Rates of MAGE-A3 and PRAME expressing tumors in FFPE tissue specimens from bladder cancer patients: potential targets for antigen-specific cancer immunotherapeutics

Abstract

Introduction: Antigen-specific active immunotherapy is an investigational therapeutic approach of potential interest for bladder cancer regardless of disease stage. Clinical development of antigen-specific immunotherapeutics against bladder cancer must be preceded by assessment of the expression of relevant genes in bladder tumors. The objectives of this study (NCT01706185) were to assess the rate of expression of the MAGE-A3 and PRAME genes in bladder tumors and to investigate the feasibility of using formalin-fixed paraffin-embedded (FFPE) tumor tissues for testing. Materials and methods: Archived FFPE bladder tumor specimens (any stage) were tested for mRNA expression of MAGE-A3 and PRAME using antigen-specific quantitative reverse transcription polymerase chain reaction assays. Data on patients and tumor characteristics were obtained from hospital records to investigate these characteristics’ possible association with the antigen expression. Results: Over 92% of the 156 tumors examined gave valid antigen test results. Of the tumors with a valid test, 46.5% were MAGE-A3-positive, 32.2% were PRAME-positive and 59.7% positive for at least one of them. Exploratory analyses of possible associations between antigen expression and patient or tumor characteristics did not identify clear associations between antigen expression and any of the variables investigated. Conclusions: Assessment of tumor antigen mRNA expression by using FFPE bladder tissues was feasible. The rates of MAGE-A3-positive and PRAME-positive tumors indicate that both antigens may be interesting targets for immunotherapeutics against bladder cancer.

Introduction

Bladder cancer is among the most common malignancies, particularly in men for whom it is the fourth most common cancer type [1]. The incidence continues to increase with estimated 386,000 new cases and 150,000 deaths worldwide in 2008 [2].

Important unmet clinical needs remain. For non-muscle invasive (< T2) disease (~75% of new cases), the standard of care is tumor resection followed by intravesical instillation of Mycobacterium bovis bacillus Calmette-Guérin (BCG) [3]. However, BCG can lead to frequent and sometimes serious side effects such as cystitis, irritative voiding symptoms and, rarely, life-threatening BCG sepsis. Moreover, 30-45% of the patients do not respond to BCG and in up to half of the responding patients it does not prevent disease relapse [3,4]. For patients with muscle-invasive (≥ T2) disease (~25% at diagnosis), survival outcomes after radical cystectomy are poor, especially for patients with non-organ confined disease with a 10-year survival rate of 23% [5].

Bladder tumors are recognized as immunogenic and the presence of tumor infiltrating lymphocytes is associated with a favorable prognosis [6,7]. The incapacity of the host immune system to eradicate the tumor despite initiating an immune response against it may be explained by several immunosuppressive mechanisms whose relative importance has not yet been established [6,8].

The immunogenicity of bladder cancer suggests that active antigen-specific immunotherapy may be a promising therapeutic approach. Tumors express antigens with varying degrees of tumor specificity and the theoretical basis for active antigen-specific immunotherapy is that tumor-associated antigens may be exploited to mobilize the patient’s own immune system to recognize and attack tumor cells. Hence, only patients whose tumor expresses the targeted antigen may benefit from an antigen-specific immunotherapy [9] so it is important to determine the rate of antigen expression in this patient population.

MAGE-A3 (Melanoma Antigen) and PRAME (PReferentially expressed Antigen of MElanoma) are two tumor-associated antigens which have been in focus of research to develop antigen-specific cancer immunotherapy. Currently they are both, in the form of a recombinant protein combined with an immunostimulant, under clinical development as immunotherapeutics against non-small cell lung cancer (NSCLC) and melanoma.

MAGE-A3 belongs to the cancer testis (CT) gene family. CT genes are abundantly expressed in a wide range of different cancers but are generally not expressed in normal adult cells except in germ cells of the immune privileged testis [10]. The lack of human leukocyte antigen class I cells on the surface of germ cells implies that MAGE-A3 derived epitopes cannot be presented to and recognized by T-cells of the immune system. In consequence, immunization against MAGE-A3 is not expected to lead to immune-related toxicities while inducing an immune response against tumor cells expressing the antigen [11]. The biological functions of MAGE-A3 are largely unknown but its presence on tumor cells has been associated with worse prognosis [12-14] suggesting a pathogenic role in the progression of disease. The MAGE-A3 gene has been found to be expressed in substantial fractions of human tumors in a wide range of different cancers.

PRAME may be characterized as an atypical CT gene [15], because it, unlike the other CT genes, is also expressed, albeit at low or minimal levels, in certain types of cells in normal adult tissues other than the testis, including endometrium, ovary, kidney and adrenal gland [16]. High levels of PRAME expression have been found in many types of cancer, including melanoma, NSCLC, breast cancer, leukemia and other hematological cancers [17]. PRAME has a well-defined immunogenicity, as it was originally identified as the protein that triggered an endogenous anti-tumor immune response in a melanoma patient [18] and has been found to induce specific killing of PRAME expressing leukemia cells by PRAME specific cytotoxic T lymphocytes [19]. Its cellular functions are not yet fully characterized [15,20] but it has a role in tumor progression by inhibiting cellular differentiation through blockade of retinoic acid receptor signaling in tumor cells. High levels of PRAME expression have been found to correlate with shorter overall survival in neuroblastoma [21] and serous ovarian adenocarcinomas [22] and with poor clinical outcomes in breast cancer [23]. However, a recent retrospective study of 1260 samples from NSCLC patients found that PRAME expression, as well as MAGE-A3 expression, had no prognostic value for overall survival [24].

The first objective of this study was to determine the rate of expression of MAGE-A3 and PRAME in tumors from bladder cancer patients in order to assess the proportion of bladder cancer patients who would potentially be eligible for inclusion in clinical studies of immunotherapeutics targeting these antigens. A second objective was to evaluate the feasibility of using archived formalin-fixed paraffin-embedded (FFPE) tumor tissues for these antigen expression tests.

Materials and methods

Ethics statement

All data and samples were taken from those already stored at the investigation site. Many patients may no longer be alive, or no longer be in contact with the investigation site. Patients were not subjected to any treatments or other procedures and no questionnaires were addressed directly to them. For these reasons, patients were not required to give their informed consent before inclusion in the study. Therefore, strict anonymity of all study data were established and maintained. The study (ClinicalTrials.gov: NCT01706185) was conducted according to Good Clinical Practice, the Declaration of Helsinki, and local rules and regulations of the country. Prior to initiation, the study and its protocol were accepted by the local ethics committee, the “Commissie Medische ethiek van de Universitaire Ziekenhuis KU Leuven”, that did not raise any concern related to the absence of informed consent in view of the appropriate protocol rules applied to maintain confidentiality.

The gene expression testing for this study was performed in 2008.

Study cohort

FFPE tissue specimens (N = 156) from bladder cancer patients were selected from the pathology archive of a single center (University Hospitals Leuven, Belgium) and tested by means of quantitative reverse transcription polymerase chain reaction (qRT-PCR) at a central laboratory. Standard non-buffered 6% formalin was used as fixative but the duration of the fixation for individual FFPE specimens was not recorded. The inclusion criteria were: 1) the presence of biopsy-proven bladder cancer (all tumor slides were reviewed by an experienced pathologist (EL) to confirm the diagnosis and assess the quality of the tissue) so patients with different disease stages were eligible, and 2) availability of sufficient tumor tissue. Patient and tumor characteristics (gender, age, resection year (1991 to 2007), T stage, tumor histology and size) were obtained from the patients’ hospital records.

Manual dissection

Manual dissection of the tumor specimens was performed by Response Genetics Inc. (RGI, Los Angeles, USA). Ten micron thick sections were mounted on plain glass slides and stained with nuclear fast red (NFR). Using a hematoxylin and eosin stained 5 µm thick section as a guide, a minimum of 50 mm² of tumor tissue with 50-80% of neoplastic cells (as required for the MAGE-A3 assay) was scraped off the NFR-stained slides and collected in a micro-centrifuge tube.

Total RNA extraction

Total RNA was extracted and purified from the manually dissected FFPE specimens using a modified version of the RNeasy FFPE kit (Qiagen, Valencia, Ca., USA). To improve the elimination of genomic DNA (gDNA), the kit procedure was modified by introducing an additional DNAse digestion step performed on the MiniElute column for 15 min at room temperature.

cDNA synthesis

Complementary DNA (cDNA) master-mix was prepared by mixing 1.5 μL M-MLV RT (Invitrogen, Carlsbad, Ca., USA), 6 μL 1st Strand buffer (5×), 3 μL of 0.1 M DDT, 0.75 μL BSA (Invitrogen) diluted to 2.9 mg/mL, 3 μL pre-mixed dNTP (Applied Biosystems, Warrington, UK), 0.3 μL pd(N) 6 random hexamers (Roche Molecular Diagnostics, Pleasanton, Ca., USA) and 0.75 mL RNAse Inhibitor (Invitrogen). cDNA synthesis was performed by mixing 15 μL cDNA master-mix and 15 μL RNA and incubating 8 min at 26°C, 45 min at 42°C and 5 min at 95°C.

qRT-PCR assays

MAGE-A3 and PRAME genes along with β-actin housekeeping gene were amplified by qRT-PCR using TaqMan chemistry (ThermoFisher) on the ABI 7900 system (Applied Biosystems) in 384-well plates. The assays were specifically designed for FFPE specimens with shorter amplicon size, 75 base pairs for MAGE-A3 and 69 for PRAME. Probe and primer sequences used are presented in Table 1.

Table 1.

Probe and primer sequences

Name	Sequence
MAGE-A3
MAGE-A3 probe	FAM-CCT-GTG-ATC-TTC-AGC-AAA-MGB
MAGE-A3 forward	TGT-CGT-CGG-AAA-TTG-GCA-GTA-T
MAGE-A3 reverse	CAA-AGA-CCA-GCT-GCA-AGG-AAC-T
PRAME
PRAME probe	FAM-TGC-TCA-GGC-ACG-TGA-T-MGB
PRAME forward	GAG-GCC-GCC-TGG-ATC-AG
PRAME reverse	CGG-CAG-TTA-GTT-ATT-GAG-AGG-GTT-T
β-ACTIN
Actin probe	TGC-TCG-CTC-CAA-CC
Actin forward	CTG-GAA-CGG-TGA-AGG-TGA-CA
Actin reverse	CGG-CCA-CAT-TGT-GAA-CTT-TG

The equivalent of 0.5 μL purified RNA was amplified in a final volume of 20 μL using the Gold Amplification System (Applied Biosystems) following the kit instructions. Primers and probes were used at concentrations of 1.2 μM and 0.2 μM, respectively. The amplification profile was 1 cycle of 2 min at 50°C, 1 cycle of 12 min at 95°C, 46 cycles of 15 sec at 95°C and 1 min at 60°C.

cDNA corresponding to 50 ng (100%) and to 0.5 ng (1%) of total RNA extracted from the human melanoma cell line MZ-2-3.0 (provided by Ludwig Institute of Cancer Research, Brussels, Belgium) was included after PCR amplification as positive control (or gene expression reference level, GERL). To detect possible PCR contamination, a negative control (H₂O) was included in each test plate. To verify successful gDNA removal for the MAGE-A3 assay, a minus reverse transcription (RT) reaction was performed for all specimens.

Calculation of tumor antigen expression levels

The relative antigen expression level within a tumor was calculated by the formula:

relative tumor antigen expression level = 2^(c-d)

where c = β-actin cycle threshold (Ct) obtained from the equivalent of 0.5 μL of tumor RNA and d = tumor antigen Ct obtained from the equivalent of 0.5 μL of tumor RNA.

A tumor specimen was defined as MAGE-A3 positive if the relative expression of MAGE-A3 to β-actin in the biopsy was ≥ 1% of the relative expression of MAGE-A3 to β-actin in the GERL cell line (4.96E-5). For PRAME, tumor specimens with expression ≥ 0.3% of the relative expression of PRAME to β-actin in the GERL cell line (1.73E-4) were categorized as PRAME-positive. The limit of quantification (LOQ) was 36.5 Ct for MAGE-A3 and 37 for PRAME.

Additional requirements for a specimen to be characterized as antigen positive were: 1) absence of PCR contamination using the negative control; 2) positivity of the GERL cell line; 3) β-actin Ct < 30; 4) a difference ≤ 2 between the Ct-values of the replicates; 5) absence of gDNA contamination for MAGE-A3 by requiring a difference of minimum two cycles between the MAGE-A3 PCR with and without RT, respectively. For PRAME, a gDNA control was not required because the primers span the intron-exon boundaries.

PCR test results were deemed invalid if there was gDNA contamination, inconsistent replicates or if the specimen had a mean β-actin Ct-value plus ΔCt exceeding the assay’s LOQ (those out-of-range specimens were ignored). A FFPE specimen was categorized as quantity-not-sufficient (QNS) if there was not sufficient tumor tissue to perform the assay or if the amount of RNA extracted was insufficient.

Statistical methods

The antigen expression results are presented as the proportion (with 95% confidence interval, CI) of antigen-positive tumors among those with a valid test result for the respective antigen. Co-expression rates were determined for tumors with valid test results for both antigens.

Results

Patient and tumor characteristics are presented in Table 2. Of the 156 tumors included, 150 (96.2%) were typed as transitional cell carcinomas and 134 (85.9%) were from muscle-invasive tumors (≥ T2).

Table 2.

Patient and tumor characteristics

Characteristics	Parameters or categories	Total (N = 156)
		Value or n	%	95% CI
Age group (years)	less than 45	3	1.9	0.4-5.5
	45-54	15	9.6	5.5-15.4
	55-64	36	23.1	16.7-30.5
	65-74	73	46.8	38.8-54.9
	75 or more	29	18.6	12.8-25.6
Gender	Female	21	13.5	8.5-19.8
	Male	135	86.5	80.2-91.5
Stage of tumor	T1	22	14.1	9.1-20.6
	T2	42	26.9	20.1-34.6
	T3	50	32.1	24.8-40.0
	T4	42	26.9	20.1-34.6
Primary tumor size (cm)	Mean	2.97	-	-
	SD	2.43	-	-
	Median	2.5	-	-
	Missing information	50
Histology	Transitional cell carcinoma	150	96.2	91.8-98.6
	Squamous cell carcinoma	4	2.6	0.7-6.4
	Other*	2	1.3	0.2-4.6

N = Number of patients; n/% = number/percentage of patients in a given category; Value = value of the considered parameter; SD = Standard deviation; 95% CI = 95% confidence interval;

^*Small cell carcinoma (one) and adenocarcinoma (intestinal type) (one).

Valid antigen expression test results for MAGE-A3 and PRAME were obtained from 144 (92.3%) and 152 (97.4%) specimens, respectively. Details of the reasons for invalid results or specimens categorized as QNS are provided in Table 3.

Table 3.

Validity of MAGE-A3 and PRAME expression tests

MAGE-A3
Characteristics	Categories	n	%	95% CI
Validity test (N = 156)	Valid	144	92.3	86.9-96.0
	Invalid	9	5.8	2.7-10.7
	QNS	3	1.9	0.4-5.5
Reason for PCR test invalidity (N = 9)	Specimen out-of-range	9	100	66.4-100
	gDNA contamination	0	-	-
	Inconsistent replicates	0	-	-
Reason for QNS (N = 3)	Not enough RNA	2	66.7	9.4-99.2
	Not enough tissue	1	33.3	0.8-90.6
PRAME
Characteristics	Categories	n	%	95% CI
Validity test (N = 156)	Valid	152	97.4	93.6-99.3
	Invalid	1	0.6	0.0-3.5
	QNS	3	1.9	0.4-5.5
Reason for PCR test invalidity (N = 1)	Specimen out-of-range	0	-	-
	Inconsistent replicates	1	100	2.5-100
Reason for QNS (N = 3)	Not enough RNA	1	33.3	0.8-90.6
	Not enough tissue	2	66.7	9.4-99.2

QNS = Quantity not sufficient; Total N = Total number of specimens for examination; N = number of specimens with a given characteristic; n/% = number/percentage of specimens in a given category of the characteristic; 95% CI = 95% confidence interval.

Of the valid specimens, 67 (46.5%) were MAGE-A3 positive while 49 (32.2%) were PRAME positive (Table 4). Valid co-expression results were obtained from 144 tumors and of these 29 (20.1%) were positive for both antigens and 86 (59.7%) were positive for at least one (Table 5). Antigen expression rates for subsets determined by patient and tumor characteristics are shown in Table 4. For both MAGE-A3 and PRAME, the rates of antigen-positive tumors within each of the subgroups considered (gender, age, tumor T stage, tumor size) were quite similar to the overall rates and no distinct pattern of variation was observed. In particular, increasing tumor T stage was not associated with higher rates of antigen-positive tumors.

Table 4.

Antigen expression rates according to patient and tumor characteristics

Characteristics and subgroups	MAGE-A3 positive				PRAME positive
	N	n	%	95% CI	N	n	%	95% CI
Overall	144	67	46.5	38.2-55.0	152	49	32.2	24.9-40.3
Gender
Men	124	56	45.2	36.2-54.3	131	39	29.8	22.1-38.4
Women	20	11	55.0	31.5-76.9	21	10	47.6	25.7-70.2
Age group (years)
less than 45	3	1	33.3	0.8-90.6	3	1	33.3	0.8-90.6
45 to 54	14	6	42.9	17.7-71.1	15	6	40.0	16.3-67.7
55 to 64	30	14	46.7	28.3-65.7	33	11	33.3	18.0-51.8
65 to 74	68	38	55.9	43.3-67.9	72	26	36.1	25.1-48.3
75 or more	29	8	27.6	12.7-47.2	29	5	17.2	5.8-35.8
Tumor stage
T1	19	12	63.2	38.4-83.7	21	8	38.1	18.1-61.6
T2	38	15	39.5	24.0-56.6	40	15	37.5	22.7-54.2
T3	50	24	48.0	33.7-62.6	50	14	28.0	16.2-42.5
T4	37	16	43.2	27.1-60.5	41	12	29.3	16.1-45.5
Size of the primary tumor
< 3 cm	55	24	43.6	30.3-57.7	58	17	29.3	18.1-42.7
≥ 3 cm	45	22	48.9	33.7-64.2	46	16	34.8	21.4-50.2

N = number of tumors in this category with valid expression test result for the antigen considered; n/% = number/proportion of tumors in this category with valid expression test result positive for the antigen considered; 95% CI = 95% Confidence Interval.

Table 5.

Rates of co-expression of MAGE-A3 and PRAME

MAGE-A3 and PRAME status	Total (N = 144)
	n	%	95% CI
MAGE-A3 positive and PRAME positive	29	20.1	13.9-27.6
MAGE-A3 positive and PRAME negative	38	26.4	19.4-34.4
MAGE-A3 negative and PRAME positive	19	13.2	8.1-19.8
Positive for at least one antigen	86	59.7	51.2-67.8

N = number of tumors with valid co-expression test; n/% = number/proportion of tumors with valid co-expression test result in a given category of MAGE-A3 and PRAME expression; 95% CI = 95% Confidence interval.

Table 6 indicates that the possibility of obtaining a valid antigen test result decreases with the age of the FFPE specimen, especially for the MAGEA3 assay.

Table 6.

Antigen test validity and rate of antigen-positive tumors according to year of resection

Year of resection	MAGE-A3						PRAME
	Valid test result			MAGE-A3-positive			Valid test result			PRAME-positive
	n	%	95% CI	n	%	95% CI	n	%	95% CI	n	%	95% CI
1991-1996 (N = 15)	8	53.3	26.6-78.7	3	37.5	8.5-75.5	13	86.7	59.5-98.3	3	23.1	5.0-53.8
1997-2001 (N = 36)	33	91.7	77.5-98.2	15	45.5	28.1-63.6	34	94.4	81.3-99.3	15	44.1	27.2-62.1
2002-2007 (N = 105)	103	98.1	93.3-99.8	49	47.6	37.6-57.6	105	100	96.5-100	31	29.5	21.0-39.2

N = number of tumors resected during this period; n/% = number/percentage of tumors resected during this period with a valid test result or positive expression for the antigen considered; 95% CI = 95% Confidence Interval; Difference between N and n is due to invalid or QNS specimens.

Discussion

We used qRT-PCR assays to investigate the expression rates of two tumor antigens in FFPE tumor specimens from patients with bladder cancer of all stages. MAGE-A3 and PRAME were expressed by 46.5% and 32.2% of the tumors, respectively, while 59.7% were positive for at least one antigen.

Previous studies of the rate of MAGE-A3 expression in bladder tumors have reported rates between 35% and 59% [25-28]. Dyrskjøt et al. [29] were the first to characterize PRAME expression in urothelial carcinoma of the bladder. They reported that 20% of the tumors expressed PRAME and 43% MAGE-A3. These studies mostly used fresh frozen (FF) tumor material, while one study used both tissue types [27] without detailing the results according to type. In contrast, we used only archived FFPE tissue specimens after manual dissection including an enrichment step to focus on tumoral content, as laser microdissection is not applicable for diagnostics in a clinical setting due to technical limitations.

RNA from snap-frozen fresh tissue is assumed intact and well-preserved and FF tissue is therefore widely recognized as being the most reliable for molecular genetic analyses [30]. However, FF specimens are difficult to collect and handle in large scale studies with centralized testing as they are complicated to process and costly to store [31]. Routine practice in hospitals worldwide is to prepare FFPE specimens by fixation in non-buffered 6% formalin to be used for histopathological diagnosis. The formalin fixative ensures an excellent preservation of the tissue morphology, which allows a finer diagnosis and a more accurate assessment of the tumor cell content than what is possible with FF tissue [32]. FFPE tissues are stable at room temperature and easy to store or collect for centralized testing. To enable molecular testing based on these established and relatively low cost procedures and protocols it is of great interest to develop molecular test methods suitable for FFPE tissues.

Methods used for the preparation of FFPE specimens lead to well documented chemical modifications and partial degradation of the RNA [33]. Fixatives react with proteins and nucleic acids causing cross-linking and leading to low yields (at least two-fold lower yield of RNA from FFPE than from FF specimens) and shorter RNA lacking in poly-A tail [32,33]. The fragmentation of RNA due to the preparatory processing of FFPE specimens leads to a loss of intact amplicon, which implies that shorter amplicons (maximum 100-200 base-pairs) must be used if the downstream analyses are based on qRT-PCR [31,34]. Other factors during the tissue fixation or paraffin embedding, such as temperature shifts and hypoxia, the time between resection and fixation plus the amount of fixative used relative to the specimen size, may also affect the quality of the RNA [33]. In particular, the duration of fixation is a critical factor with regard to the quality of the RNA extractable for downstream analyses [30].

Partly motivated by the vast potential of future studies coupling molecular analyses of archived FFPE tissues with long-term clinical and outcome data, several recent studies have assessed the comparability of molecular analysis results obtained from patient matched FFPE and FF tumor tissues from various types of tumors and using different types of test methods such as qRT-PCR and microarray [30-32,34,35]. The overall conclusion of the comprehensive analyses performed in these studies is that results obtained with both types of tissue generally concord well despite the likely real differences between FF and FFPE results (e.g., due to variations in the protocols used, [33]). These studies show that, taking FF tissue results as the reference, meaningful biological information may be obtained from the partially degraded RNA extracted from FFPE tissue [33].

More specifically, it has been found that while DNA microarray-based studies are highly sensitive to RNA degradation, qRT-PCR assay methods are more robust and tolerate some partial degradation of RNA [31,34]. To correct for differences in the quality and the quantity of RNA used as input for qRT-PCR, one or more reference gene(s) should be used for normalization. The relative expression of the gene of interest in a particular tissue specimen can then be determined as the ΔCt value, or the Ct value of the investigated gene minus the Ct value of the reference gene. The shift in Ct value as a function of increasing RNA degradation is generally observed to be similar for the test gene(s) and the reference gene, so the ΔCt values calculated are fairly constant [34]. The studies mentioned mostly report high correlation coefficients (range 0.77 to 0.91) for comparisons of relative gene expression levels measured by qRT-PCR assays of matched FFPE and FF specimens. However, low correlation has been reported for a certain number of genes, in particular genes with a low level of expression (high minimum Ct value) and genes with highly variable normalized expression levels [31]. This indicates that the similarity of gene expression levels found in FFPE and FF tissue, respectively, must be verified for each particular gene investigated and cannot be taken for granted [31]. Such verification was part of the testing and validation of the MAGE-A3 and PRAME specific assays used in this study. At the time of development of the tests, use of a single reference gene was the conventional approach, and β-actin was selected as the most appropriate considering its expression rate.

Immunohistochemistry (IHC) might be an alternative to antigen expression testing by qRT-PCR assays, but this requires prior identification of a MAGE-A3 specific antibody, in particular one that can distinguish between MAGE-A3 and MAGE-A6 (which has 98% identity with MA-GE-A3). Efforts to develop a MAGE-A3-specific antibody to be used for IHC testing of FFPE material, which can be done locally at low cost by any hospital laboratory, are ongoing.

Several relatively small studies comparing gene expression results from patient-matched tissues using both qRT-PCR and IHC generally find good concordance albeit with some discrepancies [36-39]. They generally recommend combining the two methods, as the qRT-PCR allows a quantification of the level of expression while the IHC provides information about the heterogeneous distribution of antigen expression in the tissue and even on the level of expression in individual cells. At the current state of development it is not known if the therapeutic activity of an antigen-specific immunotherapeutic is dependent on the level of gene expression, so to gain more insight on this it is important to measure the expression level with some precision.

Of note, FFPE tissue tested by a qRT-PCR assay is used for patient screening in a recently initiated randomized Phase II trial investigating an immunotherapeutic agent targeting MAGE-A3 in muscle invasive bladder cancer (the MAGNOLIA trial, NCT01435356). In prospective studies with antigen screening, instructions should be given to the study sites for the preparation of FFPE tissues (e.g., fixative, fixation time) to ensure conformity with what the antigen assays have been validated for.

We obtained valid MAGE-A3 and PRAME expression results from 92.3% and 97.4% of the tumors, respectively. Differences in the proportion of specimens giving valid test results were anticipated because tumors express the antigens at different levels and the LOQ is not the same for MAGE-A3 and PRAME. As expected based on previously reported findings [35], the age of the FFPE specimen had an impact on the proportion of valid test results, especially for MAGE-A3 (Table 6). This indicates that although RNA degradation slows down once the fixation is done, it does not stop completely [34]. For the 105 specimens resected and fixated five years or less before testing, the rate of valid tests was above 98%. For clinical use, the degradation of RNA over time would not be an issue as the antigen testing would be carried out soon after the formalin fixation.

Patard et al. [25] found that the rate of MAGE-A3 positive tumors increased with tumor stage. Likewise, Dyrskjøt et al. [29] reported statistically significant correlations between bladder tumor stage (and grade) and both PRAME and MAGE-A3 expression. For muscle-invasive tumors an association between increasing tumor stage and higher rates of antigen expression could not be confirmed in our data set but this may be due to the small subgroup samples. Another hypothetical explanation of these divergent findings may be that the amount of tumor cells after manual dissection is controlled more effectively with FFPE than with FF tissue, an additional advantage of molecular testing on FFPE tissue.

Until now, no truly tumor-specific antigen has been identified. This would be a gene expressed by all cells of a tumor and by strictly no type of normal cells and would be essential for survival and proliferation of tumor cells. With its expression in normal cells limited to germ cells in the immune privileged testis site, MAGE-A3 is sometimes considered as close to this ideal. For PRAME, however, there is a theoretical possibility that its expression in certain types of normal cells in various organs like the kidney or adrenal gland may lead to auto-immune toxicity in patients treated with a PRAME-specific immunotherapeutic [18]. However, the level of PRAME expression in normal cells is very low or minimal in non-germ line cells [18], so the theoretical probability that these normal cells will be recognized as a target by T cells is also very low. It is noticeable that early clinical studies of PRAME-specific immunotherapeutics in various cancers have reported good safety results with no issues related to immune-mediated disorders [40-42].

Our findings indicate that a large proportion of bladder cancer patients may be candidates for active immunotherapy targeted at MAGE-A3 or PRAME. As there are important unmet medical needs for all disease stages, clinical development of antigen-specific immunotherapeutics could be aimed both as a substitute for BCG for patients with non-muscle-invasive bladder cancer that are refractory or intolerant to BCG and as adjuvant therapy to radical cystectomy for patients with muscle-invasive disease. Combining an antigen-specific immunotherapeutic with BCG may presumably have synergistic effects [43].

Immunotherapeutics targeting two or more tumor-associated antigens may also be of interest. A greater proportion of patients would be able to benefit from such combinations, which would also compensate for tumor heterogeneity. Another theoretical benefit of combining several tumor antigens in immunotherapeutics is that they may diminish the risk of tumor escape through development of antigen-specific immunosuppressive escape mechanisms [6].

In conclusion, this study has established the feasibility of using archived FFPE tumor tissues for retrospective analysis of tumor antigen expression by means of qRT-PCR. The antigen expression rates observed in FFPE tissues correspond to those reported in other studies using FF tissue. Contrary to other studies, we did not find an association between disease stage and the rate of antigen expression. Our results suggest that MAGE-A3 and PRAME both may serve as targets for active immunotherapies against bladder cancer.

Disclosure of conflict of interest

Thierry Coche and Olivier Gruselle are employees of the GSK group of companies. Patrick Therasse was an employee of the GSK group of companies at the time of the study and manuscript development. Patrick Therasse and Thierry Coche own stock/stock options in GSK group of companies. All other authors declare that they have no conflict of interest. Universitaire Ziekenhuizen (UZ) Leuven received honorarium for the selection of paraffin blocks for this study and is applicant for a patent related to “Fatty Acid Enzymes as targets for Cancer Diagnostics and Therapeutics (PCT/EP2013/056790, co-applicant)”.