Study on titanium dioxide nanoparticles as MALDI MS matrix for the determination of lipids in the brain


 The structures of lipids are diverse, and thus, lipids show various biological functions. Systematic determination of lipids in organisms has always been a concern. In this paper, a methodology on the matrix-assisted laser desorption ionization mass spectrometry (MALDI MS), with titanium dioxide nanoparticles (TiO2 NPs) as the matrix, was studied for lipid determination. The results showed that the following conditions were preferable in the determination of small-molecule lipids (such as hypoxanthine, guanosine, uridine, and cytidine), lipid standards (such as GC, GM, TG, phosphatidylethanolamine, phosphatidylcholine, and ceramide), and mixed lipids (extracted from brain homogenate with methanol alone and with the B&D method): TiO2 NPs as the matrix, absolute ethanol as the solvent, 1 mg of TiO2 NPs dispersed in 1 mL of absolute ethanol as the matrix solution, NaCl as the ionization reagent, and positive mass spectrometry (MS) as the mode. Modified TiO2 NP as a new matrix for MALDI MS will be a future research direction; in addition, the characteristics of TiO2 NPs make it a potential matrix for imaging MS.


Introduction
The diversity of lipid structures endows lipids with various biological functions [1]. Lipids are indispensable in the regulation of physiological activities, and the metabolism abnormalities of lipids cause many diseases such as chronic inflammation, atherosclerosis, hypertension, diabetes, obesity, Alzheimer's disease, and cancer [2][3][4]. Lipids have diversified categories and complex structures, and the different numbers of carbon atoms or unsaturated bonds in lipid acyl chains will diversify lipids into many types and subtypes [5]. Therefore, the systematic determination of lipids in organisms has always been a difficult problem. In recent years, many methods have been developed for lipids detection, such as thin-layer chromatography, liquid chromatography [6], gas chromatography, enzyme-linked immunosorbent assay, nuclear magnetic resonance spectrometry, and mass spectrometry (MS) [7][8][9]. Among them, MS has been widely applied because it has the advantages of high sensitivity, specificity, high throughput, and high accuracy. The utilization of high-resolution mass detectors has greatly improved the ability of MS for lipid determination and promotes the research on lipids [10,11].
Lipids are the main components of the brain and perform many important functions. Brain nerve cells contain a large amount of 22 carbon hexanoic acid (DHA), and it plays an important role in the development of children's brain nerves [12]. Cholesterol metabolism, lipoprotein (a), apolipoprotein ApoE, etc., are all importantly related to Alzheimer's disease and dementia [13]. A European study showed that a lipid-containing diet can delay or prevent the occurrence of Alzheimer's disease and other dementias. A lipid diet containing Omega-3 fatty acids can inhibit the above-mentioned diseases [14]. Basic and clinical studies have shown that there is an important link between metabolic disorders such as cholesterol, fatty acids and lipids, and the pathogenesis of stroke [15]. Studying the metabolism of fatty acid endogenous compounds in the brain tissue during the course of ischemic stroke found that: There are 9 fatty acid metabolism abnormalities in the brain tissue of rats with focal cerebral ischemia, 9-hexadecenoic acid, docosa, hexaenoic acid, arachidonic acid, 9-octadecenoic acid, hexadecanoic acid, 11-octadecenoic acid, octadecanoic acid, 13-eicosene acid, 11-eicosenoic acid, can be used as potential biomarkers [16].
Matrix-assisted laser desorption mass spectrometry (MALDI MS) is a new MS technique [17][18][19]. Typically, a sample is dispersed in matrix molecules and forms crystals. When the crystals are irradiated by laser, the energy derived and accumulated leads to a rapid temperature increase, which sublimates the matrix crystals. The sample and matrix will expand and migrate into the gas phase, the matrix provides the volume flow, in this process, the energy passes on to the sample and ionized it into ions, then the ions are separated and analyzed in the MS system [20,21]. It is noteworthy that monocharged ions are dominant in MALDI, so the signals of ions in these mass spectra can match with the mass of polypeptides and proteins exclusively [22][23][24].
The laser wavelength of a commercial MALDI apparatus has been fixed already, and thereby the matrix is the most important factor affecting the detection ability of the apparatus. Common matrix materials include solid, liquid and liquid/solid biphasic materials [25][26][27]. Divided by the chemical properties, matrix materials include inorganic materials such as graphite and metal salts and organic materials (acidic, basic, and neutral) [28,29]. Currently, no matrix is suitable for the analysis of all samples, so the research and analysis of the matrix are very important.
The screening and selecting of the specific matrix are very tricky in MALDI analysis. To be considered as a good matrix, it should satisfy these requirements: (1) To reduce the force between molecules and prevent the formation of molecular clusters, the matrix should be miscible with the target. (2) The matrix is stable under vacuum conditions. (3) The matrix can absorb laser light. (4) Under laser light, the matrix can protect the sample and transfer energy to it without destroying the sample's structure. (5) The matrix can provide protons, which can promote the ionization of the sample [30][31][32]. Because of the uneven distribution of combined matrix-sample crystals on the MALDI target, the reproducibility of samples is poor, especially in the quantitative analysis of MALDI, which is also a difficult problem.
In the presence of traditional organic matrix materials, such as α-cyano-4-hydroxycinnamic acid (CHCA), 2,4,6-trihydroxyacetophenone (THAP), and 2,5-dihydroxybenzoic acid (DHB), the mass spectrum signals of analytes can be intensive. But, because the molecular weights of matrix materials are close to 200 Da, copious matrix debris with a small mass number will be generated when the samples are subjected to laser irradiation. The mass spectrum signals of debris will seriously interfere with the spectral analysis of small molecules. The authors attempted to develop a new matrix material to eliminate debris and simplify the spectrum for more accurate detection of small molecules. Currently, the research and development of nanomaterials are at the cutting edge in the MALDI-TOF MS matrix field [33,34]. Nanomaterial is a new material with a structural unit size of 1-100 nm. Due to its small particle size, large specific surface area, and strong modifiability, it has been widely applied in pharmaceuticals, catalysis, sensing, magnetic recording, and so on. The nanomaterials as the MALDI-TOF MS matrix principally include metal-organic frameworks, carbon-based nanomaterials, silicon-based nanomaterials, metal particles and their oxides, etc.: (1) Metal-organic framework matrix: Currently, the metal centers of MOFs used to prepare the matrix principally include Zr, Cr, Zn, Cu, etc., and most of the ligands are polyhydroxycarboxylic acids and pyridines, and other organic compounds containing multiple auxochromes and chromophores [35,36]. (2) Nanographene matrix: This matrix includes graphene, graphene derivatives, and their composites. Many different types of graphene derivatives and composites such as graphene sheets [37], grapheme/SiO 2 nanocomposites [38], and N-doped graphene [39], have been proven to be excellent MALDI-TOF MS matrixes. (3) Carbon nanotube matrix: The surface of carbon nanodot can be modified with abundant carboxylic acid groups, showing water solubility and so it can be evenly dispersed in the solution without discharge and pollution. In addition, the size of a carbon nanodot is small (∼3 nm), and the analyte is easy to desorb, so the detection limit can be as low as 0.2 fmol [40,41]. (4) Silicon-based nanomaterial matrix: Li et al. [42] prepared a novel MALDI matrix material by the electrochemical etching of porous silicon (PSi), and the surface of the matrix was modified with palladium nanoparticles (Pd NPs). This material can specifically enrich peptides. Due to the limitation of pore size and molecular filtration effect of PSi, the peptides in serum samples could be selectively captured and enriched in the pores, thus eliminating the interference of large protein molecules in subsequent MALDI-TOF MS detection. (5) Metal nanoparticle matrix: Co, Ag, Mo, and nanooxides such as TiO 2 and ZrO 2 ·Fe 3 O 4 /TiO 2 [43] and ZnO-CuO nanoparticles [44] have been used as the matrix for MALDI-TOF MS to detect small molecules. Functionalization and modification of metal nanomaterials are an important means to improve the performance of the metal nanomaterial matrix [45].
Compared with other biomolecules, lipids have the following characteristics: (1) The molecular weights (<1,500 U) are low and mass/charge ratios are similar, so the signals may overlap with the relevant peaks of the matrix system in MALDI-MS. This is detrimental to the recognition of phospholipids.
(2) The water solubility is low, and the lipids should be dissolved in the solvent to a certain extent during measurements.
(3) Diverse molecular structures. For example, according to the polar head, the lipids can be divided into phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylserine (PS), phosphatidylinositol (PI), phosphatidylglycerol (PG), glycerophosphatidic acid (PA), and so forth. From the perspective of the bond (ester, alkanoyl, or alkenoyl bond) between the fatty acid chain and glycerol skeleton at Sn-1 and Sn-2 (unsaturated) positions, the lipids can be divided into certain subclasses. Different types of phospholipids usually co-exist but the different polar heads determine that they cannot be detected simultaneously with the same detection mode. For example, PC and PS are subject to positive ionization, while PE, PI, PA, and PG are subject to negative ionization. (4) The lipids with different structures are distinctly contained in biological tissues.
These characteristics of lipids require specific MALDI matrix systems. Cinnamic acid derivatives such as erucic acid (SA), α-cyano-4-hydroxycinnamic acid (HCCA), and benzoic acid derivatives such as 2,5-dihydroxybenzoic acid (DHB) are the earliest matrix for phospholipid analysis. 2,6-Dihydroxyacetophenone (DHA) is the most common neutral matrix. Correspondingly, 50-70% ethanol or methanol solutions (aq) containing 0.1% TFA are the most common solvents. The addition of cesium iodide [46] or lithium chloride [47] as the ionization reagent is conducive to MS analysis. 9-Aminoacridine (9-AA) was first adopted by Vermillion Salsbury and Hercules in the year 2002 to detect metabolites or phospholipids in peptide hydrolysates [48]. During the determination of lipids, cholesterol and triacylglycerides could not be ionized in the presence of 9-AA, and thereby the analysis complexity of phospholipids in lipids could be effectively reduced [49].
Metal and metal-oxide nanoparticles include Au, Ag, Mn, Zn, Fe 2 O 3 , TiO 2 , ZnO, etc. For example, Jackson et al. sprayed gold colloid (5.5 nm) dispersed in ethanol with an artificial spray gun on rat brain slices for the determination of cerebroside [50]. Taira et al. modified two hematite (Fe 2 O 3 ) unit cells (corundum-structured) with functionalized silicate materials to obtain functionalized nanoparticles (fNPs), and sprayed the fNPs on rat brain slices for the analysis of peptides and lipids [51].
The traditional analytical methods, such as HPLC, TLC, and GC-MS, are complicated and insensitive. For instance, for GC-MS analysis, phospholipids should be hydrolyzed first and then derivatized. Moreover, only the structures of fatty acyl groups can be recognized, and those of phospholipids cannot be accurately identified. HPLC-ESI-MS and ESI-MS techniques possess the merits of simple sample pretreatment, high resolution, easy automation, and so on. Liquid chromatography coupled with mass spectrometry (LC-MS) has greatly promoted the development of phospholipids, and the core technology is ESI-MS. The chromatographic technology strengthens the separation of phospholipids in samples, so the separation and identification process of phospholipids has the advantages of high throughput, high sensitivity, and efficiency. Phospholipids can be divided into glycerophospholipids and sphingomyelins (SM) according to their alcohol domains. Glycerophospholipids can be divided into PC, PE, PS, PI, PG, and PA according to the polar head. The ionization energy of MALDI MS is relatively high, so some samples that are insoluble in highly polar solvents such as methanol and water can also be ionized. In addition, this technique can deal with different types of lipids through rapid, sensitive, and high-throughput qualitative or quantitative analysis without special treatment.
TiO 2 has many good features including being nontoxic, no-scent, environmentally friendly, anti-corrosion, photocatalytic, able to absorb UV light, and having a large specific surface area. Its even size helps the analytes form uniform crystals; TiO 2 itself does not generate jamming signals like other matrices [52,53].
In this paper, titanium dioxide nanoparticles (TiO 2 NPs) were used as the matrix in MALDI MS for lipids analysis [54], the methodology of which was studied. The feasibility of TiO 2 NPs matrix and the influences of matrix solvents, matrix compositions, and ionization reagents on the analysis of lipid samples by MALDI MS were discussed.

Materials and instruments
Titanium dioxide was purchased from Sigma-Aldrich Bligh & Dyer method [55,56], separately. An analytical balance, pipettes, and 50 mL beakers were employed. A Solarix ft ms spectrometer was used in the determination and the parameters of the mass spectrometer are listed in Table 1.

Preparation of samples, matrix, and ionization reagents
A matrix solution was prepared as follows: 1 mg of titanium dioxide was mixed with 1 mL of absolute ethanol. An ionization reagent was prepared as follows: sodium chloride was dissolved in absolute ethanol to form a 10 −5 mol·L −1 solution. Biological small-molecule solutions were prepared as follows: a sample was mixed with absolute ethanol to prepare 0.1 mL of a 0.001 mol·L −1 solution.
A lipid standard solution was prepared as follows: a standard was mixed with 1 mL of methanol, and 0.1 mL of the solution was withdrawn as a 0.001 mol·L −1 lipid standard solution.
A solution of mixed lipids was prepared as follows: 1 mL methanol was added to the mixed lipids, and 0.1 mL of the solution was withdrawn as a 0.001 mol·L −1 solution of mixed lipids.

Pretreatment of samples of mixed lipids
Two methods were separately applied for the pretreatment: the method using methanol alone and the B&D method.
The former method was performed as follows: 450 μL of methanol was mixed with 50 μL of the brain homogenate. After stirring for 1 min, the mixture was placed for 30 s and then was centrifuged for 10 min at 12,000 rpm. 100 μL of the extract was withdrawn with a pipette and  The B&D method was performed as follows: 250 μL of chloroform and 500 μL of methanol were mixed as the solvent. About 100 μL of the brain homogenate was mixed with the solvent, and the mixture was stored at −20°C for 5 min with ice. Then, 250 μL of water and 250 μL of chloroform were added. After stirring for 1 min, the mixture was centrifuged for 10 min at 12,000 rpm. Then, 100 μL of the extracts at upper and lower layers were separately withdrawn and stored at −20°C, separately (the samples contained different lipids with different molecular polarities after the removal of proteins and minerals).

Preparation of MALDI samples
With TiO 2 as the matrix, all the samples for analysis were prepared with a dry-point method implemented as follows: 3 μL of a sample solution and 3 μL of the matrix solution were uniformly mixed. If an ionization reagent was required, 1 μL of a sodium chloride solution was added to the mixture. After complete mixing, 1 μL of the mixture was added dropwise on a clean stainless-steel sample target. After the evaporation of solvent and crystallization of the sample, a qualified sample was thus prepared for analysis with the mass spectrometer.

Influence of TiO 2 matrix solvent
Water, absolute ethanol, and methanol were separately used as the solvent to prepare a matrix solution for the detection of small biological molecules. When water was used as the solvent, a suspension was formed, and the sites with concentrated samples were difficult to find, so water serving as the solvent was denied. Figures 1 and 2 show the spectra with methanol and anhydrous ethanol as the matrix solvent, respectively. After comparison, the signals in Figure 1 were distributed disorderly and contained many signals of impurities. Therefore, absolute ethanol was chosen as the solvent, and TiO 2 was chosen as the matrix for determination; the results showed clearer signals, compared with those in Figures 1 and 2, even if the spectra were of 10 −7 and 10 −8 , the signal peaks obtained by MS can be clearly separated. These spectral results show that the detection precision was high with the use of the TiO 2 matrix and absolute ethanol solvent. The precision of MS for solvent is good.

Influence of the TiO 2 matrix concentration
About 0.01, 0.1, 1, and 10 mg of TiO 2 were separately used for the preparation of matrix solutions for the determination    Figure 3 shows that the mass spectra of 18:1-12:0 glucosyl(β)ceramide (GC, C 36 H 69 NO 8 ) obtained under the conditions of these four concentrations are rather different. The mass spectrum under the condition of 1 mg of TiO 2 is the clearest, almost without interfering signals, which can be seen from Figure 3c. In contrast, the other three spectra contain intense interfering signals and show poor results. Hence, 1 mg of matrix per mL was selected in the following experiment.

Influence of ionization reagents
Li + , Na + , K + , and Se + were separately selected as the ionization reagent in the determination of 18:1-12:0 glucosyl(β)ceramide (GC, C 36 H 69 NO 8 ). The resulting mass spectra were compared for the optimization of the ionization reagent in this experiment. Figure 4 shows that the mass spectra of 18:1-12:0 glucosyl(β)ceramide (GC, C 36 H 69 NO 8 ) separately obtained in the presence of these four ionization reagents are distinct. Among them, the spectrum in the presence of Na + is the clearest, almost without interfering signals, which can be seen from Figure 4c. In contrast, the other three spectra contain intense interfering signals and exhibit poor quality. Hence, NaCl was selected as the ionization reagent in the following experiment.

Influence of option between positive and negative MS
The mixed lipids samples are usually extracted from plasma, brain homogenate, and so forth. In the present work, lipids were extracted with methanol alone and the B&D method, separately. The former lipids extracted seemed like lipids. The latter seemed like samples with high polarity and small polarity. The volumetric ratio of plasma to methanol in the former method was 1:9. The volumetric ratio of methanol/chloroform/water in the B&D method was 1:1:0.9. The positive and negative mass spectra of samples extracted by both methods were compared. The negative mass spectra shown in Figure 6 were compared with the positive spectra shown in Figure 5, respectively. It was found that the interfering signals of the positive spectra were far less than those of the negative spectra. Figure 5 shows that the positive mass spectra of lipids, whether a high or low polarity, exhibit clear signals. On the other hand, the positive and negative mass spectra of samples extracted with methanol alone show similar graphic quality (see Figure 7). In summary, positive MS was selected in the following sections.
The positive and negative mass spectra of samples extracted with the B&D method (upper and lower layers) and methanol alone are illustrated as follows: the B&D method in Figures 5 and 6 and the extraction with methanol alone in Figure 7.

Mass spectra of lipid standards
The lipid standards were determined with 1 mg of TiO 2 as the matrix, absolute ethanol as the solvent, and NaCl as the ionization reagent. The corresponding mass spectra were obtained.
Under the conditions of absolute ethanol as the matrix solvent, 1 mg of the matrix, and Na + as the ionization reagent, the mass spectra of lipid standards were almost clear with a few interfering signals, which did not appreciably interfere with the signals of lipid standards.  [M + Na] + , and [M + K] + ions. The mass spectra of GM, ceramide, GC, TG, PC, and PE show clear signals with a few interfering signals by using TiO 2 as the matrix for MS. In particular, the conventional organic matrix (cyano-4-hydroxycinnamic acid, CCA) and TiO 2 as the matrix for comparison of MS are shown in Figure 8, and the spectrum of ceramide using titanium dioxide matrix is very clear. However, for the DHB-based TOF MS measurements, some of the galactocerebroside peaks overlap with the PC peaks with the mass resolution of measurements [57]. The information on lipid standards is shown in Table 2.
MALDI MS is a soft ionization technique. Its traditional matrices create massive interference when the sample has a low mass-to-charge ratio, which makes it difficult to analyze small molecules. Comparing with a small organic molecular matrix, TiO 2 NPs perform very well at a low m/z (<500 Da) interval. Moreover, the structure of TiO 2 NP crystals is uniform that improves the reproducibility, stability, and sensitivity of the analytical

Conclusion
Lipids with diverse structures possess many biological functions and play an indispensable role not only in the regulations of various physiological activities (energy conversion, materials transport, information recognition and transmission, cells development, differentiation, and apoptosis). MS has been widely applied because of its high sensitivity, specificity, high throughput, and high accuracy.
In this study, FT-ICR-MS was used as the instrument for the determination of small biological molecules (such as hypoxanthine, guanosine, uridine, and cytidine), lipid standards (such as GC, GM, SM, PE, PC, and ceramide), and mixed lipids in the mode of positive MS, with TiO 2 as the matrix, absolute ethanol as the matrix solvent (a matrix solution was prepared by dissolving TiO 2 in absolute ethanol), 1 mg·mL −1 matrix, and NaCl as the ionization reagent. The spectra derived showed clear signals, good sensitivity, and a few interfering signals.
As a new matrix for MALDI MS, TiO 2 NPs explore the possible applications of inorganic matrices instead of traditional organic matrices. There are several points regarding TiO 2 NPs as a matrix and should be further studied. The first goal is to improving TiO 2 NPs to work for other types of analytes. Second, experiments show that the crystallization type of the matrix and the analyte has strong effects on the result, yet there is no theoretical basis or explanation for this phenomenon. This makes it difficult to make qualitative and quantitative analyses, so a theoretical basis will have a significant impact. Last, but not least, the features of TiO 2 NPs make it a potential matrix for imaging MS. Author contributions: Congnan Peng: writingoriginal draft, Qian Zhang: writingreview and editing, Jian-an Liu: data curation, Zhen-peng Wang: methodology, formal analysis; Zhen-wen Zhao: funding acquisition; Ning Kang and Yuxin Chen: validation, Qing Huo: supervision, conceptualization.

Conflict of interest:
The authors state no conflict of interest.