Stable isotope dimethyl labelling for quantitative proteomics and beyond

Abstract

Stable-isotope reductive dimethylation, a cost-effective, simple, robust, reliable and easy-to- multiplex labelling method, is widely applied to quantitative proteomics using liquid chromatography-mass spectrometry. This review focuses on biological applications of stable-isotope dimethyl labelling for a large-scale comparative analysis of protein expression and post-translational modifications based on its unique properties of the labelling chemistry. Some other applications of the labelling method for sample preparation and mass spectrometry-based protein identification and characterization are also summarized.

This article is part of the themed issue ‘Quantitative mass spectrometry’.

1. Introduction

Comparative or quantitative analysis of protein expression in two biological states (e.g. normal versus disease or treated versus non-treated states) is a key focus of proteomics studies [1,2], which aim to discover new biomarkers or gain mechanistic understandings [3]. Two-dimensional gel electrophoresis [4], separating hundreds to thousands of proteins present in a highly complicated biological sample on a single gel, is traditionally used for protein profiling. However, time-consuming procedures such as spot-by-spot in-gel digestion followed by liquid chromatography-mass spectrometry (LC-MS) analysis and the lack of sensitivity for low-abundant proteins have limited its use in large-scale proteomics study [5].

Alternatively, as shown in figure 1, stable-isotope labelling coupled with LC-MS analysis is able to simultaneously provide quantitative and qualitative (protein identification) information in a single run and has become a popular method for large-scale analysis of protein expression [1]. The labelling methods can be divided into three categories, namely metabolic labelling [6,7], enzymatic labelling [8] and chemical labelling [9–11]. Isotopic elements are incorporated into the sample at the protein level (metabolic and chemical labelling) or at the peptide level during (enzymatic labelling) or after (chemical labelling) enzymatic digestion. Using shotgun proteomics approach, the digested peptides derived from two compared samples are combined and separated by LC or two-dimensional liquid chromatography (2D-LC), and then analysed by MS and tandem mass spectrometry (MS/MS) (figure 1). Proteins are identified based on peptide sequencing (MS/MS) via database search. Using the same set of data, the same proteins derived from two biological states are quantified based on the relative peak area of extracted ion chromatograms (XICs) of the precursor ion pair (MS) or based on the relative intensity of the tandem mass tag (MS/MS) of the isobaric ion pair. As summarized in table 1, metabolic labelling is referred to as stable-isotope labelling by amino acids in cell culture (SILAC) and enzymatic labelling is mainly referred to as H₂O/H₂¹⁸O labelling.

Figure 1.

Stable-isotope labelling methods for MS-based quantitative proteomics. (Online version in colour.)

Table 1.

Stable-isotope labelling methods for MS-based quantitative proteomics.

method	labelling site	ΔMw (isotopic pair)	labelling performancea	MS or MS/MS (quantification)	sample type	duplex/multiplex	cost	refs
metabolic labelling
SILAC	Lys and Arg	6 Da	+++	MS	cell culture	duplex or triplex	$$$	[6,7]
enzymatic labelling
H2O/H218O	C-termini	4 Da	++	MS	all	duplex	$$	[8]
chemical labelling
ICAT	Cys	8 Da	+++	MS	all	duplex	$$$	[9]
photocleavable ICAT	Cys	7 Da	+++	MS	all	duplex	$$$	[10]
cICAT	Cys	9 Da	+++	MS	all	duplex	$$$	[11]
methanol-d3	Asp, Glu, C-termini	3 Da	+	MS	all	duplex	$	[12]
acetylation	N-termini and Lys	3 Da	+++	MS	all	duplex	$	[13]
2-methoxy-4,5-dihydro-1H-imidazole	Lys	4 Da	+	MS	all	duplex	$$	[14]
dimethyl labelling	N-termini and Lys	4 Da	+++	MS	all	duplex, triplex, four-plex or five-plex	$	[15–18]
tandem mass tag	N-termini and Lys	isobaric	++	MS/MS	all	duplex	$$$	[19]
iTRAQ	N-termini and Lys	isobaric	+++	MS/MS	all	four-plex or eight-plex	$$$	[20,21]

^aThe labelling performance is evaluated by the reaction rate and specificity of each labelling reagent. ‘+++’, ‘++’ and ‘+’ represent very good, good and fair, respectively.

For chemical labelling, dozens of stable-isotope reagents (table 1) mainly targeting amine or carboxylic acid groups have been reported [12–21] and reviewed recently [22–31]. Each chemical labelling method has advantages and disadvantages. For example, isotope-coded affinity tag (ICAT) was designed to incorporate selective affinity enrichment for cysteine residues but non-cysteine containing peptides were not labelled. Isobaric tags for relative and absolute quantitation (iTRAQ) are able to compare multiple samples (multiplex) up to eight-plex [20,21]. However, reactive N-hydroxysuccinimide (NHS) ester reagents contained in iTRAQ kits are not stable for long-term storage. In contrast with ICAT and iTRAQ, which are commercial kits, stable-isotope dimethyl labelling by reductive amination [15–18] uses commercial reagents and is superior to others in high reaction yield, high accuracy and high reproducibility, as well as robustness, simplicity and low cost [32,33]. It was claimed that stable-isotope dimethyl labelling, SILAC and iTRAQ are three benchmarking labelling methods for quantitative proteomics [32]. In addition to a general review of quantitative proteomics using stable-isotope dimethyl labelling, the use of unique properties of dimethyl labelling for sample preparation and for MS-based protein characterization is also included in this review.

2. Labelling chemistry for liquid chromatography-mass spectrometry based quantitative proteomics

Reductive amination is a well-known organic reaction commonly used in protein chemistry [34,35]. As depicted in figure 2a, formaldehyde forms Schiff base with the N-terminus or ϵ-amino group of Lys residue of a peptide/protein; the aldimine intermediate is then reduced by sodium cyanoborohydride (NaBH₃CN) to form a secondary amine which is more nucleophilic than the primary amine and immediately reacts with another formaldehyde molecule to form dimethylamino group. Reductive amination is a global labelling method for all peptides with one labelling site for peptides without Lys residue or 1 + N labelling sites for peptides with N Lys residues. All amino groups are dimethylated except the N-termini of Pro residue, which is a secondary amino group and is monomethylated by the reaction. It is particularly notable that dimethylated peptides exhibit remarkable a₁ ion enhancement by collision induced dissociation (CID) (figure 2b), possibly via CO neutral loss from the b₁ ion [36]. The strong a₁ fragment ion is very useful for MS-based protein characterization and will be discussed later. Because formaldehyde (CH₂O) is small and water-soluble, the labelling reaction is quick, specific and complete.

Figure 2.

Reaction mechanism of (a) reductive dimethylation (dimethyl labelling) and (b) a₁ ion formation of dimethylated peptides by CID.

Stable-isotope dimethyl labelling using reductive amination for MS-based quantitative proteomics was first reported in 2003 [15]. It is compatible with most chromatographic methods for peptide separation using reversed phase (RP) [15], strong cation exchange (SCX) [37], hydrophilic interaction chromatography (HILIC) [38], immobilized metal ion affinity chromatography (IMAC) [39] column or with multi-dimensional separation by combining two or three separation methods [37]. Because the number of positive charges borne by a peptide is normally increased by dimethylation, fragmentation efficiency of CID or electron transfer dissociation is likely to be enhanced [40]. This increases the success rate of protein identification using database search based on MS/MS spectra. There were some concerns when the method was first reported: (i) shift of the retention time caused by isotopic (deuterium) effect [41], (ii) overlapping of isotopic clusters due to the relatively small mass difference, and (iii) lack of quantification software [15,42]. The first problem was solved by using XICs for quantitative analysis instead of ion intensities [15,43–45]. The second problem was simulated by MS-Isotope (embedded in the Protein-Prospector) software and found to be negligible in affecting accuracy when compared with the ratio calculated from ‘pure’ monoisotopic peaks [44] except for extremely large peptides (more than 3 kDa) bearing single labelling site [46]. The third concern is not a problem now because many available softwares such as Proteome Discoverer (Thermo Fisher Scientific), Mascot Distiller (Matrix Science), XPRESS [47], MaxQuant [48], PVIEW [49] and MSQuant [50] can be applied for any kinds of labelling methods including dimethyl labelling.

Because dimethyl labelling is relatively fast, specific and mild, it is perfectly suitable for online or on-column protocols via in situ labelling to improve the throughput and automation. Heck and co-workers standardized three types of protocols for stable-isotope dimethyl labelling: in-solution, online and on-column using solid-phase extraction (SPE) to meet the requirements of different sample amounts [18,43]. Using online labelling, a fully automated analysis method can be developed including the following procedures: (1) the peptide sample was loaded onto a trap column connected by a six-port valve; (2) the peptides retained on the column were chemically labelled by flushing the trapping column with the labelling reagents; (3) excess reagents were washed away by online desalting; (4) the second sample underwent steps (1), (2) and (3) sequentially; (5) the valve was switched to the analytical column connected to MS for a regular LC-MS analysis. The new protocols simplified several time-consuming handling steps including removal of amine-containing ingredients (e.g. NH₄HCO₃), addition of the labelling regents, incubation and quenching for the reaction, desalting and concentration of the product, leading to less sample loss and better reproducibility [51,52]. Raijmakers et al. applied this on-column labelling protocol to determine the differences in composition between bovine liver and spleen 20S core proteasome complexes [52]. Zou et al. integrated this on-column sequential labelling protocol with a RP-SCX biphasic column for multi-dimensional separation and high-throughput quantitative proteome analysis [51].

Although dimethyl labelling is commonly used for comparative quantification of two samples, it can be extended for multiplex analysis based on MS spectra. Different isotopic forms of formaldehyde such as CD₂O or ¹³CH₂O and sodium cyanoborohydride such as NaBD₃CN or NaBH₃CN are commercially available at low cost. As summarized in table 2, multiplex labelling can be easily achieved by different combinations of reagents to generate a panel of mass differences and used to compare multiple samples acquired from different time points or different dosages [20]. For example, four-plex analysis by dimethyl labelling can be achieved by using reagent combinations of CH₂O/NaBH₃CN, CH₂O/NaBD₃CN, CD₂O/NaBH₃CN and CH₂O/NaBD₃CN [16]. The multiplex can be further extended to five-plex using an additional isotopic reagent combination like ¹³CD₂O/NaBD₃CN [17]. These combinations were coupled with Lys-C enzyme to generate at least two labelling sites (N-termini and Lys) for each peptide, ensuring at least 4 Da of the mass shift between the nearest labelled forms in order to minimize overlapping of neighbouring isotopic clusters, increasing the accuracy of quantification. However, peptides generated by Lys-C are relatively long and may not be suitable for conventional MS/MS-based protein identification strategy. Therefore, other combinations like CH₂O/NaBH₃CN, CD₂O/NaBH₃CN and ¹³CD₂O/NaBD₃CN were used to enlarge the mass shift of the neighbouring isotopic peaks [18]. Tandem mass tags may also be generated by stable-isotope dimethyl labelling using mass defect-based pseudo-isobaric dimethyl labelling (pIDL) reagents, CD₂O/NaBD₃CN and ¹³CD₂O/NaBH₃CN, coupled with LysC digestion [53]. The precursor ion pair generated by the pIDL method have an almost identical mass (isobaric tag); the quantification ratios were calculated based on the intensity of a₁ fragment ion pair which exhibit a small mass difference (0.00584 Da) resolved by high-resolution MS/MS [53].

Table 2.

Combinations of stable-isotope reagents for multiplex dimethyl labelling.

	reagent
method	CH2O + NaBH3CN + 28 Da per site	CH2O + NaBD3CN + 30 Da per site	CD2O + NaBH3CN + 32 Da per site	CD2O + NaBD3CN + 34 Da per site	13CD2O + NaBD3CN + 36 Da per site	13CD2O + NaBH3CN + 34 Da per site	enzyme	refs
duplex	▴		▴				trypsin	[15]
triplex	▴		▴		▴		trypsin	[18]
four-plex	▴	▴	▴	▴			Lys-C	[16]
five-plex	▴	▴	▴	▴	▴		Lys-C	[17]
pIDL				▴		▴	Lys-C	[53]

3. Applications for quantitative proteomics

Stable-isotope dimethyl labelling has been widely used in quantitative proteomics since the method was first reported in 2003 [54]. In 2012, Heck et al. published a review on applications of stable-isotope dimethyl labelling mainly for global (or large-scale) expression profiling, quantitative analyses of post-translational modifications (PTMs) and protein interaction. In this review, we briefly update applications either not included in or reported after Heck's review (2012 until now).

(a). Global protein-expression profiling

Like most proteomics studies, many early applications of stable-isotope dimethyl labelling focused on finding biomarkers of diseases or key proteins involved in pathways of cell models [55–59]. D'Aguanno et al. used stable-isotope dimethyl labelling to investigate the role p63 isoforms play in cancer stem cells (CSCs) [60]. Colon CSCs derived from primary tumour were transduced with lentiviral vectors carrying either p63 containing (TA) or lacking (ΔN) gene. The light and heavy labelled tryptic peptides from TAp63, ΔNp63, and control colon CSC cells were combined and separated using immobilized pH gradient isoelectric focusing followed by RP chromatographic separation and the separated peptides were analysed by LC-MS/MS. Proteins were identified and quantified using Mascot Distiller software. Differentially expressed proteins were functionally classified and annotated by bioinformatics tools. Their data showed that most modulated proteins were involved in metabolic processes and these proteins were confirmed by western blotting and targeted label-free quantitative analysis [60]. Swa et al. used stable-isotope dimethyl labelling coupled with integrative protein–protein interaction network analysis to reveal the role that annexin-1 may play in the process of developing mammary tumourigenesis [61]. Sato et al. applied triplex dimethyl labelling using CH₂O/NaBH₃CN (light), CD₂O/NaBH₃CN (medium) and ¹³CD₂O/NaBD₃CN (heavy) reagents to samples derived from the pooled, control and treated tissues, respectively. The combined samples were analysed by nanoLC-high-resolution LTQ Orbitrap Velos mass spectrometry. Among 6694 identified proteins, proteins belonging to the eukaryotic initiation factor (eIF) families were upregulated in the control compared with those in inhibitor-treated tissues which exhibited lung metastases. The data implied that eIF family members, especially eIF4A1 and eEF2, are highly correlated to the metastatic phenotype of advanced breast cancer [62]. Chiang et al. used duplex dimethyl labelling coupled with nanoLC-MS/MS to comparatively analyse urine proteins from patients with and without urothelial carcinoma [63]. Among 219 candidate proteins identified, SH3 domain binding glutamic acid-rich protein like 3 (SH3BGRL3) was identified to overexpress in patient urines. This finding positively correlated with the higher histological grading and muscle invasiveness of urothelial carcinoma, indicating SH3BGRL3 may be a potential biomarker for bladder cancer [63].

In addition to biomarker discovery and biomedical research, quantitative proteomics using stable-isotope dimethyl labelling was also applied to other fields such as food science and energy. Ho et al. used stable-isotope dimethyl labelling to quantitatively analyse signature tryptic peptide (SFMFGGLASGETR) derived from a marker protein, herbicide-resistantgene-related protein 5-enolpyruvylshikimate-3-phosphate synthase (CP4EPSPS), derived from genetically modified soya bean [64]. To enrich CP4EPSPS from other high abundant proteins such as glycinin and β-conglycinin, strong anion exchange and SDS-PAGE were used for sample preparation. The combined dimethylated tryptic peptides of CP4EPSPS were identified and quantified using matrix assisted laser desorption and ionization time of flight MS and the data showed excellent linear correlation with the content of genetic modification in soya bean samples [64]. Tolonen et al. applied duplex dimethyl labelling for quantitative proteomics of Clostridium phytofermentans in the medium of cellulosic substrates versus glucose. Their results indicated that secreted carbohydratases accompanied with glycolytic enzymes and alcohol dehydrogenases were upregulated upon the treatment of cellulosic substrates, promoting hemicellulose or cellulose degradation and ethanol production. They concluded that quantitative proteomics studies can provide valuable information to improve biomass fermentation for industrial applications [65].

(b). Quantitative analysis of protein post-translational modifications

Protein PTMs play crucial roles in modulating biological functions and more than 400 PTMs have been reported so far. However, identification and quantification of site-specific PTMs by MS are still challenging. Enrichment steps are normally required due to low abundance of the modified proteins and ion suppression effect inherent with electrospray ionization [66]. Many targeted enrichment methods, mainly for phospho/glyco peptides, have been developed [67–69] but relatively large amounts of starting samples are still required for identifying low-abundant PTMs. Owing to the low cost, stable-isotope dimethyl labelling has been a better choice than other expensive labelling kits for these studies [54]. In table 3, we summarize applications of stable-isotope dimethyl labelling for quantitative analysis of phosphoproteomics and glycoproteomics in terms of sample type and amount of total protein used, labelling approach, enrichment method, quantification software and efficiency. These applications are briefly discussed in the following.

Table 3.

Applications of stable-isotope dimethyl labelling for quantitative analysis of protein phosphorylation and glycosylation.

sample	multiplex	enrichment method	quantification software	no. modified protein/ peptide identified (total protein amount used)	refs
zebra fish embryos	duplex	SCX, online RP-TiO2	MSQuant	348 phosphoproteins (100 µg)	[70]
phosphoproteomics
MCF cells	duplex	IMAC-HILIC	Mascot distiller	2857 unique phosphorylation sites (1 mg)	[39]
HeLa cells	triplex	immunoprecipitation (PY99-agarose beads)	MSQuant	1100 unique phosphopeptides (4 mg)	[71]
membrane fraction of hESCs and NSCs	duplex	TiSH: TiO2-SIMAC-HILIC	LOESS or DanteR package	10 087 phosphopeptides (200 µg)	[72]
porcine muscle	triplex	TiSH	Proteome Discoverer	784 unique phosphopeptides (200 µg)	[73]
INS-1 cells	duplex (CH2O/NaBH3CN and 13CH2O/NaBD3CN)	TisH	Proteome Discoverer	6600 phosphopeptides (300 µg)	[74]
human liver tissues	duplex	in situ sample processing approach, Ti4+-IMAC	MaxQuant	8548 unique phosphopeptides (0.5 mg)	[75]
liver tissues from C57BL/6 mice	duplex	hydroxy acid-modified metal oxide chromatography (HAMMMOC)	Mass Navigator	1090 unique phosphopeptides (25 µg)	[76]
heart tissues from C57BL/6 mice	triplex	SCX	Proteome Discoverer	525 unique phosphopeptides	[77]
yeast mitochondria	Triplex	IMAC (Phos-Select Iron Affinity Gel)	MassChroQ	670 unique phosphopeptides (1.3 mg)	[78]
glycoproteomics
plasma from C57/Bl6 mice	duplex	lectin affinity chromatography	XPRESS	708 glycoproteins were identified (40 µl of plasma)	[79]
serum samples from patients of liver cirrhosis and HCC patients	duplex	hydrazide capture	MSQuant	1179 glycopeptides were identified (100 µl of serum)	[80]
membrane protein fraction of rat myocardial tissue	duplex	TiO2 (for sialic acid-containing glycopeptides) + ZIC-HILIC (for neutral glycopeptides)	MSQuant and StatQuant	590 non-redundant glycosylation sites were identified and quantified	[81]

Quantitative phosphoproteomics is one of the hot topics in PTM studies and is also the most common applications of stable-isotope dimethyl labelling for PTM analysis. Common phosphopeptide enrichment methods include immobilized antibody [71], Fe²⁺-IMAC [82], TiO₂ [70] and SCX [77]. In addition, Wakabayashi et al. developed hydroxy acid-modified metal oxide chromatography (HAMMMOC) to enhance the specificity of TiO₂ [76]. Impressively, HAMMMOC method was able to identify 1090 unique phosphopeptides with as little as 25 µg of total protein contained in the sample. Larsen et al. developed a so-called TiSH three-stage purification approach, in which phosphopeptides were pre-fractionated using TiO₂, followed by sequential elution from IMAC (SIMAC) to separate multi- and mono-phosphorylated peptides and the mono-phosphorylated peptides were further fractionated by HILIC [72–74]. Phosphopeptide enrichment using single step enrichment may not provide sufficient specificity and many studies have applied multi-dimensional separation and enrichment for phosphopeptides.

Huang et al. reported the first application of dimethyl labelling for quantitative phosphoproteomics in which duplex dimethyl labelling combined with IMAC was used to analyse protein phosphorylation induced by 8-bromo-cGMP in pregnant rat uteri [82]. Lemeer et al. used SCX to fractionate dimethylated peptides followed by online affinity enrichment using TiO₂ and LC-MS/MS analysis for comparative study of protein phosphorylation between Fyn/Yes knockdown zebrafish embryos and the wild-type [70]. Wu et al. combined IMAC and HILIC for the study of quantitative phosphoproteomics involved in estrogen-induced transcriptional regulation in MCF-7 breast cancer cells [39]. In their studies, a total of 2857 unique phosphorylation sites contained in 1338 phosphoproteins were identified and quantified. Boersema et al. reported an efficient quantitative profiling method using triplex dimethyl labelling coupled with peptide-level immunoprecipitation for comparative analysis of low-abundant tyrosine phosphorylated proteins in HeLa cells upon treatment with pervanadate or epidermal growth factor [71]. Song et al. compared conventional duplex with triplex dimethyl labelling and they found that the triplex labelling increased the number of quantified phosphopeptides by 50% and reduced the processing time by 50% [83]. We note that stable-isotope dimethyl labelling is able to couple with almost any phosphopeptide enrichment methods, yielding excellent quantitative results.

Protein glycosylation is another crucial PTM that controls many biological processes such as cell growth, cell migration, cell adhesiveness, immune response and tumour cell proliferation [84,85]. Stable-isotope dimethyl labelling has also been widely employed in quantitative glycoproteomics [79–81,86]. As summarized in table 3, common glycopeptide enrichment methods include lectin affinity chromatography [79], hydrazide capture chemistry [80], and TiO₂ combined with zwitterionic HILIC (ZIC-HILIC) [81,86]. Unlike specific recognition between lectin and carbohydrate epitopes, hydrazide capture method is able to catch all types of glycosylated protein/peptide and is becoming a popular method for quantitative glycoproteomics.

Her et al. demonstrated a comparative strategy to simultaneously analyse glycosylated and non-glycosylated proteins in which dimethylated glycopeptides were separated from the non-glycosylated peptides using microcrystalline cellulose. The relative abundances of the dimethylated peptide pairs, both non-glycosylated and glycosylated, were directly analysed using a LTQ linear ion trap mass spectrometer [87]. Zhang et al. developed an online method for quantitative N-glycoproteome analysis [88] by combining a HILIC column for glycopeptide enrichment, a hydrophilic PNGase F immobilized enzymatic reactor for deglycosylation, and a C₁₈ trap column for dimethyl labelling. Zou et al. integrated glycopeptide enrichment using hydrazide capture and stable-isotope dimethyl labelling on beads in sequence to comparatively quantify N-linked glycoproteins in human serum of healthy volunteers (n = 16) and liver cancer patients (n = 11) [89]. Although hydrazide capture provides high specificity (more than 90%) for glycopeptide enrichment via aldehyde group exposed by oxidative cleavage of cis-vicinal diols on carbohydrates, oxidation of the N-terminal Ser/Thr residue of a peptide may concomitantly occur. This will lead to undesired covalent binding of glycopeptides to hydrazide beads via the oxidized Ser/Thr N-termini and hinder the release of N-terminal Ser/Thr-containing glycopeptides by PNGase F [90]. As depicted in figure 3, prior dimethyl labelling can block N-termini of the Ser/Thr residue to prevent oxidation (aldehyde formation), minimizing the sample loss due to undesired covalent bonding to the beads [91]. Xia et al. applied the revised protocol for comparative glycoproteomics study of triple-negative breast cancer (TNBC) tissues [92]. They identified 550 unique N-linked glycoproteins, among which 56 were upregulated and 16 were downregulated. Furthermore, several N-linked glycoproteins such as vascular endothelial growth factor receptor 1, insulin receptor and tissue factor pathway inhibitor were validated and regarded as potential biomarkers for TNBC diagnosis [92].

Figure 3.

Flowchart of stable-isotope dimethyl labelling combined with hydrazide capture for quantitative glycoproteomics.

4. Applications beyond quantitative analysis

Heck et al. have previously demonstrated that dimethyl labelling is useful for de novo sequencing by combining the labelling with Lys-N enzyme to create a majority of Lys-N residue [93]. Dimethyl labelling is also useful for blocking active sites, increasing peptide basicity (or charge state), and enhancing peptide sequencing based on a₁ fragment ion. As depicted in figure 4, many proteomics applications beyond quantification have been reported based on these characteristics. For example, dimethyl labelling can be used to detect protein N-termini [94] (figure 4a) generated after translocation/processing events which are crucial to understand the transformation of precursors to mature proteins. Chen et al. designed a flowchart to target protein N-termini via sequentially blocking active sites by dimethyl labelling. In their study, proteins were first dimethylated at N-termini and ϵ-amino group of Lys residue and then digested by trypsin to expose free N-termini of internal peptides. These peptides were captured by aldehyde-containing resin (POROS-AL). Peptides containing dimethylated or endogenously modified protein N-termini cannot be captured by the resin and the flow-through fractions are collected for LC-MS/MS analysis. By this manner, the N-terminal sequence or modifications (e.g. acetylation and formylation) were able to be unambiguously identified [95]. Based on the same concept, a so-called terminal amine isotopic labelling of substrates (TAILS) strategy was developed by Overall et al. using dendritic polyglycerol aldehyde polymers for internal peptide depletion [95,96]. Using TAILS strategy, 731 acetylated, 132 cyclized N-termini and 288 matrix metalloproteinase-2 cleavage sites in mouse fibroblast secretomes were identified and quantified for proteome-wide analysis of N terminome [95,96]. The authors further modified the TAILS strategy to analyse C terminome (C-TAILS) [97] in which dimethyl labelling was first used to block amino groups and then carboxylic acids including carboxyl groups of the C-termini and the side chain residue of aspartate or glutamate. The latter reaction was achieved by the combination of 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (EDC)/NHS and ethanolamine. After trypsin digestion, the newly generated N-terminal amino groups of internal peptides were protected using dimethyl labelling again and their free C-termini were captured by linear polyallylamine polymer via EDC-mediated condensation; the survived protein C-termini were identified and quantified using LC-MS/MS [97]. In addition to protein N-termini and C-termini, Guo et al. used isotope-coded dimethyl labelling and chemoprecipitation to identify and quantify low-abundant protein nitration [98]. In their study, the flowchart included the following steps: (i) dimethyl labelling was used to block N-terminal amine and Lys ϵ-amine of all tryptic peptides; (ii) Na₂S₂O₄ was added to reduce nitrotyrosines in corresponding aminotyrosines; (iii) the dimethylated aminotyrosine-containing peptides were captured using solid-phase active ester reagent (SPAER) on glass beads and other peptides were washed away; and (iv) the dimethylated aminotyrosine-containing peptides were released from SPAER beads under acidic condition and analysed by LC-MS/MS [98].

Figure 4.

Applications (beyond quantification) of stable-isotope dimethyl labelling for (a) identification of protein N-terminal sequence, (b) identification and quantification of N^α-acetylated peptide via SCX fractionation and (c) identification of disulfide linkages or ubiquitation tags via enhanced a₁ ions. (Online version in colour.)

The second useful property of dimethyl labelling is based on the increased basicity of a peptide by the labelling [99], which can alter chromatographic behaviour during ionic separation (figure 4b). Based on this concept, Chen et al. compared the retention behaviour between dimethylated, non-modified and acetylated peptides (bearing the same peptide sequence) during SCX separation. They found that their retention order (acetylated peptide < non-modified peptide < dimethylated peptide) closely correlates to their basicities. Therefore, using dimethyl labelling and simple SCX SPE for sample preparation, the number of acetylated peptides identified from HepG2 cells was four times greater than that of direct analysis (without dimethyl labelling and SCX) and two times greater than that using merely SCX (without dimethyl labelling). Chen et al. further applied this approach (dimethyl labelling and SCX) to identify and quantify N^α-acetylated proteins in HepG2 cells with and without tert-butyl hydroperoxide (t-BHP) treatment [100]. In their study, 576 non-redundant N^α-acetylated peptides were identified from 50 µg of cytosolic proteins extracted from HepG2 cells. Notably, the whole process was completed in 24 h and N^α-acetylated peptides of some protein isoforms, such as β-actin/γ-actin, ERK1/ERK2, α-centractin/β-centractin and ADP/ATP translocase 2 and 3, were simultaneously detected from the SCX flow-through fraction [100].

The third useful property of dimethyl labelling is the enhanced a₁ ion by CID fragmentation [36] as depicted in figure 2b. Hsu et al. found that a₁ ion enhancement not only occurred to dimethylated peptides but also dimethylated proteins [101]. Based on multiple a₁ fragments observed in a single MS/MS spectrum, Huang et al. developed a novel method to identify disulfide linkages [102]. As depicted in figure 4c (top), multiple a₁ ions of a disulfide-linked precursor ion were detected by CID and they can be used to indicate the N-terminal residue of individual peptides that are linked by disulfide bonds. This approach was demonstrated using recombinant human pancreatitis-associated protein which contained three disulfide linkages formed by six cysteines. Once two or more a₁ ions were matched, each peptide sequence was assigned based on matching the remaining b/y ion series in the MS/MS spectra. Home-made software named rapid assignment of disulfide linkage via a₁ ion recognition (RADAR) was developed based on this algorithm [102]. Huang et al. applied RADAR to identify disulfide linkages of recombinant monoclonal antibody bevacizumab (Avastin) [103] and crude snake venom [104]. Such multiple a₁ ion recognition approach was also applied to identify ubiquitinated peptides because ubiquitinated Lys normally contains a short diglycine (GG) tag. As shown in figure 4c (bottom), the main peptide skeleton and the diglycine side chain of an ubiquitinated peptide can be readily recognized by their individual a₁ ion and the side chain's b₂′ ion. As a proof of concept, Griffiths et al. demonstrated this approach by identifying a ubiquitinated protein, UbK48, spiked in a six-protein mixture [105]. Wu et al. reported that the a₁ signal was suppressed for precursor ions with phosphorylated Ser/Thr N-termini (N-p*Ser/Thr) but no suppression was observed for the precursor ion with phosphorylated Tyr N-terminus [36]. Although somehow negative, this provides a useful hint for mapping phosphorylation sites at the N-termini [106]. Despite diverse usefulness, a₁ ion is not detectable for ion trap instruments due to the ‘1/3 rule’, which also occurs for other tandem mass tags like iTRAQ. This limitation, however, can be overcome with the use of an Orbitrap instrument equipped with a higher-energy collisional dissociation chamber outside the trap.

5. Concluding remarks

Stable-isotope dimethyl labelling based on reductive dimethylation has been well accepted and widely used for quantitative proteomics analysis (460 citations up to now). It is a useful method for the ‘poor’. In addition, its usefulness is not limited to quantification; dimethyl labelling has been used to assist developing versatile enrichment or separation protocols, de novo sequencing, and characterization of protein PTMs based on MS analysis. With no doubt, we anticipate more quantitative applications and deeper explorations of its usefulness in MS-based analysis to be demonstrated in the future.

Competing interests

We declare we have no competing interests.

Funding

This work was financially supported by Taiwan MOST grants (MOST 104-2113-M-020–001) to J.-L.H.