children. The clinical results identified
   metabolites from pathways previously noted, and pathway analysis
   indicated analogous pathways were significantly altered in both the
   rats and humans after APAP overdose. The results suggest a metabolomics
   approach may enable the discovery of specific, translational biomarkers
   of drug-induced hepatotoxicity that may aid in the assessment of
   patients.

   Keywords: acetaminophen, biomarker, mass spectrometry, metabolomics,
   overdose

1. Introduction

   Acetaminophen (APAP) is a commonly used over-the-counter drug worldwide
   with a reported regular weekly consumption of 60 million by Americans
   [[34]1]. While generally safe at the recommended dose, overdose of APAP
   accounts for half of the cases of acute liver failure (ALF) in the
   United States [[35]2]. APAP is available in various formulations
   (tablets, liquids, suppositories, etc.) and in prescription and
   over-the-counter combination drugs; its widespread availability makes
   it more likely that a patient may consume more than the recommended
   dose. Approximately half the patients with APAP hepatotoxicity
   experience mild to moderate reactions, while approximately 48% are
   diagnosed with ALF; among those with ALF diagnoses, it has been
   reported that 29% require a liver transplant, with a 28% mortality rate
   [[36]3]. Early therapeutic intervention with N-acetylcysteine (NAC) is
   critical to reduce the morbidity and mortality associated with
   APAP-induced hepatotoxicity. However, recognition of APAP
   hepatotoxicity can be challenging, as patients may be unaware of
   excessive exposure, and the early stages of toxicity are non-specific.
   Moreover, patients may present for medical evaluation after the time
   period for which NAC is most effective. ALT is often used to evaluate
   the severity of APAP-induced hepatotoxicity. However, there is a delay
   between ingestion of a toxic dose and the rise of ALT.
   Metabolomics-focused technologies can identify urinary biomarkers
   related to progression and recovery from APAP-induced hepatotoxicity.
   Such biomarkers, in conjunction with the clinical standards, may be
   useful in predicting how a patient will progress through the stages of
   APAP hepatotoxicity.

   APAP-induced hepatotoxicity has been thoroughly investigated, and the
   mechanism is generally well understood. APAP is primarily metabolized
   by Phase II metabolic pathways to glucuronidated and sulfated
   conjugated forms, which are excreted in the urine. A small portion of
   APAP (~10% depending on amount ingested) undergoes metabolism to the
   highly reactive toxic metabolite N-acetyl-para-benzoquinone imine
   (NAPQI) that can bind to proteins and form APAP-protein adducts
   [[37]4]. In addition, NAPQI can bind glutathione to form non-toxic
   cysteine and mercapturic acid conjugates, which are excreted in the
   urine. In the case of hepatotoxicity, the glucuronidation and sulfation
   pathways become saturated, resulting in the formation of excess NAPQI,
   which depletes the glutathione stores and leads to hepatocyte necrosis.
   APAP has been considered a model hepatotoxicant in metabolomics
   studies, and there have been multiple literature reports of the
   analysis of urine samples in animals
   [[38]5,[39]6,[40]7,[41]8,[42]9,[43]10,[44]11,[45]12,[46]13,[47]14,[48]1
   5,[49]16] and humans following ingestion of APAP
   [[50]5,[51]10,[52]17,[53]18,[54]19,[55]20]. Many of the studies have
   reported changes in oxidative stress, glutathione-related and citrate
   cycle (TCA) metabolites, and altered fatty acids following ingestion of
   high doses of APAP.

   The non-clinical studies can identify potential biomarkers of interest
   and provide pathway and mechanistic information. Importantly, it has
   been shown that the changes in the non-clinical models may be
   translated to the clinical setting. In this study, urine samples were
   collected from children or adolescents hospitalized with APAP overdose
   under an IRB-approved protocol [[56]10,[57]21,[58]22]. Nuclear magnetic
   resonance (NMR)- and ultra-performance liquid
   chromatography/quadrupole-time-of-flight mass spectrometry (UPLC/QToF
   MS)-based metabolomics techniques were used to evaluate the urine
   samples from six control and 13 overdose patients. In particular,
   metabolites previously identified in the non-clinical APAP study
   [[59]13] were evaluated in this study. Pathway analysis of both the
   clinical and non-clinical data was carried out and similar pathways
   were noted to be altered in both sample sets. The similarity between
   the clinical and the non-clinical data suggest that metabolomics can be
   used to identify translatable biomarkers and pathways, which may have
   relevance to the clinical assessment of patients with APAP
   hepatotoxicity.

2. Results

   In this study, there were a total of six subjects in the control group
   and 13 subjects in the overdose group, which represented a subset of
   subjects from previous publications [[60]10,[61]22]. [62]Table 1
   provides the demographic and clinical chemistry data for the 19
   subjects in the study.

Table 1.

   Demographics and clinical data for control and overdose subjects. Peak
   acetaminophen alanine aminotransferase (ALT) and adducts are reported
   as the median (range).
         Variables         Control N = 6    Overdose N = 13
        Age (years)            14.08             13.91
      PEAK ALT (IU/L)           NA         2050 (106, 6072)
   Peak Adduct (nmol/mL) 0.00 (0.00, 0.00) 1.48 (0.20, 6.69)
   [63]Open in a new tab

   Within the overdose group, three patients had urine samples collected
   at three or more timepoints, two patients had samples collected at two
   timepoints, and the remaining eight had urine samples from a single
   timepoint. For the control group, a single urine sample was acquired.
   The median age was 14.08 years for the control group, while the median
   age was 13.91 for the overdose subjects. Elevations in ALT for the
   control group were not noted; therefore, the peak ALT was not
   applicable. For the overdose group, the median peak ALT was 2050 IU/L,
   with values ranging from 106–6072 IU/L at the peak of ALT. The time to
   treatment for the APAP overdose group was variable, which may in part
   account for the large range in ALT values. Additionally, as noted, only
   five patients in the overdose group had samples collected over multiple
   timepoints, so the reported ALT value may not represent the maximum
   value in the patients with only one sample. APAP-protein adducts were
   measured in each group, with a reported median of 0.00 nmol/mL in the
   control group and a median level of 1.48 nmol/mL in the overdose group.

   [64]Figure 1A,B shows the score plots from PLS-DA of the QToF-MS data
   in both positive and negative modes, respectively.

Figure 1.

   [65]Figure 1
   [66]Open in a new tab

   Partial least squares discriminant analysis (PLS-DA) scores plots for
   both (A) positive and (B) negative ionization modes. Control samples
   are shown in blue and samples from overdose subjects are in red.

   APAP and APAP metabolite peaks were removed prior to analysis, and
   [67]Figure 1A,B shows a distinct separation between the control and
   overdose groups in both ionization modes. Closer examination of the
   scores plot reveals that urine samples from overdose patients collected
   at multiple timepoints clustered closely together. Individual
   metabolites from the QToF-MS spectra were evaluated based upon the
   loadings plot and identified based upon Human Metabolome Database or
   based upon authentic standards in an in-house database. Metabolites
   detected in the NMR spectra were quantified using Chenomx NMR Suite.

   The datasets from the identified metabolites from QToF-MS and NMR were
   combined for pathway analysis using MetaboAnalyst ([68]Figure 2).

Figure 2.

   [69]Figure 2
   [70]Open in a new tab

   Metabolome view from pathway analysis performed using MetaboAnalyst.
   Select pathways with high pathway impact and/or high p-value are
   labeled.

   Significantly altered pathways (p < 0.05) that also had high impact
   values include arginine and proline metabolism, TCA cycle, taurine and
   hypotaurine metabolism, glycine, serine and threonine metabolism, and
   glutathione metabolism. Pathway significance is determined from pathway
   enrichment analysis and based upon the values for each compound in the
   dataset. The impact value, on the other hand, is determined by pathway
   topology analysis. Impact is determined based upon the importance of a
   metabolite within a pathway; a metabolite that is found at a junction
   point within a pathway may have a greater impact on the pathway
   function if the level is altered. Pyruvate metabolism had a high
   impact, although the p-value was not significant. The impact value for
   pyruvate metabolism was high, since pyruvate, lactate, and acetate are
   all located at junctions within the pathway ([71]Figure S9); however,
   the changes in these metabolites were not significant in control versus
   overdose, resulting in a non-significant p-value for the pathway. Other
   pathways that were significantly altered but with lower impact included
   tryptophan, pyrimidine, vitamin B6, and b-alanine metabolism;
   pantothenate and CoA biosynthesis; and the pentose phosphate pathway.
   Among the high impact pathways, pyruvate metabolism, glycine, serine,
   and threonine metabolism, and arginine and proline metabolism are all
   connected to the citric acid cycle. Furthermore, taurine and
   hypotaurine metabolism as well as arginine and proline metabolism are
   connected to glutathione metabolism, which was also significantly
   altered. The results are consistent with our previous study of
   APAP-induced hepatotoxicity in Sprague Dawley rats, which indicated
   altered energy and glutathione pathways [[72]13].

   Metabolites from NMR or QToF-MS with a significant correlation to ALT
   or APAP-protein adducts are presented in [73]Table 2.

Table 2.

   Select metabolites with high correlations to alanine aminotransferase
   (ALT) or acetaminophen (APAP)-protein adducts. Metabolites with
   significant correlations (p < 0.05) to ALT or APAP-protein adducts are
   in bold.
   Metabolite Platform ALT APAP-Protein Adducts
   2-Oxoarginine ^a LCMS POS, 0.72 min, m/z 174.09 0.6192, p = 2.40 ×
   10^−2 0.4318
   Ascorbic acid ^a LCMS POS, 0.72 min, m/z 177.04 0.4361 0.8446, p = 2.81
   × 10^−4
   Ascorbic acid ^a LCMS NEG, 0.72 min, m/z 175.02 0.3889 0.8159, p = 6.72
   × 10^−4
   Alanine NMR, 1.47, 3.78 ppm 0.2600 0.6792, p = 1.073 × 10^−2
   Choline NMR, 3.19, 3.51, 4.06 ppm 0.1789 0.5993, p = 3.04 × 10^−2
   Citrulline ^a LCMS POS, 0.67 min, m/z 176.10 −0.3810 −0.6676, p = 1.27
   × 10^−2
   Cresol ^a LCMS NEG, 3.77 min, m/z 107.05 −0.5642, p = 4.46 × 10^−2
   −0.4348
   Fructose NMR, 3.55–4.11 ppm 0.3900 0.8170, p = 6.51 × 10^−4
   Glucose NMR, 3.23–3.89, 4.64 ppm 0.2778 0.6367, p = 1.93 × 10^−2
   Hippurate ^a LCMS NEG, 3.32 min, m/z 178.05 0.4858 0.8561, p = 1.88 ×
   10^−4
   Hippurate NMR, 3.96, 7.54, 7.63, 7.82 ppm 0.4116 0.8350, p = 3.82 ×
   10^−4
   Hydroxybutyrylcarnitine ^a LCMS POS, 1.05 min, m/z 248.15 −0.5674, p =
   4.46 × 10^−2 −0.4900
   Indoxyl ^a LCMS POS, 3.30 min, m/z 134.06 0.7062, p = 6.98 × 10^−3
   0.8230, p = 5.49 × 10 ^−4
   Lactate NMR, 1.32, 4.11 ppm - 0.7906, p = 1.29 × 10^−3
   Proline ^a LCMS POS, 0.83 min, m/z 116.07 −0.5733, p = 4.05 × 10^−2
   −0.5376
   Propylene glycol NMR, 1.13, 3.44, 3.54, 3.88 ppm 0.4025 0.8263, p =
   4.99 × 10^−4
   Pyruvate NMR, 2.36 ppm 0.3680 0.7134, p = 6.18 × 10^−3
   Taurocholic acid isomer ^b LCMS NEG, 4.83 min, m/z 514.28 0.4259
   0.6891, p = 9.18 × 10^−3
   Trimethylamine N-oxide NMR, 3.25 ppm 0.5807, p = 3.74 × 10^−2 0.3733
   Uracil ^a LCMS POS, 0.78 min, m/z 113.03 - 0.6349, p = 1.97 × 10^−2
   Uric acid ^a LCMS NEG, 0.83 min, m/z 167.02 - 0.5636, p = 4.49 × 10^−2
   [74]Open in a new tab

   ^a Compounds identified based upon an authentic standard from an
   in-house database; ^b Compounds identified from the Human Metabolome
   Database.

   Among the reported metabolites, many are related to the significantly
   altered pathways identified in the pathway analysis. In addition,
   taurocholic acid had a significant correlation to APAP-protein adducts.
   Taurocholic acid and other bile acids have previously been shown to be
   altered due to APAP-induced hepatotoxicity in both the rodent model and
   clinical samples [[75]5,[76]21,[77]22]. Hydroxybutyrylcarnitine had a
   significant negative correlation to ALT levels. The acylcarnitines have
   also been shown to be altered early in APAP-induced hepatotoxicity in
   the rodent model [[78]5,[79]13], with the maximum in the acylcarnitine
   occurring prior to the maximum of ALT. p-Values were adjusted for
   multiple comparison testing using the false discovery rate (FDR) and
   after adjustment; only the correlations for hippurate, ascorbic acid,
   and propylene glycol were significant.

3. Discussion

   Hepatotoxicity due to APAP overdose remains a critical issue, with high
   rates of morbidity and mortality that can occur if not recognized
   quickly [[80]1,[81]23,[82]24,[83]25]. In this study, metabolomics
   analyses have been employed to evaluate clinical urinary biomarkers of
   APAP-induced hepatotoxicity. Such non-invasive biomarkers may be useful
   in determining the stage of hepatotoxicity in the clinical setting and
   may predict prognosis when used in conjunction with the current
   clinical measures.

   The urinary metabolites reported in [84]Table 2 represent several
   different pathways that have been previously identified in non-clinical
   models of APAP-induced hepatotoxicity, including glutathione
   metabolism, bile acids metabolism, the citric acid (TCA) cycle, and
   fatty acid β-oxidation
   [[85]6,[86]14,[87]15,[88]21,[89]26,[90]27,[91]28]. Furthermore, a
   previous clinical study evaluated the ability of urine metabolites to
   predict APAP-induced toxicity in humans identified responders versus
   nonresponders based upon a subset of metabolites [[92]29]. Among the
   metabolites reported in the study by Winnike and coworkers [[93]29],
   alanine, hippurate, creatinine, and trimethyl amine-N-oxide were also
   found to be correlated with ALT or APAP protein adducts, as shown in
   [94]Table 2. Alanine and hippurate were increased in the overdose group
   in this study and in the group of responders from the study by Winnike
   et al. [[95]29], while creatinine was decreased in the overdose group
   and the responders.

   The relatively increased formation of toxic NAPQI results in depletion
   of glutathione (GSH). Glutathione is important in the detoxification of
   reactive molecules, and depletion of GSH following APAP overdose has
   been well-established. While GSH was not measured directly in this
   study, pathway analysis showed that the glutathione metabolism pathway
   was significantly altered with a high impact value. In addition,
   multiple pathways associated with glutathione metabolism were also
   significantly altered, including taurine and hypotaurine metabolism,
   arginine and proline metabolism, and glycine, serine, and threonine
   metabolism. These alterations in glutathione metabolism and other
   glutathione-relevant pathways may reflect an effort at the cellular
   level to replenish GSH stores to detoxify NAPQI and prevent further
   liver injury.

   Previous reports [[96]5,[97]6,[98]13,[99]30] have noted that APAP
   overdose can result in global energy failure potentially through the
   inhibition of fatty acid β-oxidation. This study also showed that the
   TCA cycle was altered in the overdose group compared to the control
   group. Pathway analysis indicated that the TCA cycle had a high impact
   value and significant p-value (p < 0.05). In addition, [100]Table 2
   shows that hydroxybutyrylcarnitine had a significant negative
   correlation with ALT. It was previously reported that the
   acylcarnitines were elevated in serum from rats given a toxic dose of
   APAP prior to an increase in ALT levels [[101]13]. Acylcarnitines have
   been proposed as blood-based biomarkers of reduced β-oxidation and
   mitochondrial dysfunction. In healthy individuals, urinary
   acylcarnitine concentrations are typically negligible
   [[102]31,[103]32]. It has been hypothesized that the alteration in
   fatty acid β-oxidation may reflect a shift in energy metabolism to
   glycolysis to make up for the loss of ATP. Further studies are needed
   to confirm the switch to glycolysis in the clinical samples.

   In this study, the time from subject’s overdose to presentation in the
   hospital was variable, along with the time to NAC treatment.
   Furthermore, the study population represents a small cohort with six
   control subjects and 13 overdose patients, and more studies are needed
   to verify the results and determine whether other factors like genetics
   play a role in the response to acetaminophen. However, despite these
   challenges, the results from this clinical metabolomics study were
   consistent with previous findings in a rat model of APAP-induced
   hepatotoxicity. Metabolites from arginine and proline metabolism, TCA
   cycle, taurine and hypotaurine metabolism, glycine, serine and
   threonine metabolism, and glutathione metabolism pathways were
   significant in the APAP overdose group compared with the control group.

4. Materials and Methods

4.1. Ethical Approval

   As part of a multicenter study on the toxicity of APAP in children ages
   2–18, biological samples including urine were collected from subjects.
   The study was approved by the institutional review boards of all
   participating institutions, in accordance with the guidelines of the
   1975 Declaration of Helsinki. In this study, only a subset of the urine
   samples was evaluated that fell directly under the purview of the
   University of Arkansas for Medical Sciences (UAMS) Institutional Review
   Board. Informed consent (and assent) documents were approved by the
   UAMS Institutional Review Board.

4.2. Subjects and Sample Collection

   Complete details on the subjects and sample collection can be found in
   James et al. [[104]10]. Urine samples from the control group and the
   overdose group were evaluated in this study. The control group (N = 6)
   included healthy subjects with no APAP use in the preceding 14 days,
   while the overdose group (N = 13) included children or adolescents
   hospitalized with APAP overdose as determined by published guidelines.
   The control group had a single urine sample collected, while the
   overdose group had multiple timepoints collected during
   hospitalization. Additionally, clinical and demographic data were
   collected and stored in a database that also included ALT levels.

4.3. Metabolomics Methods

4.3.1. Chemicals

   Optima LC/MS grade acetonitrile and water were purchased through Fisher
   Scientific (Pittsburgh, PA, USA). Formic acid and leucine-enkephalin
   were purchased from Sigma-Aldrich (St. Louis, MO, USA).

4.3.2. UPLC/QToF-MS Analysis

   A 5 μL volume of urine diluted by water (1:10) was injected on to a
   Waters Acquity Ultra Performance Liquid Chromatography (UPLC) system
   equipped with a Waters bridged ethyl hybrid (BEH) C18 column (2.1 mm ×
   10 cm; 1.7 μm particle size) held at 40 °C for separation of the
   metabolites. In addition to the individual urine samples, a pooled
   sample was prepared from an aliquot of each sample and run along with
   blank samples at regular intervals during the sample analysis. The UPLC
   mobile phases were 0.1% formic acid in water (solution A) and 0.1%
   formic acid in acetonitrile (solution B). Metabolites were eluted at a
   constant flow rate of 0.4 mL/min over the following gradient: 0–30% B
   from 0 to 6 min, 30–50% B from 6 to 9 min, and 50–95% B from 9–11 min.
   The final gradient composition was held for 1 min and returned to 100%
   A at 12.1 min. Following elution from the column, metabolites were
   detected with a QToF Premier mass spectrometer (Waters, Milford, MA,
   USA) operated in positive and negative ionization electrospray modes.
   Capillary voltages were 3.2 kV and 2.4 kV for positive and negative
   modes, respectively, with a cone voltage of 40 V in both modes. Data
   was collected in full scan mode from m/z 100 to 900 in both ionization
   modes. In order to assure mass accuracy, leucine-enkephalin at a
   concentration of 250 pg/μL (in 50:50 acetonitrile:0.1% formic acid) was
   used as the lock-mass in positive mode ([M + H]^+ = 556.2771) and at a
   concentration of 25 ng/μL for negative mode ([M − H]^− = 554.2615).

   Micromass MarkerLynx XS Application Version 4.1 (Waters, Milford, MA,
   USA) with extended statistical tools was employed for peak selection
   and peak alignment of the raw data. Peaks were extracted using the
   parameter settings: XIC window 0.02 Da; retention time window 0.02 min;
   peak width at 5% height 15S; peak-to-peak baseline noise 50. Prior to
   statistical analysis, the data was normalized based upon total ion
   intensity. The resulting aligned dataset was analyzed by unsupervised
   principal component analysis (PCA) and supervised partial least squares
   discriminant analysis (PLS-DA). The resulting ion intensities and
   retention times were exported as EXCEL files. The majority (>90%) of
   the metabolites detected by MS were identified based on the accurate
   mass measurement and retention time by comparison with an in-house
   database, while the rest (<10%) were putatively identified by accurate
   mass (<0.02 Da or 6 ppm) and comparison of the fragment mass spectra
   with the fragment mass spectra in the Human Metabolome Database (HMDB;
   [105]www.hmdb.ca). [106]Supplementary Figure S19 indicates how unknowns
   were identified based upon comparison of the fragmentation mass
   spectrum to the fragmentation mass spectrum of a standard compound in
   an in-house database.

4.3.3. Nuclear Magnetic Resonance Spectroscopy Analysis

   Urine samples were combined with 4,4-dimethyl-4-silapentane-1-sulfonic
   acid (DSS, chemical shift standard) in D[2]O, difluoro trimethylsilanyl
   phosphonic acid (DFTMP, pH standard), and sodium phosphate buffer for
   analysis. A pooled urine sample was also prepared for NMR and run at
   regular intervals during data analysis. Urine samples were analyzed on
   a Bruker Avance NMR spectrometer operating at 600.133 MHz for proton
   and equipped with a Bruker QNP cryoprobe. Water suppression was
   achieved using the noesy1dpr pulse sequence. For each sample, 128 scans
   were collected into 32 K data points over a spectral width of 9615.39
   Hz. Raw data was processed in ACD/Labs 1D NMR Manager (ACD/Labs,
   Toronto, ON, Canada). Processed data were grouped and integrated. The
   individual spectra exported were as *.jdx files for quantification of
   select metabolites in Chenomx NMR Suite (Edmonton, AB, Canada). The
   metabolite concentrations were normalized to the urine creatinine
   concentration in each sample.

4.3.4. Pathway Analysis

   MetaboAnalyst 3.0 was employed for pathway analysis of the clinical and
   nonclinical data
   [[107]33,[108]34,[109]35,[110]36,[111]37,[112]38,[113]39]. The NMR and
   MS data were combined and the HMDB ID used as the compound label. The
   dataset comprised a total of 118 unique compounds from the NMR and MS
   platforms for pathway analysis. Once matched, the data was
   mean-centered and divided by the standard deviation of each variable.
   The Homo sapiens pathway library was selected for the analysis with the
   global test selected as the pathway enrichment analysis method. For
   pathway topology analysis, relative betweenness centrality was selected
   to estimate node importance. The output from the analysis included
   metabolome ([114]Figure 2), pathway and compound views to explore the
   data; KEGG pathways with key metabolites highlighted and compound views
   are provided in [115]Supplementary material (Figures S1–S17). In the
   report, pathway significance is determined from pathway enrichment
   analysis and based upon the values for each compound in the dataset.
   The impact value, on the other hand, is determined by pathway topology
   analysis and the importance of each pathway calculated based upon the
   cumulative percentage from matched metabolite nodes.

4.3.5. Correlations Analysis

   The Pearson correlations between ALT or protein adducts were calculated
   in R using the rcorr function within the Hmisc package. Both the
   correlation coefficient and the p-value for the correlation were
   returned from the calculation. The FDR-adjusted p-values were
   calculated in R using the Benjamini & Hochberg correction [[116]40].

Acknowledgments