Tumour-Stromal Architecture Can Define the Intrinsic Tumour Response to VEGF-Targeted

Author Manuscript Published OnlineFirst on September 12, 2013; DOI: 10.1158/1078-0432.CCR-13-1637 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Tumour-stromal architecture can define the intrinsic tumour response to VEGF-targeted

therapy

Neil R. Smith1,3, Dawn Baker1, Matthew Farren1, Aurelien Pommier1, Ruth Swann1, Xin Wang1,

Sunita Mistry1, Karen McDaid1, Jane Kendrew1, Chris Womack1, Stephen R. Wedge1,2 and

Simon T. Barry1,3

1Oncology Innovative Medicines, AstraZeneca, Alderley Park, Macclesfield, Cheshire SK10

4TG, UK; 2Northern Institute for Cancer Research, Newcastle University, Newcastle upon Tyne,

NE2 4HH, UK.

Running title: Tumour-stromal phenotypes define VEGF sensitivity

Key words: Angiogenesis, VEGF-A, VEGFR-2

3To whom requests for reprints should be addressed at: Oncology iMed, AstraZeneca,

Mereside, Alderley Park, Macclesfield, Cheshire SK10 4TG, UK.

E-mail: [email protected] and [email protected]

Abbreviations: VEGFi, Vascular Endothelial Growth Factor inhibitor; PDGF, Platelet Derived

Growth Factor; TV, tumour vessel; SV, stromal vessel; TMA, tissue microarray; MVD,

microvessel density

Conflict of interest statement:

All authors are current or former AstraZeneca employees and shareholders. There are no other

conflicts of interest to disclose.

Downloaded from clincancerres.aacrjournals.org on September 24, 2021. © 2013 American Association for Cancer Research. Author Manuscript Published OnlineFirst on September 12, 2013; DOI: 10.1158/1078-0432.CCR-13-1637 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Translational Relevance

VEGF pathway inhibitors (VEGFi) are approved for the treatment of advanced cancer. Although

these deliver some clinical benefit, questions remain as to which tumour types are most sensitive,

which patients receive greatest benefit and why the broad pre-clinical activity seen with these

agents did not translate to the clinic. Here we differentiate human tumour types based on their

stromal architecture into tumour vessel (TV) or stromal vessel (SV) phenotypes. The TV

phenotype is associated with tumour types (renal and thyroid cancer) with better single agent

clinical response to a VEGFi than the SV phenotype (CRC and NSCLC). The TV phenotype is

also evident in many human tumour xenograft models commonly used to evaluate efficacy,

whereas a model displaying the SV phenotype was found to be refractory to treatment with the

VEGFR2 blocking antibody DC101. These data suggest an association between stromal

architecture and intrinsic tumour sensitivity to VEGFi therapy.

Abstract

Purpose: The aim of the study was to investigate the vascular and stromal architecture of pre-

clinical tumour models, and patient tumour specimens from malignancies with known clinical

outcomes to VEGFi treatment, to gain insight into potential determinants of intrinsic sensitivity

and resistance.

Experimental design: The tumour-stroma architecture of pre-clinical and clinical tumour

samples were analysed by staining for CD31 and α-SMA. Tumour models representative of each

phenotype were then tested for sensitivity to the VEGFR2 blocking antibody DC101.

Results: Human tumour types with high response rates to VEGF inhibitors (e.g. renal cell

carcinoma), have vessels distributed amongst the tumour cells (a “tumour vessel” phenotype,

TV). In contrast, those malignancies where single agent responses are lower, such as non-small

cell lung cancer (NSCLC), display a complex morphology involving the encapsulation of tumour

cells within stroma that also supports the majority of vessels (a “stromal vessel” phenotype, SV).

Only 1 out of 31 tumour xenograft models displayed the SV phenotype. TV models were sensitive

to VEGFR2 blocking antibody DC101, whereas the SV model was exclusively refractory. The TV

phenotype was also associated with a better RECIST response to bevacizumab + chemotherapy

in metastatic CRC.

Conclusion: The tumour-stromal architecture can differentiate between human tumour types that

respond to a VEGF-signalling inhibitor as single agent therapy. In addition to reconciling the

clinical experience with these agents versus their broad activity in pre-clinical models, these

findings may help to select solid tumour types with intrinsic sensitivity to a VEGFi or other

vascular-directed therapies.

Introduction

The early promise of agents that target the Vascular Endothelial Growth Factor (VEGF) signalling

axis has failed to translate widely in the clinic. Although VEGF-signalling plays a pivotal role in

angiogenesis, and some patients receive benefit, not all patients are responsive to VEGF-inhibitor

(VEGFi) therapy, and those that do respond will eventually become refractory to treatment (1, 2).

The expectation that the use of VEGFi therapy would transform the way in which many solid

human tumours are managed clinically was based largely upon the broad efficacy of these agents

in a variety of pre-clinical tumour models in vivo (3, 4). Given this discrepancy, further insight into

potential determinants of VEGFi efficacy is warranted.

The discordance between preclinical and clinical responses, and the initial lack of markers

identifying intrinsic resistance to a VEGFi has been a significant focus of research. Additional

studies in preclinical models have revealed a number of putative mechanisms that may contribute

to both intrinsic and acquired VEGFi resistance. One of the first mechanisms proposed as

mediating resistance to VEGF-A neutralisation was the presence or recruitment of inflammatory

infiltrate, specifically CD11b, GR-1 positive cells derived from the bone marrow, in response to

upregulation of tumour derived factors (5, 6). A complementary resistance mechanism is the

expression of factors that drive alternative angiogenic pathways such as high expression of the

potent angiogenic stimulus FGF-2 (7). Structural features of the vessel have also been shown to

influence the response to VEGF-signalling inhibitors, with mature vessels supported by pericytes

or myofibroblast like cells being less sensitive to VEGFi treatment. Furthermore, models of

acquired resistance to VEGFi are characterised by the recruitment of pericytes, possibly through

an elevated EGFR signalling response (8), or PDGF-mediated signals (9-11). Determining the

potential clinical relevance of such resistance mechanisms requires a closer examination of the

preclinical models to determine which aspects are representative of human cancer.

We have previously studied the diversity in the angiogenic response represented by a panel of

pre-clinical tumour xenograft models, and found that while the human tumour cells may show

different expression of angiogenic genes, the host response is similar between models (12). To

build on this study we have profiled a number of different human tumours to examine the

histological relationship between tumour and stroma, to determine whether these features are

present within a panel of histologically diverse human tumour xenografts. We found that based on

the tumour-stromal architecture human disease could be broadly sub-divided into two phenotypes

that have different sensitivity to VEGFi therapy as a single agent. The significance of the

phenotype is explored in the context of both pre-clinical modelling and clinical samples.

Material and Methods

Human and xenograft tumour tissues

Formalin fixed paraffin embedded (FFPE) human primary cancer resection blocks, both whole

and formatted into tumor microarrays (TMA), were sourced under approved legal contract from

commercial tissue suppliers, Asterand, Cytomyx and TriStar Technology Group and Wales

Cancer Bank. Appropriate consents, licensing and ethical approval was obtained for this

research. The suitability of each specimen for immunohistochemical analyses was determined by

pathology assessment of tissue morphology and preservation (H&E) and the general extent of

antigen preservation (pan p-Tyr immunostains). Tumor xenograft tissue was derived from

experiments conducted as described (13) with licenses issued under the UK Animals (Scientific

Procedures) Act 1986.

DC101 tumour growth studies

Calu-6 and Calu-3 human lung tumour xenografts, established in nude and SCID female mice

respectively. HT29 and SW620 tumours were established in nude mice. Once tumours reached

a mean volume of ~0.2cm3 to~0.3cm3 (Calu-3), mice were then intraperitoneally injected twice-

weekly with either 15mg/kg of DC101 (Cell Essentials Inc.) or an isotype control antibody for the

times indicated. DC101 is a monoclonal neutralizing antibody raised to murine VEGF receptor-2

(flk-1), that was chosen on the basis of its specificity as a VEGFi. Following treatment, tumours

from each group were excised and divided; one half being snap frozen in liquid nitrogen and

stored at -80°C until required, and the other fixed in neutral buffered formalin for 24 hours.

Histopathological staining

The following antibodies were employed in immunohistochemical (IHC) and immunofluorescence

(IF) analyses: rabbit anti-mouse CD31 (AstraZeneca, CHG-CD31-P1, 13); mouse anti-human α-

smooth muscle actin (αSMA; Sigma, 1A4); rabbit anti-human PDGFRβ (Epitomics 1469-1),

mouse anti-human CD68 (Dako, M0876); mouse anti-human Neutrophil Elastase (Dako, M0752);

rabbit anti-human E-cadherin (Cell Signaling Technology, 3195); mouse anti-human vimentin

(Dako, M0725), rat anti-mouse Gr-1 (BD Pharmingen, 550291), rat anti-mouse F4/80 (Serotec,

MACP497) and mouse anti-human Ki67 (Dako, M7240). Tissues were sectioned onto glass

slides, dewaxed and rehydrated. For both IHC and IF, all incubations were performed at room

temperature and TBS containing 0.05% Tween (TBST) used for washes. Antigen retrieval was

performed in pH 6 retrieval buffer (S1699, Dako) at 110°C for 5min in a RHS-1 microwave

vacuum processor (Milestone), then endogenous biotin (Vector, SP-2002, Neutrophil elastase, E-

cadherin, Gr-1 and F4/80 only), peroxidase activity (3% hydrogen peroxide for 10min) and non-

specific binding sites (Dako, X0909) blocked.

For single marker analyses, antibodies raised to CD31, αSMA, PDGFRβ, CD68, Neutrophil

elastase, E-cadherin, vimentin, Gr-1, F4/80 and Ki67 were diluted, 1:400, 1:1000, 1:2000,1:200,

1:200, 1:100, 1:100, 1:1000, 1:100 and 1:100, respectively in antibody diluent (Dako, S0809) and

applied to sections for 1hr. Mouse Envision secondary (Dako, K4007) for αSMA, vimentin, CD68

and Ki67, rabbit Envision secondary (Dako, K4003) for CD31, biotinylated rabbit anti-mouse IgG

(Dako, E0464) for Neutrophil elastase, goat anti-rabbit IgG (Dako, E0432) for PDGFRβ and E-

cadherin or rabbit anti-rat IgG (Dako, E0488) for Gr-1 and F4/80 were added for 30min. For

Neutrophil elastase, vimentin, Gr-1 and F4/80, Vectastain Elite ABC solution (Vector, PK-6100),

diluted as instructed in kit, was added for 30min. Sections were washed and developed in

diaminobenzidine for 10min (Dako, K3466) then counterstained with Carazzi’s hematoxylin.

Appropriate no primary antibody and isotype controls were performed for each antibody.

Chromogenic CD31-αSMA costain was performed using the Envision G/2 doublestain System

(Dako, K5361) following the manufacturers recommendations with antigen retrieval and primary

antibody dilution as described above. CD31-αSMA co-immunofluorescence was performed as

previously described (13).

Pathology scoring and computer-assisted image analysis

Chromogenic or fluorescent images were captured using the x20 objective of either an Aperio

image scan (Leica Biosystems) or Pannoramic SCAN (3DHISTECH), respectively. Scanned

images were scored by a trained pathologist and two scientists using simple subjective reporting

procedures. Both human and xenograft tumours were scored for stromal-vessel (SV) and tumour-

vessel (TV) phenotypes using H+E or CD31-αSMA stains. The TV phenotype was defined as a

tumour structure where vessels are embedded throughout the tumour cell mass and the SV

phenotype classed as tumour cell nests surrounded by well developed stromal structures which

contain the majority of the vessels. A tumour was classified as either TV or SV based on the

predominant phenotype (cut off of >60% tumour area) while a tumour composed of 40-60% of

both phenotypes was scored as intermediate. These criteria were prospectively defined by an

expert pathologist to clearly differentiate between tumours that were predominantly TV (>60%) or

SV (>60%) and those that were intermediate (40-60%) and difficult to classify as one or the other

phenotype. Based on CD68 marker immunostaining, macrophage infiltrate into the tumour

compartment was classified as negative, low (>0 but ≤1% CD68+ cells: tumour cells) or medium-

high (>1% CD68+ cells: tumour cells). Tumour EMT status was scored using epithelial (E-

cadherin) and mesenchymal (vimentin) markers to define epithelial (E-cadherin+vimentin-),

mesenchymal (E-cadherin-vimentin+) or intermediate phenotypes (E-cadherin+vimentin+).

Computer-assisted image analysis was performed on digitally acquired chromogenic images. For

images of human tumours immunostained for CD31-αSMA, the tumour compartment, including

associated stroma, was selected by hand using the Aperio image viewer software (Leica

Biosystems). Downstream image analysis of the annotated areas were performed using Aperio

image analysis software (Leica Biosystems). Microvessel density (MVD, number of vessels per

mm2 viable tumour), using CD31 as a marker, was determined using the Aperio microvessel

analysis algorithm (Leica Biosystems). Myofibroblast levels based on αSMA were determined

using the Aperio color deconvolution algorithm to measure % αSMA positive pixels/total number

of pixels. GenieTM (Leica Biosystems), a pattern recognition software tool, was trained to segment

stained images of the Calu-3 xenograft into tumour, stroma and necrotic regions and used to

determine the % of each compartment per tumour. This Genie classifier was combined with the

appropriate Aperio image analysis algorithm to analyze biomarker parameters in stromal or

tumour classified regions of Calu-3 tumours. Calu-3 stromal or tumour MVD, based on CD31

immunostain, was determined using the microvessel analysis algorithm. Stromal myofibroblast

content (αSMA and PDGFRβ) or macrophage (F4/80) and neutrophil (Gr-1) infiltrate into the

stroma or tumour was analyzed using the color deconvolution algorithm (Leica Biosystems) to

determine % positive pixels / total number of pixels. Tumour proliferative index (number of Ki67

positive tumour cells/ total number of tumour cells (haematoxylin positive nuclei)) and tumour cell

density (total number of tumour cells (haematoxylin positive nuclei)/tumour area) was determined

using a nuclear algorithm (Leica Biosystems). To measure tumour nest size, the diameter of 44-

220 distinct nests were measured for each αSMA immunostained Calu-3 section (to highlight

haematoxylin positive tumour nests) using the Aperio image viewer measuring tool (Leica

Biosystems). Biomarker data was analysed using the Student's two tailed t-test, to determine

statistical significance between treatment groups.

TaqMan® Fluidigm Gene Expression Profiling

RNA was isolated from 30-50mg frozen tumour using an RNeasy Lipid Tissue Mini Kit (QIAGEN,

74104), according to manufacturer’s protocol. On-column DNase digestion was performed using

the RNase-free DNase Kit (QIAGEN, 79254). RNA concentration was measured using the

NanoDrop ND1000 (Thermo Fisher Scientific).Human and Mouse specific assays were designed

and supplied by Applied Biosystems while eukaryotic 18S rRNA was used as the endogenous

control (Supplementary Table 4). Total RNA (50ng) was converted into cDNA using the High

Capacity cDNA Reverse transcription kit (Applied Biosystems, 4368814) in final volume of 20μl,

according to the manufacturer’s instruction. cDNA (1.25 μl) was pre-amplified using a pool of

TaqMan primers at a final dilution of 1 in 100 and Pre-amplification Master Mix (Applied

Biosystems, 4391128) in a final volume of 5μl. Samples were diluted 1 in 5 with 1x TE and stored

at -20°C.Sample and assay preparation for 48.48 dynamic arrays was performed according to the

manufacturer’s instruction (Fluidigm).

Data was collected and analysed using the Fluidigm Real-Time PCR Analysis 2.1.1 software

(v.2.1.3). Species-specific normalization of the expression data to 18S rRNA was performed as

reported (12) Gene expression values were calculated using the comparative CT (-∆CT) method

as previously described in User Bulletin #2 ABI PRISM 7700 Sequence Detection System

10/2001, using the corrected 18S rRNA Ct values for normalisation of the tumour transcript and

the original values for the stroma. To determine genes altered by DC101 treatment student and

differentially expressed between Calu-3 versus the others models t tests and fold change were

calculated with significantly altered genes identified by having a P<0.05 and a Fold Change >1.5.

Results

Tumour-stromal architecture defines human tumour types that respond to VEGF-signaling

inhibitors as monotherapy

To assess the tumour-stromal architecture of different solid human malignances we used a

number of tumour tissue micro-arrays (TMA) comprising renal cell carcinoma (RCC),

glioblastoma multiforme (GBM), head and neck squamous cell carcinoma (HNSCC), non small

cell lung carcinoma (NSCLC), ovarian (OvC), thyroid (ThC) subtypes, colorectal (CRC), prostate

(PC) and breast (BC) cancers (Fig.1, Supplementary Table 2). For the purposes of this analysis

we have viewed responsive disease as high incidence of tumour response by RECIST

assessment (tumour shrinkage) and/or a large increase in progression free survival on long term

therapy reported with different therapies targeting VEGF-signalling. RCC (14) and GBM (15, 16)

are diseases we have taken as representative of cancers known to respond to VEGFi as a single

agent therapy (Supplementary Table 1). CRC (17) and NSCLC (18) are diseases where VEGF

inhibitors show little single agent responses, and hence are trialed in combination with cytotoxic

therapy. Tumours were screened by immunofluorescence to visualize the relative distribution of

myofibroblast like cells (αSMA), blood vessels (CD31) and tumour cells (DAPI). This revealed

two dominant morphologies: (1) a “tumour-vessel” phenotype (TV) which describes a tumour

structure with vessels embedded throughout the tumour cell mass represented in RCC, GBM and

HCC (Fig. 1a, Supplementary Fig. 1, and Table 2); (2) a “stromal-vessel” phenotype (SV) in

which the tumour structure is dominated by a pattern of tumour cell nests surrounded by well

developed stromal structures containing the majority of the vessels (Fig. 1b), as evident in CRC,

NSCLC and PC (Fig. 1b and Supplementary Fig. 1 and Table 2). Using this classification, cores

from multiple tumour types were scored for either a TV or SV phenotype (Fig. 1c). The TV

phenotype was observed in RCC, GBN, OvC, HCC, ThC and HNSCC. The SV phenotype was

dominant in CRC, NSCLC, PC and BC

Since the phenotype ratios for each tumour type were derived initially from TMA cores,

full tumour sections were also classified for each phenotype to ensure the results were

representative of larger diagnostic tumour samples (RCC, HCC and PC (n=10), GBM, ThC, CRC

and BC (n=9) and NSCLC (n=19)); Fig. 2a, Supplementary Fig 2, Supplementary Table 3).

Where sections exhibit regions of both phenotypes, the predominant phenotype (>60%) was

scored (Supplementary Fig. 2a). The frequency of the phenotypes representing diagnostic

specimens agreed with the original TMA dataset (Supplementary Fig. 2b and c). These data

establish that the tumour-stromal architecture of the tumour can be used to cluster human tumour

types into two groups. Significantly, those clinical tumour types associated with response to a

VEGFi alone (RRC (14), ThC (19, 20), GBN (15, 16) were dominated by a TV phenotype. In

contrast, tumours where a VEGFi has shown modest activity in combination (CRC (17), PC (21)

NSCLC, (18), BC (22)) were dominated by the SV phenotype.

Additional biomarker analyses were performed on the diagnostic tumour panel (n=32 TV,

n=35 SV) to further examine the properties of the two phenotypes (Fig. 2b-d). Although immature

and mature-pericyte covered vessels (CD31+ endothelial cells with tightly associated αSMA+

cells) were associated with both phenotypes, tumour embedded vasculature of the TV phenotype

tended to be pericyte-free while stromal vessels in the SV phenotype were often pericyte covered

or associated with αSMA+ fibroblasts (Fig. 2b). Computer-assisted image analysis of CD31

immunostained lesions revealed that the tumour vascular density in the combined tumour and

stromal compartments was higher in TV (mean ± s.e.m = 105.3 ± 21.3) than SV (47.7 ± 7.5; Fig.

2c and Supplementary Fig. 3b). αSMA analysis showed myofibroblasts were greater in SV (s.e.m

= 10.0 ± 1.8) compared to TV (1.4 ± 0.3) phenotype tumours (Fig. 2c and Supplementary Fig 3b)

tumours. EMT status was determined by immunostaining serial sections for E-cadherin and

vimentin (Fig. 2d and Supplementary Fig 3b). TV tumours were generally more mesenchymal

(E-cadherin-/vimentin+) than SV tumours (E-cadherin+/vimentin-). Qualitative histological

assessment of CD68 immunostained tumours from the panel showed that the TV phenotype has

higher levels of macrophage tumour infiltrate (Supplementary Fig 3b, c).

Neuroendocrine pancreatic cancer has recently been shown to respond to the VEGFi

sunitinib (23) while pancreatic adenocarcinoma has proven comparatively resistant to such

approaches (24, 25, 26). Comparing available samples of neuroendocrine pancreatic cancer with

pancreatic adenocarcinoma revealed them to represent TV and SV phenotypes, respectively

(Supplementary Fig. 3a), consistent with our hypothesis on VEGFi sensitivity.

The TV phenotype is common to tumour xenograft models

To evaluate the relevance of the two phenotypes in pre-clinical models, a panel of human tumour

xenografts grown subcutaneously and representing a broad range of different tumour types of

origin (i.e. lung (8), colon (5), breast (4), prostate (3), brain (2), stomach (1), ovary (1), pancreas

(1), skin (1), pharynx (1), uterous (1), vulva (1) and blood (1) tumour and fibrosarcoma (1) cell

lines) were classified for the two phenotypes based on the approach employed for human

tumours (Fig. 3). The majority (30/31) exhibited a TV phenotype (Fig. 3a,b). Only Calu-3 (lung

adenocarcinoma) had a SV phenotype (Fig. 3a,c). This data indicates that the TV phenotype is

common to subcutaneously grown tumor xenograft models.

Intrinsic gene expression differences between SV and TV xenograft tumours

Gene expression analysis of human transcripts was used to investigate the relationship between

Calu-3, Calu-6, SW620, and HT29 tumours. The difference in expression of a range of 180

genes associated with angiogenesis, inflammation and invasive growth was determined (Fig. 3d,

e). This gene set differentiated Calu-3 from the other models; genes showing the highest

differential expression in Calu-3 were MMP7, LPAR3, CT55, IL1RN, SPP1 (osteopontin).

Compared to other models analysed, Calu-3 xenograft tumours were enriched in genes

associated with recruitment of the stromal cells, in particular FGF and PDGF ligands (Fig 3d). To

validate that Calu-3 tumours showed a different expression profile they were also compared to a

broader panel of tumour xenograft models, again the same genes were over-expressed relative

to other models. This expression analysis supports the conclusion that Calu-3 tumour cells exhibit

high expression of a number of transcripts that may contribute to their phenotype.

The TV phenotype has greater sensitivity to VEGFR-2 inhibitor, DC101 than the SV

phenotype, in vivo

To determine whether the tumour-stromal architecture could influence tumour response to VEGFi

we used the murine VEGFR-2 blocking antibody, DC101 to assess the effects of pruning

neovasculature on tumour growth in TV (Calu-6, HT-29 and SW620) compared to SV (Calu-3)

tumour xenograft models (25). DC101 was dosed intraperitoneally at 15mg/kg twice-weekly.

Calu-6, HT-29 and SW620 xenografts exhibited the classic tumour growth response to VEGF-

signaling inhibitors (Fig. 4a) seen in multiple models (27, 28, 29). In contrast Calu-3 tumours

exhibited a poor response, with DC101 having no effect on tumour growth (experiment 1), or

inducing a small initial reduction in growth (0-3 days, experiment 2) followed by a rapid return to a

growth rate similar to the controls (Fig 4b).

The gross vascular response to DC101 is similar between TV and SV tumour xenografts

To compare the total vascular response of TV and SV models to DC101, MVD analysis was

performed on CD31 immunostained control and DC101 treated samples from the Calu-6 and

Calu-3 (replicate 1 and 2) growth studies. The effect of DC101 on MVD reduction was similar

between Calu-6 and Calu-3 models (Fig. 4c). Furthermore, species specific RT-qPCR using

genes associated with angiogenesis and tumour cell invasion (Fig. 4d) revealed that human

tumour gene changes were unique to each tumour xenograft model, but a range of murine genes

associated with endothelial cells, were reduced in both. Therefore, while there is a clear vascular

response to DC101 in both models this does not translate to growth inhibition in the Calu-3

xenograft.

DC101 targets the stromal vasculature in a SV phenotype model which leads to a

reduction in the stromal compartment but with negligible effect on tumour size

Tumour embedded vessels were not detected in the Calu-3 model, however vasculature was

detected at a high density in the stroma (Fig 5a, b). Stromal vasculature was reduced by >50% in

response to DC101 with the remaining stromal vessels exhibited a greater mean vessel area than

those in untreated Calu-3 tumours (Fig 5a). Qualitative assessment of CD31-αSMA

immunofluorescent stained tumours revealed the proportion of pericyte covered stromal vessels

increased in treated tumours (Fig. 5b). The data indicates that while treatment with DC101

reduces small immature vessels it fails to reduce the larger more mature vasculature within the

stroma.

To examine the influence of the SV phenotype on the response to DC101 we performed a

detailed biomorphometric analysis of DC101 treated Calu-3 xenograft tumours. In Calu-3

tumours, DC101 treatment led to reduction in the area of the stromal compartment and an

increase in the area occupied by tumour cells with negligible effect on tumour size (Fig. 5c).

Analysis of the stromal compartment for myofibroblasts using αSMA and PDGFRβ staining

revealed a significant reduction in stromal myofibroblast levels (Fig. 5d). DC101 treatment also

led to a significant increase in the mean diameter of tumours nests (from 200 to 450μm, P=0.04),

however effects on the proliferative index and density of the tumour were negligible at this time

point (Supplementary Fig. 4a, c). The recruitment of inflammatory cells was also assessed using

F4/80 (macrophages) and Gr-1 (neutrophils). DC101 treatment resulted in a small but significant

increase in the tumour macrophage content and a significant accumulation of Gr-1 positive cells

in the stroma (Supplementary Fig. 4b, c). In conclusion, the reduction of stromal angiogenic

vasculature by DC101 in the Calu-3 model had the greatest impact on the architecture and

cellular composition of the stroma, rather than the tumour compartment, but with a negligible

effect on the overall tumour mass.

The two phenotypes correlate with different clinical efficacies to bevacizumab and

FOLFIRI in metastatic CRC.

The relationship between phenotype and RECIST response to a combination of VEGFi and

oxaliplatin-based chemotherapy in metastatic CRC patients was investigated. Two TMAs

consisting of surgical tumour samples from 56 metastatic CRC patients (Tristar) were

chromogenically stained for CD31-αSMA and scored for the TV and SV phenotypes by two

observers. 42 patient samples were classified as the SV phenotype and 14 as the TV phenotype

(Fig. 6). RECIST response information for FOLFIRI and bevacizumab (Avastin) treatment as 1st

or 2nd line therapy post surgery was available for each patient. Fisher's Exact Test was used to

determine the significant associations between the phenotype and best RECIST response

categories (PD and other RECIST outcome) during patient treatment periods (for this dataset

best RECIST responses also had longest duration). The probability of the SV phenotype group

having a poorer response (PD category) to bevacizumab and FOLFIRI than that of the TV

phenotype group was statistically significant (P=0.0487; 95% CI, 1.24-∞). Although it is not

possible to conclude that the differential response is a direct result of the treatment with

bevacizumab alone, the data suggest that the TV phenotype appears to be more sensitive to the

combination therapy than the SV phenotype.

Discussion

This study suggests that human tumour types can be broadly categorised according to

the spatial distribution of their blood vessels in relation to the tumour cells and other stromal

components. Across a broad panel of human tumour types, independent of disease type, tumours

largely adopted either a tumour vessel (TV) or stromal vessel (SV) phenotype. The TV phenotype

appears to be indicative of tumours that are likely to show greater single agent responses to anti-

angiogenic drugs targeting the VEGF signalling axis. Although we were able to separate different

tumour types using this approach no tumour type studied was exclusively one type of

architecture. For example, although renal cell carcinoma tumours are predominantly of the TV

phenotype, some exhibit a SV architecture. Conversely while CRC tumours are dominated by the

SV phenotype a subset exhibit the TV architecture. Heterogeneity is also evident within individual

tumours. In particular, tumour types with a SV architecture have regions with TV architecture,

which may represent more angiogenic regions of the tumour. TV vasculature is embedded within

the tumour cell mass and forming a major structural feature of the tumour, facilitating intra-

tumoral blood flow. In contrast, in SV tumours the vascular supply develops almost exclusively in

the stromal compartment, which separates the tumour cells from the vessels, potentially creating

more mature vessels.

The TV phenotype was associated with differences in the distribution of other cells,

specifically the presence of high levels of macrophages. In SV tumours these macrophages did

not commonly associate with the tumour cells perhaps indicating a different role for the

macrophage in these tumours. A possible explanation for these observations is that the

intratumoural vasculature observed in the TV phenotype, but absent from the SV phenotype,

would facilitate extravasation of immune infiltrate into the tumour. In contrast, in the SV

phenotype macrophages would extravasate into the dense stroma surrounding the tumour nests

which may impair their migration into the tumour. Based on E-cadherin loss and vimentin gain,

the TV phenotype was more associated with tumour cells of a mesenchymal phenotype. It is

conceivable, that the tissue of origin or subsequent progression of tumour to a more

mesenchymal phenotype could influence reciprocal signaling between the tumor and the tumour

microenvironment, thereby influencing the extent of stromal recruitment and its subsequent

architecture. However, we do not currently have a mechanism to explain this.

While it remains challenging to acquire human tumour samples with associated outcome data

following treatment with VEGF signalling inhibitors, we were able to obtain TMAs of samples from

CRC patients treated with bevacizumab + FOLFIRI. Although these samples were limited in

number, and only RECIST response category information was available, they did allow initial

exploration of the hypothesis that TV and SV tumours may respond differently to therapy. The

analysis from the TMA is valid as previous data on baseline samples sets showed good

concordance between the whole tumour sample and the TMA. The distribution of TV vs SV was

largely representative of the larger datasets for CRC. Interestingly in this dataset the TV

phenotype was associated with good response, with no examples of progressive disease

associated with this phenotype, suggesting that the TV phenotype may be more sensitive to the

combination therapy. This supports the idea that the TV or SV should be considered when

differentiating response of these tumours to angiogenic therapy. It would be interesting to

understand in late stage CRC, where VEGF inhibitors can be used as a single agent, whether

differential response or disease control is associated with the differences in phenotype. As PFS

and OS information was not available for these samples, further analyses to correlate phenotypes

with these survival outcomes are required to understand the utility of this approach to predict long

term outcome in diseases dominated by SV morphology. Only primary tumour samples were

analysed in this study. Often archival diagnostic samples from the primary tumour are the only

tissue available to evaluate the tumour biomarker status of a patient. It will be important to

establish whether the phenotypes are maintained during disease progression (e.g. metastatic

lesions) and on progression following therapeutic intervention to determine the utility of stratifying

tumour types based on phenotype using archival samples.

Pre-clinical tumour xenografts have been used to define many of the mechanisms with the

potential to influence response to VEGFi, other angiogenic factors (7), bone marrow derived cells

(5,6), or by recruitment of stromal fibroblasts/pericytes into the tumour which drive the maturation

of tumour vessels (8, 9 , 10). Our analysis suggests that those mechanisms of resistance defined

in pre-clinical tumour xenografts may be most relevant to those diseases that display a TV

phenotype. It is probable that in selecting tumour models that grow quickly to facilitate drug

testing, stromal rich tumours have been severely under-represented by virtue of their growth

characteristics.

Understanding the implications of the SV architecture on tumour cell function and response to

drugs will be important. The Calu-3 model (SV phenotype), which exhibited a slower growth rate

and showed a poor response to the specific VEGFR2 blocking antibody DC101, expressed higher

levels of human genes thought to play a role in recruitment of stroma than models representative

of the TV phenotype. These genes, PDGF-A, C and D, FGF-2, IL-8 are known to influence

fibroblast function as well as resistance to VEGF signalling inhibitors (9-12, 27-31). The high

stromal content would generate a distinct microenvironment where the mature vessels in the

stroma may be resistant to therapy, an effect observed with VEGF-signalling inhibitors in pre-

clinical tumour xenografts (3, 11), the stromal phenotype may be a more extreme representation

of this resistance mechanism. Interestingly Calu-3 tumour stroma had a high local vessel density

that was reduced following DC101 treatment in this model, but a significant number of established

mature residual vessels remained. In addition, although treatment with DC101 led to a decrease

in the density of the stroma the general architecture of the stroma was maintained. This suggests

that in Calu-3 tumours there may be a role for the vasculature in maintaining the stroma, or that

changes in the tumour compartment following a reduction in vessels lead to a reduction in

reactive stroma. Treating human SV tumours with VEGFi may change the apparent vessel

permeability, perfusion or blood flow as determined by imaging, or reduces vessels as measured

by other biomarker approaches, but without reaching a threshold that is sufficient to impact upon

tumour cell growth or survival. In SV tumours, pruning of immature vasculature and retention of

larger mature vessels in the stroma by VEGFi could alter interstitial pressure within the tumour

consistent with the vascular normalisation hypothesis (32) and influence drug delivery. Tumour

types that are predominantly SV (e.g. CRC, NSCLC, BC) do not exhibit dramatic responses to

VEGFi monotherapy in the clinic but alterations in stomal density may give benefit in combination

or with single agent therapy in late stage disease. Although we have only performed a limited

analysis it will be interesting to examine how reductions in tumour vasculature influence tumour

cell survival or growth in other models with a high stromal content. Here we focussed on testing

VEGF signalling inhibition by using an antibody that specifically antagonised VEGFR-2 on the

murine vasculature. Many small molecule VEGFR inhibitors have also been developed but these

compounds have broader pharmacology profiles and inhibit additional kinase such as PDGFR

and related family members which impart additional effects in the stroma or tumour. It will be of

future interest to also model such small molecule inhibitors preclinically taking into account their

clinical exposure profiles, to determine whether any differential effects are delivered with these

mixed pharmacology agents on TV and SV tumours.

Our findings offer an alternative way to interpret the effects we have seen with VEGFi therapy in

the clinic. If tumours with either a TV or SV architecture are viewed as distinct from the point of

view of tumour angiogenesis then it also challenges whether we have interpreted current

biomarkers appropriately. For example, while biomarker changes in vessel numbers or vascular

function in response to VEGFi treatment, could conceivably be observed in tumours with either a

TV or SV phenotype, it is possible that these may only translate into therapeutic benefit in the

former. Given that predominantly SV diseases (e.g. CRC) will still have a small proportion of

tumours with a TV phenotype, it would also be interesting to determine whether differences in

response are seen within these tumour subsets. This segmentation of disease may offer an

approach to prioritise tumour types that are likely to show a clinical response to VEGFi therapy,

and potentially other vascular directed agents. We suggest that diseases that predominantly

display a TV phenotype would show good single agent response to such agents.

This study has a number of limitations. Although the preclinical evaluation of the concept uses

several TV models treated with DC101, because of the scarcity of tumour xenografts that

represent the SV phenotype we were limited to one SV model (Calu-3). Studies using additional

VEGR antagonists at different exposures in models that represent a range of phenotypes,

including additional SV phenotypes and models heterogeneous for both phenotypes, would be

important in investigating this concept further. In addition, a more comprehensive evaluation of

the phenotypes in specific tumour types is required to better understand; i) prevalence and

heterogeneity, ii) prognostic and predictive value and iii) phenotype changes associated with

disease progression or following therapy.

The vascular architecture is not the only parameter that is likely to determine the response of an

individual tumour to VEGFi therapy. Within the context of each phenotype other factors may

further influence the outcome following therapy. For example, with a vascular targeting therapy it

is reasonable to expect factors that determine the metabolic status of the cells, or the degree of

environmental stress the tumour cell can withstand, would also influence the likelihood of

obtaining an objective tumour response. In particular, the stromal architecture may influence

features such as the intrinsic metabolic status of the tumour cells either directly, or by creating a

dependency on metabolic coupling between the stromal fibroblast and tumour to promote

anabolic growth (the “reverse Warburg effect”) (33). It is unclear what features of the tumour cell

determine the vessel phenotypes, but a better understanding of the underlying mechanisms could

help further refine patient selection strategies for VEGFi therapy.

In conclusion, we suggest that consideration of the tumour-stromal architecture may be an

important determinant of whether tumours will be therapeutically susceptible to treatment with

VEGFi monotherapy and potentially other vascular modulating agents. The ability to stratify

VEGFi responses in preclinical models and human disease on the basis of a TV or SV phenotype

warrants wider evaluation.

References

1 Crawford Y and Ferrara N, Tumor and stromal pathways mediating refractoriness/resistance to

anti-angiogenic therapies. Trends in Pharmacological Sciences 2009:30:624-30.

2 Sitohy B, Nagy JA and Dvorak HF Anti-VEGF/VEGFR therapy for cancer: reassessing the

target.

Cancer Res. 2012;72:1909-14.

3 Bergers G and Hanahan D. Modes of resistance to anti-angiogenic therapy. Nat Rev Cancer

2008;8;592-603.

4 Ivy SP, Wick JY, Kaufman BM. An overview of small-molecule inhibitors of VEGFR signaling

Nat Rev Clin Oncol. 2009;6:569-79.

5 Muramatsu M, Yamamoto S, Osawa T, and Shibuya M. Vascular endothelial growth factor

receptor-1 signaling promotes mobilization of macrophage lineage cells from bone marrow and

stimulates solid tumor growth. Cancer Res. 2010;70:8211-21.

6 Shojaei F, Wu X, Qu X, Kowanetz M, Yu L, Tan M, et al. G-CSF-initiated myeloid cell

mobilization and angiogenesis mediate tumor refractoriness to anti-VEGF therapy in mouse

models. PNAS. 2009;106:6742-7.

7 Welti JC, Gourlaouen M, Powles T, Kudahetti SC, Wilson P, Berney DM, et al. Fibroblast

growth factor 2 regulates endothelial cell sensitivity to sunitinib. Oncogene. 2011;30:1183-93.

8 Cascone T, Herynk MH, Xu L, Du Z, Kadara H, Nilsson MB, et al. Upregulated stromal EGFR

and vascular remodeling in mouse xenograft models of angiogenesis inhibitor-resistant human

lung adenocarcinoma. J Clin Invest. 2011;121:1313-28.

9 Kitadai Y, Sasaki T, Kuwai T, Nakamura T, Bucana CD, Fidler IJ, et al. Targeting the

expression of platelet-derived growth factor receptor by reactive stroma inhibits growth and

metastasis of human colon carcinoma. Am J Pathol. 2006;169:2054-65.

10 Pietras K, Pahler J, Bergers G, Hanahan D. Functions of paracrine PDGF signaling in the pro-

angiogenic tumor stroma revealed by pharmacological targeting. PLoS Med. 2008;5(1):e19.

11 Crawford Y, Kasman I, Yu L, Zhong C, Wu X, Modrusan Z, et al. PDGF-C mediates the

angiogenic and tumorigenic properties of fibroblasts associated with tumors refractory to anti-

VEGF treatment. Cancer Cell. 2009;15:21-34.

12 Farren M, Weston S, Brown H, Broadbent N, Powell S, Shaw R, et al. Expression of stromal

genes associated with the angiogenic response are not differentiated between human tumour

xenografts with divergent vascular morphologies. Angiogenesis. 2012;15:555-68.

13 Smith NR, Baker D, James NH, Ratcliffe K, Jenkins M, Ashton SE, et al. Vascular endothelial

growth factor receptors VEGFR-2 and VEGFR-3 are localized primarily to the vasculature in

human primary solid cancers. Clin Cancer Res. 2010;16:3548-61.

14 Sternberg CN, Davis ID, Mardiak J, Szczylik C, Lee E, Wagstaff J, et al. Pazopanib in locally

advanced or metastatic renal cell carcinoma: results of a randomized phase III trial. J Clin Oncol.

2010;28:1061-8.

15 Batchelor TT, Duda DG, di Tomaso E, Ancukiewicz M, Plotkin SR, Gerstner E, et al. Phase II

study of cediranib, an oral pan-vascular endothelial growth factor receptor tyrosine kinase

inhibitor, in patients with recurrent glioblastoma. J Clin Oncol. 2010;28:2817-23.

16 Friedman HS, Prados MD, Wen PY, Mikkelsen T, Schiff D, Abrey LE, et al. Bevacizumab

alone and in combination with irinotecan in recurrent glioblastoma. J Clin Oncol. 2009;27:4733-

40.

17. Winder T, Lenz HJ. Vascular endothelial growth factor and epidermal growth factor signaling

pathways as therapeutic targets for colorectal cancer. Gastroenterology. 2010;138:2163-76.

18 Das M, Wakelee H. Targeting VEGF in lung cancer. Expert Opin Ther Targets. 2012;16:395-

406.

19 Lam ET, Ringel MD, Kloos RT, Prior TW, Knopp MV, Liang J, et al. Phase II clinical trial of

sorafenib in metastatic medullary thyroid cancer. J Clin Oncol. 2010;28:2323-30.

20 Bible KC, Suman VJ, Molina JR, Smallridge RC, Maples WJ, Menefee ME, et al. Efficacy of

pazopanib in progressive, radioiodine-refractory, metastatic differentiated thyroid cancers: results

of a phase 2 consortium study.; Endocrine Malignancies Disease Oriented Group; Mayo Clinic

Cancer Center; Mayo Phase 2 Consortium. Lancet Oncol. 2010;11:962-72.

21 Antonarakis ES, Carducci MA. Targeting angiogenesis for the treatment of prostate cancer.

Expert Opin Ther Targets. 2012;16:365-76.

22 Mackey JR, Kerbel RS, Gelmon KA, McLeod DM, Chia SK, Rayson D, et al. Controlling

angiogenesis in breast cancer: A systematic review of anti-angiogenic trials. Cancer Treat Rev.

2012;38:673-88.

23 Raymond E, Dahan L, Raoul JL, Bang YJ, Borbath I, Lombard-Bohas C, et al. Sunitinib malate

for the treatment of pancreatic neuroendocrine tumors. N Engl J Med. 2011;364:501-13.

24 Kindler HL, Wroblewski K, Wallace JA, Hall MJ, Locker G, Nattam S, et al. Gemcitabine plus

bevacizumab compared with gemcitabine plus placebo in patients with advanced pancreatic

cancer: phase III trial of the Cancer and Leukemia Group B (CALGB 80303) J Clin Oncol.

2010;28:3617–22.

25 Kindler HL, Ioka T, Richel DJ, Bennouna J, Létourneau R, Okusaka T, et al. Axitinib plus

gemcitabine versus placebo plus gemcitabine in patients with advanced pancreatic

adenocarcinoma: a double-blind randomised phase 3 study. Lancet Oncol. 2011;12:256-62.

26 O'Reilly EM, Niedzwiecki D, Hall M, Hollis D, Bekaii-Saab T, Pluard T, et al. A Cancer and

Leukemia Group B phase II study of sunitinib malate in patients with previously treated metastatic

pancreatic adenocarcinoma (CALGB 80603). Oncologist. 2010;15:1310-9.

27 Prewett M, Huber J, Li Y, Santiago A, O'Connor W, King K, et al. Antivascular endothelial

growth factor receptor (fetal liver kinase 1) monoclonal antibody inhibits tumor angiogenesis and

growth of several mouse and human tumors. Cancer Res. 1999;59:5209-18.

28 Wedge SR, Kendrew J, Hennequin LF, Valentine PJ, Barry ST, Brave SR, et al. AZD2171: a

highly potent, orally bioavailable, vascular endothelial growth factor receptor-2 tyrosine kinase

inhibitor for the treatment of cancer. Cancer Res. 2005;65:4389-400.

29 Ebos JM, Lee CR, Christensen JG, Mutsaers AJ, Kerbel RS. Multiple circulating

proangiogenic factors induced by sunitinib malate are tumor-independent and correlate with

antitumor efficacy.Proc Natl Acad Sci U S A. 2007;104:17069-74.

30 Casanovas O, Hicklin D, Bergers G, Hanahan D. Drug resistance by evasion of antiangiogenic

targeting of VEGF signaling in late stage pancreatic islet tumors. Cancer Cell 2005;8:299–309.

31 Crawford Y, Kasman I, Yu L, Zhong C, Wu X, Modrusan Z, et al . PDGF-C mediates the

angiogenic and tumorigenic properties of fibroblasts associated with tumors refractory to anti-

VEGF treatment. Cancer Cell 2009;15:21–34.

32 Chauhan VP, Stylianopoulos T, Martin JD, Popović Z, Chen O, Kamoun WS, et al.

Normalization of tumour blood vessels improves the delivery of nanomedicines in a size-

dependent manner. Nat Nanotechnol. 2012 7:383-8.

33 Whitaker-Menezes D, Martinez-Outschoorn UE, Lin Z, Ertel A, Flomenberg N, Witkiewicz

AK,et al. Evidence for a stromal-epithelial "lactate shuttle" in human tumors: MCT4 is a marker of

oxidative stress in cancer-associated fibroblasts. Cell Cycle. 2011 ;10:1772-83.

ACKNOWLEDGMENTS

AUTHOR CONTRIBUTIONS

NRS, STB designed and directed experiments and their analyses and wrote the manuscript.

SM, DB, RS conducted experiments and reviewed manuscript

MF, JK designed and conducted experiments and reviewed manuscript

CW provided pathology review

XW provided statistical review

SRW interpreted study data and helped draft and critique the manuscript

COMPETING FINANCIAL INTERESTS

All authors are current or former employees of AstraZeneca.

FIGURE LEGENDS

Figure 1 Two dominant phenotypes based on tumour-stromal architecture stratify tumour

types. (a,b) diagrams highlighting: (a) tumour-vessel phenotype (TV, a tumour structure where

vessels are embedded throughout the tumour cell mass) and (b) stromal-vessel phenotype (SV,

a tumour structure dominated by a pattern of tumour cell nests surrounded by well developed

stromal structures which contain the majority of the vessels). Also shown are

immunofluorescence stained FFPE human tumours (CD31-AF488 (green); αSMA-AF555 (red)

and DAPI counterstain (blue)), indicative of TV (RCC, GBM ,HCC, ThC, OvC and HNSCC) and

SV (CRC, NSCLC, PC and BC). (c) Percentage ratios of TV:SV phenotypes for RCC, GBM,

OvC, HCC, ThC, HNSCC, CRC, NSCLC, PC and BC. Phenotypes were scored from CD31-

αSMA or H+E stained TMAs (1 TMA per tumour type, number of patient samples per tumour type

are shown).

Figure 2 Biomarker analyses of the two phenotypes in whole tumour samples. (a)

Percentage ratios of TV:SV:intermediate phenotypes for RCC, GBM, HCC, ThC, HNSCC, CRC,

PC and BC. Phenotypes were scored from CD31-αSMA costained whole FFPE tumour sections

(number of patient samples per tumour type are shown). (b) Pericyte coverage of vessels

associated with TV (RCC and ThC) and SV (CRC and NSCLC) phenotypes, white or grey arrows

indicate naked and pericyte covered vessels, respectively. Samples were analysed for: (c)

microvessel density (MVD, number of CD31+ vessels per mm2) and myofibroblast content

(percentage number of αSMA+ pixels/ total number pixels) and (d) EMT status (percentage ratio

epithelial (E-cadherin+vimentin-): intermediate (E-cadherin+vimentin+): mesenchymal (E-cadherin-

vimentin+)).

Figure 3. Prevalence and intrinsic gene expression differences between the two

phenotypes in tumour xenograft models (a) Classification of 31 subcutaneously implanted

human tumour xenograft models for TV and SV phenotypes. Data was derived from the scoring

of CD31-αSMA immunostained TMAs (4 FFPE tumours per model, 3 cores per tumour). (b,c)

Diagrams and fluorescent immunostained (CD31-AF488 (green); αSMA-AF555 (red) and DAPI

counterstain (blue)) images representing preclinical xenograft. (b) Tumour vessel (Calu-6 - low

myofibroblast content) and (c) stromal vessel (Calu-3). (d, e) Trancript profiling of TV (Calu-6,

HT29 and SW620) and SV (Calu-3) tumour xenograft models. Genes with high (d) and low (e)

expression levels in Calu-3 compared to Calu-6, HT29 and SW620 are presented. ddCT =

(Average dCT Calu-3) – (Average dCT (Calu-6 + SW620 + HT29)).

Figure 4 The tumour vessel phenotype is more sensitive to VEGFR-2 signalling inhibitor,

DC101 than the stromal-vessel phenotype in vivo. (a, b) Geometric mean tumour volume

3 growth curves (cm ) for DC101 (15mg/kg twice-weekly), (red) and IgG1 (control), (blue) treated

arms of (a) Calu-6, HT-29, SW620 and (b) Calu-3 (growth curve 1, growth curve 2) tumour

xenograft studies. Geometric mean ± s.e.m. are shown for growth curves. (c) Microvessel

+ 2 density (MVD, number of CD31 vessels per mm ) of IgG1-control and DC101 treated Calu-6 and

Calu-3 tumour studies (growth curves 1 and 2). (d) Significant gene changes induced in the

tumour (human) and stromal (stromal) vascular modulation and invasion transcript profile on

DC101 treatment of Calu-6 and Calu-3 (growth curve 1) xenograft tumors. Tumour transcripts are

normalized to human 18s while stromal transcript are normalized to total 18s. Only significant

gene changes are included (P>0.05 and fold change greater then 1.5), Each column is an

individual tumour and each colour gradient is scaled independently within each gene.

Figure 5 The stromal neovasculature is targeted by DC101 in a tumour model of the SV

phenotype, Calu-3 which leads to stromal changes with negligible effect on tumour size.

(a) Microvessel density (MVD, number of CD31+ vessels per mm2) and microvessel area (MVA,

total CD31+ area (μm2)/total number of CD31+ vessels) of Genie classified stromal compartment

of IgG1-control and DC101 treated Calu-3 tumours (growth curve study 1). (b) Chromogenic

(CD31, DAB, brown) and fluorescent (CD31-AF488 (green); αSMA-AF555 (red) and DAPI

counterstain (blue) immunostained control and DC101 treated Calu-3 tumours (growth curve 1).

For CD31 images, black or grey arrows point to small and large stromal vessels, respectively

(S=stroma, T=tumour). For CD31-αSMA images, white or grey arrows indicate naked and

pericyte covered vessels, respectively. (c) GenieTM (Aperio Technologies Ltd) classification and

quantification of tumour, stromal and necrotic areas (% number of tumour, stromal or necrotic

pixels/ total number of pixels) using digitally captured images of haematoxylin stained sections for

DC101 (15mg/kg/twice weekly, n=6) and IgG1 (control, n=6) treated arms of Calu-3 growth

studies (1 and 2 (data not shown)). Haematoxylin (blue) staining of control and DC101 treated

Calu-3 tumours and overlay of Genie classified tumour, stromal and necrotic compartments for

Calu-3 growth curves 1 and 2 (data not shown). (d) Stromal myofibroblast content (percentage

number of αSMA+ or PDGFRβ+ pixels/ total number pixels) within Genie classified regions of

control and DC101 treated tumours (growth curve 1). ). Mean ± s.e.m. and P-values are shown

for all biomarkers

Figure 6 Correlation of phenotype with RECIST response to bevacizumab and FOLFIRI in

metastatic CRC.

Data was derived from a CD31-aSMA immunostained TMA consisting of 56 metastatic CRC

samples (n=2 cores per tumour) from patients prior to treatment . Tumours were classified as

either tumour vessel (TV) or stromal vessel (SV) phenotypes by two observers and compared to

best RECIST outcome for the treatment period of each patient (CR, complete response; PR,

partial response; SD, stable disease and PD, progressive disease). Based on the dataset,

Fisher’s Exact Test demonstrated that the SV phenotype group has higher probability of having a

poor response (PD category) to bevacizumab and FOLFIRI than that of the TV phenotype group

with statistical significance (P=0.0487; CI, 1.24-∞).

Fig. 1 a b

TV SV

RCC CRC

CD31 CD31 α αSMA SMA 100μμM 200μM μ 200μM DAPI 100 M DAPI 100 M HCC NSCLC

100μM

CD31 CD31 α αSMA SMA 100μM 200μM μ 200μM DAPI DAPI 100 M ThC PC

CD31 CD31 αSMA μ αSMA 100100MμM 200μM 100μM 200μM DAPI DAPI

μ c 100 M

100%

90% TVTV SV 80% SV

) 70%

60%

50%

40%

Phenotypes (% ratio Phenotypes (% 30%

20%

10%

0% PC BC ThC CRC RCC OvC HCC GBM NSCLC HNSCC

TV SV

Fig.2

a b

RCC CRC 100%

90%

80%

70%

ratio) 60% % % 50% CD31 40% CD31 αSMA 50μM αSMA DAPI 30% DAPI Phenotype ( ThC NSCLC 20%

10%

0% PC BC ThC RCC CRC HCC GBM NSCLC

TV SV

TV 50μM 50μM Intermediate SV

c d

Intermediate 40 500 SV 100% 450 35 ) Mesenchymal 2 90% Indeterminable 400 30 80% Epithelial 350 70% 25 300 60% 50% 250 20 40% Microvessel density density Microvessel 200 SMA (%) positivity

α 15 30% EMT status (% ratio) (number vessels vessels per mm (number 150 20% 10 100 10% 0% 50 5 SV TV 0 0 BC PC RCC ThC CRC HCC BC PC

GBM Phenotype RCC ThC CRC HCC GBM NSCLC (SCC) NSCLC (SCC) NSCLC (Adeno) NSCLC NSCLC (Adeno) NSCLC

Fig. 3

Tumour xenograft models (n=31) a TV SV 0 5 6 9 o 0 3 6 7 5 0 6 9 5 3 7 5 0 0 A 7 G G C PC PC A37 LoV A431 HT2 FaDu Calu HPA Calu A549 A549 HL-6 A278 MCF- Du14 HX14 SW62 U87M HT108 MES-S NCIN8 ZR-75-1 BT474cl HCT11 Colo20 U118M NCIH46 NCIH52 NCI H197 NCI MDAMB231 CWR22Rv1 b TV c SV Calu-6 Calu-3

CD31 CD31 α−SMA α−SMA μ DAPI 200μM DAPI 200 M

d e

Fig.4

a Calu-6 HT-29 SW-620 1.6 1.0 2.0

1.4 0.9

1.2 0.8 1.5

1.0 0.7 mour volume volume mour

mour volume 1.0 mour volume volume mour e.m.) .e.m.) e.m.) u 0.8 0.6 u u u s u TV u

0.6 0.5 +/- +/- s. +/- 3 +/- s. 3 3 0.5 0.4 (cm 0.4 (cm (cm

0.2 0.3 Geometric mean t mean Geometric

Geometric mean t 0.0 Geometric meant 0.0 0.2 0 4 8 12 16 20 0 4 8 1216202428 0481216 Days of treatment Days of treatment Days of treatment

b Calu-3 (1) Calu-3 (2) 1.8 1.8 1.6 1.6 1.4 1.4 1.2 1.2 SV 1.0 1.0 0.8 0.8 +/- s.e.m.)

0.6 3 0.6 (cm3 +/- s.e.m.) (cm3 +/- 0.4 (cm 0.4 0.2 0.2 Geometric mean tumour volume 0.0 Geometric mean tumour volume 0.0 0 1020304050 0 5 10 15 20 25 30 35 Days of Treatment Days of treatment

c Calu-6 Calu-3 (1) Calu-3 (2) 300 50 300

250 ) 40 2

) 250 ) 2 2 m y

m m t t m m y y P=0. 03 P=0.0001 P= 0. 004 30 200 200 SV 150 TV 20 Microvessel densi Microvessel densit Microvessel density 150 (numberper vessels (numberm per vessels

(numberm per vessels 100 10

50 100 0

0 50 Control DC101 Treated 0 Control DC101 Treated

d Calu-6 Calu-3 (1)

Human Human Fold Fold Gene P-Value Gene P-VlValue Change Change CDH8 5.17 0.005 PLAT 1.77 0.032 SGK2 2.57 0.029 VEGFA 1.61 0.009 LPAR1 2.38 0.016 BMP2 1.53 0.037 GAS6 1.70 0.028 MST1 1.51 0.036 LPAR1 1.68 0.034 FGF2 -1.69 0.033 SPP1 -1.98 0.014 CSTA -4.06 0.010 Mouse Fold Gene P-Value Mouse Change Gene Fold P-Value Notch3 -1.50 0.034 Change Mmp9 -1.52 0.020 Cxcl2 7.01 0.006 Cdh3 -1.65 0.034 Mmp9 2.55 0.027 Flt1 -1701.70 00050.005 Il1b 2.11 0.009 Mmp13 -1.72 0.031 Itgam 1.95 0.010 Gpr116 -1.76 0.030 Tie1 -1.63 0.005 Tie1 -1.85 0.021 Flt4 -1.64 0.004 Efnb2 -1.88 0.042 Pdgfb -1.66 0.003 Flt4 -1.91 0.001 Gpr116 -1.80 0.008 Angpt2 -2.03 0.021 Efnb2 -1.92 0.002 Dll4 -2.77 0.000 Kdr -2.02 0.001 Cdh8 -4.03 0.009 Dll4 -3.32 0.001 DC101 Control DC101 Control

Fig. 5

a Calu-3 (1) b

700 T 600 )

2 S 500 Stromal MVD S T 400 P=0.0001 S essel density 1 els per mm v v s s 300 3 3

CD T 200 (number ves Stromal micro Stromal 100 S T

CD31 S 0 T

μ T 50μM 50 M

200 T 180 P=0.04 A rea a a 160 M M S 140 S Stromal MVA α

120 / S per vessel)

2 100 T m CD31/ μ

( 80 60 Stromal microvessel T 40 T 20 0 CD31 CD31 α SMA T 100μM α SMA 100μM DAPI DAPI

c Calu-3 (1) Calu-3 (1)

70 P=0.09 60 50 P=0.001 40 Haematoxylin P=0.001 30 partments (%) roportionsxenograft of m m p p μ co 20 200 M 200μM P=0.35 10 Relative 0

Tumour Stroma Overlay

Necrotic space Compartment

200μM 200μM

d T T S

T T Calu-3 (1) S S Control 100 αSMA S T 90 PDGFRβ 80 T S μ 70 αSMA 100 M PDGFRβ S 100μM 60 50 T P=0.0005 % positivity 40 T 30 T 20 P=0.009 S S T 10 0 S S S

DC101 Treated T S T 100μM αSMA PDGFRβ 100μM

Fig.6

30 PD SD 25 PR CR 20 (n=56)

Patient number Patient 10

0 TVTVSV SD Phenotype

Tumour-stromal architecture can define the intrinsic tumour response to VEGF-targeted therapy

Neil R. Smith, Dawn Baker, Matthew Farren, et al.

Clin Cancer Res Published OnlineFirst September 12, 2013.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-13-1637

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2013/09/13/1078-0432.CCR-13-1637.DC1

Author Author manuscripts have been peer reviewed and accepted for publication but have not yet been Manuscript edited.

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/early/2013/09/12/1078-0432.CCR-13-1637. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.