Characterization of the mammalian miRNA turnover landscape

Steady state cellular microRNA (miRNA) levels represent the balance between miRNA biogenesis and turnover. The kinetics and sequence determinants of mammalian miRNA turnover during and after miRNA maturation are not fully understood. Through a large-scale study on mammalian miRNA turnover, we report the co-existence of multiple cellular miRNA pools with distinct turnover kinetics and biogenesis properties and reveal previously unrecognized sequence features for fast turnover miRNAs. We measured miRNA turnover rates in eight mammalian cell types with a combination of expression profiling and deep sequencing. While most miRNAs are stable, a subset of miRNAs, mostly miRNA*s, turnovers quickly, many of which display a two-step turnover kinetics. Moreover, different sequence isoforms of the same miRNA can possess vastly different turnover rates. Fast turnover miRNA isoforms are enriched for 5′ nucleotide bias against Argonaute-(AGO)-loading, but also additional 3′ and central sequence features. Modeling based on two fast turnover miRNA*s miR-222-5p and miR-125b-1-3p, we unexpectedly found that while both miRNA*s are associated with AGO, they strongly differ in HSP90 association and sensitivity to HSP90 inhibition. Our data characterize the landscape of genome-wide miRNA turnover in cultured mammalian cells and reveal differential HSP90 requirements for different miRNA*s. Our findings also implicate rules for designing stable small RNAs, such as siRNAs.


The fast turnover rates are an intrinsic property of relevant miRNAs
For fast-turnover miRNAs, we asked whether the fast turnover rate is an intrinsic property of the relevant miRNAs. This is because global transcription inhibition, in theory, may affect potential miRNA "protectors" if they themselves turnover fast.
We modeled miR-222-5p (miR-222*) and miR-23a-5p (miR-23a*), and used a tet-off system to ectopically express them in cell lines that normally do not express the relevant miRNA at high levels. With the inducible expression, we could shut down the ectopic miRNA expression and follow their turnover without affecting global transcription in the cells (Fig S6A). Specifically, we transduced MCF-7 cells with tTA and a miR-222hairpin-expressing vector under the control of tetO promoter. After establishing ectopic expression of miR-222-5p and miR-222-3p, we added Doxycycline to shut down the ectopic transcription. Because this Tet-off system may not stop transcription as instantaneously as chemical transcriptional inhibitors, we first evaluated its performance using a short-lived luciferase, known to have very short RNA and protein half-lives (1) ( Fig S6D). The level of luciferase RNA dropped ~50% after 4 hours and ~75% after 6 hours. In comparison, the level of the fast-turning-over miR-222-5p decreased to background level after 6 hours ( Fig S6B), even faster than luciferase. In contrast, the stable miR-222-3p only showed ~20% drop after 24 hours (Fig S6B), a level expected because the proliferation of MCF-7 cells effectively diluted the pre-made stable miR-222-3p. Similar results were obtained with miR-23a in Raji cells (Fig S6C), a cell line with faster proliferation. These data indicate that the fast-turnover rates of these miRNAs are intrinsic properties of the miRNAs themselves.
In addition, we found it highly unlikely that active extracellular secretion underlies rapid miR-222-5p turnover. In established HDMYZ culture, the amount of miR-222-5p in the cell medium was only < 0.05% of the amount in cells (Fig S6E, S6F, S6G). In addition, the level of miR-222-5p in medium did not significantly change after cells were treated with ActD, whereas the intracellular level of miR-222-5p drastically dropped after treatment ( Fig S6E). Together, these data support the notion that the fast-turnover rates are properties of these miRNAs and are mediated through an intracellular mechanism.
The inducible lentiviral miR-222 and miR-23a plasmid was cloned similarly as published (2) To normalize qRT-PCR data, miRNA or c-myc level was first normalized to 18s or U6 level from the same sample, then normalized to their respective 0 hour samples or control.
For IP samples, the data for miRNAs, 18s and U6 level were normalized based on sample fraction, so that the data reflect their levels relative to those in input (i.e. 100% means complete recovery of the relevant RNA species from the IP experiment).

MCF-7 cells were transduced with retroviruses for pMIRWAY-GFP-WT-miR-222
or pMIRWAY-GFP-mut-miR-222. To control for similar transduction levels for wild-type and mutant miR-222 viruses, we controlled the transduction rate to be ~50% for both viruses. Transduced cell pools were then sorted by FACS, gating on GFP+ cells with similar GFP levels between wild-type and mutant miR-222. Cells were then treated with 10ug/ml ActD for 0, 2, 4, 6, 12, 24 hours.

Measurement of miR-222 secretion
HDMYZ cells were treated with 2.4ug/ml ActD or the same volume of DMSO for 8 hours. The same fraction of HDMYZ cells and cell culture supernatant were collected in TRIzol and TRIzol LS reagent for RNA extraction (e.g. 20% of cells and 20% of medium). All the RNA samples were diluted in 7ul of water and 5.8ul of RNA from each sample was used in Qiagen miScript RT-qPCR system. The miR-222-3p, miR-222-5p and U6 level in the cells were normalized to their level in cells with DMSO treatment.