Thioredoxin system protein expression in carcinomas of the pancreas, bile duct and ampulla.

Background: Pancreatic cancer (PC), including the ampulla and bile duct, is very aggressive, and thus difficult to treat with effective therapies. The current treatment options have failed to improve PC five-year survival rates over the last 30 to 40 years, which remain very low, at ~3%; there is, therefore, an urgent need to identify new targets and treatment modalities (1). Methods: The protein expression of thioredoxin (Trx), thioredoxin reductase (TrxR) and thioredoxin interacting protein (TxNIP) was assessed in two cancer patient cohorts by standard immunohistochemistry using tissue microarrays. The first cohort was composed of 85 pancreatic adenocarcinomas (PAD) and the second of 145 cancers of the bile duct and ampulla. Results: In the PAD cohort, high cytoplasmic TrxR expression significantly associated with lymph node metastasis (P = 0.033). High expression of cytoplasmic (P = 0.018) and nuclear (P = 0.006) Trx were significantly associated with better overall survival, with nuclear Trx expression remaining significantly associated with survival in multivariate Cox-regression (Hazard Ratio (HR) 0.316; 95% Confidence Interval (95% CI) 0.174-0.573; P < 0.0001) when potentially confounding factors were included (gender, age, tumour size, tumour grade, tumour stage, lymph node status, perineural and venous invasion). In cancers of the bile duct and ampulla, high expression of nuclear TrxR and high cytoplasmic TxNIP were associated with patients aged above 60 years (P = 0.024 and P = 0.049 respectively). Associations were also observed between high nuclear TrxR expression and the presence of venous (P = 0.001) and perineural (P = 0.021) invasion. Low cytoplasmic TxNIP expression was also associated with the presence of perineural invasion (P = 0.025). High expression of cytoplasmic TxNIP was significantly associated with better overall survival (P = 0.0002),

United Kingdom, while combination chemotherapy is used more commonly in Europe and the United Kingdom than in the United States (8). Nevertheless, the current treatment modalities for PC have failed to improve the five-year survival rate over the last 30 to 40 years and in the United Kingdom remains very low, at ~3% (1).
For radiation therapy, a number of resistance mechanisms may explain the lack of treatment success, one of these being inherent or acquired expression of redox proteins i.e. proteins that can scavenge certain chemotherapy-and/or radiation-induced radicals, mainly intracellular reactive oxygen species (ROS), making the treatment less effective.
Redox proteins have, therefore, been studied widely for the role they play in regulating therapeutic response of tumour cells, in addition to assessing the association of their expression with clinicopathological criteria and/or patient survival parameters (9).
The Trx system is an important family of redox related proteins that can regulate redox homeostasis and affect the redox state of different signaling molecules, thereby regulating numerous downstream pathways involved in regulation of cell growth, apoptosis, gene transcription, cell cycle progression and oxidative stress (10-12).
The Trx system consists of thioredoxin (Trx), the activating enzyme thioredoxin reductase (TrxR) and the endogenous inhibitor of the system, thioredoxin-interacting protein (TxNIP).
Thioredoxins are a class of low molecular weight redox proteins characterised by a conserved active site (-Cys-Gly-Pro-Cys) found in all Trx family proteins (13,14).
Trx proteins are 12 kDa proteins that include cytosolic thioredoxin-1 (Trx1), mitochondrial thioredoxin-2 (Trx2) and a larger thioredoxin-like protein (p32TrxL). Cytosolic Trx1 is the most widely studied isoform and acts as the major disulphide reductase of proteins in living cells (14). Reduced Trx1, the bioactive form, binds to apoptosis signal-regulating kinase 1 (ASK-1), a key apoptotic regulator whose activation is essential for tumour necrosis factor α (TNF α)-induced apoptosis, and transforms it to inactive form which protects cells against apoptosis (15). The growth and apoptotic regulatory effects of Trx may also be explained by the selective activation of a number of transcription factors such as NF-κB (nuclear factor kappa-light-chain-enhancer of activated B cells), glucocorticoid receptor, TP53, AP1 (activator protein 1) and AP-2 (activator protein 2) (14).
TrxR is a member of the homodimeric pyridine nucleotide-disulphide oxidoreductase family and exists in two forms: cytosol (TrxR1) and mitochondria (TrxR2) (16). TrxR is the only known enzyme that can reduce oxidised Trx by obtaining reducing equivalence from NADPH (17). As an activator of Trx, TrxR therefore plays an important role in the pathways reliant on Trx activity, with its inhibition causing oxidation of Trx leading to activation of p38 and JNK, and downstream apoptosis (18).
TxNIP, also known as vitamin D3 up-regulated protein 1 (VDUP1) or Trx binding protein 2 (TBP2), is a stress-responsive protein that inhibits Trx activity by preventing the recycling of oxidised Trx to the reduced, bioactive, form (16). TXNIP can inhibit the activity of Trx by two pathways. First, oxidised TxNIP 'Cys 247' binds reduced Trx 'Cys 32' and acts as a competitive inhibitor to remove Trx from those proteins whose function are inhibited by separation from Trx, such as ASK-1. Second, overexpression of TxNIP, by factors such as disturbed flow and high glucose, resulting in decreased activity of TrxR, leading to an increase in oxidative stress and apoptosis (19).
Numerous studies have investigated the relationship of redox proteins, including the Trx system, with clinicopathological criteria and patient survival outcomes in different types of cancer, including ovarian, breast, gastro-oesophageal, colorectal and brain cancers (20)(21)(22)(23)(24)(25)(26). However, expression of Trx system proteins has never been previously studied in any robust way, or associated with clinical characteristics, in PC patient tumours. One previous study demonstrated that pancreatic ductal adenocarcinoma (PDAC) tissues were immunohistochemically more positive for Trx expression (24/32 cases) than pancreatic cystadenocarcinoma or normal pancreas tissues but there was no link to/with clinicopathological criteria or patient survival outcomes (27).
The current study set out to determine if Trx system proteins (Trx, TrxR and TxNIP) were expressed in carcinomas of the pancreas, bile duct and ampulla, and if expression correlated with clinicopathological criteria and/or patient survival.

Clinical samples
Assessment of expression of the Trx system was conducted using tissue microarrays

Western blotting
The specificities of Trx, TrxR and TxNIP antibodies were initially assessed by Western blotting using pancreatic cell lysates from PANC-1, MIA PaC-2 and BxPc-3 pancreatic cell lines and breast cancer cell lysates of from breast cancer cell lines, MDA-MB-231 and MCF-7, that were used as positive controls. All cell lines were originally obtained from ATCC with authentication conducted every 4-6 months using Short Tandem Repeat (STR) profiling test. Sub-confluent cells were harvested and resuspended in RIPA buffer (Sigma) supplemented with protease inhibitor cocktail, phosphatase inhibitor cocktail (Thermo) and EDTA solution (Thermo). Lysates were separated by SDS-PAGE and transferred onto a nitrocellulose membrane. After blocking with 5% (w/v) milk powder in 0.1% PBS/Tween, the nitrocellulose membrane was incubated with primary antibody at 4°C overnight. The primary antibodies used in this study were: rabbit anti-human Trx antibody (1:5000 dilution; Ab133524, Abcam), mouse anti-human TrxR antibody (1:1000 dilution; Ab16847, Abcam) and rabbit anti-human TxNIP antibody (1:1000 dilution; Ab188865, Abcam).
Mouse anti human β-actin antibody (1:2000 dilution; Ab8226, Abcam) was used as internal control. Secondary antibody was anti-mouse HRP or anti-rabbit HRP-conjugated antibody (Dako), at room temperature for 1 hour. Membranes were developed with Amersham ECL reagent on hyperfilm (GE Healthcare).

Tissue microarray and immunohistochemistry
Antibody concentrations to be used in IHC were optimised, by IHC as decribed later, using three full-face PC tissue sections from each of the three PC types i.e. 3 pancreatic adenocarcinomas, 3 ductal adenocarcinomas and 3 ampullary adenocarcinomas. The optimization of antibody concentration was carried out using a Novolink Novocastra polymer detection kit (Leica, Denmark), each using three different concentrations, including one as recommended by the manufacturer in their datasheet. Additional concentrations were tried if initial concentrations did not give appropriate immunohistochemical staining. As antibodies of the Trx system were previously optimised in BC tissues (Zhang, 2014), breast tumour composite section were used as as positive and negative controls. Negative controls omitted the primary antibody from the procedure. By comparing staining patterns from each marker with the negative control, asessed by a specialist pathologist, the appropriate concetration of each antibody (anti-Trx, anti-TrxR and anti-TxNIP antibodies) was determined.
TMA construction and immunuhistochemical process has been previously described (29).
Briefly, TMA slides were initially deparaffinised in xylene, followed by rehydration in ethanol and water. Antigen retrieval was performed in 0.01molL -1 sodium citrate buffer (pH=6.0) in a microwave for 10 minutes at 750W and 10 minutes at 450W. Tissue was treated with peroxidase block, washed with Tris-buffered saline (TBS), and then treated with protein block solution. Anti-Trx, anti-TrxR and anti-TxNIP antibodies were diluted 1:2000, 1:100 and 1:250 respectively, and applied to the tissue for one hour at room temperature (for anti-Trx and anti-TxNIP) or overnight at 4°C (for anti-TrxR). Following antibody incubation, tissue was washed with TBS prior to application of post primary solution, washed with TBS followed by application of Novolink polymer solution.
Immunohistochemical reactions were developed using 3,3' diaminobenzidine as the chromogenic substrate and tissue was counterstained with heamotoxylin prior to dehydration in ethanol and fixation in xylene. Breast tumour composite sections, comprised of 6 stage 1 breast tumours of grade 1 to 3, were included as positive controls with each run, with the negative control having primary antibody substituted for PBS.
All cores were assessed semi-quantitatively using an immunohistochemical H-score using a HPF Nikon Eclipse E600 microscope at 200x magnification. Staining intensity was assessed as; none (0), weak (1), medium (2) and strong (3) over the percentage area of each staining intensity. H scores were calculated by multiplying the percentage area by the intensity grade (H score range 0-300). Each core was assessed individually by two individuals, including one specialist histopathologist, blinded to clinical data, and a consensus agreed. An average H-score was generated by taking the mean H-score of the three cores (n=187; n=72 PAD cohort, n=115 bile duct and ampullary cohort) or two cores (n=43; n=13 PAD cohort, n=30 bile duct and ampullary cohort) available for each patient included in this study (n=230).

Statistical analysis
The relationship between categorised protein expression and clinicopathological variables was assessed using Pearson Chi Square (χ 2 ) test of association. Survival curves were plotted according to the Kaplan-Meier method and significance determined using the logrank test. Multivariate survival analysis was performed by using the Cox Proportional Hazards regression model. All differences were deemed statistically significant at the level of P < 0.05. Statistical analysis was performed using SPSS 23.0 software (IBM Corporation). Stratification cut points were determined using X-Tile software (Yale School of Medicine) and were determined prior to statistical analyses (30).

Results
Antibody specificity was determined prior to immunohistochemical staining (supplementary figure 1). Figure 1 shows representative photomicrographs of different staining patterns, i.e. weak, moderate or strong staining of cytoplasmic or nuclear expressision, of Trx system protein expression from PAD TMAs. Supplementary figure 2 shows representative photomicrographs of staining of Trx system protein expression from the carcinomas of the pancreas, bile duct and ampulla TMAs.
In the PAD cohort, cytoplasmic Trx had a median H-score of 210 with values ranging from 100 to 300; nuclear Trx had a median H-score of 225 and ranged from 50 to 300; cytoplasmic TxNIP had a median H-score of 150 and ranged from 0 to 267; cytoplasmic TrxR had a median H-score of 75 and ranged from 0 to 225; nuclear TrxR had a median Hscore of 75 and ranged from 0 to 225. The X-tile cut point for cytoplasmic Trx was 160, nuclear Trx was 234, cytoplasmic TxNIP was 217, cytoplasmic TrxR was 58 and nuclear The correlation between expression levels of the proteins with one another was assessed using the Spearman rank correlation coefficient. In the PAD cohort, cytoplasmic TxNIP expression had a statistically significant, albeit weak, correlation with cytoplasmic TrxR (r = 0.234, P = 0.038) and nuclear TrxR expression (r = 0.241, P = 0.032). In addition, cytoplasmic TrxR expression was strongly correlated nuclear TrxR expression (r = 0.711, P < 0.001). Nuclear expression of Trx was also correlated with cytoplasmic Trx expression (r = 0.549, P < 0.001).

Associations with clinicopathological criteria
Levels of protein expression were assessed in light of clinicopathological criteria in the PAD and the bile duct and ampullary cancer cohorts to determine associations. In the PAD cohort, the only association observed was between high cytoplasmic TrxR expression and lymph node metastasis (χ 2 = 4.533, d.f. = 1, P = 0.033) (Supplementary Tables 1 and 2).

Relationship with clinical outcome
In the PAD cohort, high expression of both cytoplasmic and nuclear Trx were significantly associated with better overall survival (P = 0.018 and P = 0.006 respectively) (  Table   5). In the multivariate Cox-regression for this cohort the potential confounding factors of patient grade, stage, lymph node status, perineural and venous invasion were included and were significantly associated with survival, with individual Kaplan-Meier statistics of P = 0.011, P = 0.004, P = 0.003 P = 0.001 and P = 0.012 respectively.
As Trx and TrxR were expressed in both nucleus and cytoplasm, data were also analysed by grouping patients into combinations based upon expression profiles, i.e. low nuclear staining with low cytoplasmic, low nuclear with high cytoplasmic, high nuclear with low cytoplasmic and high nuclear with high cytoplasmic. In the PAD cohort, no significant correlation was observed in the combination analysis between nuclear and cytoplasmic expression of Trx or TrxR.
Equally, in the bile duct and ampullary carcinoma cohort, no significant correlation was observed from the analysis of combined nuclear and cytoplasmic TrxR expression.
However, low nuclear with high cytoplasmic expression of Trx (n=14) showed longer overall survival than other three subgroups (n=114); either against each separate subgroup (P = 0.017) ( Figure 4A) or when the three subgroups were combined together (P = 0.002) ( Figure 4B).

Discussion
The Trx system regulates the redox state of different signalling molecules and, as a result, can regulate cell growth, apoptosis, gene transcription, cell cycle progression and ability to deal with oxidative stress (10)(11)(12). Several studies have reported associations between Trx system protein expression and clinicopathological criteria and patient survival outcome in different cancer types (20)(21)(22)(23)(24)(25)(26). Nevertheless, the prognostic significance of expression of the Trx system has never been previously assessed in PC patient tumours.
One previous study demonstrated that PDAC tissues were immunohistochemically more positive for Trx expression (24/32 cases) than pancreatic cystadenocarcinoma or normal pancreas tissues (27), suggesting a possible association of Trx expression with malignant potential of PDAC.
Current findings show that high cytoplasmic TrxR expression was associated with lymph node metastasis (P = 0.049). A previous study, albeit in a different cancer type (50 patients with oral squamous cell carcinoma), described an association between low TrxR expression and lymph node metastasis (P = 0.027) (31). It may well be, therefore, that TrxR plays a role in regulating lymph node metastasis. TrxR expression was also of interest in the bile duct and ampulla cancer cohort, with current data showing that high nuclear TrxR expression was associated with the presence of venous (P = 0.001) and perineural invasion (P = 0.021), again suggesting that in cancers of the pancreas, bile duct and ampulla, TrxR may be involved in tumour invasion.
Low cytoplasmic TxNIP expression was also significantly associated with the presence of perineural invasion (P = 0.025) in the bile duct and ampulla cancer cohort. In agreement with such findings, a previous study also reported an association between high TxNIP and absence of perineural invasion (P = 0.030) in 140 gastro-oesophageal adenocarcinoma patients (23).
In the PAD patients, current findings demonstrate that high expression of cytoplasmic Trx (P = 0.018) and nuclear Trx (P = 0.06) is significantly associated with better overall survival, with nuclear Trx expression remaining significantly associated with survival in multivariate Cox-regression analysis (P < 0.0001). Such a finding, in light of expression in other tumour types, is somewhat unexpected. Raffle and colleagues observed that high expression of Trx was significantly associated with poor overall survival (P = 0.004) in 12 colorectal cancer patients (32). In 154 ovarian cancer patients, low cytoplasmic expression of Trx was significantly associated better progression-free survival (P = 0.032), whereas Trx nuclear expression did not (P = 0.455) (20). And in 65 gastric cancer patients high expression of Trx was significantly associated with poor recurrence-free (P = 0.008) and overall survival (P = 0.015) (33). However, a study in 174 Hodgkin lymphoma patients produced data similar to current findings, showing that high cytoplasmic Trx expression associated with better failure-free survival (P = 0.049) and which remained significant in multivariate Cox-regression analysis (P = 0.023) (34). The relative importance of expression may, therefore, vary from tumour type to tumour type with no overall generalisations possible. Current data also show no relationship between cytoplasmic or nuclear TrxR expression with overall survival, unlike a study in 98 locally advanced breast cancer that showed that high expression of TrxR (P = 0.021) and TxNIP (P = 0.021) were significantly associated with better distant metastasis-free survival (22).
In the bile duct and ampulla carcinoma cohort, current data reveal a strong association between high cytoplasmic TxNIP expression and better overall survival, which remained significant in multivariate Cox-regression analysis. Such data are comparable with the previous breast cancer study mentioned above, that showed high expression of TxNIP was significantly associated with distant metastasis-free survival (P = 0.021) and overall survival (P = 0.037) (22). Lim and colleagues also observed a significant correlation between high expression of TxNIP and longer relapse-free survival (P = 0.036) in 65 gastric cancer patients (33). Furthermore, high expression of TxNIP was significantly associated with a better disease specific survival (P = 0.016) in 66 gastro-oesophageal adenocarcinomas patients (23), with a recent study also showing that high-grade meningioma patients with strong TXNIP expression (n=37) had longer recurrence-free time (P = 0.02) than those with weak expression of TxNIP (n=28) (25).
Althought carcinomas of the pancreas, bile duct and ampulla have overlapping symptoms and a common treatment, such differences between the results of two cohorts may be due to the significant differences in their survival, and accumulating evidence displaying differences in their biology, including their molecular profiles (35).
A further important aspect to note is that this study describes the protein expression of Trx system family members and not their relative activity levels. Determining the activity of Trx system family members may be possible as part of future in vitro studies, using Trx/TrxR activity assays such as the insulin reduction assay (Kunkel et al., 1997).
Further analysis, in the bile duct and ampullary carcinoma cohort, showed that low nuclear with high cytoplasmic expression of Trx showed longer overall survival than other three subgroups; either against each subgroup (P = 0.017) or when the three subgroups were combined (P = 0.002). In contrast to the current findings, Woolston and colleagues investigated the associations between Trx expression and ovarian cancer patients' survival, and observed that high nuclear with low cytoplasmic expression of Trx demonstrated a significantly better overall survival either against each separate subgroup (P = 0.04) or when the three subgroups were combined together (P = 0.004). Furthermore, high nuclear with low cytoplasmic expression of Trx were associated with better progression-free survival either against each separate subgroup (P = 0.025) or when the three subgroups were combined together (P = 0.003) (20).
A number of studies have examined the differences in the subcellular localization of Trx protein in the mammalian cell, suggesting that Trx protein was in reduced state within the nucleus more than in the cytoplasm, and was imported from the cytoplasm during oxidative stress, both in normal and malignant cancer cells (36)(37)(38)(39)(40). Furthermore, a previous study has demonstrated that nuclear Trx protein offers better protection against oxidative stress than cytoplasmic or mitochondrial Trx protein in human colonic epithelial cells (41). Taken together with previous studies, current data suggest that the Trx system plays a role in the activation of nuclear transcription factors that potentially govern cellular responses to oxidative stress.
In summary, this study demonstrates that high nuclear and high cytoplasmic expression of Trx is associated with better overall survival in PAD patients, and that high cytoplasmic TxNIP expression is associated with better survival in bile duct and ampullary cancer patients, with each being potentially important independent prognostic factors.

Ethics approval and consent to participate
Ethical approval was obtained from the National Regional Ethics Service Committee East Midlands -Nottingham 1 for the use of anonymised archival specimens and the requirement for patient/relative consent was waived by the Ethics Committee.

Availability of data and material
Additional data has been submitted as supplementary information with original H-scores and images being held in a secure database. Such information can be made available via contacting the corresponding author.

Competing interests:
The authors declare that they have no conflicts of interest.

Funding
The authors wish to thank Pancreatic Cancer UK for funding this study via a Research

Supplementary Files
This is a list of supplementary files associated with the primary manuscript. Click to download.