Cancer Immunology, Immunotherapy

, 58:1657

Immune evasion mechanisms in colorectal cancer liver metastasis patients vaccinated with TroVax (MVA-5T4)

Authors

    • CR UK Immunology Group, Paterson Institute for Cancer ResearchUniversity of Manchester, Christie Hospital NHS Trust
    • Department of Medical OncologyUniversity of Manchester, Christie Hospital NHS Trust
    • Cellular Immunology SectionLaboratory of Immunology, National Institutes of Health
  • Adam Dangoor
    • CR UK Immunology Group, Paterson Institute for Cancer ResearchUniversity of Manchester, Christie Hospital NHS Trust
    • Department of Medical OncologyUniversity of Manchester, Christie Hospital NHS Trust
  • Deborah J. Burt
    • CR UK Immunology Group, Paterson Institute for Cancer ResearchUniversity of Manchester, Christie Hospital NHS Trust
  • Thomas D. Southgate
    • CR UK Immunology Group, Paterson Institute for Cancer ResearchUniversity of Manchester, Christie Hospital NHS Trust
  • Sai Daayana
    • CR UK Immunology Group, Paterson Institute for Cancer ResearchUniversity of Manchester, Christie Hospital NHS Trust
  • Richard Harrop
    • Oxford BioMedica, Medawar Centre
  • Jan W. Drijfhout
    • Department of Immunohematology and Blood TransfusionLeiden University Medical Centre
  • David Sherlock
    • Department of SurgeryNorth Manchester Healthcare NHS Trust
  • Robert E. Hawkins
    • Department of Medical OncologyUniversity of Manchester, Christie Hospital NHS Trust
    • CR UK Immunology Group, Paterson Institute for Cancer ResearchUniversity of Manchester, Christie Hospital NHS Trust
Original Article

DOI: 10.1007/s00262-009-0674-y

Cite this article as:
Elkord, E., Dangoor, A., Burt, D.J. et al. Cancer Immunol Immunother (2009) 58: 1657. doi:10.1007/s00262-009-0674-y

Abstract

We have recently reported the results of a phase II trial in which two TroVax [modified vaccinia ankara (MVA) encoding the tumour antigen 5T4] vaccinations were given to patients both pre- and post-surgical resection of liver metastases secondary to colorectal cancer (CRC). 5T4-specific cellular responses were assessed at the entry and 2 weeks after each vaccination by proliferation of fresh lymphocytes and ELISA for antibody responses; 18 from the 19 CRC patients mounted a 5T4-specific cellular and/or humoral response. Here, we present a comparison of individual and between patient responses over the course of the treatments using cryopreserved peripheral blood mononuclear cells (PBMC) samples from the baseline until after the fourth vaccination at 14 weeks. Assays used were proliferation assay with 5T4-Fc fusion protein, overlapping 32mer 5T4 peptides, MVA-LacZ and MVA-5T4 infected autologous monocytes. Responses to 5T4 protein or one or more peptide pools were pre-existing in 12/20 patients and subsequently 10 and 12 patients showed boosted and/or de novo responses, respectively. Cumulatively, 13/20 patients showed proliferative responses by week 14. We also assessed the levels of systemic T regulatory cells, plasma cytokine levels, phenotype of tumour-infiltrating lymphocytes including T regulatory cells and tumour HLA class I loss of expression. More than half of the patients showed phenotypes consistent with relative immune suppression and/or escape highlighting the complexity of positive and negative factors challenging any simple correlation with clinical outcome.

Keywords

Colorecta cancer5T4TroVaxImmune evasionT regulatory cell

Introduction

Tumours employ a plethora of mechanisms limiting natural or vaccine induced anti-tumour immunity [1]. These include, but may not be limited to, down-regulation of MHC class I molecules, lack of expression of costimulatory molecules (CD80, CD86), secretion of immunosuppressive cytokines [vascular endothelial growth factor (VEGF), interleukin (IL)-10 and transforming growth factor (TGF)-β], lack of CD95 (Fas) expression, and expression of CD95L (FasL) [24]. A role of Treg cells in controlling tumour responses has been documented in animal models [5]. Consistent with these data are the observations that increased numbers of Treg in the peripheral blood and tumour microenvironment of cancer patients can correlate with poorer clinical outcome [1, 68]. Locally in the tumour, the balance of immune effector and regulatory cells appears to be important. For example, a high CD8+ T cell tumour-infiltration, a higher CD8+/CD4+ T cell ratio, and a higher CD8+/Treg ratio were associated with the absence of lymph node metastases in patients with large early stage cervical cancer [9]. CRC patients with high CD8+/CD4+ ratio of tumour-infiltrating lymphocytes showed a better clinical course with significantly higher 5-year survival [10].

The oncofoetal antigen 5T4 is expressed by many human carcinomas but demonstrates only limited expression in normal tissue [11]. For example, the majority of colorectal carcinomas are 5T4 positive [12] with those showing strong tumour-associated expression having poorer clinical outcome [13, 14]. Other studies have implicated over-expression of 5T4 molecules with decreased adhesion and increased motility [1517]. More recently, we have shown that 5T4 is an integral part of epithelial mesenchymal transition (EMT), a combination of cellular and molecular changes associated with the development of cancers and their metastasis [18, 19]. The frequent upregulation of 5T4 in a variety of tumour types is, thus, of likely functional significance to metastasis. A cancer vaccine targeting 5T4 has been developed consisting of a highly attenuated vaccinia virus, MVA, containing the human tumour-associated antigen 5T4 under control of a modified vaccinia virus promoter, mH5. Clinical trials of this vaccine in patients with advanced colorectal cancer have established safety and proof of concept by inducing anti-5T4 immune responses measured by serological and T cell assays [2023].

We have recently reported a TroVax trial delivering two vaccinations in patients prior to and following surgical treatment of their colorectal cancer liver metastases [24]. Overall, after four immunizations 95% of patients made a 5T4-specific response as measured by either cell mediated and/or serological assay against 5T4 extracellular region protein [24]. The validated primary proliferative assay used fresh PBMC against affinity purified (Myc/His tag) 5T4 protein lacking the transmembrane and cytoplasmic domains and secreted by CHO cells. In this study, we report additional evaluation by a proliferation assay using cryopreserved PBMC samples from baseline until after the fourth vaccination at 14 weeks. This secondary assay was designed to compare responses within patients at different time points within the same assay. It used a different source of 5T4 protein expressed as a fusion of 5T4 extracellular domain with human IgG1 Fc produced by COS cells [25], overlapping 32mer 5T4 peptides [26], as well as MVA-LacZ and MVA-5T4 infected autologous monocytes [27]. We have also investigated systemic levels of T regulatory cells (Treg) and plasma cytokine levels as well as the phenotype of tumour-infiltrating lymphocytes and alterations of MHC class I tumour expression, all of which might facilitate immune escape from 5T4 and other tumour specific immunity.

Patients and methods

Patients

Twenty patients were enroled in the trial. There were 19 CRC and one hepatocellular carcinoma. All the patients’ characteristics including age, sex, primary site of disease, 5T4 phenotype of liver metastasis, prior treatments and the total number of TroVax injections given have been previously described [24]. Vaccination and blood sampling time points are depicted in Fig. 1. Sixteen out of nineteen CRC patients received 4 vaccinations or more (except nos 4, 10 and 20); of these 16 patients 3 received one more vaccination at week 24 and 8 patients received 2 vaccinations at weeks 24 and 32. At the time of surgery, tumour biopsies of the liver metastases were taken and processed as previously described [24]. The 15 patents who received at least four vaccinations and potentially curative surgery were included in survival analyses [24].
https://static-content.springer.com/image/art%3A10.1007%2Fs00262-009-0674-y/MediaObjects/262_2009_674_Fig1_HTML.gif
Fig. 1

Trial schedule. Syringe indicates vaccination time point. Below the solid arrow time points at which blood samples were taken at weeks 0, 2, 4, 10, 14, 28 and 40 for monitoring of immune responses. CT scans were performed as indicated to monitor disease progression

Immunofluorescence

Immunofluorescence estimation of CD8+, CD4+ and CD4+FoxP3+ populations was performed on 7 μm cryostat sections of tumour biopsies or human tonsil as positive control. Frozen fixed tumour sample sections were mounted on APS coated slides and fixed in 100% ice-cold methanol. Following a serum block, immunohistochemical staining of FoxP3 and CD4 was carried out by incubation with mouse IgG1 monoclonal anti-FoxP3 primary antibody (eBioscience, San Diego, CA, USA) at dilution 1:50 and mouse IgG2a monoclonal anti-CD4 (Abcam, Cambridge, UK) at dilution 1:100 overnight and detected using goat antimouse IgG1 Alexa Fluor 488 and goat antimouse IgG2a Alexa Fluor 546 (Invitrogen, Paisley, UK). Rabbit anti-CD8 (AbCam; dilution, 1:100) was detected using Goat anti Rabbit-IgG Alexafluor 488. Cell nuclei were stained with DAPI. Sections were digitally photographed using a Coolsnap HQ camera mounted on a Zeiss 200M inverted microscope. Five respective fields of view were analysed and the TIL counting was performed using Image J image analysis.

Proliferation assays

The rationale for the 5T4 proliferative responses reported in this manuscript was as a secondary assay within the Cancer Research UK approved protocol. These assays were performed from individual time points from baseline to week 14 on the same day using previously cryopreserved PBMC. A total of 1.5 × 105 PBMC/well were cultured with eight pools of three 32mer peptides(PP1–PP8), overlapping by 15 aa and spanning the whole sequence of h5T4 protein, as previously described [26] (10 μg/ml of each peptide), 10 μg/ml h5T4-Fc protein, 10 μg/ml IgG1 protein (Sigma, as a control for the 5T4-Fc protein) and medium controls with or without any equivalent dilution of DMSO (VWR Leicestershire, UK), which was used to prepare peptide stocks.

PHA-M (Sigma-Aldrich, Dorset, UK) at 1, 2 and 5 μg/ml was used as a polyclonal stimulant. Each target was investigated using six replicates except for 5T4-Fc when three replicates were used. IgG1 was used as a control for 5T4-Fc. DMSO/medium was the control for the peptide pools. Plates were incubated for 5 days at 37°C before pulsing with 1 μci/mL of [3H]TdR for an additional 18 h then harvested using a Packard cell harvester and counted on a Packard Topcount (Perkin Elmer, Massachusetts, USA). The stimulation indices (SIs) were calculated by dividing the counts per minute (cpm) of the test wells by the cpm of control wells. The definitions of a response were as follows:
  • Pre-existing response: SI >2

  • Sustained response: SI >2 pre-vaccination and rising post vaccination

  • De novo response: SI <2 pre-vaccination and rising to >2 post vaccination and by twofold from baseline value

A proliferation assay based on MVA-LacZ or MVA-5T4 infected autologous monocytes was also used to investigate the patients’ responses to the vector and 5T4, as previously described [27].

T regulatory cells in peripheral blood

T regulatory cells were identified by expression of CD4, CD25 and FoxP3 transcription factor using the human regulatory T cell staining kit from eBioscience. Peripheral blood mononuclear cells were stained for extracellular CD4 and CD25 markers using CD4/25 cocktail (a cocktail of anti-human CD4-FITC and anti-human CD25-APC). Following fixation and permeabilization, the cells were washed and blocked for non-specific binding sites using normal rat serum. Anti-human FoxP3-PE or rat IgG2a-PE isotype negative control were then added for 30 min before washing twice and flow cytometric analysis using the FACSCalibur flow cytometer (Becton Dickinson, Oxford, UK), as previously described [7].

Cytokine assays

Total human TGF-β1 was measured in plasma using an ELISA kit (R&D Systems, Minneapolis, MN USA) according to the manufacturer’s instructions. Biologically latent TGF-β1 was activated by acid treatment prior to measurement. Seven human cytokines including IL-2, IL-10, IL-12p70, IFN-γ and TNF-α were measured in 25 μl of plasma using the Th1/Th2 10plex Kit II (Bender MedSystems GmbH, Vienna, Austria) according to the manufacturer’s instructions. Samples were acquired using the FACScan (Becton Dickinson,) and analysed using a FlowCytomixPro 1.0 data analysis software.

HLA class I expression

HLA locus specific and allele specific tumour expression analysis was performed as previously described [28]. The following monoclonal antibodies (mAbs) were used: W6/32 (anti-HLA class I/β2m complex), ATCC; A131 (anti-HLA-A antigens), J. Kornbluth, University of Arkansas, Fayetteville, AK, USA; HC10 (HLA-B/C heavy chain), H. Ploegh; Massachussetts Institute of Technology, Cambridge, MA, USA; BB7.1 (anit-HLA-B7) and L368 (anti-HLA-B2M), J. Bodmer, J Radcliffe Hospital, Oxford, UK; CR11-351 (anti-HLA-A2, 28), H2-89-1 (anti-HLA-B), S. Ferrone, New York Medical College, Valhalla, NY, USA; LT129.11 (anti-HLA-A31), AG Palma-Carlos, Lisbon, Portugal; 375-1 (anti-HLA-B13), 160-30 (anti-HLA-A3), 116.5.28 (anti-HLA-Bw4) and 126.39 (anti-HLA-Bw6), K. Gelsthorpe, Sheffield BTS, UK; HA41 (anti-HLA-A24), One Lambda, Canoga Park, CA, USA. Frozen sections were labelled using a standard horseradish peroxidase labelled avidin–biotin technique following fixation in acetone for 2 min. Finally, the chromogen diaminobenzidine tetrahydrochloride with 0.03% hydrogen peroxide was applied and the colour allowed to develop before counterstaining with Mayer’s haematoxylin, followed by dehydration through graded alcohols, fixation in xylene, and mounting. All incubations were performed at room temperature in a humid chamber. The normal HLA expression of stromal tissue and lymphocytes served as a positive control within each section, and serial sections of each specimen incubated without the primary antibody were included as negative controls.

Results

5T4-proliferative responses

PHA was used to analyse the quality of PBMC from the different time points for each patient and the cryopreserved cells were able to proliferate to this polyclonal stimulus (99% of all assays) and in a dose dependent manner (91% of all assays). The variation in magnitude of SI between individual time point samples can never account for the observed vaccine related responses to 5T4 antigens. Three examples of the proliferative responses to 5T4-Fc, 5T4 peptide pools (PP) and controls between weeks 0–14 are shown in Fig. 2a, b and c. Patient 6 had pre-existing responses to PP2 and PP4 which were sustained at weeks 4, 10 and 14. There was a de novo response to PP7 at week 4 which was sustained at week 10 and 14. At week 10, de novo responses to PP1, PP3, PP5, PP6 and PP8 were seen with responses to PP1 and PP3 sustained at 14 weeks. Patient 7 showed pre-existing responses to PP2, PP5 and PP6 sustained at 4, 10 and 14 weeks, PP4 response was seen at weeks 0, 4 and 14 but not week 10; de novo responses to PP1 and PP7 were only transiently detected at week 4. Patient 11 showed a pre-existing response to 5T4-Fc sustained at week 4 only and de novo responses to PP1, PP3, PP4, PP5 and PP6 at week 14.
https://static-content.springer.com/image/art%3A10.1007%2Fs00262-009-0674-y/MediaObjects/262_2009_674_Fig2_HTML.gif
Fig. 2

Proliferation assays in response to 5T4-Fc fusion protein and 5T4-derived peptide pools. The figures show stimulation indices of proliferation in response to DMSO/media control, IgG1 control, 5T4 protein and 5T4-derived peptide pools in PBMC isolated from patient 6 (a), patient 7 (b) and patient 11 (c). white box week 0, gray box week 4, dotted box week 10, black box week 14. Overall proliferative response of all patients to all 5T4-derived peptide pools is shown in (d). Error bars show the standard error of the mean (SEM)

Overall, there was a relatively low rate of response seen against 5T4-Fc with 5 patients (nos. 2, 7, 11, 16, 18) showing pre-existing responses but only 2 patients at week 4 (no. 8 de novo; no. 11 sustained), one patient at week 10 (no. 8 sustained) and 3 patients at week 14 (nos. 5 and 19 de novo; no. 8 sustained). More common were responses to 32-mer peptide pools; 10/20 patients showed pre-existing responses to one or more peptide pools, with 8, 10 and 10 patient responses seen at weeks 4, 10 and 14, respectively. Cumulatively 13/20 patients showed proliferative responses by week 14.

A summary of the patient responses is given in Table 1. Twelve out of twenty patients showed pre-existing responses to 5T4-Fc or one or more peptide pools; the most frequent responses were to PP2 (8), PP4 (6), and 5T4-Fc (5). Ten out of twenty patients showed sustained responses to one or more antigens at one or more time points (5T4-Fc, PP2, PP4–PP7); the most frequent responses were to PP2 (7 patients at 15 time points) and PP4 (3 patients at 6 time points). Twelve out of twenty patients showed de novo responses to one or more antigens at one or more time points (5T4-Fc, PP1–PP8); the most frequent responses were made to PP2 (5 patients at 5 time points), PP3 (5 patients at 6 time points), PP4 (4 patients at 5 time points) and PP7 (5 patients at 7 time points). No responses were detected in patients 4, 12, 14 and 20. Overall the order of the prevalence of response to any PP or 5T4-Fc (number of patients) is PP2 (12), PP4 (10), 5T4-Fc (8), PP6 (7), PP3 (7), PP5 (6), PP7 (6), PP1 (5) and PP8 (3). The widest repertoire and kinetics of response was found in patients 6, 7, 11 and 18.
Table 1

Pre-existing, boosted, and de novo T cell proliferative responses

Week

Pre-existing

Boosted responses, de novo responses (bold, italics)

0

4

10

14

Patient

 1

D (week 8)

 2

5T4, A, D, G J

D

D

 3

D, J

D, S

 4

NT

 5

D, G, J

5T4

 6

D, J

D, J, S

D, J, A, G, M, P, S, V

D, J, A, G, S

 7

5T4, D, J, M, P, V

D, J, M, P, A, S

D, M, P

D, J, M, P

 8

P

5T4, D, S, V

5T4, D

P

 9

G

 10

M

 11

5T4

5T4

A, G, J, M, P

 12

 13

D

D

 14

 15

D, J

D

D, P

D

 16

5T4

D, G, J, P

J, P

 17

D

D, M

D, J

 18

5T4, A, D, G, J, M, P, S

S

J, P, S

 19

S

5T4

 20

The table summarises 5T4-specific proliferative responses of frozen PBMC isolated from CRC liver metastases patients prior to TroVax vaccination and post vaccination (weeks 4, 10 and 14)

Letter name of peptide pool, – no response detected, NT not tested

An analysis of variance (ANOVA) followed by Tukey multiple range test was used to analyse the proliferative response against 5T4 peptide pools. This suggested that the stimulation indices were significantly higher at week 10 (after 3 vaccinations and surgery) than those for weeks 0 and 4 (2 vaccinations, pre-surgery); the SIs at week 14 were higher than at week 0. SIs for PP2 were higher than those for all other peptide pools (Fig. 2d). Patient 6 demonstrated the highest level of proliferative response against 5T4-derived peptide pools. To test whether results for this patient had an undue effect on the analysis, the ANOVA was also performed excluding patient 6. This led to the week 10 response falling back so that it was no longer significantly higher than weeks 4 and 14. However, the post vaccine responses at weeks 4, 10, and 14 remained higher than the pre-vaccine response at week 0.

Before vaccination, the median proliferative response to MVA-LacZ and MVA-5T4 was 2.35 [Inter-quartile range (IQR) = 0.65–6.55] and 1.92 (IQR = 1.08–3.66), respectively. All patients excluding patients 4 and 7 developed strong proliferative responses to MVA after treatment. Following 2 vaccinations (week 4, before surgery) or 3 vaccinations (week 10, after surgery), or 4 vaccinations (week 14), compared to pre-existing levels, lymphoproliferative responses to MVA-LacZ and MVA-5T4 increased 20 and 24, or 12.7 and 19.9, or 10.7 and 12.2-fold, respectively. The pre-existing responses defined by SI >2 with both MVA-LacZ and MVA-5T4 were concordant in 17/20 cases (12 positive) and in all those showing SI >10 (nos. 11, 14, 15 and 18). It is clear that boosted or de novo responses to 5T4 protein and/or peptides after 2, 3 or 4 vaccinations are not limited by pre-existing responses to MVA of SI >2 or even greater than 10. Thus, the ability to induce a 5T4 proliferative response is apparently not influenced by previous exposure to the vector or the subsequent response to MVA.

Circulating Treg levels before and after treatment

The median frequency of CD4+CD25high Treg cells as a proportion of lymphocytes in healthy donors (n = 23) was 1.89% (IQR = 1.33–2.45). In a set of CRC liver metastases patients (n = 28), recruited by the same criteria as for the trial protocol including receiving surgery but unvaccinated, the median frequency prior to surgery was 2.45% (IQR = 1.63–4.09) which is significantly higher (Mann–Whitney U test; P = 0.026). The Treg levels in the trial patients were determined using a three-colour flow cytometric analysis of cryopreserved PBMC to determine the relative proportions of CD25highFoxP3+ cells within the CD4+ T cell population from pre-vaccination, week 10 (except no 13 at week 14) and week 26 (except no 1, 8, 17 18 at weeks 22, 18, 33, 16, respectively). This population represents more than 70% of the CD4+CD25high cells [7]. The median frequency of CD4+CD25+FoxP3+ cells prior to vaccination (16 patients) or after 3 vaccinations (15 patients excluding no 16) or 5 vaccinations (16 patients) was 3.34 (IQR = 2.54–4.48) or 3.06 (IQR = 2.36–4.20) or 3.12 (IQR = 2.42–4.34), respectively. Individual patients show fairly consistent levels across the time points. Overall, there was no significant difference in Treg frequency at baseline following either 3 or 5 vaccinations (Wilcoxon Signed Ranks; P = 0.599 or 0.298, respectively). Systemic Treg levels were determined at times defined by the trial protocol and it is possible that other significant albeit transient changes were missed.

Tumour-infiltrating lymphocytes

In our previous study, the magnitude of peritumoral CD3+ infiltration determined by immunohistochemistry (IHC) was significantly associated with prolonged survival but neither CD4+, CD8+ T cell infiltration alone nor the CD4+:CD8+ ratio showed a significant correlation [24]. In this study using immunofluorescence (n = 18 CRC), the mean ± SD of CD4+ and CD8+ T cell infiltrations were 121 ± 57 (range = 19–238) and 58 ± 32 cells/mm2 (range = 19–238), respectively. The number of CD4+ or CD8+ TIL determined by immunofluorescence were not significantly different from the results obtained by IHC (Wilcoxon Signed Rank 2 sided tests; P = 0.1202 and 0.7819, respectively). We also determined infiltration of CD4+FoxP3+ T regulatory which showed a mean density of 18 ± 4/mm2 but was very variable between patients ranging from 0 to 66 CD4+FoxP3+ cells/mm2 (median = 10.6; IQR = 5.8–29.1). When the 15 patients who received at least four vaccinations and had potentially curative surgery were stratified into those > or < median of Treg infiltration, there was no significant association with survival (Fig. 3a; log-rank, P = 0.4640). Interestingly, when FoxP3+ cells were calculated as percentage of CD4+ cells (median 14, IQR 6.8–31.9), and stratified into > or < median, there was a significant association of low Treg infiltration with increased survival (Fig. 3b; log-rank, P = 0.0234). The median of CD8+:Treg ratio for this group of patients was 3.10 (IQR 1.5–7.6, range 1.1–307), but stratification of patients above or below median did not show any correlation with survival (log-rank, P = 0.270; data not shown). Using a simple linear regression analysis, no significant inverse correlation between Treg infiltration and CD8+ infiltration (P = 0.259, r = 0.281) or CD4+ infiltration (P = 0.288, r = 0.265) was apparent.
https://static-content.springer.com/image/art%3A10.1007%2Fs00262-009-0674-y/MediaObjects/262_2009_674_Fig3_HTML.gif
Fig. 3

Kaplan–Meier survival plots. Evaluable CRC patients were stratified below (n = 9) or above (n = 6) median FoxP3+ cell infiltration into the tumour (a), and below (n = 7) or above (n = 8) median FoxP3+ cell infiltration as a proportion of CD4+ cells into the tumour (b)

Plasma concentration of cytokines before and after treatment

Prior to vaccination the mean (±SD) plasma concentration of TGF-β1 (16 patients) was 2,061 ± 1,227 pg/ml (range = 631–5,164, 95% CI = 1,407–2,715), which is about double plasma levels determined by this methodology in normal donors (1,165 ± 214 pg/ml; range, 903–1,654; R&D data sheet). After three vaccinations and surgery, TGF-β1 level was reduced in the plasma of only two patients (nos. 7 and 9; reduced to normal level in no. 9), whilst all other patients (except no 11; no change) showed higher TGF-β1 concentration (mean ± SD = 2,289 ± 1,249). At 26 weeks, the mean (±SD) of TGF-β1 concentration was 2,010 ± 768 pg/ml (range = 968–3,417, 95% CI = 1,602–2,420); with 6 patients (nos. 1, 5, 6, 7, 16, 19) showing a reduction (normal level in nos 5 and 19) in TGF-β1 levels. Overall, the patients showed relatively elevated levels of TGF-β1 at recruitment which did not show any significant reduction following three vaccinations and surgery or later at week 26.

Detectable levels of IL-10, IFN-γ and IL-2 were not found in any patient before or after vaccination with the exception of patient 5 where at week 10, IFN-γ = 380.2 pg/ml, IL-2 = 39.9, IL-10 = 69.4 pg/ml were detected. TNF-α was detected in 9, 4 and 6 of 15 patients at 0, 10 and 26 weeks, respectively, and there was no significant change in levels compared to pre-treatment (10 and 26 weeks; P = 0.598 and 0.183, respectively).

HLA class I expression

All tumour and normal liver tissue were positive for HLA class I using the W6/32 antibody recognizing a monomorphic determinant of HLA class I molecules. The IHC analyses using a combination of HLA class I locus and allele specific monoclonal antibodies and the tissue type of the individual patients allowed the detection of changes in HLA expression in the tumour compared to the associated normal tissue. This approach cannot exclude HLA downregulation since the reagents available are not comprehensive but can provide clear evidence of HLA loss in many cases. Thus, 17 CRCs were investigated and no loss of HLA class I expression could be detected in the tumours of patients nos. 1, 3, 5, 6, 12, 14 and 18 (Table 2). Locus specific antibody analyses documented loss of expression at HLA-A (nos. 7, 8, 11, 16, 17 and 20), HLA-B (nos. 11, 16, 17 and 20) and specific alleles (nos. 8, HLA-B7; 9, HLA-B8; 13, HLA-A3,-B7; 15, HLA-B7 and 20, HLA-B7), as shown in Table 2.
Table 2

HLA class I expression in CRC liver metastases

Patient

HLA class I tissue type

Heavy chain (W6/32)

Locus B + C (HC-10)

Locus A (A131)

Locus B (H289.1)

Bw4/Bw6

Alleles

Result

PT001

NT

++

++

NT

++

NT

NT

No loss detected

PT003

A2, A29; B35 (Bw6), B44 (Bw4); Cw4, Cw16

++

++

++

++

Bw4 ++

Bw6 ++

A2 ++

No loss detected

PT005

A29; B44 (Bw4), B57 (Bw4); Cw6, Cw16

++

++

++

++

Bw4 ++

NT

No loss detected

PT006

A2, A23; B40 (Bw6), B49 (Bw4); Cw3, Cw7

++

+

+

+

Bw4 +

Bw6 +

A2 ++

A23 ++

No loss detected

PT007

A1; B8 (Bw6); Cw7

++

++

±

++

Bw6 ++

NT

Locus A heterogeneous

PT 008

A1, A3; B7 (Bw6), B8 (Bw6); Cw7

++

++

±

Bw6 ±

A3 (−)

B7 (−)

Locus A negative and B7 allele negative

PT009

A1; B8 (Bw6), B57 (Bw4); Cw6; Cw7

++

++

++

++

Bw4 ++

Bw6 (−)

NT

Allele B8 negative

PT 011

A1, A66; B7 (Bw6), B8 (Bw6); Cw7

++

++

±

±

Bw6 ±

B7 ±

Loci A and B heterogeneous

PT012

NT

++

++

++

+

NT

NT

No loss detected

PT013

A2, A3; B7 (Bw6), B44 (Bw4); Cw5, Cw7

++

±

++

±

Bw4 ++

Bw6 ±

B7 (−)

A2 ++

A3 ±

Alleles B7 and A3 negative

PT014

A2, A24; B27 (Bw4), B35 (Bw6); Cw1, Cw4

++

++

++

++

Bw4 ++

Bw6 ++

A2 ++

A24 NT

No loss detected

PT 015

A2, A3; B7 (Bw6), B35 (Bw6); Cw4, Cw7

++

±

++

±

Bw6 ++

A2 ++

A3 NT

B7 (−)

Allele B7 negative

PT016

A11, A24; B7 (Bw6), B39 (Bw6); Cw7

++

±

Bw6 (−)

A24 ±

B7 (−)

Loci A and B negative

PT017

A11, A68, B44 (Bw4), B51 (Bw4); Cw7, Cw15

++

++

Bw4 +

NT

Loci A and B negative

PT 018

A2, A26; B18 (Bw6), B38 (Bw4); Cw7, Cw12

++

++

++

++

Bw4 ++

Bw6 ++

A2 ++

No loss detected

PT019

A2, A30; B7 (Bw6), B13 (Bw4); Cw6, Cw7

++

±

NT

NT

Bw4 NT

Bw6 ±

B7 (±)

A2 (++)

A30 (++)

Allele B7 negative

PT020

NT

++

NT

NT

Loci A and B negative

++ Positive, + weak positive, ± heterogeneous, − negative

NT not tested

Immune evasion and escape phenotypes

Table 3 summarises HLA tumour loss, systemic and local Treg levels in the patients. Overall there were 10/17 CRC tumours with evidence of locus and/or allele loss of HLA class I expression; 8/18 CRC tumours showed >10.6% (median) frequency of FoxP3+ cells within the CD4+ T cell population; only 4 patients showed >10 CD8+/FoxP3+ ratios, whilst 13/16 patients showed elevated (arbitrarily >3.0%) Treg level in the peripheral blood at weeks 0 and/or 10 and/or 26 weeks. Plasma TGF-β1 levels were relatively high (>1,650 pg/ml) in 9/16 patients but this did not correlate with either peripheral (P = 0.407, r = −0.223) or local Treg levels (P = 0.107, r = 0.418). Figure 4 illustrates patient 14 with very high systemic and tumour Treg levels but no evidence of HLA loss by the tumour cells whereas patient 17 had reducing systemic and lower levels of tumour-associated Treg but complete absence of expression by the tumour of HLA-A and B locus molecules.
Table 3

Summary of HLA tumour loss, systemic and local Treg levels, and CD8:FoxP3 ratio of all patients

Patient

HLA-I loss

FoxP3/CD4 (%)

CD8:FoxP3

W0

Treg

W10

W26

1

N

0

>15.3

2.5

4

1.21

3

N

14.0

13.2

1.9

1.2

1.5

5

N

6.8

5.1

2.5

3.4

2.8

6

N

39.7

1.8

2.4

3.8

4.2

7

Y

31.9

1.3

1.9

1.9

1.2

8

Y

18.8

3.1

3.4

4.7

3.5

9

Y

9.3

3.8

3.8

3.7

3.4

10

NT

0

>76.7

NT

NT

NT

11

Y

9.7

7.6

2.6

2.3

2.8

12

N

4.8

9.8

3.3

1

2

13

Y

7.7

5.9

4.5

4.2

3.7

14

N

76

1.7

10.2

10.2

10.4

15

Y

19.7

1.1

5.6

4.3

5

16

Y

0.7

122.0

3.7

NT

4.8

17

Y

3.5

1.5

4.5

2.5

2.4

18

N

34.1

1.2

3.3

3

2.5

19

Y

16.7

3.1

3.1

3.1

3.1

20

Y

5.0

1.8

NT

NT

NT

Total

Loss (10/17)

>10.6% (8/18)

<10 (14/18)

>3 (10/16)

>3 (10/15)

>3 (8/16)

Cut off level for immune evasion mechanisms: any loss HLA; FoxP3/CD4 (%) in tumour > median (10.6); CD8:FoxP3 defined as <10 (arbitrary); systemic Treg level defined as >3 (arbitrary)

N no loss detected, Y loss detected; NT not tested

https://static-content.springer.com/image/art%3A10.1007%2Fs00262-009-0674-y/MediaObjects/262_2009_674_Fig4_HTML.gif
Fig. 4

Immune modulation and escape. The figures show Treg frequency in peripheral blood (a) and tumour (b), and HLA class I expression in the tumour (c) of two patients. Patient 14 showed high systemic and tumour-infiltrating Treg levels with no evidence of HLA loss by the tumour cells. Patient 17 had low systemic and tumour-infiltrating Treg levels but complete absence of expression by the tumour of HLA-A and B locus molecules

Discussion

In this study, we performed additional evaluation of cellular immune responses to 5T4 in cryopreserved PBMC from CRC liver metastasis patients vaccinated with TroVax. Responses to 5T4-Fc or one or more peptide pools were pre-existing in 12/20 patients and subsequently 10 and 12 patients showed boosted and/or de novo responses, respectively. Cumulatively 13/20 patients showed proliferative responses by week 14. Comparing these 5T4-Fc and 5T4 peptide pool proliferative responses with those of the primary assay previously reported [24], there was a concordance of 65, 53, 59 and 57% for any 5T4 specific response at baseline, 4, 10 and 14 weeks, respectively; there was no bias for positive or negative responses. The two independent measures of proliferation do not show a very high concordance. This may reflect the limitations of arbitrary cut off levels of negativity or positivity although it is obviously necessary to start with some preset values when defining potential trial outcomes. It seems likely that the difference in results obtained in the two assays reflects a combination of differential sensitivity and specificity. Thus, in the primary assay the use of fresh lymphocytes probably increases the sensitivity but the 5T4 protein is heavily glycosylated and may not be optimally processed in vitro. By contrast using cryopreserved lymphocytes reduces sensitivity but overlapping pools of peptides increases the range of potential 5T4 specific responses that could be measured with a lack of optimal antigen processing. The latter may of course include responses generated to 5T4 epitopes by vaccination but not necessarily by natural tumour-associated expression. In support of these speculations, 5T4 protein responses were detected on 35 occasions using fresh PBMC but only 2 and 8 times, respectively for cryopreserved PMBC or both assays. On 11 occasions, responses were only detected with overlapping peptides with cryopreserved PBMC.

Not surprisingly given the size of protein and the diversity in tissue type of the patients it is difficult to match individual peptide responses with potential HLA class II restricting element. However, of the five patients showing a response to PP5 (M, N, O peptides), three had a HLA-DR4 allele which has been shown previously to restrict response to a 5T4 epitope within the N and O peptides [26]. The 10mer ELISPOT responses reported previously [24] were found in eight patients PBMC at various points post vaccination and 6/7 that were tested showed tumour HLA class I loss by immunohistochemistry. This is consistent with these effectors being expanded by the vaccination having previously driven evolution of the tumour phenotype. They could have been maintained at low levels by cross priming to 5T4 glycoprotein through local antigen presenting cells. One tumour biopsy also had detectable 5T4 specific T cells in the TlLs isolated after two vaccinations and this patient’s tumour did show downregulation of HLA class I expression (patient 19).

We have previously reported an association of prolonged survival with fresh PBMC 5T4 specific proliferation and the density of CD3+ TIL at surgery; indeed 7/8 patients who had pre-existing fresh PBMC proliferative response to 5T4 were long term survivors [24]. Where pre-existing 5T4 responses were detected in the cryopreserved PMBC (n = 5), four patients were represented in the evaluable survival group and three were longer term survivors. The interpretation of the proliferative peptide responses is likely to be complex. It clearly reflects the immunogenicity of the vaccine but how the individual peptide specificity correlates with outcome is challenging in a study of this size.

It is possible that the mechanisms underlying the prolonged survival may include both serological and cellular immunity manifest through either direct or indirect mechanisms. Typically, cancer vaccines are aimed at generating CD8+ T effectors likely to be able to directly kill as well as produce anti-tumour cytokines such as IFN-γ or TNF. However, the actions of the latter may be confounded by immune escape mechanisms dependent on HLA loss in the tumour targets or the influence of T regulatory cells locally and/or systematically or by the presence of immunosuppressive cytokines. The complexity of factors likely to influence vaccine or natural tumour specific cell mediated immunity in the patients in this trial is evidenced by considering the proportion of patients with immune escape phenotypes. Thus, 59% demonstrated HLA class I tumour loss, 44% elevated FoxP3+ Treg as percentage of CD4+ T cells in situ, 78% showed <10 CD8+:FoxP3+ ratios whilst 81% had >3% Treg level in the peripheral blood. Overall, these observations reflect the complexity of interactions likely to influence the balance of immune factors contributing to clinical outcome. However, in terms of prolonged survival only elevated levels of Treg as a proportion of CD4+ infiltrate, but not as an absolute number, were significantly associated with poorer survival.

There are some published data on TIL influence in CRC. One study stratified primary CRC by high and low Treg infiltration but did not find any significant difference in survival between patients [29]. Another documented a reduction in tumour-infiltrating T cells in CRC liver metastases compared to normal liver tissue and this was speculated to be as a result of increased tumour-infiltration by Treg [30] but our work does not support this conclusion. In another report, CD8+ T cell infiltration was shown as a positive prognostic factor in CRC [31] but this was not apparent in this study where most of the tumours (14/18) showed <10 CD8+/Treg ratios. Our previous IHC work showed that CD3+ levels of infiltration but not CD4+ or CD8+ were associated with better survival. Taken together and with the rider that this is a small study, these data suggest the possibility that manipulation of local CD4+ infiltration could influence immune control of the cancer.

Acknowledgments

This trial was sponsored by Cancer Research UK and monitored by the Cancer Research UK Drug Development Office. EE, AD, DB, TS and PLS were supported by CR UK; SD by Wigan Cancer Research Fund as a Joseph Starkey Fellow. We thank Ester Martin for her contribution to the HLA expression studies. We are very grateful for all patients who participated in this trial.

Copyright information

© Springer-Verlag 2009