NanoLC MALDI MS/MS based quantitative metabolomics reveals the alteration of membrane biogenesis in oral cancer

Abstract

Cancer cells use aberrant metabolic process for proliferation and metastasis. High-throughput separation and quantification of metabolites from bio-samples is crucial in this aspect. Matrix Assisted Laser Desorption Ionization Mass Spectrometry (MALDI MS), a prime technique used for metabolite identifications, suffers from several limitations in quantification studies. In this study, we have established a novel approach using conventional nanoLC-MALDI MS/MS interface for separation, identification and label free quantifications of the metabolites from biopsies. Quantification of metabolite using MALDI requires a homogeneous distribution of the matrix mixed samples on target plate for proportional time-of-flight (ToF) with the concentration of a particular metabolite present in that sample under same laser intensity. Here, crude metabolites extracted from cancer biopsies are separated and eluted as ‘fraction spots’ on MALDI target plate using nanoLC. Introducing a novel parameter, %aAUC (percentage of average area under the curve), comparative quantification of separated metabolites isolated from normal, pre-cancer and cancer biopsies has been explored. Such comparative quantification is only possible with multiple AUCs of a particular metabolite obtained from this nanoLC-MALDI MS approach. Further, selected metabolite peaks have been analyzed through MS/MS fragmentation. This approach was validated using known concentrations of an internal standard (thiourea) with its corresponding aAUC (average area under the curve). Interestingly, such comparative quantification has revealed a significant change in expression of crucial lipid metabotypes like triglyceride, phospatidyl inositol, phosphatidyl choline, glycerophospholipid, cytidine diphosphate-diacylglycerol and phosphatidylinositol bisphosphate indicating altered lipid metabolism associated membrane biogenesis in oral pre-cancer (oral submucous fibrosis) and cancer. Hence, our study successfully established an altered membrane biogenesis in oral cancer using a novel and distinct approach for separation, identification and label-free quantification of metabolites in fast growing metabotype-biomarker identification era.

---

Introduction

Reprogramming of metabolism, a crucial phenomenon in cancer cells, leads to altered metabolite production of different regulatory and cellular processes.^1,2 Information on the overview of altered metabolites production in oral cancer is of key importance for the analysis of molecular basis of cancer metastasis and theragnosis. Unfortunately, very limited information are available till date on oral cancer metabolomics whereas genomic and proteomic profiling are well characterized.^3–5 Previous protein and gene level study in oral cancer has revealed a number of novel biomarkers which have added a great value in clinicopathological diagnosis of oral cancer patients.^6,7 In addition to proteomics and genomics, study of cancer metabolomics has proved a great assurance in understanding altered cell metabolism leading to improved theragnosis in oral cancers.^8,9 Poinsignon et al., 2016 reported the DATAN derivatization based quantification of L and D 2-hydroxyglutaric acid in biological fluid (serum and plasma) using HPLC-ESI-MS/MS method.¹⁰ Again Uchiyama et al., 2012 studied the distribution of phophatidyl choline in cancer and stromal region using imaging mass spectrometry and MS/MS approach in oral squamous cell carcinoma.¹¹ But because of technical difficulties in instrument capability, sensitivity, lacking of sophisticated method and complicated data handling, metabolomics remain a highly challenging and evolving area.^9,12,13 In this connection, Wang et al., 2003 reported a quantification technique of metabolites and protein using the mono-isotopic peak intensity of the corresponding metabolites.¹⁴ Again 43 intra-cellular metabolites and 32 extra-cellular metabolites in protozoa parasite T. brucei. were quantified using LC-MS based ¹³C labeling approach by Kim et al., 2015.¹⁵ Another study reported the down-regulation of 16 metabolites (valine, alanine, proline, inositol mono-phosphate, anthranillic acid, citrulline etc.) and up-regulation of 5 metabolites (glutathione, phenyl alanine, 5-adenosine methionine quantification) in stationary phase of S. enterica by HPLC based isotope ratio (¹²C and ¹³C) approach.¹⁶ Although numbers of analytical methods such as NMR, MS coupled with HPLC or GC are used as key techniques in metabolomics, comparative metabolite quantification between normal and cancer tissues remains as a gap in cancer metabonomics.^17–21

Matrix assisted laser desorption ionization (MALDI) mass spectrometry (MS) technique is a prime technique involved in proteomics, peptidomics as well as metabolomics study from very long time.^22,23 In addition to LC/MS and GC/MS techniques MALDI MS is gradually being used as a platform for non-targeted metabolomics.²⁴ But with respect to quantification of different analytes using MALDI faces big challenges till now.^12,13 To facilitate the metabolite identification and quantification study different commercially available advanced instrument like quadrupole-time-of-flight (QTOF), nano liquid chromatography (nanoLC), gas chromatography (GC) are connected additionally with different mass analyser for the online separation of crude sample and to give spectra at lower range which are the possible range for different metabolites and lipid molecules.^24,25 In Liu et al., 2012 has showed that nanoLC/MS is much more sensitive than HPLC/MS in metabolites identification studies.²⁶ Nano liquid chromatography (nanoLC) coupled MALDI allows to separate sample components such as metabolites and small molecules with the help of C₁₈ reverse phase columns prior to MALDI MS or MS/MS analysis.²⁷ But proper systematic metabolic quantification through MALDI in normal and cancer tissues remains a challenge in cancer theragnosis. Explication of cancer metabonomics has been continuing as a hurdle from 1920's Warburg effect to modern era in molecular biology. The plasticity of cancer cell metabolic pathway in totality has not yet been fully explored.^1,28 Aerobic glycolysis through increased lactate production, Wargburg effect in cancer cell is a well known phenomenon.²⁹ But, recently, dual/reverse Wargburg effect was extensively reviewed.^30,31 Our previous NMR based study also indicated the minimal Warburg effect with up-regulation of fatty acid synthesis and altered choline metabolism in oral cancer.³² The study elucidated an altered metabolic pathway in cancer by qualitative assessment and identification of NMR peak from HMDB database. Evaluation of altered metabolic pathway in cancer, comparative quantification of the metabolites amongst different cellular aberrancies is very crucial. Spectral regions obtained from conventional spectroscopy (NMR, MS, FTIR) are generally overlapped due to heavy crowds of metabolites present in crude biological samples.¹⁸ From such spectral pattern, it is difficult to perform the comparative quantification of a particular metabolic peak present in different study groups. To address this problem, we have proposed a nanoLC method for preliminary separation of metabolite from a mixture followed by comparative quantification of the metabolites. Subsequently, a novel quantitative approach by introducing a new parameter, percentage of average area under the curve (%aAUC) using MALDI MS has been adopted.

Hence, the present study successfully established a label free quantification technique for non-targeted metabolomics of oral pre-cancer and cancer through nano-LC MALDI MS/MS. Such label-free quantification approach reveals a remarkable alteration in lipid metabolic pathway in oral cancer. Importantly, this quantitative metabolomic approach can lead the way to move mass spectrometry to the clinic for better pre-cancer and cancer molecular diagnosis.

Materials and methods

Sample collection

Total 45 oral biopsy samples having 15 numbers from each study groups [normal oral mucosa (NOM), oral submucous fibrosis with dysplasia (OSFWD) and oral squamous cell carcinoma (OSCC)] were collected from Department of Oral and Maxillofacial Pathology, Guru Nanak Institute of Dental Science and Research (GNIDSR), Kolkata, India under the informed consent of patients with the approved ethical clearance (ref. no. GNIDSR/IEC/ECC/2015/02). Control and disease states samples were histopathologically (gold standard haematoxylin and eosin staining) confirmed by expert oral onco-pathologists. Selection criterions of the patients were (i) clinico-pathological signs and symptoms of OSF (presence of fibrotic bands with firm coarse oral mucosa as well as partial trismus), (ii) addiction in betel quid, pan masala, areca nut, tobacco smoking etc. (iii) Co-morbid cases or having cancer on other anatomical sites were excluded. The light microscopic attributes such as nuclear pleomorphism of the surface epithelial cells, nuclear hyperchromatia, increased and abnormal mitosis, dyskeratosis were considered for the confirmation of dysplastic OSF. On the other hand, non dysplastic OSF samples were confirmed by the presence of atrophic surface epithelium with hyalinized avascular sub-epithelial connective tissue etc. Presence of dysplastic epithelium and its active invasion to the sub-epithelial connective tissue having concomitant formation of neoplastic islands confirmed the OSCC samples.

Hematoxylin and eosin staining

Tissue samples were fixed in phosphate buffered formalin (10%) and subsequently microtomed to obtain 4–5 μm thick paraffin sections. Tissue sections were placed on albumin-coated (chicken-egg) glass slide. Slides were then de-paraffinated by 15–20 min of xylene treatment. Harris' hematoxylin (cat. no. AG2AF62372, Merck, Mumbai, India) and counter-stained with eosin-yellowish (cat. no. MI7M572117, Merck, Mumbai, India) were used for staining the de-paraffinated sections.

Metabolite extraction from tissue samples

Metabolites from normal, pre-cancer (OSFWD) and cancer (OSCC) tissue samples were isolated according to Want et al., 2012.³³ Briefly, tissue samples (60 mg) were taken from liquid nitrogen container and crushed using pre-processed and pre-chilled pestle and motor by adding 1.5 ml of pre-chilled methanol/water (1 : 1). The tissue lysate was centrifuged at 16 000g for 10 min. The solid precipitate were then mixed with 1.6 ml of pre-chilled dichloromethane/methanol (3 : 1) and subsequently centrifuged at 16 000g for 10 min. Supernatant was dried in fume hood.

Sample preparation for nanoLC and MALDI

Dried samples were dissolved in 120 μl of methanol/water (1 : 1). Two micro liter of this sample was mixed with 2 μl 2,5-dihydroxybenzoic acid (DHB) matrix (10 mg per ml of TA-30 solution (3 : 7 v/v acetonitrile: 0.1% v/v trifluoroacetic acid in H₂O) and spotted on MTP 384 ground steel (Bruker Daltonics, Geramany) MALDI plate and allowed to stand for 30 min at room temperature for proper crystallization prior to direct MALDI data acquisition. Furthermore, 12 μl of this solution was used for nanoLC separation.

Method development for chromatographic separation of metabolites through nanoLC-MALDI system

Metabolites separation and quantification were performed using an automated system composed of two separate devices (Table 1). Auto sampler (Bruker, Germany), a nano scale liquid chromatographic system containing a C₁₈ nano column (Magic C₁₈ AQ, 200 Å, 3 μm, Michrom) was used for separation. The automated spotting system (Proteineer fc II, Bruker, Germany) which enables the mixing of proper matrix with elute from nanoLC followed by spotting on MALDI target plate (MTP AnchorChip™ 384BC). The MALDI spotter was first calibrated against the spot position of the MTP target plate for equipped transponder antennas associated automated target identification and sample tracking (Bruker, Germany). Spotter was also programmed to take 0.25 min for sample deposition on each spot. Chromatographic method development and the related experimental details have been depicted in Table 1. NanoLC was equipped with MALDI TOF/TOF MS system (Bruker, Germany). 12 μl sample was loaded onto C₁₈ nano column using a nano pump. Elution was performed at a flow rate of 600 nL min⁻¹ for 55.02 minute using a nano pump and an auto-sampler (Proteineer fc II, Bruker Germany). Valve positions were switched on for direct connection between gradient mixer and column. Linear gradient was set to 37 min for 5% to 45% acetonitrile. Data acquisition was performed using HyStar 3.2 software (Bruker Daltonics, Germany).

Table 1 Experimental details

Instrument	NanoLC MALDI MS/MS
Column specification	Magic C18 AQ 3 μm 200 Å (0.1 × 150 mm), Bruker, Germany
Autosampler specification	Proteineer fc II, Bruker Germany
Nano-LC MALDI plate	MTP AnchorChip™ 384BC
Mobile phases	Mobile phase A: 0.1% TFA in LC-MS grade water
	Mobile phase B: 0.1% TFA in acetonitrile
	Mobile phase C: 0.1% TFA in LC-MS grade water
Injection volume	12 μl
Pressure	Minimum = 0 bar, maximum = 550 bar
Flow rate	600 nL min^−1
LC time program/method development [time (min) : conc. of mobile phase B (%)]	[0.00 : 5.00], [7.00 : 5.00], [37.00 : 45.00], [38.00 : 100.00], [47.00 : 100.00], [48.00 : 5.00], [55.02 : 5.00]

---

Mass spectrometry and data acquisition

MALDI TOF MS was performed using Ultraflextreme mass-spectrometer (Bruker Daltonics, Germany). Data acquisition in the positive-ion reflection mode by the smart beam laser (Bruker, Germany) was performed using FlexControl 2.0 software. Resolving power of mass spectrometer was 40 000 and mass accuracy was less than 5 ppm. Laser repetition rate for MS/MS condition was 1000 Hz. Data analysis was performed by FlexAnalysis 2.0, BioTools 2.3 (Bruker) and metabolites were identified using Human Metabolome Database (HMDB) (http://www.hmdb.ca/).

aAUC standard curve preparation using internal standard

To establish the aAUC based quantification of metabolite using MALDI AnchorChip technology a standard curve is prepared. Different concentration of an internal standard (thiourea, m/z 77.7 Da) was mixed with DHB matrix and subsequently spotted into the hydrophilic core surrounded by hydrophobic ring of the AnchorChip targeted MALDI plate as well as on the normal MALDI targeted plate, Bruker (Fig. 1). Thiourea (MERCK, cat. no. QGIQ610875) was dissolved in MS grade water to prepare 2, 4, 6, 8, 10 and 12 μg μl⁻¹ concentrations. Two micro liter of each concentration of internal standard was mixed with 2 μl of DHB matrix solution [10 mg DHB matrix per ml of TA-30 solution (3 : 7 v/v acetonitrile: 0.1% v/v trifluoroacetic acid in H₂O)] and subsequently spotted into the hydrophilic core of the plate for homogeneous metabolite-matrix crystal distribution. Since the hydrophilic core is surrounded by a hydrophobic ring it will prevent the sample spreading and guide the sample to remain into the hydrophilic core for proper metabolite-matrix crystal distribution. Each concentration was spotted in triplicate to obtain the corresponding aAUC. Data acquisition was using FlexControl 2.0 software v and aAUCs were calculated using FlexAnalysis 2.0 software. The standard curve was prepared by plotting concentration (μg μl⁻¹) on X-axis and aAUC on Y-axis to establish the relationship between aAUC and corresponding concentration of metabolite.

Fig. 1 (A) Upper panel shows heterogeneous distribution of metabolite-matrix crystals in normal MALDI targeted plate; lower panel shows the same without sample (B) upper panel shows homogeneous distribution of metabolite-matrix crystals on AnchorChip targeted MALDI plate; lower panel shows the hydrophilic cores at the centre without sample.

Results

Diagnostic confirmation of the biopsy specimen

Confirmation of tissue conditions were performed by gold standard haematoxylin and eosin staining as well as clinical diagnostic acumen of expert onco-pathologist. The histopathological attributes like, presence of inflammatory cells in sub-mucosal region, shape of reti-ridges and abnormal nuclei were considered to confirm the disease states. Fig. 2 represented the H & E staining of NOM, OSFWD and OSCC.

Fig. 2 Microphotography (200×) of hematoxylin and eosin staining: (A) Normal Oral Mucosa (NOM); (B) Oral Submucous Fibrosis with Dysplasia (OSFWD); (C) Oral Squamous Cell Carcinoma (OSCC).

MALDI MS of crude metabolites prior to nanoLC separation

Before going to nanoLC separation, mass spectra of the crude metabolites were acquired from NOM, OSFWD and OSCC samples. MALDI spectra of NOM, OSFWD and OSCC samples showed an overall qualitative difference in number and intensities of metabolite peaks observed in range of 600–1000 m/z (Fig. 3). But all the three spectra showed huge spectral load showing the presence of significant number of metabolites in crude sample. These results yielded a considerable spectral overlap in all the three samples which may be one important reason of the suppression effect discussed above. The qualitative analysis of the crude spectra showed that in case of NOM and OSFWD from 610–820 m/z, and in OSCC from 870–960 m/z the spectral load was prominent (Fig. 3). These difference peak patterns of metabolites in different study groups clearly gave us a qualitative indication of altered metabolism in oral cancer.

Fig. 3 MALDI MS spectra of crude metabolites extracted from normal (A), oral pre cancer (B) and cancer (C) biopsies prior to nanoLC separation.

NanoLC separation of the metabolites mixture

The metabolites from crude sample (mixture of metabolites) were separated on the MALDI target plate using nanoLC. As mentioned above, the crude spectra of metabolites showed huge complexity of metabolites peak on single spot (Fig. 3). But after separation using nanoLC the complexity or overcrowded peaks is dramatically decreased as the metabolites present in the crude sample as separated and distributed over 96 spots instead of a single spot. As a consequence, almost five to ten metabolites peak were obtained in a single spot. Acquiring MS data from all the 288 spots (96 for NOM, 96 for OSFWD and 96 for OSCC) a number of lipid phenotypes (PC, PI, PIP₂, PG, TG, CDP-DAG) were identified involving membrane biogenesis. In Fig. 4, the graphical overview of ‘spot elution’ of the all the above lipid phenotypes is presented. It is important to notice that one metabolite (e.g. phosphatidylcholine in OSCC sample) was eluted within a particular gradient of acetonitrile (12–13%) and a particular retention time (12.25–13 min). As the flow rate or elution rate was 600 nL min⁻¹ and one spot took 0.25 min, so a particular metabolite was eluted within a spot range [e.g. 4 spots (K1–K4) in case of phosphatidylcholine in OSCC] (Fig. 4 and 6). This spot elution of particular metabolite was named as fraction spot. Spectral elution profile of a representative metabolite (phosphotidylcholine, m/z 851 in OSCC sample) in Fig. 6 showed the variation of AUC with its retention time and fraction spots. The separated metabolites were identified by MALDI MS in combination with HMDB database.

Fig. 4 (A) Spotting of eluted metabolite on MALDI target plate by spotter; (B) schematic diagram of the target plate showing direction of metabolite elution (spot elution); (C) schematic representation of separation and ‘spot elution’ of metabolites on MALDI target plate showing the presence of a particular metabolite (indicated by a particular color) in eluted fraction spots of three conditions: normal (A to D); pre-cancer (E to H); cancer (I to L).

Fig. 5 Overlapped MS spectra of PG (A) and PC (B) with isotopic distributions obtained from eluted fraction spots showing variation of AUCs in three conditions: normal (NOM), pre-cancer (OSF) and cancer (OSCC). MS/MS spectra of PG and PC along with their parent peak are shown in (C) and (D) respectively.

Fig. 6 MALDI MS spectra with isotopic distributions of a representative metabolite (phosphotidylcholine, m/z 851) present in four fraction spots (K1 to K4) showing gradual increase and decrease of AUC as it eluted from the nano column. Spectral region of six consecutive eluted fraction spots (J1 to K5) are shown here.

Validation of quantification approach with an internal standard

Validation of aAUC (obtained from AnchorChip guided target spot) concept to determine the concentration of metabolite revealed the direct proportional relationship of aAUC with the concentration of the sample. Present study also established a linear equation (Y = 10551X + 7438.4) in between aAUC and concentration having significant R² value (0.9613) (Fig. 7A). But the spots on normal MALDI plate showed non-significant (R² = 0.3151) relation in between concentration of internal standard and corresponding aAUC (Fig. 7B).

Fig. 7 Standard curve of aAUC with respect to concentration of internal standard spotted on AnchorChip MALDI target plate (A); normal MALDI target plate (B).

Quantification of the metabolites by %aAUC in different study groups

Quantification of the metabolites was performed by nanoLC coupled MALDI system. Before MALDI MS data acquisition, machine was calibrated to less than 5 ppm for high mass accuracy. Previous study used nanoLC/MS in metabolite identification.²⁶ But, here the non-targeted metabolites were identified and quantified by implementing an idea of spot elution. Multiple AUCs of a particular metabolite were eluted as sequential spots from nano column onto the target plate. The fraction spot elution of metabolites was conceptualized into a parameter %aAUC (percentage of average area under the curve) which enabled us in quantification study of non-targeted metabolomics. It includes the following steps:

(i) The AUC of a particular metabolite peak present in eluted spots are summed up to obtain its average AUC (aAUC).

aAUC = area under the curve of a particular metabolite, n = number of spots in which particular metabolite is present.

(ii) The percentage of aAUC (%aAUC) is calculated by compiling all aAUCs of the metabolite present in different study groups for significant comparative quantification.

where, 1 ≤ r ≤ Φ; (r, Φ = integers) (aAUC)r = aAUC of the rth study group; %aAUC = percentage of (aAUC)r; Φ = total number of groups.

The %aAUC concept was further explained with a representative metabolite, phosphatidylglycerol (PG). The nanoLC elution profile of PG (Fig. 4C indicated by green arrow, Table 2), showed that it was eluted on MALDI target plate from spot B4 to C9 in NOM group with retention time 11.25–14.25 min, F1 to G9 in OSFWD group with retention time 12–14.25 min and J6 to K12 in OSCC group with retention time 10.75–15 min (Fig. 4B for the direction of the elution on target plate). After MS data acquisition from all spots multiple AUCs (i.e. one AUC from each spot in which PG is present) for PG (Fig. 5A) was obtained. To get aAUC (average AUC) for each group (i.e. NOM, OSFWD and OSCC), total AUC is divided by total number of spots from the corresponding group using eqn (a) (Table 2 and Fig. 4C). Finally %aAUC was calculated using the eqn (b) calculating the percentage of aAUC of PG considering aAUC of PG in all groups which yielded %aAUC of PG in NOM is 20.98361, in OSFWD 27.54423 and in OSCC 51.47216 (Table 2). Expressional fold change of a particular metabolite was calculated by dividing its %aAUC of disease group (OSFWD/OSCC) with the control counter part (NOM). In such way the fold change of PG is 1.31 for OSFWD and 2.45 for OSCC (Table 2). %aAUC and other parameters of identified lipid phenotypes were depicted in Table 2. This method yielded fold change >2 for PI, PG, TG and timnodonyl CoA and fold change >1.9 for PC and PIP₂ in cancer sample (OSCC).

Table 2 Quantification of nanoLC separated metabolites with their corresponding retention times, %aAUC and fold change^a

Metabolites	m/z	Retention time in nanoLC			Quantitation
		NOM (n = 15)	OSFWD (n = 15)	OSCC (n = 15)	NOM			OSFWD		OSCC
					aAUC	%aAUC	aAUC	%aAUC	Fold change	aAUC	%aAUC	Fold change
a — indicates absence of metabolite peak. Arrow colors correspond to the particular metabolite shown in Fig. 4.
PC (phosphatidylcholine)	851	13.75–14.25 min	13.75–14.25 min	12.25–13 min	878	25.1594	903.5	25.89011	1.02	1708.25	48.9505	1.95
PI (phosphatidylinositol)/beta-analyl Co-A	839.2	13.75 min	13.75–14.25 min	12.5–13 min	259.5	19.41639	461	34.49308	1.77	616	46.09053	2.37
PIP2 (phosphatidylinositol bisphosphate)	997.3	15.25–15.75 min	16.5–17 min	15–15.5 min	238	25.68807	231	24.93254	0.98	457.5	49.37938	1.92
PG (phosphatidylglycerol)	769.1	11.25–14.25 min	12–14.25 min	10.75–15 min	646.4	20.98361	848.5	27.54423	1.31	1585.6	51.47216	2.45
TG (triglycerides)	869.2	11.25–17.75 min	12.75–16.5 min	11.25–15.5 min	493.7	30.52241	344.5	21.2983	0.69	779.3	48.17929	2.26
CDP-DAG (cytidine diphosphate diacylglycerol)	1008.5	—	13.75 min	12–12.75 min	—	—	262	28.39339	—	660.75	71.60661	—
Timnodonyl CoA (timnodonyl coenzyme A)	1052.5	13.25–14.25 min	12.75–13.75 min	11.75–13.25 min	1815.66	28.19939	396.66	6.160608	0.21	4226.33	65.64	2.33
3-oxo-Docosa-7,10,13,16-all-cis-tetraenoyl-CoA	1096	13.75 min	—	12.25–13 min	346	49.07801	—	—	—	359	50.92199	1.08
Ganglioside	1166	—	—	13.5–15 min	—	—	—	—	—	503.25	100	—

---

MS/MS fragmentation of separated metabolites

Some of the well separated metabolites were chosen for further de novo fragmentation using MALDI MS/MS. For MS/MS fragmentation of metabolites, parent ion was first detected and subsequently it was selected for the generation of product ions. Fragmentation pattern of PG, PC, PI, PIP₂, CDP-DAG, timnodonyl CoA, ganglioside showed significant number of product ions (Fig. 5C and D and S2†) when acquired in lift mode of the mass spectrometer (Bruker, Germany). Each metabolite showed distinct spectra upon laser fragmentation and the representative MS/MS spectra of some metabolites (phosphatidylcholine, phosphatidylglycerol) have been shown in Fig. 5C and D. Remaining MS/MS spectra are reported in Fig. S2.†

Statistical performance and reproducibility of the analysis

To perform the reproducibility of the analysis, Coefficient of Variance (CV) of AUCs of a particular metabolites in different study groups have been evaluated. CV values (Fig. 9) depicted that the low coefficient of variance of each metabolites in between their AUCs for each group (CV < 0.3 for each metabolite of NOM group; CV < 0.25 for each metabolite of OSFWD group; CV < 0.5 for each metabolite of OSCC group) which indicated the high reproducibility of the analysis. For reliability testing Intraclass Correlation Coefficient (ICC) amongst repeated measurements (aAUC) of metabolites for different study groups (NOM, OSFWD and OSCC) were evaluated through IBM SPSS Statistics 20 software. The ICC for each of the study groups [NOM-0.982 (single measure) and 0.999 (average measure); OSFWD-0.944 (single measure) and 0.996 (average measure); OSCC-0.972 (single measure) and 0.998 (average measure)] revealed the high intra-rate reliability [ICC > 0.8] for each group repetitive measurement (Table 3). Descriptive statistics like principal component analysis were also performed by XLSTAT, 2016 software to reduce the original variables into a lower number of orthogonal variables. PCA Biplot amongst different factors have been represented in Fig. 10. PCA scree plot (Fig. 11A and B) showed the highest eigenvalue for Factor 1 (F1) (6.519) which depicted that F1 explained more percentage of variability (77.42%) (Fig. 11B). Moreover, Discriminant Analysis (DA) (All-Groups Scatter Plot) was represented in (Fig. 11C) showing the discrete grouping of the metabolites according to the tissue conditions.

Fig. 8 Schematic diagram of altered lipid metabolic pathway in OSCC. Up regulation of PC, PI, PG, TG suggests enhanced membrane biogenesis observed in dysregulated cell division causing cancer progression.

Fig. 9 Reproducibility analysis of results by Coefficient of Variance (CV) of aAUCs of the metabolites (PC, PI, PIP₂, PG, TG, CPD-DAG, TCoA-timnodonyl CoA, 3-ODTCoA-3-oxo-docosa-7,10,13,16-all-cis-tetraenoyl-CoA) present in different study groups [NOM (A), OSFWD (B) and OSCC (C)].

Table 3 Intraclass correlation coefficient analysis of metabolites aAUCs present in different study groups (NOM, OSFWD and OSCC)^d

Intraclass Correlation Coefficient (ICC) for NOM
	Intraclass correlation^b	95% confidence interval		F-test with true value 0
		Lower bound	Upper bound	Value	df1	df2	Sig
a The estimator is the same, whether the interaction effect is present or not.b Type A intraclass correlation coefficients using an absolute agreement definition.c This estimate is computed assuming the interaction effect is absent, because it is not estimable otherwise.d Two-way mixed effects ICC model.
Single measures	0.982^a	0.956	0.996	798.825	6	84	0.000
Average measures	0.999^c	0.997	1.000	798.825	6	84	0.000

---

Intraclass Correlation Coefficient (ICC) for OSFWD
	Intraclass correlation^b	95% confidence interval		F-test with true value 0
		Lower bound	Upper bound	Value	df1	df2	Sig
Single measures	0.944^a	0.867	0.988	256.946	6	84	0.000
Average measures	0.996^c	0.990	0.999	256.946	6	84	0.000

---

Intraclass Correlation Coefficient (ICC) for OSCC
	Intraclass correlation^b	95% confidence interval		F-test with true value 0
		Lower bound	Upper bound	Value	df1	df2	Sig
Single measures	0.972^a	0.938	0.992	564.402	8	112	0.000
Average measures	0.998^c	0.996	0.999	564.402	8	112	0.000

---

Fig. 10 PCA Biplot: (A) Biplot (axes F1 and F2: 86.38%); (B) Biplot (axes F2 and F3: 19.40%); (C) Biplot (axes F3 and F4: 8.98%) and (D) Biplot (axes F4 and F5: 5.33%) for the aAUCs of metabolites (PC, PI, PIP₂, PG, TG, CPD-DAG, TCoA-timnodonyl CoA, 3-ODTCoA-3-oxo-docosa-7,10,13,16-all-cis-tetraenoyl-CoA) present in different study groups (NOM, OSFWD and OSCC).

Fig. 11 PCA scree plot (A) and eigenvalue table (B) of aAUCs of the metabolites present in different study groups showed the highest eigenvalue for Factor 1 (F1) (6.51) which depicted that F1 explained more percentage of variability (77.42%); (C) discriminant analysis (All-Groups Scatter Plot) for the identified metabolites aAUCs present in different study groups (NOM, OSFWD and OSCC).

Discussion

Recent years, metabonomics is emerging as an alternative approach for understanding the dysregulated molecular and cellular ambience in cancer.^1,2 Amongst the various techniques, MALDI-MS has become an important tool for metabolic profiling.^34,35 Although MALDI MS is highly used for metabolomics but quantification of metabolites using MALDI MS has several limitations.³⁶ One of the limitations is ionisation efficiencies of components to be analysed by MALDI MS for quantification.³⁷ The laser power is also a crucial variable in quantification study. Moreover, metabolite identification, annotation and quantification from crude biological samples (e.g. tissue extracts) using MALDI technique is not very conclusive because of heavy peak crowds resulting overlapping spectra. Quantification using MALDI is also hindered by the suppression effect leads to data distortion of complex biological samples where thousands of different components are present with a vast range of concentrations.³⁷

New technologies based on chromatography coupled MS approach are evolving faster to overcome these challenges. Here we have developed nanoLC coupled MALDI MS method for successful separation of metabolites from crude biological samples and their comparative quantification with newly introduced parameter %aAUC. The nano-scale chromatography method (Table 1) using crude metabolite samples from pre-cancer and cancer biopsies has been developed for separation of metabolites to overcome the suppression effect due to variable concentrations and wide range of different metabolites present in bio-samples.³⁶ Most of the metabolites have been separated in the retention time from 7 to 22 min (Table 1 and Fig. 4) where the linearly increased acetonitrile gradient is from 5% to 25% (Table 1). Spot elution profile of each metabolite separated through nano C₁₈ column represents the elution pattern of each metabolite with their corresponding retention time (Table 2 and Fig. 4C). As for example, binding of phosphatidyl choline in OSCC sample with C₁₈ nano-bead and its subsequent elution with acetonitrile gradient (12% to 13%; RT: 12.25 min to 13 min) indicates the binding-interaction of the metabolite with nano-column (Fig. 6).

Further the advantage of nanoLC MALDI over conventional MALDI MS for label free comparative quantification of metabolites has been proposed by formulating a new parameter, %aAUC as well as the expressional fold change of the metabolites amongst different study groups (NOM, OSF and OSFWD). Such comparative quantification is only possible with multiple AUCs of a particular metabolite obtained from nanoLC separated fraction spots (Fig. 4 and 6). %aAUC represents a normalised parameter, which is proportional to the total amount of the particular metabolite present in a sample. Since the same metabolite is chosen for comparative quantification in different study groups, the data distortion due to different ionization efficiencies is eliminated in this protocol for calculation of %aAUC. To minimize the variation of laser power throughout the experiment, MS data has been acquired from six thousands laser shots per spot.

This unique approach, validated by an internal standard (Fig. 7), proved remarkable expressional changes in lipid phenotypes in different study groups (Table 2). It showed that PG, PI, PC, PIP₂, CDP-DAG are up-regulated in OSFWD and OSCC (Table 2, Fig. 5 and S1†). Further, an altered metabolic pathway (up-regulation of lipid anabolism) has been proposed on the basis of the experimental finding (Fig. 8). Uncontrolled proliferation of the cancer cells needs effluent supply of ATP and bio-synthetics as a raw material for newly growing cell.^2,28,38 Therefore, both glycolytic flux as well as de novo biosynthesis of lipids, nucleotides and non-essential amino acids is crucial for cancer cell proliferation.³⁹ Malate and oxaloacetate (OAA), the TCA cycle intermediates, are main raw materials for the lipid (especially TG, PG, DAG, and PI) biosynthesis. These are very essential for membrane biogenesis for newly dividing cells.^39,40 Present study has also depicted the up-regulation of lipid anabolism rather than passing through the normal TCA cycles (Fig. 8). In this context, up-regulation of trimethyl amine N-oxide (TMAO), a choline break down product, in oral cancer was reported in our previous NMR based study.³² Here another choline derivative, PC was up-regulated in cancer (Table 2). Such expressional changes of choline derivatives indicate altered choline metabolism in oral cancer.³² Again, MS/MS fragmentation of well separated metabolites would further enrich available metabolomic database (Fig. 5 and S2†).

Hence, present study demonstrates a novel approach in quantitative cancer metabolomics using nanoLC coupled MALDI MS/MS. First time we report the separation and comparative quantification of crude metabolite extract from oral pre-cancer and cancer biopsies by developing a nano liquid chromatographic method. In the light of newly designed parameter, %aAUC a significant up regulation of lipid anabolic pathway indicating increased membrane biogenesis has been proposed in oral cancer having its' overlap with pre-cancer.

Conclusion

MALDI MS/MS is a powerful and readily acceptable means for the mass spectrometric characterization of bio-molecules such as proteins, peptides, DNA as well as chemical compounds and polymers. But label free quantification of small molecules such as metabolites using MALDI is a big challenge till now. Here we show that nanoLC coupled with MALDI MS/MS can be used for non-targeted metabolomics which includes the separation, identification and cost effective label free quantification of metabolites extracted from oral cancer samples by calculating the a newly developed parameter %aAUC. %aAUC is calculated considering multiple AUC of a particular metabolite obtained from the MALDI spectra of each spots (in which the particular metabolite is present) after nanoLC separation of crude metabolite sample. This study has revealed a remarkable alteration in lipid metabolic pathway as evidenced by the up-regulated expression of lipid phenotypes [phosphatidylinositol (PI), phosphatidylcholine (PC), phosphatidylglycerol (PG), phosphatidylinositol-bisphosphate (PIP₂), cytidinediphosphate-diacylglycerol (CDP-DAG)] indicating increased membrane bio-genesis in oral pre-cancer and cancer. Present study uses oral cancer samples but it may applicable to any cancer samples. Hence, this approach may facilitate the attempts for better understanding of altered cancer cell metabolism and it provides a great assurance to move mass spectrometry to the clinic.

Conflict of interest

The authors declare that they have no conflicts of interest.

Author's contributions

SB and DD designed, performed, analyzed the experiments and wrote the paper; BCS and AC provided technical support and analysis of the results; MP and RRP contribute in onco-pathological evaluation; AKD, AB, RB, AKR and JC evaluated the results and corrected the manuscript. All authors reviewed the results and approved the final version of the manuscript.

Acknowledgements

This work was carried out with financial assistance from the IAN project (Sanction No. IIT/SRIC/SMST/IAN/2013-14/222) of IIT, Kharagpur, MHRD, Govt. of India. The authors would like to acknowledge the Central Research Facility (CRF) of the Indian Institute of Technology (IIT), Kharagpur for establishing the nanoLC-MALDI facility. SB, DD thank IIT Kharagpur for individual fellowships.

References

1. G. J. Patti, O. Yanes and G. Siuzdak, Nat. Rev. Mol. Cell Biol., 2012, 13, 263.
2. R. J. DeBerardinis, J. J. Lum, G. Hatzivassiliou and C. B. Thompson, Cell Metab., 2008, 7, 11.
3. S. Hu, T. Yu, Y. Xie, Y. Yang, Y. Li, X. Zhou, S. Tsung, R. R. Loo, J. A. Loo and D. T. Wong, Cancer Genomics Proteomics, 2007, 4, 55.
4. F. D. Shah, R. Begum, B. N. Vajaria, K. R. Patel, J. B. Patel, S. N. Shukla and P. S. Patel, J. Clin. Biochem., 2011, 26, 326.
5. S. Bag, M. Pal, A. Chaudhary, R. K. Das, R. R. Paul, S. Sengupta and J. Chatterjee, J. clin. Pathol., 2015, 2014, 202791.
6. Y. S. Cheng, T. Rees and J. Wright, Clinical and Translational Medicine, 2014, 3, 3.
7. A. Katakura, N. Yamamoto, T. Sakuma, K. Sugahara, T. Onda, S. Noguchi and T. Shibahara, Journal of Oral and Maxillofacial Surgery, Medicine, and Pathology, 2015, 27, 1.
8. V. M. Asiago, L. Z. Alvarado, N. Shanaiah, G. N. Gowda, K. Owusu-Sarfo, R. A. Ballas and D. Raftery, Cancer Res., 2010, 70, 8309.
9. R. J. DeBerardinis, J. J. Lum, G. Hatzivassiliou and C. B. Thompson, Cell Metab., 2008, 7, 11.
10. V. Poinsignon, L. Mercier, K. Nakabayashi, M. D. David, A. Lalli, V. Penard-Lacronique, C. Quivoron, V. Saada, S. De Botton and S. Broutin, J. Chromatogr. B: Anal. Technol. Biomed. Life Sci., 2016, 1022, 290.
11. Y. Uchiyama, T. Hayasaka, N. Masaki, Y. Watanabe, K. Masumoto, T. Nagata, F. Katou and M. Setou, Anal. Bioanal. Chem., 2014, 406, 1307–1316.
12. M. W. Duncan, H. Roder and S. W. Hunsucker, Briefings Funct. Genomics Proteomics, 2008, 7, 355.
13. M. Bucknall, K. Y. Fung and M. W. Duncan, J. Am. Soc. Mass Spectrom., 2002, 13, 1015.
14. W. Wang, H. Zhou, H. Lin, S. Roy, T. A. Shaler, L. R. Hill, S. Norton, P. Kumar, M. Anderle and C. H. Becker, Anal. Chem., 2003, 75, 4818.
15. D. H. Kim, F. Achcar, R. Breitling, K. E. Burgess and M. P. Barrett, Metabolomics, 2015, 11, 1721.
16. W. Lu, E. Kimball and J. D. Rabinowitz, J. Am. Soc. Mass Spectrom., 2006, 17, 37.
17. R. Kaddurah-Daouk, B. S. Kristal and R. M. Weinshilboum, Annu. Rev. Pharmacol. Toxicol., 2008, 48, 653.
18. C. Zheng, S. Zhang, S. Ragg, D. Raftery and O. Vitek, Bioinformatics, 2011, 27, 1637.
19. J. K. Nicholson, J. C. Lindon and E. Holmes, Xenobiotica, 1999, 29, 1181.
20. M. Coen, E. Holmes, J. C. Lindon and J. K. Nicholson, Chem. Res. Toxicol., 2008, 21, 9.
21. Y. Sato, I. Suzuki, T. Nakamura, F. Bernier, K. Aoshima and Y. Oda, J. Lipid Res., 2012, 53, 567.
22. M. A. M. Rodrigo, O. Zitka, S. Krizkova, A. Moulick, V. Adam and R. Kizek, J. Pharm. Biomed. Anal., 2014, 95, 245.
23. A. Nawrocki, K. Thorup-Kristensen and O. N. Jensen, J. Proteomics, 2011, 74, 2810.
24. W. Zou, X. Wen, X. Sheng, Y. Zheng, Z. Xiao, J. Luo, S. Chen, Y. Wang, Z. Cheng and D. Xiang, Exp. Ther. Med., 2016, 11, 1653.
25. R. Kaddurah-Daouk, B. S. Kristal and R. M. Weinshilboum, Annu. Rev. Pharmacol. Toxicol., 2008, 48, 653.
26. J. Liu, Z. Zhao and Y. Teffera, Rapid Commun. Mass Spectrom., 2012, 26, 320.
27. P. Taylor, P. A. Nielsen, M. B. Trelle, O. B. Horning, M. B. Andersen, O. Vorm, M. F. Moran and T. Kislinger, J. Proteome Res., 2009, 8, 1610.
28. R. A. Cairns, I. S. Harris and T. W. Mak, Nat. Rev. Cancer, 2011, 11, 85.
29. M. G. Vander Heiden, L. C. Cantley and C. B. Thompson, Science, 2009, 324, 1029.
30. X. D. Xu, S. X. Shao, H. P. Jiang, Y. W. Cao, Y. H. Wang, X. C. Yang and H. T. Niu, Oncol. Res. Treat., 2015, 38, 117.
31. J. Xie, H. Wu, C. Dai, Q. Pan, Z. Ding, D. Hu and X. Hu, Sci. Rep., 2014, 4, 4927.
32. S. Bag, D. R. Banerjee, A. Basak, A. K. Das, M. Pal, R. Banerjee and J. Chatterjee, Biochem. Biophys. Research Commun., 2015, 459, 574.
33. E. J. Want, P. Masson, F. Michopoulos, I. D. Wilson, G. Theodoridis, R. S. Plumb and J. K. Nicholson, Nat. Protoc., 2013, 8, 17.
34. Z. Lei, D. V. Huhman and L. W. Sumner, J. Biol. Chem., 2011, 286, 25435.
35. D. Yukihira, D. Miura, K. Saito, K. Takahashi and H. Wariishi, Anal. Chem., 2010, 82, 4278.
36. E. Szajli, T. Feher and K. F. Medzihradszky, Mol. Cell. Proteomics, 2008, 7, 2410.
37. G. Schlosser, G. Pocsfalvi, E. Huszar, A. Malorni and F. Hudecz, J. Mass Spectrom., 2005, 40, 1590.
38. E. M. Griner and M. G. Kazanietz, Nat. Rev. Cancer, 2007, 7, 281.
39. O. E. Owen, S. C. Kalhan and R. W. Hanson, J. Biol. Chem., 2002, 277, 30409.
40. A. Nohturfft and S. C. Zhang, Annu. Rev. Cell Dev. Biol., 2009, 25, 539.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra07001a

‡ Share the first authorship.

---