APRIL is a novel clinical chemo-resistance biomarker in colorectal adenocarcinoma identified by gene expression profiling

The study was approved by the North of Scotland Research Ethics Committee. Patients provided informed consent in accordance with the regulations and instructions of the North of Scotland Research Ethics Committee for study participation, including use and publication of results. Full clinicopathological details are provided in table 1 and 2 and in Additional File 1. Patients were selected for either SCRT or CRT based upon MRI staging features [12]. All the radiotherapy was CT planned, using a 3 field technique (posterior and two lateral fields), multileaf collimation and with patients having a full bladder during the radiotherapy. Surgery was performed either the following week, for SCRT patients, or 6 to 8 weeks after completion of chemo-radiotherapy.

Tumour biopsies were collected at the time of endoscopic diagnosis of rectal adenocarcinoma and placed immediately into RNAlater (800 μl) (Ambion, Austin, Texas). Tumour biopsies collected at time of curative surgical resection were placed immediately into normal saline and a pathologist provided a representative tumour biopsy, which was placed immediately into RNAlater within 30 minutes (800 μl). Tissues were stored in RNALater at 4°C overnight (16-18 hours), then washed in 500 μl ice cold RNase free PBS (Ambion, Austin, TX) and snap frozen in liquid nitrogen. Long-term storage of tissues was at -80°C. Before RNA extraction, histological diagnosis and features were confirmed by frozen section histology. Extraction and purification of total RNA was performed using TRIZOL reagent (Invitrogen, Carlsbad, CA) and RNeasy Microkits (Qiagen, Venlo, The Netherlands), according to the manufacturer's instructions. Quantification of total RNA was performed by spectrophometry (260/280 ratio 1.9 to 2.2 for all samples). Quality of total RNA and cRNA was assessed using a BioAnalyser 2100 (Agilent technologies, Palo Alto, CA). Target preparation for the Affymetrix Genechips™ was according to manufacturer's instructions (Affymetrix, Santa Clara, CA). Specifically, 4 μg of total RNA was used for reverse transcription and synthesis and amplification of biotin labelled cRNA using the One cycle target labelling and control reagents. Clean-up of biotin-cRNA was performed with RNeasy Minikits (Qiagen, Venlo, The Netherlands). Fragmentation was performed using 20 μg of biotin-labelled cRNA. A hybridisation cocktail was prepared from 15 μg which was first hybridised to Test 3 GeneChips™ to assess sample quality (GAPDH 3':5' < 3 and Actin 3': 5' < 3) and then to HGU133 Plus2.0 GeneChips™ (10 μg) for gene expression analysis. Procedures for hybridisation, washing, staining and scanning of chips were carried out according to standard protocols (Affymetrix, Santa Clara, CA).

Analysis of the gene expression data is described in detail in Additional file 2 and as described previously [11, 13]. Raw data for gene expression is provided in MIAME complaint format in Array express, accession number E-MEXP-1901

Description of the Tissue Microarray (TMA) is provided in previous publications [14]. A total of 268 colorectal tumours and 50 normal colon cores are represented, with 1 core per case. During the staining procedure 34 (13%) tumour cores were lost, leaving cores from 234 patients available for assessment. Antigen retrieval was performed by microwaving in 10 mM citrate (pH 6.0) for 20 minutes. An autostainer (Dakocytomation, Glostrup, Denmark) was used for staining the sections using a mouse monoclonal primary antibody for human APRIL/TNFSF13 (1:60 dilution, Abcam, Cambridge, UK) and Chemate-Envision detection system (Dakocytomation, Glostrup, Denmark), according to the manufacturer's instructions. All sections were double scored by 2 independent investigators who were blinded to the clinical data. Scoring discrepancies were resolved by examination of sections at a double-headed microscope. Sections were scored positive or negative for tumour and/or stromal staining. In addition tumour staining intensity was scored as weak, moderate or strong.

Continuity corrected χ² test, with Fisher's exact test where appropriate, was used for binary categorical variables, Pearson's χ² test for non-binary categorical variables and Student's t-test for numerical variables. Kaplan-Meier curves were constructed to assess survival and the log rank test to assess statistical significance. The Cox proportional hazards model was used for multivariate analysis of survival. Two-sided p values of less than 0.05 were considered significant. All analyses were performed using SPSS for Windows, version 13.0 (SPSS Inc, Chicago, IL).

In a pilot transcriptomics study of rectal cancer patients, we used oligonucleotide microarrays to profile the expression of over 47000 transcripts representing 38562 human genes in rectal tumour biopsies before and after pre-operative treatment with CRT (n = 4 patients); table 1). Rectal tumour biopsies before and after SCRT (n = 4 patients; table 1) were also analysed to enable comparison of gene expression changes in patients treated with 5FU-based chemo-radiotherapy with those observed in patients receiving radiotherapy alone. Rectal tumour biopsies, at diagnosis and surgical resection, from two patients who did not undergo any pre-operative treatment (table 1) were used to identify treatment-independent gene expression changes.

SOPs were developed and validated to allow collection of tissues at endoscopic diagnosis and at surgical resection, whilst preserving RNA integrity. Total RNA extracted from these tissues (10-30 mg) in this pilot study provided sufficient yield (8 to 40 ug) and quality total RNA for gene expression analysis on Affymetrix oligonucleotide microarrays. Raw gene expression data is provided in MIAME complaint format in Array express, accession number E-MEXP-1901.

Threshold and probabilistic filtering of the data (see Additional file 2) identified 86 genes (91 probe sets) consistently, significantly and specifically altered following 5FU-based CRT and 51 genes (58 probe sets) following SCRT (see Additional File 3 for details of genes and fold change following therapy). Hierarchical cluster analysis, highlights 2 distinct clusters of genes up-regulated or down-regulated following CRT (figure 1A) or SCRT (figure 1B). The expression profiles of each of these gene sets clearly separates pre- and post-treatment samples into two primary clusters for each treatment group (figure 1). A matrix analysis (DMTv1.0, Affymetrix, CA) of therapy-altered gene sets identified using threshold filtering alone (see Additional file 2; 697 probe sets in CRT group and 570 in SCRT group, including 86 overlapping), reveals that these genes sets are significantly non-overlapping (p = 0.010), demonstrating highly distinct alterations to the tumour transcriptome following treatment with SCRT or 5FU-based CRT.

The biological functions of the CRT and SCRT altered gene sets were evaluated (additional file 4). While many of the same key biological pathways are identified in each treatment group, consistent with a co-ordinated transcriptional response, there are some pathways only altered following CRT and some pathways (cell death and cell cycle) where there is numerically significantly more change in gene expression in the CRT treated patients (additional file 4).

This represents an initial pilot study of the first samples in our rectal cancer patient cohort. It is important to note that the small sample size, necessitates validation of these candidate gene expression changes in a larger cohort. The primary aim of this study was to identify candidate 5FU resistance markers in rectal tumours, in a pilot discovery study using a transcriptome-wide approach and to validate key candidate/s that may have mechanistic relevance in a larger cohort. Identification and validation of one such marker is described below.

As we were interested in potential mediators of 5FU resistance in rectal tumours in vivo, we further mined the gene expression analysis using a pathway focussed analysis of cell deaths pathways, including those involved in regulation or execution of caspase-dependent apoptotic, caspase-independent and necrotic cell death genes (n = 2177 genes; additional file 5). Threshold and probabilistic filtering of the gene expression data identified 17 cell death genes consistently and significantly altered in rectal tumours following chemo-radiotherapy (additional file 6). Several of these genes have been implicated in colorectal cancer pathogenesis and the pathogenesis of other cancers, and also radioresistance, but none previously in 5FU chemoresistance (for more details see additional file 6). Comparison of the 17 cell death genes altered in response to 5FU based CRT in tumours from rectal cancer patients, with gene expression changes identified in our previous study of 5FU resistant cancer cell lines [11], demonstrated 4 of the 17 genes up-regulated following CRT (but not radiotherapy alone) in rectal cancer patients and in 5FU-resistant cancer cells compared to the sensitive parental lines(See additional file 6, Table S6.1). This included the TNF superfamily ligand, APRIL (TNFSF13).

APRIL has been characterised as promoting cell survival and cell proliferation and this involves NFκB activation [15‐19]. In addition, APRIL mRNA has been shown to be increased in colorectal tumours compared to normal mucosa [17]. These data supported further investigation of a putative functional role for APRIL in clinical 5FU chemo-resistance.

APRIL protein expression was evaluated in 234 resected colorectal adenocarcinomas and 50 normal colon or rectal mucosa specimens (table 2). APRIL protein was not expressed in normal colon tissues but was, as expected, expressed in both colorectal tumour cells and the tumour stroma (Table 3 and figure 2). Tumour cell staining was observed in the cytosol and membrane of tumour cells (figure 2). Stromal staining was evident in both the extracellular matrix and also in stromal cells (figure 2).

We examined the relationship between APRIL protein expression and survival after surgical resection. We prospectively determined that we would evaluate both tumour cell and tumour stromal expression of APRIL protein due to its characterized biological function as a secreted autocrine and/or paracrine molecule. There was no significant relationship between APRIL protein expression in tumour cells and survival (Additional file 7). In contrast, expression of APRIL protein in the tumour stroma was associated with poor survival (n = 234, p = 0.019, figure 3a), including in stage III patients (n = 102, p = 0.016, figure 3b), but was not associated with survival in Stage I or II (n = 46 p = 0.601 and n = 86 p = 0.440, respectively, Additional File 7).

In light of our hypothesised role of APRIL in 5FU resistance, we stratified the Stage III patients according to whether or not they received adjuvant chemotherapy with 5FU following surgical resection of their primary tumour. Stage I and II patients did not receive adjuvant chemotherapy in this series. Tumour stroma expression of APRIL protein is only associated with worse survival in those patients treated with adjuvant 5FU and there is no relationship with survival in Stage III patients not treated with adjuvant chemotherapy (n = 102, p < 0.001, figure 3c). In 5FU treated Stage III patients (n = 63), median survival for stroma positive is 36 months with predicted 5 year survival 42.0% (95% confidence interval 11.8% - 72.2%); median survival not yet reached for stroma negative and predicted 5 year survival is 85% (95% confidence intervals 71.7%-98.6%). Multivariate analysis confirms expression of APRIL protein in the tumour stroma as an independent prognostic factor in chemotherapy treated Stage III patients, with a HR of 6.25 (95% CI 1.48-26.32, p = 0.013, table 4).

The survival of the 5FU treated Stage III colorectal cancer patients who express APRIL protein in the tumour stroma parallels survival observed in Stage III patients who did not receive adjuvant therapy (treatment decision due to patient or physician preference), irrespective of APRIL protein expression (figure 3c). In contrast, the APRIL negative patients have an excellent predicted 5 year survival and have a clear and statistically significant (p < 0.001) survival benefit compared to untreated or APRIL positive 5FU treated patients (figure 3c). These data suggest that APRIL has no prognostic impact in colorectal cancer treated by surgical resection alone, but has predictive impact for benefit from adjuvant 5FU in colorectal cancer patients.