Distinct miRNA profiles have been described for many cancers including hematologic and solid malignancies 123456789101112. Many reports have shown that patterns of miRNA expression differ between normal and cancerous tissues 12345678910121314151617181920. Gene expression profiling of traditional mRNA targets in whole blood or fractionated leukocytes has also shown correlations with many types of both neoplastic and non-neoplastic human disease, for example renal cancer and Crohn's disease 21222324252627282930. To investigate whether miRNA patterns in blood correlated with tumorigenesis, we measured by qRT-PCR a panel of miRNAs in MYC-induced transgenic models of tumorigenesis.

First, we developed a protocol optimized for collection, storage and shipping of whole mouse blood, RNA extraction from a small volume of stored sample, and qRT-PCR assays for mouse blood miRNA profiling. To enable blood to be collected from mice at different time points and stored so that total RNA extraction and miRNA quantitation could be batch analyzed, mouse blood was mixed with an RNA stabilizing reagent (RNAlater^® Tissue Collection:RNA Stabilization Solution, Ambion), transported, and stored at -20 deg C. Total RNA extraction was subsequently performed using the Mouse RiboPure™ Blood kit (Ambion).

Total RNA yields from cardiac puncture samples from 6 mice averaged 114 μg by NanoDrop measurement (Figure 1A). To determine the quality of total RNA extracted, Agilent bioanalyzer scans were performed. RNA was found to be intact as shown by the high RIN values observed in the Agilent traces and the sharp 18S and 28S ribosomal RNA (rRNA) peaks and lack of significant species between those peaks (Figure 1B, C). The bioanalyzer traces also showed a distinct peak that correlates with the recovery of low molecular weight RNA including miRNA. The high yields of RNA from mouse blood suggested that sufficient RNA for analysis might be extracted from lower volume tail vein bleeds that do not require sacrifice of the mouse. Indeed, blood acquired via tail vein from eight mice yielded an average of 15.8–33.7 μg RNA (Figure 1A). RNA yields from even the lowest amounts of tail vein blood were sufficient to run several hundred miRNA assays (as described below). Our initial experiments thus established the feasibility of carrying out blood miRNA profiling on individual animals during time-course experiments.

To determine which miRNAs could be readily detected in mouse blood, a panel of 111 quantitative reverse transcriptase PCR (qRT-PCR) SYBR miRNA assays was run on initial samples of blood from normal mice (see Additional file 1). The qRT-PCR assays were normalized via several methods, including 5S rRNA, U6 RNA, and global mean. Results were validated through the detection of several miRNA using TaqMan^® qRT-PCR assays. High concordance was observed in relative levels of blood miRNAs as determined by the SYBR-based and TaqMan-based assays, with a correlation value of 0.86 (Figure 1D). We conclude that qRT-PCR is a highly sensitive approach for identifying mouse blood miRNA profiles.

To characterize blood miRNA patterns in mice with cancer, we utilized our previously described conditional transgenic mouse models of MYC-induced lymphoma, hepatocellular carcinoma and osteosarcoma 31323334353637. 30 miRNAs that were reproducibly detectable using our methodology and known to be involved in the pathogenesis of cancer were selected for quantitation 514161718203839. Significant differences were found in 12/30 miRNA in blood of mice with lymphoma, and 8/30 miRNA in blood of mice with liver tumors compared with normal controls (Figure 2). 25/30 miRNA for a mouse with osteosarcoma were also significantly different from normal controls, suggesting that further studies might be warranted in other tumor models (see Additional file 2). We conclude that changes in specific miRNAs can be detected in the blood of transgenic mice with lymphoma, hepatocellular carcinoma, and osteosarcoma. It is possible that we would have detected additional miRNA changes in mice with liver cancer if we had measured a broader panel of miRNAs. Alternatively, blood miRNA changes may be more readily detected in hematopoietic and mesenchymal tumors than in epithelial tumors.

Next, we attempted to identify specific changes in miRNA profiles during lymphomagenesis. We performed qRT-PCR from blood specimens collected from transgenic mice before tumorigenesis (preneoplastic, n = 3), early tumor formation (subclinical lymphoma, n = 2) and after development of clinical lymphoma (overt lymphoma, n = 4). The results for a panel of 27 blood miRNA in all groups were compared to values from mice without MYC activation (normal control, n = 5) (Figure 3A). Unsupervised hierarchical clustering analysis revealed that overt lymphoma miRNA profiles grouped together. In mice with overt lymphoma, we identified a decrease in 20 miRNAs, including let-7d and miR-16 (Figure 3A). Notably, both of these miRNAs have been implicated in the pathogenesis of cancer 5. Importantly, we found that the miRNA profile in preneoplastic mice was markedly different from mice with overt lymphoma (Figure 3A). Hence, our results suggest the possibility that the performance of miRNA measurements in mice might be useful to monitor tumor development.

An important feature of our method is that its sensitivity should allow for the analysis of serial blood specimens from the same mouse. We obtained serial tail vein blood samples from mice with MYC-induced lymphoma before and after oncogene inactivation. The majority of miRNAs, including members of the let-7 family, were detectable at low levels and then increased upon tumor regression (Figure 3B) as early as 3 days after MYC inactivation even though tumors did not fully regress until after 14 days. Thus, changes in blood miRNA profiles can be determined from sequential peripheral blood samples drawn from individual mice.

An important caveat of our studies is that we cannot determine the source of the miRNA changes. A priori, we could be detecting miRNAs in rare circulating tumor cells or from host cells that are influenced by tumor growth. Regardless, the changes in miRNAs we see correlated with tumor progression, and thus may be useful as biomarkers of tumorigenesis.

Compared to microarray-based global miRNA profiling, the real-time PCR assays used in our studies require several magnitudes less input RNA and can be performed in substantially less time. These features of our method make it tractable for the detection of miRNA expression to identify early metastasis or minimal residual neoplastic cells. Moreover, we found that specific blood miRNA profiles mirrored specific types and states of cancer. Whether our approach will also be useful for the analysis of human patients remains to be determined. Finally, our method appears to be applicable to the sequential analysis of changes in miRNA expression in other models and in clinical materials. Our approach may be useful to identify biomarkers to detect early disease states and to predict clinical response.

At the time the experiments were performed, Marianna Goldrick, Jennifer Ho and Yu Liang were scientists employed by Ambion/Applied Biosystems.

ACF, MMG and DWF conceived and designed the experiments. ACF, MMG, JH and PB performed experiments. MMG, JH, and YL analyzed data. ACF, MMG, PB and DWF wrote the manuscript. All authors read and approved the final manuscript.