One-Pot Two-Nanoprobe Assay Uncovers Targeted Glycoprotein

Mar 21, 2017 - †Department of Chemistry and ⊥Graduate Institute of Clinical Medicine, College of Medicine, National Taiwan University, Taipei, Tai...
1 downloads 11 Views 2MB Size
Article pubs.acs.org/ac

One-Pot Two-Nanoprobe Assay Uncovers Targeted Glycoprotein Biosignature Mira Anne C. dela Rosa,†,‡,§ Wei-Chun Chen,§,∥ Yi-Ju Chen,§ Rofeamor P. Obena,§ Chih-Hsiang Chang,§ Rey Y. Capangpangan,§,# Tung-Hung Su,⊥ Chi-Ling Chen,⊥ Pei-Jer Chen,⊥ and Yu-Ju Chen*,†,§ †

Department of Chemistry and ⊥Graduate Institute of Clinical Medicine, College of Medicine, National Taiwan University, Taipei, Taiwan ‡ Nano Science and Technology Program, Taiwan International Graduate Program and §Institute of Chemistry, Academia Sinica, Taipei, Taiwan ∥ Department of Chemistry, National Taiwan Normal University, Taipei, Taiwan # Caraga State University, Butuan City, Philippines S Supporting Information *

ABSTRACT: We report a one-pot two-nanoprobe approach coupled to mass spectrometry for simultaneous quantification and post-translational modification (PTM) profiling of targeted protein in biofluid. Using N-glycoprotein as model, the assay employs two nanoprobes, antibody-conjugated SiO2 nanoparticles and lectin-conjugated magnetic Fe3O4 nanoparticles, to achieve target glycoprotein isolation from biofluid and subsequent glycopeptide enrichment in a single tube. As demonstrated on α-fetoprotein (AFP), a serum biomarker for hepatocellular carcinoma (HCC), the assay has high purification specificity (20 glycopeptides) with 2-fold and 10fold superior total glycopeptide intensity compared to nonone-pot method (9 glycopeptides) or without enrichment (6 glycopeptides), respectively. By multiple reaction monitoring mass spectrometry (MRM-MS) analysis of the nonglycopeptides, the assay can quantify low abundant AFP expression (0.5 ng) with good correlation with conventional ELISA method (Pearson’s r = 0.987). Furthermore, we present the first study revealing AFP glycopeptide signatures of individual HCC patients, comprised of 23 heterogeneous glycoforms of bi- and triantennary, core and terminal fucosylation, and mono- to trisialylation. In addition to 12 novel AFP glycoforms, our quantification result uncovers five abundant glycoforms in HCC, including 3 core-fucosylated (CF) forms. These identified CF forms may be evaluated in future studies as potential targets in a glycopeptide biomarker panel to further improve accuracy of conventional AFP-L3 tests. Through this one-pot assay, a comprehensive target protein profile comprised of protein expression and glycosylation pattern was achieved in simple protocol with high sensitivity, reduced analysis time, and minute starting material. This assay can be extended to other PTM biosignatures by conjugation of other affinity ligands on the nanoprobe.

A

found to be more specific (97%) for early disease diagnosis. Thus, a combination of the protein concentration and its disease-specific modifications may comprise a combinatorial biosignature that can be more accurate at diagnosis and monitoring of the disease. Among the PTMs, glycosylation has gained considerable attention because of its ubiquitous presence and critical role in cell−cell recognition, membrane transport, and signal transduction.8 From the structural point of view, the folding and stability of a protein can be changed by a single alteration of a glycan to a specific site, which, in turn, affects the protein’s

lthough quantification of protein biomarker concentration in biofluid is used as a disease diagnosis tool, its utility is limited because many biomarkers do not have sufficient specificity and sensitivity to diagnose the disease.1,2 Alternatively, aberrations in protein post-translational modifications (PTMs) are increasingly being studied because of their close association with cancer and other diseases.3,4 Recently, it was proposed that a more efficient strategy to discover biomarkers was through comprehensive profiling of all modifications of the targeted protein under disease state.5 For example, αfetoprotein (AFP), an N-glycosylated protein and clinically adopted serum marker for hepatocellular carcinoma (HCC),6 has insufficient specificity (80−94%) and high false negative rate (35−59%) at any stage7 when considering its concentration alone. Meanwhile, its glycosylated form, AFP-L3, was © XXXX American Chemical Society

Received: November 8, 2016 Accepted: March 9, 2017

A

DOI: 10.1021/acs.analchem.6b04396 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

Figure 1. Schematics for one-pot enrichment of target protein and its glycopeptides for glycosylation profiling and protein quantification. (A) Protein enrichment from serum by SiO2@Ab; (B) On-particle trypsin digestion; (C) Glycopeptide enrichment by MNP@ConA; (D) Nonglycopeptide isolation by centrifugation; (E) Protein quantification by MRM-MS; and (F) Glycosylation profiling by LC-MS/MS. MRM-MS: multiple reaction monitoring mass spectrometry; LC-MS/MS: liquid chromatography tandem mass spectrometry.

function in biological systems.3 Thus, tools to analyze the changes on the glycan structure, site occupancy, and glycoform concentration may provide new insight into the role of glycosylation in disease development and progression.9 Mass spectrometry (MS) is a powerful tool routinely employed for the analysis of glycosylation in complex mixtures,10 especially for the discovery of unknown phenotype-dependent glycosylation profiles under pathological conditions. However, comprehensive characterization of a glycoprotein still presents an analytical challenge. First, due to sample matrix complexity of clinical samples (e.g., body fluids or tissues)11 and the low abundance of glycoprotein biomarkers (pg/mL−low ng/mL),6 various strategies, such as abundant protein depletion, multidimensional separation, or enrichment strategies, are usually employed prior to mass analysis to facilitate the target protein isolation.10,12 To further characterize the incredibly high degree of glycan structure diversity, afterward, enrichment of glycopeptides have to be employed, followed by either detached glycan analyses13−15 or intact glycopeptide analyses,16,17 to obtain information on glycan structure and site occupancy. To retain site-specific information, intact glycopeptide analysis is imperative. Methods based on glycan-specific recognition (lectin affinity, hydrazide/ boronic acid chemistry) or glycan physicochemical properties (hydrophilic interaction (HILIC), metal oxide affinity chromatography) are commonly employed prior to MS analysis.18 Affinity-based platforms, such as antibody-based detection of protein-specific glycosylation,19 lectin-derivatized chips,20 and metal organic framework-based probes,21 have been continuously developed, indicating that glycopeptide enrichment is crucial to sample preparation. As a result, multiple timeconsuming steps with different purification materials at both protein and glycopeptide levels are required for site-specific glycosylation profiling of the targeted protein biomarker. For quantification of biomarker concentration, furthermore, other approaches such as enzyme-linked immunosorbent assay (ELISA) have to be independently performed.

To circumvent these issues, we designed a one-pot enrichment strategy to achieve sequential target glycoprotein and glycopeptide purification. By employing dual nanoprobes with divergent separation properties and functionality, a multistep assay is achieved in a single container (“one-pot”; Figure 1). Combined with liquid chromatography−tandem MS (LC-MS/MS) analysis for quantification and glycosylation profiling, we demonstrated this strategy to AFP as a proof-ofconcept. Currently, it is analyzed using ELISA,22,23 thus, its PTM information cannot be obtained. Moreover, although enzymatically released glycans of AFP have been previously discovered in ascetic fluid, HCC Hep G2 cell lines and human serum,24−26 so far, the intact AFP glycopeptides from human serum have not been fully characterized, despite its clinical utility. Hence, we also implemented the assay for profiling and quantification of AFP in HCC patients in order to demonstrate its potential clinical utility and to discover novel HCCassociated glycan structures that could aid HCC biomarker development.



EXPERIMENTAL SECTION Materials. Purified α-fetoprotein from human cord serum (AFP) was purchased from Abnova and anti-α-fetoprotein antibody (anti-AFP) from Thermo Fisher Scientific. Purified human haptoglobin and hemoglobin were purchased from Sigma-Aldrich. Concanavalin A from Canavalia ensiformis (Jack bean) was obtained from Sigma-Aldrich. Tris(2-carboxyethyl)phosphine hydrochloride (TCEP, ≥98%), S-methylmethanethiosulfonate (MMTS, 97%), and triethylammonium bicarbonate buffer (TEABC, 1M) were purchased from SigmaAldrich. Sequencing grade porcine trypsin was purchased from Promega. Sodium cyanoborohydride (NaBH3CN, ≥95%), formaldehyde (36.5−38% in H2O), and formaldehyde-d2 solution (∼20 wt % in D2O, 98 atom % D) were purchased from Sigma-Aldrich. ZipTip pipet tips were purchased from Merck-Millipore. Detailed synthetic protocols for antibodyconjugated silica nanoparticles, lectin-conjugated magnetic B

DOI: 10.1021/acs.analchem.6b04396 Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry



Article

RESULTS AND DISCUSSION One-Pot Two-Nanoprobe Assay. Dual nanoprobes with contrasting separation property and functionality are crucial to achieve this one-pot, multistep assay. At the protein level, we fabricated nonmagnetic SiO2 NP (Figure S1 and Table S1) decorated with antibodies (SiO2@Ab) to isolate the target glycoprotein from human serum by centrifugation. Subsequently, at the glycopeptide level, we fabricated Fe3O4 magnetic nanoparticles (MNP, Figure S1 and Table S1) decorated with glycan-specific lectin Concanavalin A (MNP@ConA) to isolate the glycopeptides using a magnet from the nonglycopeptides and SiO2 NPs. As shown in Figure 1A, the first step consists of isolating the target glycoprotein from human serum using SiO2@Ab, followed by washing and centrifugation to remove nonantigenic serum proteins in supernatant. After trypsin digestion of the isolated protein on the SiO2 NP in the same tube (Figure 1B), MNP@ConA is then added to the same tube to selectively isolate the glycopeptides by magnetic extraction (Figure 1C). Finally, the glycopeptides are eluted from the MNP@ConA and analyzed by LC-MS/MS to determine the glycosylation patterns (Figure 1F). Meanwhile, the remaining nonglycopeptides in the supernatant are also collected after centrifugation of the SiO2 NPs and analyzed by multiple reaction monitoring mass spectrometry (MRM-MS) for protein quantification (Figure 1D,E). The efficiency of this approach arises from the binding specificities of the two nanoprobes: antibody on SiO2 NPs has high specificity to isolate the low abundant target protein, while the lectin on MNP has different glycotoperecognition capability to purify the glycopeptides (inset, Figure 1). In the current study, ConA, which specifically binds to recognize high mannose oligosaccharides and complex-type biantennary glycans,30 was used. Specificity and Sensitivity of Protein Level and Glycopeptide Enrichment. The specificity of the antibody toward its target molecule is a very crucial issue to determine specificity and sensitivity in antibody-based assays. We first evaluated the specificity for protein level isolation and determined whether AFP is the major protein isolated from human serum using the one-pot strategy. Using Mascot Search31 to identify the proteins from the LC-MS/MS data, AFP was purified as the major protein isolated from SiO2@Ab, which garnered the highest Mascot score (1293) and Exponentially Modified Protein Abundance Index (emPAI = 3.38),32 demonstrating highly specific protein purification in complex biofluid by one-pot assay (Table S2). The Mascot score is a measure of statistical confidence in protein identification,31 while the emPAI approximates the quantity of the proteins in a mixture.32 Some known interacting proteins,33,34 such as serum albumin (emPAI = 3.31), prothrombin (emPAI = 0.86), and vitamin K-dependent protein S (emPAI = 0.82), were also copurified and identified from the mixture. Notably, AFP has similar physicochemical properties and high sequence homology with albumin, and thus, it was copurified from serum. Despite the copurification of minor serum proteins, we employed MS, which is a powerful tool to detect and differentiate proteins on the basis of their mass-to-charge ratio (m/z). Thus, in this work, the specificity and confidence in identification and quantification can be ensured by sequence determination for biomarker in complex clinical sample.

nanoparticles, and titanium dioxide-coated magnetic nanoparticles are presented in the Supporting Information. Human Serum Samples. For clinical application of onepot method, serum from hepatocellular carcinoma (HCC) patients were obtained with approval from Institutional Review Board of National Taiwan University Hospital in Taipei, Taiwan, and with informed consent from the patients. One-Pot Enrichment by Dual Nanoprobes. Antibodyconjugated silica nanoparticles (SiO2@Ab, 600 μg) were added to the standard protein solution in PBS or diluted human serum (150 μL). The solution was incubated at room temperature for 1 h in a rotary mixer. After extraction, the SiO2@Ab were separated by centrifugation at 7500 rpm for 5 min, washed twice with Tween-TBS (TTBS, 200 μL), and then once with deionized water (200 μL). The extracted target protein−SiO2@Ab complex were resuspended in TEABC (25 mM, 5 μL) reduced and alkylated with TCEP (5 mM) and MMTS (2 mM), and then digested with trypsin in a shaker at 37 °C for 5 h. The resulting mixture of peptides and SiO2 nanoparticles (SiO2 NPs) were then dried in a centrifugal evaporator. For glycopeptide enrichment, Concanavalin A lectin-conjugated Fe3O4 magnetic nanoparticles (MNP@ ConA, 360 μg in ABC) were first washed twice with HEPESCa2+ buffer (0.1 M HEPES, 0.01 M CaCl2, 150 μL). MNP@ ConA were resuspended in HEPES−Ca2+ buffer (300 μL) and were added directly to the peptide and SiO2 NPs mixture. The solution was incubated at room temperature for 1.5 h in a shaker. After extraction, the MNP@ConA were separated by a magnet and the SiO2 NPs were removed and collected in a separate tube. The MNP@ConA were washed thrice with HEPES−Ca2+ buffer (150 μL), ensuring that all SiO2 NPs were removed. After which, the glycopeptides were eluted from the MNP@ConA by addition of ACN/H2O/TFA = 50:49.9:0.1 (300 μL) and incubation for 30 min in a shaker. After magnetic separation, the glycopeptides were collected, dried in a centrifugal evaporator, and desalted with ZipTip. For peptide quantification, nonglycopeptides were separated from the SiO2 NPs collected previously by centrifugation at 7500 rpm for 5 min. The supernatant was collected and dried in a centrifugal evaporator. Liquid Chromatography Tandem Mass Spectrometry (LC-MS/MS) Analysis. The enriched glycopeptides from standard protein solutions (proof-of-concept) were analyzed on a nanoAcquity UPLC system (Waters, Milford, MA) coupled to a Synapt G1 Q-ToF mass spectrometer (Waters) by a nanospray ion source. For glycosylation profiling of HCC samples, LC-MS/MS was performed on linear trap quadrupole (LTQ)-Orbitrap Velos (Thermo Fisher Scientific, Bremen, Germany) coupled to a nanoACQUITY UPLC System (Waters) through a PicoView (New Objective) nanospray interface. Multiple reaction monitoring (MRM) was performed on a nanoAcquity UPLC system (Waters) coupled to a QTrap 5500 (AB Sciex) by a nanospray ion source. Detailed LC-MS/ MS analysis parameters and methods are provided in Supporting Information, S1.5−S-1.7. Glycopeptide structures were identified with the aid of MAGIC software (http://magic. iis.sinica.edu.tw/index.html)27 and Byonic (Protein Metrics),28 followed by manual validation. MRM data were analyzed with the aid of Skyline software (MacCoss Lab, University of Washington).29 C

DOI: 10.1021/acs.analchem.6b04396 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

method. Additionally, only six AFP glycopeptides were observed using a workflow that does not involve glycopeptide level enrichment (online LC-MS/MS only). Compared to the one-pot approach, this is inferior by 10-fold in terms of accumulated abundance, highlighting the necessity of enrichment at the glycopeptide level. The results showed that LCMS/MS alone cannot provide sufficient separation resolution and comprehensive identification of the nonglycopeptides and glycopeptides of AFP. Low ionization efficiency of glycopeptides due to their high molecular weight and negative charges still present challenges for glycopeptide detection. In addition, the presence of more abundant nonglycopeptides causes ion suppression during mass analysis. It should be noted that AFP is a 66-kDa glycoprotein with a single glycosylation site. Thus, a typical AFP tryptic digest would consist of a majority of nonglycopeptides, which easily overwhelms the glycopeptides. Quantification Performance. For quantification of AFP extracted by the one-pot method, the nonglycopeptide fraction was collected and analyzed by MRM-MS. Nonglycopeptides are chosen to represent the quantity of the protein in the sample. Only nonglycopeptides with 8−15 amino acids (average length of ∼10)37 were selected. To ensure that the peptides chosen are not glycosylation sites, we consulted UniProt database to filter all potential glycosylation sites of AFP. Peptide sequences with “NXS/T” motif may be potentially N-glycosylated and were also excluded. We chose confidently identified nonglycopeptides with high and reproducible intensity and did not contain any cysteine and methionine residues and modifications. Three peptides were ultimately chosen for quantification: TFQAITVTK, YIQESQALAK, and GYQELLEK. A representative quantification curve from a nonglycopeptide fragment GYQELLEK/y5 was plotted (Figure 3A) to show wide

Additionally, we determined whether the AFP antibody has affinity to glycosylated AFP glycoforms, which may influence the enrichment recovery of glycopeptides. To determine binding sites, the AFP was digested with trypsin, followed by enrichment of antibody-binding peptides by incubation with SiO2@Ab. After LC-MS/MS analysis, seven nonglycopeptides were identified as the possible binding sites of AFP antibody, whereas no AFP glycopeptide was identified. Enrichment of AFP and direct on-SiO2@Ab digestion were also performed to analyze the antibody-bound peptides to confirm the result. These imply that AFP antibody does not have binding specificity to glycopeptides. At the glycopeptide level, 20 intact glycopeptides, composed of eight unique glycoforms, were identified from AFP isolated by MNP@ConA through the one-pot approach (Table S3). To show general applicability of the method, we also applied onepot purification to another clinically relevant N-glycoprotein, Haptoglobin (Hp), where we identified 17 glycopeptides, composed mostly of sialylated glycoforms (Table S4). Using both proteins, we are able to identify glycoforms with exposed mannose residues as well as biantennary complex-type glycoforms, which are known to bind weakly to ConA.30 The use of nanoparticle-based immuno-extraction has previously demonstrated better purification specificity and efficiency compared to traditional immunoprecipitation methods.35,36 In addition to the strength of using nanoprobe, the competitiveness of this approach was further evaluated by comparison with traditional “non-one-pot” assay, where the enriched AFP was eluted from the antibodies, transferred to a different tube, and dried prior to trypsin digestion. By the non-one-pot method, 11 out of 20 intact AFP glycopeptides failed to be detected. In addition, the total abundance (sum of peak areas) of the AFP glycopeptides from non-one-pot also significantly decreased (Figures 2 and S2). Consistently, similar superiority of the onepot method was demonstrated on Hp with ∼2-fold higher number and abundance of glycopeptides (Figures 2 and S2). Thus, in addition to shorter sample preparation time, our result demonstrates the superior efficiency and sensitivity of one-pot method, which likely reduced sample loss from the additional elution and drying steps, and tube transfers of the non-one-pot

Figure 3. Quantification curves of AFP. (A) Linearity of a representative MRM transition (GYQELLEK(+2)/y5) for protein quantification. (B) Correlation of protein quantification data obtained from ELISA and one pot-MRM. MRM concentrations were calculated from an average of three peptides.

dynamic range (0.5−500 ng) with good linearity (r2 = 0.9999) and good precision (4.2−17.5%). Such detection limit is capable of providing quantification of very low concentration of AFP for healthy individual (