Relationship between human tumour angiogenic profile and combretastatin-induced vascular shutdown: an exploratory study

Combretastatin-A4-phosphate (CA4P) acts most effectively against immature tumour vasculature. We investigated whether histological angiogenic profile can explain the differential sensitivity of human tumours to CA4P, by correlating the kinetic changes demonstrated by dynamic MRI (DCE-MRI) in response to CA4P, with tumour immunohistochemical angiogenic markers. Tissue was received from 24 patients (mean age 59, range 32–73, 18 women, 6 men). An angiogenic profile was performed using standard immunohistochemical techniques. Dynamic MRI data were obtained for the same patients before and 4 h after CA4P. Three patients showed a statistically significant fall in Ktrans following CA4P, and one a statistically significant fall in IAUGC60. No statistically significant correlations were seen between the continuous or categorical variables and the DCE-MRI kinetic parameters other than between ang-2 and Ktrans (P=0.044). In conclusion, we found no strong relationships between changes in DCE-MRI kinetic variables following CA4P and the immunohistochemical angiogenic profile.

Combretastatin-A4-phosphate (CA4P) is a tubulin-binding, selective vascular disruptive agent (VDA) that causes rapid shutdown of blood flow to tumours with minimal effects on normal vasculature. Tumour endothelial cells are often immunohistochemically immature, lacking smooth muscle and pericyte coverage, and more sensitive to the effects of VDAs probably because the intracellular cytoskeleton is less developed (Tozer et al, 2005). Our aim was to determine human tumour vascular maturity using an immunohistochemical angiogenic profile, and to study any relationship with CA4P response assessed by change in DCE-MRI kinetic variables.

MATERIALS AND METHODS
Following ethics approval, patients enrolled in Mount Vernon phase I CA4P studies (CR-UK PH1/066 & PH1/092) who underwent DCE-MRI analysis were identified. Records were obtained from 26 identified patients and tissue was available for 24 patients (Table 1). A full set of DCE-MRI data was available for 21. Tissue blocks were coded to conceal patient identity. Sections (4 mm) were cut onto slides, and stained with the panel of antibodies at concentrations and with control samples illustrated in Table 2.

Immunohistochemical procedures
All sections were deparaffinised in xylene and rehydrated through graded alcohols to water.
CD34 and aSMA double staining Dako's Cytomation Envision double stain system (K5361) (Dako Ltd., Cambridgeshire, UK), was used performed according to the manufacturer's instructions. Specimens were counterstained using Gills haematoxylin (01500E; Surgipath Europe Ltd.) for 7 s and washed. The samples were mounted and cover slipped using Dako Faramount aqueous mounting medium (S3025).
CD105 We used Dako's Catalysed Signal Amplification staining system (K1497) and followed the manufacturer's instructions exactly.
Ang-1, Ang-2, PDGF, and CD61 Benchside manual staining procedures were used. Endogenous peroxidase activity was quenched with 1.5% hydrogen peroxide in methanol for 10 min. Antigen retrieval was performed by microwaving in 0.01 M citric acid pH 6 for 10 min (ang-1) and 20 min (ang-2). After washing in Tris-buffered saline tween-20 (TBST) primary antibody was applied as shown in Table 2, and incubated overnight at 41C. After rinsing in TBST, Dako Envision anti-rabbit secondary antibody (K5007) with peroxidase was applied for 30 min. Antigen was developed with diaminobenzidine, slides counterstained with haematoxylin, dehydrated, cleared, and mounted. For CD61 and PDGF Vectastain's mouse IgG normal horse serum (Vector Labs, Burlingame, CA, USA) diluted in TBS was applied for 30 min. A horse anti-mouse biotinylated secondary was then applied, followed by incubation with streptavidinperoxidase reagent.

Immunohistochemical profile analysis
Each slide was assessed using a Zeiss Axioskop microscope, with Visilog 5.0 (Noesis) image capture software. The random field   (2000) were used. Any CD34 þ endothelial cells that were distinctly separate from other microvessels at high power were considered individual vessels, and recorded as the number of vessels per high power field. aSMA staining was analysed concurrently -any CD34 þ vessel also aSMA þ was recorded. Mean microvessel density (MVD per mm 2 ) and pericyte coverage index (PCI) were calculated after analysing at least 20 random fields per slide (Table 3). To determine the Ki-67 proliferation index (%) or Glut-1 hypoxia index (%) each slide was scanned at low and high magnification. A semiquantitative visual assessment was made across the slide of the proportion of tumour cells staining positively (Table 3).
Random results were checked by an experienced immunohistochemist (FD), who had not seen the initial data. In case of disagreement as to the correct result, that slide was analysed by a consultant histopathologist.

DCE-MRI data
DCE-MRI involves the rapid acquisition of sets of T1-weighted images through tumour as an intravenous bolus of contrast agent is injected. The change in signal intensity over time can be analysed with mathematical models (Tofts, 1997). Quantitative kinetic variables are derived that provide information about tumour microcirculation as they are indirectly related to perfusion, vascular permeability, and vessel surface area (d' Arcy et al, 2006).
The IAUGC (initial area under the gadolinium concentrationtime curve -units: mM min) is calculated for the first 60 s following Gd-DTPA injection (Evelhoch, 1999;Evelhoch et al, 2004). It has the advantage of being a quantitative variable that is obtained without mathematical modelling or knowledge of the arterial input function. It is a measure of how much contrast agent is taken up by the tumour in the first 60 s post-Gd-DTPA injection and is influenced by tumour blood flow rate and tumour vessel permeability. By using AUC from the Gd-DTPA concentration time curve rather than from the signal intensity time curve, the problem of dependence of signal enhancement on tissue T1 levels is avoided.
The transfer constants for Gd-DTPA diffusion into the tumour extravascular extracellular space (EES) from blood plasma (K trans ) and back again (k ep ) can be calculated using Tofts' mathematical model (Tofts and Kermode, 1991;Tofts et al, 1999). K trans (and k ep ) have both blood flow rate and permeability components as Gd-DTPA is not freely diffusible and their biological meaning is dependent on the balance between capillary permeability and blood flow in the tissue of interest. In high permeability tissues, K trans is equal to the blood plasma flow per unit volume of tissue. In low permeability tissues such as the brain, it is equal to the permeability surface area product between the blood plasma and the EES per unit volume of tissue (Tofts, 1997). In most tumours the situation is likely to be somewhere between these two extremes.
Studies were performed on a 1.5 T scanner as previously described . Images were analysed using specialist software (Magnetic Resonance Imaging Workbench (MRIW), Institute of Cancer Research, London) (d 'Arcy et al, 2006). A dynamic analysis was performed using MRIW for each pixel within the region of interest on each slice. The MRIW converts signal intensities of the T1-weighted DCE-MRI data set into T1 relaxation rates and then into Gd-DTPA concentrations for individual voxels, using the methods described by Parker et al (1997). Standardised terms were employed as defined in Tofts et al (1999). The percentage change from mean baseline kinetic variables following CA4P was calculated (Table 4).

Effects of CA4P
A Gaya et al

Statistical analysis
Measurement error (Bland and Altman, 1996) was used to define confidence intervals for the DCE-MRI variables. Linear regression and correlation analysis were used to test for relationships between the continuous variables and MRI data. Analysis of Variance was used to test for relationships between categorical variables and MRI data.

RESULTS
Tissue was received from 24 patients (mean age 59, range 32 -73, 18 women, 6 men). Eighteen patients were taken from the PH1/066 single agent UK CA4P trial , and 6 from the PH1/092 CA4P þ anti-CEA antibody study (Gaya AM et al, 2008, unpublished). The most common histological subtypes (Table 1) were colorectal adenocarcinoma (7), uterine leiomyosarcoma (4), and ovarian adenocarcinoma (3). The age of the tumour blocks varied between 6 months and 21 years. All but one block contained primary tumour from surgical specimens, not metastatic tissue as patients were not rebiopsied before entrance to the phase I trials.
We only studied patients who received doses of CA4P between 40 and 114 mg m À2 that have been associated with changes in DCE-MRI kinetic variables. There was a wide range of DCE-MRI variable change following CA4P (Table 4). Percentage change in K trans from mean baseline following CA4P varied between þ 111 and À75%, and IAUGC 60 between þ 186 and À90%. In reality, only falls in these parameters (consistent with tumour vascular shutdown) are clinically significant. The 95% confidence interval for K trans change (determined from repeatability analysis in the individual studies and repeated with the combined data) was À62.5 to þ 166.5%, IAUGC 60 ±80%. There was no obvious visual relationship seen between CA4P dose and fall in kinetic parameters (Table 4). Three patients (3, 10, 12) showed significant falls in K trans following CA4P (outside 95% CI), and only 1 patient (12) a significant fall in IAUGC 60 .
No statistically significant correlations were seen between the continuous or categorical variables and the DCE-MRI kinetic variables (Tables 5a and b) other than between ang-2 and K trans (P ¼ 0.044). Inspecting Tables 3 and 4, there are no obvious relationships between the angiogenic profile and DCE-MRI data.

DISCUSSION
We found no obvious relationships between changes in DCE-MRI variables and the immunohistochemical angiogenic profile. This suggests immunohistochemical analysis of patients' tumours is unlikely to be useful in predicting response to VDA therapy. There was enormous variability seen both in the DCE-MRI kinetic variables and the immunohistochemistry. It is difficult to know whether the variability is related to the MRI and immunohistochemical techniques or is due to specific tumour characteristics that are of interest to us ie vascular maturity.
The variability of CD34 and CD105 MVD is consistent with previous studies. Mean microvessel density has been correlated with risk of metastasis, prediction of efficacy of anticancer therapy, and prognosis in a variety of tumour types (Des Guetz et al, 2006;Gadducci et al, 2006;Zhao et al, 2006;Barresi et al, 2007;Calvin et al, 2007;Gulubova and Vlaykova, 2007;Kavantzas et al, 2007;Trivella et al, 2007). The PCI results agree with Eberhard's data (Eberhard et al, 2000). We have found no evidence that mature vasculature with higher PCI responds poorly to CA4P, or that immature CD105 þ vasculature responds better. Tumours with higher CD34 MVD may also have higher CD105 þ MVD, however it does not necessarily follow that active angiogenesis takes place.
There are a number of aspects of this study that undermine the power of our conclusions. We obtained only small numbers of samples with a heterogeneous mix of different tumour types. However, our sample was limited to patients who had received CA4P and undergone DCE-MRI. Within the ANOVA categorical analysis especially, this greatly reduced the statistical power. Our paraffin blocks, up to 21-year old, were taken from the primary surgical procedure, whereas patients received CA4P during phase I trials many years later after metastatic relapse. It is still unclear whether antigenicity within paraffin blocks reduces over time (Fergenbaum et al, 2004). There are also suggestions that the angiogenic phenotype might continuously evolve from primary tumour through to metastasis (Sullivan and Graham, 2007).
There were widespread falls in MRI kinetic variables following CA4P, however for most patients these did not achieve statistical significance. For some this was because the CA4P dose received was below the threshold level for DCE-MRI detectable drug activity of 50 mg m À2 ) -for others the CA4P simply did not have the expected tumour blood flow effect. The reason for administration of the lower CA4P doses in a few patients was the dose escalation protocol for the phase I trial that revealed this phenomenon. 95% CI for K trans change was À62.5 to þ 166.5%, IAUGC 60 ± 80%. Conversely, taking the two separate trials individually, in PH1/066 the lower 95% confidence limit for K trans was À44% and in PH1/092 it was À76.5%. This wide DCE-MRI parameter variability is in line with previously published data (Padhani et al, 2002;Taylor et al, 2004Taylor et al, , 2006. Combining DCE-MRI data can adversely affect statistical confidence intervals due to the slight differences in sequences and calibration methods used (Taylor et al, 2006). There was wide intrapatient variability, patient motion, and an estimated arterial input function. Modified Fritz-Hansen arterial input coefficients, which are currently used, demonstrate better reproducibility (Fritz-Hansen et al, 1996;Taylor et al, 2007). Averaging the kinetic parameters over a whole lesion also dilutes the effect of CA4P central vascular shutdown because of relative sparing at the tumour periphery.
There was no obvious visual relationship between CA4P dose and fall in MRI parameters. Most patients treated with  Table 4 there is clearly no correlation, confirmed statistically. Several patients received lower, possibly inactive doses of CA4P. A threshold relationship for DCE-MRI detectable CA4P activity, with patients receiving 450 mg m À2 exhibiting greater parameter falls has previously been demonstrated Rustin et al, 2003). Microvessel density does not reflect the functional properties of vessels, including permeability, which contribute to DCE-MRI variables. Studies reporting correlation between MRI variables and MVD have therefore found only moderate associations (Su et al, 2003;Tuncbilek et al, 2004). Several studies (Buckley et al, 1997;Stomper et al, 1997;Su et al, 2003) show conflicting data relating immunohistochemical expression with DCE-MRI data. The different analysis methods used may have affected the results. Hawighorst et al (1998) showed that MRI parameters were better indicators of patient survival than MVD or VEGF.
The significant correlation between ang-2 and K trans is probably a statistical quirk due to the number of variables investigated. However, it might be explained as: angiopoietins are involved in the angiogenic switch (Ahmad et al, 2001;Tanaka et al, 2003). Ang-1 stabilises vessels by maintaining pericyte coverage; ang-2 promotes pericyte removal, which in the presence of VEGF facilitates the angiogenic response and in the absence of VEGF induces vessel regression (Moon et al, 2006;Bach et al, 2007;Shim et al, 2007;Winter et al, 2007). Depending on the tumour model, stabilisation of blood vessels by ang-1 may either promote tumour angiogenesis or reduce tumour growth (Metheny- Barlow and Li, 2003). Wang (Wang et al, 2007) has shown ang-2 correlates with active angiogenesis in human colorectal carcinomas, and that ang-2 also correlates with VEGF. Thus, unstable immature ang-2 expressing vasculature undergoing neoangiogenesis is probably more susceptible to the effects of VDAs and thus detectable by DCE-MRI.
In conclusion, there was wide intrapatient variability in angiogenic profile expression and DCE-MRI kinetic variable changes following CA4P. There was no obvious correlation between falls in MRI variables and vascular markers, however sample numbers were small. Immunohistochemical analysis of patients' tumours is unlikely to be useful in predicting response to VDA therapy. This was an exploratory study, which we hoped would suggest some leads for further analysis. Although one could postulate that this study should be expanded to look at larger patient numbers and use fresh tumour samples from a homogeneous group, such a study is unlikely ever to be performed.

Effects of CA4P
A Gaya et al