Boronic Acid-Modified Magnetic Fe3O4@mTiO2 Microspheres for Highly Sensitive and Selective Enrichment of N-Glycopeptides in Amniotic Fluid

Although mesoporous materials and magnetic materials are used to enrich glycopeptides, materials sharing both mesoporous structures and magnetic properties have not been reported for glycopeptide analyses. Here we prepared boronic acid-modified magnetic Fe3O4@mTiO2 microspheres by covalent binding of boronic acid molecules onto the surfaces of silanized Fe3O4@mTiO2 microspheres. The final particles (denoted as B-Fe3O4@mTiO2) showed a typical magnetic hysteresis curve, indicating superparamagnetic behavior; meanwhile, their mesoporous sizes did not change in spite of the reduction in surface area and pore volume. By using these particles together with conventional poly(methyl methacrylate) (PMMA) nanobeads, we then developed a synergistic approach for highly specific and efficient enrichment of N-glycopeptides/glycoproteins. Owing to the introduction of PMMA nanobeads that have strong adsorption towards nonglycopeptides, the number of N-glycopeptides detected and the signal-to-noise ratio in analyzing standard proteins mixture both increased appreciably. The recovery of N-glycopeptides by the synergistic method reached 92.1%, much improved than from B-Fe3O4@mTiO2 alone that was 75.3%. Finally, we tested this approach in the analysis of amniotic fluid, obtaining the maximum number and ratio of N-glycopeptides compared to the use of B-Fe3O4@mTiO2 alone and commercial SiMAG-boronic acid particles. This ensemble provides an interesting and efficient enrichment platform for glycoproteomics research.

As a biologically broad and significant post-translational modification, protein glycosylation is involved in many physiological activities and disease states, such as protein folding, cell division, intracellular secretion, inflammation, and congenital disorders [1][2][3][4] . Moreover, it has been discovered that more than half cancer biomarkers are glycosylated peptides or proteins 5 . In the present glycoproteome research, the discovery of glycosylation site occupancy and identification of glycan heterogeneity at each glycosite have been largely undertaken by the use of mass spectrometry (MS) or tandem MS/MS 6,7 . Glycoproteins and glycopeptides in real samples are much lower in amount compared to nonglycosylated ones, and besides, their ionization efficiency is rather poor, both of which result in negative signal suppression in MS analysis. Therefore, in order to obtain high-resolution profiling of endogenous glycoproteins in serum or tissues, use of efficient strategies for specific isolation and enrichment of the targets is indispensable.
For the past years, methods for glyco-specific enrichment divide into several categories by means of different mechanisms, including lectin affinity 8,9 , size exclusion 10 , hydrazide chemistry 11,12 , hydrophilic interaction 13,14 , and boronic acid-derived matrixes [15][16][17][18][19] , with the last one receiving more attention recently. With phenylboronic acids typically employed, the boronic acid moiety can form cyclic ester with cis-diol group of glycoconjugates in an alkaline medium and at acidic pH the ester dissociates 19 , making boronic acid a unique ligand for reversibly collecting and detaching glycopeptides. Besides, this method isolates both N-and O-glycopeptides in an unbiased manner, thus complementing some limitations of other methods, such as biased glycol-enrichment with respect to lectin affinity and hydrazide chemistry, and insufficient selectivity and recovery related to hydrophilic interaction. Several types of materials have been developed to conjugate boronic acid groups for glycopeptide analyses, including agarose resin, mesoporous silica 15 , polymer particles 16 , magnetic particles [18][19][20][21][22][23][24] , carbon nanotubes 25 , and graphene oxide 26 . Each type of material has its own merit. Boronic acid-agarose resin has already been commercialized, boronic acid-mesoporous silica shows large pore volume and pore size, boronic acid-magnetic particles display facile separation by external magnetic field, and boronic acid-carbon nanotubes or graphene oxide possess exceptionally large specific surface area and high density of boronic acid groups.
We wish to confer mesoporous structure combined with magnetic property to the boronic acid-modified composite for the enrichment of glycopeptides. Mesoporous structure can bring about large surface area and tunable pore size for the flux of targets, and meanwhile, the response to magnetic field enables convenient separation from the complex matrixes in real applications. To our best knowledge, there is no report on the enrichment of glycopeptides by such kind of material. Recently, Ma et al. developed a method to prepare magnetic core/shell Fe 3 O 4 @mTiO 2 microspheres for highly efficient enrichment of phosphopeptides 27 . The microsphere owns the features which meet well with what we entail, including a mesoporous crystalline TiO 2 layer ensuring a large absorption capacity and a high mass transport efficiency, and a Fe 3 O 4 colloidal cluster core with excellent magnetic response. After post-functionalization with boronic acid group, we anticipate the composite to capture and separate glycopeptides efficiently. In order to obtain a better selectivity, we employed a synergistic strategy by adopting conventional PMMA nanobeads as the second enriching material to reduce the effect of nonglycopeptides 28 . Finally, we tested the method in the glycopeptides analysis of human amniotic fluid samples with remarkable results.

Preparation of
(3) Obtain Fe 3 O 4 @mTiO 2 microspheres by hydrothermal treatment of the Fe 3 O 4 @TiO 2 microspheres to form mesoporous TiO 2 shell. The product obtained in the second step was ultrasonically dispersed in ethanol/H 2 O (40 mL/20 mL), followed by the addition of 3 mL of NH 3 · H 2 O. It was transferred into a Teflon-lined stainless-steel autoclave (100 mL capacity) and maintained at 160 °C for 20 h. After cooled down to RT, the product was obtained by magnetic precipitation, washed with ethanol several times, and dried at 60 °C. Such brick-colored material was made up of Fe 3 O 4 @mTiO 2 microspheres. Preparation of PMMA Nanobeads. The PMMA nanobeads were prepared according to a previous method 29 . 5 g of MMA monomer and 0.15 g (NH 4 ) 2 S 2 O 8 were added to 65 mL of water and reacted at 75 °C for 4 h with magnetic stirring. The obtained PMMA nanobeads were centrifuged at a speed of 6,000 rpm, sufficiently rinsed with water, and dried at 80 °C.

Sample Preparation. Amniotic fluid samples were obtained from Nanjing Maternity and Child Health
Care Hospital with written consent and approval of the ethics board. The experimental protocols were approved by Nanjing Medical University. All methods were performed in accordance with the relevant guidelines and regulations. All amniotic acid specimens (8-10 mL) from women at 16-18 weeks of gestation, carrying out prenatal diagnosis mostly due to advanced maternal age ranging from 30 to 45 years, were centrifuged at 12,000 rpm for 15 min at 4 °C to remove insoluble debris after thawing on ice. Afterwards, the cell-free supernatants were vacuum-dried with a SpeedVac system (RVC 2-18, Marin Christ, Osterode, Germany). Then they were added with acetone (1:4, V/V), briefly vortexed, and incubated at −20 °C for 60 min. After centrifugation, the sediments were collected and dissolved in ABC buffer (50 mM), and the protein concentrations were measured by using the Pierce BCA protein assay kit (Rockford, IL, USA). Finally, they were stored at −80 °C for further LC-MS/MS analysis.

Digestion of Standard Proteins and Protein Mixture from Amniotic Fluid. The standard proteins,
i.e. HRP, MYO, and fetuin, were each dissolved in ABC buffer (50 mM) at 1.0 mg/mL, and denatured at 100 °C for 5 min. After cooled down, trypsin was added at an enzyme-to-protein ratio of 1: 40 (w/w) for hydrolysis overnight at 37 °C, respectively.
For the digestion of amniotic fluid, proteins of amniotic fluid (1 μL) were reduced with 10 mM of DTT for 30 min at 60 °C and alkylated with 20 mM of IAA for 30 min at 37 °C in dark. Then trypsin digestion was applied to the sample, as described above. Finally, desalting of the sample was conducted on C18 columns before it was stored at −20 °C for further use.
Synergistic Enrichment of N-Glycopeptides. The procedure is illustrated in Fig. 1b. The digestion products of standard proteins and peptides mixture from amniotic fluid were diluted with ABC buffer (50 mM) to 200 μL, followed by the addition of 10 μL of B-Fe 3 O 4 @mTiO 2 suspension (10 mg/mL) and 40 μL of PMMA nanobeads (10 mg/mL). After shaking at RT for 1 h, a magnet was used to separate the glycopeptides-captured

MALDI-TOF/TOF MS Analysis.
For the analysis of enriched glycopeptides, 1 μL of eluate was deposited on a MALDI plate, and then 1 μL of DHB matrix (12.5 mg/mL in ACN/H 2 O/TFA, 50:49.9:0.1 by volume) was spotted onto 600 μm anchorchips (Bruker Daltonics, Bremen, Germany). The Bruker peptide calibration mixture was spotted for external calibration. MALDI-TOF/TOF MS was carried out on a time-of-flight Ultraflex Extreme mass spectrometer (Bruker Daltonics, Bremen, Germany). Peptide mass maps were acquired in positive reflection mode, averaging 800 laser shots per spectrum. Resolution was 15000-20000. The Bruker calibration mixtures were used to calibrate the spectrum to a mass tolerance within 0.1 Da. Each acquired mass spectrum (m/z range 1000-5000) was processed using the software FlexAnalysis v.2.4 supplied by Bruker Daltonics. The peak detection algorithm was SNAP (Sort Neaten Assign and Place), signal-to-noise (S/N) threshold was 3, and the quality factor threshold was 50. Characterization. Transmission electron microscopy (TEM) was carried out on a JEOL-2100F transmission electron microscope operating at 200 kV. Scanning electron microscopy (SEM) was carried out on a Zeiss Supra 40 field-emission scanning microscope at an acceleration voltage of 5 kV. N 2 adsorption-desorption analyses were conducted using a Micrometritics ASAP 2020 accelerated surface area analyzer at 77 K, using Barrett-Emmett-Teller (BET) calculations for the surface area. Before measurements, the samples were degassed in a vacuum at 120 °C for at least 6 h. Fourier transform infrared (FTIR) spectra were measured on a Bruker Vector-22 FTIR spectrometer from 4000 to 400 cm −1 at room temperature. Power X-ray diffraction (PXRD) data were recorded on a Philips X'Pert PRO SUPER X-ray diffractometer equipped with graphite-monochromatized Cu Kα radiation. X-ray photoelectron spectroscopic (XPS) study was performed on an ESCALAB 250 spectrometer (Thermo-VG Scientific). The magnetization curve was measured with a superconducting quantum interference device (SQUID) magnetometer (Quantum Design MPMS XL).

Results and Discussion
Synthesis and Characterization of B-Fe 3 O 4 @mTiO 2 microspheres. It was found that both the volume ratio of ethanol to water and the amount of ammonium were critical to pore evolution of the TiO 2 shell. To acquire a relatively large pore for the flux of glycopeptides, the volume ratio of ethanol to water was set at 40:20 and 3 mL of NH 3 · H 2 O was applied. We analyzed the size and morphology of the particles by SEM and TEM. From the SEM images (Fig. 2a,d,g), we observed an obvious increase of size for the Fe 3 O 4 @TiO 2 microspheres compared to bare Fe 3 O 4 , and little change of size after their transformation to Fe 3 O 4 @mTiO 2 . From the enlarged SEM images (Fig. 2b,e,h), we could observe that the Fe 3 O 4 @TiO 2 microspheres had smooth surfaces as those for the bare Fe 3 O 4 particles, suggesting a homogeneous shell of amorphous TiO 2 around the magnetic core. This was confirmed by the PXRD study, showing no characteristic TiO 2 crystal peaks (Fig. S1). After hydrothermal reaction, the surfaces of those Fe 3 O 4 @mTiO 2 particles became uneven, which was owing to the formation of crystalline TiO 2 and generation of mesoporous structure. This was further demonstrated by TEM images (Fig. 2c,f,i), with a shell of TiO 2 crystals with sizes falling within 20-30 nm surrounding the core for the Fe 3 O 4 @mTiO 2 microspheres. PXRD also showed the appearance of a new diffraction peak near 25° corresponding to the (101) plane of anatase-phase TiO 2 (Fig. S1) 33 . After modification with boronic acid, the SEM images (Fig. 2j,k) showed a little aggregation of particles, and the TEM result (Fig. 2l) revealed a layer of organic substance adsorbed onto the surfaces of the mesoporous particles.
We identified the successful modification of boronic acid by means of FTIR and XPS. Compared to the Fe 3 O 4 @mTiO 2 microspheres, the sample after reactions possessed several new bands in its FTIR spectrum, as shown in Fig. 3a. A weak band at 2923 cm −1 was ascribed to the -CH 2 absorption from the silane agent APTMS. The strong peaks at 878 and 1440 cm −1 were attributed to benzene ring vibrations from FBBA, and the presence of an obvious peak at 1335 cm −1 clearly corresponded to the B-O stretching 28 . In addition, another strong peak at 1028 cm −1 should be caused by the absorption of Si-O band 34 . The survey XPS spectrum exhibited the presence of C, N, O, B, Fe, and Ti in the sample, the binding energy (BE) for Fe 2p3/2 locating at 710.2 eV and the BE for Ti 2p3/2 at 458.1 eV (Fig. 3b-d). The low intensity of the Fe 2p signal was due to core encapsulation by TiO 2 . The BE for N 1 s was centered at 398.9 eV obviously (Fig. 3e), implying the adsorption of APTMS. Lastly the BE for B 1 s was observed at 192.0 eV (Fig. 3f), which was a solid proof for the binding of FBBA. From the above results, we can say that the modification of Fe 3 O 4 @mTiO 2 microspheres by boronic acid is successful.
We further investigated the mesoporous character of the microspheres before and after boronic acid modification by N 2 sorption analyses. It clearly showed that both particles manifested typical type IV gas sorption isotherms (Fig. 4a), hence implying that the mesoporous structure was not destroyed by the post-modification. Using the BET model for calculations, the specific surface area dropped from 146.0 m 2 /g for Fe 3 O 4 @mTiO 2 to 19.6 m 2 /g for B-Fe 3 O 4 @mTiO 2 , and meanwhile, the pore volume decreased from 0.31 to 0.11 cm 3 /g appreciably. However, from the pore size distribution curves (Fig. 4b), the sizes of the pore cavities kept barely changed, centered at 19.6 and 19.5 nm, respectively. The reduced surface area and pore volume should be explained by the organic grafting in the particles, which blocked some of the tiny slits between the neighboring TiO 2 crystals in the shell. This was also observed in the boronic acid-functionalized mesoporous silica 15 . Despite the reduction in surface area and pore volume, the relative large pore size was still favorable for the permeation of targeted glycopeptides during the enrichment process. Then the magnetic property was assessed, as shown in Fig. 4c. The magnetic hysteresis curve showed the saturation magnetization (Ms) value decreasing from 67 emu/g for the bare Fe 3 O 4 to 16 emu/g for the B-Fe 3 O 4 @mTiO 2 , which was caused by the surface modification. Despite the drastic reduction of Ms value, the superparamagnetic feature of the B-Fe 3 O 4 @mTiO 2 microspheres expedited their efficient separation in 30 s with a magnet. Such a fast response was beneficial to practical applications.

Specific Enrichment of N-Glycopeptides by a synergy of B-Fe 3 O 4 @mTiO 2 Microspheres with
PMMA Nanobeads. The enrichment of N-glycopeptides for the B-Fe 3 O 4 @mTiO 2 particles was first tested with the tryptic digest of HRP, a standard glycoprotein, by MALDI-TOF/TOF MS. In the absence of enriching material, only signals for nonglycopeptides were detected (Fig. 5a), while after enrichment by B-Fe 3 O 4 @mTiO 2 , five peaks were identified corresponding to N-glycopeptides from the digestion of HRP (Fig. 5b). This detected number of glycopeptides is comparable with those by using enriching materials such as FDU-12-GA (5) and Fe 3 O 4 @SiO 2 -APB (3) 15, 28 , but much less than those by using core-satellite composite (17) and APBA-MCNTs (21) 23,25 . We assume that this difference originates from the difference in experimental conditions, including content of boronic acid, instrument, and concentration of HRP used, etc. Apart from the increase in the number of N-glycopeptides detected, the use of B-Fe 3 O 4 @mTiO 2 also lowered the intensities from the nonglycopeptides, thereby improving the signal-to-noise (S/N) ratio of the N-glycopeptides greatly. This result demonstrated the excellent specificity of the B-Fe 3 O 4 @mTiO 2 to N-glycopeptides.  To acquire more number of N-glycopeptides signals, we employed PMMA nanobeads as a second material for enriching nonglycopeptides. With the introduction of PMMA, the number of detected N-glycopeptides increased to 11 (Fig. 5c). Not only this, the peak intensities were much higher with a cleaner background in the mass spectrum, suggesting that the S/N ratio could be dramatically increased. Peak 3 at m/z = 2445 was chosen to evaluate the sensitivity of the method. At the present concentration of HRP (0.1 ng/μL), the S/N ratio was larger than 100, implying that the limit of detection for our method was at the level of 10 pg/μL. This value is comparable with those obtained by using core-satellite composite, Fe 3 O 4 @SiO 2 -APB, and APBA-MCNTs 23,25,28 . All the results revealed the striking advantage of the synergistic enrichment strategy. The detailed sequence information of all the glycopeptides identified is listed in Table S1. A comparable sensitivity of detection was also observed with fetuin as the standard N-glycoprotein (Fig. S2). This suggests that N-glycopeptides from other model proteins can be also enriched by a combination of the B-Fe 3 O 4 @mTiO 2 microspheres and PMMA nanobeads.
We then evaluated the selectivity of the method by mixing the standard N-glycopeptides (from HRP) with the standard nonglycopeptides (from MYO) at a molar ratio 1:100 of HRP to MYO. Direct analysis generated only signals for nonglycopeptides with complicated background in the spectrum (Fig. 6a). By the use of B-Fe 3 O 4 @ mTiO 2 , 8 peaks for N-glycopeptides were detected, and the interfering peaks related to nonglycopeptides vanished completely (Fig. 6b). After the introduction of PMMA nanobeads, 4 more peaks were detected and a much larger S/N ratio was also expected (Fig. 6c). This is owing to the unspecific adsorption of nonglycopeptides by PMMA nanobeads to create more opportunities for B-Fe 3 O 4 @mTiO 2 to interact with targeted peptides. Besides, the traditional washing step was avoided for this approach, thus minimizing the loss of glycopeptides to the least. The binding capacity of this method was measured to be 120 mg/g (Fig. S3). Next the performance of B-Fe 3 O 4 @ mTiO 2 was compared to that of commercial particles SiMAG-boronic acid. Under identical conditions, there were 4 peaks related to N-glycopeptides in the spectrum with using bare SiMAG-boronic acid (Fig. 6d). By combining PMMA nanobeads, 3 more peaks for glycopeptides were detected with a higher sensitivity as well (Fig. 6e). Taken together, we can expect better performance of detection by employing the synergistic enrichment method and the enriching capability of B-Fe 3 O 4 @mTiO 2 is superior to that of commercial SiMAG-boronic acid particles.
We further investigated the recovery of N-glycopeptides. A pre-prepared tryptic digested HRP was divided into two equivalent parts. One was treated with PNGaseF in H 2 18 O to release the glycans, and the other involved capturing of the glycopeptides with B-Fe 3 O 4 @mTiO 2 , eluting them, and then treating with PNGaseF in H 2 16 O. Through mixing the two parts, we could profile the products with MS to comparatively study the abundances of the glycopeptides from different oxygen isotopes according to peak areas. Inferred from the MALDI-TOF results ( Fig. S4 and Table S2), the recovery of N-glycopeptides by the synergistic method was determined to be 92.1%, much improved than from B-Fe 3 O 4 @mTiO 2 alone which was 75.3%. This recovery value exceeds that enriched by a combination of Fe 3 O 4 @SiO 2 -APB with PMMA (90%) 28 , and thus it makes the best among the boronic acid-based methods for enriching glycopeptides to date.
To test the applicability of the new method, amniotic fluid was examined as a model biological sample for identifying the N-glycopeptides and N-glycoproteins. The quantitative and qualitative analysis of amniotic acid may help to identify patients who will develop pregnancy complications or to discover fetal-disease specific markers 35 . After pretreatment, the proteins in the sample were trypsin-digested and incubated with enriching materials for capturing glycopeptides. Three replications using three enriching ensembles were performed to test the applicability and optimize the usage of the presented material in the enrichment of glycopeptides. First, a total of 126 N-linked glycopeptides corresponding to 97 glycoproteins were identified (Table S3). Second, the maximum number and ratio of N-glycopeptides detected were obtained by the synergistic enrichment method (Table S4). 9 more N-glycopeptides and 4 more N-glycoproteins were identified by the combined materials than by the B-Fe 3 O 4 @mTiO 2 alone. We also found that the ratio of enriched glycopeptides obtained by the combined materials was significantly higher than SiMAG-boronic acid using Fisher's exact test (Fold change = 1.5 and P value = 0.04). Besides, the ratio was not significantly improved (P = 0.56) compared to that of using B-Fe 3 O 4 @ mTiO 2 alone, but we did identify more N-glycopeptides using the synergistic method. Hence the use of combined materials showed an enrichment of about 13 folds for N-linked glycopeptides. These results clearly demonstrate the advantage of the synergistic enrichment method over using one enrichment material alone in detecting glycopeptides, especially in complex biological samples.

Conclusion
We prepared boronic acid-modified, mesoporous TiO 2 -coated magnetic Fe 3 O 4 nanoparticles. By combining the B-Fe 3 O 4 @mTiO 2 with PMMA nanobeads, we have developed a new synergistic method for enriching N-glycopeptides specifically. The coverage of boron moieties onto the Fe 3 O 4 @mTiO 2 decreased the specific surface area and pore volume greatly, but did not affect the pore size, thereby maintaining the permeability of the enriching material towards glycopeptides. Although the saturation magnetization value of the material was much diminished than that of bare Fe 3 O 4 , the separation of the material by an external magnet could be finished within 30 s, indicating its excellent superparamagnetic property. The results of enriching standard glycopeptides showed that, compared to using B-Fe 3 O 4 @mTiO 2 alone, the synergistic approach could detect more number of glycopeptides in MS spectra with a cleaner background, and thus the signal-to-noise ratio increased dramatically and the sensitivity improved greatly. The combined materials also outperformed the commercial SiMAG-boronic acid particles in the identification of more glycopeptides from standard samples. The recovery of glycopeptides