IJC International Journal of Cancer

Proangiogenic tumor proteins as potential predictive or prognostic biomarkers for bevacizumab therapy in metastatic colorectal cancer Maressa A. Bruhn1, Amanda R. Townsend2, Chee Khoon Lee3, Aravind Shivasami1, Timothy J. Price2,4, Joe Wrin1, Georgia Arentz1,5, Niall C. Tebbutt6, Christopher Hocking2, David Cunningham7, and Jennifer E. Hardingham1,5 on behalf of the BHI in collaboration with AGITG 1

Haematology-Oncology Department, Basil Hetzel Institute,The Queen Elizabeth Hospital, Woodville, SA, Australia Medical Oncology, The Queen Elizabeth Hospital, Woodville, SA, Australia 3 NHMRC Clinical Trials Centre, University of Sydney, Sydney, NSW, Australia 4 School of Medicine, Haematology- Oncology Department, University of Adelaide, North Terrace, Adelaide, SA, Australia 5 School of Medical Sciences, Department of Physiology, University of Adelaide, North Terrace, Adelaide, SA, Australia 6 Ludwig Oncology Unit, Austin Health, Heidelberg, VIC, Australia 7 Royal Marsden Hospital, Department of Medicine, Royal Marsden NHS Foundation Trust, Sutton, Surrey, United Kingdom 2

Key words: angiogenic proteins, predictive biomarkers, bevacizumab therapy Additional Supporting Information may be found in the online version of this article Conflict of interest: Professor Cunningham has declared a potential financial conflict of interest as he is in receipt of research funding from Amgen, Roche, Celgene, Sanofi Aventis, Merck Serono, Novartis and Astra Zeneca. No other authors have any conflict of interest. Grant sponsor: South Australian Health and Medical Research Institute/Cancer Council of South Australia; Grant number: 1028595 DOI: 10.1002/ijc.28698 History: Received 30 Oct 2013; Accepted 5 Dec 2013; Online 26 Dec 2013 Correspondence to: Dr. J. Hardingham, Level 1 BHI, The Queen Elizabeth Hospital, 28 Woodville Road, Woodville, SA 5011, Australia, Fax: 161-8-82227872, E-mail: [email protected]

C 2013 UICC Int. J. Cancer: 135, 731–741 (2014) V

Metastatic colorectal cancer (mCRC) is a leading cause of cancer death worldwide, and the second most common cause of cancer death in Australia.1 Tumor growth and metastasis depends on the formation of new blood vessels (angiogenesis) for the supply of oxygen and nutrients and this process is largely driven by tumor production of vascular endothelial growth factor (VEGF).2 Recently, the use of targeted therapy to treat mCRC, such as the monoclonal antibody bevacizumab, offers promise in improving patient outcomes. Bevacizumab is a recombinant humanized monoclonal IgG1 antibody that selectively binds to and ablates the proangiogenic activity of the major human VEGF, VEGF-A,3 and is now commonly incorporated into combination chemotherapy regimens for mCRC. Although statistically significant, the improvement in progression-free survival (PFS) and overall survival (OS) with anti-VEGF therapy added to chemotherapy is modest, with key trials demonstrating a 2- to 5-month survival improvement compared to chemotherapy alone.4–6 A significant number of patients do not achieve any meaningful

Cancer Therapy

Tumor biomarkers to more accurately predict a patient’s response to a given therapy are much needed in oncology practice. For metastatic colorectal cancer the anti-vascular endothelial growth factor (VEGF) monoclonal antibody bevacizumab is now commonly included in first-line therapy regimens and has led to modest but significant improvements in patient outcomes compared with chemotherapy. Given the modest gains there is a pressing need for predictive biomarkers to better identify patients who would benefit from this targeted therapy. We used a multiplex protein assay to determine the tumor expression levels of the proangiogenic proteins IL-6, IL-8, bFGF, PDGF-BB and VEGF-A in formalin-fixed paraffin-embedded tumors from the MAX clinical trial patients with available tissue samples. Patients were dichotomized into “low” vs. “high” expression subgroups based on median baseline levels to correlate with objective response rate (ORR), progression-free survival (PFS) and overall survival (OS). “Low” tumor VEGF-A level was predictive of better ORR for bevacizumab [ORR (low) 53% vs. (high) 19%, interaction p 5 0.03] but not for PFS [hazard ratio, HR (low) 0.73 vs. (high) 0.62, interaction p 5 0.68] in the comparison of capecitabine (C) versus C and bevacizumab (CB) and CB plus mitomycin (M). When analyzed as a dichotomized variable, “high” VEGF-A was prognostic for shorter PFS (unadjusted HR 1.34, p 5 0.06; adjusted HR 1.55, p 5 0.008). The other four proteins were neither predictive of bevacizumab benefits nor prognostic for ORR, PFS or OS. “Low” tumor VEGF-A was associated with longer PFS after adjustment for other baseline factors. Proangiogenic proteins were not predictive of benefit with bevacizumab for PFS.

732

Proangiogenic proteins and bevacizumab therapy in mCRC

Cancer Therapy

What’s new? There is a pressing need for predictive biomarkers to better identify metastatic colorectal cancer patients who would benefit from anti-VEGF monoclonal antibody bevacizumab therapy. This study is the first to measure the expression levels of a panel of angiogenic proteins from FFPE tumors and to also use a multiplex assay platform--an advantage given the limited amount of tissue available from clinical trials. Low tumor VEGF-A was associated with significantly longer progression free survival after adjustment for other baseline factors. However neither VEGF-A, nor the other angiogenic proteins IL-6, IL-8, bFGF or PDGF-BB, were predictive of outcome for bevacizumab therapy.

benefit. This could be due to inherent resistance mechanisms to evade the inhibition of VEGF via activation of pathways leading to tumor expression of compensatory proangiogenic factors7,8 or by tumor recruitment of myeloid cells secreting proangiogenic factors.9 Such factors may prove useful as predictive biomarkers to identify those patients likely to show tumor resistance so that this therapy can be used in the most appropriate and cost-effective way. In addition, these biomarkers may also provide prognostic information. In contrast to anti-VEGF antibody therapy, selection of patients for antiepidermal growth factor receptor (EGFR) antibody treatment with predictive biomarkers is mandatory, as KRAS exon 2 mutation reliably identifies patients who do not benefit from this therapy.10,11 More recent evidence suggests that KRAS exon 3 or 4 mutation12 or NRAS mutation13 also predicts for lack of benefit, and is likely to be included in future routine clinical practice. There remains, however, an urgent need for predictive biomarkers for anti-VEGF therapies. Given the cross-talk between activated EGFR signaling pathways and tumor angiogenesis,14 we have previously investigated the predictive and prognostic impact of KRAS and BRAF gene mutation status and PTEN copy number status in patients receiving capecitabine with bevacizumab or capecitabine without bevacizumab in the phase III AGITG (Australasian Gastrointestinal Trials Group) MAX study: neither KRAS or BRAF mutation nor PTEN loss was predictive of bevacizumab treatment outcome.15,16 For this analysis, we proposed to investigate the association of activated EGFR pathway-induced proangiogenic molecules on the response to bevacizumab. The downstream activation of mTORC1 and induction of hypoxiainducible factor 1 (HIF-1) a17 results in translation of proangiogenic proteins including VEGF-A, platelet-derived growth factor (PDGF)-BB, interleukin (IL)-8, basic fibroblast growth factor (bFGF), angiopoietin (Ang)-2 and IL-6.18–20 Further, NF-kB is activated via PI3K-AKT signaling or by tumor hypoxia, independently of HIF-1a, and also results in production of angiogenic factors.21 In tumor-associated fibroblasts the EGFR pathway is activated by the increased amounts of the EGFR ligand TGFa in the tumor microenvironment resulting in increased angiogenic factors including IL6, IL8 and PDGFBB.22 In addition, IL8 is known to regulate endothelial cell proliferation and tube formation,23 and has been reported to rescue tumor angiogenesis from VEGF inhibition.24 We hypothesized that overexpression of such proangiogenic factors would diminish the efficacy of anti-VEGF ther-

apy and represent biomarkers of resistance to such therapy. One aim of this study was to develop a robust multiplex assay to measure the levels of the proangiogenic proteins IL-6, IL-8, bFGF, PDGF-BB and VEGF-A that may be associated with bevacizumab response using archived formalin-fixed paraffinembedded (FFPE) tumor tissue from the MAX study patient cohort. These markers were chosen as representative of the angiogenic molecules activated in this setting, and that were of a suitable molecular weight (MW) to be effectively solubilized in FFPE tissue lysates. A further consideration was that the corresponding antibody had been validated for use in a commercially available multiplex suspension array format. The second aim was to assess whether a defined level of one or more proangiogenic proteins was predictive of bevacizumab benefit and associated with better prognosis.

Material and Methods Trial design and tumor collection

Tissue samples were obtained from the MAX clinical trial conducted by the AGITG with mCRC patients recruited in Australia, New Zealand and the United Kingdom. The MAX study design and results have been reported previously.6 The study included 471 patients with previously untreated mCRC and 314 patients gave written informed consent for translational biomarker studies at the time of study enrolment. Ethics approval for translational studies was obtained centrally from the Cancer Institute of NSW Ethics Committee. Eligible patients were enrolled in this trial between July 2005 and June 2007 and randomly assigned to receive capecitabine (C); capecitabine and bevacizumab (CB) and capecitabine, bevacizumab and mitomycin C (CBM) (Fig. 1). Patients were evaluated for tumor response or progression every 6 weeks by means of radiologic imaging. Treatment was continued until the disease progressed or until the patient could not tolerate the toxic effects. The addition of bevacizumab to capecitabine, with or without mitomycin, significantly improved PFS and OS. FFPE sections of tumor tissue from archival specimens collected at the time of enrolment were retrieved from storage at various participating hospital pathology departments by MAX trial investigators. Assays of tissue samples were performed by investigators blinded to trial endpoints. Protein preparation

Protein lysates were prepared from 2–3 3 10 mm FFPE sections from each primary tumor mounted on glass slides. C 2013 UICC Int. J. Cancer: 135, 731–741 (2014) V

733

Bruhn et al.

Paraffin was removed by xylene and the tissue rehydrated through graded ethanol to Milli-Q water. The tissue was scraped from the slide into a sterile eppendorf tube, weighed and incubated with 2% w/v lysis buffer EB (2% SDS, 200 mM DTT and 20 mM Tris-HCL, pH 8.8),25 at 100 C for 20 min and then at 80 C for 2 hr with agitation. The lysates, mixed with 13 HaltTM protease inhibitor cocktail (EDTAfree) (Pierce Biotechnology, Rockville, IL), were clarified by centrifugation at 14,000 rcf at 4 C for 30 min and transferred into Bio-PlexV lysis buffer (Bio-Rad, Carlsbad, CA) using the ReadyPrepTM 2-D Cleanup kit (Bio-Rad). Protein was quantified using the EZQ protein assay kit (Life Technologies, Carlsbad, CA). R

(WB). Primary antibodies used were anti-b actin mouse monoclonal (MAB47778) and anti-STAT3 mouse monoclonal (MAB1799) antibodies (both from R&D Systems, Minneapolis, MN). WB was performed using the ECL Plus Western Blotting Detection reagents (GE Healthcare, Buckinghamshire, UK) and an HRP-conjugated goat-anti-mouse secondary antibody (R&D Systems). Chemiluminescence was detected using the LAS4000 imager (GE Healthcare). Protein MW was estimated by comparison to a prestained protein ladder (Benchmark, Life Technologies). R

Bio-PlexV suspension array analyses

Protein lysates were assayed for the candidate biomarkers using custom Bio-PlexV 5-plex arrays (Bio-Rad) to detect IL6, IL-8, bFGF, PDGF-BB and VEGF-A. Protein lysates were diluted to a final concentration of 450 mg/mL, and samples were assayed in duplicate using the Bio-PlexV 200 System instrument (Bio-Rad) on the low photomultiplier tube (PMT) setting according to the manufacturer’s instructions. Data were analyzed using the Bio-PlexV Manager 6.0 software R

SDS PAGE and Western blotting

To determine the effectiveness of protein solubilization, each protein lysate (20 lg) was run on replicate 4–20% gradient SDS-PAGE gels (NuSep, Sydney, Australia) for both protein visualization after staining with CoomassieV Brilliant Blue G (Sigma-Aldrich, Sydney, Australia) and for Western blotting R

C 2013 UICC Int. J. Cancer: 135, 731–741 (2014) V

R

R

Cancer Therapy

Figure 1. CONSORT diagram MAX AGITG TRIAL: proangiogenic protein biomarker analysis.

734

Proangiogenic proteins and bevacizumab therapy in mCRC

(Bio-Rad). A five-parameter logistic regression model (5PL) was used to generate standard curves and results were expressed as pg/mL. Noting that immunohistochemistry (IHC) is generally used for assessing protein overexpression and the BioplexV method is novel, we also examined for concordance between VEGF-A level and IHC expression of VEGF-A from tissue microarrays generated from the MAX trial tissue blocks for an earlier project. R

remaining patients could not be retrieved or were not suitable for Bio-PlexV analyses. Amongst those who were excluded from this biomarker study (Table 1), there were more patients with ECOG 2 performance status (p 5 0.002), more patients with locally advanced disease of the colon or rectum (p < 0.001) and fewer patients had resection of their primary tumor (p < 0.001). Baseline characteristics were otherwise similar for patients included in this analysis, and for those excluded owing to missing Bio-PlexV data (Table 1). R

R

Statistical analysis R

All randomly assigned patients where Bio-PlexV data were available were included in the analysis. PFS, the primary endpoint, was defined as the time from randomization until documented evidence of disease progression, the occurrence of new disease or death from any cause. The secondary endpoints were OS, defined as the time from randomization until death from any cause, and objective response rate (ORR), defined according to the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.0. The Bio-PlexV analytes were analyzed as continuous and categorical dichotomized variables. As there are no widely accepted cutoff points adopted for any of the Bio-PlexV analytes in colorectal tumors, we used the medians of distribution of each of these analytes as cutoff points for dichotomization into “high” vs. “low” level. We also examined for association between VEGF-A level and IHC expression of VEGF-A from tissue microarray generated from the tissue blocks for a previous analysis. The PFS of patients according to each protein level and treatment groups were summarized with the use of Kaplan– Meier curves, and the difference between these groups was compared (C vs. CB and CBM) with the use of the log-rank test. A proportional hazards model with treatment covariate (C vs. CB and CBM), protein level and their interaction was used to assess whether any of the proteins were predictive of bevacizumab treatment efficacy. To assess whether any of the proteins were independent prognostic factors, multivariate proportional hazards regression models were fitted to data for all patients, with protein level and other previously identified significant prognostic baseline covariates.6 Similar methodologies were adopted for the assessment of the predictive and prognostic values of these biomarkers for OS and ORR. All reported p values were two sided and there was no correction for multiple comparisons. R

Cancer Therapy

R

Results Characteristics of the patients

Of 471 patients who underwent random assignment, a total of 196 primary tumor specimens (n 5 60 from the C group, n 5 68 from the CB group and n 5 68 from CBM group) were examined for proangiogenic protein biomarkers (accounting for 41.5% of the total study population; Fig. 1). The median follow-up time of these patients was 30.2 months (range 0.6–42.4 months). Tumor specimens from the

Isolation of total protein from FFPE and Western blot analysis

The median protein yield from 2–3 3 10 mm sections was 95.4 mg (range 16–274.5 mg) yielding sufficient protein for the assays for all tumors. The stained SDS-PAGE gels showed that in the main, only protein of MW < 70 kDa was solubilized and appeared as bands on the gel. A substantial amount of crosslinked protein still remained in or close to the wells (Supporting Information Fig. 1A). A strong single band for b-actin of the expected MW of 43 kDa was present in the FFPE samples (Supporting Information Fig. 1B), whereas no discrete band for the higher MW protein STAT3 (MW 89 kDa) was obtained (data not shown). Because of the nature of the samples (FFPE) it was inevitable that some proteins would not be suitable for analysis owing to the extensive cross-linking of proteins, particularly the higher MW proteins, as a result of varying lengths of time in formalin fixative before paraffin embedding25; hence, only proangiogenic proteins < 70 kDa were selected for BioplexV analysis. R

Proangiogenic protein levels

The median concentrations of IL-6, IL-8, bFGF, PDGF-BB and VEGF-A were 0.91 pg/mL (range 0.06–4.29), 1.21 pg/mL (0.05–9.74), 5.12 pg/mL (0.34–10.05), 4.99 pg/mL (0.09–15.28) and 5.73 pg/mL (1.12–18.55), respectively. The distributions of each biomarker according to treatment arms are summarized in Supporting Information Figure 2. Between C vs. CBM treatment arms there were significant mean differences for IL-6 (1.35 pg/mL, p < 0.0001), IL-8 (0.81 pg/mL, p 5 0.001) and VEGF-A (2.24 pg/mL, p < 0.0001) concentrations. There were no significant differences for bFGF (0.16 pg/mL, p 5 0.67) and PDGF-BB (0.90 pg/mL, p 5 0.18) concentrations between the treatment arms. Between C vs. CB treatment arms, there were no significant mean differences for IL-6 (0.14 pg/mL, p 5 0.50), IL-8 (0.001 pg/mL, p 5 1.00), bFGF (0.11 pg/mL, p 5 0.78), PDGF-BB (0.11 pg/mL, p 5 0.88) and VEGF-A (0.79 pg/mL, p 5 0.18) concentrations. When comparing protein concentration to IHC 0, 11, 21 and 31 there was no correlation between VEGF-A protein level and VEGF-A IHC expression (r 5 0.03, p 5 0.70; Supporting Information Fig. 3). Predictive value of the proangiogenic proteins

The benefit of bevacizumab on PFS was not significantly different among the patients with “high” vs. “low” levels of the C 2013 UICC Int. J. Cancer: 135, 731–741 (2014) V

735

Bruhn et al.

Table 1. Patient demographic and clinical characteristics Number of all patients (%), N 5 472

Included patients (%), N 5 196

Excluded patients (%), N 5 276

p-Value1

Median

67

68

67

0.62

Range

32–86

36–84

32–86

Male sex

295 (63)

114 (58)

181 (66)

0–1

263 (56)

126 (64)

137 (50)

2

209 (44)

70 (36)

139 (50)

Capecitabine dose 2 g per m2 per day

315 (67)

129 (66)

186 (67)

0.72

Disease-free interval > 12 months

125 (27)

59 (31)

66 (24)

0.09

Prior adjuvant chemotherapy

104 (22)

44 (22)

60 (22)

0.86

Prior radiotherapy

59 (13)

18 (9)

41 (15)

0.07

Caecum

49 (10)

20 (10)

29 (11)

0.92

Ascending colon

48 (10)

26 (13)

22 (8)

0.06

Age (years)

ECOG performance status

0.10 0.002

Primary site of cancer

Transverse colon

28 (6)

12 (6)

16 (6)

0.88

Descending colon

16 (3)

8 (4)

8 (3)

0.48

Sigmoid colon

139 (29)

62 (32)

77 (28)

0.38

Rectosigmoid colon

54 (11)

25 (13)

29 (11)

0.45

Rectum

107 (23)

36 (18)

71 (26)

0.06

Primary tumor resected

371 (79)

185 (94)

186 (67)

5.73: 30 vs. 48%; p 5 0.02 for the interaction between VEGF-A level and the assigned treatment). There were no significant differences in the treatment effect on ORR for the other proangiogenic proteins; however, PDGF-BB “low” patients showed a trend towards a better ORR (Table 2).

C 2013 UICC Int. J. Cancer: 135, 731–741 (2014) V

Prognostic value of the proangiogenic proteins

When VEGF-A was analyzed as a categorical variable, the median PFS was 9.3 months among the patients with tumor VEGF-A “low” (5.73 pg/mL) compared with 7.2 months among those with VEGF-A “high” (>5.73 pg/mL), HR 1.34, 95% CI 0.99–1.81, p 5 0.06 (Fig. 3). In multivariate analysis, VEGF-A was a significant prognostic factor for PFS after adjustment for other baseline factors (HR of VEGF-A “high” vs. “low,” 1.55, 95% CI 1.12–2.13, p 5 0.008) (Table 3). When VEGF-A was analyzed as a continuously measured variable, it was not significantly associated with PFS (unadjusted HR 1.03, 95% CI 0.98–1.08, p 5 0.31; adjusted HR 1.03, 95% CI 0.98–1.08, p 5 0.30 for 1 pg/mL increase). There were no significant differences in PFS or OS between different levels of the other proangiogenic proteins in unadjusted univariate or adjusted multivariate analyses

Cancer Therapy

Extent of disease at baseline

736

analyzed as a categorical variable (Table 3) or continuous variables (results not shown).

Discussion

“low” VEGF-A tumor expression appeared to be an independent prognostic factor for longer PFS, this finding could not be confirmed when analyzed as a continuous variable for every 1 pg/mL increase in unadjusted or adjusted analyses. The lack of predictive value of proangiogenic proteins in our study is consistent with prior studies that failed to demonstrate the predictive ability of any biomarker, including VEGF-A, for anti-VEGF therapy. Jubb et al. assessed VEGFA expression by IHC and in situ hybridization in 278 mCRC patients (153 bevacizumab and 125 placebo) and found that

Cancer Therapy

In this analysis of the MAX clinical trial cohort, baseline tumor VEGF-A “low” was associated with a higher ORR in bevacizumab treatment group but this finding was not translated into PFS or OS benefit of bevacizumab anti-VEGF therapy. The other four proteins were not predictive of bevacizumab benefit on ORR, PFS and OS outcomes. Although

Proangiogenic proteins and bevacizumab therapy in mCRC

Figure 2. Forest plots showing hazard ratios (HRs) for (a) progression-free survival and (b) overall survival subgroup analyses by proangiogenic biomarker levels.

C 2013 UICC Int. J. Cancer: 135, 731–741 (2014) V

737

Cancer Therapy

Bruhn et al.

Figure 2. Continued.

VEGF-A was not a predictive marker for bevacizumab efficacy.26 A recent study of plasma VEGF-A levels (measured by ELISA) dichotomized into “low” and “high” by median cutoff amongst 813 previously untreated mCRC patients enrolled in phase III trials of bevacizumab failed to show plasma VEGF-A was predictive for PFS or OS.27 Our analysis has shown a significantly improved ORR for patients with VEGF-A “low” compared to “high” tumors in those treated with bevacizumab compared to chemotherapy alone. Given the MAX clinical trial randomized patients to three arms, the two bevacizumab arms were pooled for this C 2013 UICC Int. J. Cancer: 135, 731–741 (2014) V

analysis. Analysis of the CBM and CB groups individually suggests that the differences in ORR in the VEGF-A “low” vs. “high” groups appear driven by the patients treated with CBM (ORR 55 vs. 18%). In the CB-treated patients there is no difference in ORR between the groups (ORR 26 vs. 35%). This raises the possibility of an interaction between mitomycin, bevacizumab and VEGF-A levels. In a preclinical study, mitomycin C treatment has been shown to increase the levels of VEGF-A in rat bladder cancer cells, suggesting that mitomycin C may increase the angiogenic potential of cancer cells.28 In contrast, mitomycin C was shown to enhance

738

Proangiogenic proteins and bevacizumab therapy in mCRC

Table 2. Response rate by proangiogenic protein levels Treatment

n

%

N

IL-6  0.91, n 5 97 C

4

%

p3

51

0.20

IL-6 > 0.91, n 5 97 10

23

CB

8

19

19

42

CBM

30

71

3

7

Overall

42

431

45

461

IL-8  1.21, n 5 63

IL-8 > 1.21, n 5 92

C

4

16

13

31

CB

8

32

10

24

CBM

13

52

19

45

Overall

25

401

42

461

bFGF  5.12, n 5 94 C

13

bFGF > 5.12, n 5 94 28

14

35

CB

17

37

10

25

CBM

16

35

16

40

40

431

Overall

46

1

49

PDGF-BB  4.99, n 5 78 10

26

15

41

CB

13

33

13

35

CBM

16

41

9

24

Overall

39

501

37

411

VEGF-A  5.73, n 5 96 9

0.29

PDGF-BB > 4.99, n 5 91

C

C

0.27

0.09

VEGF-A > 5.73, n 5 97 19

19

48

CB

12

26

14

35

CBM

26

55

7

18

Overall

47

491

40

411

0.032

1

% indicates the proportion of responders according to the different levels of proangiogenic proteins. Response rate was not evaluable in 18 V patients who had Bioplex analysis. 2 Results were confirmed when additional analysis comparing C versus CBM only was undertaken (p 5 0.02 for the interaction between VEGF-A level and the assigned treatment). 3 p-Value for interaction between biomarker status and the assigned treatment (C vs. CB and CBM).

Cancer Therapy

R

apoptosis of T24 bladder cancer cells when VEGF-C was inhibited by RNAi29: this may be the mechanism for a better response in our patient cohort treated with bevacizumab and mitomycin C. However, as VEGF-A was measured as a pretreatment level in our study, the exact interaction with mitomycin C remains unclear and will require confirmation on a larger cohort of patients treated with bevacizumab with and without mitomycin C. Confirmatory studies are also important as the benefit seen for ORR in this analysis was not translated into PFS or OS benefit of bevacizumab therapy. It seems that the biology of VEGF and effect of antiVEGF antibody is different among different tumor types30 and in different populations. Among patients with metastatic gastric cancer in the AVAGAST trial, patients with “high” (above the median) baseline plasma levels of VEGF-A were more sensitive to bevacizumab treatment as measured by OS (HR 0.72; 95% CI 0.57–0.93) than those with “low” VEGF-A levels (HR 1.01; 95% CI 0.77–1.31; p interaction 5 0.07).

Figure 3. Progression-free survival according to VEGF-A level. Dashed line represents VEGF-A “high” > 5.73 pg/mL, solid line represents VEGF-A “low”  5.73 pg/mL.

C 2013 UICC Int. J. Cancer: 135, 731–741 (2014) V

739

C 2013 UICC Int. J. Cancer: 135, 731–741 (2014) V

However, in the non-Asian subgroup this OS benefit from bevacizumab in those with “high” VEGF-A was even more pronounced (p 5 0.04).31 In contrast, in our study in Australian (predominantly non-Asian) mCRC patients the subset with “low” tumor baseline VEGF-A showed an improved ORR for bevacizumab treatment arms, but not overall longer PFS, possibly reflecting the difference in tumor types. In our study, “high” baseline tumor VEGF-A was a significant independent prognostic factor for shorter PFS when analyzed as a dichotomized variable but not as a continuous variable. The study by Hegde et al. also reported that higher plasma VEGF-A was prognostic for shorter OS.27 High VEGF expression, as determined by IHC, has been associated with poorer prognosis of CRC in two small studies32,33; however, this was not confirmed in other studies using IHC.34–36 Furthermore, Martins et al. reported a series of 672 CRC patients in which tumor IHC for several angiogenic proteins was performed: they found VEGF-C but not VEGF-A to be a prognostic factor for OS in stage III rectal cancer.37 We found, using a multiplex suspension array platform and solubilized tumor protein, that baseline VEGF-A level was not correlated with VEGF-A IHC expression as reported in a prior analysis from this trial cohort38 (Supporting Information Fig. 3). The discrepancy is most likely due to the fact that protein from the entire tumor tissue including tumor stroma was assayed by BioplexV in this study compared to the IHC study where only the malignant epithelial glands were scored. This would account for the lack of correlation as stromal cells within the tumor contribute to the level of VEGF-A (and other proangiogenic molecules)8,39,40 and this aspect needs to be considered when attempting to validate other methods of protein expression analysis such as BioplexV with IHC, considered the “gold standard.” In support of our finding several studies did show that circulating VEGF-A was a prognostic biomarker in CRC. In a study analyzing serum samples from 132 CRC patients undergoing curative resection, Kwon et al. showed that high preoperative VEGF levels were associated with increased tumor size and higher CEA levels, and in multivariate analysis, high VEGF-A was an independent prognostic factor for shorter OS (HR 4.779, 95% CI 1.15–19.94, p 5 0.032).41 More recently, in a retrospective subset analysis of more than 2,000 mCRC patients, high baseline levels of serum/plasma VEGF-A and CEA were confirmed as prognostic biomarkers for poorer PFS and OS in patients with mCRC, independent of treatment.42 These findings in serum or plasma concur with the results reported here for solubilized tumor protein. Although there have been many studies in a variety of cancers suggesting a role for other cytokines and angiogenic factors as potential predictive factors for anti-VEGF efficacy, none have been translated into clinical use.24,43–46 We also did not find any significant association with “high” or “low” levels of IL-6, IL-8, bFGF or PDGF-BB with ORR, PFS or OS, although PDGF-BB “low” patients showed a trend towards a better ORR to regimens containing bevacizumab. R

R

Cancer Therapy

Analysis unadjusted (univariate) or adjusted (multivariate) for factors known to be statistical significant at baseline for progression-free survival: treatment assignment, performance status, resection of primary tumor, number of metastatic site of disease, serum alkaline phosphatase and serum bilirubin. 2 Analysis unadjusted (univariate) or adjusted (multivariate) for factors known to be statistically significant at baseline for overall survival: performance status, resection of primary tumor, serum alkaline phosphatase and BRAF mutation status. 3 When VEGF A analyzed as a continuous variable, unadjusted HR 1.03, 95% CI 0.98–1.08, p 5 0.31; adjusted HR 1.03, 95% CI 0.98–1.08, p 5 0.30 for 1 pg per mL increase.

1

1.00 1.00 (0.71–1.41) 0.89 (0.64–1.25) 0.0083 1.55 (1.12–2.13) 193 VEGF-A > 5.73 vs.  5.73 pg/mL

1.34 (0.99–1.81)

0.063

0.50

0.97

0.64 0.92 (0.65–1.30)

0.99 (0.69–1.43) 0.99

0.73 0.94 (0.67–1.32)

1.00 (0.70–1.43)

0.70

0.54

1.06 (0.78–1.45)

1.11 (0.79–1.57)

0.46

0.50

1.12 (0.83–1.52) 188

169

bFGF > 5.12 vs.  5.12 pg/mL

PDGF-BB > 4.99 vs.  4.99 pg/mL

1.12 (0.81–1.53)

0.47

0.30 0.83 (0.59–1.17)

0.87 (0.59–1.28) 0.34

0.18 0.80 (0.57–1.11)

0.83 (0.57–1.22)

0.42

0.32

0.85 (0.58–1.26)

0.82 (0.57–1.20) 0.50

0.99 1.00 (0.74–1.34)

0.89 (0.63–1.25)

194

155

IL-6 > 0.91 vs.  0.91 pg/mL

IL-8 > 1.21 vs.  1.21 pg/mL

p-Value HR (95% CI) p-Value HR (95% CI) p-Value HR (95% CI) p-Value HR (95% CI) N

Adjusted analysis Unadjusted analysis

1

Progression-free survival

Table 3. Baseline proangiogenic proteins and the risk of disease progression and death for all patients

1

Unadjusted analysis

Overall survival

2

Adjusted analysis2

Bruhn et al.

740

Proangiogenic proteins and bevacizumab therapy in mCRC

There was no evidence for a prognostic role for any of these four proteins as measured in the MAX trial patients. Our results based on dichotomization of VEGF-A above and below the median could still represent a chance finding as there is no significant association with PFS when analyzed as a continuous variable for every 1 pg/mL. As alluded to in a recent review one of the challenges in biomarker assessments is the validation of cutoff values.44 We used the median of the samples for each analyte as others have done.31,42,45 One study showed that when analyzed by quartile, the patients with plasma VEGF-A levels in the highest two quartiles experienced most benefit from bevacizumab.31 Whether alternate methods of categorization would have changed the results reported here is unclear, and future research is required to address this. To our knowledge, this is the first study to measure protein levels of proangiogenic proteins in FFPE colorectal tumor tissue using a multiplex suspension array system, and has validated the use of archival FFPE tissue as a source of protein suitable for this platform. Measuring protein levels

from FFPE tissue is important as for most clinical trials to date, FFPE tissue has been the only option for retrospective translational biomarker studies and as the tissue is often in limited amount, the ability to multiplex the analytes is a distinct advantage. Importantly, assaying tumor cell lysates for angiogenic factors should result in much less variation owing to platelet activation, when compared to plasma samples. Platelet-derived angiogenic cytokines such as VEGF and PDGF are released from activated platelets and particularly from EDTA-activated platelets. As there is a large variation in susceptibility to platelet EDTA activation among individuals this would make plasma measurements of angiogenic cytokines difficult if not impossible to interpret.46 In this study, none of the proangiogenic proteins were predictive of benefit with bevacizumab for PFS. Ongoing research that correlates biomarker expression with clinical outcomes in patients with mCRC treated with bevacizumab in further patient cohorts is required to improve patient selection for this treatment.

References

Cancer Therapy

1.

Australian Institute of Health and Welfare & Australasian Association of Cancer Registries 2010. Cancer in Australia: an overview, 2010. Cancer series no. 60, 2010. 2. Folkman J. Tumor angiogenesis: therapeutic implications. N Engl J Med 1971;285:1182–6. 3. Ferrara N, Hillan KJ, Gerber H-P, et al. Discovery and development of bevacizumab, an antiVEGF antibody for treating cancer. Nat Rev Drug Discov 2004;3:391–400. 4. Hurwitz H, Fehrenbacher L, Novotny W, et al. Bevacizumab plus irinotecan, fluorouracil, and leucovorin for metastatic colorectal cancer. N Engl J Med 2004;350:2335–42. 5. Saltz LB, Clarke S, Dıaz-Rubio E, et al. Bevacizumab in combination with oxaliplatin-based chemotherapy as first-line therapy in metastatic colorectal cancer: a randomized phase III study. J Clin Oncol 2008;26:2013–19. 6. Tebbutt NC, Wilson K, Gebski VJ, et al. Capecitabine, bevacizumab, and mitomycin in first-line treatment of metastatic colorectal cancer: results of the Australasian Gastrointestinal Trials Group Randomized Phase III MAX Study. J Clin Oncol 2010;28:3191–8. 7. Li A, Dubey S, Varney ML, et al. IL-8 directly enhanced endothelial cell survival, proliferation, and matrix metalloproteinases production and regulated angiogenesis. J Immunol 2003;170:3369–76. 8. Abdollahi A, Folkman J. Evading tumor evasion: current concepts and perspectives of antiangiogenic cancer therapy. Drug Resist Updat 2010;13:16–28. 9. Shojaei F, Ferrara N. Refractoriness to antivascular endothelial growth factor treatment: role of myeloid cells. Cancer Res 2008;68:5501–4. 10. Karapetis CS, Khambata-Ford S, Jonker DJ, et al. K-ras mutations and benefit from cetuximab in advanced colorectal cancer. N Engl J Med 2008; 359:1757–65. 11. Amado RG, Wolf M, Peeters M, et al. Wild-type KRAS is required for panitumumab efficacy in patients with metastatic colorectal cancer. J Clin Oncol 2008;26:1626–34.

12. Peeters M, Oliner KS, Parker A, et al. Massively parallel tumor multigene sequencing to evaluate response to panitumumab in a randomized phase III study of metastatic colorectal cancer. Clin Cancer Res 2013;19:1902–12. 13. Oliner KS, Douillard J-Y, Siena S, et al. Analysis of KRAS/NRAS and BRAF mutations in the phase III PRIME study of panitumumab (pmab) plus FOLFOX versus FOLFOX as first-line treatment (tx) for metastatic colorectal cancer (mCRC). J Clin Oncol 2013;31:abst 3511. 14. Ciardiello F, Troiani T, Bianco R, et al. Interaction between the epidermal growth factor receptor (EGFR) and the vascular endothelial growth factor (VEGF) pathways: a rational approach for multi-target anticancer therapy. Ann Oncol 2006; 17 (Suppl 7):vii109–vii114. 15. Price TJ, Hardingham JE, Lee CK, et al. Impact of KRAS and BRAF gene mutation status on outcomes from the phase III AGITG MAX trial of capecitabine alone or in combination with bevacizumab and mitomycin in advanced colorectal cancer. J Clin Oncol 2011;29:2675–82. 16. Price TJ, Hardingham JE, Lee CK, et al. Prognostic impact and the relevance of PTEN copy number alterations in patients with advanced colorectal cancer (CRC) receiving bevacizumab. Cancer Med 2013;2:277–85. 17. Shaw RJ, Cantley LC. Ras, PI(3)K and mTOR signalling controls tumour cell growth. Nature 2006; 441:424–30. 18. Pugh CW, Ratcliffe PJ. Regulation of angiogenesis by hypoxia: role of the HIF system. Nature Med 2003;9:677–84. 19. Semenza GL. Targeting HIF-1 for cancer therapy. Nat Rev Cancer 2003;3:721–32. 20. Mizukami Y, Kohgo Y, Chung DC. Hypoxia inducible factor-1 independent pathways in tumor angiogenesis. Clin Cancer Res 2007;13:5670–4. 21. Lee CH, Jeon YT, Kim SH, et al. NF-kappaB as a potential molecular target for cancer therapy. Biofactors 2007;29:19–35. 22. De Luca A, Gallo M, Aldinucci D, et al. The role of the EGFR ligand/receptor system in the secre-

23.

24.

25.

26.

27.

28.

29.

30.

31.

tion of angiogenic factors in mesenchymal stem cells. J Cell Physiol 2011;226:2131–8. Li A, Dubey S, Varney ML, et al. IL-8 directly enhanced endothelial cell survival, proliferation, and matrix metalloproteinases production and regulated angiogenesis. J Immunol 2003;170: 3369–76. Gyanchandani R, Sano D, Ortega Alves MV, et al. Interleukin-8 as a modulator of response to bevacizumab in preclinical models of head and neck squamous cell carcinoma. Oral Oncol 2013; 49:761–70. Addis MF, Tanca A, Pagnozzi D, et al. Generation of high-quality protein extracts from formalin-fixed, paraffin-embedded tissues. Proteomics 2009;9:3815–23. Jubb AM, Hurwitz HI, Bai W, et al. Impact of vascular endothelial growth factor-A expression, thrombospondin-2 expression, and microvessel density on the treatment effect of bevacizumab in metastatic colorectal cancer. J Clin Oncol 2006;24: 217–27. Hegde PS, Jubb AM, Chen D, et al. Predictive impact of circulating vascular endothelial growth factor in four phase III trials evaluating bevacizumab. Clin Cancer Res 2013;19:929–37. Verma A, DeGrado J, Hittelman AB, et al. Effect of mitomycin C on concentrations of vascular endothelial growth factor and its receptors in bladder cancer cells and in bladders of rats intravesically instilled with mitomycin C. BJU Int 2011;107:1154–61. Zhang HH, Qi F, Shi YR, et al. RNA interference-mediated vascular endothelial growth factor-C reduction suppresses malignant progression and enhances mitomycin C sensitivity of bladder cancer T24 cells. Cancer Biother Radiopharm 2012;27:291–8. Jayson GC, Hicklin DJ, Ellis LM. Antiangiogenic therapy—evolving view based on clinical trial results. Nat Rev Clin Oncol 2012;9:297–303. Van Cutsem E, de Haas S, Kang Y-K, et al. Bevacizumab in combination with chemotherapy as first-line therapy in advanced gastric cancer: a

C 2013 UICC Int. J. Cancer: 135, 731–741 (2014) V

741

Bruhn et al.

32.

33.

34.

35.

37.

38.

39.

40.

expression in colon carcinomas. BMC Cancer 2011;11:277. Martins SF, Garcia EA, Luz MAM, et al. Clinicopathological correlation and prognostic significance of VEGF-A, VEGF-C, VEGFR-2 and VEGFR-3 expression in colorectal cancer. Cancer Genomics Proteomics 2013;10:55–67. Weickhardt AJ, Williams D, Lee C, et al. Vascular endothelial growth factors (VEGF) and VEGF receptor expression as predictive biomarkers for benefit with bevacizumab in metastatic colorectal cancer (mCRC): analysis of the phase III MAX study. ASCO Annual Meeting Proceedings (PostMeeting Edition). J Clin Oncol 2011;29:3531. Barbera-Guillem E, Nyhus JK, Wolford CC, et al. Vascular endothelial growth factor secretion by tumor-infiltrating macrophages essentially supports tumor angiogenesis, and IgG Immune complexes potentiate the process. Cancer Res 2002;62: 7042–9. Ferrara N. Pathways mediating VEGFindependent tumor angiogenesis. Cytokine Growth Factor Rev 2010;21:21–6.

41. Kwon KA, Kim SH, Oh SY, et al. Clinical significance of preoperative serum vascular endothelial growth factor, interleukin-6, and C-reactive protein level in colorectal cancer. BMC Cancer 2010;10:203. 42. Jurgensmeier JM, Schmoll HJ, Robertson JD, et al. Prognostic and predictive value of VEGF, sVEGFR-2 and CEA in mCRC studies comparing cediranib, bevacizumab and chemotherapy. Br J Cancer 2013;108:1316–23. 43. Jain RK, Duda DG, Willett CG, et al. Biomarkers of response and resistance to antiangiogenic therapy. Nat Rev Clin Oncol 2009;6:327–38. 44. Maru D, Venook AP, Ellis LM. Predictive biomarkers for bevacizumab: are we there yet? Clin Cancer Res 2013;19:2824–7. 45. Llovet JM, Pe~ na CEA, Lathia CD, et al. Plasma biomarkers as predictors of outcome in patients with advanced hepatocellular carcinoma. Clin Cancer Res 2012;18:2290–300. 46. Zimmermann R, Ringwald J, Eckstein R. EDTA plasma is unsuitable for in vivo determinations of platelet-derived angiogenic cytokines. J Immunol Methods 2009;347:91–2.

Cancer Therapy

36.

biomarker evaluation from the AVAGAST randomized phase III trial. J Clin Oncol 2012;30: 2119–27. Zafirellis K, Agrogiannis G, Zachaki A, et al. Prognostic significance of VEGF expression evaluated by quantitative immunohistochemical analysis in colorectal cancer. J Surg Res 2008;147:99– 107. Bendardaf R, Buhmeda A, Hilska M, et al. VEGF-1 expression in colorectal cancer is associated with disease localization, stage, and longterm disease-specific survival. Anticancer Res 2008;28:3865–70. Zheng S, Han MY, Xiao ZX, et al. Clinical significance of vascular endothelial growth factor expression and neovascularization in colorectal carcinoma. World J Gastroenterol 2003;9:1227–30. Liang JF, Wang HK, Xiao H, et al. Relationship and prognostic significance of SPARC and VEGF protein expression in colon cancer. J Exp Clin Cancer Res 2010;29:71. Li Q, Wang D, Li J, et al. Clinicopathological and prognostic significance of HER-2/neu and VEGF

C 2013 UICC Int. J. Cancer: 135, 731–741 (2014) V