Circulating mRNA profiles differ before and after CRT

Using an FDR of 0.5% and a fold-change more than 2, Significance Analysis of Microarrays (SAM) identified 136 genes differentially expressed between before and after CRT (Fig. 1a). Among the differentially expressed genes, 46 genes were up regulated and 90 genes were down regulated. The gene ontology (GO) term enrichment and functional annotation analysis by Database for Annotation, Visualization, and Integrated Discovery (DAVID) identified 5 significant clusters with the one-tail Fisher exact probability value of 0.025 in the annotation category of all GO biological progress terms. The enriched clusters and related gene groups were summarized in Supplementary Table e1. The annotation cluster one consists of 41 genes and the gene-annotation association map is shown in Fig. 1b. The immune system process and immune response were the two most significant functions altered by CRT (Bonferroni adjusted p value<0.001). The supervised hierarchical clustering analysis revealed nearly perfect segregation of the pre-CRT from the post-CRT samples and the resulting heat map is shown in Fig. 1c.

Open in a separate window

Figure 1

The circulating mRNA profiles were significantly altered between before and after preoperative chemoradiation (CRT). (a) Significance Analysis of Microarray (SAM) plot. (b) The gene ontology term enrichment analysis was performed by the Database for Annotation, Visualization, and Integrated Discovery. Red and white colors denote the positively and negatively reported corresponding gene-annotation association, respectively. (c) Supervised hierarchical cluster analysis identified 136 mRNAs with expression that differed significantly between before and after CRT. Green and red colors denote down- and up-regulated genes, respectively.

Expressed circulating mRNAs differ between complete responder and non-complete responder

By using stringent statistical methods, there was no significant difference in the expression profiles between the CR group and non-CR group. Since the neoadjuvant CRT significantly altered the gene expression signature of peripheral blood cells, we hypothesized that changes in expression of circulating mRNA with neoadjuvant CRT may be predictive for pathological response. BAM identified ten candidate mRNAs (AFTPH, C10ORF76, CCNL1, FAM13A1, FAM84B, HIST1H4H, HIST2H4A, IFI27, KCNRG, and SEPT4) for further evaluation (Table 2). A second analysis showed that AFTPH (aftiphilin), CCNL1 (Cyclin-L1), FAM13A1 (family with sequence similarity 13, member A1), FAM84B (family with sequence similarity 84, member B), HIST2H4A (histone cluster 2, H4a), and SEPT4 (septin 4) were truly differentially expressed between patients with pCR and non-pCR.

The three most differentially expressed circulating mRNAs (CCNL1, FAM84B, and SEPT4) were assayed by quantitative RT-PCR. As shown in Fig. 2a, the change in the expression of FAM84B (p=0.02) but not CCNL1 (p=0.98) or SEPT4 (p=0.41) was significantly different between the pCR and non-pCR groups.

Open in a separate window

Figure 2

Circulating mRNA in predicting response to and outcome of preoperative chemoradiation. (a) The change in circulating levels of CCNL1, FAM84B, and SEPT4 mRNAs between pathological complete responders (pCR) and non-complete responders (non-pCR) measured by quantitative reverse-transcriptase polymerase chain reaction. Data shown as a scatter plot and the intersecting line shows the median value with the interquartile range. (b) Receiver-operating characteristic curve shows the performance of fold-change in FAM84B mRNA expression in predicting the pathological complete response, with the area under curve being 0.73. (c) Kaplan-Meier curves of the disease-free survival stratified by fold-change in FAM84B mRNA expression using a cutoff value of 0.3 (p=0.15).

FAM84B mRNA as a novel biomarker for ESCC

The fold change (2^-CT) of circulating FAM84B mRNA obtained by qRT-PCR was used in ROC curve analysis to evaluate its predictive ability. The median value of fold change was 0.3 (ranged 0.1 to 43.1). Figure 2b shows that the area under the two-class ROC curve was 0.73 (p=0.02). With a cutoff value of 0.3, the sensitivity, specificity, positive predictive value, and negative predictive value were 75%, 67%, 63%, and 78%, respectively. Patients with fold change <0.3 had a significantly higher chance to achieve pCR (p=0.01). In survival analysis, patients with fold change <0.3 had a trend toward longer progression-free survival (median not reached versus 18 months, p=0.15; Fig. 2c).

Serum FAM84B protein level change is associated with treatment response

To verify that change in circulating FAM84B mRNA is predictive of treatment response, we performed multiplex PLA of serum samples from 79 patients receiving combined modality therapy¹⁵. Overall, the changes in serum FAM84B protein level showed a trend toward association with pathological response (p=0.08). After removal of two outliers, decrease in serum FAM84B level after CRT was greater in the pCR group than non-pCR group (p=0.02; Fig. 3a). Among the 46 patients in an independent validation cohort, the pCR group (14 patients) had a strong trend toward greater reduction of serum FAM84B protein (p=0.06; Fig. 3b).

Open in a separate window

Figure 3

Box-and-Whisker plots at the 5th and 95th percentiles of FAM84B serum protein level quantified by proximity ligation assay. (a) The changes after chemoradiation between pathological complete responders (pCR) and non-complete responders (non-pCR) (p=0.02). (b) The changes between pCR and non-pCR groups for an independent validation cohort (p=0.06).

Overexpression of FAM84B in tissue samples of ESCC

Thirty-two patients had pretreatment tissue sample available for immunohistochemistry analysis, and 26 samples (81%) showed ≥ moderately staining intensity for FAM84B (Table 1). Seventeen patients had post-esophagectomy residual tumors available for immunohistochemistry analysis, and 16 samples (94%) showed ≥ moderately staining intensity for FAM84B (Table 1). The FAM84B protein was highly expressed in cancerous tissue while the paired normal esophageal epithelium was negative for IHC staining (Fig. 4a). The distributions of FAM84B staining intensities on pretreatment tumor biopsies were not different between pCR and non-pCR groups (p=0.99; Fig. 4b). Interestingly, patients with high intensity on pretreatment tumor biopsies had non-significant worse progression-free survival than those with low intensity (median 18 months versus not reached, p=0.098; Fig. 4c)

Open in a separate window

Figure 4

FAM84B was overexpressed in esophageal squamous cell carcinoma (ESCC) tumor biopsies. (a) The immunohistochemistry analysis of FAM84B from patients with paired specimens of the cancerous lesion and normal epithelium of the esophagus. (b) Box-and-Whisker plot of FAM84B staining intensities between pathological complete responders (pCR) and non-complete responders (non-pCR) (p=0.99). (c) Kaplan-Meier curves of the disease-free survival stratified by FAM84B staining intensities (p=0.098).

Overexpression of FAM84B in ESCC cell lines

Ten human cell lines were evaluated for FAM84B protein expression by Western blotting. The human esophageal non-neoplastic epithelial cell line Het-1A showed negative or weak expression. Expression by the 9 esophageal SCC cell lines was strong in CE-81T/VGH, KYSE-30, KYSE-410, and OE-21, moderate in KYSE-150, and weak or absent in KYSE-70, KYSE-270, CE-48T/VGH (data not shown), and CE-146T/VGH (data not shown; Fig. 5a). Overall, 56% of esophageal SCC cell lines were positive for FAM84B overexpression.

Open in a separate window

Figure 5

FAM84B was overexpressed in esophageal squamous cell carcinoma (ESCC) cell lines and associated with tumor progression. (a) The cropped Western blot of FAM84B protein from lysates of the esophageal non-neoplastic squamous epithelial cell line (Het-1A) and ESCC cell lines (CE46T/VGH & CE146T/VGH are not shown). (b) Growth curves of ESCC CE81T/VGH xenograft tumors. Mean volume ±standard deviation (n=5 per group) are plotted as a function of time since injection. Cropped immunoblots of CE81T/VGH cells transfected with shRNA against FAM84B. Immunohistochemistry stains of tumor xenografts confirmed the FAM84B knockout tumor has no FAM84B protein expression. Images were taken at 40X.

Neoadjuvant treatment

All patients underwent combined modality therapy in accordance to approved procedures. Patients received conformal radiotherapy with 40Gy in 20 daily fractions, 5 fractions per week over 4 weeks and were treated with concurrent cisplatin-based chemotherapy. The details of chemotherapy regimens are shown in Table 1. Curative surgery with radical esophagectomy and en-bloc lymph node dissection was performed 2–10 weeks (median 7 weeks) after completion of CRT.

Definition of pathological response

Pathological complete response is defined as no residual tumor is the esophagus or dissected lymph nodes. Presence of any residual esophageal tumor regardless of the percentage of regression, any metastatic lymph node(s), or metastatic disease is considered non-pCR.

Blood sampling

Whole blood was withdrawn in accordance to an IRB-approved procedure within one week before the start of CRT and within 4 weeks after the completion of CRT.

Microarray analysis

Total RNA was prepared using the PerfectPure RNA blood kit (5 Prime Inc., Gaithersburg, MD) according to the protocol provided by the manufacturer of Tempus™ tubes. Globin mRNA was depleted from each total RNA sample using the GLOBINclear™-Human kit (Ambion, Austin, TX). RNA concentration and quality were determined using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE) and Agilent 2100 Bioanalyzer/RNA 600 LabChip kit (Agilent Technologies, Palo Alto, CA). RNA of sufficient quality with an RNA integrity number >7.0 was used to synthesize cRNA for microarray analysis.

Globin depleted RNA (500ng) was primed with the T7 Oligo(dT) primer and amplified using an Illumina TotalPrep RNA Amplification kit (Ambion) to synthesize the first strand cDNA containing a T7 promoter sequence. The cDNA then underwent second strand synthesis, RNA degradation by DNA Polymerase and RNase H, and a clean-up process to remove excess RNA, primers, enzymes, and salts that would inhibit in vitro transcription. In vitro transcription was employed using the double-stranded cDNA as a template and T7 RNA polymerase to synthesize multiple copies of biotinylated-cRNA. The labeled cRNA was purified by a Filter Cartridge and quantified by a NanoDrop spectrophotometer (NanoDrop Technologies, Wilmington, DE). Labeled cRNA (1.5μg) was hybridized to Illumina Human WG-6 v3 BeadChips (Illumina, San Diego, CA). The intensity of the bead’s fluorescence was detected by the Illumina BeadArray reader and the data analyzed using Bead Studio software. The data were pre-processed using the lumi R/BioConductor package. Array background adjustment was performed using a quantile normalization algorithm, and the data were logarithm base-2 transformed for further statistical analysis.

Probes were filtered out if their detection call p-value was >0.01 in all samples (eliminating 31,757 [65%] of 48,804 probes). All data were minimum information about a microarray experiment (MIAME) compliant, and the raw data have been submitted to the Gene Expression Omnibus database (accession number GSE43519).

Quantitative reverse transcription polymerase chain reaction (qRT-PCR)

Reverse transcription of RNA was performed with a High Capacity cDNA RT kit (Applied Biosystems, Foster City, CA) using random primers and 1μg of total RNA from samples as template. The Taqman® gene expression assays are used to validate selected candidate genes. Taqman® quantitative real time PCR was carried out using an ABI 7900HT Fast Real-time PCR System (Applied Biosystems, Foster City, CA). The PCR conditions were as follows: 50 cycles of denaturation at 95°C for 15sec and 1min of annealing and elongation at 60°C. Beta-actin (Hs99999903_m1) was selected as an internal control. All reactions for each target gene and endogenous control were measured in triplicate. Relative gene expression data was calculated using the -cycle threshold (CT) method.

Proximity ligation assay

Multiplex proximal ligation assay (PLA) was performed on serum samples as described previously^15,33. Briefly, samples were thawed and mixed in a 1:1 ratio with buffer (Olink AB, Uppsala, Sweden) for undiluted assay. For probing, 2μL of the buffered serum sample was mixed with 2μL of probe, and incubated for 2hours at 37°C to allow the probe to bind analytes. Ligation was achieved by incubating the reaction mixture with the probed samples for 15minutes at 30°C to dilute and separate any free probes. To stop ligation, 2μL of uracil-DNA excision mix (Epicentre, Madison, WI) was added and incubated for 15minutes at room temperature. Preamplification of the bar-coded amplicons was performed with pool-PCR primer (Platinum Taq kit, Invitrogen Corp., Carlsbad, CA) for 13 cycles with a 4-minute extension at 60°C, and the preamplification products were diluted 10-fold in 1×TE. For each protein assayed, a separate quantitative PCR reaction was required in a 384-well plate with target protein–specific quantitative detection primer lyophilized at the bottom of each well, to which sample and iTaq SYBR Green Supermix with Rox (Bio-Rad, Hercules, CA) were added. Real-time quantitative PCR was performed with a sample volume of 10μL per well for 40 cycles at 95°C for 15seconds and 60°C for 1minute. To ensure standardization of values for each biomarker investigated, all samples were simultaneously probed and evaluated on a single 384-well plate with a PBS-BSA blank well. Cycle threshold (CT) values from quantitative PCR were converted into an estimated number of starting amplicons, or PLA units, by calculating 10 (-0.301×CT+11.439).

Immunohistochemistry analysis

Four-μm sections were cut from the paraffin-embedded tissue samples. Antigen retrieval was performed using Trilogy™ (Cell Marque, Rocklin, CA) at 121°C for 10minutes. The sections were treated with 3% hydrogen peroxidase to quench endogenous hydrogen peroxidase activity, treated with Power Block (BioGenex, San Ramon, CA) for 20minutes to block nonspecific reactions to other antigens, incubated with the commercial FAM84B rabbit polyclonal antibody (1:100; Proteintech Group, Chicago, IL) overnight at 4°C, stained with diaminobenzidine (DAB) (BioGenex), washed additional times, and counter-stained with hematoxylin for 3minutes. The staining intensity was scored from 0 to 3+ as follows: 0, no staining; 1+, weak staining; 2+, moderate staining; 3+, strong staining. High intensity was defined as staining intensity of 2~3+ and 3+, and low intensity was defined as staining intensity from 0 to 2+.

Cell culture and western blotting

The human non-neoplastic esophageal epithelial cell line Het-1A (American Type Culture Collection [ATCC], Rockville, MD) was maintained in RPMI 1640 (Gibco BRL, Paisley, UK). The human esophageal squamous cell lines CE-48T/VGH, CE-81T/VGH, and CE-146/VGH (Bioresource Collection and Research Center [BCRC], Hsiuchu City, Taiwan) were maintained in DMEM (Gibco), while KYSE-30, KYSE-70, KYSE-150, KYSE-270, KYSE-410, OE-21 were gifts of Dr. Chih-Hung Hsu and cultured in RPMI 1640. All culture media were supplemented with 10% (v/v) fetal bovine serum and 1% antibiotics (penicillin-streptomycin-amphotericin B). Cells were cultured at 37°C in a humidified atmosphere of 5% CO2/95% air. Mycoplasma contamination was tested on a regular basis. Whole cell lysates were prepared using cell lysis buffer (Cell Signaling, Beverly, MA) plus protease inhibitor cocktail (Thermo Fisher Scientific, Waltham, MA). Protein concentration was determined with Bradford reagent (Bio-Rad, Hercules, CA). Proteins in samples (each containing 100μg of total protein) were separated by 10% sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE), transferred to polyvinylidenedifluoride membranes, and immunoblotted with antibodies against human FAM84B (Abcam, Cambridge, MA) and β-actin (Santa Cruz Biotechnology, Santa Cruz, CA). Bound antibodies were detected using the appropriate horseradish peroxidase-coupled secondary antibodies followed by enhanced chemiluminescence with luminol substrate (Millipore, Bedford, MA).

Gene expression knockdown

To establish a stable clonal cell population with knockout FAM84B, specific shRNA constructs cloned into plasmid pLKO.1-puro are obtained from The RNAi Consortium (TRC) via National RNAi Core Facility (Academia Sinica, Taiwan ROC). The shRNA-encoding plasmid delivery is conducted by a 3-plasmid lentiviral vector system. Lentiviruses are prepared according to the standard protocol. Cells are infected with FAM84B-targeting pseudovirus in the presence of 8μg ml^–1 polybrene followed by 0.5-1.0μg ml^–1 puromycin selection at 24hours after transduction. Knockdown efficiency of the target cells is validated by Western blot. The target sequence used for FAM84B knockdown is “CACCTAAGTTACAAGGAAGTT”.

In vivo ectopic xenograft model

Male SCID mice (6 weeks of age) were obtained from the National Laboratory Animal Center and used for ectopic (subcutaneous) xenograft implantation. Ectopic tumors were established by subcutaneous injection 1×10⁶ of wide type CE81T/VGH cells, CE81T/VGH cells with control shRNA, and CE-81T/VGH cells with FAM84B shRNA into the right hind leg of mice. There were 5 mice in each group. Tumor volumes were measured with a set of calipers and calculated using a standard formula: width×length×depth/2. All experimental procedures using these mice were performed in accordance with protocols approved by the National Taiwan University Institutional Animal Care and Use Committee.

We feel grateful with the kindness from Dr. Chuhsing K Hsiao for providing statistical consultation of data analyses in this project. We thank to Cancer Registry and Medical Information Management Office at National Taiwan University Hospital in providing great assistance for accessing patient survival data. We thank the staff of the Eighth Core Lab, Department of Medical Research, National Taiwan University Hospital for technical support during the study. This work was financially supported by grant NCTRC200730 from the National Center of Excellence for Clinical Trial and Research, Taiwan and grant 102-2314-B-002-155-MY3 from Ministry of Science and Technology, Taiwan.