We used a lentiviral vector encoding an shRNA targeting the 3′UTR of murine Parn to generate a stable clonal C2C12 myoblast line with reduced expression of PARN (PARN KD). Parn mRNA and protein abundance were evaluated by qRT-PCR () and western blotting () in the PARN KD cell line and in a cell line generated with a control lentiviral vector lacking shRNA sequences (CTRL). The PARN KD cell line showed a robust reduction in PARN expression ().
Determination of global mRNA decay rates in PARN knockdown cells.
We first wanted to assess mRNA decay rates in the PARN KD cells and compare them to those we obtained previously in the CTRL cell line 
. Briefly, both cell lines were treated with Actinomycin D (Act-D) for 30 minutes and samples were collected at 0, 10, 50, 110 and 230 minutes after transcription inhibition. Total RNA was isolated from each sample and used to generate cDNA probes for hybridization to microarrays. The experiment was repeated in triplicate and three independent half-lives were generated for each transcript in each cell line by plotting the abundance at each time point and fitting to an exponential decay curve. As an example the half-lives for the Gpsm1
mRNA in the two cell lines are shown in . Each half-life was considered reliable if the data fit well to the curve (p<0.05) and the 95% confidence interval was less than twice the half-life. We required that the half-life met these criteria for at least two of the three replicates. Both PARN KD and CTRL cells were assayed at the same time but analysis of the results from the CTRL cells was published previously 
Reliable half-lives were generated for 1581 mRNAs in the PARN KD cells (Dataset S1A
). Although this dataset is somewhat smaller than that previously obtained for the CTRL cell line [7398 mRNAs; 23]
, it is nevertheless large enough to be informative. Overall, we obtained half-lives in both cell lines for 1389 mRNAs (Dataset S1B
). Comparison of half-lives in CTRL and PARN KD cells allowed us to identify 64 transcripts that showed a statistically significant difference in decay rate between the two cell lines with 40 transcripts showing stabilization and the remaining 24 being destabilized ( and Table S1
, respectively). To ascertain that the microarray analysis reflected true changes in mRNA decay rates, we assayed half-lives following Act-D treatment for four of the stabilized transcripts (Adora2b
) by qRT-PCR (). These transcripts were selected because they have relatively short half-lives (less than 2 hours) allowing us to assess their decay over a time frame that minimizes the toxic effects of Act-D on the cell. All four transcripts were significantly more stable following PARN knockdown, as predicted by the microarray analysis. Moreover, instability of the Zfp36l2
mRNA was restored by transfection of an expression vector encoding shRNA-resistant human PARN demonstrating that stabilization was not caused by off-target effects of the shRNA on expression of unrelated genes (). Thus, we conclude that the PARN deadenylase influences decay rates of a subset of mRNAs in mammalian cells.
Forty mRNAs are stabilized in PARN knockdown cells.
Independent assays validate the changes in decay in PARN KD cells and verify that stabilization is due to PARN depletion.
Given that PARN is a deadenylase, we predicted that mRNAs stabilized by PARN KD would show effects on the length of their poly(A) tail. We investigated this possibility for the Zfp36l2
mRNA using an RNase H/northern blotting approach. Briefly, total RNA isolated from CTRL and PARN KD cells was treated with an oligonucleotide and RNase H to induce cleavage ~120 nt upstream of the poly(A) tail. After separation on a polyacrylamide gel followed by electroblotting, the 3′ fragment was detected using a radiolabeled probe complementary to the 3′UTR. As shown in , the poly(A) tail of Zfp36l2
mRNA was clearly elongated in PARN KD cells compared to the CTRL cells. In fact, in the CTRL cells the vast majority of Zfp36l2
mRNA had a surprisingly short poly(A) tail of just 20–30 nt. In the PARN KD cells the amount of Zfp36l2
mRNA with a long poly(A) tail of up to ~190 nt was two to three fold more than in the CTRL cells. This was not a general effect on all mRNAs as the β-Actin (Actb
) mRNA showed no difference in poly(A) tail length between the two cell lines (Figure S1
). Although abundance of Zfp36l2
mRNA was similar in CTRL and PARN KD cells (), western blotting () demonstrated a small increase in abundance of ZFP36L2 protein which would be consistent with enhanced translation resulting from the elongation of the poly(A) tail. We also saw evidence for increased abundance of ZFP36L2 protein by immunofluorescence (Figure S2
PARN modulates Zfp36l2 poly(A) tail length to reduce expression of ZFP36L2 protein.
In order to determine whether the effects of PARN on the Zfp36l2
mRNA are mediated by sequences in the 3′UTR we cloned the 3′UTR into a luciferase reporter construct (Luc-36L2) and measured luciferase activity following transfection into CTRL and PARN KD cells. The empty vector (Luc) was used as a control and gave very similar activity regardless of whether expressed in the CTRL or PARN KD cells (). Interestingly, the Luc-36L2 reporter produced significantly less luciferase activity than the Luc reporter in the control cell line suggesting that the sequences contained therein either repress translation or promote decay of the reporter mRNA. Importantly, PARN KD cells reproducibly exhibited a two-fold higher luciferase activity than the control cells () and this was also seen when PARN was knocked down with a different shRNA (Figure S3
) showing that this effect is PARN-specific. Interestingly, the clear increase in luciferase activity following PARN depletion is mediated predominantly by enhanced translation as there was little effect on abundance of either reporter mRNA in PARN KD cells (). The increase in luciferase expression is in the same range as the increase in abundance of endogenous ZFP36L2 protein in PARN KD cells (). Together these results indicate that the action of PARN on the Luc-36L2 reporter results in translation repression presumably through poly(A) shortening. Moreover, factors associated specifically with the 3′UTR of Zfp36l2
mRNA are likely responsible for the effects of PARN on Zfp36l2
gene expression. At this time we do not know what factor might be responsible for recruiting PARN, but the Zfp36l2
3′UTR does have AU-rich elements like those reported to bind proteins such as TTP/ZFP36; a protein that induces PARN-mediated deadenylation in vitro
PARN acts through the 3′UTR of Zfp36l2 to repress translation.
The relatively small number of mRNAs affected by PARN at the level of mRNA stability precluded a meaningful analysis of Gene Ontology (GO) terms or sequences that might impacted by reduced PARN activity. Still, we did note that several of the stabilized transcripts encode proteins with roles in mRNA metabolism (Toe1/Caf1z, Edc3, Zfp36l2, Dgcr14, Nufip1
) and transcription (Gata2, Zfp219, Klf14
) indicating that PARN may influence gene expression at multiple levels and impact a wider range of genes. To investigate this possibility we used the 0 minute time point from the array experiments to estimate global mRNA abundances in CTRL and PARN KD cells. We found that of the 18,201 transcripts detected, 1199 showed a 1.5-fold or greater change in mRNA abundance in PARN KD cells (Dataset S2
). Surprisingly, given that PARN KD was expected to increase expression of its target mRNAs, the majority (63.7%) of the affected mRNAs were down-regulated. We verified the abundance changes for several transcripts by qRT-PCR and found that of 14 mRNAs examined, all but one (Lama2
) showed changes similar to those predicted by the array (). Moreover, there was generally a good correlation between the change predicted by the microarray and that observed by qRT-PCR in untreated cells although the qRT-PCR indicated changes of a greater magnitude than the array (Figure S4
). This confirms that Act-D treatment did not globally affect our mRNA abundance measurements and that the 0 minute time point mRNA abundances are generally an acceptable indicator of relative differences in mRNA abundance between PARN KD and CTRL cell lines.
mRNA stabilization is not correlated with increases in mRNA abundance.
We next took advantage of the availability of both mRNA abundance and decay data to analyze the impact of changes in mRNA stability on overall mRNA levels. We were surprised to discover that for the 40 transcripts showing clear evidence for stabilization following PARN knockdown, there was generally only a small effect on mRNA abundance and in many cases abundance was reduced rather than increased (). There was a similar inverse correlation for the mRNAs that were destabilized (). In order to verify this observation, we measured the abundance of three transcripts that were stabilized by PARN depletion in proliferating myoblasts (). Interestingly, Adora2b
mRNA (1.4-fold stabilized ( and )) showed ~2-fold reduced abundance in PARN KD cells, while Ankrd54d
mRNA (1.85-fold stabilized) showed no statistically significant change in abundance (). In contrast, Gpsm1
mRNA (1.96-fold stabilized) did show a small increase in abundance by this assay. As described earlier (), there was no significant change in abundance of the Zfp36l2
transcript despite a ~2.4-fold increase in stability. Taken together, these results strongly suggest the existence of coupling between transcription and decay for many transcripts such that changes in mRNA decay rate are compensated for by opposing effects on transcription 
In order to further support this idea we assessed the abundance of newly transcribed pre-mRNAs for each of the four stabilized transcripts. Briefly, C2C12 cells were labeled for a short time with 4-thiouridine (4sU) and total RNA was prepared. Newly transcribed 4sU-labeled RNAs were biotinylated and isolated on streptavidin beads. Pre-mRNAs were detected and quantified by qRT-PCR using one intronic primer and one exonic primer. As shown in , all four pre-mRNAs exhibited significantly reduced abundance in the PARN KD cells, consistent with slower transcription rates for these transcripts in this cell line. To summarize, each of the four mRNAs we evaluated showed increased stability following PARN KD () but reduced levels of pre-mRNAs indicating reduced transcription (). This change in the relative balance of decay and transcription results in only small changes in mRNA abundance ().
GO analysis using DAVID 
revealed that the transcripts whose expression was most affected by PARN shared some interesting features (Tables S2
). In particular, amongst the down-regulated genes there was a significant enrichment of mRNAs encoding proteins required for blood vessel development, cell adhesion, cell motion and axon guidance (Table S2
). This is supported by the observation that a large proportion (~15%) of the down-regulated mRNAs encoded extracellular proteins including several collagens (Col1a1, Col1a2, Col6a1, Col6a2, Col3a1, Col12a1
), biglycan (Bgn
) and matrix metalloproteases (Mmp19, Mmp2
). In contrast, the up-regulated mRNAs were more likely to encode components of large ribonucleoprotein complexes such as the ribosome and spliceosome (Table S3
Our GO analysis suggested that PARN knockdown might influence cell motility as cell movement requires extensive interactions with the extracellular matrix and is required for processes such as axon guidance and blood vessel development. We used a wound healing assay to investigate the ability of CTRL and PARN KD cells to migrate. Briefly, CTRL and PARN KD cells were grown to near confluence and then deprived of serum to prevent cell division. The monolayer was scratched to remove cells and incubated for eight hours to permit cells to migrate into the wound. Wound healing was assessed by counting the number of cells present within the boundaries of the wound. There was a clear increase in the wound healing capacity of PARN KD cells compared to the CTRL cells () indicating that PARN KD cells migrate more rapidly. Moreover wound healing was restored to near normal levels following transfection of a plasmid encoding human PARN (). We conclude that PARN modulates processes required for cell motility in C2C12 myoblasts.
PARN knockdown results in enhanced wound healing.