Diabetic microangiopathy has become a heavy social burden worldwide, but at present it is still difficult to predict and diagnose this ailment at an early stage. Various proteomics approaches have been applied to the pathophysiological study of diabetic microangiopathy. Conventional proteomics methods, including gel-based methods, exhibit limited sensitivity and robustness and have typically been used in high- or middle-abundance biomarker discovery. Clinical samples from patients with diabetic microangiopathy, such as biopsy samples, are minute in size. Therefore sample preparation, quantitative labelling and mass spectrometry technologies need to be optimized for low-abundance protein detection, multiple-sample processing and precision quantitation. In this review, we briefly introduce the recent technological developments in proteomics methods and summarize current proteomics-based, translational research on diabetic microangiopathy. Recent technological developments in proteomics tools may shed new light on the pathogenesis of diabetic microangiopathy and biomarkers and therapeutic targets related to this condition.
isobaric tagging for relative and absolute quantitation
linear ion trap
non-proliferative diabetic retinopathy
proliferative diabetic retinopathy
surface-enhanced laser desorption/ionization
stable isotope labelling by amino acids in cell culture
Proteomic analyses have been used for the integrative profiling of proteins expressed in cells, tissues and organs. These studies have included protein identification, quantification, intracellular localization and the examination of protein interactions [1, 2]. Translational medicine is defined by the National Institutes of Health as ‘the process of applying ideas, insights, and discoveries generated through basic scientific inquiry to the treatment or prevention of human disease’ . Proteomics research has provided promising and powerful approaches for the systematic study of proteins that are differentially expressed between healthy individuals and patients or between different disease stages for disease prognosis, diagnosis and therapy . Recently, proteomics methods have also been successfully applied in translational studies on cancer , Alzheimer's disease  and other maladies . Recent technological developments in proteomic sample preparation, labelling and mass spectrometric analysis will be described in this review, as will the contribution of these new proteomics tools to translational research on diabetic microangiopathy.
Diabetes mellitus (DM) is a chronic metabolic disease caused by β-cell dysfunction and/or insulin resistance. The prevalence of diabetes among adults (aged 20–79 years) worldwide was estimated to be 6.4% in 2010 and could potentially rise to 7.7% in 2030 . Indeed, the total number of people with diabetes has been projected to increase from 285 million in 2010 to 439 million in 2030. In patients with DM, chronic hyperglycaemia can damage capillaries and lead to diabetic microangiopathy, such as retinopathy and nephropathy . The prevalence of diabetic complications has been reported by the American Diabetes Association to be 98% for patients diagnosed with diabetes for 10 years or more. Compared with type 2 DM patients without complications, the direct medical costs for type 2 DM patients with diabetic microangiopathy are tripled. Thus, predicting and diagnosing diabetic microangiopathy has become a global problem. Notably, proteomics analyses could advance the study of the pathogenesis of diabetic microangiopathy, as well as the search for diagnostic biomarkers and therapeutic targets for addressing this condition.
Recent technological developments in proteomics
The greatest challenge for the application of proteomics methods to translational research has been the quantitative analysis of the entire proteome to uncover biomarker candidates for use in clinical diagnosis and therapy. Proteomic approaches for protein sample processing, identification and quantification have been developed to meet the requirements of translational research (Fig. 1).
Sample preparation techniques
Proteomics studies rely on both gel-based techniques, such as two-dimensional gel electrophoresis, and gel-free approaches, such as multidimensional protein identification technology. However, current gel-free technologies exhibit serious drawbacks in terms of their time consumption, automation, sensitivity and robustness. Furthermore, hydrophobic proteins (such as membrane proteins), post-translationally modified proteins, low-abundance proteins and proteins present in minute amounts still provide major challenges to current technologies. Because of these challenges, there is a great demand for novel technological developments to allow the systematic integration of cell/tissue protein extraction, pre-concentration and protein fractionation. The development of new technologies, such as the filter-aided sample preparation (FASP) method and proteomic reactors , may pave the way to achieving this goal. Wiśniewski et al. developed the FASP method for processing proteomic samples . The method uses a 10K filter to remove detergents and small molecules, such as dithiothreitol (DTT) and iodoacetamide (IAA), to exchange buffers and to filter out undigested proteins as well as trypsin. FASP is combined with the StageTip SAX fractionation method for tissue-membrane protein analysis , and FASP can also be employed for post-translational modification analysis of formalin-fixed and paraffin-embedded tissues as well as normal tissues . Ethier et al. have developed a single microfluidic device for minute sample processing referred to as a proteomic reactor using a small bed of packed, strong cation-exchange material for sample processing . Protein samples are acidified and loaded into the proteomic reactor at a low pH (pH < 3). The proteins are positively charged and absorb onto the reactor material because the pI of the loaded proteins is higher than the pH in the reactor, and the non-ionic detergents are then readily washed away. The proteins are subsequently reduced and alkylated by adding the reagents DTT and IAA to the reactors, after which they are digested by increasing the pH to 8 which activates the trypsin loaded into the reactor with the protein sample. The obtained peptides are finally eluted using a buffer compatible with HPLC-ESI-MS/MS. Zhou et al. further developed a simplified and user-friendly reactor compatible with bench-top centrifuges , which can also be used for clinical sample processing. The development of new sample preparation techniques has the potential to provide powerful and promising tools for translational research.
Sample labelling and label-free for quantitative proteomics
Proteomics analyses were initially focused primarily on protein expression profiling. Scientists expected to identify as many proteins in a gel spot as possible using two-dimensional gel electrophoresis. Upon further development of the field, proteomics research began to focus on the identification of proteins with significant biological activity, rather than simple protein profiling. Such protein molecules are generally referred to as biomarkers. Quantitative proteomics has therefore arisen as a good way to study the differential expression of proteins for biomarker discovery.
According to the difference-of-quantitation method, quantitative proteomics can be divided into relative and absolute quantitation. Relative quantitative methods (such as SILAC and iTRAQ) are used to compare protein and peptide differences between samples, while absolute quantitation of target proteins is achieved by comparison with synthetic peptides containing isotopic labels serving as internal standards.
Stable isotope labelling by amino acids in cell culture (SILAC) is a popular alternative means of conducting relative quantitative proteomics analyses. Stable-isotope-labelled amino acids, typically arginine and lysine, are introduced during cell growth and can achieve close to 100% labelling efficiency. Through comparison of the relative abundance of the labelled and unlabelled proteins from two or three samples based on MS, quantitative data can be easily obtained [15, 16] and biomarkers, such as for cancer , identified. Recently, a so-called ‘spike-in’ or ‘super’ SILAC method has been developed to measure tissue proteomes. Due to the complexity of tissues, using single cell lines in traditional SILAC has not been found to be adequate for tissue proteomics analyses. This has been particularly true in cancer research, where it has not been possible to model the diversity of tumour samples with single cell lines. To improve SILAC, Geiger et al. developed a new method for using a mixture of cell lines as an internal ‘spike-in’ standard . They chose cell lines from different sources, different stages and with a variety of molecular markers. The mixture of these cell lines was referred to as a super SILAC mixture. As a result of using the super SILAC mixture, the accuracy and robustness of their analyses were improved compared with the results using a single cell line internal standard.
Isobaric tagging for relative and absolute quantitation (iTRAQ) is a post-biosynthetic labelling strategy that has been applied to many sample types, such as cell cultures, biofluids and tissues . The iTRAQ reagent uses tags of different masses to label proteins and peptides. The labelled samples are then pooled and analysed via MS to identify peptides and hence proteins. iTRAQ is also a popular way to quantitate changes in the proteome .
The primary absolute quantitation strategy employs a synthetic, stable, isotope-labelled peptide that is added to samples as an internal standard. The selected-reaction-monitoring function of MS has been applied to detect and quantitate proteins based on the intensities of the related peptides and internal standards . Absolute quantitation methods have permitted the detection of subtle changes in the proteome  and have received increasing attention in proteomics research .
The initial technology for MS-based quantitative proteomics involved different isolabelling methods with stable isotopes. Although these methods are still in use, they have some limitations. One is expense. The other is that some isolabelling methods are applicable to certain biological sample types. Due to shortcomings of isolabelling methods, labelling-free quantitative proteomics techniques emerged and became an attractive technique in biomarker discovery.
In label-free quantitative proteomics study, no label is introduced into either of the samples. The same amount of each sample is trypsinized individually and analysed in separated LC-MS experiments. MS1 and MS2 spectra for each sample are acquired and the individual peptide properties are compared between the individual measurements. Spectral counting and intensity-based measurements are two major methods for label-free MS-based quantitative proteomics . Moreover, the method named the total protein approach was developed. Then the protein copy numbers per cell could be determined from the MS data on the basis of individual intensities compared with the total MS signal of the measured proteome . The generic nature of label-free quantification is not restricted to any model system and can also be used for clinical samples such as tissue and body fluids. Mann's laboratory performed proteomics analysis of patient primary carcinoma, nodal metastases and paired normal tissue through the label-free quantitative technique. A total of 8000 proteins were identified and 1808 proteins changed significantly. Their research provided novel insights into proteome remodelling during cancer development .
Advanced mass spectrometry technology
Mass spectrometry (MS) has become an irreplaceable tool for the identification and quantitation of peptides and proteins in proteomics analyses because of its speed, sensitivity and accuracy. The recent rapid development of biological mass spectrometry has primarily been the result of two soft ionization techniques: matrix-assisted laser desorption ionization (MALDI)  and electrospray ionization (ESI) . A MALDI source is often combined with time of flight (TOF) or TOF/TOF MS. Due to the compatibility of the raw materials and surfaces involved, MALDI MS has been employed directly for tissue imaging. Using MALDI, crude clinical samples have been fractionated via simple one-step affinity chromatography on activated solid phases prior to MS analysis . The MALDI MS technique has been applied in the study of cancer in both fresh  and formalin-fixed, paraffin-embedded tissue . Compared with MALDI MS, ESI MS coupled with LC (LC-MS) has been the preferred method for the analysis of complex samples. There are also other varieties of high resolution MS combined with ESI; the major ESI MS methods used for proteomics studies are QTOF and orbitrap . With the development of proteomics technology, scientists have been pursuing MS with both a higher resolution and speed. Orbitrap Elite and Q Exactive, which exhibit resolutions of 240 000  and 140 000 , respectively, are the newest members of the orbitrap family. In addition to its high resolution, Orbitrap Elite provides several fragmentation methods, such as collision induced dissociation, high energy collision dissociation and electron transfer dissociation. These new methods should greatly increase the information derived from MS/MS, especially for modified peptides . Ahlf et al. applied Orbitrap Elite to a top-down proteomics analysis and developed improved separation at a higher resolution. These authors identified 690 unique proteins from H1299 human lung cancer cells, and over 2000 proteoforms were identified from proteins with intact masses < 50 kDa . Q Exactive is a benchtop orbitrap with a resolution of 140 000. Due to its special features, such as its ability to perform multiplex, narrow-mass-range analyses and provide different fragmentation spectra, the Q Exactive system has expanded the strategies available for proteomics research . By combining the Q Exactive system with nano UHPLC, Nagaraj et al. developed an analytical method for examining the yeast proteome. They identified more than 4000 proteins, representing nearly all the proteins expressed in a yeast cell . As mentioned previously, higher scan speeds, in addition to higher resolutions, have been an important development in high resolution MS. The scan speeds of traditional QTOF and orbitrap are below 50 Hz. The most recent Triple TOF has attained a scan speed of 100 Hz, which is considered adequate for carrying out peptide and protein identifications . A new strategy to analyse the proteome referred to as SWATH has been developed using Triple TOF due to its high scan rate. SWATH is a novel MS/MS scanning process in which data are acquired via repeated cycling in 25-Da precursor-isolation windows in Triple TOF, and fragment information for all ions is captured. The data obtained have demonstrated that SWATH is a promising method for qualitative and quantitative proteome probing in samples using a single MS injection .
The use of proteomics approaches in translational research on diabetic microangiopathy
Diabetic retinopathy (DR) and diabetic nephropathy (DN) are regarded as the most common diabetic microangiopathies. There is currently no way for clinical workers to diagnose and treat DR and DN at an early stage. However, with the development of proteomics tools, researchers are hopeful that application of these methods can solve these problems (Table 1).
Table 1. Summary of the proteomics approaches in translational research on diabetic microangiopathy
DR has become a leading cause of visual impairment to differing extents. This condition begins as non-proliferative diabetic retinopathy (NPDR) and can progress to proliferative diabetic retinopathy (PDR) if the blood glucose level cannot be controlled within the normal range. Although photocoagulation and vitrectomy have been applied to treat DR, these interventions are often not effective in halting the loss of visual acuity . The best way to prevent blindness in diabetic patients is early diagnosis. Therefore, proteomics approaches have been widely used in DR research to find early diagnostic biomarkers (Fig. 2).
The analysis of retina proteins is a direct way of identifying the occurrence and development of DR. Takada et al. collected membrane samples from PDR patients and patients with idiopathic epiretinal membranes (ERMs) undergoing vitrectomy. They included three cases in each LC/MS/MS analysis, and a total of 225 and 154 proteins were found in the PDR and ERM groups, respectively. Of these proteins, 123 were common to both groups, and 102 and 31 proteins were unique to the PDR and ERM groups, respectively. Ten of the identified proteins showing highly normalized spectral-abundance factors (NSAFs) were selected for comparison between the two groups. Of these proteins, only periostin and thrombospondin-1 were unique to patients with PDR. Ultimately, these researchers chose periostin as a candidate marker because it exhibited the highest NSAF expression value. This result was confirmed via RT-PCR . However, the authors could not demonstrate conclusively whether the highly expressed periostin caused PDR or was overexpressed due to the tissue damage caused by PDR. As a result, it is still unknown whether periostin plays a role in the aetiology of PDR. Further research on the retina proteome has been performed using animal models. Li et al. treated diabetic rats with grape-seed proanthocyanidin extract (GSPE), and the retinas from three groups were analysed using MALDI-TOF-TOF/MS following two-dimensional electrophoresis (2-DE). A total of 18 proteins were found to be differentially expressed to a significant degree in diabetic rats compared with normal rats. However, αB-crystallin, βA4-crystallin, αA-crystallin, ubiquitin carboxy-terminal hydrolase L1, G protein β1, Igγ-2A and pyruvate kinase returned to normal levels following treatment with GSPE. The authors hypothesized that the downregulated proteins might play key roles in the development of and recovery from DR . In another study, Fort et al. used quantitative two-dimensional difference gel electrophoresis (2D-DIGE), iTRAQ analysis and MALDI-TOF/TOF MS to analyse samples from mice that had been separated into five different groups. Three of the five groups were treated with insulin in different ways: one was injected with insulin both locally in the eyes and systemically, and the other two groups were injected with insulin either locally in the eyes or systemically. Members of both families of crystallin, the α-crystallins (αA and αB) and βγ-crystallins (βA2, βA3, βB4, γB and γD), were found to be overexpressed, but the levels of each crystallin changed during the course of diabetes treatment. Crystallins have been implicated in the regulation of cell death in lens epithelial cells and central nervous system cells . In this study, both systemic and local insulin treatments reversed the upregulation of α- and γ-crystallin. Therefore the authors suggested that insulin treatment could reduce the level of cell apoptosis observed in DR and noted the possible role that crystallins may play in DR.
In addition to retinas, ocular fluids have been widely used in proteomics studies on DR as they can be obtained from patients more easily than retinal tissue. Moreover, protein changes occurring in this structure may be more important than systemic alterations for the pathogenesis of DR [42, 43].
Vitreous humour is the acellular, highly hydrated extracellular matrix which is surrounded by and attached to the retina and behind the lens of the eye . Because of its close contact with the retina, the proteome and the biochemical properties of the vitreous humour can reflect the physiological and pathological conditions of the retina. García-Ramírez et al. used fluorescence-based DIGE and MALDI-MS to separate and identify vitreous humour proteins from PDR patients and non-diabetic patients with macular holes. They found that zinc-α2-glycoprotein, apolipoprotein (APO) A1, APOH, fibrinogen A and complement factors C3, C4b, C9 and factor B were highly expressed in PDR patients compared with non-diabetic patients. On the other hand, pigment epithelial-derived factor, interstitial retinol-binding protein and inter-α-trypsin inhibitor heavy chain were under-expressed. A subset of these results was confirmed via western blot analysis . Other researchers also observed that APOA1 and APOH expression levels were elevated in PDR patients. As APOA1 and APOH are proteins involved in nutrition transport, an increase in their expression may lead to nutritional disorders and induce disease .
In other studies, two or more mass-spectrometric and chromatographic processes have been combined to identify additional proteins. For example, Kim et al. used immunoaffinity subtraction, (IS)/2-DE/MALDI-MS, nano-LC-MALDI-MS/MS and nano-LC-ESI-MS/MS to identify vitreous humour proteins from PDR patients and non-diabetic controls. They identified 531 proteins in the vitreous proteome, of which 415 were found in PDR patients and 346 were present in controls. Some proteins from this group are known to be related to the pathogenesis of DR, such as the members of the insulin-like growth factor binding protein family, carbonic anhydrase, osteopontin, angiotensinogen and clusterin. These researchers also identified 240 proteins that had not previously been associated with PDR . In a later study, Gao et al. divided patients into non-diabetic, no-DR and PDR groups. Soluble proteins from vitreous samples of these patients were separated via 12% SDS/PAGE, and the resulting gels were divided into slices 1–2 mm in width. They then digested each gel slice and analysed the obtained peptides using nano-LC/MS/MS with a linear ion trap (LTQ) mass spectrometer. In this study, 252 proteins were identified in human vitreous humour, including 37 proteins that were differentially expressed between the three different groups. These proteins were associated with many metabolic mechanisms, including the kallikrein-kinin system, the coagulation cascade, catalytic activities, enzyme-regulatory activities and cell lysis . These new findings from the vitreous proteome may allow us to explore pathogenesis, therapeutic targets and new diagnostic biomarkers for PDR.
Aqueous humour and tears
To obtain additional information about DR, other ocular fluids have also been studied. Although some ocular fluids such as the aqueous humour (AH) and tears do not directly contact the retina, they can still reflect the state of health of the blood vessels in the eye . The AH is located in the space between the lens and the cornea and is divided into anterior and posterior chambers by the iris. Chiang et al. collected AH from 22 DM subjects, 11 with DR (DR group) and 11 without DR (control group). They analysed the proteins desalted from these AH samples via 2-DE. Eleven protein spots showed a significant difference in the 2-DE gels based on the results of both t tests (P <0.05) and sequential goodness of fit tests (P <0.05) between the control and DR groups. Then these researchers analysed samples that were manually excised and subjected to in-gel digestion through MALDI-TOF MS and identified proteins based on peptide mass fingerprinting. They found that the levels of APOA1, serotransferrin, keratin type I cytoskeletal 9, keratin type I cytoskeletal 10, growth factor receptor-bound protein 10, brain-specific angiogenesis inhibitor 1-associated protein 2, cystathionine β-synthase and retrotransposon gag domain-containing protein 1 expression increased, while those of matrix metalloproteinase 13, podocan and selenoprotein P were reduced in the DR group .
As early as 2000, Herber et al. used 2-DE to analyse tears from diabetic patients and healthy volunteers. They found a group of proteins that were expressed at significantly different levels between the two groups. Unfortunately, they did not determine the identity of these proteins . Recently Kim et al. collected tear samples from subjects with DM, no-DMR (diabetes mellitus without the retinopathy) and NPDR, as well as from healthy subjects. A total of 20 differentially expressed spots were detected using 2-DE. Following identification of the differentially expressed proteins through ESI-Q-TOF, the results were confirmed by western blotting. Remarkably, the levels of lipocalin 1 (LCN-1) and heat shock protein 27 in the tears decreased in both the no-DMR and NPDR groups, whilst the levels of these proteins increased in the blood in previous studies . Detection of the opposite expression patterns for these proteins may be useful in the diagnosis of PDR.
Additionally, with the development of these proteomics techniques, more proteins have been identified than ever before. Csősz et al. collected tear samples and analysed them using LC-MS/MS. The samples were labelled with an iTRAQ fourplex reagent and digested in solution. A total of 53 proteins were identified, 50 of which were identified with more than 99% confidence. Changes in the expression levels of LCN-1, lactotransferrin, lacritin, lysozyme C, lipophilin A and the Ig lambda chain C region have not been reported in other diseases. These changes appear to be specific for DR, suggesting that these proteins may be useful biomarkers for this ailment .
The data currently indicate that DN is the most important single factor in adult end-stage renal disease, and half of new haemodialysis patients are attributed to DN. Microalbuminuria is a non-invasive marker that can be used for the diagnosis of DN. There is global consensus that an annual test should be performed to assess urine albumin excretion in type 1 diabetic patients with a diabetes duration of ≥ 5 years and in all type 2 diabetic patients starting at diagnosis [9, 54]. However, many patients are normoalbuminuric but show advanced changes in renal histopathology. Once proteinuria occurs, glomerular function will irreversibly and progressively decline . Therefore, we need to find better methods for the early diagnosis and treatment of DN. In pursuit of this goal, proteomics approaches have been widely applied to investigate DN in the past decade (Fig. 3).
Urine is the most commonly used and non-invasive type of clinical specimen. Sharma et al. analysed the urine proteome via 2-D DIGE to detect proteins that were differently expressed between three patients with DN and five non-diabetic control subjects. They found 99 spots that were significantly different between the two groups: 63 were upregulated, and the others were downregulated. Alpha 1 antitrypsin (AAT) was identified as the most highly abundant upregulated spot using surface-enhanced laser desorption/ionization (SELDI) TOF-MS. The increase in AAT in the urine reflected the local production of AAT in the kidney and may have been involved in inflammation and matrix accumulation . Jiang et al. divided patients into four groups: patients with normoalbuminuria (DM group), microalbuminuria (DN1 group), macroalbuminuria (DN2 group) and a control group (n =8 for each group), and used SDS/PAGE and MALDI-TOF-MS to analyse and identify proteins. It was found that urinary E-cadherin was upregulated 1.3-fold, 5.2-fold and 8.5-fold in the DM, DN1 and DN2 groups, respectively, compared with the control group. The results were further verified via western blotting and ELISA. Urinary-soluble E-cadherin showed a high sensitivity and specificity for DN diagnosis. The sensitivity and specificity were calculated to be 78.8% (95% CI, 74%–83%) and 80% (95% CI, 65%–91%), respectively . A study examining a large number of samples was designed by Dihazi et al. in which SELDI-TOF-MS and strong anion exchange (SAX2) protein arrays were employed to compare urine proteins between four groups. A total of 162 subjects were divided into the following groups: DM patients without nephropathy and microalbuminuria (DM-WNP, n =45); DM patients with macroalbuminuria or microalbuminuria (DM-NP, n =38); patients with proteinuria due to non-diabetic renal disease (n =34); and healthy controls (n =45). The authors found a protein with an m/z value of 6188 (P <0.0000004) that showed significantly reduced expression and a protein with an m/z value of 14 766 (P <0.00008) that was selectively excreted in the urine of DM-NP patients . The m/z 6188 and m/z 14 766 proteins were identified as a processed form of ubiquitin and the ubiquitin ribosomal fusion protein UbA52, respectively. Urine UbA52 may serve as a diagnostic marker for DM-NP patients, and the processed form of ubiquitin may be involved in the pathological mechanism of DN. Papale et al. analysed urine from 20 healthy subjects and 20 normoalbuminuric and 18 microalbuminuric diabetic patients. They also analysed urine from 132 patients with biopsy-proven nephropathy [65 DN patients, 10 diabetic patients with non-diabetic chronic kidney disease (CKD) and 57 non-diabetic patients with CKD]. The authors identified two proteins, with m/z values of 11 700 and 8589, that were the most prominent predictors in the classification tree of DN based on SELDI-TOF-MS analysis . As protease inhibitors were used in this study, the results were different from those of Dihazi et al. The excretion of β2 microglobulin and ubiquitin was observed to be increased in the urine of patients with DN compared with levels in diabetic patients without DN or with non-diabetic CKD. These results were confirmed using either immunoprecipitation or ELISA .
Capillary electrophoresis MS (CE-MS) is considered to be a suitable method for the analysis of urine because low molecular weight proteins and peptides can be analysed directly without additional manipulation [60, 61]. Rossing et al. used CE-EI-MS to analyse urine samples from patients in a blinded study. A total of 102 urinary proteins were found to differ significantly between patients with normoalbuminuria and nephropathy. These authors then created a support vector, machine-based model including 65 of these defined proteins. This model correctly identified DN with 97% sensitivity and specificity and distinguished DN from other chronic renal diseases with 81% sensitivity and 91% specificity . This finding could have great clinical value for patients with type 2 diabetes and nephropathy. Although microalbuminuria may predict the occurrence of DN in DM patients, it cannot be equated directly with DN in diabetic patients because many other nephropathies can also cause microalbuminuria. Rossing et al. employed a biomarker model to test whether the examination of a urinary proteome using CE-MS could accurately identify subjects with DN. The model for DN showed 93.8% sensitivity and 91.4% specificity, with an area under the curve of 0.948 (95% CI, 0.898–0.978). Sixty of the proteins in the model were confirmed to show differential expression between diabetic patients with DN and diabetic controls . Zürbig et al. applied CE-MS to profile the low molecular weight proteome in urine to identify proteins and peptide markers of DN. They examined urine samples from a longitudinal cohort of type 1 and type 2 diabetic patients (n =35) using CKD as a biomarker classifier. They found that a reduction of collagen fragments preceded an increase of urinary albumin excretion. The risk for the development of DN could be evaluated with urinary proteomics methods at an early stage . These results demonstrated the potential for accurate urinary proteome analysis for DN detection and thus represented a breakthrough for the diagnosis and prediction of DN.
Furthermore, the fact that urine can be easily recovered and stored allowed an interesting case–control study to be conducted using proteomic profiling of urine to predict the development of DN 10 years prior to its onset. The case subjects included type 2 diabetes patients who were normoalbuminuric and progressed to DN within 10 years (n =31) and type 2 diabetic patients who remained normoalbuminuric after 10 years (n =31). The baseline characteristics were well matched, with the exception of HbA1c. Urine proteins were analysed in this study via SELDI-TOF-MS, and the researchers observed a 12-peak urine signature that distinguished patients developing DN from unafflicted patients. More importantly, the 12-peak signature was dependently associated with DN and may predict DN earlier than microalbuminuria .
Plasma contains large amounts of protein, and many diseases can be detected based on changes in proteins in the plasma prior to other clinical manifestations. However, the plasma proteome is more complex than the urine proteins, so it is more difficult to find useful markers in plasma. Overgaard et al. studied the plasma from 123 Caucasian, type 1 diabetic patients who were divided into three groups: normoalbuminuric (n =42), microalbuminuric (n =40) and macroalbuminuric (n =41). Either anion exchange or hexapeptide fractionation was used to reduce plasma complexity prior to SELDI-TOF-MS analysis. Following data analysis, 16 peaks were selected for further investigation. Four of the 16 peaks were identified as transthyretin, apolipoprotein C1 (APOC1), apolipoprotein A1 (APOA1) and cystatin C. The identity of these proteins was confirmed through either immunoprecipitation and sequencing or western blotting . These proteins were differentially expressed in patients with DN; however, the data have not yet been corroborated in a longitudinal cohort study [67, 68]. Another study employed iTRAQ labelling and LC-MS/MS to further analyse the same samples. In this study, it was found that APOA2, B, C3, D and E were present in one of the largest protein clusters and may play roles in cardiovascular disease, genetic disorders and metabolic diseases. The researchers then examined APOA1, A2, C3, E, H and J in a series of diabetic patients (n =30), divided by the change in glomerular filtration rate (ΔGFR) over the course of 4 years. However, no correlation was found between ΔGFR and the concentrations of APO lipoproteins [66, 69]. This negative result may be related to the small sample size, and tests in longitudinal sample sets are needed to confirm the results.
While kidney tissues can be collected incidentally during surgery, patients must assume more risk and pain to undergo a pathological kidney examination. Therefore animal models have been used to investigate early pathological changes and potential therapeutic targets. Tilton et al. used 2-DE combined with MALDI-TOF, MALDI-TOF/TOF and LC-MS/MS to analyse rat renal cortices. They classified differentially expressed proteins based on their modulated physiological functions and produced systemic information about DN mechanisms . GSPE was used to treat rats with DN. Then samples from kidneys were analysed via 2-DE, followed by MALDI-TOF-TOF MS and LTQ-ESI-MS/MS. A total of nine proteins, confirmed to be involved in oxidative stress, glycosylation damage and amino acid metabolism in the kidneys of diabetic rats, were observed to return to normal levels as a result of GSPE therapy [40, 71]. This study has improved our understanding of the mechanisms underlying DN and has assisted in the discovery of new therapeutic targets.
Proteomics in diabetic microangiopathy clinical practice
Despite the progress described above, there has yet to be any great breakthrough using proteomics techniques to examine the pathogenesis and therapeutic targets of diabetic microangiopathy, and the great concern in this context is the methods used by clinical researchers. Label-free quantitation represents a major recent advance in quantitative proteomics methods and can be applied to any proteomic sample, without the need for labelling with iTRAQ. This method may be of great help in analysing clinical samples in the future. However, only half of the proteins in a typical sample can be identified with the method, and low-abundance proteins are particularly difficult to identify. Furthermore, signal transduction pathways are difficult to study as they contain many low-abundance proteins. This has prompted researchers to find new ways to identify proteins in signalling pathways. And biomarkers used in a clinical setting must be convenient and show high sensitivity and specificity. At present, there are no biomarkers that have been identified with proteomics methods that can be applied to diagnose diabetic microangiopathy. Notably, the high cost of proteomics methods has also been a barrier to their widespread use in clinical settings.
Although proteomics methods are increasingly used in clinical diagnosis and therapy [5-7], no significant breakthroughs have been reported in diabetic microangiopathy. While some researchers have successfully applied proteomics methods to the study of diabetic microangiopathy [66, 69], it will take time to verify whether these methods can be widely used in a clinical setting. As MS resolution continues to increase and the development of sample separation techniques and quantitative proteomics methods proceeds, increasing numbers of post-translationally modified and low-abundance proteins will be identified in the future. Theoretically, we should be able to identify all of the proteins in a sample with proper optimization of our methodology. Such advances would represent a tremendous leap forward in the study of the occurrence, development and treatment of diabetes-associated diseases.
This study was supported by grants from the Natural Science Foundation of Shanghai (No. 10ZR1424100), the ‘One Hundred Talent Program’ and the instrumental development project (No. YG2012106) of the Chinese Academy of Sciences. We would like to thank Kun Liu from the Department of Ophthalmology, Shanghai Jiao Tong University Affiliated First People's Hospital, and Zhigang Zhang and Wenjiao Zeng from the Department of Pathology, School of Basic Medical Sciences, Fudan University, for providing the figures. We are also grateful to Guocheng Shi and Lingfeng Chen from the Shanghai Jiao Tong University School of Medicine for their critical review of the manuscript. The authors would like to acknowledge editorial help from Dr Ettore Appella and Dr Lisa Jenkins.