Article pubs.acs.org/molecularpharmaceutics
Potential Markers of Cisplatin Treatment Response Unveiled by NMR Metabolomics of Human Lung Cells I. F. Duarte,*,† A. F. Ladeirinha,‡ I. Lamego,† A. M. Gil,† L. Carvalho,§,∥,⊥ I. M. Carreira,‡,⊥ and J. B. Melo‡,⊥ †
CICECO, Department of Chemistry, University of Aveiro, Campus Universitário de Santiago, 3810-193 Aveiro, Portugal Laboratory of Cytogenetics and Genomics, Faculty of Medicine, University of Coimbra, 3000 Coimbra, Portugal § University Hospitals of Coimbra, 3000-075 Coimbra, Portugal ∥ Institute of Pathological Anatomy, Faculty of Medicine, University of Coimbra, 3000 Coimbra, Portugal ⊥ CIMAGO, Faculty of Medicine, University of Coimbra, 3000 Coimbra, Portugal ‡
S Supporting Information *
ABSTRACT: In this work, 1H high resolution magic angle spinning (HRMAS) nuclear magnetic resonance (NMR) spectroscopy was used to characterize the variations in the metabolome (small metabolites and mobile lipids) of A549 human lung cells in response to exposure to the alkylating drug cisplatin. Multivariate analysis and signal integration of spectral data were carried out to unveil exposure-induced effects and follow their time course. Parallel and strongly correlated increases in lipids (particularly unsaturated triglycerides) and nucleotide sugars (particularly uridine diphosphate Nacetylglucosamine) were found in cisplatin-treated cells, highlighting these compounds as potential biomarkers of treatment response. Other significant changes upon drug exposure comprised an increase in sorbitol and decreases in niacinamide and several amino acids (glutamine, alanine, lysine, methionine, citrulline, phenylalanine and tyrosine). These results show that in vitro NMR metabolomics is a powerful tool for detecting variations in a range of intracellular compounds upon drug exposure, thus offering the possibility of identifying candidate metabolite markers for in vivo monitoring of tumor responsiveness to treatment. KEYWORDS: A549 lung cells, cisplatin, HRMAS NMR, metabolic profile, metabolomics
■
INTRODUCTION The development of sensitive approaches to effectively monitor the response to chemotherapy is critical for the timely adjustment of therapeutic strategies and the improvement of clinical outcomes in cancer treatment. Based on the premise that the interaction of drugs with cells consistently affects metabolism, thus leading to altered levels of endogenous metabolites, metabolic profiling (or metabolomics) has the potential to unveil new end point markers of drug toxicity and/ or efficacy.1,2 In this context, metabolic profiling of cultured cells is a particularly useful approach.3,4 Indeed, in spite of limitations such as oversimplicity and dosimetry mismatch compared to in vivo conditions, in vitro cell studies allow direct relationships between cells’ metabolic variations and drug exposure conditions (e.g., treatment duration and dose) to be established, without the confounding influences present in animal models or in humans. The knowledge of such relationships may contribute to a better understanding of drugs’ mechanisms of action and pave the way for metabolites (identified in vitro) to be explored as in vivo markers of therapy response. Proton nuclear magnetic resonance (1H NMR) © XXXX American Chemical Society
spectroscopy is one of the most useful techniques to characterize the metabolome of cultured cells, as it allows the rapid, nondestructive screening of major metabolites involved in key metabolic pathways, with high analytical reproducibility.5−7 Several studies have demonstrated that the NMRdetected intracellular composition is exquisitely sensitive to cell phenotypes such as drug resistance, as well as to exposure to different drugs. In particular, NMR metabolic profiling of several types of mammalian cells has been applied to study the pharmacodynamic behavior of a number of cytotoxic and cytostatic compounds, including alkylating agents (e.g., cisplatin),8−12 antitumor antibiotics (e.g., doxorubicin),13,14 mitotic inhibitors (e.g., paclitaxel)15,16 and several molecularly targeted agents (e.g., imatinib).17,18 In those studies, significant dose- and time-dependent effects have been unveiled in a range Received: June 7, 2013 Revised: September 18, 2013 Accepted: September 19, 2013
A
dx.doi.org/10.1021/mp400335k | Mol. Pharmaceutics XXXX, XXX, XXX−XXX
Molecular Pharmaceutics
Article
dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) salt to formazan, as measured spectrophotometrically at 570 nm.21,22 In brief, MTT was dissolved in sodium medium (pH 7.4) at 5 mg/mL and incubated with the cells for 2 h at 37 °C. After this period an equal volume of acid isopropanol (0.04 M HCl in isopropanol) was added to each well and mixed until all the formazan crystals were dissolved. MTT reduction was expressed as a percentage of control cells’ absorbance. Cell Sampling for NMR. Cell pellets were suspended in 1 mL of PBS/D2O (0.14 M NaCl, 0.0027 M KCl, 0.0015 M KH2PO4, 0.0081 M Na2HPO4 in deuterated water, pH 7.4), centrifuged and resuspended in 40 μL of PBS/D2O, to which 5 μL of PBS/D2O containing 3-(trimethylsilyl)propionic-2,2,3,3d4 acid (TSP) 0.25% (chemical shift referencing and shimming) was added. Then, a 3-fold cycle of liquid nitrogen dipping and ultrasonication was performed to obtain a suspension of lysed cells. Each sample was transferred to a NMR disposable insert with sealing cap and stored at −80 °C until analysis. NMR Measurements. The packed disposable inserts were placed in standard 4 mm MAS rotors and spun at the magic angle at 4 kHz. A Bruker Avance spectrometer operating at 500 MHz for 1H observation and a 4 mm HRMAS probe were used. Standard 1D spectra with water presaturation (pulse program “noesypr1d”, Bruker library) were acquired at 296 K with a 6510 Hz spectral width, 32 K data points, a 4 s relaxation delay (d1), 100 ms mixing time (d8) and 256 scans (4 dummy scans). All 1D spectra were processed with a 0.3 Hz line broadening, zero filling to 64 K data points, manual phase and baseline correction. Chemical shifts were referenced internally to the alanine signal at δ 1.48 ppm (as this provided better spectral alignment than calibration by the TSP signal). Spectral assignment was based on 2D total correlation spectroscopy (TOCSY) and heteronuclear single quantum coherence (HSQC) spectra and consultation of spectral databases (Bruker Biorefcode database and the human metabolome database, HMDB)23 and further supported by statistical total correlation spectroscopy (STOCSY)24 (MATLAB version 7.12.0, The MathWorks Inc.). Data Processing and Multivariate Analysis. Spectra were aligned using recursive segmentwise peak alignment25 to minimize chemical shift variations and normalized by probabilistic quotient normalization (PQN)26 to account for dilution-independent effects on spectral area (MATLAB version 7.12.0, The MathWorks Inc.). Normalized data were scaled by either unit variance (UV) or Pareto scaling (SIMCAP 11.5). Principal component analysis (PCA) and partial-leastsquares discriminant analysis (PLS-DA) were applied to the 1H NMR spectra (δ 9.1−0.25 ppm, water region excluded, 42641 variables) using SIMCA-P 11.5 (Umetrics, Sweden), with a default 7-fold internal cross validation, from which Q2 and R2 values, respectively reflecting predictive capability and explained variance, were extracted. In order to estimate the relative variations of individual compounds over exposure time, selected signals in the quotient normalized 1D spectra were integrated using the Amix-Viewer software (Bruker, version 3.9). For each metabolite, the statistical significance of the difference between the means of the two groups (control and treated) was assessed using the two-sample t test or the nonparametric analogue Wilcoxon rank sum test with continuity correction. A p-value 0.9) (Figure S2a, Supporting Information). Moreover, these resonances showed strong correlations with lipid signals, as will be discussed later. STOCSY was also useful to confirm the assignment of the relatively intense signals in the δ 3.60−3.86 ppm range to sorbitol (Figure S2b, Supporting Information), the major polyalcohol detected, which appeared elevated in response to drug treatment. Furthermore, visual comparison of the spectra in Figure 1 suggests that, compared to control cells, treated cells presented decreased levels of several amino acids (e.g., alanine, lysine, glutamine, tyrosine, phenylalanine) and of lower abundance metabolites detected at low-field, namely, uracil and niacinamide. Time Course of Metabolic Variations. In order to screen for differences between control and treated cells at all time points (12, 18, 24, 36 and 48 h), multivariate analysis was applied to the corresponding spectra (n = 37). In both PCA and PLS-DA scores scatter plots (Figures 2a and 2b, C
dx.doi.org/10.1021/mp400335k | Mol. Pharmaceutics XXXX, XXX, XXX−XXX
Molecular Pharmaceutics
Article
Figure 2. Multivariate analysis of 1H HRMAS NMR spectra from control and cisplatin-treated cells (Pareto scaled): (a) PC1 vs PC2 scores scatter plot obtained by principal component analysis (PCA), (b) LV1 vs LV2 scores scatter plot obtained by partial-least-squares discriminant analysis (PLS-DA) (2 LVs, R2X 0.723, R2Y 0.703, Q2 0.56), (c) PLS-DA LV1 loadings profile colored according to variable importance in the projection (VIP).
Table 1. List of Metabolites/Metabolic Features Altered in Cisplatin-Treated Cells Compared to Control Cellsa variation (%) relative to control cells metabolic feature cholesterol total TG f.a. chain length u.f.a. PTC GPC UDP-GlcNAc UDP-GalNAc sorbitol niacinamide glutamine alanine lysine methionine phenylalanine tyrosine citrulline
NMR signals (δ, ppm)b
24 h
36 h
48 h
C18 (0.71) CH3 (0.89) CHOR (5.22) (CH2)n (1.28) to CH3 (0.89) CHCH (5.33) to CH3 (0.89) N(CH3)3 (3.26) N(CH3)3 (3.23) Glc-C1H (5.52) Gal-C1H (5.55) CH2OH (3.82) pyridine C4H (8.94) γCH2 (2.45) βCH3 (1.48) δCH2 (1.72) γCH2 (2.64) ring (7.32,7.38,7.43) ring C3,5H (6.89) δCH2 (3.14)
+31.9 ± 5.3* +42.3 ± 5.2** +45.2 ± 6.0** +48.5 ± 4.9** +39.3 ± 0.7** −9.7 ± 8.3 +65.1 ± 35.8 +31.4 ± 10.4 +12.3 ± 4.6 c −8.3 ± 4.2 c c −7.2 ± 2.3 −9.8 ± 3.5 −7.0 ± 5.4 −10.3 ± 6.6 −15.1 ± 4.8*
+42.5 ± 7.0 +64.2 ± 13.2** +89.1 ± 15.1 +53.0 ± 7.3 +52.4 ± 5.1** −13.5 ± 8.8 +91.8 ± 10.7* +51.9 ± 11.3* +14.6 ± 7.2 +11.1 ± 5.2 −11.2 ± 7.7 −11.1 ± 5.1 c −9.0 ± 1.9* −8.0 ± 4.1 c c −19.8 ± 6.4
+42.9 ± 13.1 +57.2 ± 20.7 +59.6 ± 24.0 +40.7 ± 7.8** +51.5 ± 7.7** −20.6 ± 9.4 +52.8 ± 8.2* +95.4 ± 13.2* +91.8 ± 25.4 +29.7 ± 6.6 −29.4 ± 7.3** −17.0 ± 6.8* −22.5 ± 5.1* −19.3 ± 2.3** −13.3 ± 3.9* −26.3 ± 7.0** −36.2 ± 8.0** −30.4 ± 6.9**
*p < 0.05, **p < 0.01. Abbreviations: u.f.a. unsaturated fatty acids, PTC phosphatidylcholine, GPC glycerophosphocholine, UDP-Glc-NAc uridine diphosphate N-acetylglucosamine, UDP-Gal-NAc uridine diphosphate N-acetylgalactosamine. bNMR signals selected to represent each metabolite or metabolic feature. cVariation
