Open Access

Why some tumours trigger neovascularisation and others don’t: the story thus far

Chinese Journal of Cancer201635:18

https://doi.org/10.1186/s40880-016-0082-6

Received: 21 October 2015

Accepted: 20 December 2015

Published: 12 February 2016

Abstract

Background

Angiogenesis is not essential for tumours to develop and expand, as cancer can also grow in a non-angiogenic fashion, but why this type of growth occurs is unknown. Surprisingly, our data from mRNA transcription profiling did not show any differences in the classical angiogenic pathways, but differences were observed in mitochondrial metabolic pathways, suggesting a key role for metabolic reprogramming. We then validated these results with mRNA profiling by investigating differential protein expression via immunohistochemistry in angiogenic and non-angiogenic non-small cell lung cancers (NSCLCs).

Methods

Immunohistochemical staining for 35 angiogenesis- and hypoxia-related biomarkers were performed on a collection of 194 angiogenic and 73 non-angiogenic NSCLCs arranged on tissue microarrays. Sequencing of P53 was performed with frozen tissue samples of NSCLC.

Results

The non-angiogenic tumours were distinguished from the angiogenic ones by having higher levels of proteins associated with ephrin pathways, mitochondria, cell biogenesis, and hypoxia-inducible factor 1 (HIF1) regulation by oxygen and transcription of HIF-controlled genes but lower levels of proteins involved in the stroma, cell–cell signaling and adhesion, integrins, and Delta-Notch and epidermal growth factor (EGF)-related signaling. However, proteins classically associated with angiogenesis were present in both types of tumours at very comparable levels. Cytoplasmic expression of P53 was strongly associated with non-angiogenic tumours. A pilot investigation showed that P53 mutations were observed in 32.0% of angiogenic cases but in 71.4% of non-angiogenic tumours.

Conclusions

Our observations thus far indicate that both angiogenic and non-angiogenic tumours experience hypoxia/HIF and vascular endothelial growth factor (VEGF) pathway protein expression in a comparable fashion. However, angiogenesis does not ensue in the non-angiogenic tumours. Surprisingly, metabolic reprogramming seems to distinguish these two types of neoplastic growth. On the basis of these results, we raise the hypothesis that in some, but not in all cases, initial tissue remodeling and/or inflammation could be one of the secondary steps necessary to trigger angiogenesis. In the non-angiogenic tumours, in which neovascularisation fails to occur, HIF pathway activation could be the driving force toward metabolic reprogramming.

Keywords

Cancer Angiogenesis Hypoxia Blood vessels

Background

In 1927, Otto Warburg described what would be called the “Warburg effect,” in which tumour cells exhibited characteristic changes in metabolism, particularly the use of glycolysis rather than oxidative phosphorylation, despite the presence of adequate amounts of oxygen [1, 2]. Warburg believed that this process was the actual cause of neoplastic transformation [3].

Tumour development is now known to be driven by genetic damage. However, mutations in some metabolic enzymes, such as succinate dehydrogenase (SDH) and fumaratehydratase (FH), both parts of the tricarboxylic acid (TCA) cycle, can drive neoplastic transformation, and intermediate products of metabolism can also promote neoplastic progression [4].

Proliferating cancer cells have a high energy requirement to maintain homeostatic cellular processes. The shift in energy production to aerobic glycolysis, while allowing for more rapid production of adenosine triphosphate (ATP), yields far less energy than oxidative phosphorylation: there are two net molecules of ATP per glucose molecule in glycolysis versus 36 molecules of ATP via oxidative phosphorylation [5].

The reasons for the glycolytic energy dependence of proliferating tumour cells are still being debated. Initially, it was believed that the mitochondria in tumours were intrinsically defective. However, it was determined that tumour mitochondria are actually functional, retaining the capacity for oxidative phosphorylation and consuming oxygen at similar rates to normal tissues [6], although it should be appreciated that a degree of variability in mitochondrial activities exists across different neoplasms. Alternatively, high rates of glycolysis might be co-selected with factors that promote the expression of hypoxia-related genes (such as those required for angiogenesis) as an oxygen-independent energy source. Finally, increased intermediate products of glycolysis can easily be shunted into the biosynthetic pathways required for serine and nucleotide synthesis [7].

According to Folkman’s original theory [8], the onset of hypoxia in tumour triggers angiogenesis, which in turn is essential for supplying neoplastic cells with nutrients and oxygen and evacuating metabolic waste and carbon dioxide. The best understood hypoxia signaling mechanism is the stabilization and post-transcription activation of the hypoxia-inducible factor (HIF) proteins, which lead to the activation of many different pathways, including the vascular endothelial growth factor (VEGF) pathway. The VEGF pathway prompts and supports neoangiogenesis and glycolysis. Hypoxia-inducible pathway activation also has other effects, which include reducing the activity of mammalian target of rapamycin (mTOR), which in turn can reignite autophagy and promote survival under stress [9].

HIF is a heterodimer of an alpha subunit that is unstable in normoxia and a constitutively present and stable beta subunit. The hypoxia activation of HIF causes the heterodimer to bind to DNA at specific locations, called hypoxic response elements (HREs), eliciting the transcriptional up-regulation of genes required to respond appropriately to hypoxia [9].

In addition to triggering the VEGF pathway, the ubiquitously expressed HIF1 isoform promotes the transcription of glucose transporter 1 (GLUT1), which activates glucose transport inside the cell, lactate dehydrogenase-A (LDH-A), which is involved in the glycolytic pathway, erythropoietin (EPO), which enhances erythropoiesis, and nitric oxide synthase (NOS), which promotes angiogenesis and vasodilatation [9].

HIF1 also prevents the entry of pyruvate into the TCA cycle by inducing the expression of pyruvate dehydrogenase kinase 1 (PDK1), thus altering the expressed isoform of cytochrome c and inhibiting mitochondrial biogenesis. This process causes reduced levels of oxygen consumption and a shift away from oxidative phosphorylation. Interestingly, HIF1 can also be activated under normoxic conditions by a variety of oncogenic pathways, such as phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA), and by mutations in von Hippel-Lindau tumour suppressor (VHL), SDH, and FH [10].

In the classic angiogenic pathway, VEGF binds to VEGF receptor 2 (VEGFR2) on endothelial cells, increasing the expression of the Notch ligand Delta-like 4 (DLL4) on the same cells. DLL4 then binds to its receptor Notch on the adjacent endothelium. Further expression of VEGFR2 and VEGFR1, as well as a smaller amount of VEGFR3, then follows, leading to triggering/amplification of the downstream phospholipase C familyγ (PLCγ)–protein kinase C (PKC)–Raf kinase–MAP kinase-ERK kinase (MEK)–mitogen-activated protein kinase (MAPK) pathway, concomitantly prompting cell proliferation and cell survival throughout the phosphoinositide 3-kinase (PI3 K)/protein kinase B (AKT) pathway [11].

The switch to glycolysis in neoplasia was, according to Warburg, irreversible [3], yet a more complex picture has emerged over the last decade. There have been observed instances in which oxidative phosphorylation predominates during neoplastic transformation [12]. This variation between OxPhos and glycolysisin cancer cells has been increasingly linked to specific disturbances in cell signaling pathways [13].

Additionally, tumours of the same genetic lineage can develop different metabolic adaptations depending on the host tissue from which they arise, suggesting that the stromal environment might play a crucial role in shaping the metabolic profile [14]. The different molecular mechanisms being postulated to explain this variability of the Warburg effect include the following: inhibition of pyruvate dehydrogenase (PDH) by PDK1, reduction of mitochondrial biogenesis and inhibition of oxidative phosphorylation, both are caused by P53 inactivation and mutations [15].

Warburg raised two important issues: first, how tumour cells are supplied with glucose; and second, how they are supplied with oxygen [1]. Folkman’s work addressed the latter question with the hypothesis that tumour growth is strictly angiogenesis-dependent [16]. The work undertaken to test this hypothesis led to the inclusion of “angiogenesis” as one of the hallmarks of cancer [8].

Although there is strong evidence that angiogenesis frequently occurs in cancer, we also now know that this event does not always occur. Indeed, some tumours, called “non-angiogenic tumours,” can grow without triggering new vessel formation by co-opting preexisting vessels [17, 18].

Non-angiogenic growth was first identified by histology in primary and metastatic lung carcinomas because neoplastic cells filled the alveolar spaces, co-opting the pre-existing capillary network and exhibiting a characteristic “chicken-wire” appearance [17]. A gene expression signature for non-angiogenic non-small cell lung cancer (NSCLC) was published in 2005 [19]. Surprisingly, rather than the classic angiogenesis-related genes, the differentially expressed genes were involved in mitochondrial metabolism, transcription, protein synthesis, and the cell cycle. Lack of differential mRNA expression between tumour phenotypes was noted for genes classically associated with hypoxia and angiogenesis. This result suggested that the response to hypoxia does not necessarily trigger neovascularisation, as would be observed in angiogenic tumours, but could actually be dependent on the genetic background of neoplastic cells, and in some instances, it could lead to metabolic reprogramming [19]. We therefore postulated that the degree to which a tumour will rely on angiogenic or non-angiogenic growth could be associated with a variety of events, including hypoxia, pseudo-hypoxia, and metabolic re-programming.

In the first part of the present study, we investigated whether there were truly no differences in the expression of hypoxia-and angiogenesis-related proteins between angiogenic and non-angiogenic tumours, as suggested by mRNA profiling. We also investigated the degree of these proteins expression and the expression of some mitochondrial biogenesis proteins via immunohistochemistry. Notably high cytoplasmic P53 expression in non-angiogenic tumours, compared to angiogenic tumours, was found after completing the first part of this study. We therefore performed a second investigation, in which we examined and sequenced the p53 gene in these tumours.

Methods

Tissue specimens

Clinical specimens of NSCLCs were obtained from a series of consecutive patients who underwent surgical treatment at the John Radcliffe Hospital, Oxford, UK. This collection had ethical committee approval (study number C02.216—The pathophysiology of human neoplasia). The tissues were formalin-fixed and paraffin-embedded. Tissue microarrays were constructed using the Beecher Instrument MTA-1 manual arrayer (Beecher Instruments, Inc., Sun Prairie, WI, USA). Up to four suitable areas of appropriate tumour were chosen from a slide stained with hematoxylin and eosin (H&E), avoiding areas of necrosis.

Immunohistochemical staining of tissue sections

The 4-μm sections were cut from paraffin blocks and mounted on glass histology slides. Non-specific protein binding was blocked by incubating with 2.5% normal horse serum (Vector Laboratories, Burlingame, CA, USA), and the primary antibody was then applied. Details of all of the antibodies used are presented in Table 1. Substitution of the primary antibody with phosphate-buffered saline (PBS) served as a negative control. The slides were then counterstained with hematoxylin for 20 s.
Table 1

Antibodies used for immunohistochemistry

Antigen

Clone

Source

HIF1

ESEE122

Abcam (Cambridge, UK)

HIF2

EP109b

NDCLS (University of Oxford, John Radcliffe Hospital, Oxford, UK)

CA9

M75

BioScience Slovakia s.r.o. (Bratislava, Slovakia)

VEGFA

VG-1

NDCLS

TYMP

PGF44

NDCLS

KDR

FLK1(A3)

Santa Cruz Biotechnology, Inc. (Dallas, TX, USA)

KDRp34

34a

NDCLS

FIH

FIH162c/D6

NDCLS

PHD1

PHD112/G

NDCLS

PHD2

366G/76

NDCLS

PHD3

EG188c

NDCLS

DLL4

D4/37

NDCLS

TSP1

8A6B-TSP-1

Leica Microsystems (UK) Ltd. and Novocastra Reagents (Milton Keynes, UK)

CXCR4

MAB 172

R&D System (Minneapolis, MN, USA)

EPHB2

AF496

R&D System

EPHB3

R&D AF432

R&D System

EPHB4

R&D AF446

R&D System

SOD1

30F11

Novocastra

BCL2

124

Dako (Cambridgeshire, UK)

FOS

Polyclonal 27436

Abcam

EGF

EGF 10

Abcam

EGFR

F4

Abcam

BNIP3

Ana40

Sigma-Aldrich Company Ltd. (Dorset, UK)

P53

DO7

Dako

PI3

Ab 40755

Abcam

SP1

SP1 polyclonal

Abcam

STAT3

E121-21

Abcam

LON

20H1

NDCLS

MEF2D

MEF2D polyclonal

Abcam

JMY

HMY117A

NDCLS

TRAP1

TR-1A

LabVision (TorsbySjöväg, Värmdö, Sweden)

GST

GST3/GST pi

Abcam

NCAM

IB6

Novocastra

CHGA

FLEX polyclonal

Dako

SYP

299

Novocastra

HIF hypoxia-inducible factor; CA9 carbonic anhydrase 9; VEGFA vascular endothelial growth factor A; TYMP thymidine phosphorylase; KDR vascular endothelial growth factor receptor 2; KDRp34 vascular endothelial growth factor p34; FIH factor-inhibiting HIF; PHD prolyl hydroxylase dehydrogenase; DLL4 Delta-like 4; TSP1 thymidine phosphorylase 1; CXCR4 chemokine (C-X-C motif) receptor 4; EPH ephrin; SOD1 superoxide dismutase 1; BCL2 B cell lymphoma 2; FOS FBJ murine osteosarcoma viral oncogene homolog; EGF epidermal growth factor; EGFR epidermal growth factor receptor; BNIP3 BCL2/adenovirus E1B 19 kDa interacting protein 3; PI3 peptidase inhibitor 3; SP1 Sp1 transcription factor; STAT3 signal transducer and activator of transcription 3; LON Lon protease; MEF2D myocyte enhancer factor 2D; JMY junction-mediating and regulatory protein, p53 cofactor; TRAP1 TNF receptor-associated protein 1; GST glutathione S-transferase; NCAM neural cell adhesion molecule; CHGA chromogranin A; SYP synaptophysin

Scoring

The immunohistochemical staining was scored for cytoplasmic and nuclear localization and, when present, for membrane staining. Two observers scored the slides. Intensity was scored on a scale of 0–3 (0 = no staining, 1 = weak, 2 = moderate, 3 = strong staining). The percentage of positive cells was recorded on a scale from 0 to 4 (1 = 1%–10%, 2 = 11%–50%, 3 = 51%–80%, 4 = 81%–100%) or alternatively on a continuous scale from 0% to 100%. The intensity and percentage values were then multiplied to provide a score called the “intensity percentage score” (IPS), the maximum score of which ranged from 12 to 300 [20].

Disputed scores were discussed and a consensus reached. For tissue microarrays, cores that did not contain tumour tissue or that were more than 50% incomplete were excluded. The number of cases scored for each marker was between a minimum of 73 and a maximum of 194 cases for angiogenic tumours and between a minimum of 48 and a maximum of 73 cases for non-angiogenic tumours.

Statistical analysis

To evaluate the association between protein biomarker expression and angiogenic status, the Mann–Whitney two-tailed non-parametric test was performed using GraphPad Prism statistical analysis software, version 4 (GraphPad Software Inc., San Diego, CA, USA).

Pathway visualization

Three lists were made: two of proteins more highly expressed in angiogenic or non-angiogenic tumours at least in one subcellular location and the third of proteins always equally expressed in all of their subcellular localizations (Table 2). To visualize the pathways associated with these proteins, each list was imported into the Web-based Enricher facility (http://amp.pharm.mssm.edu/Enrichr/) [21]. The data were visualized from the online databases for gene ontologies (GO biological process and GO cellular component) and for pathways [Kyoto Encyclopedia of Genes and Genomes (KEGG) 2015, WikiPathways 2015, Reactome 2015, and Panther]. The results were visualized by bar graph sorted by combined score.
Table 2

Lists of proteins used for pathway visualization using the EnrichrWebd facility [21]

Expression status

Proteins

Equal expression in both tumour types

HIF2, VEGFA, TYMP, KDR, KDRp34, FIH, PHD1, SOD1, EPHB4, BCL2, EGFR, FGF, SP1, LON, MEF2D, RPSAa, CHGA, SYP

Higher expression in angiogenic tumours than in non-angiogenic tumours in at least one subcellular location

HIF1, PHD2, PHD3 (cytoplasmic), CXCRN, TSP, DLL4, BNIP3, PI3, EGF, FOS, STAT3, ITGB3a, ITGAVa

Higher expression in non-angiogenic tumours than in angiogenic tumours in at least one subcellular location

CA9, PHD3 (nuclear), EPHB2, EPHB3, NCAM, P53, TRAP1, JMY, GST

FGF fibroblast growth factor; RPSA ribosomal protein SA; ITGB3 integrin, beta 3; ITGAV integrin alpha-V; HIF hypoxia-inducible factor; CA9 carbonic anhydrase 9; VEGFA vascular endothelial growth factor A; TYMP thymidine phosphorylase; KDR vascular endothelial growth factor receptor 2; KDRp34 vascular endothelial growth factor p34; FIH factor-inhibiting HIF; PHD prolyl hydroxylase dehydrogenase; DLL4 Delta-like 4; TSP1 thymidine phosphorylase 1; CXCR4 chemokine (C-X-C motif) receptor 4; EPH ephrin; SOD1 superoxide dismutase 1; BCL2 B cell lymphoma 2; FOS FBJ murine osteosarcoma viral oncogene homolog; EGF epidermal growth factor; EGFR epidermal growth factor receptor; BNIP3 BCL2/adenovirus E1B 19 kDa interacting protein 3; PI3 peptidase inhibitor 3; SP1 Sp1 transcription factor; STAT3 signal transducer and activator of transcription 3; LON Lon protease; MEF2D myocyte enhancer factor 2D; JMY junction-mediating and regulatory protein, p53 cofactor; TRAP1 TNF receptor-associated protein 1; GST glutathione S-transferase; NCAM neural cell adhesion molecule; CHGA chromogranin A; SYP synaptophysin

aData for RPSA are from Ref. [18] and for ITGB3 and ITGAV are from Ref. [23]

P53 sequencing

Genomic DNA was isolated from 33 specimens of NSCLC (25 angiogenic cases, seven non-angiogenic cases, and one undetermined case) and five specimens of peri-tumour lung tissues. The coding regions corresponding to exons 3–9 (amino acids 25–331) were sequenced. DNA was whole-genome amplified using GenomiPhi (GE Healthcare, Piscataway, NJ, USA). The primers for polymerase chain reaction (PCR) amplification and subsequent sequencing reactions are described in Table 3. PCR was performed using ThermoStart PCR Master Mix (Thermo Fisher Scientific, Waltham, MA, USA), following the manufacturer’s protocol. PCR products were purified and bidirectionally sequenced using the BigDye Terminator cycle sequencing kit, version 1.1 (Applied Biosystems, Foster City, CA, USA), and an ABI 3100 Genetic Analyzer (Applied Biosystem, Paisley, UK). Sequence data were analyzed using Mutation Surveyor, version 3.25 (Softgenetics, State College, PA, USA). Predicted effects on protein were assessed in silico using PolyPhen2 software (http://genetics.bwh.harvard.edu/pph2/). Fisher’s two-sided exact test was performed to compare mutation frequencies in angiogenic versus non-angiogenic cases. All sequencing experiments were performed in duplicate.
Table 3

Primers for polymerase chain reaction (PCR) amplification and subsequent P53 sequencing reactions

Primer

Sequence

Annealing temperature (°C)

Amplicon length (bp)

Exon 3–4

Forward: GTGGGAAGCGAAAATTCCAT

Reverse: GCCAGGCATTGAAGTCTCAT

60

506

Exon 5–6

Forward: TGTTCACTTGTGCCCTGACT

Reverse: TTAACCCCTCCTCCCAGAGA

60

467

Exon 7

Forward: GAGCTTGCAGTGAGCTGAGA

Reverse: GGGATGTGATGAGAGGTGGA

EX7F_seq CCTGCTTGCCACAGGTCT (to be used for sequencing instead of Exon 7 forward primer)

61.5

390

Exon 8–9

Forward: GACAAGGGTGGTTGGGAGTA

Reverse: GCCCCAATTGCAGGTAAAAC

60

500

Results

Immunohistochemical staining

The complete results for the cytoplasmic and membranous expression are reported in Table 4, and those for nuclear expression are reported in Table 5, while a summary of these results appears in Table 6. The complete original results of ontology and pathway visualization appear in Additional file 1.
Table 4

Cytoplasmic and membranous expression of the proteins in angiogenic versus non-angiogenic non-small cell lung cancers (NSCLCs)

Protein

Intensity percentage score (IPS) of protein expression

P value

Tumour type with higher expression

Angiogenic tumours

Non-angiogenic tumours

HIF1

6.22 ± 0.33

6.00 ± 0.48

0.87

HIF2

0.00 ± 0.00

0.00 ± 0.00

Not applicable

CA9 cytoplasm

2.38 ± 0.16

3.50 ± 0.23

<0.001

Non-angiogenic

CA9 membrane

4.00 ± 0.40

3.30 ± 0.88

0.62

VEGFA

8.83 ± 0.35

7.38 ± 0.78

0.07

TYMP

3.20 ± 0.54

2.21 ± 0.32

0.28

KDR

94.56 ± 1.64

97.80 ± 1.54

>0.05

KDRp34

10.40 ± 0.30

10.55 ± 0.62

0.71

FIH

9.10 ± 0.32

10.00 ± 0.47

0.22

PHD1

3.74 ± 0.27

4.20 ± 0.48

0.47

PHD2

3.50 ± 0.30

2.46 ± 0.53

0.02

Angiogenic

PHD3

2.70 ± 0.09

3.38 ± 0.17

<0.001

Non-angiogenic

DLL4vessels

2.34 ± 0.10

1.53 ± 0.09

<0.001

Angiogenic

TSP stroma

16.98 ± 2.75

2.00 ± 1.74

<0.001

Angiogenic

CXCR4

2.47 ± 0.20

2.34 ± 0.31

0.92

 

EPHB2

7.78 ± 0.27

9.06 ± 0.33

<0.01

Non-angiogenic

EPBH3 cytoplasm

159.3 ± 8.84

202.20 ± 9.95

<0.01

Non-angiogenic

EPBH3 membrane

23.83 ± 7.23

86.51 ± 12.78

<0.001

Non-angiogenic

EPHB4

7.82 ± 0.26

7.40 ± 0.38

0.57

SOD1

4.36 ± 0.45

9.39 ± 4.22

0.27

BCL2

1.05 ± 0.25

0.68 ± 0.35

0.54

FOS

8.05 ± 0.37

5.31 ± 0.55

<0.001

Angiogenic

EGFR

203.40 ± 10.79

166.30 ± 25.74

0.33

EGF

18.90 ± 3.03

14.44 ± 4.15

0.72

FGF

9.20 ± 0.42

9.72 ± 0.71

0.74

BNIP3

6.08 ± 0.36

5.92 ± 0.82

0.88

P53

0.28 ± 0.12

2.28 ± 0.56

<0.001

Non-angiogenic

PI3

4.62 ± 0.36

4.92 ± 0.61

0.51

SP1

3.70 ± 0.37

3.38 ± 0.55

0.86

STAT3

8.78 ± 0.39

7.11 ± 0.79

0.03

Angiogenic

LON

102.50 ± 11.20

134.20 ± 17.10

0.18

MEF2D cytoplasm

50.11 ± 4.16

65.67 ± 9.66

0.25

MEF2Dmembrane

60.00 ± 7.12

67.50 ± 12.67

0.47

JMY

1.84 ± 0.18

2.42 ± 0.31

0.05

Non-angiogenic

TRAP1

4.22 ± 0.28

5.48 ± 0.31

<0.01

Non-angiogenic

GST cytoplasm

236.20 ± 7.16

249.80 ± 11.40

0.23

GSTmembrane

28.44 ± 5.52

69.38 ± 14.03

<0.01

Non-angiogenic

NCAM

0.40 ± 0.09

3.92 ± 0.07

<0.001

Non-angiogenic

CHGA

0.08 ± 0.04

0.27 ± 0.14

>0.05

SYP

0.12 ± 0.05

0.44 ± 0.15

>0.05

All data are presented as mean ± standard error

HIF hypoxia-inducible factor; CA9 carbonic anhydrase 9; VEGFA vascular endothelial growth factor A; TYMP thymidine phosphorylase; KDR vascular endothelial growth factor receptor 2; KDRp34 vascular endothelial growth factor p34; FIH factor-inhibiting HIF; PHD prolyl hydroxylase dehydrogenase; DLL4 Delta-like 4; TSP1 thymidine phosphorylase 1; CXCR4 chemokine (C-X-C motif) receptor 4; EPH ephrin; SOD1 superoxide dismutase 1; BCL2 B cell lymphoma 2; FOS FBJ murine osteosarcoma viral oncogene homolog; EGF epidermal growth factor; EGFR epidermal growth factor receptor; BNIP3 BCL2/adenovirus E1B 19 kDa interacting protein 3; PI3 peptidase inhibitor 3; SP1 Sp1 transcription factor; STAT3 signal transducer and activator of transcription 3; LON Lon protease; MEF2D myocyte enhancer factor 2D; JMY junction-mediating and regulatory protein, p53 cofactor; TRAP1 TNF receptor-associated protein 1; GST glutathione S-transferase; NCAM neural cell adhesion molecule; CHGA chromogranin A; SYP synaptophysin

Indicates equal expression

Table 5

Nuclear expression of the proteins in angiogenic versus non-angiogenic NSCLCs

Protein

Intensity percentage score (IPS) of protein expression

P value

Tumour type with higher expression

Angiogenic tumours

Non-angiogenic tumours

HIF1

5.49 ± 0.31

3.79 ± 0.39

0.01

Angiogenic

VEGFA

6.09 ± 0.52

5.62 ± 0.32

0.74

TYMP

1.94 ± 0.22

1.16 ± 0.21

0.42

KDR

231.20 ± 8.55

229.80 ± 9.88

>0.05

KDRp34

11.69 ± 0.16

11.45 ± 0.39

0.77

FIH

7.50 ± 0.35

8.15 ± 0.72

0.50

PHD1

3.10 ± 0.34

2.50 ± 0.35

0.99

PHD2

2.56 ± 0.27

1.41 ± 0.35

0.02

Angiogenic

PHD3

2.18 ± 0.13

0.65 ± 0.20

<0.001

Angiogenic

CXCR4

5.13 ± 0.23

3.72 ± 0.37

<0.01

Angiogenic

EPHB2

3.26 ± 0.26

4.91 ± 0.35

<0.001

Non-angiogenic

EPBH3

99.27 ± 7.41

147.30 ± 9.23

<0.001

Non-angiogenic

EPHB4

5.56 ± 0.30

5.73 ± 0.34

0.81

SOD1

4.21 ± 0.45

5.00 ± 0.91

0.44

C-FOS

7.50 ± 0.52

5.53 ± 0.75

0.05

Angiogenic

EGF

235.20 ± 5.60

214.8 ± 8.68

0.01

Angiogenic

FGF

7.51 ± 0.38

8.68 ± 0.76

0.11

BNIP3

3.96 ± 0.40

0.00 ± 0.00

<0.001

Angiogenic

P53

2.66 ± 0.36

2.32 ± 0.53

0.91

PI3

0.53 ± 0.13

0.00 ± 0.00

<0.02

Angiogenic

SP1

5.31 ± 0.46

4.04 ± 0.60

0.24

STAT3

7.80 ± 0.39

6.31 ± 0.81

0.09

MEF2D

251 ± 4.62

243.10 ± 8.04

0.38

JMY

2.01 ± 0.17

2.60 ± 0.27

0.07

TRAP1

1.57 ± 0.18

4.58 ± 0.32

<0.001

Non-angiogenic

GST

243.30 ± 6.15

244.10 ± 10.91

0.55

CHGA

0.00 ± 0.00

0.34 ± 0.12

>0.05

All data are presented as mean ± standard error

HIF hypoxia-inducible factor; CA9 carbonic anhydrase 9; VEGFA vascular endothelial growth factor A; TYMP thymidine phosphorylase; KDR vascular endothelial growth factor receptor 2; KDRp34 vascular endothelial growth factor p34; FIH factor-inhibiting HIF; PHD prolyl hydroxylasedehydrogenase; DLL4 Delta-like 4; TSP1 thymidine phosphorylase 1; CXCR4 chemokine (C-X-C motif) receptor 4; EPH ephrin; SOD1 superoxide dismutase 1; BCL2 B cell lymphoma 2; FOS FBJ murine osteosarcoma viral oncogene homolog; EGF epidermal growth factor; EGFR epidermal growth factor receptor; BNIP3 BCL2/adenovirus E1B 19 kDa interacting protein 3; PI3 peptidase inhibitor 3; SP1 Sp1 transcription factor; STAT3 signal transducer and activator of transcription 3; LON Lon protease; MEF2D myocyte enhancer factor 2D; JMY junction-mediating and regulatory protein, p53 cofactor; TRAP1 TNF receptor-associated protein 1; GST glutathione S-transferase; NCAM neural cell adhesion molecule; CHGA chromogranin A; SYP synaptophysin

Indicates equal expression

Table 6

Protein biomarker expression in angiogenic and non-angiogenic NSCLCs according to their intracellular localization

Expression status

Intracellular expression localization

Proteins

Equal expression in both tumour types

Nuclear

VEGFA, TYMP, KDR, KDRp34, FIH, EPHB4, PHD1, SOD1, FGF, P53, SP1, STAT3, MEF2D, JMY, GST, CHGA

Cytoplasm

HIF1, HIF2, VEGFA, TYMP, KDR, KDRp34, FIH, PHD1, SOD1, CXCR4, EPHB4, BCL2, EGF, EGFR, FGF, BNIP3, PI3, SP1, LON, MEF2D, GST, CHGA, SYP

Membrane

CA9, MEF2D

Higher expression in angiogenic tumours than in non-angiogenic tumours in at least one subcellular location

Nuclear

HIF1, PHD2, PHD3, CXCR4, BNIP3, PI3, EGF, FOS

Cytoplasm

PHD2, TSP (stroma), DLL4 (endothelium), FOS, STAT3

Membrane

None

Higher expression in non-angiogenic tumours than in angiogenic tumours in at least one subcellular location

Nuclear

EPHB2, EPHB3,TRAP1

Cytoplasm

CA9, PHD3, EPHB2, EPHB3, NCAM, P53, TRAP1, JMY

Membrane

EPHB3, GST

HIF hypoxia-inducible factor; CA9 carbonic anhydrase 9; VEGFA vascular endothelial growth factor A; TYMP thymidine phosphorylase; KDR vascular endothelial growth factor receptor 2; KDRp34 vascular endothelial growth factor p34; FIH factor-inhibiting HIF; PHD prolyl hydroxylase dehydrogenase; DLL4 Delta-like 4; TSP1 thymidine phosphorylase 1; CXCR4 chemokine (C-X-C motif) receptor 4; EPH ephrin; SOD1 superoxide dismutase 1; BCL2 B cell lymphoma 2; FOS FBJ murine osteosarcoma viral oncogene homolog; EGF epidermal growth factor; EGFR epidermal growth factor receptor; BNIP3 BCL2/adenovirus E1B 19 kDa interacting protein 3; PI3 peptidase inhibitor 3; SP1 Sp1 transcription factor; STAT3 signal transducer and activator of transcription 3; LON Lon protease; MEF2D myocyte enhancer factor 2D; JMY junction-mediating and regulatory protein, p53 cofactor; TRAP1 TNF receptor-associated protein 1; GST glutathione S-transferase; NCAM neural cell adhesion molecule; CHGA chromogranin A; SYP synaptophysin

In Table 7, a selection of pathways shows that proteins expressed in both types of tumours are associated with angiogenesis-, VEGF-, and oxidative stress-related pathways. The non-angiogenic tumours are distinguished from the angiogenic tumours by having higher levels of proteins related to ephrin pathways, response to hypoxia, HIF1 regulation by oxygen, and transcription of HIF-controlled genes but lower levels of proteins involved in stromal cell–cell signaling and adhesion, integrins, and Delta-Notch- and EGF-related signaling.
Table 7

Selection of visualized pathways

Proteins equally expressed in angiogenic and non-angiogenictumours

Proteins up-regulated in angiogenic tumours

Proteins up-regulated in non-angiogenic tumours

Pathway

Databases

Pathway

Databases

Pathway

Databases

Focal adhesion

KEGG, WikiPathways

Focal adhesion

KEGG, WikiPathways

VEGF-related

KEGG, Panther, Reactome

Angiogenesis

Panther, WikiPathways

Angiogenesis

Panther

Angiogenesis

Panther

Itegrin signalling

Panther

Oxidative stress

Panther, WikiPathways

Cell response to stress

Reactome

Delta-Notch

WikiPathways

Collagen-related pathways

Reactome

Cell–cell communication

Reactome

Integrin signaling

Panther

  

EGF-related pathways

WikiPathways, Panther

Response to Hypoxia

Reactome

Ephrins signaling

Reactome

Ephrins-related pathways

Reactome

Regulation of HIF by oxygen

Reactome

Regulation of genes by HIF

Reactome

Nitrogen metabolism

KEGG

Gene ontology analysis confirmed that proteins usually associated with angiogensis were present in both types of tumours, whereas higher levels of proteins associated with Notch, extracellular matrix, and cell adhesion were present in the truly angiogenesis-producing tumours. However, in the tumours that grow in a non-angiogenic fashion, proteins associated with mitochondria, cell biogenesis, carbonic dehydratase, ephrins, and axon guidance-related functions were more commonly detected (Table 8).
Table 8

Selection of visualized ontologies

Proteins equally expressed in angiogenic and non-angiogenic tumours

Proteins up-regulated in angiogenic tumours

Proteins up-regulated in non-angiogenic tumours

Ontology

GO-ontology database

Ontology

GO-ontology database

Ontology

GO-ontology database

Cell migration/sprouting angiogenesis

GO.0002042

Biological process

Immune response-activating signals

GO.0002757

Biological process

Regulation of synapses

GO.0051965

Biological process

Endothelial cell migration

GO.0043534

Biological process

Immune response-regulating cell signaling

GO.0002768

Biological process

Axon guidance and neuronal regulation

GO.0031290

GO.0007413

GO.0021952

GO.0021955

GO.0008038

Biological process

VEGF signaling pathways

GO.0038084

Biological process

Mesodermal cell differentiation

GO.0048333

Biological process

Activation of immune response

GO.0002253

Biological process

Positive regulation cell biogenesis

GO.0044089

Biological process

Basement membrane

GO.0005604

Cell component

Extracellular matrix part

GO.0044420

Cell component

Mitochondrial intermembrane space and matrix

GO.0005758

GO.0005759

Cell component

Complex involve in cell adhesion

GO.0098636

Cell component

PDGF receptor-binding

GO.0005161

Molecular function

Fibronectin- and extracellular matrix-binding

GO.0001968

GO.0050840

Molecular function

Ephrin receptor activity

GO.0005003

Molecular function

Extracellular matrix-binding

GO.0050840

Molecular function

Notch-binding

GO.0005112

Molecular function

Carbonate dehydratase activity

GO.0004089

Molecular function

Axon guidance

GO.0008046

Molecular function

Sequencing

The higher level of p53 expression in the cytoplasm of the non-angiogenic tumours, as compared with the angiogenic tumours, was one of the most striking results (Table 9) and raised the question of whether non-angiogenic cells had wild-type p53 or a different set of mutations from the angiogenic tumours. Our results showed that p53mutations were observed in eight of 25 angiogenic cases (32.0%) but in five of seven non-angiogenic cases (71.4%) (P = 0.091 by Fisher’s two-tailed exact test) (Table 9).
Table 9

Summary of p53 mutations detected in non-angiogenic and angiogenic NSCLC cases

Sample ID

Sample type

Mutation location

Mutation type

Domain

Predicted effect on protein activity

104

Non-angiogenic

c.761T>TA; p.I254S

Missense

HCD IV

Damaging

105

Non-angiogenic

c.734G>GA; p.G245D

Missense

HCD IV

Damaging

121

Non-angiogenic

c.488A>AG; p.Y163C

Missense

DNA binding

Damaging

152

Non-angiogenic

c.634T>TG; p.F212 V

Missense

DNA binding

Benign

249

Non-angiogenic

c.314G>GA; p.G105D

Missense

DNA binding

Damaging

133

Angiogenic

c.511G>GT; p.E171X

Nonsense (truncated protein)

HCD III

Truncating

138

Angiogenic

c.824G>GA; p.C275Y

Missense

HCD IV

Damaging

139

Angiogenic

c.het_del216C; p.V73 Wfs48X

Frameshift (truncated protein)

Prolinerich

Truncating

141

Angiogenic

c.407A>AC; p.Q136P

Missense

HCD II

Damaging

98

Angiogenic

c.524G>GA; p.R175H

Missense

HCD III

Potentially damaging

147

Angiogenic

c.524G>GA; p.R175H

Missense

274

Angiogenic

c.471_472TC>GA; p.V157G

Missense

DNA binding

Damaging

Affected domains are listed. The software used to predict the functional effect of the detected sequence changes was PolyPhen_2 (http://genetics.bwh.harvard.edu/ggi/pph2/c2ea64efde6f039a5ca76a2a264ae4f3cf922360/1121012.html )

HCD highly conserved domain

All of the detected mutations were heterozygous, and almost all of them corresponded to hot spots previously reported in different tumour types (http://genetics.bwh.harvard.edu/ggi/pph2/c2ea64efde6f039a5ca76a2a264ae4f3cf922360/1121012.html). The locations of the mutations were randomly distributed across the sequenced region. No specific pattern of mutation location seemed to be related to tumour subtype, angiogenic or non-angiogenic (Fig. 1 and Table 9).
Fig. 1

Localization of the mutations detected on the p53 gene. The sequenced region is indicated with an orange line in the figure above. Mutation locations are indicated with arrowheads (purple mutations found in angiogenic samples; gray: mutations found in non-angiogenic samples). Human P53 protein (HP53) can be divided into five domains, each corresponding to specific functions: yellow is the highly conserved domain I (HCD I)/transactivation domain; red is the second transactivation domain, which is proline-rich; blue is the DNA-binding domain essential for p53-DNA interactions that also contains HCD II-V and is the target of 90% of the p53 mutations found in human cancers; green is the nuclear export signal (NES) localized in the oligomerization domain of p53

Discussion

As initially suggested by our mRNA profiling work [19] and also confirmed by the immunohistochemical data presented here, we failed to reveal significant differences between angiogenic and non-angiogenic NSCLCs as far as the expression of proteins associated with the classic hypoxia/angiogenesis pathway is concerned. Because neovascularisation is found in some tumours but not in others, we suggest that activation of the classical angiogenic pathways is necessary, but not sufficient, to induce the sprouting of new vessels in cancer.

The higher levels of expression of proteins associated with extracellular matrix, cell adhesion, and inflammation were in agreement with the observed presence of tumour-associated stroma and chronic inflammation in many angiogenic tumours [18, 22]. mRNA profiling of these tumours has equally shown that stromal remodeling, cell adhesion, and inflammation are enhanced in angiogenesis [19]. Both mRNA and immunohistochemical data suggest a crucial role in angiogenic cancers for FBJ murine osteosarcoma viral oncogene homolog (FOS), a protein involved in cell proliferation, remodeling, and inflammation. The question remains of whether tissue remodeling is a consequence or a cause of the triggering of angiogenesis. Because non-angiogenic tumours usually preserve the pre-existing architecture, we hypothesize that the triggering of tissue destruction could be a secondary step necessary for the activation of angiogenesis. Nonetheless, this process would be far from a general rule because we have observed, in a subset of NSCLC, that angiogenesis can occur in the absence of tissue destruction [23]. Clearly, other mechanisms must exist.

If classical angiogenesis pathways are similarly active in both angiogenic and non-angiogenic tumours, how does their biology differ? Our previous transcriptional profiling work demonstrated higher levels of mRNA coding for molecules associated with oxidative phosphorylation and mitochondrial biogenesis in non-angiogenic tumours [19]. In the present study, we did not examine components of the oxidative phosphorylation pathway, although we did investigate the expression of some proteins involved in mitochondrial functions. Our results showed that proteins related to mitochondria and to cell biogenesis-promoting processes were more highly expressed in non-angiogenic tumours than in angiogenic tumours. This feature was consistent with higher mitochondrial regulatory activity and a possible metabolic switch in non-angiogenic tumours. Again, whether this switch is a cause or an effect of HIF activation remains unclear.

Interestingly, our data suggested increased involvement of response to hypoxia and to HIF regulation by oxygen in non-angiogenic tumours. We hence speculate that, in these tumours, although the HIF pathway failed to induce new vessel formation, it could well be involved in metabolic reprogramming.

Finally, we noted markedly higher levels of cytoplasmic P53 expression in non-angiogenic tumours than in angiogenic tumours via immunohistochemistry. In a pilot study, we sequenced p53 in a limited number of angiogenic and non-angiogenic NSCLCs. Non-angiogenic tumours had a higher incidence of mutations, which were all missense mutations, whereas angiogenic tumours had an amalgam of frame shifts and missense and nonsense mutations. Because p53 also affects mitochondrial respiration [24], it will be necessary to investigate further how the observed p53 mutations could functionally affect its ability to regulate respiration and/or angiogenesis.

Conclusions

On the basis of our observations collected so far, from the mRNA profiling, immunohistochemical and histopathologic data, we conclude that all tumours, angiogenic and non-angiogenic, experience hypoxia/HIF and VEGF pathway activation. However, angiogenesis does not always ensue. Based on these findings, we suggest that in non-angiogenic tumours, HIF pathway activation could be the driving force toward metabolic reprogramming.

Declarations

Authors’ contributions

OA designed the study with FP, contributed most of the writing of the manuscript and the laboratory work; RL helped with immunohistochemical staining and data analysis; MFM provided training and technical support for sequencing; JH, CS, KCG, and ALH helped with reviewing the literature and the writing of the manuscript; FP designed the study and contributed to most of the writing along with OA. All authors read and approved the final manuscript.

Acknowledgements

OA was supported by a scholarship from the Pathological Society of Great Britain and Ireland. MFM was supported by Leukaemia and Lymphoma Research program (UK) and Associacion Espanola Contra el Cancer (AECC, Spain).

Competing interests

The authors declare that they have no competing interests.

Ethical approval

C02.216 The pathobiology of neoplasia in human tissues.

Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Authors’ Affiliations

(1)
Radcliffe Department of Medicine, Nuffield Division of Laboratory Science, John Radcliffe Hospital, University of Oxford
(2)
Radcliffe Department of Medicine, Leukaemia and Lymphoma Research Molecular Haematology Unit, Nuffield Division of Laboratory Science, John Radcliffe Hospital
(3)
Biodonostia Research Institute, Oncology Area
(4)
Department of Medical Oncology, Molecular Oncology Laboratories, Weatherall Institute of Molecular Medicine, John Radcliffe Hospital

References

  1. Warburg O. The metabolism of tumours in the body. J Gen Physiol. 1927;8:519–30.PubMed CentralView ArticlePubMedGoogle Scholar
  2. Dupuy F, Tabaries S, Andrzejewski S, Dong Z, Blagih J, Annis MG, et al. PDK1-dependent metabolic reprogramming dictates metastatic potential in breast cancer. Cell Metab. 2015;22:1–13.View ArticleGoogle Scholar
  3. Warburg O. On the origin of cancer cells. Science. 1956;123(3191):309–14.View ArticlePubMedGoogle Scholar
  4. Pollard PJ, Wortham NC, Tomlinson IP. The TCA cycle and tumorigenesis: the examples of fumarate hydratase and succinate dehydrogenase. Ann Med. 2003;35(8):632–9.View ArticlePubMedGoogle Scholar
  5. Gatenby RA, Gillies RJ. Why do cancers have high aerobic glycolysis? Nat Rev Cancer. 2004;4(11):891–9. doi:https://doi.org/10.1038/nrc1478.View ArticlePubMedGoogle Scholar
  6. Frezza C, Gottlieb E. Mitochondria in cancer: not just innocent bystanders. Semin Cancer Biol. 2009;19(1):4–11. doi:https://doi.org/10.1016/j.semcancer.2008.11.008.View ArticlePubMedGoogle Scholar
  7. Vander Heiden MG, Cantley LC, Thompson CB. Understanding the Warburg effect: the metabolic requirements of cell proliferation. Science. 2009;324(5930):1029–33. doi:https://doi.org/10.1126/science.1160809.PubMed CentralView ArticlePubMedGoogle Scholar
  8. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell. 2011;144(5):646–74. doi:https://doi.org/10.1016/j.cell.2011.02.013.View ArticlePubMedGoogle Scholar
  9. Harris AL. Hypoxia–a key regulatory factor in tumour growth. Nat Rev Cancer. 2002;2(1):38–47. doi:https://doi.org/10.1038/nrc704.View ArticlePubMedGoogle Scholar
  10. Denko NC. Hypoxia, HIF1 and glucose metabolism in the solid tumour. Nat Rev Cancer. 2008;8(9):705–13. doi:https://doi.org/10.1038/nrc2468.View ArticlePubMedGoogle Scholar
  11. Kerbel RS. Tumor angiogenesis. N Engl JMed. 2008;358(19):2039–49. doi:https://doi.org/10.1056/NEJMra0706596.View ArticleGoogle Scholar
  12. Jose C, Bellance N, Rossignol R. Choosing between glycolysis and oxidative phosphorylation: a tumor’s dilemma? Biochim Biophys Acta. 2011;1807(6):552–61. doi:https://doi.org/10.1016/j.bbabio.2010.10.012.View ArticlePubMedGoogle Scholar
  13. Schulze A, Harris AL. How cancer metabolism is tuned for proliferation and vulnerable to disruption. Nature. 2012;491(7424):364–73. doi:https://doi.org/10.1038/nature11706.View ArticlePubMedGoogle Scholar
  14. Yuneva MO, Fan TW, Allen TD, Higashi RM, Ferraris DV, Tsukamoto T, et al. The metabolic profile of tumors depends on both the responsible genetic lesion and tissue type. Cell Metab. 2012;15(2):157–70. doi:https://doi.org/10.1016/j.cmet.2011.12.015.PubMed CentralView ArticlePubMedGoogle Scholar
  15. Obre E, Rossignol R. Emerging concepts in bioenergetics and cancer research: metabolic flexibility, coupling, symbiosis, switch, oxidative tumors, metabolic remodeling, signaling and bioenergetic therapy. Int J Biochem Cell Biol. 2015;59:167–81. doi:https://doi.org/10.1016/j.biocel.2014.12.008.View ArticlePubMedGoogle Scholar
  16. Folkman J. Tumor angiogenesis: therapeutic implications. N Engl J Med. 1971;285(21):1182–6. doi:https://doi.org/10.1056/NEJM197111182852108.View ArticlePubMedGoogle Scholar
  17. Pezzella F, Di Bacco A, Andreola S, Nicholson AG, Pastorino U, Harris AL. Angiogenesis in primary lung cancer and lung secondaries. Eur J Cancer. 1996;32A(14):2494–500.View ArticlePubMedGoogle Scholar
  18. Pezzella F, Pastorino U, Tagliabue E, Andreola S, Sozzi G, Gasparini G, et al. Non-small-cell lung carcinoma tumor growth without morphological evidence of neo-angiogenesis. Am J Pathol. 1997;151(5):1417–23.PubMed CentralPubMedGoogle Scholar
  19. Hu J, Bianchi F, Ferguson M, Cesario A, Margaritora S, Granone P, et al. Gene expression signature for angiogenic and nonangiogenic non-small-cell lung cancer. Oncogene. 2005;24(7):1212–9. doi:https://doi.org/10.1038/sj.onc.1208242.View ArticlePubMedGoogle Scholar
  20. Bates GJ, Fox SB, Han C, Launchbury R, Leek RD, Harris AL, et al. Expression of the forkhead transcription factor FOXP1 is associated with that of estrogen receptor-beta in primary invasive breast carcinomas. Breast Cancer Res Treat. 2008;111(3):453–9. doi:https://doi.org/10.1007/s10549-007-9812-4.View ArticlePubMedGoogle Scholar
  21. Chen EY, Tan CM, Kou Y, Duan Q, Wang Z, Meirelles GV, et al. Enrichr: interactive and collaborative HTML5 gene list enrichment analysis tool. BMC Bioinformatics. 2013;14:128. doi:https://doi.org/10.1186/1471-2105-14-128.PubMed CentralView ArticlePubMedGoogle Scholar
  22. Ferguson M. Angiogenesis in human lung tumours. Oxford: Oxford University; 2008.Google Scholar
  23. Passalidou E, Trivella M, Singh N, Ferguson M, Hu J, Cesario A, et al. Vascular phenotype in angiogenic and non-angiogenic lung non-small cell carcinomas. Br J Cancer. 2002;86(2):244–9. doi:https://doi.org/10.1038/sj.bjc.6600015.PubMed CentralView ArticlePubMedGoogle Scholar
  24. Matoba S, Kang J, Patino W, Wragg A, Boehm M, Gavrilova O, et al. p53 regulates mitochondrial respiration. Science. 2006;312(5780):1650–3.View ArticlePubMedGoogle Scholar

Copyright

© Adighibe et al. 2016

Advertisement