Expression of TXNIP is associated with angiogenesis and postoperative relapse of conventional renal cell carcinoma

One of the common mediator of tumour progression is the oxidative stress induced by inflammatory tumour microenvironment (TME). Activated fibroblasts, local and immune cells produce reactive oxygen species (ROS) supporting tumour cell proliferation and pave the way for metastatic tumour growth. TXNIP regulates ROS generation by inhibiting the antioxidative function of thioredoxin (TXN). The shift of TXNIP/TXN balance towards overexpression of TXNIP is associated with proliferation of endothelial cells during tumor angiogenesis. The oxidative stress activates the hypoxia inducible factor-1 (HIF-1), which plays an important role in the biology of conventional RCC (cRCC). Under oxydative stress TXNIP interacts with NLRP3 inflammasome leading to maturation and secretion of inflammatory cytokine IL1β. To establish the role of TXNIP and downstream genes HIF1α and IL1β in the biology of cRCC, we have applied immunohistochemistry to multi-tissue arrays containing tumours of 691 patients without detectable metastases at the time of operation. We found that cRCC displaying a fine organised capillary network with nuclear translocation of TXNIP and expressing IL1β have a good prognosis. In contrary, we showed a significant correlation between cytoplasmic TXNIP expression, inefficient vascularisation by unorganized and tortuous vessels causing tumour cell necrosis and postoperative tumour relapse of cRCC.

Controversial results have been published on TXNIP expression and tumour progression. High level of TXNIP expression was associated with a significantly shorter survival of patients with non-small cell lung cancer and invasive growth of hepatocellular carcinoma 22,23 . On the other hands, lack or reduced expression of TXNIP in cancer cell lines, experimental mouse models as well as in tumour tissues suggested that TXNIP is a tumour suppressor gene [24][25][26][27][28][29][30] . It has also been reported that decreased TXNIP RNA expression is associated with poor prognosis of patients with clear cell renal cell carcinoma 31 . To establish the role of TXNIP and downstream genes HIF1α and IL1β in the biology of cRCC, we have applied immunohistochemistry to tissue multi-arrays containing tumours of patients without detectable metastases at the time of operation.

Results
Expression of TXNIP in normal kidney. TXNIP showed a weak expression at the luminal surface of proximal tubular cells and a strong cytoplasmic expression in connecting and collecting duct cells in normal adult kidneys, whereas the loop of Henle and distal convoluted tubules were negative (data not shown). Vascular smooth muscle cells and endothelial cells of small kidney arteria displayed a weak TXNIP expression. We did not find nuclear TXNIP expression in normal adult kidney. No expression of HIF1α and IL1b has been seen in normal kidney tissue.
Expression of TXNIP in conventional RCC . Results of immunohistochemistry of 691 cRCC separated two groups of tumours. One group of 512 tumours (74%) without cytoplasmic expression of TXNIP (Fig. 1A), and the other group displaying medium or strong cytoplasmic TXNIP staining in 95 (14%) and 84 (12%) cases, respectively. (Fig. 1B). As the first Kaplan-Meier analysis revealed that patients with medium or strong cytoplasmic TXNIP expression have similar disease-free survival, we have evaluated the results of cytoplasmic TXNIP immunohistochemistry in correlation to clinical and pathological parameters as negative or positive.
Expression of IL1β and HIF1α in conventional RCC . We used the same series of TMA as for TXNIP immunohistochemistry. A granular cytoplasmic IL1β expression has been seen in 601 of the 691 conventional RCCs (Fig. 1C), whereas 90 cases showed a negative immunreaction with IL1β antibody (Fig. 1D). We did not find HIF1α expression in any of the tumour cells analysed in this study. Positive HIF1α staining has been detected in some endothelial cells and in tumour infiltrating immune cells.
TXNIP expression in tumour microvessels. Irrespectively of positive or negative cytoplasmic expression in tumour cells TXNIP showed a strong expression in endothelial cells of the tumour stroma. In highly differentiated cRCC resembling the classic "Grawitz" tumour a fine network of TXNIP positive endothel cells and tumour cells without TXNIP staining was seen ( Fig. 2A). The overwhelming majority of endothelial cells displayed a nuclear TXNIP expression (Fig. 2B). In the second group of tumours with medium or high TXNIP expression in tumour cells only few vessels has been recognized. The majority of TXNIP positive tumours displayed unorganized, irregular or tortuous tumour vessels (Fig. 2C) as compared to those seen in well differentiated cRCC (Figs. 1A and 2A). Around small necrotic tumour areas tumour cells displayed strong cytoplasmic TXNIP immunoreaction (Fig. 2D). As vascular TXNIP expression has been detected in all tumours analysed in this cohort, we evaluated only the cytoplasmic TXNIP expression in relation of tumour the progression.
Correlation analysis. The TXNIP expression in tumour cells was significantly correlated with tumour size, grade and T-classification and tumour necrosis as well as postoperative progression of cRCCs (Table 1, all p < 0.001). Kaplan-Meier analysis revealed that patients having a cRCC with cytoplasmic expression of TXNIP protein have a significantly shorter disease-free survival compared to those without TXNIP expression (Fig. 3A). Univariate Cox regression analysis revealed that tumor size, grade, T classification, necrosis as well as cytoplasmic TXNIP positivity were significantly associated with postoperative tumour progression (all p < 0.001). However, in multivariate Cox regression analysis only cytoplasmic TXNIP expression and necrosis remained independent predictor of cancer progression indicating two times higher risk of postoperative tumour relapse ( Table 2). Kaplan-Meier analysis showed that patients having a cRCC without cytoplasmic expression of IL1β protein have a shorter disease-free survival compared to those with cytoplasmic IL1β expression (Fig. 3B). In multivariate cox regression analysis no correlation between IL1β expression and postoperative relapse has been found. IL1β protein level showed a significant correlation only with histological grade (p < 0.05).

Discussion
We analysed the expression of TXNIP protein in a large cohort of cRCC without detectable metastasis at the time of operation. The Kaplan-Meier survival analysis indicated that patients with tumours dysplaying cytoplasmic TXNIP protein expression have significantly shorter tumour free survival (p < 0.001). Multivariate analysis revealed that these patients have a more than two times higher risk to develop a metastatic disease during the median follow-up of 73 months (p = 0.001).
The only report on TXNIP and renal cancer evaluated RNA expression data deposited in Cancer Genome Atlas (TCGA) and concluded that decreased expression of TXNIP predicts a poor prognosis 31 . Reduced expression or lack of expression of TXNIP was associated with the development of oxidative stress induced experimental RCC in rat 29 . Decrease of TXNIP RNA in bladder cancer and development of bladder cancer in TXNIP-KO mice has been described 27 25 . In experimental mouse model the silencing of TXNIP increased the predisposition to hepatocellular carcinoma (HCC) 30 . In two studies expression of TXNIP was detected in all of primary human HCC, but the expression was decreased in 66% and 50% of the cases, respectively 28,32 . Based on these data TXNIP was suggested to be a tumour suppressor gene. However, in line with our results a significantly increased expression of TXNIP and elevated ROS level was associated with invasive growth in human hepatocellular carcinoma and high expression of TXNIP was an independent prognostic factor in non-small lung cancer 22,23 . TXNIP is a member of α-arrestin protein family and by directly binding to the antioxidant TXN blocks its reducing potential 13 . TRXs are small redox active proteins that play an important role to maintain the cellular redox balance under normal conditions. Elevated TXNIP expression and oxidative stress influence several biological functions, regulates cell growth, differentiation and apoptosis 33,34 . By supporting the ROS production TXNIP promotes endothelial cells proliferation and angiogenesis by activating the transcription factor NF-kB and regulating the vascular endothelial growth factor (VEGF), and vascular endothelial growth factor receptor 2 (VEGFR2) signalling 18,35,36 . The importance of TXNIP in the maintenance of endothelial homeostasis was recently demonstrated by a TXNIP-KO mouse experiment 37 . ROS are also involved in the induction of HIF family transcription factors the major signalling components downstream of hypoxia 38 . Many of the HIF target proteins are also involved in the angiogenesis. In our study no HIF1α expression has been observed the tumours cells. We showed in this study that slowly growing, differentiated cRCC with excellent clinical outcome displays an organized vascular network consisting of TXNIP positive endothelial cells. The translocation of TXNIP into the nuclei of endothelial cells of tumour supporting vessels indicates an increased level of ROS. It was shown previously that forced expression of TXNIP in isolated microvascular endothelial cells results in its nuclear translocation and activates the NF-kB pathway leading to expression of pro-inflammatory cytokines including IL1β 20,39,40 . Based on our results that IL1β expressed preferentially in a group of tumours showing strong TXNIP nuclear expression in supporting endothelial cells indicates that the IL1β expression is mediated through the NFkB pathway, rather than direct activation of NLRP3 inflammasome. Recently, it was shown that IL1β mediates epithelial to mesenchymal transition of proximal tubular cells of kidney 21 . Conventional RCC derives from proximal tubular cells of the kidney. However, we found IL1β expression in the group of cRCC, with excellent outcome of disease and not in the group of tumours with increased cell motility leading to high risk of postoperative metastatic growth.
We showed in this study that rapid growth of tumour cells in highly malignant cRCC with metastatic capacity exceeded the proliferation of oxygen supplying capillaries. The tumour cell/blood vessel relation is strongly shifted in favour of tumour cells and lead to inefficient vascularisation by unorganized and tortuous vessels. Elevated ROS level can impair the function of important proteins and interfere with metabolic pathways. In hypoxemic microenvironment the ATP is mostly produced via high rate of anaerobic glycolysis and the Warburg effect leads to lactic acid fermentation and further increases oxidative stress 41,42 . The hypoxemic and acidic TME may dramatically change the oxidative-reductive balance by producing high level of ROS which in increased concentration may inhibit tumour growth [43][44][45] . The high level of oxidative stress may cause serious damage of   www.nature.com/scientificreports/ tumour DNA and proteins leading to death of cancer cells 46 . We showed in this study that the highest expression of TXNIP protein occurred in tumour areas with small circumscribed tumour cell necrosis. In multivariate analysis, in addition to TXNIP positivity, tumour necrosis was presented as independent prognostic factor of postoperative tumour relapse (p = 0.004).
In conclusion, our current study demonstrates that cRCC displaying a fine organised capillary network with nuclear translocation of TXNIP and expressing IL1β have a good prognosis. In contrary, we showed a significant correlation between cytoplasmic TXNIP expression, inefficient vascularisation by unorganized and tortuous vessels causing tumour cell necrosis and postoperative tumour relapse of cRCC.

Material and methods
Patients. We have analysed a cohort of cRCCs obtained from 691 patients operated between 2000 and 2013 without clinically detectable metastasis at the first observation described previously 47 . Data on regular follow-up and tumour specific death was obtained from Tumour Registry of the Department of Urology in accordance with the relevant institutional guidelines and regulations. Follow-up was defined as a time from the operation until the last recorded control or cancer specific death. The pertinent clinical and pathological data in are presented in Table 1. Of the 691 patients, 406 (59%) were males and 285 (41%) females, the mean age of the cohort was 61.3 ± 11.2 years (range 23-88 years). The average tumour size was 50.2 ± 25.8 mm. During the median follow-up of 73 ± 28 months, tumour relapse was observed in 112 patients (16%). Of 691 tumours, 511 (74%) were classified as pT1 including 308 (45%) pT1a tumour. The overwhelming majority of cRCCs (456 of 691) displayed G1 tumour grade. Regarding to the tumour stage, 671 (97%) of tumours were designed to stage I and II.
Tumour samples and tissue microarray (TMA). The histological diagnosis and TNM classification was established by a genitourinary pathologist (GK) according to the Heidelberg and TNM classification and by applying 3 trier grading 48,49 . The collection and use of all tissue samples for this study was approved by the Ethics Committee of the University Pecs, Hungary (No. 5343/2014). All participants sign an informed consent that after establishing the histological diagnosis the rest of formalin fixed and paraffin embedded material can be used for immunohistochemistry. We have identified representative tumour areas on haematoxylin and eosin stained slides and the corresponding paraffin blocks were used for TMA construction. Three to five core biopsies with a diameter of 0.6 mm were taken from each tumour and were placed in the recipient block using a Manual Tissue Arrayer (MTA1, Beecher Instruments, Inc., Sun Prairie, USA). We have also analysed foetal and adult kidneys for establishing the TXNIP, HIF1α and IL1β expression in normal renal tissue.
Immunohistochemistry. After removing the paraffin and rehydration the 4 μm thick sections were subjected to heat-induced epitope retrieval in citrate buffer, pH 6.0 for IL1β and EnVision FLEX Target Retrieval at the dilution of 1:600. EnVision FLEX horse-radish-peroxydase conjugated secondary antibody (DAKO) was applied for 30 min at room temperature and colour was developed using the DAB substrate (DAKO). Tissue sections were counterstained with Mayer's haematoxylin (Lillie's modification, DAKO) and after 10 s bluing were mounted with Pertex. For negative control the primary antibody was omitted and for positive control we used the normal kidney biopsies included into the TMAs. We have scored the TXNIP staining as negative, medium and high expression. If at least one of the 3 or more core biopsies was positive, we evaluated the tumour as positive.

Statistical analysis.
Statistical analysis was carried out as described earlier 47 . Correlations between categorical variables were estimated with Fisher's exact test. Estimates of the cumulative survival distributions were calculated by the Kaplan-Meier method, and the differences between the groups were compared using the logrank test. The significance of clinical-pathological variables was evaluated using the univariate and multivariate Cox proportional hazard regression model. Analysis was performed using IBM SPSS Statistics v.27 for Windows (Inc. Chicago IL, USA). p-value < 0.05 was considered the limit of statistical significance.

Data availability
The datasets are available from the corresponding authors on reasonable request.