TiO₂ nanocrystals – assisted laser desorption and ionization time-of-flight mass spectrometric analysis of steroid hormones, amino acids and saccharides. Validation and comparison of methods

Abstract

In the present study, the possibility for the application of TiO₂ nanocrystals of various shapes and sizes, for substrate-assisted laser desorption and ionization time-of-flight mass spectrometric (SALDI TOF MS) quantitative analysis of small molecules (steroid hormones, amino acids and saccharides) was investigated. Parameters, such as homogeneity of the substrate/analyte distribution, reproducibility of the measurements, within-day, and day-to-day repeatability, were determined. The homogeneity of different nanocrystal/analyte combinations on the target plate were compared based on the signal-to-noise values of several analyte signals. Obtained results show that all TiO₂ nanocrystals, regardless of their shape, have great potential for the detection and determination of steroid hormones, amino acids and saccharides with good analytical parameters and detection limits. On the other hand, the reproducibility of the S/N ratio and detectability of the analytes recorded in various modes differ depending on the substrate. All examined molecules were detectable in negative ion mode with TiO₂ NTs, in contrast to all other organic matrices and substrates, and the best reproducibility was obtained with the larger nanocrystals, TiO₂ PNSs and TiO₂ NTs, making them good candidates for the quantitative determination of small molecules.

---

Introduction

The interference of matrix with analyte signals in the low mass range and non-uniform analyte distribution within the matrix during their co-crystallization limit the application of conventional MALDI for the analysis of low molecular weight compounds (M < 1500 Da, “small molecules”).^1,2 One of the most important features of substrate-assisted laser desorption and ionization mass spectrometry, SALDI MS, on the other hand, is the absence of matrix intrusion in the low-mass region because the use of organic matrices is not required. SALDI MS thus extends the detectable mass range of small molecules to less than m/z 500.

The term SALDI MS was first introduced by Sunner et al. using graphite as a matrix.³ Since that time, numerous substrates have been tested as matrices and in general, they enable efficient ionization with minimal analyte fragmentation, and with the possibility of achieving high selectivity, sensitivity and reproducibility of the analysis due to appropriate surface chemistry and morphology.² Substrates for SALDI MS techniques often utilize nanoparticles, which absorb the laser energy and efficiently transfer it to the analyte.

The performance of SALDI has been improved in terms of a soft LDI process, with the detection of both polar and nonpolar compounds. Literature highlights indicate the signal enhancement factors for molecular ions in the SALDI mass spectra.⁴ In contrast to MALDI, the SALDI process has a disadvantage based on the fact that the efficiency of generating protonated molecular ions is low.⁴

The use of metal nanoparticles for SALDI MS was originally inspired by Tanaka et al., exploiting a suspension of 30 nm cobalt nanoparticles in glycerol to analyze proteins and synthetic polymers.⁵ Besides metals, semiconductors with good UV absorbance are promising candidates for SALDI MS.⁶ In the UV region, TiO₂ NPs exhibit strong absorption characteristics (band gap of bulk anatase TiO₂: 3.2 eV).⁶ TiO₂-based substrates afford the advantages of being chemically stable in air, chemically modified and readily available.^4,7

The usual approach in LDI mode for the detection of molecules that do not have sufficient proton donating activity is to apply higher laser power, which in turn might result in the increased onset of side-processes, such as aggregation or fragmentation.⁸ However, LDI mass spectrometry when applicable has limitations related to the thermal decomposition of analytes, sensitivity, and variety of compound classes (mainly, low molecular weight organic substances) and efficiency of absorbing UV laser radiation.^2,9

In our previous study, we demonstrated the advantages of the application of colloidal TiO₂ nanoparticles (5 nm diameter) for the SALDI mass spectrometric analysis of transition metal complexes,¹⁰ and the influence of the presence of inorganic salts on the quality of obtained mass spectra.¹¹ Those results showed that although suffering from lower sensitivity in the detection of transition metal complexes compared to organic matrices, TiO₂-assisted LDI spectra were much simpler to interpret, and their quality was not affected by the presence of inorganic salts. The current study is extended to the investigation of the possibility for the quantification of small physiologically-relevant molecules, which are not detectable by the matrix-free approach, by the assistance of three TiO₂ nanocrystals (TiO₂ nanoparticles (TiO₂ NPs), TiO₂ prolate nanospheroids (TiO₂ PNSs) and TiO₂ nanotubes (TiO₂ NTs)) with different sizes and shapes. The following groups of small molecules were used as the model system:

• Sex steroid hormones play important roles in maintaining normal reproductive and non-reproductive functions.

• Aminothiols such as glutathione (GSH), cysteine (Cys), and homocysteine (Hcy) and other amino acids are important in the control of vital metabolic processes and in the defense against reactive oxygen species.

• Carbohydrates are energy sources for all cells in the organism, especially for the brain and the nervous system, and changes in their metabolism could lead to severe disease states.

The usefulness of TiO₂-based substrates towards the sensitivity, reproducibility, rapidity and simplicity throughout the analyses of substrate and analyzed molecule combination, is validated in this study.

Materials and methods

Materials

All tested chemicals, β-cyclodextrin, D-(+)-maltose, glutathione D-(+)-glucose, testosterone, progesterone, estradiol, arabinose, raffinose, DL-methionine, L-alanine, and L-cysteine were purchased from Sigma-Aldrich, Taufkirchen, Germany.

Methods

Synthesis of TiO₂ nanoparticles, TiO₂ prolate nanospheroids and TiO₂ nanotubes. A dispersion of TiO₂ NPs was prepared following a procedure described previously.¹² Briefly, TiCl₄ solution (cooled to −20 °C) was added to cold bidistilled water (at 4 °C) in a drop wise manner under vigorous stirring and left at this temperature for 30 min. The solution pH was between 0 and 1. The slow growth of the particles was achieved by dialysis of the colloidal dispersion of TiO₂ against the solution (4 °C), until its pH reached 3.5. The concentration of TiO₂ in the colloid was determined as described previously.¹³

Titania nanotubes were synthesized according to the procedure reported by Kasuga et al.¹⁴ The hydrothermal processing (90 min at 250 °C, Teflon vessel – Parr acid digestion bomb) of the dispersion (pH ∼ 5–6) of scrolled titania nanotubes (2.5 mg mL⁻¹) was carried out for the synthesis of TiO₂ PNSs.

X-ray. The X-ray diffraction (XRD) measurements were carried out on a Bruker D8 ADVANCE diffractometer in theta/theta reflection geometry with parallel beam optics obtained by a multilayer Göbel mirror. Diffraction data for structure analysis were obtained in the 2θ range from 10° to 80° with steps of 0.05° and 10 s counting time per step.

SALDI TOF MS. A small volume of the sample solutions (0.5 μL) and then the same volume of TiO₂ substrate solutions were applied onto the MALDI target plate. Aqueous suspensions of TiO₂ NPs, PNSs and NTs were used at concentrations of 0.22 mol L⁻¹, 0.03 mol L⁻¹ and 0.12 mol L⁻¹, respectively.

Experiments were performed in positive and negative-ion reflector mode on a Voyager DE Pro MALDI TOF mass spectrometer. The samples were irradiated with a 20 Hz nitrogen laser at 337 nm. Spectra were obtained automatically for the m/z range of 1–1000, except for β-cyclodextrin wherein m/z was in the range of 500–2000. Ions produced by laser desorption were energetically stabilized during a delay extraction period of 150 ns. Accelerating voltage was set at ±20 kV. To obtain good resolution and signal-to-noise (S/N) ratios, laser fluence was adjusted to slightly higher than the threshold (for substrates TiO₂ NPs, TiO₂ PNSs and TiO₂ NTs it was 2400, 1950 and 2000 LI, respectively).

Statistical analysis. One sample spot on the MALDI plate was visually split into eight sectors, and each sector was shot with 120 laser shots (960 shots per one spot). Based on S/N values obtained in these sets of experiments, the homogeneity of each analyte/substrate combination was calculated. For calculations of within-day repeatability, this procedure was carried out in triplicate. The same procedure was repeated for three consecutive days for the calculation of day-to-day precision.

The within-day precision (variance of repeatability) for each combination of substrate/molecule and intermediate precision (a day-to-day repeatability) over three days were analyzed using statistical testing (ANOVA). In one day measurements, all three sets of eight measurements (each set was the sum of 960 spots) were grouped.

Results and discussion

In spite of the efforts made toward quantitative applications of MALDI MS, there are still significant challenges that include poor shot-to-shot and batch-to-batch reproducibility, caused by the inhomogeneous co-crystallization of analytes with the matrix, and interference of matrix signals.^2,6

Herein, the applicability of TiO₂ nanocrystals of different diameters and shapes, such as colloidal TiO₂ nanoparticles (NPs, average diameter ∼5 nm), prolate nanospheroids (PNSs, length: 40–50 nm, lateral dimension: 14–16 nm) and nanotubes (NTs, length: 100–150 nm, average diameter 11 nm), were tested as substrates for SALDI TOF MS quantitative analysis of three groups of low mass molecules (twelve molecules in total). The analyzed molecules were amino acids (L-cysteine (L-Cys), L-alanine (L-Ala), DL-methionine (DL-Met), and tripeptide glutathione (GSH)); sex steroid hormones (estradiol – E2, testosterone – T, and progesterone – PRG); and carbohydrates (D-(+)-glucose – D-(+)-Glu, D-(+)-maltose – D-(+)-Malt, raffinose – Raff, arabinose – Ara, β-cyclodextrin – β-CD). The identities of the signals of the tested molecules in the SALDI spectra are given in Table S1.† The NPs and NTs substrates have anatase crystalline structures as shown in our previous reports.^15,16 XRD patterns of PNSs revealed the anatase crystal structure of these particles as well (Fig. S1†).

The selection of small molecule analytes for investigation was driven by their clinical significance and the necessity for their fast detection and quantification. Carbohydrates are the exception and they were chosen due to the absence of reliable and fast methods for their analysis. Concerning the chemical structures, members of each group of molecules bear the same functional groups.

T and its metabolite dihydrotestosterone (DHT), PRG and E2 are classified as sex-steroids and they decrease with age. Sex steroid hormones are implicated in the cognitive processes of the adult brain.¹⁷ In the last two decades it has been shown that receptors for estrogens, progesterone and androgens are expressed in non reproductive tissue/organs (bone, brain, cardiovascular system) playing a role in their function. Therefore, it is critical to evaluate the impact of sex steroid hormones in the pathophysiology of certain diseases (osteoporosis, alzheimer's, and atherosclerosis).¹⁸

The second group is represented by GSH, L-Cys, L-Ala and DL-Met. The level of GSH and Cys in blood are correlated with several diseases, such as hyperhomocysteinemia as a risk factor for atherosclerosis, cardiovascular and chronic kidney disease.¹⁹ Thiol groups are reducing agents, existing at a concentration around 5 mM in animal cells. The normal level of the Cys in the cell is 9.5–11.5 μM.²⁰ The reduced form of GSH is the most abundant intracellular low molecular weight thiol in the cell, and plays an essential role in protecting the cell from toxic species.²¹ Met and Cys are involved in many vital catalytic reactions that maintain our bodies functioning properly. Ala can be easily formed and thus has close links to metabolic pathways such as glycolysis, gluconeogenesis, and the citric acid cycle. It also arises together with lactate and generates glucose from protein via the Ala cycle. Alterations in the Ala cycle that increase the levels of serum alanine aminotransferase (ALT) are linked to the development of type II diabetes.²²

Carbohydrates are energy sources, but so far, there are no methods for their analysis that are fast and do not require their modification.

In the matrix-free approach (LDI mode) to analysis, none of the analytes yield signals in the mass spectra, most likely due to insufficient proton donating activity and absorptivity of the analytes in the range of the laser irradiation wavelength. This indicates that the presence of the assisting agent on the plate is required to obtain the corresponding mass spectra.

SALDI TOF mass spectra of steroid hormones, amino acids and carbohydrate

In the first set of experiments, we tested whether the TiO₂ nanocrystals were suitable for the detection of selected molecules. In fact, only the peaks that were clearly not present in the LDI spectra of the substrate (Fig. S2†), showing no overlap with the substrate peaks were considered for further statistical analysis. It is important to emphasize that the analytes do not yield any detectable signals under LDI conditions (substrate and matrix-free approach); signals appeared when nanocrystals were applied.

The reasons why nanocrystals made of titania enable mass spectrometric detection of selected compounds and the exact processes that occur on the substrate surface under laser irradiation, are not yet fully clarified. Numerous authors assume that thermal effects play an important role,^4,23 but there are also those who emphasize the electric/charge transfer and/or photocatalytic processes on the surface of titania-based nanocrystals.^24,25 Further and carefully planned experiments are required to clarify these fundamental processes in the vacuum/MALDI chamber.

Analyte concentration was in the range of 1–11 mM, depending on the mass. In Table 1 is presented the detectability of individual compounds with the peak list and assignment; specifically, analytes that are obtained only with the increased concentrations are presented herein. Representative SALDI TOF mass spectra in positive and negative ion mode are given in Fig. 1a and b, respectively. We have chosen the most representative mass spectra from among the spectra acquired with all TiO₂ nanocrystals, with almost all detectable signals. The signals taken for statistical evaluation are indicated according to their m/z ratio (Table S1† shows the identities of detected signals).

Table 1 Detectability of molecules with TiO₂ substrates. The table comprises all signals (m/z ratio) arising from the analytes detectable with a particular substrate^a

Molecule	Substrate TiO2	Detectability		m/z values			Detectable at higher concentration
		Positive	Negative	Positive	Negative		Positive	Negative
		Ion mode		Ion mode			Ion mode
a *It is detectable, but S/N values are lower than 50, which is our empirical value.
E2	NPs	−	−	—		—	407.6, 423.6, 439.7, 439.7*		—
	PNSs	+	+	407.6, 423.6*		328.9
	NTs	+		407.6, 423.6, 439.7		328.9
T	NPs	+	−	401.6, 423.6		—	401.6*		399.6, 399.6
	PNSs	+	−	423.6		—
	NTs	+	+	401.6, 423.6		399.6
PRG	NPs	+	−	315.5, 337.5		—	315.5*		313.5, 313.5
	PNSs	+	−	337.5		—
	NTs	+	+	337.5		313.5
L-Ala	NPs	+	+	90.1, 112.1, 128.2		88.1	112.1		88.1
	PNSs	−	+	—		88.1*
	NTs	+	+	112.1, 128.2		88.1
L-Cys	NPs	+	−	121.2, 144.2*		—	144.2, 144.2		120.2, 120.2
	PNSs	−	−	—		—
	NTs	+	+	144.2, 167.1		120.2
DL-Met	NPs	+	−	172.2*		—	150.2, 172.2, 188.2		148.2
	PNSs	+	+	172.2, 195.2		148.2
	NTs	+	+	172.2, 188.3, 195.2		148.2
GSH	NPs	+	−	330.3*		—	308.3, 330.3, 346.4		306.3, 306.3
	PNSs	+	−	308.3, 330.3, 346.4*		—
	NTs	+	+	330.3, 346.4		306.3
D-(+)-Glu	NPs	−	−	—		—	203.2, 219.3		179.2*, 179.2*
	PNSs	+	−	203.2		—
	NTs	+	+	203.2, 219.3		179.2
D-(+)-Malt	NPs	−	−	—		—	365.3		222.0, 222.0
	PNSs	+	−	365.3		—
	NTs	+	+	365.3		222.0
β-CD	NPs	−	−	—		—	1158.0, 1174.1
	PNSs	+	−	1158.0		—
	NTs	+	+	1158.0, 1174.1		1134.0
Raff	NPs	+	−	203.2, 365.3, 527.4, 543.5		—			222.0
	PNSs	+	+	185.1, 203.2, 365.3, 384.3, 527.4		222.0, 503.4	185.1*, 203.2*, 543.5
	NTs	+	+	185.1, 203.2, 365.3, 527.4, 543.5		222.0, 503.4	185.1*, 203.2*
Ara	NPs	−	+	—		149.1*	173.1, 189.2, 173.1, 189.2*		149.1, 149.1
	PNSs	−	+	—		149.1*
	NTs	+	+	173.1, 189.2		149.1

---

Fig. 1 (a) Positive ion mode SALDI TOF mass spectra of E2, T, PRG, L-Ala, L-Cys, DL-Met, GSH, D-(+)-Glu, D-(+)-Malt, β-CD, Raff, Ara, with one of the substrates that gave the major number of peaks of analyzed molecules, for which S/N values were analyzed further. (b) Negative ion mode SALDI TOF mass spectra of E2, T, PRG, L-Ala, L-Cys, DL-Met, GSH, D-(+)-Glu, D-(+)-Malt, β-CD, Raff, Ara with one of the substrates that obtained the major number of peaks of analyzed molecules for which S/N values were analyzed further.

Investigated hormones were detectable with TiO₂ PNSs and TiO₂ NTs in positive ion mode, whereas only E2 was unidentified with TiO₂ NPs, at least at the tested concentrations. Similarly, T and PRG had low S/N ratios with TiO₂ NPs. In contrast, TiO₂ NTs as substrate obtained good quality spectra with hormones in negative ion mode (Fig. 1b). With all three TiO₂ substrates we detected only adducts of E2 or T molecules, their corresponding fragments, and alkali (Na⁺ or K⁺) adducts, whereas in the case of PRG, only the proton and sodium adducts were detectable.

The spectra of amino acids did not show regularities like those of the hormones. For instance, L-Cys and L-Ala were not detected with TiO₂ PNSs, whereas the molecular ion of L-Cys was measurable with TiO₂ NPs. In general, when TiO₂ NTs were exploited, adducts of all amino acids with alkali ions were detectable in positive ion mode, whereas in negative ion mode only deprotonated molecules were obtained (Table 1).

In the case of the spectra of L-Ala measured in negative ion mode, all three substrates were demonstrated to be applicable. When the S/N intensities of the signals arising from L-Ala were compared, the following order was obtained as follows: TiO₂ NPs < TiO₂ PNSs < TiO₂ NTs.

To this end, the Na⁺ adduct of reduced GSH was easily detected with TiO₂ PNSs and TiO₂ NTs, and showed no signal with NPs.

Literature highlights that cationization by sodium and potassium is the predominant ionization mechanism for amino acids in SALDI MS analysis, whereas negative ions could be obtained by the abstraction of a proton.²⁶ This represents the advantage of the used method, due to the fact that the spectra are much simpler for interpretation because no adducts could be formed with the matrix.

Furthermore, ions arising from all five carbohydrates were detected with TiO₂ NTs in both ion modes, and Ara was observed only with this substrate, whereas no signals were detectable with others. TiO₂ NPs appeared to be good only for the detection of Raff in positive ion mode. Briefly, in positive ion mode with the assistance of all three substrates, we detected only alkali adducts for all analyzed molecules, i.e. [M + Na]⁺ and [M + K]⁺, whereas negative ion mode spectra of carbohydrates were more easily interpreted and TiO₂ NTs were demonstrated to be a good substrate. Only the signals arising from β-CD were rather low.

As compared to MALDI MS of the tested compounds (data not shown), SALDI MS with TiO₂ substrates gave better sensitivity and reproducibility in negative-ion mode, at least in the case of the investigated biological molecules. The background generated from TiO₂ NTs is very low, making the TiO₂ NTs substrate suitable for the analysis of small molecules in positive ion mode (Fig. 1a), whereas in negative ion mode it exhibited a slightly higher background (Fig. 1b).

As the obtained data suggests, some molecules (T, PRG, L-Ala, L-Cys, and Raff) could be analyzed with TiO₂ nanocrystals with satisfying detectability.

It is possible that when the substrate concentration increases, for instance a higher concentration of TiO₂ NPs, the number of nanoparticles around the analyzed molecules augments as well, resulting in their better detectability.²⁷ This occurs due to sufficient energy from the laser and thus, high ionization efficiency. As a side effect of the increased concentration of the substrate, a high background noise might be generated and poor sensitivity is usually obtained. In accordance with obtained results, we have decided to limit the concentration of NPs to gain a satisfactory response with the low background signal.

Homogeneity of analyte/substrate distribution on the sample plate

The homogeneity of the sample/matrix co-crystals in MALDI and the distribution of the analyte over the substrate surface in the SALDI approach, define the reproducibility and accuracy of the measurement.^1,2 Therefore, in the next set of experiments, the homogeneity of our systems was monitored, and as illustrated in Fig. S3,† one sample spot was divided into eight sectors. For calculation and presentation of homogeneity, we used the mean value of the coefficient of variation of eight S/N values with nine repetitions within three days for each analyte/substrate combination in positive and negative ion mode (Table 2 and 3). These values were used as homogeneity measures for different analyte/substrate combinations.

Table 2 Homogeneity of samples on the SALDI plate in positive ion mode

Molecule	m/z	RSD, %
		NPs	PNSs	NTs
E2	407.6	—	9.1–21.3	10.4–18.7
T	401.6	20.3–61.3	—	27.3–45.8
	423.6	22.1–56.5	6.0–16.3	13.2–27.8
PRG	315.5	59.6–107.0	—	—
	337.5	65.0–94.0	7.9–21.9	11.3–24.2
L-Ala	90.1	10.6–24.2	—	—
	112.1	9.5–17.1	—	5.5–17.5
	128.2	10.9–17.1	—	14.0–25.2
L-Cys	121.2	9.7–38.8	—	—
	144.2	5.7–27.3	—	31.8–21.7
	167.1	—	—	11.3–34.0
DL-Met	172.2	—	6.6–14.7	5.9–38.9
	188.3	—	—	5.4–27.9
	195.2	—	6.4-18.6	7.1–48.5
GSH	308.3	—	7.2–38.8	—
	330.3	—	11.5–78.4	6.2–17.6
	346.4	—	—	8.2–31.01
D-(+)-Glu	203.2	—	3.3–20.9	14.4–45.3
	219.3	—	—	11.1–56.7
D-(+)-Malt	365.3	—	6.0–17.4	9.8–28.1
β-CD	1158.0	—	9.1–17.8	17.8–48.1
	1174.1	—	—	11.3–54.5
Raff	185.1	—	5.4–23.1	16.1–65.5
	203.2	21.3–57.2	6.4–25.1	23.9–65.2
	365.3	32.3–74.8	5.4–12.6	5.3–29.4
	384.3	—	4.4–18.2	—
	527.4	32.3–62.8	7.7–14.8	9.8–30.7
	543.5	29.6–58.2	—	18.8–49.9
Ara	173.1	—	—	8.2–26.4
	189.2	—	—	18.4–39.3

---

Table 3 Homogeneity of samples on the SALDI plate in negative ion mode

Molecule	m/z	RSD, %
		NPs	PNSs	NTs
E2	328.9	—	5.3–27.3	11.0–30.8
T	399.6	—	—	16.5–34.0
PRG	313.5	—	—	18.6–36.1
L-Ala	88.1	3.8–13.5	—	5.0–27.3
L-Cys	120.2	—	—	6.5–19.8
DL-Met	148.2	—	8.9–17.2	11.0–29.3
GSH	306.3	—	—	16.5–27.8
D-(+)-Glu	179.2	—	—	14.2–25.1
D-(+)-Malt	222.0	—	—	6.3–37.1
Raff	222.0	—	6.3–14.6	14.4–32.7
	503.4	—	12.6–33.2	26.8–63.0
Ara	149.1	—	—	16.8–41.8

---

The best homogeneity was achieved with TiO₂ PNSs; the values of relative standard deviation (RSD) were around 20%, except for GSH, wherein TiO₂ NTs obtained better results and quite satisfactory values of RSD (6.2–17.6%). TiO₂ NPs showed the worst homogeneity with analyzed molecules, although L-Ala (m/z 112.1) and L-Cys (m/z 144.2) were exceptions with achieved RSD of 9.5–17.1% and 5.7–27.3%, respectively. In the last combination, the best mixing was achieved among all samples studied with TiO₂ NPs.

In negative ion mode (Table 3), the molecules were detectable in great majority with TiO₂ NTs, as discussed previously. However, three analytes, which are detectable with TiO₂ PNSs, gave better homogeneity compared to the other analytes: E2 (m/z 328.9, RSD 5.3–27.3%), Raff (m/z 222.0, RSD 6.3–14.6%, m/z 503.4, RSD 12.6–33.2%), and DL-Met (m/z 148.2, RSD 8.9–17.2%). The only one detectable with TiO₂ NPs was L-Ala (m/z 88.1) along with much better reproducibility (3.8–13.5%). The negative ion mode in MALDI TOF MS was far less sensitive than that in SALDI TOF MS. The matrix by itself yields a number of peaks, especially in negative ion mode when the detection of small molecules is very difficult.²⁸

Day-to-day repeatability

To compare the accuracy of measurements with respect to signal intensity and potential use of tested systems for quantitative analysis, within-day precision and day-to-day repeatability were calculated.

In Tables 4 and 5, calculated F values for all combinations are shown (F ratio is the ratio of two mean square values. If the null hypothesis is true, the calculated F value does not exceed F_crit. For eight measurements, F_crit is 5.1. A large F ratio (higher than F_crit) means that the variation in the group average values is more than one expects to observe by chance). Day-to-day variation exceeded within-day variation in those cases wherein the calculated F value exceeds F critical value (F_crit = 5.1). Under applied experimental conditions, within-day variation for the most analyte/substrate pairs was very good (less than 10%) (Tables 4 and 5) when compared to literature data.^6,27,29 The best results obtained for molecule/substrate pairs in positive ion mode were T (m/z 423.6)/PNSs – RSD 2.1% and Raff (m/z 365.3)/PNSs RSD – 2.7%. In negative ion mode, the best results were obtained for L-Ala (m/z 88.1)/NPs – RSD 2.9%.

Table 4 The values of variations in day-to-day and within-day repeatability in the positive ion mode

Molecule	m/z	ANOVA day-to-day, %			ANOVA within-day, %			F, Fcrit 5.143
		NPs	PNSs	NTs	NPs	PNSs	NTs	NPs	PNSs	NTs
E2	407.6	—	7.4	8.1	—	13.3	14.5	—	1.9	0.1
T	401.6	26.4	—	33.5	53.5	—	13.9	0.3	—	18.5
	423.6	17.3	12.9	10.1	33.0	2.1	7.0	0.2	115.6	7.3
PRG	315.5	19.6	—	—	36.6	—	—	1.9	—	—
	337.5	58.4	20.5	8.8	101.3	9.7	11.4	0.0	14.5	2.8
L-Ala	90.1	4.1	—	—	28.0	—	—	0.9	—	—
	112.1	9.7	—	2.4	26.2	—	7.5	0.6	—	1.3
	128.2	11.7	—	27.3	27.2	—	7.0	0.4	—	46.5
L-Cys	121.2	14.2	—	—	34.6	—	—	1.5	—	—
	144.2	16.2	—	4.5	32.6	—	20.1	0.3	—	0.9
	167.1	—	—	7.3	—	—	20.8	—	—	0.6
DL-Met	172.2	—	13.4	9.4	—	13.8	12.6	—	3.9	2.7
	188.3	—	—	16.3	—	—	13.5	—	—	5.4
	195.2	—	30.2	21.1	—	5.3	13.2	—	98.1	8.6
GSH	308.3	—	33.1	—	—	27.6	—	—	5.1	—
	330.3	—	45.4	6.4	—	62.0	12.2	—	2.6	0.2
	346.4	—	—	29.0	—	—	20.7	—	—	6.9
D-(+)-Glu	203.2	—	5.7	18.2	—	10.2	25.9	—	0.0	2.5
	219.3	—	—	8.0	—	—	23.9	—	—	1.3
D-(+)-Malt	365.3	—	4.9	18.3	—	9.9	13.0	—	0.3	7.0
β-CD	1158.0	—	12.3	16.5	—	22.0	32.2	—	1.3	0.2
	1174.1	—	—	35.9	—	—	29.6	—	—	5.4
Raff	185.1	—	18.9	18.7	—	16.3	38.6	—	5.0	0.3
	203.2	39.0	14.0	19.4	43.9	11.5	34.7	3.4	5.4	0.1
	365.3	22.5	14.1	5.8	53.5	2.7	16.6	1.5	81.3	0.6
	384.3	—	13.5	—	—	9.0	—	—	7.7	—
	527.4	25.5	11.0	6.6	57.8	7.7	16.2	1.6	7.0	0.5
	543.5	24.7	—	24.3	36.6	—	31.3	2.4	—	2.8
Ara	173.1	—	—	16.7	—	—	8.9	—	—	11.5
	189.2	—	—	18.6	—	—	26.3	—	—	2.5

---

Table 5 The values of variations in day-to-day and within-day repeatability in negative ion mode

Molecule	m/z	ANOVA day-to-day, %			ANOVA within-day, %			F, Fcrit 5.143
		NPs	PNSs	NTs	NPs	PNSs	NTs	NPs	PNSs	NTs
E2	328.9	—	14.7	28.0	—	7.2	15.2	—	13.6	11.2
T	399.6	—	—	35.8	—	—	18.3	—	—	12.5
PRG	313.45	—	—	20.0	—	—	16.1	—	—	5.6
L-Ala	88.1	3.7	—	10.0	2.9	—	7.3	6.0	—	6.7
L-Cys	120.2	—	—	7.9	—	—	9.7	—	—	3.0
DL-Met	148.2	—	15.9	9.8	—	9.1	11.3	—	10.2	3.2
GSH	306.3	—	—	5.7	—	—	21.5	—	—	1.2
D-(+)-Glu	179.2	—	—	21.6	—	—	14.1	—	—	8.0
D-(+)-Malt	222.0	—	—	26.3	—	—	25.3	—	—	4.3
Raff	222.0	—	7.8	6.0	—	5.2	10.4	—	7.7	0.0
	503.4	—	3.7	16.7	—	10.5	30.3	—	1.4	0.1
Ara	149.1	—	—	5.9	—	—	30.0	—	—	1.1

---

Based on the abovementioned results, TiO₂ PNSs and TiO₂ NTs are potential matrices for quantitative determination of molecules.

In general, the best agreement of values for day-to-day variations of the same molecule/substrate pair was achieved with TiO₂ PNSs, with RSD values lower than 25%. This repeatability was excellent and also in accordance with data for combination of catechin-modified TiO₂ NPs and steroid hormones wherein the batch-to-batch variations were less than 15%, n = 7.²⁹

TiO₂ NTs as potential substrates for quantitative analysis of molecules in positive ion mode obtained minimal values of day-to-day variation for particular molecules. The variation for most of the analyzed molecules with TiO₂ NTs was less than 10% (for example: Raff (m/z 365.3) – RSD 5.8%, L-Ala (m/z 112.1) – RSD 2.4%), whereas for other molecules, RSD was around 20%. Similar results were presented in the study by Lee et al.,⁶ wherein the within-day variation was less than 10% over 50 spots in the same sample, and day-to-day variation was less than 15% for three different batches (each for 50 spots). These results also confirm that TiO₂ NTs are very good substrates for the detection and determination of amino acids.

In negative ion mode, the smallest values for day-to-day variation with TiO₂ NTs were achieved for GSH (m/z 306.3) – RSD 5.7%, Ara (m/z 149.1) – RSD 5.9% and Raff (m/z 222.0) – RSD 6.0%. The only sample detected in negative mode with TiO₂ NPs that had very low values of day-to-day variation and within-day variation was L-Ala, with RSD 3.7% and 2.9%, respectively. When TiO₂ PNSs was used as substrate, RSD values for day-to-day variations were from 3.7% (Raff) to 15.9% (DL-Met), and for within-day were from 5.2% (Raff) to 9.1% (DL-Met).

Limits of detection

Finally, the limits of detection (LODs) of analyte/substrate combinations were investigated. Under optimal conditions, we obtained a calibration curve of S/N values for all twelve analyzed molecules against their concentrations in the range of 0.02–112 mM. LODs and the linear range of analytes are summarized in Table 6 for positive ion mode and Table 7 for negative ion mode. The correlation coefficients (R²) in the linear range were from 0.637 to 0.999. The majority of the obtained correlation coefficients were around 0.9, but for some peaks the value was even lower. TiO₂ PNSs obtained the best results for the correlation coefficients.

Table 6 Limits of detection, linearities and correlation coefficients for all analyzed molecules with TiO₂ substrates in positive ion mode^a

Molecule	m/z positive ion mode	LOD mM			Linear range mM			R^2
		NPs	PNSs	NTs	NPs	PNSs	NTs	NPs	PNSs	NTs
a —This signal cannot be detected with this TiO2 substrate, * it cannot be calculated from the graph.
E2	407.6	3.11	6.12	0.45	*	0.3–41.0	0.3–1.6	*	0.955	0.866
	423.5	2.71	1.15	0.29	*	0.3–4.1	0.3–1.6	*	0.849	0.938
	439.7	3.05	1.48	0.30	*	0.3–8.2	0.3–1.6	*	0.925	0.936
T	401.6	0.47	—	5.11	*	—	0.1–20.0	*	—	0.816
	423.6	5.28	0.31	2.35	0.5–50.0	0.04–2.0	2.0–20.0	0.974	0.927	0.987
PRG	315.5	3.33	0.23	0.46	1.0–10.0	*	0.5–10.0	0.833	*	0.985
	337.5	2.80	2.15	0.46	0.5–10.0	0.5–10.0	0.5–10.0	0.837	0.910	0.985
L-Ala	90.1	2.81	—	—	0.6–56.1	—	—	0.993	—	—
	112.1	8.98	2.28	0.06	2.2–56.1	0.6–11.2	0.6–2.2	0.962	0.918	0.999
	128.2	11.02	—	0.55	0.6–56.1	—	0.6–2.2	0.899	—	0.917
L-Cys	122.2	16.05	—	35.22	4.1–82.5	—	4.1–123.8	0.925	—	0.839
	144.2	4.22	26.62	0.69	4.1–41.3	8.3–123.8	4.1–16.5	0.981	0.908	0.997
	167.1	—	—	5.12	—	—	4.1–16.5	—	—	0.872
DL-Met	150.2	1.39	—	0.10	6.7–33.5	—	0.3–3.4	0.998	—	0.998
	172.2	3.46	1.07	0.96	6.7–33.5	0.3–13.4	0.3–3.4	0.987	0.984	0.869
	188.3	1.61	—	0.78	6.7–33.5	—	0.3–3.4	0.997	—	0.910
	195.2	—	0.64	0.87	—	0.3–3.4	0.3–3.4	—	0.938	0.890
GSH	308.3	3.55	—	3.74	6.5–32.5	—	0.7–16.3	0.984	—	0.896
	330.3	5.43	0.09	0.32	3.3–32.5	3.3–16.3	0.2–1.6	0.952	0.999	0.937
	346.4	2.19	7.94	0.56	6.5–32.5	6.5–32.5	0.3–3.2	0.994	0.925	0.950
D-(+)-Glu	203.1	4.85	0.42	0.57	5.6–55.5	0.3–2.8	0.3–5.6	0.986	0.960	0.978
	219.3	3.73	0.12	1.04	5.6–55.5	0.3–2.8	0.3–5.6	0.992	0.997	0.929
D-(+)-Malt	365.3	7.75	0.08	0.84	5.6–27.8	0.1–2.8	0.1–2.8	0.935	0.968	0.881
β-CD	1158.0	0.83	—	0.17	0.9–8.8	—	*	0.984	—	*
	1174.1	2.29	—	0.03	0.9–8.8	—	*	0.891	—	*
Raff	185.1	3.34	0.04	0.03	1.0–9.9	0.1–0.4	0.1–0.4	0.908	0.988	0.994
	203.1	2.67	0.09	0.02	1.0–9.9	0.1–1.0	0.1–0.4	0.883	0.931	0.997
	365.3	17.73	0.22	0.70	9.9–39.7	0.1–1.0	0.1–0.4	0.637	0.921	0.957
	384.3	—	0.18	0.60	—	0.1–1.0	0.4–2.0	—	0.943	0.999
	527.4	10.97	0.24	0.70	9.9–39.7	0.1–1.0	0.1–0.4	0.821	0.907	0.962
	543.5	1.28	0.02	0.10	*	0.2–1.0	0.1–0.4	*	0.999	0.910
Ara	173.1	4.9	0.70	0.57	0.3–33.3	0.3–6.7	0.3–3.3	0.941	0.977	0.960
	189.2	3.66	1.38	0.5	0.3–33.3	0.3–6.7	0.3–3.3	0.966	0.916	0.949

---

Table 7 Limits of detection, linearities and correlation coefficients for all analyzed molecules with TiO₂ substrates in negative ion mode^a

Molecule	m/z negative ion mode	LOD mM			Linear range mM			R^2
		NPs	PNSs	NTs	NPs	PNSs	NTs	NPs	PNSs	NTs
a —This signal cannot be detected with this TiO2 substrate.
E2	328.9	—	6.64	0.36	—	0.3–41.0	0.3–1.6	—	0.921	0.909
T	399.6	—	3.24	1.45	—	0.5–10.0	1.0–5.0	—	0.815	0.902
PRG	313.5	0.82	0.77	0.20	1.0–5.0	0.5–20.0	0.5–5.0	0.967	0.987	0.997
L-Ala	88.1	0.55	1.88	0.26	0.6–5.6	0.6–5.6	0.6–2.2	0.983	0.829	0.981
L-Cys	120.1	10.00	16.83	7.54	4.1–123.8	8.3–82.5	4.1–16.5	0.989	0.930	0.759
DL-Met	148.2	3.12	0.36	0.23	6.7–33.5	0.3–3.4	0.3–1.3	0.997	0.979	0.958
GSH	306.3	6.10	1.02	0.02	6.5–48.8	3.3–16.3	0.2–1.6	0.969	0.995	0.999
D-(+)-Glu	179.2	23.00	0.89	0.54	5.6–55.5	0.3–5.6	0.3–5.6	0.763	0.948	0.980
D-(+)-Malt	222.0	12.95	0.45	0.11	2.8–41.6	0.1–2.8	0.1–2.8	0.826	0.962	0.998
Raff	222.0	3.97	0.22	0.21	1.0–9.9	0.4–2.0	0.1–1.0	0.972	0.983	0.927
	503.4	—	0.08	0.73	—	0.4–2.0	0.4–2.0	—	0.998	0.842
Ara	149.1	8.41	1.12	1.61	0.3–66.6	0.3–6.7	0.3–13.3	0.974	0.943	0.965

---

In positive ion mode (Table 6), the minimal values of LODs for L-Ala, Ara, β-CD, L-Cys, E2, GSH, DL-met, PRG, and Raff were achieved with TiO₂ NTs in the range of 0.02–0.69 mM. TiO₂ PNSs gave low values of LODs with D-(+)-Glu, GSH, DL-Met, PRG, and Raff in the range of 0.02–0.64 mM. The sensitivity of detection of these compounds might however be increased with modified nanoparticles, such as catechin modified TiO₂ nanoparticles,²⁹ but this is beyond the scope of the present study.

Because of the poorer ionization efficiencies of Cys, relative to GSH, and greater noise in the low-mass region (m/z < 200), their LODs were higher than those of GSH.²⁷

In negative ion mode (Table 7), both TiO₂ PNSs and TiO₂ NTs obtained satisfactory LOD values for most of the analyzed molecules; however, TiO₂ NTs obtained the smaller values in the range of 0.02 for GSH to 7.54 mM for L-Cys. Similar to that observed in the positive ion mode, the LOD for L-Cys is higher than that of GSH for all substrates. TiO₂ NPs among the analyzed molecules were the most suitable for detection of PRG and L-Ala with LODs of 0.82 and 0.55 mM, respectively.

Conclusions

We have investigated the possibility of using TiO₂ nanocrystals of different shapes and sizes as substrates for the SALDI TOF mass spectrometric analysis of selected physiologically relevant small molecules such as steroid hormones, amino acids and carbohydrates. According to the obtained data, all TiO₂ nanocrystals could be used as substrates regardless of their size and shape. However, there are variations in spectral quality with respect to the homogeneity, repeatability, sensitivity and detection limit, and in the potential for quantitative MS analysis.

In conclusion, our results show the following:

- Larger titania based nanocrystals (PNSs and NTs) produced spectra with much higher reproducibility and ease of interpretation.

- Steroid hormones were best detected when TiO₂ PNSs were used as substrates.

- Amino acids, GSH and carbohydrates were detectable both as positive and negative ions with TiO₂ NTs.

- The best homogeneity was achieved with TiO₂ PNSs for most of the analyte molecules (exception was GSH wherein TiO₂ NTs obtained better results).

- In general TiO₂ NTs have the highest potential for quantification of the most molecules in positive ion mode.

Acknowledgements

This study was supported by the Serbian Ministry of Education, Science and Technological Development, Grant No. 172011 and 172056. Authors are thankful to Dr Dunja Drakulić from the Laboratory of Molecular Biology and Endocrinology from the VINČA Institute of Nuclear Sciences, University of Belgrade, for the critical reading of the manuscript.

Notes and references

1. L. H. Cohen, F. Li, E. P. Go and G. Siuzdak, in MALDI MS: A Practical Guide to Instrumentation, Methods and Applications, ed. F. Hillenkamp and J. Peter-Katalinić, Wiley, Münster, 1nd edn, 2007, vol. 1, ch. 9, pp. 299–337.
2. P. A. Kuzema, J. Anal. Chem., 2011, 66, 1227–1242.
3. J. Sunner, E. Dratz and Y.-C. Chen, Anal. Chem., 1995, 67, 4335–4342.
4. R. Arakawa and H. Kawasaki, Anal. Sci.: Int. J. Jpn. Soc. Anal. Chem., 2010, 26, 1229–1240.
5. G. L. Hortin, Clin. Chem., 2006, 52, 1223–1237.
6. K.-H. Lee, C.-K. Chiang, Z.-H. Lin and H.-T. Chang, Rapid Commun. Mass Spectrom., 2007, 21, 2023–2030.
7. C.-T. Chen and Y.-C. Chen, Rapid Commun. Mass Spectrom., 2004, 18, 1956–1964.
8. D. S. Peterson, Mass Spectrom. Rev., 2007, 26, 19–34.
9. C. Dass, in Fundamentals of Contemporary Mass Spectrometry, ed. W. J. Pesce and P. B. Wiley, Wiley, New Jersey, 1st edn, 2007, vol. 1, ch. 2, pp. 15–60.
10. M. Radisavljević, T. Kamčeva, I. Vukićević, M. Radoičić, Z. Šaponjić and M. Petković, Rapid Commun. Mass Spectrom., 2012, 26, 2041–2050.
11. I. Popović, M. Nešić, M. Nišavić, M. Vranješ, T. Radetić, Z. Šaponjić, R. Masnikosa and M. Petković, Mater. Lett., 2015, 150, 84–88.
12. T. Rajh, D. M. Tiede and M. C. Thurnauer, J. Non-Cryst. Solids, 1996, 205–207, 815–820 , Part 2.
13. R. C. Thompson, Inorg. Chem., 1984, 23, 1794–1798.
14. T. Kasuga, M. Hiramatsu, A. Hoson, T. Sekino and K. Niihara, Adv. Mater., 1999, 11, 1307–1311.
15. M. Radoičić, Z. Šaponjić, I. Janković, G. Ćrić-Marjanović, S. Ahrenkiel and M. Čomor, Appl. Catal., B, 2013, 136–137, 133–139.
16. M. Vranješ, Z. Šaponjić, L. Živković, V. Despotović, D. Šojić, B. Abramović and M. Čomor, Appl. Catal., B, 2014, 160–161, 589–596.
17. E. B. Drake, V. W. Henderson, F. Z. Stanczyk, C. A. McCleary, W. S. Brown, C. A. Smith, A. A. Rizzo, G. A. Murdock and J. G. Buckwalter, Neurology, 2000, 54, 599–603.
18. B. C. Fauser, J. S. Laven, B. C. Tarlatzis, K. H. Moley, H. O. Critchley, R. N. Taylor, S. L. Berga, P. G. Mermelstein, P. Devroey, L. Gianaroli, T. D'Hooghe, P. Vercellini, L. Hummelshoj, S. Rubin, A. J. Goverde, V. De Leo and F. Petraglia, Reprod. Sci., 2011, 18, 702–712.
19. B. J. Mills, M. M. Weiss, C. A. Lang, M. C. Liu and C. Ziegler, J. Lab. Clin. Med., 2000, 135, 396–401.
20. C. Carru, A. Zinellu, S. Sotgia, R. Serra, M. F. Usai, G. F. Pintus, G. M. Pes and L. Deiana, Biomed. Chromatogr., 2004, 18, 360–366.
21. P. Ghezzi, B. Romines, M. Fratelli, I. Eberini, E. Gianazza, S. Casagrande, T. Laragione, M. Mengozzi, L. A. Herzenberg and L. A. Herzenberg, Mol. Immunol., 2002, 38, 773–780.
22. N. Sattar, O. Scherbakova, I. Ford, D. S. O'Reilly, A. Stanley, E. Forrest, P. W. Macfarlane, C. J. Packard, S. M. Cobbe and J. Shepherd, Diabetes, 2004, 53, 2855–2860.
23. R. A. Kruse, S. S. Rubakhin, E. V. Romanova, P. W. Bohn and J. V. Sweedler, J. Mass Spectrom., 2001, 36, 1317–1322.
24. T. Watanabe, K. Okumura, H. Kawasaki and R. Arakawa, J. Mass Spectrom., 2009, 44, 1443–1451.
25. A. L. Linsebigler, G. Lu and J. T. Yates, Chem. Rev., 1995, 95, 735–758.
26. S. Nitta, H. Kawasaki, T. Suganuma, Y. Shigeri and R. Arakawa, J. Phys. Chem. C, 2013, 117, 238–245.
27. N.-C. Chiang, C.-K. Chiang, Z.-H. Lin, T.-C. Chiu and H.-T. Chang, Rapid Commun. Mass Spectrom., 2009, 23, 3063–3068.
28. M. Petković, J. Schiller, M. Müller, S. Benard, S. Reichl, K. Arnold and J. Arnhold, Anal. Biochem., 2001, 289, 202–216.
29. T.-C. Chiu, Talanta, 2011, 86, 415–420.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra20042c

---