Abstract

   Non-invasive measures of the response of individual patients to cancer
   therapeutics is an emerging strategy in precision medicine. Platelets
   offer a potential dynamic marker for metabolism and bioenergetic
   responses in individual patients since they have active glycolysis and
   mitochondrial oxidative phosphorylation and can be easily isolated from
   a small blood sample. We have recently shown how the
   bioenergetic-metabolite interactome can be defined in platelets
   isolated from human subjects by measuring metabolites and bioenergetics
   in the same sample. In the present study, we used a model system to
   assess test the hypothesis that this interactome is modified by
   xenobiotics using exposure to the anti-cancer drug doxorubicin (Dox) in
   individual donors. We found that unsupervised analysis of the
   metabolome showed clear differentiation between the control and Dox
   treated group. Dox treatment resulted in a concentration-dependent
   decrease in bioenergetic parameters with maximal respiration being most
   sensitive and this was associated with significant changes in over 166
   features. A metabolome-wide association study of Dox was also
   conducted, and Dox was found to have associations with metabolites in
   the glycolytic and TCA cycle pathways. Lastly, network analysis showed
   the impact of Dox on the bioenergetic-metabolite interactome and
   revealed profound changes in the regulation of reserve capacity. Taken
   together, these data support the conclusion that platelets are a
   suitable platform to predict and monitor therapeutic efficacy as well
   as anticipate susceptibility to toxicity in the context of precision
   medicine.

Graphical abstract

   [37]Image 1
   [38]Open in a new tab

Highlights

     * •
       Chemotherapeutic drug doxorubicin dose-dependently inhibited
       bioenergetic parameters in platelets.
     * •
       Doxorubicin induced metabolite alterations in multiple pathways
       that modulate cellular bioenergetics.
     * •
       The interactome revealed novel interactions between bioenergetic
       parameters and metabolites in platelets.
     * •
       The integration of platelet bioenergetics with the metabolome can
       be utilized for precision medicine strategies.

1. Introduction

   Doxorubicin (Dox), a non-selective class I anthracycline, also known as
   Adriamycin, is a common chemotherapeutic used to treat a broad range of
   cancers including leukemia, multiple myeloma and cancers of the breast
   [[39][1], [40][2], [41][3]]. Dox was first isolated from Streptomyces
   peucetius in the 1960s and approved for medical use in the United
   States in 1974 [[42]4]. Although several formulations of Dox are in
   use, the occurrence of dose limiting side effects remains relatively
   high [[43]5]. The antineoplastic properties of Dox are attributed to
   its ability to intercalate into nucleic acids [[44]6] leading to
   DNA/RNA damage and inhibition of macromolecule synthesis [[45]7]. Dox
   crosses the cell membrane by passive diffusion [[46][8], [47][9],
   [48][10]] and accumulates in the nuclear compartment and the
   mitochondria [[49]11]. In addition to passive diffusion the organic
   cation transporter SLC22A16, which is expressed in many cancers, also
   contributes to cellular uptake of Dox.

   Pharmacokinetic studies of Dox show a rapid uptake by tissue following
   a single intravenous injection but a slow tissue elimination with a
   half-life of up to 48 h and a maximum serum concentration (C[max]) in
   the low micromolar range [[50]7]. Doxorubicin undergoes three metabolic
   fates with 50% of the administered dose remaining as the parent
   molecule. Dox is primarily metabolized to doxorubicinol (Dox-ol) via a
   two-electron reduction mediated by carbonyl reductase [[51]12,[52]13].
   A second metabolic route involves an enzymatically driven one-electron
   reduction of the quinone functional group by nitric oxide synthase,
   NADPH cytochrome P450 reductase and other oxidoreductases, leading to
   the formation of a semiquinone radical. Molecular oxygen (O[2]) can
   then oxidize the Dox semiquinone radical to form superoxide (O[2]˙^-)
   and hydrogen peroxide (H[2]O[2]) [[53]2,[54]14] contributing to
   oxidative stress and macromolecule-damage. The final metabolic route of
   Dox involves the deglycosidation of the parent compound by microsomal
   reductases to form doxorubicinone (Dox-one), doxorubicinolone
   (Dox-olone), and 7-deoxy-doxorubicinolone (7-Dox-olone)
   [[55]7,[56]15,[57]16]. This route accounts for a small fraction of Dox
   metabolism and has been linked to changes in human erythrocytes energy
   metabolism as seen by a shift in the pentose phosphate pathway and
   inhibition of antioxidant enzymes [[58]17].

   Although Dox is widely used as an anti-tumor agent, its use is
   associated with serious side effects ranging from cardiotoxicities to
   thrombocytopenia [[59]5,[60][18], [61][19], [62][20]]. Dox mediates
   tumor cell death via inhibition of topoisomerase I and II and
   activation of apoptosis [[63]6]. This mechanism also occurs in
   non-cancer cells, mediated by a p53-dependent signaling pathway
   [[64]10], and leads to cardiomyopathy. Dox-induced cardiotoxicity has
   also been ascribed to mitochondrial cytochrome c release and
   permeability transition pore opening [[65]21]. Longer exposures have
   been linked to decreased expression of mitochondrial proteins, and the
   redox signaling of the semiquinone exacerbates reactive oxygen species
   production further activating the apoptotic cascade [[66]21,[67]22].
   Furthermore, Dox can form complexes with cardiolipin, a mitochondrial
   inner membrane phospholipid important for allosteric regulation of
   mitochondrial enzymes [[68]2,[69]23] and intercalation with
   mitochondrial DNA [[70]11].

   Dox-induced thrombocytopenia and destruction of mature platelets
   [[71]6,[72]18,[73]21] suggests that they may be useful to monitor Dox
   effects on mitochondrial function. Platelets are enucleated cells
   derived from megakaryocytes and have a lifetime circulation of
   approximately 10 days in the blood [[74]24,[75]25]. Although platelets
   do not contain a nucleus, they do contain RNA transcripts and other
   RNAs acquired during circulation [[76]26] and contain mitochondria
   which are essential for platelet function [[77]27]. Interestingly, it
   is now becoming clear that platelets can serve as biomarkers for
   mitochondrial dysfunction and the metabolome is predictive of
   bioenergetic function [[78]28,[79]29]. Indeed, platelet bioenergetics
   have been shown to be potential biomarkers for the clinical severity of
   sickle cell disease, asthma, Alzheimer's and Parkinson's disease
   [[80][30], [81][31], [82][32]]. These findings raise the possibility
   that platelet bioenergetics can also serve to monitor the potential
   dose limiting toxicity of therapeutics, such as Dox, which have side
   effects mediated by their effects on metabolism.

   In a recent study combining the mitochondrial stress test (MST) with
   high-resolution metabolomics, we found that subsets of metabolites,
   including fatty acids and xenobiotics correlated with mitochondrial
   parameters, establishing platelets as a platform to integrate
   bioenergetics and metabolism for analysis of mitochondrial function in
   precision medicine [[83]29]. In the current study, we reasoned that the
   mitochondrial response to stress will be modulated by the activity of
   metabolic pathways prior to exposure to a therapeutic agent. If so,
   differences in the platelet metabolome between individuals could
   contribute to variability in the bioenergetic profiles of the intact
   platelet and therefore support use of platelet bioenergetic-metabolite
   interactome to predict and monitor therapeutic effects. This concept
   was addressed by measuring mitochondrial bioenergetic parameters among
   platelet donors in the presence and absence of Dox and integrating
   these data with non-targeted metabolomics. We found that the
   pre-treatment basal metabolome was correlated with maximal respiration,
   the response to Dox and its metabolism. These data provide the
   foundation for not only understanding Dox-mediated platelet toxicity
   but also contribute to precision medicine-based chemotherapies.

2. Materials and methods

2.1. Chemicals

   All reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA)
   unless otherwise specified. A mixture of internal standard stable
   isotopic chemicals [[84]33,[85]34] was purchased from Cambridge Isotope
   Laboratories, Inc. (Andover, Pennsylvania).

2.2. Platelet isolation

   Platelet concentrates collected from individual donors were obtained
   from the University of Alabama at Birmingham blood bank and used
   between days 6–8 after collection. Platelet activation during
   preparation is suppressed by the inclusion of Prostaglandin I2 and any
   potential activation was assessed by microscopy [[86]35]. In this study
   no platelet samples were excluded due to activation. Collection and use
   of these samples was approved by the University of Alabama at
   Birmingham Institutional Review Board (Protocol #X110718014). Platelets
   used for these studies were between day 6 and 8 after collection or
   freshly isolated as described previously [[87]35,[88]36]. In brief,
   platelets were pelleted by centrifuging at 1500 g for 10 min then
   washed with PBS containing prostaglandin I[2] (1 μg/ml) and platelet
   number was determined by turbidimetry [[89]37]. Prostaglandin I2 is
   included in the washing buffer to prevent activation during platelet
   preparation. The platelets for the experiments were isolated from 14
   individual donors and data are reported for all patients for
   metabolomics and for 13 for cellular bioenergetics due to the failure
   of one assay. Platelet aggregation using the 96-well plate reader was
   measured as previously described [[90]38].

2.3. Doxorubicin treatment

   For both bioenergetics and metabolomics studies, Dox stocks (20 mM in
   MilliQ water) were diluted to a 10-fold working solution in XF DMEM
   assay media (pH adjusted to 7.3). All assays described below were
   performed in parallel with each other. Vehicle controls was XF assay
   media alone.

2.4. Platelet bioenergetics and mitochondrial function

   The 96-well format Seahorse extracellular flux analyzer (Seahorse
   Bioscience, MA, USA) was used to measure bioenergetics [[91]36].
   Platelets were diluted to a concentration of 1 × 10^7 in 75 μl XF DMEM
   assay buffer (DMEM with 1 mM pyruvate, 5.5 mM d-glucose, 4 mM
   l-glutamine, pH 7.4) and were seeded onto Cell-Tak coated XF96
   microplates and the mitochondrial stress test was performed as
   described [[92]39]. The mitochondrial complex assay is performed using
   Plasma Membrane Permeabilizer (PMP) with injection of respiratory
   substrates with or without ADP [[93]40].

2.5. High-Resolution Metabolomics (HRM)

   For metabolic measurements, a protocol similar to the bioenergetic
   measurements was used [[94]29], and untargeted metabolomics was
   performed using previously established HRM methods [[95][41], [96][42],
   [97][43], [98][44]]. Washed platelets were diluted to a concentration
   of 1 × 10^8/well in 0.75 ml DMEM assay buffer (DMEM with 1 mM pyruvate,
   5.5 mM glucose, 4 mM glutamine, pH 7.4) and treated and were incubated
   with vehicle control or Dox (25 μM) for 3 h at 37 °C in a non-CO[2]
   humidified incubator. Platelets were then washed with cold PBS and the
   proteins precipitated using acetonitrile (50 μl) containing a mixture
   of stable isotope-labeled internal standard [[99]33,[100]34]. Pooled
   platelets (3 × 10^8 platelets from 3 wells) in 150 μl of acetonitrile
   containing internal standard were incubated on ice for 10–15 min and
   precipitated proteins removed by centrifugation for 10 min at
   13,000 rpm. The extraction protocol and the solvent are chosen to
   effectively clamp the metabolites which is verified by our previous
   study using oligomycin as a positive control [[101]29]. Samples were
   randomized to minimize effects of instrumental drift during analysis,
   and 10 μL aliquots were analyzed with three technical replicates using
   reverse-phase C[18] liquid chromatography (Targa C[18] 2.1  mm × 50  mm
   x 2.6 μm, Higgins Analytical) combined with a High-Field Q-Exactive
   mass spectrometer (Thermo Fisher). Mass spectral detection completed in
   negative mode electrospray ionization (ESI) at 120,000 (FHWM)
   resolution over a mass-to-charge ratio (m/z) range of 85–1250. A
   quality control pooled reference plasma sample (Q-Std3) was included at
   the beginning and end of each batch of 25 samples for quality control
   and quality assurance [[102]45]. Raw data files were extracted using
   apLCMSv6.3.3 [[103]46] with xMSanalyzerv2.0.7 [[104]47], followed by
   batch correction with ComBat [[105]48]. Uniquely detected ions
   consisted of m/z, retention time (RT) and ion abundance, are referred
   to as metabolic features.

2.6. Data processing and metabolic feature selection

   Prior to data analysis, triplicate injections were averaged and only
   m/z features with at least 80% non-missing values in either of the
   groups and more than 40% non-missing values across all samples were
   retained. After filtering based on missing values, data were log2
   transformed and quantile normalized [[106]49]. Selection of
   differentially expressed m/z features was performed based on one-way
   repeated measures ANOVA, using the limma package in R [[107]50]. For
   the metabolome-wide association study of Dox, metabolic features which
   were correlated with Dox were selected using linear regression analysis
   and transformed as described above. Benjamini-Hochberg false discovery
   method was used for multiple hypothesis testing corrections at a
   FDR<0.2 threshold [[108]51]. Visualization of the data, which was based
   on similarity in expression, was performed using unsupervised two-way
   hierarchical clustering analysis (HCA) utilizing the hclust() function
   in R to determine the clustering pattern of selected m/z features and
   samples. Principal component analysis (PCA) was performed using the
   pca() function implemented in R package pcaMethods.

2.7. Pathway enrichment analysis

   To evaluate metabolic alterations at a systems level metabolome-wide
   association analysis was performed for discriminatory metabolites at
   p < 0.05 and characterized for pathway enrichment using mummichog
   v1.0.10 software [[109]52]. For this analysis, features differing at
   p < 0.05 were selected to protect against type 2 error, and permutation
   testing (p < 0.05) was used in pathway enrichment analysis to protect
   against type 1 error [[110]53]. Pathways including minimum 4 matched
   metabolites in total size were selected and annotated using the
   criteria described below.

2.8. Metabolite annotation

   Metabolic features were annotated using xMSannotator [[111]54];
   confidence scores for annotation by xMSannotator are derived from a
   multistage clustering algorithm. Identities of selected metabolites
   were confirmed by co-elution relative to authentic standards and ion
   dissociation mass spectrometry (Level 1 identification by criteria of
   Schymanski et al. [[112]55]. Supplemental annotations were made based
   on high or medium confidence (≥4) with M-H adducts detected in the
   negative mode. Lower confidence annotations were made using KEGG,
   (Kyoto Encyclopedia of Genes and Genomes) [[113]56]; HMDB (Human
   Metabolome Database) [[114]57]; T3DB [[115]58], and Lipid Maps
   [[116]59] databases at 5 ppm tolerance.

2.9. xMWAS

   Bioenergetic and HRM data from the same set of samples were integrated
   by using xMWAS based on the sparse partial least-squares (sPLS)
   regression method for data integration [[117]60]. sPLS is a
   regression-based modeling approach which performs simultaneous variable
   selection and data integration, and is designed for problems where the
   sample size (n) is much smaller than the number of variables (p) and
   the variables are highly correlated [[118]61]. In addition, xMWAS
   performs community detection using the multilevel community detection
   algorithm [[119]62] to identify groups of nodes that are heavily
   connected with other nodes in the same community, but have sparse
   connections with the rest of the network. The input for xMWAS included
   the cellular bioenergetics (13 samples × 6 energetic parameters) and
   the metabolome (13 samples x 3240 metabolic features) for the vehicle
   control group and (13 samples × 6 energetic parameters) and the
   metabolome (13 samples x 3263 metabolic features) for the Dox treated
   group, which had been quantile normalized and log-transformed).
   Thresholds for determining significant associations met the correlation
   threshold criteria (|r| > 0.6) and p < 0.05 as determined by Student's
   t-test.

2.10. Bubble plots

   Generation of the bubble plot, a visualization tool for metabolic
   pathways found to be associated with bioenergetic parameters, was
   performed using the R script corrplot() [[120]63]. For these analyses,
   metabolic features which were previously found to be significant using
   the xMWAS correlation criteria, previously described in the xMWAS
   methods, were input into mummichog v1.0.10. Pathway analysis was
   performed for each bioenergetic parameter independently, and
   significant pathways were selected using the criteria previously
   described within the pathway enrichment methods. Both the size as well
   as the color of the bubble represent the pathway significance level
   based on the -Log[10]P value.

2.11. Statistical analysis

   The data reported in the metabolomics analyses are derived from
   platelets isolated from 13 or 14 different donors. Each platelet group
   was comprised of 3–5 technical replicates, and the data is presented as
   mean ± SEM. Statistical significance was determined using either a
   T-TEST or ANOVA with Tukey's post hoc test for data with more than 2
   groups, and p < 0.05 was considered significant. The linear correlation
   between multiple pairs of bioenergetic parameters were determined using
   the multivariate function of the JMP statistical program (JMP®, Version
   13, SAS Institute Inc., Cary, NC). A correlations (r-values) table that
   summarizes the strength of the linear relationships between each pair
   of bioenergetic parameters and a table with corresponding p-values were
   generated to identify significant dependencies between parameters. Data
   for |r| ≥0.4 and p ≤ 0.01 were considered significant.

3. Results

3.1. Effect of Dox treatment on platelets bioenergetics

   To assess the impact of Dox on bioenergetics platelets were pretreated
   for 3 h with increasing concentrations of Dox before the mitochondrial
   stress test (MST) was used to measure the oxygen consumption rate (OCR)
   and extracellular acidification rate (ECAR) [[121]64]. A representative
   trace for a single individual is shown in [122]Fig. 1A. The control
   platelets showed the expected response to oligomycin with a suppression
   of basal OCR which was reversed on addition of the uncoupler FCCP
   [[123]64]. Antimycin inhibited OCR below post-oligomycin levels and
   2-DG, an inhibitor of glycolysis, had no additional effect on OCR. Also
   shown in [124]Fig. 1A and B is the concentration dependent effect of
   Dox (0–25 μM) on intact platelet bioenergetics. The MST parameters were
   calculated to characterize cellular bioenergetics [[125]65] and showed
   ([126]Fig. 1B) that the lowest concentrations of Dox suppressed reserve
   capacity and maximal respiration and an effect on ATP linked
   respiration was apparent at the 25 μM concentration. This concentration
   is a reasonable approximation for the higher range in human subjects
   treated with Dox which have a reported C[max] of 11–19 μM [[127]66].
   Platelet ECAR or platelet aggregation was not significantly changed by
   Dox treatment (results not shown).

Fig. 1.

   [128]Fig. 1
   [129]Open in a new tab

   Doxorubicin inhibits mitochondrial respiration. (A) Platelets from a
   representative donor are shown with and without incubation with
   Doxorubicin (Dox) (0–25 μM) for 3 h after which a mitochondrial stress
   test was performed by first measuring basal OCR, followed by sequential
   injection of oligomycin (Oligo) (1 μg/ml), FCCP (0.6 μM), antimycin A
   (AntiA) (10 μM) and 2-deoxyglucose (50 mM). (B) The Dox concentration
   dependent inhibition of Basal, ATP linked, Maximal and Reserve
   Capacity. Data is represented as the mean ± SEM for 5–6 technical
   replicates for this donor. Data from 13 individual donors are shown for
   (C) basal (basal OCR – AA OCR), (D) ATP-linked (AL) (basal OCR –
   oligomycin OCR), (E) maximal (FCCP OCR – AA OCR), (F) reserve capacity
   (RC) (FCCP OCR – basal OCR) (G) proton leak (PL) (oligomycin OCR – AA
   OCR), calculated and are expressed as mean ± SEM with each dot
   representing the average data from a single donor, n = 5–6 replicates
   for each parameter. (H–K) Decrease (% of pre-Dox value) for each donor
   for basal (H), ATP linked (I) Maximal (J) and Reserve Capacity (K)
   after treatment with 25 μM Dox.

   The same protocol shown in [130]Fig. 1 was used to determine the
   average response to Dox in platelets from 13 individual donors and is
   shown for the 25 μM Dox concentration in [131]Fig. 1 (Panels C–G). All
   bioenergetic parameters were significantly decreased after 25 μM Dox
   treatment. We noted that both the pre and post-Dox treatment OCR values
   varied significantly between individuals. For example, the extent of
   Dox-dependent inhibition ranged from 30 to 70% decrease in basal OCR,
   30–75% decrease in ATP-linked OCR, 50–80% decrease in maximal OCR and
   as little as a 20% decrease to complete loss of reserve capacity among
   the 13 individuals tested ([132]Fig. 1H–K). Our previous studies
   suggested that the parameters from the mitochondrial stress test
   represent an integrated metabolic profile [[133]67] and this can be
   demonstrated using a multi-variate analysis for the parameters from the
   MST in the presence or absence of 25 μM Dox (see [134]Table 1 and
   [135]Fig. 2).

Table 1.

   Platelet bioenergetic parameters with and without Dox treatment.
   Basal AL PL Maximal RC NM
   Basal
   AL 0.9505*
   PL 0.4892 0.1945
   Maximal 0.7315* 0.7994* 0.0748
   RC 0.2584 0.3927 −0.2737 0.8477*
   NM
     __________________________________________________________________

   0.5060
     __________________________________________________________________

   0.4800
     __________________________________________________________________

   0.2532
     __________________________________________________________________

   0.1407
     __________________________________________________________________

   −0.1967
     __________________________________________________________________

   1.0000
     __________________________________________________________________

   Correlation between the platelet bioenergetic parameters following Dox
   (25 μM) treatment
     __________________________________________________________________

   Basal
     __________________________________________________________________

   AL
     __________________________________________________________________

   PL
     __________________________________________________________________

   Maximal
     __________________________________________________________________

   RC
     __________________________________________________________________

   NM
     __________________________________________________________________

   Basal
   AL 0.9465*
   PL 0.5790* 0.2852
   Maximal 0.8108* 0.7877* 0.4258
   RC 0.0719 0.1005 −0.0281 0.642*
   NM 0.3779 0.3204 0.3031 0.0649 −0.3818
   [136]Open in a new tab

   Table 1 provides the Pearson correlation coefficients (r values)
   calculated using the multivariate platform of JMP statistical program
   and shows the strength of the linear relationships between each pair of
   variables. * indicates p ≤ 0.05.

Fig. 2.

   [137]Fig. 2
   [138]Open in a new tab

   Relationships between bioenergetic parameters from the mitochondrial
   stress test (MST) with and without doxorubicin. Using the data shown in
   [139]Fig. 1 a correlation analysis was performed using the JMP
   statistical program to assess the linear relationships between the
   bioenergetic parameters derived from the MST with and without 25 μM
   Dox. The bivariate plots of basal OCR and ATP-linked OCR (A), reserve
   capacity and maximal OCR (B), basal OCR and maximal OCR (C), ATP-linked
   OCR and maximal OCR (D) and basal OCR and reserve capacity (E) are
   shown. The r values, the measure of strength of linear relationships
   and results of the significance tests for these data are reported in
   [140]Table 1.

   In untreated control platelets significant relationships were evident
   between basal vs ATP linked respiration, basal vs. maximal, ATP linked
   vs maximal, and maximal vs reserve capacity ([141]Fig. 2, [142]Table
   1). Basal vs reserve capacity OCR were not significantly related. Taken
   together these data confirm our previous findings that there is a wide
   normal range for platelet bioenergetics in healthy donors with several
   interesting features [[143]29]. After Dox treatment, these
   relationships were largely preserved, but with a lower range of
   activities, as shown graphically in [144]Fig. 2. Analysis of this data
   revealed that after Dox treatment the range of OCR for ATP linked and
   basal was narrowed (also evident from [145]Fig. 1) consistent with a
   lower capacity to generate ATP. Interestingly significant relationships
   exist between the Dox metabolite and the % change in the bioenergetic
   parameters in individual platelet samples ([146]Table 2). Dox and the
   Dox metabolite, Doxorubicinone) correlate positively with Basal and
   ATP-linked OCR whereas Reserve Capacity is negatively correlated with
   Dox levels. These data are consistent with a mechanism in which the
   Reserve Capacity is being used in response to the redox cycling of Dox.
   In addition, these data confirm our previous findings in other cell
   types that in response to oxidative stress, basal OCR increases with a
   corresponding decrease in Reserve Capacity [[147]67,[148]68]. Dox
   levels in the individual platelet samples ranged from 3.54 × 10^8 ±
   2.06 × 10^8 to 1.43 × 10^9 ± 1.1 × 10^8 and its aglycone metabolite
   doxorubicinone from 2.71 × 10^7 ± 9.15 × 10^5 to 1.73 × 10^8 ±
   3.83 × 10^7 which suggest that a significant variability exist between
   individuals in the ability to metabolize doxorubicin and is related to
   the susceptibility to mitochondrial damage.

Table 2.

   Relationships between bioenergetic parameters in the individual
   platelet samples prior to Dox treatment and the Dox metabolites.
   Correlations
     __________________________________________________________________

   Basal AL PL Max Res Cap NonMito
   Dox 0.6504 0.698 −0.0738 0.0518 −0.6783 0.2246
   Doxorubicinone 0.5413 0.5686 −0.3219 0.3002 −0.2234 0.2219
   


     __________________________________________________________________

   Correlation Probability
     __________________________________________________________________

   Basal
     __________________________________________________________________

   AL
     __________________________________________________________________

   PL
     __________________________________________________________________

   Max
     __________________________________________________________________

   Res Cap
     __________________________________________________________________

   NonMito
     __________________________________________________________________

   Dox 0.022 0.0116 0.8197 0.873 0.0153 0.4828
   Doxorubicinone 0.0692 0.0537 0.3076 0.3431 0.4851 0.4881
   [149]Open in a new tab

   Table 2 demonstrates the relationship between bioenergetic parameters
   and the unchanged Dox parent compound or Dox metabolites, such as
   doxorubicinone. The Pearson correlation coefficients were calculated
   using the JMP statistical program. * indicates p ≤ 0.05.

3.2. The effect of Dox treatment on the platelet metabolome

   Samples of platelets were prepared for untargeted metabolomics with and
   without exposure to 25 μM Dox for 3 h under the same conditions used
   for the bioenergetic measurements. 2831 features (distinct m/z values)
   were identified in both the control and Dox treated samples. Using the
   KEGG metabolic pathway mapping program a total of 48 pathway matches
   were found which included metabolites associated with platelet
   activation as well as fatty acid and arachidonic acid metabolism which
   are strongly associated with platelet function after Dox treatment
   ([150]Supplementary Table 1).

   Next, we examined metabolic responses of the platelets to Dox
   treatment. ANOVA on the 2831 features revealed 166 features were
   changed after Dox treatment (p < 0.05 at FDR of 0.2) and are presented
   by hierarchical clustering analysis-heat map ([151]Fig. 3A) and PCA
   plots ([152]Fig. 3B). Manhattan plots, based upon RT, m/z and abundance
   of metabolites show that of the 166 metabolites, 68 metabolites were
   higher and 98 metabolites were lower in Dox-treated platelets compared
   to the control group ([153]Fig. 3C,D). [154]Supplementary Table 2 shows
   detailed information on 68 annotated metabolites. To examine the
   metabolic pathways altered by Dox treatment, pathway enrichment
   analysis was performed using mummichog. The results showed that
   metabolites from the prostaglandin formation from both arachidonate, as
   well as CoA, TCA cycle, fatty acid, purine, cholesterol, and urea cycle
   pathways were enriched by Dox treatment ([155]Fig. 3E). The detailed
   information on metabolites associated with these pathways is provided
   in [156]Supplementary Table 3. These data are consistent with
   inhibition of bioenergetic function by Dox in intact platelets.
   Selected metabolites from these pathways are shown in [157]Fig. 4. For
   prostaglandin formation we observed that PGC1 (4A) was decreased, and
   similar observations were made for purine metabolite hypoxanthine (4B)
   which was also decreased. The CoA catabolism metabolite pantothenate
   (4C) was found to be increased after Dox treatment, which was also
   observed for acetoacetate (4D), and the TCA cycle metabolites,
   oxalosuccinate (4E) and cis-aconitate (4F). We also observed Dox (4G)
   and its metabolite doxorubicinone in the Dox treated group, showing
   that platelets metabolize Dox. We also found a number of metabolic
   features which were highly enriched in the Dox samples; however, many
   did not have known matches to public available databases, and are
   reported in [158]Table 3.

Fig. 3.

   [159]Fig. 3
   [160]Open in a new tab

   Metabolic separation of features after treatment to Dox in human
   platelets. (A) Unsupervised hierarchal clustering heatmap indicate that
   intensity of 166 features which drive the separation between vehicle
   control and Dox treated groups. Subject sub clusters 1 and 2 are
   indicated in the vehicle group. (B) PCA plot showing separation of the
   vehicle control group (shown in green) and BaP exposed group (shown in
   red), through the 1st and 2nd principal components. (C) Type I
   Manhattan plot of m/z features plotted against the –Log[10]P value.
   Shown in gray are the 2831 features identified after filtering and
   normalization. 166 features were found to be different between the two
   groups using the criteria (p < 0.05, FDR < 0.2) as indicated by the
   blue dotted line. Shown in red were features identified to be increased
   after Dox exposure (68/166) and in blue features which were decreased
   (98/166). (D) Type II Manhattan plot using time plotted against
   –Log[10]P value. (E) Pathway enrichment analysis of stored human
   platelets after Dox exposure compared to vehicle control. A total of
   (7/119) enriched pathways were determined (Filled gray bars indicate
   significance and the cutoff (p < 0.05) is indicated by the dotted
   line). (For interpretation of the references to color in this figure