Decreased miR-199a-5p expression in HCC has been repeatedly reported, but its functional relevance has not been elucidated to date [11, 12, 15]. A pivotal role for miRNAs in the process of malignant transformation has been suggested in the literature [9, 31]. However, the precise molecular mechanisms by which miRNAs modulate tumor cell biology are largely unknown. MiRNAs from animals were first reported to repress translation without affecting mRNA levels . More recent evidence indicated that miRNAs and siRNAs can control post-transcriptional gene expression by directing the endonuclease cleavage of target mRNA, which is referred to as "slicer" activity . Endonucleolytic cleavage is generally favored by perfect base-pairing between miRNA and mRNA. Some mismatches, however, can be tolerated and still allow endonucleolytic cleavage to occur . The majority of animal miRNAs are only partially complementary to their targets . Several reports have shown that animal miRNAs can also induce significant degradation of target mRNAs despite imperfect mRNA-miRNA base-pairing [36, 37], referred to as "slicer"-independent decay. This phenomenon emphasizes mRNA degradation as an important aspect of miRNA-mediated repression of gene expression. There is also some evidence that "slicer"-independent mRNA decay induced by miRNAs might occur through promotion of mRNA decapping and 5'-3'-degradation . The contribution of translational repression or mRNA degradation to gene silencing appears to differ for each miRNA:target pair and is likely to depend on the particular set of proteins bound to the 3'UTR of the mRNA . MiR-199a-5p, which is partially complementary to the 3' UTR of DDR1 mRNA, induced significant degradation of DDR1 mRNA in hepatoma cells in our study. Thus, DDR1 has been experimentally validated as a target gene of miR-199a-5p. MiR-199b-5p, which has a very similar nucleotide sequence, is also predicted to target DDR1. However, miR-199b-5p was only detectable in four out of 24 tissue samples in a miRNA microarray assay (data not shown) and it has never been reported to be de-regulated in HCC in literature. Therefore, the regulatory function of miR-199b-5p was not further assessed in this study.
DDR1 is a tyrosine kinase receptor for collagen  and its activation can cause tumor invasion which appears to be mediated by matrix metalloproteinases (MMP)2 and MMP9 [19, 23]. Indeed, we found a positive correlation between the expression of DDR1 and MMP2 in our patient cohort (data not shown), further supporting the important role of DDR1 for tumor invasion. Consistent with recently published findings  and our results indicating a critical role for DDR1 in HCC progression, we found a significant correlation between the expression of DDR1 and AJCC/UICC tumor stage. Analyzing predictive factors for advanced tumor stage at the time of HCC diagnosis, we found a positive correlation between AJCC/UICC stage III-IV and poor tumor differentiation, presence of macrovascular invasion, and high DDR1 expression. In addition, multivariate logistic regression analysis identified high DDR1 expression as the single independent factor associated with advanced tumor stage, and, hence, poor prognosis. In line with this clinicopathological observation, we found that DDR1 gene silencing by transfection of miR-199a-5p into HepG2 cells significantly decreased tumor cell invasion in vitro. However, our studies also indicate that DDR1 is rather involved in the process of tumor invasion than in tumor growth. These results demonstrated that miR-199a-5p is capable to modulate tumor cell invasion at least in part by targeting DDR1. Although in our study DDR1 has been validated as a target gene of miR-199a-5p, no significant correlation between miR-199a-5p and DDR1 mRNA expression was found in tumor samples from our patient cohort. In addition, SNU-182 hepatoma cells exhibited increased levels of expression of both miR-199a-5p and DDR1 mRNA. Transfection of miR-199a-5p did not induce a change in DDR1 mRNA expression, but significantly down-regulated DDR1 protein in SNU-182 cells. However, no significant inhibitory effect on tumor invasion was noted. Considering the preexisting high expression of miR-199a-5p in SNU-182 cells, our results might hint at a certain independence of DDR1 to miR-199a-5p-mediated gene regulation and function in these cells. Our data hint at a more complex regulation network between DDR1 and miR-199a-5p in HCC. One reason might be that HCC is a very heterogeneous tumor entity and distinct cellular components might interfere with the effect of miR-199a-5p on DDR1 . For instance, by the use of miRNA target prediction algorithms other 23 miRNAs were also predicted to target DDR1 . In addition, DDR1 has been demonstrated to be a direct transcriptional target of the p53 tumor suppressor gene  and, therefore, the p53 status in tumor cells may also affect the expression of DDR1. Finally, recent findings reveal a more diverse role for small RNA molecules in the regulation of gene expression than previously recognized. For instance, miRNAs can act also as translation activators under specific cellular conditions . In addition, double strand RNAs can activate rather than repress gene expression by targeting non-coding regulatory regions in gene promoters . To date, the only reported stimulatory effect of miRNA on RNA expression is represented by the interaction between miR-122 and replication of hepatitis C virus RNA in hepatocytes . MiR-199a-5p expression has also been shown to be diversely deregulated in other cancer types. For instance, miR-199a-5p was also found to be down-regulated in ovarian cancer  and oral squamous cell carcinoma , but up-regulated in cervical carcinoma  and bronchial squamous cell carcinoma . Moreover, increased expression of miR-199a-5p has been considered a signature for high metastatic risk and a poor prognosis in uveal melanoma .
Thus, the complexity of the regulation of mRNA by miRNA encountered in our and other studies indicates that the effect of miRNA on its target gene is cell type and environment dependent. However, our study demonstrates a previously uncharacterized biological function of miR-199a-5p such as the ability to inhibit tumor invasion through targeting DDR1.
Less than half of patients with HCC are eligible for potential curative treatment including liver resection and transplantation because of advanced tumor stage at time of diagnosis. The combination of clinical and biological predictors may increase diagnostic accuracy of tumor staging, thus permitting optimized therapeutic management of HCC patients. Although DDR1 expression was shown to be the only predictive factor for advanced HCC, our study was clearly limited by the small sample size which may tend to overestimate the prognostic value of DDR1 expression. Thus, prospective studies that seek and independently validate the prognostic utility of DDR1 expression for patients with HCC in a larger and carefully selected cohort should be conducted. Patient survival after surgical treatment is hampered by frequent tumor recurrence and systemic chemotherapy is largely ineffective . In recent years, kinase inhibitors have become an attractive target class for drug development , and it was shown recently that systemic application of a multikinase inhibitor improves survival of patients with HCC . Further investigation of therapeutic strategies targeting the miR-199a-5p-DDR1 signaling network is therefore warranted. In conclusion, identification of the miR-199a-5p:DDR1 target pair and its crucial role in tumor cell invasion highlight the translational relevance for both prognostic prediction and targeted molecular therapy for patients with HCC.