Aletta's group reported that protein methyltransferase inhibitors completely blocked NGF-induced neurite extension of PC12 cells without affecting their viability and proliferative activity and that the required methyltransferase activity for neurite extension might be due chiefly to asymmetric arginine dimethylation and not to methylation of nucleotides or other amino acid residues (16
). Among the substrates for arginine methyltransferases, we noted RBPs, some of which proved to be involved in neuronal differentiation. Fortunately, one of the RBPs, HuR, was shown to be methylated in vitro and in vivo by CARM1 arginine methyltransferase (41
). HuR belongs to the Hu/ELAV RBP family and regulates ARE-mediated mRNA decay by competing with AUF decay-promoting factors for ARE-containing mRNA (38
). The alignment of the amino acid sequence around the dimethylated arginine residue of HuR displays a high sequence homology with that of the other mammalian Hu members (28
). Then we focused on a neuron-specific Hu family protein, HuD, as the methylated substrate and examined whether or how its methylation regulates the turnover of the bound mRNAs in PC12 cells.
Using an in vitro protein methylation assay, we showed that HuD is a specific substrate for CARM1. 3
H-labeled HuD was blotted with an antibody against mono- and dimethylarginine, which was used to demonstrate in vivo methylation of HuR (41
). The anti-M/DMA antibody immunoprecipitated endogenous HuD from the lysate of native PC12 cells but not that of CARM1−
cell lines that we established (Fig. ). It also precipitated exogenous HuD from HuDwt-transfected PC12 cell lysate but not that from R236K-transfected cells (Fig. ). These data demonstrate that Arg236
is a major CARM1 methylation site, both in vitro and in vivo, which corresponds to Arg217
of HuR. Since the antibody towards methylarginine residues used in this study was incapable of sufficiently identifying the style of arginine methylation, we had to determine it by mass spectrometry.
Our data argued against the previous reports on CARM1 subcellular localization. In HeLa cells and primary fibroblasts, CARM1 was shown to reside predominantly in the nucleus (24
). If this were the case with PC12 cells, CARM1 would be sequestered from HuD by the nuclear membrane. However, our DAB-stained preparation and fluorescence immunocytochemistry analysis revealed substantial CARM1 reactivity in the cytoplasm of PC12 cells (Fig. ). CARM1 cytoplasmic staining was somewhat obscure compared to the intense nuclear immunoreactivity and was left at the level reminiscent of the background staining in CARM1−
cells, which completely lost the nuclear staining. To exclude the possibility that the cytoplasmic staining is an artifact, antiserum adsorbed with GST-CARM1 was used for CARM1 immunodetection by the DAB staining method. Because the adsorbed antiserum failed to react with any epitopes in the cytoplasm or in the nucleus (Fig. ), the observed cytoplasmic staining with PC12 cells reflected CARM1 immunoreactivity. A recent study demonstrated cytoplasmic localization of CARM1 in C2C12 myoblasts (13
), indicating that CARM1 localization is cell type dependent. We further demonstrated a colocalization of CARM1 and HuD in the cytoplasm, which surrounded the nucleus and was extinguished around the cell periphery. We further investigated whether methylation affects the subcellular localization, because Arg236
is located in the hinge region that determines the subcellular localization (32
). As shown in Fig. , the Arg236
methylation level had no effect on the subcellular localization of HuD.
We next examined the effects of HuD methylation on innate function, such as nucleic acid-binding property. Since GAR domains are methylated in a variety of RBPs (18
), their methylation possibly influences the protein-RNA interactions. For example, the unmethylated form of hnRNP A1 has a higher binding capacity to single-stranded nucleic acids than the methylated form (55
). Arginine methylation of HuD may also modulate its direct interaction with ARE-containing mRNAs; otherwise, the methylation event may regulate an interaction with another protein which controls the RNA-binding interface of HuD, RRM2 and/or RRM3 (53
). We then investigated expression levels of the genes whose transcripts are regulated by HuD in CARM1−
PC12 cells. Although CARM1 knockdown had no effect on the protein and transcript levels of GAP43, tau, and p27, it exclusively upregulated those of p21cip1/waf1
(Fig. ). We confirmed that the rise in p21cip1/waf1
transcripts was due not to the enhancement of its transcription (Fig. ) but to the elongation of its half-life (Fig. ). We further demonstrated that the methylation-resistant HuD mutant R236K made a complex with p21cip1/waf1
mRNA to a greater extent than did methylated HuD (Fig. ). These results indicated that unmethylated HuD could lower the decay rate of p21cip1/waf1
transcript by forming a tighter complex with it. HuD was reported to act on a variety of ARE-containing mRNAs, but the only gene product affected by CARM1 loss was p21cip1/waf1
, according to our examination. Though ARE-containing GAP43 mRNA is stabilized by HuD modifying its poly(A) tail length (10
) and is associated with HuD in the growth cones of PC12 cells (59
), CARM1 depletion had no effect on GAP43 transcript level and decay rate (Fig. ). We inferred that RNP complex components other than HuD might be different among p21cip1/waf1
and GAP43 transcripts and that those of p21cip1/waf1
are influenced primarily by methylation of HuD. Since the methylation state of HuD could possibly affect the turnover of unknown transcripts, we attempted to make a catalogue of transcripts with differential binding characters to methylated and unmethylated HuD by use of a microarray-based method.
To investigate whether HuD is the sole mediator for producing such gene expression patterns in CARM1− cells, we introduced an excess amount of the methylation-resistant HuD mutant R236K into native PC12 cells (Fig. ). The overexpression of naked HuD remarkably induced p21cip1/waf1 protein without affecting the GAP43 protein level, demonstrating that a higher ratio of unmethylated HuD only by overexpression of methylation-defective HuD could reproduce the CARM1-defective phenotype. Though there remained the possibility that other CARM1 substrates, especially other neuron-specific Hu family proteins HuB and HuC, are relevant to the induction of p21cip1/waf1 transcripts, our data strongly suggested that the unmethylated protein HuD is sufficient to cause the same effect as CARM1 depletion. Even though HuB and HuC are similarly regulated by CARM1, there is no direct evidence of their interaction with p21cip1/waf1 transcripts.
Interestingly, methylated HuD was extinct 3 days after NGF treatment with PC12 cells, even when total HuD was kept at the steady-state level (Fig. ). This observation indicated that NGF could produce the same effect on HuD as CARM1 depletion, in keeping with our result that NGF induced p21cip1/waf1
transcript to an extent similar to that induced by CARM1 depletion (Fig. ). Note that CARM1−
cells exhibited a slow growth rate (Fig. ) and accelerated neurite extension (Fig. ), which also occurs in the NGF-treated PC12 cells; thus, it is possible that the loss of methylated HuD and the resultant rise in p21cip1/waf1
mRNA may be one of the major processes for the NGF signaling pathway. To explain how NGF decreases the methylated population of HuD with the total amount being unchanged, we assume that, once methylated, HuD is enzymatically demethylated at the onset of NGF signaling. Recently, monomethyl-arginine residues on H3 and H4 histone proteins were shown to be methyl-deiminated to citrulline by protein arginine deiminase 4 (PAD4) (19
). However, since this kind of reaction has been shown to act on monomethylated arginine, a novel system might be required for elimination or oxidization of the methyl group from arginine-methylated HuD.
Altogether, the methylated state of HuD determines the p21cip1/waf1
expression level, and the inhibited methyltransferase activity of CARM1 can arrest the growth of PC12 cells to represent a partially differentiated cell shape by propagating unmethylated HuD. Accordingly, our results are inconsistent with previous reports that protein arginine methylation is required for NGF-induced differentiation of PC12 cells (16
). In fact, these reports evaluated the whole effect of protein methylation by using methyltransferase inhibitors with broad spectra, whereas we focused on the effect of selective inhibition of CARM1. Compared with PRMT1, CARM1 activity covers a limited species of proteins (39
). This is why our results are discrepant with the general inhibition of arginine methylation that abrogated NGF-induced neuritogenesis of PC12.
In conclusion, our data demonstrated that CARM1 methylates Arg236 of HuD in vitro and in vivo and that the methylation of HuD raised the vulnerability of p21cip1/waf1 transcripts to direct PC12 cells to the proliferative state, whereas unmethylated HuD endowed the cells with differentiated phenotypes in an NGF-independent manner. Our final observation, that CARM1-expressing cells were localized in the VZ of the lateral ventricle of the adult mouse brain and overlapped with BrdU-positive cells (Fig. ), supports the notion that CARM1 keeps the cells in the proliferative state.