Profiling lysophosphatidic acid levels in plasma from head and neck cancer patients

Chemicals and reagents

16:0 LPA, 17:0 LPA, 18:0 LPA, 18:1 LPA, 18:2 LPA and 20:4 LPA were purchased from Avanti Polar Lipids (Alabaster, AL, USA). HPLC-grade solvents were purchased from Fisher Scientific (Pittsburgh, PA, USA). Plasma for standards was obtained from the University of Malaya Medical Centre (UMMC) blood bank. Activated charcoal was purchased from R & M Chemicals (Essex, UK).

Plasma samples

Plasma samples of OSCC patients and control subjects (a total of 120 samples) were obtained from the Malaysian Oral Cancer Database and Tissue Bank System (MOCCTBS) and managed by the Oral Cancer Research and Coordinating Centre (OCRCC), University of Malaya. Plasma from control subjects were collected concurrently from healthy individuals with no history of cancer. A second set of plasma samples was collected from NPC patients and control subjects (a total of 98 samples) from a public hospital. All cancers were diagnosed by histopathological analysis of biopsy specimens. Information regarding the gender, age, risk factors and cancer stage are shown in Table 1. All participants provided written informed consent. Ethical approval for this study was obtained from the Medical Ethics Committee, Faculty of Malaya (DF OB1403/0009L) and the Medical Research Ethic Committee of the Ministry of Health Malaysia (KKM/NIHSEC/08/0804/P-12-182).

For sample collection, whole blood (three mL) was obtained preoperatively in EDTA tubes and transported on ice at 4 °C prior to processing. Samples were centrifuged at 1500 rpm for 15 min. Plasma was immediately aliquoted into micro-tubes (0.3 mL each tube) and stored at −80 °C until preparation for mass spectrometry analysis.

Sample preparation for LC-MS/MS

All LPA stocks and internal standard (IS; 17:0 LPA) were prepared in 50% (v/v) ethanol at a concentration of 1.0 mg/mL, which is equivalent to 1,000 parts per million (ppm). For the preparation of calibration curves for quantitative analysis, LPA stocks were diluted in blank plasma (without any endogenous LPA). Blank plasma was prepared by adding 10% (w/v) activated charcoal to plasma to deplete LPA. LPA depletion was monitored daily using LCMS. The activated charcoal was replaced with fresh activated charcoal at day 4. Complete depletion was achieved after 9 days. Blank plasma samples were also treated with three freeze/thaw cycles. The entire procedure was performed with continuous shaking on ice. The plasma-charcoal mixture was centrifuged at 16000 × g for 10 min and the supernatant was aspirated and filtered using a 0.22 µm nylon filter. The filtrate (blank plasma) was then used to prepare the LPA standards for the calibration curves.

A diluted working stock of 2000 ppb was finally prepared in blank plasma to prepare 100 ppb, 200 ppb, 350 ppb, 500 ppc, 700 ppb, 1000 ppb and 1500 ppb as the standard curve for quantification. Quality control (QC) samples were prepared at 300 ppb (QC1), 750 ppb (QC2) and 1350 ppb (QC3). Calibration standards and QC samples were prepared fresh daily. 17:0 LPA was added to all samples as an internal standard (IS) at a final concentration of 5ppm. Representative MRM chromatograms showing retention times of the individual LPAs added to blank plasma are shown in Fig. S1.

LPA extraction from plasma

LPA was extracted from the plasma of OSCC and NPC patients and controls, as described previously (Zhao & Xu, 2010) with minor modifications. LPA standards and QC samples were extracted using the same extraction method. Briefly, 10 µL of plasma were added to 300 µL of methanol (MeOH) together with 20 µL of 500 ppb of 17:0 LPA (non-naturally occurring LPA species) as the internal standard. Samples were vortexed and incubated on ice for 10 min and later centrifuged at 16000 × g for 10 min at room temperature. 180 µL of the supernatant were transferred to a fresh glass vial and subjected to LC-MS/MS analysis.

Liquid chromatography with tandem mass spectrometry (LCMS/MS)

The LPA species were separated by reversed-phase liquid chromatography using a C6 Phenyl analytical column (Gemini, particle size 5 µm, pore size 110 Å, 150 × 2.0 mm, Phenomenex, Torrance, CA, USA) and an injection volume of 10 µL with a 10 min separation time. The mobile phase consisted of H20 containing 0.01% NH3 and Acetonitrile: Methanol (50:50, v/v) and was delivered at a flow rate of 0.5 mL/min. The LC-MS/MS system comprised of an Applied Bioscience (AB) SCIEX QTRAP 5500 mass spectrometer, equipped with an ultra-fast liquid chromatography Shimadzu LC system and auto sampler (Shimadzu UFLC XR). The electrospray ion source was run in a negative ionization mode. Quantification was performed in the multiple reaction-monitoring (MRM) mode. Calibration standards and QC samples were extracted and quantitated daily together with approximately 20–30 plasma samples. Quantitative data measured against a linear standard curve was analysed with Analyst data acquisition system.

Method validation

The analysis method was validated according to the bioanalytical method validation procedures outlined by the USFDA (FDA, 2018). A set of five replicates of each QC (QC1, QC2 and QC3) were prepared and analysed each day to determine the intra-day or inter-day precision and accuracy of the method. These analyses were repeated daily over three days using freshly prepared calibration curves. Freeze and thaw stability was tested with QC1 and QC3 where all samples were kept frozen at −40 °C for 25 mins and thawed at room temperature for three cycles. Analysis was carried out for every cycle. Sample stability was tested after 4 and 6 h at room temperature and at 4 °C. Post-preparative auto sampler stability was determined by reinjection of QC1 and QC3 after 4 and 6 h following the initial injection. Long term stability of QC samples was analysed after 48 days. Intra-day accuracy and precision were calculated from % bias (% (measured –theoretical)/theoretical concentration] and relative standard deviation [%RSD = % standard deviation/mean] respectively for the QC points. Inter-day accuracy was calculated similarly for each QC points from 3 validation runs. Absolute recovery was calculated by comparing the peak area of QC samples in plasma to the peak area of samples in untreated neat solution (methanol).

Statistical analyses

Bivariate analysis was carried out in order to understand any association between two variables (LPA species). Box plots describe the average values of median, minimum, maximum and the variability of the datasets using the first and third quartiles. The Kruskall–Wallis test was used in testing the differences in LPA plasma concentrations between controls, OSCC (early and advanced stage) and NPC. Correlation between pairwise LPA species was shown using scatterplots. Visual presentation using heat maps was drawn to show how all LPA values varied across all samples. Bivariate analyses were performed using the R programming language. Multivariate analysis (Principal Component Analysis (PCA) was performed using the SIMCA-P v12 (UMETRICS) software. PCA was performed on all five LPA species for all 218 samples from the two groups.

Levels and profile of LPA species in plasma from OSCC and NPC patients

To examine possible variations in individual LPA levels, LPA species were quantified in plasma samples from OSCC and NPC patients together with control subjects using LCMS/MS. Total LPA levels in control plasma were higher than those in plasma from cancer patients for both OSCC (Table 2) and NPC (Table 3). The levels of individual LPA species (16:0, 18:0, 18:1, 18:2 and 20:4) in plasma from cancer patients were compared to those in healthy control individuals using boxplots and are shown in Fig. 1 (OSCC) and Fig. 2 (NPC). For OSCC, the levels of three individual LPAs (18:1 LPA, 18:2 LPA and 20:4 LPA) were significantly different between the three groups (normal, early and advanced) (p < 0.05; Fig. 1). Post hoc tests revealed that both 18:1 LPA and 18:2 LPA levels were significantly lower in the early and advanced cancer groups compared with normal controls and 20:4 LPA levels were significantly lower in the early OSCC group than in the controls (Table 4). The levels of all LPA species examined were lower in plasma from NPC patients compared with the controls (p < 0.05). (Fig. 2; Table 3). In the present study, the relative abundance of the LPA species in the normal plasma samples was similar to previous reports, with the order of abundance being LPA 18:2>16:0>18:0>18:1>20:4 (Baker et al., 2002; Shan et al., 2008). Interestingly, in both cancer groups the relative order of abundance for both cancer types was different from the controls with LPA 16:0>18:0>18:2>18:1>20:4.

The distributions of LPA in plasma samples using normalized values of all five LPA concentrations were highlighted and visualised by the construction of Heatmaps (Fig. 3). For control oral samples, the majority of samples demonstrated LPA values similar or close to the normal median value (0.8<X<1.2). For both early and advanced OSCC groups, a different profile was observed with a subset of samples for which LPA concentrations were generally lower than in normal samples. Similar results were obtained for NPC, with all LPA concentrations being generally lower in most of the tumours.

Correlation between individual LPA species

In order to study the relationship between each pair of LPA species, correlation studies were conducted using Pearson correlation tests (Table 5). For OSCC, strong to moderate positive correlations were observed between the three groups (control, early and advanced) for all LPA species (Table 5). A similar result was observed in the NPC samples with strong positive correlations observed between the two sample groups (control, tumour) for all LPA species (Table 6). To further investigate these correlations, scatterplots were generated for all possible pairings of the five LPA species measured. The results for the OSCC samples are shown in Fig. 4. For three pairs of LPA species (18:2 LPA paired with 16:0 LPA, 18:0 LPA or 18:1 LPA), the data points for normal samples were clustered together allowing possible differentiation between control and cancer groups (both early and advanced). For NPC samples, clustering of control samples was observed in all plots (Fig. 5).

Clustering of samples by all LPA values

PCA was performed to further evaluate the separation between normal and OSCC samples according to their LPA values. The concentrations of five LPA species (16:0 LPA, 18:0 LPA, 18:1 LPA, 18:2 LPA and 20:4 LPA) were considered in this analysis. In PCA analysis, total percentage of PCA variation between samples of the first two principal components was 94.0% with values of 75.3% and 18.7% for the respective components (Fig. 6A). The PCA score plot of these first two components provided a good separation of the normal samples from both early and advanced tumour samples. The loading plot graph shows the coefficients of each variable for the first component versus the second component. This identifies which variables have the largest effect on each component. A scrutiny of the PCA loading plot, showed that all LPA have positive loadings on component 1 (x-axis) with the strongest influence from 16:0 LPA and 18:0 LPA. As for component 2 (y-axis), 18:2 LPA has the highest influence among the variables.

In NPC samples (Fig. 6B), the total percentage of PCA variation between samples of the first two principal components was 96.5% with values of 89.9% and 6.6% for the respective components. The PCA score plot showed that these two components provided a good separation between control and tumour samples. Similar to OSCC, the NPC showed that all LPA have positive loadings on component 1 (x-axis) with the strongest influence from 16:0 LPA and 18:2 LPA. As for component 2 (y-axis), 18:2 LPA has the highest influence among the variables.