In a large unselected cohort, we have used high-throughput TMA-technique and in situ hybridization to evaluate the prognostic impact of miR-21 expression in tumor tissue and stroma of NSCLC. To our knowledge, we are the first to use ISH to study miR-21 expression and outcome in a large NSCLC cohort, facilitating analyses discriminating specifically between tumor cells and cells of the tumor stroma. We find high tumor miR-21 expression in node-positive patients to be an independent positive prognostic indicator. In univariate analyses, we find high stromal miR-21 expression to be a negative prognostic factor in the total material and in node-negative patients. There was no correlation between miR-21 and angiogenesis-related markers.
miRNAs have a large impact on gene regulation, and are considered major players in tumor development and metastasis . These nucleotides act as both oncogenes and tumor suppressor genes, as they may be both up- and down-regulated in tumors. miR-21 is known to be abundantly expressed in a variety of cancers, and is in many tumor types associated with a reduced overall survival . In a recent array screening study, including 20 NSCLC patients and 10 controls, we found miR-21 to be one of the most upregulated miRNAs in tumor tissue, when compared to normal tissue, by both microarray hybridization and quantitative real time polymerase chain reaction (qRT-PCR) technique .
In recent years, a few studies have explored the prognostic impact of miR-21 in NSCLC. Markou et al. studied 48 patients, where 67% where in stage I/II and 33% stage III/IV. They found miR-21 to be an independent negative prognostic factor for OS . Gao et al. observed the same in 47 NSCLC samples. In their cohort, 47% were stage I, 25% stage II and 28% stage III, respectively . In three cohorts from Maryland, US (64% stage I, 25% stage II, 11% stage III), Norway (57% stage I, 14% stage II, 29% stage III), and Japan (74% stage I, 26% stage II), Saito and colleagues showed miR-21 to be an independent negative prognostic factor. In the Norwegian and American material, overall survival was the endpoint, while in the Japanese cohort relapse free survival was . Landi and colleagues used an oligo array with 440 human miRNAs to evaluate differences in miRNA expression depending on histology and clinical outcome in 290 NSCLC tissues, constituted by 40% stage I, 29% stage II, 26% stage III and 4% stage IV cancers. They found miR-21 to differentiate between adenocarcinoma and squamous cell carcinoma, but there was no difference in survival according to miR-21 expression rate . In a large study on 639 patients (35% stage I, 23% stage II and 42% stage III), Voortman et al. observed no prognostic impact of miR-21 on NSCLC survival. There was a tendency towards a better prognosis for high miR-21 expression, but this finding was not significant (P = 0.06) .
There were some differences in methodology between these studies, as fresh frozen tissue was used in three [11, 13, 14] and paraffin-embedded material in two studies [12, 15]. For quantification of miRNA, all except the Landi study used qRT-PCR. In the studies where miR-21 was associated with a worse prognosis, subgroup analyses were not performed. The Landi and Gao studies [11, 12] were numerously too small and the Saito study consisting of three cohorts (89, 37 and 189 patients respectively) , was not suited for subgroup analyses.
Yang et al. performed a meta-analysis based on the studies mentioned above. They also included two other studies analyzing miR-21 in serum. The conclusion was that high miR-21 expression was significantly associated with poor survival .
When using the summarized score of tumor cell and stromal expression in the whole cohort, miR-21 was without any significant prognostic impact. In the stromal compartment, miR-21 expression was a negative prognosticator in univariate, but not in the multivariate analysis. In the studies mentioned above [11, 13, 14], except for the studies by Voortman  and Landi , miR-21 appears to be a negative prognostic factor in NSCLC, corroborating our stromal results. In contrast, we found miR-21 expression in tumor cells to be an independent positive prognostic factor in node positive lung cancer patients. One may speculate if the differences we observe between our results and some of the other studies are caused by the different methodologies used. As we use ISH-technique, miR-21 expression can be assessed separately in tumor and stromal cells. When using the qRT-PCR method without prior microdissection, it can not be differentiated between the tumor cells and the stromal compartment. Gregg and colleagues performed microdissection on a prostate cancer material to separate tumor and stromal cells, and showed a large difference regarding gene expression between the two compartments . Our findings show that there is a prognostic difference in expression between the two compartments. Using qRT-PCR the contribution of miR-21 from the stromal compartment may override the contribution from tumor cells, especially at significant differences in expression. Consequently, the data will reflect the situation in stroma, and a divergent situation in the tumor cells will not be detected. The conclusion in the Yang meta-analysis reflects the findings in these studies, and does not take into account possible expression differences between the compartments.
The mechanistic functions of miR-21 are still being explored, but some functions have recently been suggested. In human umbilical vein endothelial cells (HUVECs), Sabatel and colleagues found miR-21 to be a potential inhibitor of angiogenesis via inhibition of RhoB, resulting in a reduction in endothelial proliferation, migration and vessel formation . On the other hand, Liu et al. demonstrated miR-21 to induce angiogenesis in human prostate cancer cells through up-regulating HIF-1α and VEGF and through activating the Akt and ERK pathways . Hence, miR-21 may have both pro- and anti-angiogenic functions. Angiogenesis is an important trait of cancer progression , and inhibiting angiogenesis may contribute to slow down cancer growth. In the light of these opposing findings and our data, we may speculate if miR-21 has opposite impacts in different stages of the disease. Could high miR-21 in early stages (node-negative) act pro-angiogenic and contribute to a faster progression, while in a node positive stage, it acts anti-angiogenic and protects against further progression?
We did not find any correlations between miR-21 and angiogenic markers of pathways earlier described for miR-21 and angiogenesis [9, 10]. These pathways are possibly tissue and cell type specific. Studies exploring miR-21 as a modulator of angiogenesis have in some cases used endothelial cells in their models . In our material, we have not specifically studied endothelial cells, so the connection between miR-21 and angiogenetic markers seen in endothelial cells will not necessarily be mirrored by our tissue samples as we assessed the sum off all stromal cell types.