Previous data in the field of mRNP granule biology established that stress granules are dynamic storage facilities for mRNAs and translation factors. In this view SGs are positioned downstream of many stress detection signals that first restrict translation (Kwon et al., 2007
; Ohn et al., 2008
; Leung et al., 2011
), causing accumulation of stalled translation initiation complexes, which are then assembled into stress granules via microtubule-dependent molecular motors (Ivanov et al., 2003
; Kwon et al., 2007
). This positions SGs as a consequence of stress signaling; however, the reverse role of SGs functionally signaling out to the translation apparatus to maintain a state of translational repression had not previously been established. Although the idea of SGs preceding eIF2α phosphorylation had not been previously documented, the conceptual distinction between large and small granules had been made for both arsenite and azide stressors (Buchan et al., 2011
; Zhang et al., 2011
). In the case of arsenite stress in which samples were examined every 5 min after arsenite application, we did not see appearance of eIF2α phosphorylation before assembly of granules (unpublished data). On the contrary, eIF2α phosphorylation is prominent before even small granules are assembled.
Our data indicate that G3BP-induced SGs can signal to the translation apparatus by stimulating eIF2α phosphorylation in a PKR-dependent manner (). This model explains how granule formation can be observed in both the PKR-knockout and S51A mutant MEFs that do not have induced eIF2α phosphorylation. Although we only have evidence that phosphorylated eIF2α is present at the same time as large SGs, we predict PKR activation and eIF2α phosphorylation as a consequence of large SG assembly because eIF2α phosphorylation is not observed with small SGs (). We hypothesize that the PKR activation and eIF2α phosphorylation may be involved in maintenance of SGs during other stresses after an initial phase of eIF2α phosphorylation by other kinases (e.g., HRI activation during arsenite stress).
Some points are not resolved within this model for induction of eIF2α phosphorylation (). For example, it is not known what is being sensed in the cell that triggers the coalescence of small granules into large granules. We conjecture that the depletion of initiation factors or RNA-binding proteins into small granules is sensed because it is unlikely that substantial protein synthesis can occur in the absence of accessible translation initiation factors. Sensing of small granules may also depend on localization of sufficient carrier proteins that allows high-affinity interaction between the small granule and molecular motors (Loschi et al., 2009
). This would then permit coalescence of the small granule and initiation of subsequent signaling. Another point that is unresolved is how PKR is activated after large-granule assembly. We hypothesize that compaction of RNA in a granule that is sensed by PKR could result in eIF2α phosphorylation. Alternatively, derepression of PACT/Rax may occur, which is the only cellular activator of PKR, which in turn may phosphorylate PKR in the absence of double-stranded RNA (Patel and Sen, 1998
; Ito et al., 1999
; Garcia et al., 2006
). Finally, stress granules could activate a signaling molecule upstream of PKR by such as MyD88 or IRAK1 (i.e., toll-like receptor 3, 4, or 9 signaling; Garcia et al., 2006
Our finding that PKR is responsible for eIF2α phosphorylation is supported by earlier stress granule work in which Tia1 was overexpressed (Kedersha et al., 1999
). Kedersha and colleagues showed that expression of either the nonphosphorylatable S51A mutant of eIF2α or the adenoviral VAI gene, which inhibits PKR activity, was sufficient to reduce induction of SGs by Tia1 overexpression. However, they concluded that overexpression of transgenes introduces so much exogenous RNA that PKR is activated and eIF2α is phosphorylated, resulting in SGs (Kedersha et al., 1999
). Our data provide new insight that forces us to reexamine these conclusions. Specifically, expression of G3BP in MEFs expressing the nonphosphorylatable S51A mutant as the sole source of eIF2α still induced stress granules. Furthermore, expression of GFP-λN, PABP, and eIF4G do not induce stress granules despite similar expression from the cytomegalovirus (CMV) transcription promoter.
The observation that with high transgene expression eIF2α phosphorylation was only observed with large granules also argues against activation of PKR by exogenous RNA. Finally, we tested many deletion mutants of G3BP that do not induce eIF2α phosphorylation when expressed from a CMV promoter, indicating that a specific sequence in the exogenous RNA is not responsible for PKR activation and resulting SGs (unpublished data).
Based on the prominent role of SGs and PKR in antiviral innate immunity, the finding that PKR mediates eIF2α phosphorylation coinciding with large G3BP-induced SG assembly makes sense. Many viruses target SGs for disassembly and encode proteins that are known inhibitors of PKR activity (Garcia et al., 2006
; White and Lloyd, 2012
). Of interest, PKR participates in several toll-like receptor signaling pathways, including toll-like receptors 3, 4, and 9 (Horng et al., 2001
; Jiang et al., 2003
). PKR activity also regulates NF-κB transcriptional activity (Kumar et al., 1997
; Gil et al., 2001
) and can induce c-Jun N-terminal kinase (JNK) activity (Goh et al., 2000
; Taghavi and Samuel, 2012
). Therefore it is reasonable that these proteins may be activated by assembly of large G3BP-induced granules.
JNK signaling has recently become a focal point in mRNP granule biology. Arimoto et al. (2008
) documented that Rack1, an activator of JNK signaling and subsequently apoptosis, is sequestered in arsenite-induced SGs, thereby preventing JNK activity. Rack1 is also sequestered in G3BP-induced stress granules, and G3BP inhibits activation of MTK1, an upstream kinase important for JNK activation. Active JNK has also been shown to be recruited to arsenite and heat-induced SGs along with a scaffolding molecule WDR62 (Wasserman et al., 2010
). WDR62 and active JNK recruitment was not observed in G3BP-induced granules, whereas both Tia1 and TTP-induced granules colocalize with active JNK. It will be interesting for future work to examine Rack1, WDR62, and JNK activation in cells with small versus large G3BP-induced granules to determine whether differing localization and intensity exists. Phosphorylation of the decapping regulator Dcp1a by JNK has also been shown to reduce inclusion of Dcp1 in P-bodies (Rzeczkowski et al., 2011
). Because JNK is known to act downstream of PKR to mediate innate immune responses (Garcia et al., 2006
), the PKR–JNK signaling axis is an intriguing candidate pathway for our future work because it may be influenced by signaling from stress granules. Our finding that PKR is activated in cells with large G3BP-induced granules introduces a new signaling component into area of SG-dependent signaling that may help to clarify the role of JNK in mRNP biology. These questions are of interest for our future work.