Serum samples were collected from 30 women: 10 in the first pregnancy trimester (6–12 weeks of gestational age), 10 in the third pregnancy trimester (34–41 weeks of gestational age) and 10 age-matched non-pregnant women. Eligibility for the study was limited to normal uncomplicated singleton pregnancies with no known fetal malformation. All women provided written informed consent and the local institutional review board approved the study. 8 ml of blood was collected from each woman directly into serum collection tubes (Greiner Bio-one, VACUETTE® Serum Tubes 455071). The whole blood was allowed to stand for about 1 h at RT before being centrifuged at 1800 g for 10 minutes at RT. The resultant serum was aliquoted into eppendorf tubes and stored at −80°C.

About 4 ml of urine samples were collected from each individual in a urine container. The urine was then aliquoted into eppendorf tubes and kept frozen at −80°C until it was used for RNA extraction.

100 µl serum or urine was incubated at 56°C for 1 h with 0.65 mg/ml Proteinase K (Sigma P2308). Two synthetic RNAs (IDT) were spiked-in as controls before acid phenol∶chloroform extraction and then RNA was ETOH precipitated ON at −20°C. Next, DNase treatment was performed to eliminate residual DNA fragments. Finally, after a second acid phenol∶chloroform extraction, the pellet was re-suspended in DDW and two additional synthetic RNAs were spiked-in as controls.

RNA was subjected to a polyadenylation reaction as described previously [17]. Briefly, RNA was incubated in the presence of poly (A) polymerase (PAP; Takara-2180A), MnCl2, and ATP for 1 h at 37°C. Then, using an oligodT primer harboring a consensus sequence, reverse transcription was performed on total RNA using SuperScript II RT (Invitrogen). Next, the cDNA was amplified by real time PCR; this reaction contained a microRNA-specific forward primer, a TaqMan probe complementary to the 3′ of the specific microRNA sequence as well as to part of the polyA adaptor sequence, and a universal reverse primer complementary to the consensus 3′ sequence of the oligodT tail.

We have developed a proprietary qRT-PCR based platform for evaluating microRNA levels that detects specifically mature microRNA molecules (see materials and methods and Figure 1A). Initially, we validated that our platform is capable of discriminating between homologous microRNA family members that differ by only a single nucleotide. Three such microRNAs were generated synthetically (hsa-let 7a, c & d) and each one subjected to three independent qRT-PCR reactions, where in each reaction there were present PCR primers specific to only one of the 3 family members. In most of these reactions, we observed amplification only of the appropriate family member matching the specific primer, indicating that this qRT-PCR platform detects microRNA with accuracy at the single nucleotide level (Figure 1B). Next, we mixed the three synthetic microRNAs and subjected the mixture to qRT-PCR in the presence of the let-7d primer-probe set. In parallel, we took the same amount of synthetic let-7d microRNA as used in the mixture but subjected it alone to qRT-PCR. These parallel PCR reactions were repeated two further times, using reducing concentrations of let-7d synthetic microRNA. We observed that the qRT-PCR amplified let-7d equivalently whether it was alone or in the presence of homologous family members at all tested concentrations of let-7d (Figure 1C). In an additional experiment, we wished to confirm the capability of our platform to detect specific microRNAs in a biologically relevant complex background. Decreasing concentrations (from 100% down to 0%) of total RNA extracted from liver tissue were diluted into total RNA extracted from brain tissue and 0.1 ng of these mixtures subjected to PCR reaction, each reaction containing primers specific for hsa-miR-122a, a liver specific microRNA [18]. Hsa-miR-122a was detected, even when only 0.03% of liver RNA was spiked into brain RNA (Figure 1D). Importantly, hsa-miR-122a detection was linear and there was no detection in the absence of liver RNA (100% brain RNA).

Finally, to validate the reproducibility of our methods, at two independent times, two different researchers subjected the same mix of 32 microRNAs to qRT-PCR. Similar levels of abundance were determined for each microRNA on both occasions, such that the two abundance profiles (for all 32 microRNAs) were within less than 1 C_T difference of one another (Figure 1E). Summarily, our qRT-PCR based platform is sensitive, reproducible and specific, capable of detecting accurately a few molecules of microRNA present in a complex RNA background. Such sensitivity makes it possible to use this platform to monitor the minute amount of microRNA present in cell-free body fluids.

When we initiated this study, there had been no demonstration that microRNAs are present in cell-free body fluids. While this manuscript was in preparation, it was reported that microRNAs are detectable in serum [19]. Here, a reliable protocol is presented that we developed for extracting microRNAs from body fluids, microRNA that is cell-free and DNA-free (see Materials and Methods) and that can serve as the template in our qRT-PCR based platform.

To validate that microRNAs are indeed reproducibly detectable in serum, RNA was extracted from the sera of two healthy unrelated individuals and the levels of highly abundant microRNAs examined (Figure 2A). We observed that microRNAs are present at similar levels in both serum samples. This finding not only establishes the reproducible presence of microRNAs in serum, but also indicates that in general microRNA levels are similar among individuals. Moreover, this finding supports our premise that changes in the levels of specific microRNAs might allow detection of clinical conditions. In parallel, using the same methods we confirmed that microRNAs are also detectable in other body fluids, such as urine, saliva, amniotic fluid and pleural fluid (Figure 2B and data not shown). Of note, serum and urine display different microRNA abundance profiles as might be expected for two dissimilar biological fluids, further supporting our hypothesis that bodily fluid microRNA profiles reflect physiology.

In order for microRNAs in serum to be useful as clinical biomarkers, they must be stable for reasonable periods of time, and preferably during freeze-thaw cycles, to allow for routine processing of clinical samples. Therefore, we investigated the stability of serum microRNAs at room temperature and checked the influence of freeze-thaw cycles on serum microRNA levels. We found that the levels of different microRNAs in unfrozen serum do not change substantially over a 4 hour period at room temperature (Figure 3A). Moreover, serum microRNA levels are little affected by twice freezing and re-thawing of serum samples (Figure 3B). Thus, microRNAs in serum are sufficiently robust to serve as practicable clinical biomarkers.

Finally, as a proof of concept, we investigated whether circulating microRNAs can be used to identify clinical conditions. It has been established that circulating maternal RNA contains placental embryonic RNA [2]. Therefore, we chose to compare the serum microRNA abundance profiles of non pregnant versus pregnant women, the latter in either their first or third trimester. We measured the serum levels of 28 microRNAs, including microRNAs reported to be placenta-specific [10], [20] as well as broadly expressed microRNAs. Box plots show relative microRNA levels in the sera of 10 non pregnant women, 10 women in the first trimester and 10 women in the third trimester (Figure 4A). The median fold changes in microRNA levels comparing third trimester pregnant women to non pregnant women are detailed in Table 1. MicroRNAs expressed equivalently across all samples were used for normalization. All of the placental microRNAs are found at higher levels in the sera from pregnant women, their levels rising with gestational age, and the levels of 12 microRNAs increased by more than 5-fold. Specifically, amounts of hsa-miR-526a and hsa-miR-527 are dramatically higher in the sera of third trimester pregnant women (elevated by more than 600 fold). Indeed, we found that the levels of three placental microRNAs (hsa-miR-526a, hsa-miR-527 and hsa-miR-520d-5p) could be used to accurately distinguish pregnant from non pregnant women (Figure 5).

While we were preparing these results for publication, another research group reported that hsa-miR-141 and hsa-miR-149 are present in serum and display increased abundance during pregnancy [19]. Therefore, we examined also the levels of these two microRNAs in our serum samples. We found hsa-miR-141 and hsa-miR-149 levels to be higher by 4 and less than 2 fold, respectively, in third trimester pregnant women relative to non pregnant women (Table 1 and Figure 4B). These two microRNAs are associated with epithelial tissues [16] that are not unique to pregnancy and this may explain the small changes in their amounts during pregnancy, which are insufficient to discriminate pregnant from non pregnant women.