Biomarkers of angiogenesis and their role in the development of VEGF inhibitors.

Abstract

Vascular endothelial growth factor (VEGF) has been confirmed as an important therapeutic target in randomised clinical trials in multiple disease settings. However, the extent to which individual patients benefit from VEGF inhibitors is unclear. If we are to optimise the use of these drugs or develop combination regimens that build on this efficacy, it is critical to identify those patients who are likely to benefit, particularly as these agents can be toxic and are expensive. To this end, biomarkers have been evaluated in tissue, in circulation and by imaging. Consistent drug-induced increases in plasma VEGF-A and blood pressure, as well as reductions in soluble VEGF-R2 and dynamic contrast-enhanced MRI parameters have been reported. In some clinical trials, biomarker changes were statistically significant and associated with clinical end points, but there is considerable heterogeneity between studies that are to some extent attributable to methodological issues. On the basis of observations with these biomarkers, it is now appropriate to conduct detailed prospective studies to define a suite of predictive, pharmacodynamic and surrogate response biomarkers that identify those patients most likely to benefit from and monitor their response to this novel class of drugs.

Full text

Page 1

Minireview
Biomarkers of angiogenesis and their role in the development
of VEGF inhibitors
N Murukesh1, C Dive2and GC Jayson*,1
1Department of Medical Oncology, Manchester Academic Health Science Centre, The Christie NHS Foundation Trust, Cancer Research UK and
University of Manchester, Wilmslow Road, Withington, Manchester M20 4BX, UK;2Cancer Research UK and Clinical and Experimental Pharmacology
Group, Manchester Cancer Research Centre, Paterson Institute of Cancer Research, Withington, Manchester M20 4BX, UK
Vascular endothelial growth factor (VEGF) has been confirmed as an important therapeutic target in randomised clinical trials in
multiple disease settings. However, the extent to which individual patients benefit from VEGF inhibitors is unclear. If we are to
optimise the use of these drugs or develop combination regimens that build on this efficacy, it is critical to identify those patients who
are likely to benefit, particularly as these agents can be toxic and are expensive. To this end, biomarkers have been evaluated in tissue,
in circulation and by imaging. Consistent drug-induced increases in plasma VEGF-A and blood pressure, as well as reductions in
soluble VEGF-R2 and dynamic contrast-enhanced MRI parameters have been reported. In some clinical trials, biomarker changes
were statistically significant and associated with clinical end points, but there is considerable heterogeneity between studies that are to
some extent attributable to methodological issues. On the basis of observations with these biomarkers, it is now appropriate to
conduct detailed prospective studies to define a suite of predictive, pharmacodynamic and surrogate response biomarkers that
identify those patients most likely to benefit from and monitor their response to this novel class of drugs.
British Journal of Cancer (2010) 102, 8–18. doi:10.1038/sj.bjc.6605483
Published online 15 December 2009
& 2010 Cancer Research UK
www.bjcancer.com
Keywords: VEGF; angiogenesis; biomarkers
? ?????????????????????????????????????????????
Angiogenesis, the process of new blood vessel formation, is critical
for the growth and metastasis of tumours. Early in tumourigenesis,
an ‘angiogenic switch’ is activated by hypoxia, activated oncogenes
and/or metabolic stress. The previously closely maintained
physiological balance that keeps adult vasculature in a relatively
quiescent state is then tipped in favour of angiogenesis through
the expression of pro-angiogenic growth factors, such as vascular
endothelial growth factor (VEGF) (Hanahan and Folkman, 1996).
Vascular endothelial growth factor has been confirmed in multiple
clinical trials as an important target for solid tumour treatment,
but it is not clear who benefits from this class of drugs; this is an
issue of increasing importance, bearing in mind the toxicity and
expense of VEGF inhibitors in addition to the need to generate
combination regimens that include VEGF inhibitors.
Although several biomarkers associated with angiogenesis
measured before treatment have shown to provide prognostic
value, and a few biomarkers are pharmacodynamic, there is
minimal information on biomarkers that are true surrogates for
clinical response to VEGF inhibitors. As VEGF-targeted therapies
progress through clinical development pipelines, it may take
several years to determine their clinical efficacy and overall
response data. Thus, there is a pressing need for predictive and
pharmacodynamic biomarkers and for those that are true
surrogates of clinical response.
VASCULAR ENDOTHELIAL GROWTH FACTOR
Vascular endothelial growth factor is a homodimeric glycoprotein
with a molecular weight of 45kDa. The VEGF family includes
VEGF-A (usually referred to as VEGF), VEGF-B, VEGF-C, VEGF-D
and a structurally related molecule, placental growth factor (PlGF).
Through alternative splicing of VEGF mRNA, 12 isoforms of VEGF
have been identified (Nowak et al, 2008), most of which can activate
the signal transducing receptors. However, through alternative
splicing of exon 8, anti-angiogenic variants, designated as VEGF-
Axxxb, can be formed, where xxx denotes the number of amino acids
in the mature protein. Although present in the malignant colonic
epithelium, the inhibitory effect of VEGF-A165bis overcome by an
excess of VEGF-A165, a relationship that contributes to the relative
efficacy in vivo of anti-VEGF antibodies that bind both isoforms
(Bates et al, 2002; Varey et al, 2008). The clinical significance of the
ratio of activating and inactivating isoforms of VEGF to the activity
of VEGF inhibitors remains unclear.
Three high-affinity VEGF tyrosine kinase receptors have been
identified: VEGF receptor (VEGFR)-1 (flt-1), VEGFR-2 (flt-1/KDR)
and VEGFR-3 (flt-4). The binding of VEGF to these receptors
initiates a cascade of signalling pathways that mediate endothelial
cell (EC) migration, proliferation, survival and permeability.
Additional co-receptors include neuropilins that have traditionally
been implicated in cell guidance (Kawasaki et al, 1999) and
increased binding of VEGF to its signalling receptor (Soker et al,
1998). However, recent data have suggested that NRP-1 may
regulate EC function independently of VEGFR-2 (Murga et al,
2005), and that VEGF121can directly interact with NRP-1 without
forming an NRP-1–VEGFR-2 complex (Pan et al, 2007).
Received 23 June 2009; revised 5 November 2009; accepted 18
November 2009; published online 15 December 2009
*Correspondence: Professor GC Jayson;
E-mail: gordon.jayson@christie.nhs.uk
British Journal of Cancer (2010) 102, 8–18
& 2010 Cancer Research UKAll rights reserved 0007– 0920/10$32.00
www.bjcancer.com

Page 2

Vascular endothelial growth factor-A interacts with both
VEGFR-1 and VEGFR-2 to mediate angiogenesis, whereas VEGF-
B and PlGF have high affinity for only VEGFR-1. Vascular
endothelial growth factor-C and VEGF-D bind both VEGFR-2
and VEGFR-3 (Joukov et al, 1996; Achen et al, 1998) to regulate
angiogenesis and have been implicated in lymphangiogenesis
(Shibuya and Claesson-Welsh, 2006). Vascular endothelial growth
factor receptor-2 is the principal receptor that promotes the
pro-angiogenic action of VEGF-A and has been the principal
target of anti-angiogenic therapies, although additional studies
have underlined the importance of signalling through VEGFR-1
(Carmeliet et al, 2001).
Various strategies for inhibiting VEGF have been investigated
over the last decade. These include neutralising antibodies to
VEGF (Hurwitz et al, 2004), low-molecular-weight VEGFR tyrosine
kinase inhibitors (TKis) (Motzer et al, 2006; Llovet et al, 2008) and
soluble VEGFR constructs (VEGF-Trap) (Riely and Miller, 2007)
(Table 1).
Angiogenesis and VEGF have been confirmed as targets of anti-
cancer therapeutics in multiple disease settings. Randomised
clinical trials in the first- and second-line treatment of metastatic
colorectal cancer (Hurwitz et al, 2004; Giantonio et al, 2007),
breast cancer (Miller et al, 2007), non-small-cell lung cancer
(Sandler et al, 2006), renal cancer (Motzer et al, 2007) and
hepatocellular carcinoma (Llovet et al, 2008) have demonstrated
an improvement in response, progression-free survival (PFS)
and/or overall survival (OS) when conventional therapy was
supplemented by VEGF inhibitors.
The demonstration of a survival advantage conferred by
VEGF inhibitors is of great importance. However, the initial
promise of anti-angiogenic agents, namely the reduced prevalence
of drug resistance and durable stabilisation of disease, has not
been realised. Vascular endothelial growth factor inhibitors also
have a range of host toxicities. They are expensive and optimal
development of combination anti-vascular regimens requires the
identification of those patients most likely to benefit from
treatment with this class of drugs. Despite attempts, this goal has
eluded us. In this study, we review the use of candidate predictive
and/or pharmacodynamic biomarkers pertinent to VEGF inhibi-
tion and highlight approaches that are yet to be investigated.
BIOMARKERS
A biomarker is a characteristic that is objectively measured and
evaluated as an indicator of a normal biological process, a patho-
genic process or of pharmacological response to a thera-
peutic intervention (Atkinson et al, 2001). Several categories of
biomarkers have been described that are pertinent to cancer,
namely screening, diagnostic, prognostic, predictive, pharmaco-
logical (pharmacodynamic, proof of mechanism and of concept),
surrogate response and safety biomarkers. Biomarker assays need
to be carefully validated and be robust, reliable and reproducible
when applied in clinical contexts. As VEGF inhibitors are already
licensed, the most important question to be addressed using
biomarkers is who should be treated with this class of agent; that
is, at this point in the development of VEGF inhibitors, there is a
clear need to identify predictive, pharmacological and surrogate
response biomarkers, and in particular, to discriminate between
these and prognostic biomarkers (which inform on the progres-
sion of disease irrespective of treatment).
A number of confounding issues recur in the literature
regarding the use of biomarkers: Are there enough samples from
sufficiently large trials to detect a statistically significant result?
Have assays been performed within each patient before drug
administration to determine whether a change in a biomarker can
be ascribed to the drug, that is, is baseline variation characterised?
Have biomarker studies been carried out according to the
standards of Good Clinical Laboratory Practice required by the
EU clinical trial directive (2001/20/EC). Such issues are of greater
than theoretical importance, as, for example, we know that
inappropriate blood sample handling can lead to platelet activation
and ex vivo release of PDGF and VEGF. Therefore, debate is
ongoing regarding the optimal choice of specimen for the
measurement of these biomarkers. Serum seems to be a popular
choice; however, the release of the above factors during clotting
can influence the values measured. However, considering the low
sensitivity of ELISAs to detect plasma levels and the proposed
scavenging of VEGF by platelets (George et al, 2000), serum levels
might represent the truer picture. In this study, we discuss trials
that have incorporated circulating molecular and cellular, tissue,
genetic and/or imaging biomarkers that are related to VEGF and
its inhibition. Hypertension is one of the most common toxicities
in patients having VEGF inhibitors and, in this study, we also
examine the differential benefits seen in patients experiencing
hypertension.
CIRCULATING CANDIDATE BIOMARKERS OF
ANGIOGENESIS
The majority of clinical trials that have evaluated VEGF inhibitors
have involved investigation of either anti-VEGF antibodies or
VEGFR TKis. The prototypic VEGF inhibitor is the monoclonal
anti-VEGF antibody, bevacizumab. Table 2a summarises the data
collected from studies of circulating biomarkers in cancer patients
treated with bevacizumab and other antibody-based therapeutics.
One of the first biomarkers to be evaluated was the plasma
concentration of VEGF-A. However, of multiple trials, only the
E4599 trial in non-small-cell lung cancer reported that the pre-
treatment plasma concentration of VEGF was of prognostic
significance (Dowlati et al, 2008). Intuitively, one would predict
that the pre-treatment plasma concentration of VEGF would be
most helpful in diseases that respond to single-agent VEGF
inhibitors (e.g., renal, ovarian and hepatic cancer). However, the
extent to which we can interpret such data is limited by, for
example, studies that have been too small (Siegel et al, 2008) or in
cases in which the limit of quantitation of ELISA was at a
concentration that precluded interpretation of a significant
proportion of biomarker data (Yang et al, 2003).
The initial phase I trials of anti-VEGF antibodies demonstrated
that there was a logarithmic increase in the total plasma
concentration of VEGF after drug administration (Gordon et al,
2001; Jayson et al, 2005). The source of this cytokine is not clear,
but could reflect extensive loading of the extracellular matrix with
VEGF in patients with advanced cancer. Thus, one hypothesis
would be that the magnitude of the anti-VEGF antibody-induced
change in plasma VEGF concentration might predict patient
benefit. However, despite an interesting report in one small study
(Siegel et al, 2008), this has not been confirmed in other trials,
many of which were also compromised by their insufficient size.
The other possibility is that anti-VEGF antibodies form inert
complexes, causing a false increase in circulating VEGF levels. One
study conducted using immunodepleted plasma has supported this
assumption by showing a significant decrease in VEGF levels after
treatment with bevacizumab (Loupakis et al, 2007).
Table 1VEGF-targeted therapies
Mechanism of actionExamples of drugs
Anti-VEGF monoclonal antibody
Small molecule receptor tyrosine kinase inhibitor (TKi)
VEGF-Trap (soluble VEGF receptor)
Bevacizumab
Sunitinib
Aflibercept
Abbreviation: VEGF¼vascular endothelial growth factor.
Biomarkers of angiogenesis and VEGF inhibitors
N Murukesh et al
9
British Journal of Cancer (2010) 102(1), 8–18
& 2010 Cancer Research UK

Page 3

The increase in plasma VEGF concentration in patients treated
with anti-VEGF antibodies has also been seen in those receiving
low-molecular-weight VEGFR TKis (Table 2b). A VEGFR TKi
biomarker signature has emerged, in which the drugs induce an
increase in plasma VEGF and PlGF, as well as reductions in soluble
VEGFR-2 and VEGFR-3. Presumably, this biomarker signature
reflects the larger repertoire of receptors targeted by VEGFR TKis
compared with anti-VEGF antibodies. If true, one might not expect
to see an increase in VEGFR-3 concentrations in patients receiving
bevacizumab, although this has not been formally reported.
Although a principal aim of biomarker studies in patients
receiving VEGF inhibitors is to identify those patients who are
most likely to benefit, it is equally important to detect the onset of
drug resistance and ideally the factors mediating this resistance,
which is an area of increasing importance, given recent data that
indicate the potential value of continuing treatment with VEGF
inhibitors until disease progression (Hurwitz et al, 2004; Saltz et al,
2008). To date, few biomarker studies have identified clinically
tractable mediators of resistance, but some recent data highlighted
FGF-2 and SDF-1a as potential targets (Batchelor et al, 2007).
Circulating ECs (CECs) are believed to arise from vessel walls,
either of mature vessels or the tumour vasculature. A subset of
them is thought to originate from the bone marrow and represents
circulating endothelial progenitor cells (CEPCs) (Lin et al, 2000).
Normal adults have 1–20 CECs per ml in their peripheral blood,
and the levels are shown to increase significantly in patients with
advanced cancer (Rowand et al, 2007). After successful treatment,
their concentration tends to normalise (Willett et al, 2005), in
contrast to the situation in progressive disease (Beerepoot et al,
2004). In patients with breast cancer, one study demonstrated
that the pre-treatment high concentration of CECs was a good
prognostic factor (Dellapasqua et al, 2008), whereas another
showed that in patients receiving metronomic doses of cytotoxic
chemotherapy, an increase in circulating apoptotic CECs was
associated with a better outcome (Mancuso et al, 2006).
Very few studies have evaluated changes in the number of CECs
and CEPCs in the peripheral blood of patients receiving VEGF
inhibitors. In general, CEC concentrations decrease after admin-
istration of VEGF inhibitors (Table 2a and 2b). However, this is
not a consistent observation. When patients with rectal cancer
were treated with bevacizumab, the concentration of CECs reduced
(Willett et al, 2005), whereas in patients with gastrointestinal
stromal tumour treated with sunitinib (a broad spectrum receptor
TKi that inhibits c-kit and VEGFR), there was a transient increase
in CECs that was associated with a better outcome (Norden-Zfoni
et al, 2007).
Only a few studies have enumerated the number of CECs and
CEPCs in the circulation of patients receiving VEGF inhibitors.
However, significant methodological problems have to be over-
come before these biomarkers can be incorporated routinely into
multi-centre trials. To date, repeated pre-treatment samples have
not been collected and therefore confidence intervals for
individuals have not been clearly established, obscuring decisions
with regard to treatment-induced effects. These problems are
Table 2aAnti-VEGF antibodies and circulating biomarkers
References Drug, disease and trial BiomarkersNDrug-induced changesPrognostic and predictive values
Dowlati et al (2008)Carboplatin and Paclitaxel±
Bevacizumab
NSCLC (E4599); Phase 2/3
VEGF
E-selectin
FGF-2
ICAM
VEGF
SDF-1
HUVEC
VEGF
sVEGFR-2
Urine VEGF
160
kE-selectin
mFGF-2
Baseline VEGF predicts response (P¼0.01)
-Low baseline VEGF: better PFS (P¼0.04)
-Low ICAM: better OS (P¼0.00005),1
year survival and high RR (P¼0.02)
mVEGF and SDF-1 on progression
-VEGF and SDF-1 levels correlate with
angiogenic score
NS
NS
Siegel et al (2008)Bevacizumab
Unresectable HCC
Phase 2
Bevacizumab mRCC; Phase 2
Bevacizumab+Erlotinib
Recurrent ovarian cancer
Phase 2
Gemcitabine+Cisplatin+Bevacizumab
Pancreatic cancer; Phase 2
Cyclophosphamide+Bevacizumab
Ovarian cancer; Phase 2
8
kVEGF and SDF-1
kHuVEC angiogenic score
Yang et al (2003)
Nimeiri et al (2008)
113
11
mVEGF
NS
Ko et al (2008)VEGF
FGF-2
VEGF
E-selectin
TSP-1
VEGF
sVCAM-1 sVEGFR-
2
VEGF
FGF-2
46
mVEGF
mFGF-2
mVEGF and kTSP-1
NS
Garcia et al (2008)70NS
Denduluri et al
(2008)
Bevacizumab
Breast cancer
Pilot study
Bevacizumab±IFNa-2b
Malignant melanoma
Phase 2
Octreotide+INFa-2b+
Bevacizumab
NET; Phase 2
HuMV833
Advanced cancer
Phase 1
Cyclophosphamide+Capecitabine+
Bevacizumab
Breast cancer; Phase 2
Chemoradiotherapy+Bevacizumab
Rectal cancer; Phase 1
21
msVCAM-1
msVEGFR-2
NS
Varker et al (2007)
32NSNS
Yao et al (2008)
FGF-2
IL-8
36
kFGF-2
m IL-8
NS
Jayson et al (2005)VEGFR-1, IL-8,
sVCAM-1, FGF-2,
E-selectin, HGF
CEC
CEPC
20
mVEGF
kFGF, HGF
NS
Dellapasqua et al
(2008)
46
kCECHigh baseline CECs correlate with OR
(P¼0.02), clinical benefit (P¼0.01) and
improved PFS (P¼0.04)
NSWillett et al (2005)CEC
CEPC
6
kCEC/CEPC
Abbreviations: VEGF¼vascular endothelial growth factor; CEC¼circulating endothelial cell; CEPC¼circulating endothelial progenitor cell; FGF-2¼fibroblast growth factor-2;
HCC¼hepatocellular carcinoma; HGF¼hepatocyte growth factor; HUVEC¼human umbilical vein endothelial cell; ICAM¼intercellular adhesion molecule; INFa-
2b¼interferon a-2b; mRCC¼metastatic renal cell carcinoma; NET¼neuroendocrine tumour; NS¼not significant; NSCLC¼non-small-cell lung cancer; OR¼overall
response; OS¼overall survival; RR¼response rate; SDF-1¼stromal cell-derived factor-1; sVCAM¼soluble vascular cell adhesion molecule; sVEGFR-2¼soluble VEGF
receptor 2; TSP-1¼thrombospondin-1.
Biomarkers of angiogenesis and VEGF inhibitors
N Murukesh et al
10
British Journal of Cancer (2010) 102(1), 8–18
& 2010 Cancer Research UK

Page 4

Table 2b
VEGF RTKi and soluble biomarkers
References
Drug, disease and trial
Biomarkers
N
Drug-induced changes
Prognostic and predictive values
Rini et al (2008a)
Sunitinib
Bevacizumab refractory RCC
Phase 2
VEGF-A, VEGF-C, sVEGFR-3
PlGF
61
mVEGF-A and PlGF
kVEGF-C and sVEGFR-3
Low baseline VEGF-C (P¼0.0006) and VEGF–R-3
(P¼0.006) associated with longer PFS and ORR
Burstein et al (2008)
Sunitinib
Metastatic breast cancer
VEGF, sVEGFR-2, sVEGFR-3, sKIT
64
mVEGF, kVEGFR-2, VEGFR-3
kVEGFR-3 and mOS (P¼0.07)
ksKIT and mTTP and mOS (P¼0.0194)
Bello et al (2006)
Sunitinib
Neuroendocrine tumour
VEGF, IL-8, sVEGFR-2, sVEGFR-3
109
mVEGF
kVEGFR-2, ksVEGFR-3
kVEGFR-3 with PR (n¼11)
mIl-8 with decreased tumour size
Norden-Zfoni et al (2007) Sunitinib (SU11248)
Imatinib-resistant GIST
Phase1/2
VEGF, sVEGFR-2
73
mVEGF
ksVEGFR-2
NS
Azad et al (2008)
Sorafenib+Bevacizumab
Advanced cancer
Phase 1
VEGF
IL-6IL-8
28
mVEGF
No change: IL-6, IL-8
NS
Batchelor et al (2007)
Cediranib (AZD2171)
Glioblastoma
Phase 2
VEGF, PlGF, sVEGFR-1, sVEGFR-2, FGF-2, SDF-1a,
16
mVEGF, mPlGF
ksVEGFR-2
PD-associated variables: kPlGF, msVEGFR2, mbFGF,
mSDF-1a
Drevs et al (2007)
Cediranib (AZD2171)
Advanced cancer
Phase 1
sVEGFR-1sVEGFR-2
FGF-2, PlGF,Tie-2, E-selectin, IL- 8
83
ksVEGFR-2
mVEGF, mPlGF
NS
Rosen et al (2007)
AMG 706
Advanced cancer
Phase 1
PlGF, VEGF, FGF-2, sVEGFR-1
71
mPlGF
ksVEGFR-2
mPlGF (P¼0.003) and kVEGFR-2 (P¼0.001)
associated with change in tumour size
Jonker et al (2007)
Brivanib (BMS-582664)
Advanced cancer; Phase 1
sVEGFR-2, Collagen IV
50
ksVEGFR-2
kcollagen IV
NS
Heymach et al (2008)
Vandetanib (AZD6474)
NSCLC: 3; Phase 2 trials
VEGF
251
NS
Low baseline VEGF: low risk of disease progression
Kiura et al (2008)
Vandetanib (AZD6474)
NSCLC; Phase 2
VEGF, sVEGFR-2, Tie-2
53
mVEGF
ksVEGFR-2
Low baseline VEGF and TTP Low baseline VEGF in patients with PR
Yamada et al (2008)
E7080
Advanced cancer; Phase 1
VEGF, FGF-2
PDGF
24
mVEGF
NS
Drevs et al (2005)
Vatalanib (PTK/ZK)
Advanced cancer, liver mets; Phase1/2
VEGF, FGF-2, sTie-2, and E-selectin
30
mVEGF, FGF-2
Change in VEGF correlates with outcome
(P¼0.027)
Heymach et al (2004)
Semaxinib (SU5416)
Soft tissue sarcoma
Phase 2
Urine VEGF Urine FGF-2
13
mVEGF, FGF-2
High baseline VEGF associated with kOS (P¼0.04)
Lara et al (2003)
IFNa+Semaxinib (SU5416)
RCC; Phase 2
VEGF
PAI-1
25
kVEGF, kPAI-1
kPAI-1 levels in patients with SD
kbaseline VEGF in pts with PD
Stopeck et al (2002)
Semaxinib (SU5416)
Advanced cancer; Phase 1
Urine VEGF Urine FGF-2
22
mUrine VEGF
mUrine FGF-2
mUrine VEGF in responders (Po0.05)
Fury et al (2007)
Semaxinib (SU5416)
Head and neck cancer; Phase 2
VEGF
35
mVEGF
NS
Peterson et al (2004)
Semaxinib (SU5416)
Melanoma; Phase 2
VEGF
13
mVEGF
NS
Stadler et al (2004)
Semaxinib (SU5416)
Homone refractory prostate cancer; Phase 2
VEGF
bFGF
22
NA
NS
Norden-Zfoni et al (2007) Sunitinib (SU11248)
Imatin-resistant GIST
Phase1/2
CEC
PBMC
90
mCECs
kPBMC
Significant clinical correlation (P¼0.03)
Batchelor et al (2007)
Cediranib (AZD2171)
Glioblastoma
Phase 2
CEC CEPCs
16
kCEC
kCEPC
mCECs with tumour progression.
mCEPCs with relapse after drug holidays
Abbreviations: VEGF¼vascular endothelial growth factor; RTKi¼receptor tyrosine kinase inhibitor; CEC¼circulating endothelial cell; CEPC¼circulating endothelial progenitor cell; FGF-2¼fibroblast growth factor-2;
GIST¼gastrointestinal stromal tumour; HGF¼hepatocyte growth factor; RCC¼renal cell carcinoma; NA¼not applicable; NS¼not significant; NSCLC¼non-small-cell lung cancer; OR¼overall response; ORR¼objective
response rate; OS¼overall survival; PAI-1¼plasminogen activator inhibitor-1; PBMC¼peripheral blood mononuclear cells; PD¼progressive disease; PDGF¼platelet-derived growth factor; PFS¼progression-free survival
PlGF¼placental growth factor; PR¼partial response; RR¼response rate; SD¼stable disease; SDF-1¼stromal-cell derived factor-1; sVEGFR¼soluble VEGF receptor; TTP¼time to progression.
Biomarkers of angiogenesis and VEGF inhibitors
N Murukesh et al
11
British Journal of Cancer (2010) 102(1), 8–18
& 2010 Cancer Research UK

Page 5

compounded by multiple platforms and antigen-selection suites
(Mutin et al, 1999; Mancuso et al, 2001; Rowand et al, 2007;
Dellapasqua et al, 2008) used to characterise CECs/CEPCs. The
common technique used for analysis is immunomagnetic bead
isolation or immunophenotyping using the flow cytometer. How-
ever, there is a lack of consensus regarding the choice of markers to
be used. Real-time PCR using CD146 mRNA has been proposed as
an alternative method for enumerating CECs (Furstenberger et al,
2005). At present, no markers are considered specific for CECs
and a clearer antigen profile is required for isolating these cells.
In summary, the studies in Tables 2a and 2b identify a bio-
marker signature observed in patients treated with VEGF inhibi-
tors. This includes an increase in VEGF (with or without VEGF-C),
a decrease in VEGFR-2 (and sometimes in VEGFR-3) and, in some
studies, a decrease in CECs. Occasionally, these biomarker changes
have been of prognostic significance but none have been qualified
as having predictive value. Although it is possible that the
predictive potential of these biomarkers has not been tested in
appropriately designed studies, it is important to note that many of
the early studies focused on drugs with relatively high IC50
(e.g., semaxanib, SU5416), which therefore were less potent and
perhaps less effective than the VEGFR TKis currently in the clinic,
thereby reducing the chances of measuring a change in biomarker
concentration on drugs that had more rapid clearance (e.g.,
vatalanib, PTK/ZK) or on trials that were too small to generate
statistically significant results.
IMAGING BIOMARKERS FOR VEGF INHIBITORS
Conventional radiological reporting systems for new drugs rely on
one-dimensional (Response Evaluation Criteria in Solid Tumours)
or two-dimensional (WHO criteria) response-assessment schemes.
Neither is well suited to the assessment of anti-angiogenic agents,
the principal effect of which is tumour cytostasis. Thus, early
clinical trials of VEGF inhibitors sought pharmacological proof of
concept by examining changes in the tumour vasculature,
predominantly through the use of MRI, which is a technology
that is non-invasive, sensitive and avoids ionising radiation.
Of all the biomarkers that have been tested in trials of VEGF
inhibitors, the most consistent findings have been achieved with
dynamic contrast-enhanced MRI (DCE-MRI), in keeping with the
proposed mechanism of action of the drugs (Tables 3a and 3b).
Transfer constants such as Ktrans, a composite of the vascular
permeability and endothelial surface area, are reduced in patients
receiving VEGF inhibitors, and multiple studies have demon-
strated a dose level–response relationship. In addition, a second
relationship, the correlation between the magnitude of reduction
in transfer constants and the attainment of stable or better disease,
has been widely reported (Morgan et al, 2003; Mross et al, 2005;
Thomas et al, 2005; Hahn et al, 2008). Although many of these
studies were small and confounded by inter-patient heterogeneity;
generally data show that patients whose tumours undergo at least a
50% reduction in DCE-MRI parameters attain stable or better
disease. Thus, DCE-MRI perhaps holds the greatest promise as a
prognostic and/or predictive biomarker for VEGF inhibitors, and
recent data have highlighted the potential of another DCE-MRI-
derived parameter, vp(the fractional plasma volume), as a further
candidate biomarker that may show clinical utility in trials of
VEGF inhibitors (Hahn et al, 2008).
Dynamic contrast-enhanced MRI is more complex than com-
puted tomography (CT) in terms of the ease with which it can be
incorporated into multi-site studies; for this reason, many centres
are testing the relationship between dynamic CT and DCE-MRI. On
the basis of prognostic data gathered from analysis of the tumour
vasculature seen in CT scans of patients with advanced ovarian
cancer treated with conventional cytotoxic therapy (O’Connor et al,
2007), together with ongoing comparisons between dynamic CT
and DCE-MRI, it is likely that there will be an expansion in research
to assess whether these techniques can serve as predictive
biomarkers in patients treated with VEGF inhibitors.
Recent interest in MRI techniques that do not require contrast
has highlighted blood oxygenation level-dependent (BOLD)
imaging and arterial spin labelling (ASL) (de Bazelaire et al,
2008). Arterial spin labelling, a technique in which protons
entering the zone of interest are magnetised, was developed for
imaging the vasculature of the brain. Although initial results with
ASL in bodies of patients treated with VEGF inhibitors show
promise as a prognostic biomarker (de Bazelaire et al, 2008), ASL
is technically challenging when used to image the body and usually
requires 3-T MRI machines. Blood oxygenation level-dependent
imaging, a technique that relies on the paramagnetic effects of
deoxyhaemoglobin, can be used to provide information on the
oxygenation status of the patient’s tumour, and in particular the
oxygen status in tumour vessels. However, although anatomical
resolution is good, the signal-to-noise ratio is relatively low and,
although this can be increased by administering carbogen, such a
procedure can be unpleasant for the patient (Padhani et al, 2007).
Arterial spin labelling and BOLD are both attractive techniques,
but have not been fully evaluated as biomarkers for VEGF
inhibitors and for now are likely to remain confined to specialist
imaging centres.
Hypoxia is a key mediator of angiogenesis, at least in part
because it induces the expression of VEGF. However, there have
been very few attempts to use hypoxia-imaging strategies as
biomarkers for VEGF. In addition to BOLD, which has not been
used to evaluate VEGF inhibitors in the clinic, positron emission
tomography (PET) imaging tracers have been used to image
hypoxia in the clinic. Both18F-MISO, which is retained in hypoxic
cells through electron transfer that prevents egress from cells, and
60Cu-ATSM, which is retained through mechanisms that are
unclear but depend on the redox state of cells, have been used to
image hypoxia (Padhani et al, 2007). No studies have been
reported to date on the imaging of hypoxia using these tracers in
Table 3a Antibody-based VEGFi and imaging biomarkers
References Drug, disease and trialDCE-MRI biomarkersN Drug-induced changes Prognostic and predictive values
Overmoyer et al (2004)Docetaxel±Bevacizumab
Breast cancer; Phase 2
Bevacizumab
Breast cancer; Phase 2
HuMV833 (Anti-VEGF)
Advanced cancer; Phase 1
CDP791 (Anti-VEGFR-2)
Advanced Cancer; Phase 1
kep
26
kkep
NS
Wedam et al (2006)Ktrans, ve
20
kKtrans
NS
Jayson et al (2005)KtranskeprBV 20
kkep
NS
Ton et al (2007)Ktrans
31 No DCE-MRI change
Dose-related volumetric change
NS
Abbreviations: VEGF¼vascular endothelial growth factor; DCE-MRI¼dynamic contrast-enhanced magnetic resonance imaging; kep¼rate constant; Ktrans¼bi-directional
transfer coefficient; rBV¼regional blood volume; ve¼volume of the extravascular extracellular space (EES).
Biomarkers of angiogenesis and VEGF inhibitors
N Murukesh et al
12
British Journal of Cancer (2010) 102(1), 8–18
& 2010 Cancer Research UK