• Open Access

Quantitative mass spectrometry-based proteomics in angiogenesis


  • Colour Online: See the article online to view Figs. 2 and 3 in colour.

Correspondence: Dr. Sara Zanivan, Vascular Proteomics Lab, The Beatson Institute for Cancer Research, Garscube Estate, Switchback Road, Glasgow G61 1BD, UK

E-mail: s.zanivan@beatson.gla.ac.uk

Fax: +44-141-942-6521


The process of new blood vessel formation from pre-existing ones is called angiogenesis. Beyond playing a critical role in the physiological development of the vascular system, angiogenesis is a well-recognised hallmark of cancer. Unbiased system-wide approaches are required to complement the current knowledge, and intimately understand the molecular mechanisms regulating this process in physiological and pathological conditions. In this review we describe the cellular and molecular dynamics regulating the physiological growth of vessels and their deregulation in cancer, survey in vitro and in vivo models currently exploited to investigate various aspects of angiogenesis and describe state-of-the-art and most widespread methods and technologies in MS shotgun proteomics. Finally, we focus on current applications of MS to better understand endothelial cell behaviour and propose how modern proteomics can impact on angiogenesis research.


basement membrane


delta-like protein


endothelial cell


extracellular matrix


human umbilical vein endothelial cells


linear trap quadrupole


neurogenic locus notch homolog protein




vascular endothelial growth factor

1 Introduction

Angiogenesis is the growth of new vessels from pre-existing ones and is a fundamental process for the development of the vascular system during embryogenesis and to promote and sustain diseases such as cancer [1]. In the past 10 years the number of publications in the angiogenesis field has consistently grown reflecting the tremendous increase in understanding of cellular and molecular mechanisms that regulate vessel growth. Similarly, in the past decade, MS technology has improved enormously, resulting in the development of MS-based proteomic approaches and their application to address, in an unbiased manner, a wide variety of biological questions [2]. However, in spite of the progress made in both fields, proteomic methods are still poorly exploited to study angiogenesis (Fig. 1).

Figure 1.

Angiogenesis and proteomics. The development and achievements in proteomics and angiogenesis in the last decade are simplistically represented by the number of publications, according to PubMed, per year (y-axis on the left) containing the term proteomic* and angiogenesis, respectively. Proteomic methods applied to the field of angiogenesis are represented by the number of publications per year (y-axis on the right) containing both terms. Up to 2000 indicates the number of publications from 1900 to 2000.

In this review we define the complexity of the angiogenic process, survey in vitro and in vivo models currently used to successfully investigate various aspects of angiogenesis and state-of-the-art strategies and methods used in MS shotgun proteomics. Finally, we summarise how MS has currently been applied to better understand endothelial cell (EC) behaviour and propose how modern proteomics can impact on angiogenesis research.

2 Sprouting angiogenesis in physiology and cancer

Blood vessels deliver oxygen and nutrients and remove waste products from every tissue of the body. Thus, the expansion of the vascular system by sprouting angiogenesis is crucial for the growth of organs during embryogenesis and, in the adult, to maintain the homeostasis of tissues, in wound healing, inflammation and endometrial growth during the menstrual cycle. Beyond having an important role in physiology, angiogenesis can fuel malignant diseases, such as cancer [3]. The role of vessel growth in cancer was first suggested in 1971 by Folkman, when he proposed that blood vessels nourish tumour cells, promote their growth and play a vital role in the progression of the tumour to malignancy [4]. Today, tumour angiogenesis is a well-recognised hallmark of cancer [5], and strategies to interfere with the angiogenic process have been implemented in anticancer therapies [6]. The observations made by Folkman have been followed by a boost in research in the field of angiogenesis to learn more about the molecular basis of vessel growth in the physiological and pathological environment. As the field progressed it became evident that angiogenesis is an incredibly complex system that involves a multitude of dynamically coordinated distinct cellular processes, where ECs are the most active players. ECs line the inner layer of the vessel wall and, as such, are in direct contact with the blood. At their abluminal surface they are surrounded by basement membrane (BM) and mural cells – pericytes and vascular smooth muscle cells [7]. In mature vessels ECs are quiescent, however, in the presence of pro-angiogenic stimuli, ECs switch to an angiogenic phenotype and initiate the formation of new vessels.

Sprouting angiogenesis (for other types of angiogenesis see [8]) is a highly dynamic process that consists of several sequential steps. These include (i) detachment of the pericytes, local degradation of the BM and dissociation of EC-cell junctions; (ii) sprouting and migration of endothelial tip cell towards angiogenic stimuli; (iii) cell alignment; (iv) proliferation of the stalk cells – behind the leading tip cell; (v) lumen formation; (vi) anastomosis; (vii) recruitment of pericytes and deposition of novel sub-endothelial BM (see Fig. 2A and, for a more detailed overview, see reviews [3, 9]). These individual steps are tightly controlled by many different factors. Major examples are vascular endothelial growth factor (VEGF), neurogenic locus notch homolog protein (NOTCH) and platelet-derived growth factor B (PDGFB) (other factors are reviewed in [1, 10, 11]). VEGF is the predominant regulator of angiogenesis and signals through VEGF receptors (VEGFRs), which are largely expressed in ECs. In the presence of a VEGF gradient, ECs commence the process of sprouting angiogenesis. ECs expressing high levels of VEGFR2 differentiate into tip cells, which generate thin membrane extensions, known as filopodia, to sense angiogenic cues, polarise and migrate towards the stimulus. To promote the elongation of the newly formed sprout, tip cells express high levels of delta-like protein 4 (DLL4), which signals to neighbouring ECs via its receptor NOTCH. NOTCH signalling down-regulates the expression of VEGFR2 and determines the acquisition of the stalk cell phenotype. Stalk cells begin to proliferate, elongate and eventually form the lumen of the growing endothelial tube whilst having reduced cell motility [12]. To promote the maturation of the nascent vessel, ECs secrete PDGFB to recruit PDGF receptor beta expressing pericytes [13]. Attracted by the growth factor, pericytes incorporate into the wall of the forming vessel and establish cell–cell interaction with the ECs. Finally, ECs and pericytes produce a new BM. This step, in combination with the tightening of EC-cell junctions, provides a structural support to the new vessel and restores the quiescent state of the ECs [14].

Figure 2.

Shotgun proteomics in angiogenesis. (A) Schematic overview of the major steps involved in the formation of a new mature blood vessel from preexisting one. Colour of the text reflects the specific cell type or basement membrane (BM) as follows: dark brown: BM; light brown: newly forming BM; yellow: tip cell; blue: stalk cell; light green: phalanx cell (quiescent EC); dark green: pericyte. The arrows indicate the direction of the cell movement. (B) Schematic overview of a MS shotgun proteomic approach, where models representing different aspects of the angiogenic process (A) are used in combination with SILAC for quantitative proteomic analysis. In a typical workflow, peptides from digested SILAC-containing protein mixture are resolved using LC and electrosprayed into LTQ-Orbitrap. Ions fly through the ion trap and are measured at high resolution in the Orbitrap (red arrows). The most intense ions are then fragmented in the HCD collision cell and acquired at high resolution in the Orbitrap (blue arrows). ESI, electrospray ionisation.

In order to generate a functional vascular network, the angiogenic process requires the tight regulation and correct sequential coordination of EC behaviour. For this reason, a balance of angiogenic stimuli is crucial in angiogenesis. As an example, in mouse, an imbalance of VEGF and NOTCH signalling via insufficient DLL4 results in an increase in EC sprouting, giving rise to dysfunctional vessels [15]. Conversely, in the absence of VEGF, mouse embryos die because of delayed endothelial differentiation and impaired angiogenesis [16]. Additionally, in the past decade numerous factors have been discovered to be negative regulators of angiogenesis. The best known of which are semaphorins (SEMAs) and their receptors neuropilins and plexins, ephrin ligands and Eph receptors, and slits and roundabouts [17]. Despite having been discovered in the nervous system as axon guidance cues, these protein families are now well recognised to also play a role in blood vessel guidance. As an example, Sema3a -/- mouse embryos show defects in cranial vessel remodelling [18], and the lack of functional neuropilin 1 leads to severe vascular abnormalities including vessel branching and sprouting defects [19].

Notably, an imbalance of pro- and anti-angiogenic stimuli can fuel pathological situations. A major example is angiogenesis in tumours.

Most tumours initially develop as avascular tissues. However to sustain their uncontrolled growth and progress beyond a certain size, they require nutrients and oxygen. To overcome these needs, the tumour activates the ‘angiogenic switch' [20]: tumour cells start releasing VEGF and other factors that diffuse into the microenvironment and activate the surrounding cells. ECs emerge from their quiescent state and acquire an angiogenic phenotype. Tumour angiogenesis goes through a similar set of cellular processes as those involved in physiological angiogenesis. However, due to the continuous and unbalanced supply of angiogenic stimuli produced by the tumour and stromal cells, the controlled sequential coordination of the ECs behaviour is disrupted. As a consequence, the tumour vessels have a highly aberrant organisation, structure and function [21]. At a structural level, the EC-cell adhesion is often weaker, mural cell coverage is inefficient and the deposited BM has irregular thickness and composition. As a result, tumour vessels are leaky and have chaotic blood flow [22]. The consequences of this include the intravasation and dissemination of tumour cells to form new metastatic sites, and insufficient oxygen supply to the tumour, resulting in further tumour-derived VEGF production and further deregulation of tumour vessel growth. Eventually, tumour angiogenesis can negatively impact on conventional anti-cancer treatments because it compromises chemotherapeutic delivery and induces resistance to radiation therapy [22]. Targeting the tumour vasculature is therefore an appealing anti-cancer therapeutic approach. Because VEGF was identified as a key tumour angiogenic factor, several anti-VEGF(R) blockers have been developed and approved by US Food and Drug Administration for use in the clinic [23]. Despite promising initial results of prolonged survival in patients, VEGF-targeted anti-angiogenic therapies failed to provide striking beneficial effects in the long term, and a portion of the patients developed resistance to the treatment [24]. However, several preclinical studies have shown rather than destroying the tumour vasculature, the blockade of VEGF signalling can transiently normalise the vessels by inducing the recruitment of pericytes and normalisation of the BM [25]. This restores the regular blood flow, enhances tumour oxygenation and improves the delivery of chemotherapy and efficacy of radiation therapy [26]. Additionally, recent studies have shown that promoting tumour vessel stabilisation reduces tumour metastasis [27, 28]. The normalisation of the tumour vasculature is therefore emerging as an intriguing new strategy to complement existing anticancer therapies, and can be a promising approach to increase the efficacy of drug delivery. Last but not least, tumour ECs may differ in their gene expression profile compared to ECs in normal vessels. This feature is of particular interest because surface markers for tumour ECs might enable targeting of therapeutic agents specifically to the tumour.

Because of its complexity, a comprehensive understanding of the molecular mechanisms regulating angiogenesis is a challenging and still incomplete task. However, because of its critical role in pathologies such as cancer, the potential benefit of increasing our knowledge of angiogenesis in disease justifies efforts to unravel the complexity of the process and offers approaches to manipulate the underlying mechanisms. For this purpose, in the past decades, in vitro and in vivo models have been developed to tackle different aspects of the vessel growth, and we describe them in detail below.

3 Models of angiogenesis

Models to study angiogenesis can be divided into two major categories, in vitro and in vivo.

In vitro cell-based models are essential in research because they are rapid, quantifiable, cheap and reproducible, and can be implemented for high-throughput assays. Furthermore, they can be easily manipulated, for example, by gene silencing and drug treatment. This provides a powerful tool to investigate protein function, especially when these cell alterations or treatments are not compatible with in vivo systems [29]. Since the 1970s, most of the in vitro assays applied to the study of angiogenesis used human umbilical vein ECs (HUVECs) owing to their ease of isolation and culture. More recently, improved methods in cell isolation have enriched the models with a wide diversity of ECs from different vessel types and organs [30]. In particular, microvascular ECs are the optimal cell type to investigate angiogenesis, as vessel growth is a microvascular process in vivo [31]. It should be noted, however, that certain primary ECs tend to dedifferentiate in culture, but with adequate monitoring of EC markers expression over passages, primary ECs are a powerful in vitro model since they resemble ECs in vivo more than current immortalised EC lines. A step towards a more realistic representation in vitro of the angiogenic process in vivo has been provided by the development of co-culture systems [30, 32]. These models allow the investigation of the bi-directional signalling between ECs, stromal cells and tumour cells, as well as the interaction with the surrounding matrix, which are important factors regulating angiogenesis.

In vitro models can be sub-classified depending on the dimensionality of the environment surrounding cells. These include classic 2D assays, where cells grow parallel to the culture surface, and the more sophisticated 3D systems, where cells are surrounded by matrix. 2D assays are extensively used because of their versatility and high reproducibility over a short term. In these systems, the cells are seeded on plastic dishes coated with extracellular matrix (ECM) proteins such as type I or IV collagen, fibronectin or the more complex tumour matrix (matrigel) and cellular mechanisms, such as response to angiogenic stimuli, proliferation, adhesion, migration and formation of basic tubule-like networks, can be investigated. 3D cultures provide a microenvironment that better mimics the in vivo situation. Indeed, they take into account the multidimensional interactions between ECs and the ECM, which can strongly influence EC behaviour. Additionally, 3D cultures represent a broader spectrum of processes that cannot be investigated by means of 2D systems, including sprouting, tip/stalk cell differentiation and lumen formation. Another relevant difference between 2D and 3D systems are the mechanical properties of their matrix, which is rigid in 2D and softer in 3D. Because increasing ECM rigidity has been shown to force tumour progression [33], these two models can be exploited to investigate the role of ECM stiffening on EC behaviour, to better understand the molecular mechanisms regulating angiogenesis in cancer. Of interest, cells can generally be isolated from these systems and used for downstream analysis such as Western blots or transcriptomics [32].

A further step of complexity can be achieved with ex vivo assays (organ culture), where a piece of tissue, usually an artery or vein ring section, is cultured in a 3D matrix, such as type I collagen. These assays represent a closer model of physiological angiogenesis because they include the different cell types that support the formation of new vessels. Additionally, these systems have been valuable for high-throughput screening of active molecules [30, 32].

Over the years, in addition to the information provided by in vitro assays, a variety of animal models that recapitulate the complexity of angiogenesis in humans have been established. Currently, the most widely used are the chick, zebrafish and mouse. These models can effectively contribute to our knowledge of vessel growth because of the conserved cellular and molecular mechanisms involved in angiogenesis among higher vertebrates. Additionally, these animals can be easily manipulated at gene level.

The veteran in vivo model is the chick chorioallantoic membrane whose main advantage is the easy accessibility of the vascular system to evaluate vessel development under the influence of factors with angiogenic activity. Importantly, the chick chorioallantoic membrane model has provided a valuable insight into the mutual influences of tumour and vasculature during their growth [34].

Another well-characterised in vivo model is zebrafish. The beauty of this system is that it allows high-resolution imaging of the vascular system during embryonic development. In combination with genetic manipulation, it is an incredibly powerful system to characterise the dynamics of the vascular network assembly in vivo, and as such enables the dissection of specific mechanisms, such as cell guidance, endothelial tip/stalk cell differentiation and vessel branching [35]. Importantly, despite the small size of the zebrafish, they can be used to isolate specific cellular subpopulations, including ECs, in the range of hundreds of thousands of cells [36]. This aspect has been exploited for genome profiling and makes this model very attractive for system-wide approaches in vivo.

The most widely used animal model is the mouse, and a great deal of effort has been devoted to generating robust tools to investigate vessel growth in this model. One of the most well characterised and utilised system to study sprouting angiogenesis is the dynamic remodelling of the retina in the neonatal mouse. Combined with imaging and loss and gain of function approaches, this model significantly contributed to the characterisation of the cellular and molecular cascades that regulate angiogenesis in development [37]. Interestingly, a comparative transcriptome analysis performed on tip and stalk ECs of sprouting blood vessels in mouse post-natal retina found that tip cells were enriched in mRNAs whose protein products are involved in axon guidance signalling, including SEMA3A and SEMA3F [38], and transcriptome analysis of retinas isolated from DLL4 mutant mice suggested novel proteins and mechanisms regulating tip cell function [39]. Other in vivo mouse models involve the delivery of substances via subcutaneous implantation of sponges or polymers. Although relatively invasive, they are useful for the fast evaluation of pro- and antiangiogenic molecules [40]. Also cells or pieces of tissues can be transplanted into mice. Of particular interest is the tumour xenograft model, where human cancer cells are transplanted into immunocompromised mice. This provides the opportunity to investigate tumour angiogenesis taking into account a microenvironment that, in comparison with in vitro co-culture models, better recapitulates the stroma in situ [41]. As for the zebrafish, xenografts are valuable models to isolate specific subpopulations of cells and this has been successfully exploited for transcriptomic comparison between normal and tumour ECs [42]. With modern gene manipulation and transfer methods researchers can deliberately mutate, deplete or overexpress genes in a precise manner [43], and this has been extensively applied to investigate selective gene functions in angiogenesis and to develop mouse models that recapitulate vascular diseases. Such approaches have helped to elucidate multi-step diseases such as cancer, as they offer the possibility to investigate cellular and molecular dynamics over the progression of the disease. For example, transgenic mouse models, such as the pancreatic islet cell carcinoma (RIP-Tag) and the epidermal squamous carcinoma (HPV16), are characterised by an ‘angiogenic switch' that precedes the appearance of the tumour, and also extensive vascularisation and angiogenesis of the solid tumour [44]. Both events are important for the development and progression of the tumour to malignancy. Notably, these models and the possibility to perturb individual stages of the progression by means of angiogenic and anti-cancer treatments are providing invaluable knowledge on tumour angiogenesis, with the potential to impact greatly on the development of anti-cancer therapies [30].

The in vitro and in vivo models described above have been most extensively explored with targeted techniques. However, because of the complexity of vessel growth in terms of dynamicity, coordination and integration of the different cellular processes, unbiased and system-wide approaches would provide the ideal perspective to complement the current knowledge and intimately understand the molecular mechanisms regulating this process. This has been done in the past with genome-based techniques, while proteome-based approaches were hindered by the high complexity and limited amount of sample. Below we describe the current proteomic technologies developed to tackle this challenge.

4 MS-based quantitative proteomics

Given that proteins are the major end product of the genome and the key functional units in the cell, it is of great interest to approach biological questions, including EC signalling in angiogenesis, using proteomics. For this purpose, MS-based technology has emerged as a dominant technique in the proteomic field since it provides robust methods to study protein abundance, interactions, subcellular localisation and modification state in a dynamic and unbiased manner. Currently the ambitious challenge of MS-based proteomics is to study biological systems in a comprehensive manner by attaining a complete proteome in a reasonable time frame whilst analysing only a small quantity of sample [45, 46]. To tackle this, MS shotgun proteomics is the peptide-based approach most widely used for protein identification from complex mixtures. In a typical shotgun approach, peptides from protein digestion are separated using LC, electrosprayed and analysed with high-resolution MS/MS (Fig. 2B). To successfully identify as many eluting peptides as possible it is critical to reduce the complexity of the sample and to use a mass analyser with high resolution, speed and dynamic range.

Sample complexity can be reduced by at least one-step fractionation at the protein or peptide level with, for example, 1D SDS gel [47], or ion exchange [48], respectively. Additionally, the online LC itself has the potential of very high separation power. Although this feature has been known for some time, its potential has been rediscovered in the MS field only recently, being compatible with the improving performance of mass analysers. Optimisation of the flow rate, size of packaging material and length of the column and gradient have all contributed to successfully increasing the number of identified proteins in single MS runs [49].

Incredible advances in the technology have given rise to major improvements in the performance of mass analysers. One of the most widespread high-resolution MS platforms in shotgun proteomics is based on the Orbitrap technology, first introduced to the proteomics field in 2005 [50]. To fully exploit the resolving power of the Orbitrap, despite the low speed, this mass analyser has been used in combination with high speed and high sensitivity linear trap quadrupole (LTQ)-Orbitrap [51, 52]. The newest generation of these hybrid instruments is the LTQ-Orbitrap Elite, which is characterised by ultra-high resolution (up to 240 000 resolution at m/z 400 Th) and high-speed MS/MS acquisition in the LTQ [53]. Interestingly, the Orbitrap has recently been combined with a quadrupole mass filter with high-performance precursor selection to generate the so-called Q Exactive [54]. This combination confers an incredible speed in selecting ions and makes it faster than the Orbitrap Elite. Additionally, the Q Exactive is a benchtop instrument, which brings high-resolution MS-based proteomics closer to a routine lab technology.

Together, high-resolution chromatography and Orbitrap technology have the proven capability to achieve close to comprehensive proteome analysis from a small quantity of sample, even in the absence of fractionation prior to LC-MS analysis [55].

Importantly, thanks to the high resolution and the different fragmentation methods available in the Orbitrap-based instruments – collision-induced dissociation, high collision dissociation and electron transfer dissociation – hybrid instruments perform extremely well to detect PTMs. In combination with enrichment protocols for specific PTMs it is indeed possible to perform global survey of cellular PTMs [56].

To understand the molecular dynamics that drive cellular processes, instead of measuring merely protein presence within a sample, it is often more intriguing to evaluate expression changes between experimental conditions. For this reason, MS-based technology is used in combination with quantitative methods. Currently, the two approaches most utilised are isotope labelling and label-free. In isotope-based methods, the intensity difference measured at the mass spectrometer between two peptides with the same chemical properties, but different stable isotope composition, represents relative abundance. The most widely used technique, which has become a gold standard in high accuracy quantitative proteomics, is metabolic labelling using SILAC [57]. In SILAC, cellular proteomes are labelled by incorporating non-radioactive stable isotope (13C, 15N, 2H)-containing amino acids, such as arginine and lysine (or other (semi)essential amino acids), in newly synthesised proteins. Another approach using stable isotopes is by chemical modification of proteins or peptides with tags containing stable isotope. The most well known techniques are dimethyl labelling and iTRAQ whose major advantages are the low cost and the possibility of multiplexing, respectively (these and other methods have been extensively reviewed in [58]). However, in comparison to other stable isotope methodologies, SILAC compares proteomes that are labelled from the beginning of the experiment and has the great advantage of being able to mix samples at the cell or lysate level, avoiding the possibility of introducing error during sample preparation to the accuracy of the quantification. Additionally, SILAC has proved to be an extremely versatile technology that, in combination with high-resolution MS, can be applied to perform a multitude of proteomic studies where accurate quantification is needed. These include PTMs at a global scale [56], temporal analysis [59] – crucial to investigate cellular dynamics – protein–protein and protein–nucleic acid interactions [60, 61] and sub-proteomes [62]. Although originally applied only to cell culture, SILAC has been extended to whole organisms, such as fly [63], fish [64] and mouse [65] allowing in vivo applications. Additionally SILAC can be applied in an indirect ‘spike-in’ format [66], where it is used to produce a heavy-labelled reference proteome, which is added as an internal standard to the proteomes to be compared. This further extends SILAC application to tissue samples that cannot be directly labelled, or to multiple sample analysis. In addition, SILAC has been adapted to perform absolute protein quantification, which determines the copy number of a specific protein per cell, important in biomarker research [67] and systems biology [68]. In this context, SILAC has been also used in combination with targeted proteomics, where a specific subset of proteins can be selectively detected and quantified in a large set of different conditions. This technique is based on multiple reaction monitoring and uses triple quadrupole instruments instead of the Orbitrap. The power of this technique is demonstrated by the very high sensitivity that allows identification of proteins down to few copies per cell in yeast [46].

Label-free quantification does not involve any particular manipulations of the sample prior to MS analysis and for this reason it is very simple. Importantly, it has reduced complexity and a greater dynamic range compared to label-based methods because it does not involve the mixing of proteomes. Label-free is becoming more used thanks to the recent development of sophisticated algorithms for quantification – based on spectral counting or peptide precursor ion intensity [69] – which provide more robust quantifications than previously achievable [60, 70]. However, compared to isotope-based methods, it is less accurate and, at present, label-free is suitable mainly when high changes in protein level are expected, such as with affinity purification of proteins.

Finally, MS shotgun quantitative proteomics is becoming well supported by computational platforms such as MSQuant [71] and MaxQuant [72] that, in combination with robust search engines used for peptide identification such as Mascot [73] and Andromeda [74], accomplish robust and highly accurate quantification analyses of high-resolution MS data. Importantly, also the speed of the MS data analysis has significantly improved, a prerequisite for dealing with the increasing amount of data. Of note, MaxQuant workflow is completely open source and because of its versatility – initially developed to support SILAC data, today it has extended its applicability to other quantification methods including label-free [60] – it is becoming widely used by the MS proteomic community. Altogether, these improvements in computational proteomics are making MS data analysis much more accessible than in the past.

In the modern proteomic era, high-resolution quantitative shotgun proteomics is therefore becoming a robust and successful tool to progress the understanding of molecular mechanisms in complex biological systems and with the promise to elucidate still unknown aspects of EC behaviour in angiogenesis.

5 MS-shotgun proteomics and ECs

In 2003 the first proteomic study of HUVECs in culture was published, where 53 proteins were identified from total lysate separated on 2D gel and analysed with TOF MS instrument [75]. A few years later, in a follow-up study, the number of identified proteins increased to a total of 162, and eight were found differentially expressed upon pro-apoptotic stimulation [76]. Since then, MS technology has been dramatically improved and MS shotgun proteomics has also started to impact on the research field of angiogenesis. Below, we review the use of shotgun proteomics focusing in particular on its application to improve our understanding of EC behaviour.

Many of the studies in the present literature have compared total proteomic changes of ECs in different states. The most extensive so far identified more than 5000 proteins, of which hundreds were found to be regulated according to label-free quantification, and provided a deep quantitative characterisation of the HUVECs response to inflammatory cytokines TNFα/IFNγ and ILβ [77]. In another study, the HUVEC proteome has been profiled to a depth of more than 3800 proteins upon stimulation with the pro-angiogenic factor VEGF for 4 and 8 h [78]. By means of stable 18O labelling, 1300 were quantified, where expression changes were observed for proteins known to be influenced by VEGF but, interestingly, also for proteins previously unknown to be involved in VEGF signalling.

EC stimulation by soluble factors has been recently approached in a time-resolved manner at the level of PTMs; upon angiotensin stimulation, 1288 phosphorylation sites have been identified, and isotope-based quantification established an unknown link between angiotensin and the transcription factor FOXO1 [79]. Because PTMs are critical to regulate protein function, their accurate quantification upon stimulation, although still underrepresented in literature, has the potential to bring tremendous insights into the understanding of EC signalling. In the context of angiogenesis, a detailed investigation of PTMs dynamics by angiogenic stimuli may identify proteins and mechanisms previously neglected in angiogenic signalling. Furthermore, the integration of information from single-factor stimulation will provide the opportunity to build up comprehensive functional maps of endothelial signalling and investigate possible crosstalk. Beyond single-factor stimulation also more complex stimuli, such as conditioned media and 3D microenvironments, may be used to stimulate ECs. In the context of cancer, the different response of ECs to conditioned medium generated from normal and tumour cells may be used to obtain insights on specific signalling for tumour angiogenesis to identify novel potential targets for therapies.

Another aspect that can be explored with shotgun proteomics and that is a unique feature for MS-based methods is the proteome of subcellular compartments [80]. Importantly, the quantitative and dynamic profile of subcellular compartments allows to investigate in a system-wide fashion how the distribution of proteins are altered, whose function are dependent on their subcellular localisation. So far, the secretome, including Weibel–Palade bodies secretory granules of ECs in culture [81], has been profiled by several groups [82, 83]. Conveniently, a SILAC-based strategy can overcome the problem of serum proteins from interfering with secretome sample analysis. Indeed, heavy proteins secreted by SILAC-labelled cells can be easily discriminated from the light serum proteins during MS data analysis [82]. The cellular secretome is of particular interest because it allows investigation of paracrine communication between ECs and surrounding cells. Interestingly, it has recently been shown that stromal EC secretions can significantly affect tumour progression [84, 85]. Many other subcellular compartments can be investigated by combining MS and specific isolation and enrichment protocols [80]. However, the literature lacks such information for ECs, where mainly membranome [86] and lipid rafts [87] have been profiled to date. Subcellular proteome dynamics also allow the investigation of emerging cellular regulatory mechanisms such as protein trafficking. Recently, Simons and co-workers have revealed that the endosomal trafficking of growth factor receptors, such as VEGFR2, regulates their function in ECs and that this is relevant during vessel growth [88]. It would be therefore intriguing to combine high accuracy quantitative shotgun proteomics with purification of endosomal compartments for in-depth investigation into trafficking dynamics and their functions in angiogenesis. Subcellular proteomics may also provide important insights for clinical research applications. As an example, since ECs are in direct contact with the blood, comparative analysis of secreted proteins from normal and tumour-derived ECs are a resource of candidate biomarkers, which can be used in monitoring the progression of disease or therapeutic efficacy. Conversely, differentially expressed plasma membrane proteins between normal and tumour ECs, can potentially identify novel surface markers for tumour ECs. While this has been previously tested indirectly with phage display technology [89], measuring the EC surface proteome will likely improve our capability to efficiently target anti-cancer therapies.

Technical advances to isolate specific cell populations from tissues, such as laser capture microdissection, have allowed EC extraction from fixed tissues and investigating EC proteomics in vivo. This is particularly relevant in the field of cancer research for identifying markers and mechanisms specifically regulated in tumour ECs. Due to the difficulty of extracting intact mRNA from fixed tissues, MS-based methods offer a valuable alternative to exploit formalin fixed tumour tissues for angiogenesis research. The applicability of such a technique has been recently shown in a recent publication where less than 3000 ECs were obtained by laser capture microdissection from gliomas and normal tissues and were measured with high-resolution MS (by using first generation LTQ-Orbitrap) to a depth of almost 700 proteins [90]. Importantly, tumour angiogenesis can be studied by exploiting the variety of mouse models that recapitulate defined angiogenic stages during tumour progression. A clear advantage of these models is that mice are an accessible source of fresh tissues and, in combination with antibody-based affinity purification or FACS sorting, ECs can be isolated from normal and tumour tissues. It is therefore very tempting to imagine measuring in-depth proteomic changes in ECs using this process to better understand angiogenic deregulation in cancer. This analysis can be further extended to mice treated with angiogenic and anti-cancer therapeutics, to pinpoint molecular changes in the vasculature, critical for tumour progression. Recently, benefits provided by vessel normalisation have been investigated in preclinical trials to improve radio and chemotherapeutic efficacy, and improved characterisation of the mechanisms responsible for induced normalisation would be enormously beneficial to improve therapies.

Due to the restricted amount of sample available, in vivo systems can be investigated mainly at the proteome level, while the analysis of global PTMs and sub-proteomes, where enrichment protocols often require milligrams of starting material, are still a challenging task. However, very sophisticated chromatographic settings applied to phosphoproteomics have recently reached very high sensitivity and profiled the phosphoproteome of 1 μg of cell lysate to a depth of 1011 phosphorylation sites [91]. This makes PTM analysis for small sample amount an achievable goal in the future and an exciting perspective for studying EC signalling in vivo.

Despite the fact that most of the work so far has not been done specifically in the context of angiogenesis, it is clear that ECs are a source of valuable information to be approached with MS shotgun proteomics and that this can provide novel and valuable insights into EC behaviour. So far, the depth and accuracy of the MS proteomic results that have been reported in the literature are not yet comparable to state-of-the-art results obtained in pioneering proteomic studies. This is because the last generation of high-resolution mass analysers and accurate quantifications using SILAC have not yet been used together to investigate EC behaviour. We believe that this can be achieved in the immediate future because high-resolution MS is becoming a more accessible technology and because of the possibility to apply SILAC, in a direct or indirect spike-in fashion, to ECs. Indeed, it has been reported that ECs, both primary [82, 92] and cell lines [93], can be SILAC labelled and in Fig. 3 we show that complete SILAC-labelling of primary human macro- and microvascular ECs can be achieved in a short time, without major interference in the expression of EC markers such as von Willebrand factor and the platelet EC adhesion molecule. SILAC may also be used for accurate quantification in vivo as an internal standard [66], where SILAC-labelled ECs are used as a reference sample for indirect comparison. SILAC-ECs are required to have a proteome similar, but not quantitatively identical, to the samples to be quantified. For this reason, primary or immortalised ECs can be SILAC-labelled in culture, and lysates spiked-in in equal amount to the isolated ECs to be compared. For mouse ECs, an additional possibility for quantification is represented by ECs isolated from the 13C6 lysine-labelled SILAC mouse [65]. However, because of the higher costs compared to in vitro labelling, the SILAC mouse is an appropriate option when low amount of standard is needed.

Figure 3.

SILAC labelling of primary ECs. HUVECs and human ovarian microvessel ECs were cultured in EGM-2 medium lacking arginine and lysine, with supplemented 13C14N lysine (Lys8, 87.5 μg/mL) and 13C14N arginine (Arg 10, 28 μg/mL) and 2% and 10% 10 kDa dialysed FBS, respectively. For incorporation check (A), cells at the indicated passage (P) were lysed in urea buffer and digested in solution prior to MS (LTQ-Orbitrap Velos) and data (MaxQuant environment) analysis. Each incorporation curve represents the distribution of 600–8000 quantified peptides. The incorporation, x-axis, has been calculated based on the peptide ration SILAC/non-SILAC. y-axis is in arbitrary units. For immunofluorescence (B), immunostaining was carried out on paraformaldehyde fixed cells (P3 for HUVECs and P6 human ovarian microvessel ECs), with primary mouse anti-von Willebrand factor and sheep anti-platelet endothelial cell adhesion molecule followed by secondary Alexa fluor-555 and Alexa fluor-488 labelled antibodies and DAPI for the nuclei. VWF, von Willebrand factor; PECAM1, platelet endothelial cell adhesion molecule. Immunofluorescence was analysed with a Zeiss 710 confocal microscope at 40´ magnification. Bar = 50 μm.

The application of these technologies, in combination with PTMs and subcellular enrichment protocols described above, to the variety of available in vitro and in vivo models of angiogenesis holds the promise of shedding new light on the cellular and molecular mechanisms that govern angiogenesis (Fig. 4). Finally, system-wide analysis of proteomes, sub-proteomes and PTMs of ECs combined with improved techniques and softwares for absolute quantification have the potential to deliver an even deeper understanding of signalling in angiogenesis. Although at an early stage of development, systems biology is another rapidly growing research field, which aims to combine system-wide quantitative data with sophisticated mathematical modelling to investigate protein networks in time and space. This technique will enable a greater understanding of endothelial signalling dynamics during physiological and importantly, pathological angiogenesis.

Figure 4.

In vitro and in vivo models of angiogenesis and their potential in quantitative MS. Schematic representation of the most used models ordered by increasing complexity and similarity to human. For the in vitro models, the individual cellular processes that can be studied are indicated, while for the in vivo systems, whether the mechanisms can be investigated in physiological or pathological angiogenesis (column ‘process'). As indicated in the column ‘MS-based quantification', most of the models can be approached with SILAC-based proteomics, though in different format. In the last column, high (+++), medium (++) and low (+) refers to the estimated protein amount available for MS analysis. (−) no literature was found of vessels/cells isolated from the model.

6 Conclusion

The incredible advances in MS technology and its versatile and robust applications to a variety of biological questions make MS-based proteomics a leading technology to explore complex biological systems. One of those is angiogenesis, where ECs need to coordinate a plethora of different cellular and molecular mechanisms in a dynamic and ordered manner to drive the formation of new functional vessels. Although still in its infancy, modern MS proteomics applied to existing robust in vitro and in vivo models of angiogenesis has great potential to significantly impact on the current knowledge of the complexity and dynamicity of vessel growth and open new perspectives in the context of pathologies such as cancer. We believe therefore that proteomics applied to the study of angiogenesis will initiate an exciting new era of discoveries in the field.


The work was funded by Cancer Research, UK, and we gratefully acknowledge this generous support.

The authors have declared no conflict of interest.