reversed-phase liquid chromatography with porous graphitic carbon liquid chromatography for simultaneous proteomics and N-glycomics analysis

G Model

ARTICLE IN PRESS

CHROMA-356768; No. of Pages 10

Journal of Chromatography A, xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma

Online coupling of hydrophilic interaction/strong cation exchange/reversed-phase liquid chromatography with porous graphitic carbon liquid chromatography for simultaneous proteomics and N-glycomics analysis Yun Zhao a , Henry C.H. Law a , Zaijun Zhang b,∗ , Herman C. Lam a , Quan Quan a , Guohui Li a , Ivan K. Chu a,∗ a

Department of Chemistry, The University of Hong Kong, Hong Kong, China Institute of New Drug Research and Guangdong Province Key Laboratory of Pharmacodynamic Constituents of Traditional Chinese Medicine, Jinan University College of Pharmacy, Guangzhou 510632, China b

a r t i c l e

i n f o

Article history: Received 6 December 2014 Received in revised form 20 July 2015 Accepted 10 August 2015 Available online xxx Keywords: N-Glycans N-Glycoproteins HILIC–SCX–RP Three-dimensional liquid chromatography Proteomics

a b s t r a c t In this study we developed a fully automated three-dimensional (3D) liquid chromatography methodology—comprising hydrophilic interaction separation as the first dimension, strong cation exchange fractionation as the second dimension, and low-pH reversed-phase (RP) separation as the third dimension—in conjunction downstream with additional complementary porous graphitic carbon separation, to capture non-retained hydrophilic analytes, for both shotgun proteomics and N-glycomics analyses. The performance of the 3D system alone was benchmarked through the analysis of the total lysate of Saccharomyces cerevisiae, leading to improved hydrophilic peptide coverage, from which we identified 19% and 24% more proteins and peptides, respectively, relative to those identified from a twodimensional hydrophilic interaction liquid chromatography and low-pH RP chromatography (HILIC–RP) system over the same mass spectrometric acquisition time; consequently, the 3D platform also provided enhanced proteome and protein coverage. When we applied the integrated technology to analyses of the total lysate of primary cerebellar granule neurons, we characterized a total of 2201 proteins and 16,937 unique peptides for this primary cell line, providing one of its most comprehensive datasets. Our new integrated technology also exhibited excellent performance in the first N-glycomics analysis of cynomolgus monkey plasma; we successfully identified 122 proposed N-glycans and 135 N-glycosylation sites from 122 N-glycoproteins, and confirmed the presence of 38 N-glycolylneuraminic acid-containing N-glycans, a rare occurrence in human plasma, through tandem mass spectrometry for the first time. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Tandem mass spectrometry (MS/MS) has become a vital and powerful analytical tool for highly sensitive, high-throughput proteomics studies of complex biological systems [1]. The complexity and wide dynamic range of protein abundances and the sheer number of proteins in biological samples have made the identification of full proteomes from a single experiment a considerable scientific challenge—partly because the finite scanning speed and dynamic

∗ Corresponding authors. Tel.: +852 2859 2152. E-mail addresses: [email protected] (Z. Zhang), [email protected] (I.K. Chu).

range of mass spectrometry (MS) can result in under-sampling problems. The abundances of the tens of thousands of peptides that can result after protein digestion can span over 10 orders of magnitude [2]. The proteome coverage can be improved using various approaches, including sub-cellular compartment analysis, repetitive analysis of samples, data-dependent acquisition, and advanced sample separation techniques, such as the combination of two or more dimensions of liquid chromatography (LC). Multidimensional liquid chromatography (MDLC) can be used to expand the separation spaces to overcome the limited peak capacity of onedimensional liquid chromatography, which utilizes only a single type of column chemistry [3]. The coupling of MDLC with MS/MS has become an indispensable technique in MS-based proteomics. Most MDLC systems at present are based on two-dimensional

http://dx.doi.org/10.1016/j.chroma.2015.08.017 0021-9673/© 2015 Elsevier B.V. All rights reserved.

Please cite this article in press as: Y. Zhao, et al., Online coupling of hydrophilic interaction/strong cation exchange/reversed-phase liquid chromatography with porous graphitic carbon liquid chromatography for simultaneous proteomics and N-glycomics analysis, J. Chromatogr. A (2015), http://dx.doi.org/10.1016/j.chroma.2015.08.017

G Model CHROMA-356768; No. of Pages 10 2

ARTICLE IN PRESS Y. Zhao et al. / J. Chromatogr. A xxx (2015) xxx–xxx

(2D) LC to enhance the chromatographic resolving power for peptide separation, thereby minimizing peptide co-elution and ion suppression during downstream electrospray MS/MS analyses. A prime example of a 2DLC system frequently employed for shotgun proteomics combines strong cation exchange (SCX) with reversedphase (RP) chromatography (SCX–RP) to separate peptides based mainly on their charge and hydrophobicity [4]. Among the various possible MDLC systems, the combination of hydrophilic interaction liquid chromatography and low-pH RP chromatography (hereafter simplified as “HILIC–RP”) offers higher resolution in the first-dimension HILIC column [5] as well as greater orthogonality than that of conventional 2D SCX–RP LC [3]. These features arise because two different separation modes operate in the HILIC–RP platform: mainly hydrophilic partitioning, but also dipole–dipole, hydrogen bonding, and weak electrostatic interactions, in the HILIC column, and separation based on peptide hydrophobicity in the second-dimension RP LC column. These two modes lead to increased analytical performance and an increased diversity of chromatographic modalities and separation chemistries [3,6]. SCX separation is based on the charges of the peptides. It is well established that multiply charged tryptic peptides typically carry no more than five charges, with doubly and triply protonated ions predominating (>80% in total). Thus, SCX separation, especially when using a salt gradient, has limited resolution, leading to peptide spillover and uneven peptide distributions across successive fractions in the first dimension of the 2D SCX–RP system [3,7,8]. HILIC–RP processes have been employed mainly offline, involving fraction collection from the first dimension and subsequent re-injection onto the next chromatographic dimension. Offline HILIC–RP offers the advantage of being simple to implement, as well as flexibility. The major drawbacks of the offline approach to MDLC are its labor-intensity, tedious sample manipulations, and potential for sample losses. Online coupling would be attractive because it allows automated and unattended analyses of samples of very small sizes with minimal sample losses. The development of online HILIC–RP systems has been difficult, however, because of solvent incompatibility between the HILIC and RP dimensions, as well as the problem of peptide dissolution in the high content of organic solvent in the HILIC starting gradient [9]. The high content of organic solvent that elutes from the first-dimension HILIC column would cause flow-through of the peptides retained in the second-dimension RP column, leading to sample loss in the void volume. Our group recently developed an online 2D HILIC–RP LC system for proteomics profiling, positioning an RP trap column before the first-dimension HILIC column to overcome the problem of sample dissolution in the highly organic buffer used for HILIC separation, while mitigating solvent incompatibility between the two online LC dimensions through use of a solvent mixing loop in addition to an SCX trap column [10]. The separation efficiency of 2DLC remains inadequate, however, when probing highly complex samples, such as trypsinized proteins of total cell lysates. Three-dimensional liquid chromatography (3DLC), which incorporates an additional dimension of separation, can provide higher peak capacities—a very desirable feature for the separation and identification of complex peptides [11–13]. Herein, we selected a combination of liquid chromatographic modalities—hydrophilic interaction chromatography, strong cation exchange fractionation, and conventional low-pH RP chromatography—that separate peptides according to different principles. We describe an online 3D HILIC–SCX–RP platform that we have operated in conjunction downstream with additional complementary porous graphitic carbon (PGC) separation to capture non-retained hydrophilic analytes. Relative to the 2D HILIC–RP system, the performance of this newly developed online MDLC platform was enhanced through the addition of the SCX column as the second LC dimension, thereby allowing additional charge

separation of tryptic peptides between the first HILIC and third RP LC dimensions; each sub-fraction of peptides in the SCX column was further separated using a stepwise salt gradient, followed by low-pH RP separation in the third dimension. In short, our new strategy integrates three different prevalent peptide separation technologies (HILIC, SCX, RP) into a single online platform for peptide separation (according to hydrophilic partitioning interactions, charge, and hydrophobicity, respectively). We configured the additional PGC column after the RP trap column to trap all of the hydrophilic flow-through, thereby extending the utility of the online 3D HILIC–SCX–RP/PGC LC system to the identification of Nglycans (by the PGC LC component) and de-glycan peptides (by the 3D HILIC–SCX–RP LC part) in the same run. Relative to many common LC packing materials, PGC has interesting properties, including the ability to separate very hydrophilic analytes that cannot be trapped in an RP column [14]. Retention on PGC is determined by the balance between the hydrophobicity and the interactions of polarizable and polarized groups in the analytes with the PGC surface [14]. The latter interactions depend on both the nature of the functional groups, the contact area between the analytes, and the surface of the stationary phase. Because of its special retention mechanism, PGC LC can be applied in the analyses of very polar compounds, including glycans [15]. Synchronization of the fractionations between the two modules (3D HILIC–SCX–RP and PGC) provides an enabling approach for identification of both hydrophobic and hydrophilic compounds in analyses of complex mixtures, featuring a diverse range of hydrophobicities, from a single sample injection. Thus, the integrated methodology minimizes losses of non-retained hydrophilic flow-through peptides and glycans through recapturing of the hydrophilic effluent online after sample loading into the 3D HILIC–SCX–RP module for shotgun proteomics analysis. We have evaluated the qualitative and quantitative proteomics performance of this online 3D HILIC–SCX–RP/PGC system through analyses of the tryptic digests of cynomolgus monkey brain and primary cerebellar granule neurons (CGNs) using isobaric tags for relative and absolute quantitation (iTRAQ) technology [16]. We have also, for the first time, performed detailed N-glycoproteomics and N-glycomics analyses—in the same run—of cynomolgus monkey plasma using our new MDLC system with only a single sample injection event.

2. Experimental 2.1. Chemicals and materials Protease inhibitor cocktail tablets (EDTA-free), modified sequencing-grade trypsin and N-glycosidase F (PNGase F) were obtained from Roche (Switzerland). iTRAQ Reagents-8plex was purchased from AB Sciex (Framingham, MA, USA). Bradford assay reagent was supplied by Bio-Rad (Hercules, CA, USA). The high-capacity multiple affinity removal column-Hu6 (4.6 mm × 100 mm) and the reagent kit (buffers, spin filters, spin concentrators) were obtained from Agilent Technologies (Wilmington, DE, USA). Glycoprotein isolation kits, including lectin concanavalin A (Con A) and wheat germ agglutinin (WGA), were acquired from Thermo Scientific (Rockford, IL, USA). All other chemicals and protein standards were purchased from Sigma–Aldrich (St. Louis, MO, USA). Hypercarb PGC packing materials (3-␮m particles, 250-Å pores) were obtained from Thermo Scientific. TSKgel Amide-80 packing materials (3-␮m particles, 300-Å pores) were supplied by Tosoh Corporation. Jupiter C18 packing materials (3-␮m particles, 300-Å pores) were acquired from Phenomenex (Torrance, CA, USA). SCX packing materials (PolySulfoethyl A, 5␮m particles, 300-Å pores) were purchased from PolyLC (Columbia, MD, USA). Electronically actuated six- and ten-port, two-position

Please cite this article in press as: Y. Zhao, et al., Online coupling of hydrophilic interaction/strong cation exchange/reversed-phase liquid chromatography with porous graphitic carbon liquid chromatography for simultaneous proteomics and N-glycomics analysis, J. Chromatogr. A (2015), http://dx.doi.org/10.1016/j.chroma.2015.08.017

G Model CHROMA-356768; No. of Pages 10

ARTICLE IN PRESS Y. Zhao et al. / J. Chromatogr. A xxx (2015) xxx–xxx

switching valves were acquired from Valco Instruments (Houston, TX, USA). 2.2. Sample preparation 2.2.1. Standard proteins A standard protein (albumin bovine serum, myoglobin (from horse heart), ␤-lactoglobulin (from bovine milk), ␣-casein (from bovine milk), ␤-casein (from bovine milk)) or a standard glycoprotein (ribonuclease B (from bovine pancreas), fetuin (from fetal calf serum), chicken ovalbumin) (100 ␮g) was dissolved in 100 mM ammonium bicarbonate (100 ␮L) and reduced with 50 mM DTT at 60 ◦ C for 60 min. The sample was then alkylated in 100 mM IAA for 60 min at room temperature in the dark, followed by incubating with modified trypsin (2 ␮g) at 37 ◦ C overnight. Subsequently, the trypsin digestion was terminated by boiling the sample at 99 ◦ C for 10 min. The resulting protein digests were stored at −20 ◦ C until required for further use. After terminating the trypsin digestion, N-deglycosylation was performed by incubating the digested standard glycoproteins with PNGase F at a ratio of 1 mg of glycoproteins to four units of PNGase F at 37 ◦ C for 19 h to release the N-glycans. Finally, the samples were stored at −20 ◦ C until required for use. 2.2.2. Saccharomyces cerevisiae (yeast) proteins Yeast cells (INVSc1) were grown and processed using methods similar to those reported previously [17,18]. After the lysed cells had been centrifuged (14,000 rpm, 4 ◦ C, 15 min), the proteins from the supernatant were precipitated by incubating in six volumes of cold acetone for 2 h at −20 ◦ C. The precipitated proteins were then dissolved in 8 M urea and 100 mM ammonium bicarbonate for protein quantitation using the Bradford assay reagent and subsequent digestion by trypsin. The digestion procedure was the same as that mentioned above for the standard proteins, except that the samples were diluted to a final concentration of 2 M urea by adding 100 mM ammonium bicarbonate prior to adding trypsin. The protein digests were then stored at −20 ◦ C until required for use. 2.2.3. Cynomolgus monkey brain proteins Adult (ca. 4–5 years old) male cynomolgus monkeys (Macaca fascicularis) were selected for the study, each weighing 3.0–4.0 kg. All the animal procedures were performed following the testing facility standard of practice (SOP) and regulations from the Institutional Animal Care and Use Committee of Kunming Biomedical International (KBI, Yunnan Province, China) for humane and proper care of the research animals. 2.2.4. CGNs iTRAQ proteins CGNs were isolated from Sprague-Dawley rats that were eight days old, as described previously [19]. All the animal procedures

3

were performed in accordance with the guidelines of the Experimental Animal Care and Use Committee of Jinan University, and approved by the Ethics Committee for Animal Experiments of Jinan University. 2.2.5. Glycoproteins from cynomolgus monkey plasma Only 30 ␮g of enriched glycoprotein sample was required for the online 3D HILIC–SCX–RP LC analysis. The whole analysis workflow for plasma glycoproteins is displayed in Fig. 1. The animal experiments were also conducted according to the testing facility SOP and regulations from the Institutional Animal and Use Committee of KBI (Yunnan Province, China). Blood (ca. 1 mL) was taken from the hind limb vein of a healthy cynomolgus monkey and placed into a heparinized tube; it was centrifuged immediately (3000 rpm, 4 ◦ C, 10 min) for separation of the plasma. The monkey plasma was then processed by removing six highly abundant proteins (albumin, antitrypsin, IgA, IgG, haptoglobin, transferrin) using an Agilent multiple affinity removal column-Hu-6, following the instructions in the manufacturer’s manuals. Next, Con A and WGA glycoprotein isolation kits (Thermo Scientific) were used to enrich the N-glycoproteins. Detailed procedures are provided in the Supplementary Information. Detailed procedures for protein extraction and subsequent protein quantitation from monkey brain tissues and CGNs cells are available in the Supplementary Information. The extracted proteins were sequentially precipitated with acetone, re-dissolved in urea, quantified using the Bradford assay reagent, and finally digested by trypsin—in all cases following the same procedures as those mentioned above for the yeast proteins. For glycoproteins from monkey plasma, after terminating the trypsin digestion, N-deglycosylation was performed using the same procedure as that for standard glycoproteins. For CGNs iTRAQ samples, an aliquot (100 ␮g) of the proteins from each treatment was reduced with 50 mM DTT at 60 ◦ C for 30 min, and then alkylated in 100 mM IAA for 60 min at room temperature in the dark. The samples were subsequently centrifuged (4000 rpm) using Microcon® centrifugal filters for protein purification and desalting, followed by digestion with 1:33 trypsin in 0.5 mM triethylammonium bicarbonate (TEAB) buffer at 37 ◦ C overnight. For the iTRAQ experiments, all eight digested samples were then labeled with their corresponding iTRAQ tags at room temperature for 2 h, according to the iTRAQ protocol. Ultimately, the CGNs iTRAQ samples were stored at −80 ◦ C until required for further use. 2.3. Liquid chromatography 2.3.1. RP–PGC sequential trap LC The RP–PGC sequential trap LC experiment was performed using an Agilent 1200 series capillary pump equipped with a one-well plate auto-sampler and an Agilent 1200 series nano pump (Agilent

Fig. 1. Workflow for the analysis of N-glycoproteins from cynomolgus monkey plasma enriched through ConA and WGA lectin columns after removing six highly abundant proteins and, finally, separating through the 3D HILIC–SCX–RP/PGC LC system.

Please cite this article in press as: Y. Zhao, et al., Online coupling of hydrophilic interaction/strong cation exchange/reversed-phase liquid chromatography with porous graphitic carbon liquid chromatography for simultaneous proteomics and N-glycomics analysis, J. Chromatogr. A (2015), http://dx.doi.org/10.1016/j.chroma.2015.08.017

G Model CHROMA-356768; No. of Pages 10 4

ARTICLE IN PRESS Y. Zhao et al. / J. Chromatogr. A xxx (2015) xxx–xxx

Technologies, Wilmington, DE, USA). As displayed in Supplementary Fig. 1, the two LC pumps and all the LC columns were connected by two four-port and one six-port two-position electrically actuated switching valves. All the LC columns in this study were packed in-house using an ultrahigh-pressure 65 D syringe pump (Isco, Lincoln, NE, USA). The RP trap column (150 ␮m i.d. × 5 cm length) and analytical column (75 ␮m i.d. × 15 cm length) were packed with Jupiter C18 materials. The PGC trap column (150 ␮m i.d. × 5 cm length) and analytical column (75 ␮m i.d. × 15 cm length) were packed with Hypercarb PGC materials. Solvent A delivered from capillary pump 1 was 2% ACN in 20 mM ammonium formate adjusted to pH 6.8 with ammonium hydroxide. Solvent C (2% ACN and 0.5% formic acid in water) and solvent D (2% H2 O and 0.5% formic acid in ACN) at pH 2 were delivered by nano pump 2. Briefly, 3-␮g mixtures of the N-glycans released from the standard glycoprotein digests of ribonuclease B, fetuin, and ovalbumin, combined with protein digests from albumin bovine serum, myoglobin, ␤-lactoglobulin, ␣-casein, and ␤-casein, were first injected and trapped in the sequential RP–PGC trap columns by 100% solvent A (pH 6.8) from pump 1 at 2 ␮L/min for 30 min, while the capillary between the RP analytical column and the port of valve 2 was disconnected only during the sample trapping to avoid the pH 6.8 solvent going to the RP analytical column (Supplementary Fig. 1A). The pH 6.8 LC solvent used here was applied to mimic the actual sample trapping conditions of the 3D HILIC–SCX–RP LC system. The flow of pump 1 was stopped after sample trapping. Subsequently, the RP trap column was reconditioned with 100% solvent C (pH 2) from pump 2 at 500 nL/min for 30 min (Supplementary Fig. 1B). The RP LC separation then began at 300 nL/min with the efficient gradient of 5–35% D for 60 min (Supplementary Fig. 1C). Afterward, the PGC trap column was also reconditioned with 100% solvent C from pump 2 at 500 nL/min for 30 min (Supplementary Fig. 1D), followed by PGC LC separation at 300 nL/min with the efficient gradient of 5–30% D for 60 min (Supplementary Fig. 1E).

2.3.2. Online 3D HILIC–SCX–RP/PGC LC The experiments with the fully automated online 3D HILIC–SCX–RP/PGC LC system were performed using one Agilent 1200 series capillary LC pump and two Agilent 1200 series nano LC pumps, among which the capillary pump and one of the nano pumps were both equipped with a one-well plate auto-sampler. For the capillary LC pump 1 at pH 6.8, the eluents

were solvent A as mentioned in Section 2.3.1 and solvent B (10% solvent A and 90% ACN). For the nano LC pump 2 and pump 3 at pH 2, solvent C and solvent D was used as mentioned in Section 2.3.1. As displayed in Fig. 2, all the LC pumps and columns were connected by two six-port and two ten-port, two-position (position A or B) electrically actuated switching valves. The 3D HILIC–SCX–RP/PGC LC platform consisted of a HILIC column as the first LC dimension (eight HILIC fractions), an SCX LC as the second dimension (two SCX sub-fractions), an RP LC as the third dimension, and a PGC column as an additional LC system to trap and separate all the very hydrophilic flow-throughs from the RP trap column connected to pump 1 of the 3D HILIC–SCX–RP LC system. The online 3D HILIC–SCX–RP/PGC LC platform was operated automatically, as presented in detail in the Supplementary Information. The workflow and system design are displayed in Fig. 2, while the positions of the four multiport valves at each step and the gradient composition details of each LC dimension are listed in Supplementary Tables 1 and 2. The total analysis time of the 3D HILIC–SCX–RP/PGC LC platform was 2410 min. 2.3.3. Online 2D HILIC–RP LC The set-up of the online 2D HILIC–RP LC system has been reported in detail [10]. Detailed workflows, valve positions, and solvent gradients for the online 2D HILIC–RP LC are listed in the Experimental Section of the Supplementary Information, Supplementary Tables 3 and 4. The total analysis time of the 2D HILIC–RP LC platform was 2505 min. 2.4. Mass spectrometry The MS data for the protein digests of yeast were acquired using a QSTAR XL Q-TOF mass spectrometer (AB Sciex, Foster City, CA, USA). The MS data for other experiments, except the yeast experiments, were all acquired on a TripleTOF 5600 system (AB Sciex, Framingham, MA, USA). Both mass spectrometers were equipped with nanospray sources. Detailed MS acquisition parameters are presented in the Experimental Section of the Supplementary Information. 2.5. Data analysis The acquired MS raw data were searched against the theoretical protein databases downloaded from UniProt (http://www.uniprot. org) [20] by the ProteinPilot 4.5 software (AB Sciex, Framingham,

Fig. 2. (A) Workflow and (B) system design of the 3D HILIC–SCX–RP/PGC LC system.

Please cite this article in press as: Y. Zhao, et al., Online coupling of hydrophilic interaction/strong cation exchange/reversed-phase liquid chromatography with porous graphitic carbon liquid chromatography for simultaneous proteomics and N-glycomics analysis, J. Chromatogr. A (2015), http://dx.doi.org/10.1016/j.chroma.2015.08.017

G Model CHROMA-356768; No. of Pages 10

ARTICLE IN PRESS Y. Zhao et al. / J. Chromatogr. A xxx (2015) xxx–xxx

MA) [21], as presented in detail in the Experimental Section of the Supplementary Information. In addition, a 1% global FDR was adopted as the criterion for both protein and peptide identification [22]. For N-glycomics analysis, the glycopeptides and glycans were identified through manual inspection of both the MS and MS/MS spectra; the glycans were analyzed individually from the raw data by the software GlycoWorkbench 2.1 [23] (setting the precursor mass accuracy at 20 ppm in the GlycomeDB database), as well as through manual confirmation. Because of homology between cynomolgus monkeys and humans, the proteins identified from the cynomolgus monkey were first blasted to the corresponding human proteins. If the human protein after the blast was an N-glycoprotein and its N-glycosites also appeared in the corresponding cynomolgus monkey protein sequence in the Uniprot database, then this cynomolgus monkey protein was considered as a probable N-glycoprotein in this study. For each of the normal peptides identified, the value of pI was calculated using the tool in the ExPASy bioinformatics resources portal (http://web. expasy.org/compute pi/). The hydrophobicity indices (HIs) of the peptides separated by the C18 RP LC system at pH 2 were calculated using the Sequence Specific Retention Calculator (http://hs2. proteome.ca/SSRCalc/SSRCalcX.html) [24,25]. The grand average of hydropathy (GRAVY) of each peptide was calculated as the sum of hydropathy values of all the amino acids divided by the length of the peptide [26] (http://www.gravy-calculator.de/). The net charge of the peptide separated by the SCX column was calculated as the sum of the charge contributions from the constitutive amino acid side chains of D, E, H, Y, K, and R residues as well as the N- and C-termini at pH 2.4, according to the equation reported by Sims [27].

3. Results and discussion 3.1. Setup of online 3D HILIC–SCX–RP/PGC LC platform Initial successes in overcoming the issues of solvent incompatibility and sample solubility and, thereby, implementing an online 2D HILIC–RP LC system for proteomics analyses [10] led us to pursue the development of an online 3D HILIC–SCX–RP/PGC system, as displayed in Fig. 2. Peptide separation in the 3D HILIC–SCX–RP system is based upon three different chromatographic retention mechanisms: the retention in the HILIC column is based predominantly on hydrophilic partitioning and that in the RP column is related to hydrophobicity, whereas additional charge-centric separation in the SCX column is due to Coulombic interactions [6], making it highly orthogonal to both the HILIC and RP columns. Thus, we expected the additional SCX dimension of sub-fractionation of peptides (based on their net charges through stepwise salt elution for subsequent low-pH RP separation, instead of just serving as a trap column for peptide focusing alone) to enable more extensive separation and further decrease the complexity of the peptide mixtures. One of the challenges in the characterization of hydrophilic glycopeptides and glycans for online mapping of N-glycosylation when using standard bottom-up proteomics techniques is overcoming the limitations experienced in the most commonly used RP-based front-end separation technologies; the hydrophobic stationary phase of the RP column and the hydrophilic mobile phase allow most of the peptides, mainly hydrophobic peptides, to be efficiently trapped and separated. The addition of a PGC trap column overcomes the loss of the highly hydrophilic analytes, which are not retained by the original RP trap column design [28,29], allowing the simultaneous capture of highly hydrophilic peptides, N-glycans, and glycopeptides for subsequent PGC analyses.

5

3.2. Evaluation of the RP–PGC sequential trap LC design To optimize the PGC performance after RP LC, a mixture of N-glycans released from the well-characterized N-glycoproteins ribonuclease B, fetuin, and chicken ovalbumin, combined with the protein digests from albumin bovine serum, myoglobin, ␤lactoglobulin, ␣-casein, and ␤-casein, was trapped and analyzed using the RP–PGC tandem sequential trap LC system, in which the bulk of the sample was focused on the RP trap column; the non-retained highly hydrophilic effluents were trapped by the PGC column downstream. The trapped peptides and relatively hydrophilic peptides/glycans in the RP and PGC trap columns, respectively, were then transferred onto the HILIC–SCX–RP system and PGC column online for further separation and subsequent MS/MS analysis. Supplementary Table 5 lists the representative Nglycans identified from the three standard glycoproteins by the PGC column. Supplementary Fig. 2A displays a typical product-ion spectrum of the doubly charged protonated glycan Hex6 HexNAc2 identified at m/z 699.2617 with typical B or Y ions [30] arising from glycosidic bond cleavages under low-energy CID conditions. As expected, all the N-glycans could be identified only from the PGC column, demonstrating the efficacy and necessity of this integrated-PGC system to mitigate one of the major drawbacks of traditional RP-based MDLC: identification and characterization of very hydrophilic N-glycans and peptides that flow through in the void volume. In addition to allowing N-glycomics analysis, hydrophilic peptides could also be trapped and further separated by the PGC column. Among the total of 1382 unique peptides identified, 223 were detected only from the PGC dimension, as revealed in Supplementary Fig. 3A; the enriched unique hydrophilic peptides identified through the PGC module were substantially lower in average HI (by 7.99) than those from the RP LC system (4.41 for PGC, 12.40 for RP; Supplementary Fig. 3B); Supplementary Fig. 2B presents a typical MS/MS spectrum of the doubly charged protonated octopeptide EGIHAQQK, which forms a series of y and b ions through peptide bond cleavages; it was uniquely eluted and identified by the PGC module with an HI of only −0.49. As it turns out, a substantial increase in protein coverage, by 0.17–24.78%, occurred for the different standard proteins, due to the additional hydrophilic peptides (average HI: <4.5) identified in a single PGC fraction analysis (Table 1). We attribute most of the not-yet-covered peptide sequences (ca. 1–16%) to the excluded very short peptides that do not fall within the predefined IDA m/z range criteria (Table 1 and Supplementary Table 6 in bovine fetuin as an example). Thus, the integrated PGC chromatography chemistry improved the identification of extremely hydrophilic peptides from a single sample injection event, thereby providing desirable extended protein coverage information. Table 1 Protein coverage of standard proteins identified using the single RP LC and RP–PGC sequential trap LC systems, respectively. Coverage (%) represents the percentage of matched amino acids from the identified peptide with 1% FDR divided by the total number of amino acids in the protein sequence. Protein

Ribonuclease B Fetuin Ovalbumin Albumin bovine serum Myoglobin ␤-Lactoglobulin ␣-Casein ␤-Casein

Coverage (%) Increase (b − a)

Beyond pre-defined IDA m/z range

(a) Without PGC

(b) With PGC

97.58 79.94 97.93 76.11

100 86.63 98.96 79.90

2.42 8.37 1.05 4.98

0 5.29 1.04 16.47

100 92.92 100 96.37

96.75 63.48 64.95 63.73

96.75 79.21 72.43 63.84

0 24.78 11.52 0.17

2.60 3.93 10.75 10.27

99.36 83.14 93.18 74.11

Please cite this article in press as: Y. Zhao, et al., Online coupling of hydrophilic interaction/strong cation exchange/reversed-phase liquid chromatography with porous graphitic carbon liquid chromatography for simultaneous proteomics and N-glycomics analysis, J. Chromatogr. A (2015), http://dx.doi.org/10.1016/j.chroma.2015.08.017

G Model CHROMA-356768; No. of Pages 10 6

ARTICLE IN PRESS Y. Zhao et al. / J. Chromatogr. A xxx (2015) xxx–xxx

3.3. Comparison of 3D HILIC–SCX–RP and 2D HILIC–RP LC platforms Thus far, we have discussed only the performance of the PGC module. We evaluated the analytical enhancement of the additional SCX separation column, positioned between the HILIC and RP dimensions, through the analysis of yeast tryptic peptides using the same total LC–MS/MS run time and RP fractions: 8 HILIC × 2 SCX fractions for the 3D HILIC–SCX–RP system and 16 HILIC fractions for the 2D HILIC–RP system—a total of 16 fractions. The column dimensions (length, particle size, i.d.), HILIC fraction number, and gradient time of the RP LC were selected according to our previous technical experiences, optimization and study from 1D LC to 2D HILIC–RP nano-LC [10]. All the comparisons are based on strictly control experimental conditions with similar column length, particular size, LC gradient, MS duty cycle, IDA criteria, same rolling collision energy, etc. All the factors were based on a balance between good separation performance and practical application, like the limited MS analysis time, upper limit of the LC system pressure, etc. For the 3D HILIC–SCX–RP system, we employed eight fractionation steps for the second-dimension SCX column sequentially; each fraction was eluted with just two plugs of salt solution of increasing strength, with the expectation that this process would be sufficient to fractionate the dominant charges of the tryptic peptides (our results demonstrate that most of the tryptic peptides existed in +2 and +3 charge states, with a minority of +4 charge states or greater; first SCX fraction for +2-charged peptides, second SCX fraction for peptides with charge ≥3) by striking a balance between more-extensive fractionation and the increase in experimental throughout. We identified 1113 proteins and 7632 peptides using the 3D HILIC–SCX–RP LC system, whereas we identified only 935 proteins and 6218 peptides when

using the 2D HILIC–RP LC system and a Qstar XL mass spectrometer; thus, we obtained 19% and 24% increases in the number of identified proteins and peptides, respectively, when using the 3D HILIC–SCX–RP LC system (Fig. 3A). The extended protein and proteome coverages were contributed by peptides with greater hydrophilicities (Fig. 3B) and higher values of pI (Fig. 3C). In addition to changes in overall performance, we also observed slight differences in the physical properties among the peptides identified in the two experiments. The unique hydrophilic peptides identified by the 3D LC system had lower contents of their nonpolar amino acid residues leucine, isoleucine, methionine, tryptophan, and phenylalanine, by 5.2% (isoleucine) to 26.5% (methionine). There were overall increases in the contents of most of their polar uncharged amino acid residues, including serine, threonine, asparagine, and glutamine, as well as great increases (by 9.8% and 42.2%, respectively) in the contents of positively charged lysine and histidine residues (Fig. 3D), which possess the highest values of pI (9.7 and 7.6, respectively) among amino acid residues. A high number of these positively charged residues would be expected in the later SCX fractions, because they would interact strongly with the polysulfoethyl aspartamide surface; this phenomenon could be ascribed to more positively charged basic peptides with higher values of pI tending to compete more favorably in the second SCX fraction under the acidic conditions (Fig. 4A; Average pI = 6.95 and 1229 peptides identified uniquely in 3D SCX 2; Average pI = 5.36 and 1240 peptides identified uniquely in 3D SCX 1). Thus, the additional SCX fractionations categorized acidic and basic hydrophilic peptide populations into individual compartments for the downstream low-pH RP analyses. More importantly, we observed enhanced signals of peptides featuring higher values of pI after additional SCX sub-fractionations. Fig. 4B presents the extracted ion chromatogram for the moderate-intensity peptide

Fig. 3. (A) Venn diagram of identified proteins and peptides; (B) distribution of unique normal peptides identified across the different HI ranges; (C) distribution of unique normal peptides identified across the different pI ranges; and (D) differences in the amino acid average percentages in the unique normal peptides identified between the online 3D HILIC–SCX–RP (8 HILIC × 2 SCX fractions) and 2D HILIC–RP (16 HILIC fractions) LC systems from yeast digests (different colors for the columns represent different kinds of amino acids: nonpolar, polar uncharged, positively charged, and negatively charged).

Please cite this article in press as: Y. Zhao, et al., Online coupling of hydrophilic interaction/strong cation exchange/reversed-phase liquid chromatography with porous graphitic carbon liquid chromatography for simultaneous proteomics and N-glycomics analysis, J. Chromatogr. A (2015), http://dx.doi.org/10.1016/j.chroma.2015.08.017

G Model CHROMA-356768; No. of Pages 10

ARTICLE IN PRESS Y. Zhao et al. / J. Chromatogr. A xxx (2015) xxx–xxx

7

Fig. 4. (A) Venn diagram of normal peptides identified from the yeast digests and corresponding comparison of average values of pI in the 2D HILIC–RP and 3D HILIC–SCX–RP SCX1 and SCX2 fractions. (B, C) Extracted ion chromatogram (XIC) spectra of the doubly protonated peptide [SLFGKDPSK + 2H]2+ from yeast digests identified in the (B) 3D HILIC–SCX–RP SCX2 fraction and (C) 2D HILIC–RP LC system.

SLFGKDPSK at m/z 489.7477 (46,621 counts), corresponding to the Heat shock protein SSC1 (mitochondrial), from the SCX 2 fraction in the 3D HILIC–SCX–RP system. The total integrative area of the same peptide, which appeared in only a single SCX fraction, through 2D HILIC–SCX–RP was remarkably lower (Fig. 4C, 9991 counts); other investigated peptides exhibited even more substantial increases in their total intensities, by 87–545%, through 3D HILIC–SCX–RP analyses (Supplementary Table 7), leading to higher product ion scan signal and confidence (Supplementary Fig. 4A and 4B, Supplementary Table 7). The enhanced survey scan signal for those peptides (many are basic peptides) in the SCX sub-fractions improves the chances for data-dependent product ion scan acquisitions of potentially yet-to-be-sampled peptides,

given that the mass spectrometer has finite duty cycles, as a result of increases in the number of unique identified peptides. 3.4. Proteomics applications of the 3D HILIC–SCX–RP/PGC LC system We applied our established 3D HILIC–SCX–RP/PGC system for routine proteomic analyses using CGNs and cynomolgus monkey brain; each analysis required a total LC–MS/MS acquisition time of approximately 30 h. We identified a total of 2201 unique proteins and 16,937 peptides in duplicate analyses of the CGNs—providing one of the largest CGN proteomes [31,32] (Fig. 5A). From a single 3D HILIC–SCX–RP/PGC separation, we identified 1816 proteins and

Fig. 5. (A) Venn diagram of proteins and peptides identified by the 3D HILIC–SCX–RP/PGC LC system from the CGNs digests with iTRAQ tags. (B) Distribution of GRAVY values across the PGC and HILIC fractions, represented by a box-whisker plot (box range: 25–75%; whisker range: 5–95%); on the x-axis, the numbers 1–8 represent HILIC fractions. (C) Average net charges at pH 2.4 of SCX fractions 1 and 2 across the eight HILIC fractions. (D) Linear fitting plot of the HI with respect to the retention time (efficient LC gradient range), in the 16 third-dimension RP LC fractions, of the unique normal peptides identified from monkey brain digests by the 3D HILIC–SCX–RP/PGC LC system.

Please cite this article in press as: Y. Zhao, et al., Online coupling of hydrophilic interaction/strong cation exchange/reversed-phase liquid chromatography with porous graphitic carbon liquid chromatography for simultaneous proteomics and N-glycomics analysis, J. Chromatogr. A (2015), http://dx.doi.org/10.1016/j.chroma.2015.08.017

G Model CHROMA-356768; No. of Pages 10 8

ARTICLE IN PRESS Y. Zhao et al. / J. Chromatogr. A xxx (2015) xxx–xxx

14,116 peptides from the protein digests of the cynomolgus monkey brain sample; there were 535 additional hydrophilic peptides (ca. 3.95% with average HI of 3.88) identified solely from the single PGC fraction. The 3D HILIC–SCX–RP component features three complementary peptide separation processes across the first, second, and third dimensions. Our current observation of a good fraction of protein/proteome coverage provides us with an opportunity to reevaluate the chromatographic properties of the system. The mechanisms for chromatographic separations of peptides through the SCX and RP columns are based mainly on charge and hydrophobicity, respectively. The retention mechanism of the HILIC column at pH 6.8 can be described by the peptide hydrophilicity using the GRAVY value—calculated as the sum of the hydropathy values of all the amino acids divided by the length of the peptide at neutral pH; hydrophobic molecules having higher GRAVY values, meanwhile, are less hydrophilic. Fig. 5B indicates that the GRAVY value distribution of the identified peptides decreased across most of the HILIC fractions (except for the first fraction), as depicted by the box-whisker plot; these results further illustrate the strong correlation between the increased hydrophilicity of the peptides and the degree of HILIC retention. The peptides from the PGC fraction had lower GRAVY values than those in all of the other HILIC fractions, as expected, confirming the capability of the PGC column to retain very hydrophilic analytes from the RP trap flow-through in our new LC system design. The GRAVY values of the identified peptides in the first HILIC fraction were lower than those in some latter HILIC fractions, probably because of the limited solvent compatibility mitigation capacity of the solvent mixing loop at the high organic solvent content (ca. >80% ACN, Supplementary Table 2). Nevertheless, the online solvent conversion worked well for the subsequent seven HILIC fractions, judging from the normal trend with relatively even numbers of proteins and peptides identified across all the HILIC fractions. The retention mechanism of the second-dimension SCX column was revealed after plotting the peptide average net charge at pH 2.4 of the two SCX sub-fractions, respectively, across the eight HILIC fractions. Instead of separating peptides with increasing peptide hydrophilicity in the successive seven HILIC fractions, the average net charge of SCX sub-fraction 2 (SCX 2) was higher than that of SCX sub-fraction 1 (SCX 1) throughout the eight HILIC fractions, demonstrating that the retention mechanism of the SCX column was clearly based on the charge states of the peptides (Fig. 5C). As expected from the hydrophobicitybased retention mechanism of the third-dimension RP column at pH 2, the HIs of the 12,351 identified peptides in the 16 RP fractions correlated well with the separation time (R2 = 0.86; Fig. 5D). These results from our newly developed 3D HILIC–SCX–RP system are in accordance with previous interpretations that the separation mechanisms of the three different LC dimensions—HILIC, SCX, and RP—are based on hydrophilic partitioning interactions, charge, and hydrophobicity, respectively [3,28], but the exact chromatographic mechanisms might share some commonalities. As shown in Supplementary Fig. 5, different separation characteristics among the HILIC, SCX, and RP dimensions were revealed by the varying trends observed in their hydrophobicity index (HI) value, pI and molecular weight (MW) peptide elution profiles. The HILIC peptide elution profile did not feature any prominent trends in terms of the physiochemical properties examined. In contrast, the SCX profile showed substantial decreases in HI values and MW, along with a trend of increasing pI values as the fraction number increased. Examination of the RP column separation characteristics revealed that the peptides eluted with increasing HI values, which was opposite to the trend observed in SCX. The correlation between peptide pI values and MW with respect to their normalized retention time was not appreciable, although a modest upward trend was observed in the MW elution profile. With the observation that

the elution profiles of the three LC dimensions were quite distinct from one another, this suggested that their selectivity is based on different combinations of the physiochemical properties of the peptides being fractionated. To further demonstrate the orthogonality between the HILIC and SCX dimensions, the identified peptides in each fraction of 3D HILIC–SCX–RP were plotted as a function of the first- and second-dimension eluent conditions (Supplementary Fig. 6) [33]. The least-squares fit correlation coefficient was very low (R2 = 0.0004), demonstrating orthogonal fractionation (i.e. poor correlation between the first- and second-dimension elution) was achieved when these two chromatographic stages were combined and utilized for peptide separation. 3.5. N-Glycomics and N-glycoproteomics analysis of cynomolgus monkey plasma using the 3D HILIC–SCX–RP/PGC LC system N-Glycosylation of proteins is one of the most common posttranslational modifications used to render them as functionally active [34]. The structural complexity of glycans of glycoproteins cannot be predicted directly from known genomes; thus, the ability to reliably identify and determine inherent heterogeneous glycan structures is an important step toward exploring the functional roles of N-glycosylation processes in glycoproteins [35]. The ability to determine the exact residues that are N-glycosylated, the associated glycan compositions, and, ideally, the peptide section sequence from a single LC run would greatly facilitate the characterization of N-glycoproteins. After deglycosylation by PNGase F, the masses of the N-glycosylated peptides in the NXS/T sequence increased by 0.98 Th, due to the conversion of the attachment site of the asparagine residue to an aspartic acid unit [36]. Because of their very hydrophilic nature, the N-glycans were flow-throughs not retained by the RP trap column, but were trapped by the PGC trap column for subsequent downstream PGC analyses. The greatest potential use that we envision for this 3D HILIC–SCX–RP/PGC methodology is the large-scale characterization of N-glycoproteins in complex samples, allowing efficient, concomitant analyses of both peptides and glycopeptides/glycans from the PGC and HILIC–SCX–RP systems, respectively, after a single injection event. Unlike the abundance of data for humans, there is little N-glycomics information for cynomolgus monkeys in the Uniprot database [20]. Interestingly, N-glycolylneuraminic acid (NeuGc)-containing N-glycans are seldom identified in healthy human plasma, but have been found in cynomolgus monkey plasma; these common mammalian sialic acid NeuGc residues

Fig. 6. MS/MS spectrum of the N-glycan [Man3 GalGlcNAc4 NeuGc + 2H]2+ , with m/z 893.8226, eluting at 53.22 min, identified from cynomolgus monkey plasma in the PGC dimension of the 3D HILIC–SCX–RP/PGC LC system.

Please cite this article in press as: Y. Zhao, et al., Online coupling of hydrophilic interaction/strong cation exchange/reversed-phase liquid chromatography with porous graphitic carbon liquid chromatography for simultaneous proteomics and N-glycomics analysis, J. Chromatogr. A (2015), http://dx.doi.org/10.1016/j.chroma.2015.08.017

G Model CHROMA-356768; No. of Pages 10

ARTICLE IN PRESS Y. Zhao et al. / J. Chromatogr. A xxx (2015) xxx–xxx

9

Fig. 7. Analysis of N-glycoproteins identified from cynomolgus monkey plasma by the 3D HILIC–SCX–RP/PGC LC system. (A) Distribution of N-glycoproteins with different numbers of N-glycosites from cynomolgus monkey plasma. (B) Concentrations of human plasma N-glycoproteins that were blasted from the corresponding cynomolgus monkey plasma N-glycoproteins identified according to the PeptideAtlas HPPP database [40].

cannot be synthesized by humans [37–39]. Fig. 6 displays a typical MS/MS spectrum of the doubly protonated N-glycan [Man3 GalGlcNAc4 NeuGc + 2H]2+ , with the molecular ion at m/z 893.8226 eluting after 53.22 min; peaks matching m/z 290.0867 [NeuGc − H2 O + H]+ and m/z 308.0954 [NeuGc + H]+ verified the presence of NeuGc sialylation in this N-glycan. Supplementary Table 8 summarizes the N-glycan profiling results determined from the PGC system; from a single PGC fraction we identified 38 NeuGc-containing N-glycans out of the 122 N-glycans including 5 high-mannose, 37 hybrid, and 80 complex types; for the first time, we used MS/MS to also confirm the presence of NeuGc residues in healthy cynomolgus monkey plasma (Fig. 6). As listed in Supplementary Table 9, there were, in total, 132 probable N-glycopeptides and 135 probable N-glycosites belonging to 62 probable N-glycoproteins identified from the monkey plasma. Supplementary Table 9 lists the exact N-glycosite motifs in the cynomolgus monkey proteins—valuable N-glycoproteomics information for the cynomolgus monkey. There were, in total, 168 proteins corresponding to 10,627 unique peptides identified from cynomolgus monkey plasma through a single analysis; among them, 122 proteins were probable N-glycoproteins with potential N-glycosylation sites of NXS/T (Supplementary Table 10). The N-glycopeptide sequences were first identified from the MS/MS spectra. Each of the proposed N-glycopeptides was subsequently confirmed by blasting with the corresponding sequence of the human protein in the Uniprot database; an N-glycosite was predicted by the NetNGlyc 1.0 server (http://www.cbs.dtu. dk/services/NetNGlyc/). In this current study, such a motif and the constituted peptide were counted as a probable N-glycosite and N-glycopeptide, respectively, of the cynomolgus monkey proteins. On the basis of our results above, 35.5%, 37.1%, 12.9%, 6.5%, 4.8%, and 3.2% of the cynomolgus monkey plasma N-glycoproteins contained one, two, three, four, five, and six N-glycosites, respectively (Fig. 7A). We also compared and confirmed the N-glycan and deglycosylated peptide data with our previous identified intact Nglycopeptide results for the same sample, to provide orthogonal information leading to more unambiguous identification [36,40], including details regarding the N-glycosylation sites and the associated N-glycan compositions, as well as sequences at each site. Supplementary Table 11 lists all the proposed N-glycopeptides including 68 peptide backbone sequences, 68 N-glycosites, and 54 detailed site-specific N-glycan structures from 44 N-glycoproteins identified, for the first time. Parenthetically, we further explored the performance of this new system by investigating the dynamic range of the lysate of the cynomolgus monkey plasma. Because of the scarcity of information available regarding protein concentrations in the cynomolgus monkey plasma and the homology between cynomolgus monkeys and humans, for our estimations we used the concentrations of the

corresponding human plasma proteins after blasting. We successfully mapped 112 out of 122 identified plasma N-glycoproteins to the human proteins in the PeptideAtlas HPPP database [41] (Supplementary Table 10). Our current estimated overall dynamic range of the 3D HILIC–SCX–RP platform spans over approximately seven orders of magnitude; the lowest-abundance N-glycoprotein that we identified was present at 0.29 ng/mL (collagen alpha-1(I) chain), while the highest was present at 4.0 × 107 ng/mL (P02768). More importantly, the distribution of the entire range of identified Nglycoproteins in Fig. 7B demonstrates that up to 10 N-glycoproteins (8.2%) were mapped with abundances of less than 10 ng/mL, implying not only the deep proteome penetration of the platform but also its remarkable ability to identify low-abundance N-glycoproteins.

4. Conclusions We have developed a fully automated online 3D HILIC–SCX–RP/PGC LC system that enhances both proteomics and N-glycomics sampling in two ways: (i) applying efficient 3D orthogonal chromatographic separation mechanisms—HILIC, SCX, and RP dimensions—in the 3D HILIC–SCX–RP LC system for the analyses of complex peptide mixtures featuring a diverse range of hydrophobicities and, thereby, maximize the proteome and protein coverage that can be identified in a given period of time; (ii) integrating the system downstream with a PGC column to extend the range of extremely hydrophilic peptides and glycans that can be identified from a single sample injection and, thereby, provide additional structural information. We have demonstrated the excellent fidelity and applicability of the 3D system through qualitative and quantitative proteomics, as well as through analyses of the N-glycoproteins of various complex biological samples, including the total lysates of S. cerevisiae and primary CGNs and cynomolgus monkey brain tissue. We have used this MDLC technology to identify one of the largest CGNs proteomes at low-microgram levels. The new 3D HILIC–SCX–RP integrated system features an additional online PGC column to trap and analyze very hydrophilic analytes, which would have been non-retained in the 3D HILIC–SCX–RP LC system alone, resulting in simultaneous identifications of glycans and their associated deglycosylated peptides, as well as the exact sites of N-glycosylation, after only one sample injection event. With such a 3D HILIC–SCX–RP/PGC platform in hand, we performed the first detailed simultaneous N-glycomics and N-glycoproteomics analyses of cynomolgus monkey plasma, establishing a glycan library containing 122 proposed N-glycans with detailed complementary information regarding proposed N-glycosylation sites and N-glycoproteins. This newly available information should be useful for future related research into cynomolgus monkeys.

Please cite this article in press as: Y. Zhao, et al., Online coupling of hydrophilic interaction/strong cation exchange/reversed-phase liquid chromatography with porous graphitic carbon liquid chromatography for simultaneous proteomics and N-glycomics analysis, J. Chromatogr. A (2015), http://dx.doi.org/10.1016/j.chroma.2015.08.017

G Model CHROMA-356768; No. of Pages 10 10

ARTICLE IN PRESS Y. Zhao et al. / J. Chromatogr. A xxx (2015) xxx–xxx

Acknowledgments We thank the Hong Kong Research Grants Council (project nos. HKU 701613P and 17306015P), the University of Hong Kong and the Science and Technology Program of Guangzhou (2014J4100097), the 111 Project (No. B13038) for financial support. We thank Prof. Yuqiang Wang, Prof. Simon M. Y. Lee, Dr. Ricky P. W. Kong, Dr. Samuel S. W. Szeto and Ms. Yuko P. Y. Lam for their assistance in the preparation of the monkey sample and CGNs lysates and many helpful discussions. We thank Ms. I-Sheng Chang and Dr. Maggie P. Y. Lam for their help in the preliminary N-glycomics data analysis. We also thank the School of Biological Science, the University of Hong Kong, for access to the TripleTOF 5600 mass spectrometer, and Professor R. S. S. Wu and Ms. Apple Chu for helpful discussions. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.chroma.2015.08. 017. References [1] B. Domon, R. Aebersold, Review: Mass spectrometry and protein analysis, Science 312 (2006) 212–217. [2] M.X. Gao, D.W. Qi, P. Zhang, C.H. Deng, X.M. Zhang, Development of multidimensional liquid chromatography and application in proteomic analysis, Expert Rev. Proteomics 7 (2010) 665–678. [3] M. Gilar, P. Olivova, A.E. Daly, J.C. Gebler, Orthogonality of separation in twodimensional liquid chromatography, Anal. Chem. 77 (2005) 6426–6434. [4] M.P. Washburn, D. Wolters, J.R. Yates, Large-scale analysis of the yeast proteome by multidimensional protein identification technology, Nat. Biotechnol. 19 (2001) 242–247. [5] P.J. Boersema, N. Divecha, A.J.R. Heck, S. Mohammed, Evaluation and optimization of ZIC–HILIC–RP as an alternative MudPIT strategy, J. Proteome Res. 6 (2007) 937–946. [6] S. Di Palma, M.L. Hennrich, A.J.R. Heck, S. Mohammed, Recent advances in peptide separation by multidimensional liquid chromatography for proteome analysis, J. Proteomics 75 (2012) 3791–3813. [7] E. Lau, M.P.Y. Lam, S.O. Siu, R.P.W. Kong, et al., Combinatorial use of offline SCX and online RP-RP liquid chromatography for iTRAQ-based quantitative proteomics applications, Mol. Biosyst. 7 (2011) 1399–1408. [8] M. Gilar, P. Olivova, A.E. Daly, J.C. Gebler, Two-dimensional separation of peptides using RP–RP–HPLC system with different pH in first and second separation dimensions, J. Sep. Sci. 28 (2005) 1694–1703. [9] H. Malerod, E. Lundanes, T. Greibrokk, Recent advances in on-line multidimensional liquid chromatography, Anal. Methods-UK 2 (2010) 110–122. [10] Y. Zhao, R.P.W. Kong, G.H. Li, M.P.Y. Lam, et al., Fully automatable twodimensional hydrophilic interaction liquid chromatography–reversed phase liquid chromatography with online tandem mass spectrometry for shotgun proteomics, J. Sep. Sci. 35 (2012) 1755–1763. [11] A.W. Moore, J.W. Jorgenson, Comprehensive three-dimensional separation of peptides using size exclusion chromatography/reversed phase liquid chromatography/optically gated capillary zone electrophoresis, Anal. Chem. 67 (1995) 3456–3463. [12] J. Wei, J. Sun, W. Yu, A. Jones, et al., Global proteome discovery using an online three-dimensional LC–MS/MS, J. Proteome Res. 4 (2005) 801–808. [13] H.J. Issaq, K.C. Chan, G.M. Janini, T.P. Conrads, T.D. Veenstra, Multidimensional separation of peptides for effective proteomic analysis, J. Chromatogr. B 817 (2005) 35–47. [14] L. Pereira, Porous graphitic carbon as a stationary phase in HPLC: Theory and applications, J. Liq. Chromatogr. Relat. Technol. 31 (2008) 1687–1731. [15] C.S. Chu, M.R. Ninonuevo, B.H. Clowers, P.D. Perkins, et al., Profile of native N-linked glycan structures from human serum using high performance liquid chromatography on a microfluidic chip and time-of-flight mass spectrometry, Proteomics 9 (2009) 1939–1951. [16] P.L. Ross, Y.L.N. Huang, J.N. Marchese, B. Williamson, et al., Multiplexed protein quantitation in Saccharomyces cerevisiae using amine-reactive isobaric tagging reagents, Mol. Cell. Proteomics 3 (2004) 1154–1169. [17] S. Ghaemmaghami, W. Huh, K. Bower, R.W. Howson, et al., Global analysis of protein expression in yeast, Nature 425 (2003) 737–741.

[18] F. Zhou, T.W. Sikorski, S.B. Ficarro, J.T. Webber, J.A. Marto, Online nanoflow reversed phase–strong anion exchange–reversed phase liquid chromatography–tandem mass spectrometry platform for efficient and indepth proteome sequence analysis of complex organisms, Anal. Chem. 83 (2011) 6996–7005. [19] Z.J. Zhang, W. Cui, G.H. Li, S. Yuan, et al., Baicalein protects against 6-OHDAinduced neurotoxicity through activation of Keap1/Nrf2/HO-1 and involving PKC alpha and PI3K/AKT signaling pathways, J. Agric. Food Chem. 60 (2012) 8171–8182. [20] R. Apweiler, M.J. Martin, C. O’Donovan, M. Magrane, et al., Update on activities at the Universal Protein Resource (UniProt) in 2013, Nucleic Acids Res. 41 (2013) D43–D47. [21] I.V. Shilov, S.L. Seymour, A.A. Patel, A. Loboda, et al., The Paragon Algorithm, a next generation search engine that uses sequence temperature values and feature probabilities to identify peptides from tandem mass spectra, Mol. Cell. Proteomics 6 (2007) 1638–1655. [22] W.H. Tang, I.V. Shilov, S.L. Seymour, Nonlinear fitting method for determining local false discovery rates from decoy database searches, J. Proteome Res. 7 (2008) 3661–3667. [23] A. Ceroni, K. Maass, H. Geyer, R. Geyer, et al., GlycoWorkbench: A tool for the computer-assisted annotation of mass spectra of glycans, J. Proteome Res. 7 (2008) 1650–1659. [24] O.V. Krokhin, V. Spicer, Peptide retention standards and hydrophobicity indexes in reversed-phase high-performance liquid chromatography of peptides, Anal. Chem. 81 (2009) 9522–9530. [25] O.V. Krokhin, Sequence-specific retention calculator Algorithm for peptide retention prediction in ion-pair RP-HPLC: Application to 300- and 100angstrom pore size C18 sorbents, Anal. Chem. 78 (2006) 7785–7795. [26] J. Kyte, R.F. Doolittle, A simple method for displaying the hydropathic character of a protein, J. Mol. Biol. 157 (1982) 105–132. [27] P.A. Sims, Use of a spreadsheet to calculate the net charge of peptides and proteins as a function of pH: An alternative to using “canned” programs to estimate the isoelectric point of these important biomolecules, J. Chem. Educ. 87 (2010) 803–808. [28] M.P.Y. Lam, E. Lau, S.O. Siu, D.C.M. Ng, et al., Online combination of reversedphase/reversed-phase and porous graphitic carbon liquid chromatography for multicomponent separation of proteomics and glycoproteomics samples, Electrophoresis 32 (2011) 2930–2940. [29] U. Lewandrowski, A. Sickmann, Online dual gradient reversed-phase/porous graphitized carbon nanoHPLC for proteomic applications, Anal. Chem. 82 (2010) 5391–5396. [30] B. Domon, C.E. Costello, A systematic nomenclature for carbohydrate fragmentations in FAB–MS MS spectra of glycoconjugates, Glycoconjugate J. 5 (1988) 397–409. [31] J.W. Yang, R. Rodrigo, V. Felipo, G. Lubec, Proteome analysis of primary neurons and astrocytes from rat cerebellum, J. Proteome Res. 4 (2005) 768–788. [32] E. Thouvenot, S. Urbach, O. Vigy, M. Seveno, et al., Quantitative proteomic analysis reveals protein expression changes in the murine neuronal secretome during apoptosis, J. Proteomics 77 (2012) 394–405. [33] F. Zhou, Y. Lu, S.B. Ficarro, G. Adelmant, W. Jiang, C.J. Luckey, J.A. Marto, Genomescale proteome quantification by DEEP SEQ mass spectrometry, Nat. Commun. (2013), http://dx.doi.org/10.1038/ncomms3171. [34] K.W. Moremen, M. Tiemeyer, A.V. Nairn, Vertebrate protein glycosylation: Diversity, synthesis and function, Nat. Rev. Mol. Cell Biol. 13 (2012) 448–462. [35] F. Tousi, W.S. Hancock, M. Hincapie, Technologies and strategies for glycoproteomics and glycomics and their application to clinical biomarker research, Anal. Methods-UK 3 (2011) 20–32. [36] S. Pan, R. Chen, R. Aebersold, T.A. Brentnall, Mass spectrometry based glycoproteomics: From a proteomics perspective, Mol. Cell. Proteomics (2011), http:// dx.doi.org/10.1074/mcp.R110.003251. [37] A. Varki, Loss of N-glycolylneuraminic acid in humans: Mechanisms, consequences, and implications for hominid evolution, Yearb. Phys. Anthropol. 44 (2001) 54–69. [38] A. Varki, Multiple changes in sialic acid biology during human evolution, Glycoconjugate J. 26 (2009) 231–245. [39] S.L. Diaz, V. Padler-Karavani, D. Ghaderi, N. Hurtado-Ziola, et al., Sensitive and specific detection of the non-human sialic acid N-glycolylneuraminic acid in human tissues and biotherapeutic products, PLoS One (2009), http://dx.doi. org/10.1371/journal.pone.0004241. [40] I.M. Lazar, A.C. Lazar, D.F. Cortes, J.L. Kabulski, Recent advances in the MS analysis of glycoproteins: Theoretical considerations, Electrophoresis 32 (2011) 3–13. [41] E.W. Deutsch, G.S. Omenn, D.S. Campbell, et al., A high-confidence human plasma proteome reference set with estimated concentrations in PeptideAtlas, Mol. Cell. Proteomics (2011), http://dx.doi.org/10.1074/mcp.M110.006353.

Please cite this article in press as: Y. Zhao, et al., Online coupling of hydrophilic interaction/strong cation exchange/reversed-phase liquid chromatography with porous graphitic carbon liquid chromatography for simultaneous proteomics and N-glycomics analysis, J. Chromatogr. A (2015), http://dx.doi.org/10.1016/j.chroma.2015.08.017