Dysregulation of tristetraprolin and human antigen R promotes gastric cancer progressions partly by upregulation of the high-mobility group box 1

Aberrant expression of ARE-binding proteins (ARE-BPs) plays an important role in several diseases, including cancer. Both tristetraprolin (TTP) and human antigen R (HuR) are important ARE-BPs and always play opposite roles in regulating target mRNAs. Our previous work has demonstrated that TTP expression is decreased in gastric cancer (GC). In this study, we reported that HuR was elevated in GC cell lines and gastric cancer patients and that decreased TTP expression partly contributed to the elevated HuR levels by regulating its mRNA turnover. We also observed that dysregulation of TTP and HuR elevated the high-mobility group box 1 (HMGB1) expression in different ways. HuR promoted HMGB1 expression at translational level, while TTP regulated HMGB1 mRNA turnover by destabilizing its mRNA. Increased HuR promoted cancer cell proliferation and the metastasis potential partly by HMGB1. Using immunohistochemistry, we observed that both positive cytoplasmic and high-expression of nuclear HuR were associated with poor pathologic features and survival of GC patients. In conclusion, this study demonstrated that dysregulation of the TTP and HuR plays an important role in GC. Moreover, high HuR nuclear expression or aberrant cytoplasmic distribution may serve as a predictor of poor survival.


Results
The mRNA and protein expression of HuR is increased in GC cell lines and GC patients. To understand whether HuR is highly expressed in GC, expression of the HuR mRNA and protein were examined in four GC cell lines (MGC-803, BGC-823, SGC-7901,  and in non-malignant GES-1 cells by quantitative real-time PCR (qRT-PCR) and western blotting. We observed that HuR mRNA expression was evidently higher in the MGC-803, SGC-7901 and MKN-45 cell lines than GES-1. However, there was no significant difference in expression between BGC-823 and GES-1 cells (Fig. 1a). Because HuR is an important nuclear/cytoplasmic shuttling protein, we extracted the cytoplasmic and nuclear HuR proteins and examined them by western blotting. Positive cytoplasmic HuR protein expression was detected in MGC-803, SGC-7901, MKN-45, and BGC-823 cells, whereas no cytoplasmic expression of HuR was found in GES-1 cells (Fig. 1b). These results demonstrated that the mRNA and protein expression of HuR were evidently higher in GC cell lines.

Tristetraprolin downregulates HuR expression in GC cells. TTP and HuR are two important
cancer-related ARE-BPs, and researchers have also demonstrated that TTP can downregulate the mRNA of HuR 12 . Previously we found that TTP is down-regulated in GC. To investigate whether TTP regulates the expression of HuR and whether a TTP deficiency contributes to the overexpression of HuR in GC, we transfected TTP expression plasmids (pcDNA-TTP) and the empty pcDNA3.1 control vectors into MGC-803 cells (relatively low expresser of TTP) and then examined HuR expression by qRT-PCR and western blotting. We observed that the HuR mRNA and protein expression were evidently decreased in MGC-803/TTP cells (Fig. 2a,c). Another report demonstrated that silencing of HuR induced an increase in TTP in human pulmonary microvascular endothelial cells 22 . We transfected pCMV6-HuR plasmids or pCMV6-Entry vectors into BGC-823 cells (relatively low expresser of HuR) and measured TTP expression. However, no changes in TTP mRNA or protein expression were detected (data not shown).
We further knocked down TTP expression in SGC-7901 cells (relatively high expresser of TTP) using TTP-siRNAs and examined the mRNA and protein expression of HuR, and the results confirmed the regulatory effect of TTP on HuR (Fig. 2b,c). Because TTP downregulates HuR mRNA levels, we analyzed the mRNA expression of HuR and TTP in 70 GC patients. Although the results of the Chi-squared test demonstrated no significant  The data were represented as the mean ± SD, r 2 : goodness of fit, **p < 0.01. (f) Luciferase assay in 293 T cells revealed that AREs within the HMGB1 mRNA 3′UTR mediated the inhibitory of TTP. And the schematic showed the potential binding sites of TTP within the HMGB1 mRNA 3′ UTR. Results shown on the graph represent the mean ± S.D. of three independent experiments, **p < 0.01. difference between TTP and HuR mRNA expression, we observed that 48.6% (34/70) of these GC patients displayed a pattern of high expression of HuR mRNA accompanied with decreased TTP mRNA levels (Table 1).
HuR and TTP regulate the expression of HMGB1 in different ways. Overexpression of HMGB1 in cancer cells and elevated serum levels in GC have been demonstrated in several studies 17,23,24 . Elevated HMGB1 expression is associated with poor prognosis in patients. However, the mechanisms by which HMGB1 is overexpressed in cancer were not clear 20 . It was reported that in the progression of myogenesis, HuR increased the expression of HMGB1 by suppressing the translational inhibition mediated by miR-1192 25 . To explore whether elevated HuR contributes to overexpression of HMGB1 in GC, we analyzed the mRNA and protein expression of HMGB1 after HuR had been overexpressed or knocked down in BGC-823 and MGC-803 cells. Expression of HMGB1 protein but not mRNA (data not shown) was affected by the changes in HuR expression, which indicated that elevated HuR expression in GC was a contributor to HMGB1 overexpression (Fig. 2d).
Next, we analyzed the 3′UTR of the HMGB1 mRNA and observed that it contained 14 potential binding sites (AUUUA) for TTP. Therefore, we explored whether TTP regulates HMGB1 expression. We observed that increased TTP significantly decreased the expression of HMGB1 at both the mRNA and protein levels, while knocking down TTP increased HMGB1 expression (Fig. 2a-c). To determine whether the decreased HMGB1 mRNA was the result of the regulation of its mRNA stability by TTP, we examined the half-life of HMGB1 mRNA in actinomycin D-treated MGC-803 cells transfected with TTP plasmids or the empty vectors. After 6 hours of actinomycin D treatment, approximately 33.2% of HMGB1 mRNA remained in MGC-803/TTP cells, whereas approximately 79.6% of HMGB1 mRNA was retained in the control cells (Fig. 2e). To further explore whether TTP could interact with 3′ UTR of HMGB1 mRNA directly, luciferase assay was performed using the wild and mutant types of HMGB1 ARE. The results of luciferase assay verified that TTP could regulate HMGB1 by binding to its ARE directly (Fig. 2f). Collectively, these results suggested that TTP destabilizes the mRNA of HMGB1 by binding to its ARE in GC cells. Next, we also detected the HMGB1 mRNA levels in the 70 paired cancer samples and analyzed their correlation with TTP mRNA. However, no significant relation was observed between HMGB1 and TTP expression at the mRNA level (Table 1). These results indicated that elevated HMGB1 in GC was partly due to the dysregulation of TTP and HuR, and that TTP regulated the expression of HMGB1 at the mRNA level by binding to its ARE, while elevated HuR promoted HMGB1 expression at translational level.
Elevated HuR promotes GC cell proliferation partly mediated by HMGB1 in vitro. HuR plays an important role in cancer cell proliferation by increasing the levels of proliferation related regulators (including cyclins, epidermal growth factor, c-Myc, eukaryotic translation initiation factor and others) and regulating several anti-apoptotic factors and signaling pathways 3 . To further explore whether elevated HuR expression influences the proliferation of GC cells, we transfected HuR plasmids and the control empty vectors into BGC-823 cells. Using the Cell Counting Kit-8 (CCK-8) assay, we observed that overexpression of HuR led to a significant increase in the proliferation of cancer cells (Fig. 3a). Next, we knocked down the expression of HuR in MGC-803 cells using siRNA-HuR and siRNA-control and analyzed by CCK-8. The results revealed that decreased HuR expression suppressed GC cell proliferation (Fig. 3b).
Because HMGB1 acts as a promoter of cell proliferation, we tested whether HuR-mediated acceleration of cell proliferation was correlated with extracellular HMGB1 levels. We treated BGC-823/HuR and MGC-803/ siRNA-control cells with human recombinant HMGB1 (rp-HMGB1) (100 ng/ml) and analyzed them using the CCK-8 assay. The results revealed that the proliferation of GC cells treated with rp-HMGB1 increased (Fig. 3a,b). Furthermore, we cotransfected HuR plasmids and siRNA-HMGB1 (or siRNA-control) into BGC-823 cells and analyzed by CCK-8. We observed that knockdown of HMGB1 inhibited the proliferation induced by HuR overexpression (Fig. 3c). This evidence suggested that elevated HuR in GC cells promoted cell proliferation and was partly mediated by HMGB1.

Overexpression of HuR contributes to GC cell invasion and migration in vitro.
HuR plays an important role in cancer invasion and metastasis 8,26 . Studies have revealed that HuR stabilizes several metastasis-related factors, such as matrix metalloproteinase 9 (MMP9), and epithelial-mesenchymal transition (EMT)-promoting factors, such as Snail and ZEB2. To investigate whether HuR overexpression promotes GC metastasis, we analyzed the influence of HuR expression changes on the invasion and migration of GC cells using transwell assays. We observed that knockdown of HuR in MGC-803 cells impaired invasion and migration (Fig. 3d). Meanwhile upregulation of HuR promoted invasion and migration in BGC-823/HuR cancer cells compared to BGC-823/pCMV6 cells (Fig. 3e). Studies also demonstrated that HMGB1 plays an important role in cancer metastasis. Therefore, we tested the influence of exogenous rp-HMGB1 on migration and invasion of BGC-823/HuR cells and MGC-803/si-NC. The results indicated that elevated extracellular HMGB1 level helped to increase migration and invasion ability of cells (Fig. 3d,e).

Overexpression of HuR promotes GC cell growth in vivo. To investigate whether overexpression of
HuR promotes GC growth in vivo, we established xenograft tumor models in nude mice. We injected equal numbers of stable HuR-overexpressing BGC-823/HuR cells and BGC-823/vector cells into the subcutaneous tissue of the armpits of nude mice. The size of the xenograft tumors was monitored once per week. Following 5 weeks of growth, the tumors were stripped and measured (Fig. 4a,b). We observed that the average size and weight of the xenograft tumors in models implanted with HuR-overexpressing BGC-823/HuR cells (556.2 ± 45.34 mm 3 ; 0.571 ± 0.096 g, respectively) were much higher than those observed in controls (182.0 ± 43.05 mm 3 , p < 0.01; 0.166 ± 0.034 g, p < 0.01, respectively) after 5 weeks of injection (Fig. 4c,d). Elevated HuR and HMGB1 expression were verified by western boltting (Fig. 4e). These results indicated that HuR overexpression induced significant promotion of GC cell growth in vivo.

The relationship between HuR and clinical features of GC patients.
To further explore the clinical role of HuR in GC, we measured the expression of HuR in 251 cases of GC specimens by immunohistological staining (representative images shown in Fig. 5) and analyzed the correlation between the HuR expression levels and clinicopathological features of GC. Several studies have demonstrated that aberrant cytoplasmic/nuclear distribution of HuR plays an important role in various cancers. Therefore, we analyzed the relationships between cytoplasmic or nucleus HuR expression and pathological features; the results are as shown in Tables 2 and 3. We observed that positive cytoplasmic expression of HuR was closely related to the depth of invasion (p = 0.006), lymph node metastasis (p = 0.019), distant metastasis (p < 0.001) and TNM stage (p < 0.001) ( Table 2). Using the median IHS score of 6 as a division point, these samples were divided into the high-nuclear HuR group (IHS > 6, n = 166) and low-nuclear HuR group (IHS ≤ 6, n = 85). As summarized in Table 3, statistical analysis demonstrated that high-expression levels of nuclear HuR were correlated with the depth of invasion (p = 0.040), TNM stage (p = 0.009), tumor size (p = 0.019) and TTP expression (p = 0.017), but were not correlated with sex, age, distant metastasis, lymph node metastasis, differentiation status or aberrant cytoplasmic HuR expression (Table 3).
Expression level of HuR is correlated with overall survival of GC patients. Lastly, to further elucidate whether elevated HuR expression could be used as a prognosis factor in GC, we analyzed the correlation between the overall 5-year survival and HuR expression in 190 cases of GC who received the operations between July 2007 and December 2011 using Kaplan-Meier analysis. As shown in Fig. 6, the results demonstrated that both positive cytoplasmic HuR and high-expression of nuclear HuR always predicted poor survival. The mean survival time in the positive cytoplasmic HuR group was 44.58 months (95% CI: 38.54-50.62), while for the negative cytoplasmic HuR group, it was 68.10 months (95% CI: 63.74-72.46, p < 0.001) (Fig. 6a). Furthermore, the mean survival time of patients in the high nuclear HuR expression group was 56.18 months (95% CI: 51.32-61.05), and was evidently shorter than that of the low nuclear HuR expression group (mean survival time: 71.37 months, 95% CI: 65.81-76.93, p = 0.002) (Fig. 6b).
In a previous study, we demonstrated that a reduced TTP expression level always predicted poor prognosis in GC patients 16 . As HuR was an important downstream target of TTP, we further analyzed whether the combination of the two factors can be used as a prognostic predictor for the overall 5-year survival of GC patients. As demonstrated, patients with a high nuclear-HuR expression combined with a low-TTP expression pattern had the poorest survival rates (n = 82, 5-year survival rate: 40.2%, mean survival time: 51.951 months, 95% CI: 45.732-58.171), whereas patients with a low nuclear-HuR expression combined with high-TTP expression pattern had the best survival rates (n = 32, 5-year survival rate: 71.9%, mean survival time: 73.625 months, 95% CI: 66.412-80.838) (Fig. 6c). Moreover, we analyzed the survival of patients with different cytoplasmic HuR and TTP expression models. Our results demonstrated that patients with a positive cytoplasmic HuR expression always had poor survival. Furthermore, patients with negative cytoplasmic HuR expression and relatively high TTP levels had the best survival (n = 56, 5-year survival rate: 76.8%, mean survival time: 77.107 months, 95% CI: 72.665-81.549) (Fig. 6d). To further evaluate whether elevated HuR expression represented a prognostic parameter, Cox's analysis was applied. Univariate analysis revealed that the depth of invasion, lymph node metastasis, distant metastasis, TNM stage, tumor size, differentiation status, TTP expression, and nuclear and cytoplasmic HuR expression were associated with poor survival rates. Furthermore, Cox's multivariate analysis revealed that distant metastasis, TNM stage, and both the positive cytoplasmic HuR and relatively high expression of nuclear HuR were independent prognostic factors (Table 4).

Discussion
Post-transcriptional regulation plays an important role in orchestrating the fates of RNAs. The progression of mRNAs through splicing, maturation, transportation, localization, and degradation is tightly controlled by post-transcriptional regulatory factors, including miRNAs, long non-coding RNAs, RNA-binding proteins and other factors 27 . Transcripts of a lot of cytokines, chemokines, growth factors, and so on are highly enriched of AREs, which interact with ARE-BPs to regulate the stability and translation of these mRNAs 28 27 .

. Mutations in binding regions, dysregulation of ARE-BPs expression, aberrant interplay networks with miRNAs and other factors may alter the interaction between ARE-BPs and ARE-containing transcripts and result in elevated expressions of tumor-promoting targets
Aberrant expression and the function of ARE-BPs have been reported in several types of cancers. Both TTP and HuR are two closely related ARE-BPs and are always dysregulated in cancers 3 . TTP is an mRNA-destabilizing factor and acts as a tumor suppressor, while HuR stabilizes target mRNAs and helps promote the translation of some mRNAs. Expression of TTP is decreased, while HuR is overexpressed or aberrantly distributed in the cytoplasm in several types of cancers. Interestingly, both TTP and HuR are highly subject to extensive posttranslational modifications, especially phosphorylation. Various protein kinase, for example p38 MAPK, ERK/MAPK, and MK2 are associated with their phosphorylation 5 . Phosphorylated TTP interacts with the multifunctional 14-3-3 adaptor proteins to form complexes, which protect TTP proteins from degradation and inhibit TTP mediated target mRNA decay 29 . As for HuR, modifications including phosphorylation and methylation at different residues affect its subcellular localization and interaction with target mRNAs 14 . Moreover, TTP downregulates HuR mRNA levels 11 . According to a recent study, thousands of overlapping binding sites of TTP and HuR were found in more than one thousand genes, and most of the TTP sites in the 3′UTRs overlapped with HuR binding sites 13  bind to AU rich elements and its optimum binding sequence is the AUUUA motif flanked by additional uridylate residues, whereas HuR can bind to the AU-, CU-and U-rich elements of target mRNAs. Dysregulation of the TTP-HuR axis may increase a series of pro-tumorigenesis factors that are associated with the development and progression of several of cancers, which we have previously reviewed 3 . An aberrant TTP-HuR axis is associated with proliferation, apoptosis, angiogenesis, metastasis, invasion, and resistance to chemotherapy. For example, elevated HuR in GC facilitated DNA synthesis, the G 1 to S phase transition of cell cycle and acted as an anti-apoptotic factor 30 . And HuR interacts with the 3′ UTR of matrix metalloproteinase 9 (MMP-9), urokinase plasminogen activator (uPA) and its cell surface receptor uPAR to increase cancer invasion ability. TTP is a regulator of MMP-1, -2, -9, uPA/uPAR and other factors which play important roles in tumor metastasis 3,11 . Our previous work has demonstrated that TTP is downregulated in GC. Decreased expression of TTP is closely related to the depth of invasion, TNM stage, lymph node metastasis and survival in GC patients. Restoring TTP expression in GC cells inhibits the proliferation, invasion and migration of cancer cells 16 . In this study, elevated expression of HuR and aberrant cytoplasmic translocation were observed in GC. Using immunohistochemistry staining, we found that both positive cytoplasmic localization and elevated nuclear HuR expression were closely related to poor clinicopathological features and survival of GC patients.
Furthermore, we observed that dysregulation of TTP and HuR induced elevated expression of HMGB1 in different ways. Increasing evidence indicates that HMGB1 is a multifunctional factor with diverse biological functions that depend on the context, modification and its location. As a non-histone chromosomal protein in the nucleus, HMGB1 participates in DNA replication, repair, and transcription. Elevated HMGB1 in the cytoplasm and extracellular environment is associated with a variety of activities, including inflammation, autophagy, anti-apoptosis, proliferation, metastasis, among others 31 . HMGB1 plays a paradoxical role in cancers. It acts as both a cancer-promoter and a tumor-suppressor 19 . For example, acting as an anti-tumor factor, intracellular HMGB1 sustains autophagy and stabilizes the genome. Overexpressed HMGB1 in breast cancer was reported to bind to the tumor suppressor Rb and induce cell cycle-arrest and apoptosis 32 . However, extracellular HMGB1 interacts with different receptors, including RAGE, TLR-2, and TLR-4, and acts as a cytokine, chemokine, and  growth factor to promote carcinogenesis 33 . Elevated HMGB1 levels are associated with poor survival of cancer patients 23,34 . Several studies have certified that HMGB1 is overexpressed in GC 17,19,24,35 . And higher serum HMGB1 levels have also been reported in GC patients compared to those in normal patients 24 . However, the mechanism by which HMGB1 is overexpressed in cancers is not fully elucidated 20 . A recent study reported that HuR bound to U-rich elements in the 3′UTR of HMGB1 and promoted its translation in the progression of myogenesis. In accordance with this report, we observed that elevated HuR had no effects on mRNA expression of HMGB1, but promoted its expression at translational level. In this study, we observed that downregulated TTP-expression increased HMGB1 expression at mRNA level by binding to its ARE directly. Dysregulation of the TTP and HuR in GC promoted the proliferation and metastasis of GC cells in part by elevated levels of HMGB1. Soluble RAGE, the HMGB1 antibody, and other agents, such as glycyrrhizin and quercetin that target HMGB1 showed a promising prospect for anti-cancer therapy 20 . Thus, rebalancing the TTP-HuR axis showed potency in reducing cancer. For example, inhibition of miR-29a, a miRNA that recognizes a seed target site on the 3′UTR of TTP, rebalanced the TTP-HuR axis and reduced the invasiveness of breast cancer cells 11 .
Collectively, our results demonstrated that dysregulation of TTP and HuR played an important role in GC. The increased HuR levels were partly due to the downregulation of TTP. Positive cytoplasmic HuR expression was closely related to the depth of invasion, lymph node metastasis, distant metastasis and TNM stage of GC patients. Elevated nuclear HuR levels were correlated with the depth of invasion, TNM stage, tumor size and TTP expression in GC patients. Aberrant cytoplasmic HuR and elevated nuclear HuR levels predicted a poor survival in GC patients. Additionally, we demonstrated that elevated expression of HMGB1 was correlated with dysregulation of TTP and HuR. Furthermore, different mechanisms of regulating HMGB1 expression by TTP and HuR were observed in GC. TTP regulated the expression of HMGB1 at the mRNA level by binding to its ARE, while elevated HuR mainly promoted its expression at translational level. Elevated HuR expression promoted proliferation, invasion and migration of GC cells partly through HMGB1. Increased HuR levels in a xenograft tumor model promoted tumor growth of GC cells in vivo. The underlying mechanisms by which the dysregulation of TTP and HuR promotes cancer progression are worthy for further research. And the TTP-HuR axis might serve as a potential therapeutic target and prognostic indicator for GC patients.

Materials and Methods
Ethics statement. This study was approved by the ethics committee of Nanjing medical university and the written informed consents were obtained from all participated patients. All experiments with mice were approved RNA preparation, reverse transcription-polymerase chain reaction (RT-PCR) and quantitative real-time PCR (qRT-PCR). Total RNA was extracted by TRIzol Reagent (Invitrogen, CA, USA) from cells or fresh samples according to the manufacturer's protocol. Next, RT-PCR was performed with 1 µg of total RNA using a PrimeScript RT Reagent Kit with a gDNA eraser (Takara, Dalian, China). Target mRNA expression was determined by qRT-PCR using the QuantiFast SYBR Green PCR Kit (Qiagen, Germany) on an ABI Step One Plos Fast real-time PCR system (Applied Biosystems, Austin, USA). The primer sequences used in this study were as follows: TTP-forward, 5′-TTCGCCCACTGCAACCTC-3′; TTP-reverse, 5′-CGCCCACTCTCTGAGAAGGTC-3′; HuR-forward, 5′-CCGTCACCAATGTGAAAGTG-3′; HuR-reverse, 5′-TCGCGGCTTCTTCATAGTTT-3′; HMGB1-forward, 5′-TATGGCAAAAGCGGACAAGG-3′; HMGB1-reverse, 5′-CTTCGCAACATCACCAATGGA-3′; Glyceraldehyde-3-phosphate dehydrogenase (GAPDH)-forward, 5′-AAGGTGAAGGTCGGAGTCAA-3′; GAPDH-reverse, 5′-AATGAAGGGGTCATTGATGG-3′. The specificity of primers was verified by the melting curve and agarose gel electrophoresis. The results of target mRNA expression were normalized to GAPDH and quantified by the 2 −ΔΔCT method. Transcription of mRNAs was blocked by actinomycin D (ActD) (5 µg/ml) to estimate the stability of target mRNAs. All samples were measured at least in triplicate. DNA transfection and RNA interference. The pcDNA-TTP plasmid and pcDNA 3.1 empty vector were constructed as previously described 16 . The pCMV6-HuR (RC201562) expression plasmids as well as the pCMV6-Entry (PS100001) empty vector were purchased from OriGene (OriGene Technologies, Beijing, China). Small interfering RNAs (siRNA) of TTP, HuR, HMGB1 and the control siRNA (si-NC) were purchased from Riobo Bio (Riobo Bio, Guangzhou, China). Plasmids and siRNAs were transfected into cells cultured on 6-well plates using Lipofectamine2000 (Life Technologies, USA). Next, total RNA and protein were harvested after 24 h or 48 h for further analysis. The efficiency of the plasmids and siRNAs were verified by either qRT-PCR or western blotting.
Luciferase assay. The wild type sequence of HMGB1 3′UTR (wtARE) and the mutant oligonucleotides in which all of the 14 ATTTA motifs were substituted with AGGTA (mutARE) were synthesized and subcloned into the pmirGLO Vector by Gene Pharma company (Suzhou, China). The 293 T cells were cotransfected with pmirGLO-wtARE or pmirGLO-mutARE vectors (0.5 µg) and TTP plasmids (0.5 µg). The lysates of the transfected cells were mixed with luciferase assay reagent in accordance with the manufacturer's instructions (Promega, USA),  Table 4. Univariate and multivariate analyses for the overall survival of GC patients. P value < 0.05 was indicated in bold.