It has long been recognized that certain RNA-BPs undergo nuclear/cytoplasmic shuttling and are capable of localizing in discrete intracellular compartments. Based on the seminal work of Anderson and others [42
], it is also well documented that certain RNA-BPs can colocalize. However, colocalization, as identified by microscopy and/or by co-immunoprecipitation, can be limited in interpretation. More precisely, colocalization identified by conventional and/or confocal microscopy can be vague and does not necessarily permit the conclusion that two proteins that appear to colocalize are actually in sufficiently close proximity to physically or functionally interact. This is also true for immunoprecipitation experiments that generally do not necessarily distinguish between direct and indirect protein/protein interaction. Therefore, one objective of the current work was to identify the intracellular localization and potential interactions of several key RNA binding proteins, both endogenous and exogenous, using FRET. Although a positive FRET signal does not guarantee that proteins are interacting functionally, it does provide strong evidence that two (or more) proteins are in sufficiently close proximity to physically interact.
Although immuno-FRET of ARE binding proteins is being described herein for the first time, there is certainly ample precedent demonstrating the usefulness and veracity of this methodology [31
]. A major consideration in analyzing immuno-FRET data generated with fluorescently labeled antibodies (primary or secondary), is an extraordinary attention to detail necessary to obviate issues related to background signal and non-specific fluorescence. For example, certain antibodies (and cell types) demonstrate high levels of background or non-specific fluorescence that can constitute a major limitation. Another obvious limitation of immuno-FRET is the increased probability of a false-negative signal due to the position and/or orientation of one secondary antibody fluorophor relative to the other. As with any FRET measurement, the acquired raw fluorescence data must be corrected by subtracting the appropriate channel bleed-through determined specifically for the experimental system being used.
In previous investigations [15
], we demonstrated that p37AUF1/hnRNP D and HuR, were colocalized in both nucleus and cytoplasm and physically interacted, as detected by FRET. Additionally, and in support of existing biochemical data [63
], we demonstrated that p37AUF1 and HuR can both homo and heterodimerize [15
]. The current study extends the identification of heterologous permutated interactions to include those between AUF1, HuR, KSRP, TIA-1, and Hedls, proteins variably associated with increased RNA turnover, RNA stabilization, and translational suppression. Our current findings demonstrate that both endogenous and exogenous, fluorescently-tagged AUF1, HuR, TIA-1 and KSRP are dominantly nuclear, Figures and , and that each protein pair, AUF1/HuR, KSRP/AUF1, TIA-1/HuR, KSRP/TIA-1, AUF1/TIA-1, and KSRP/HuR, demonstrates nuclear FRET. This is in contrast to the lack of FRET between each of these ARE-BPs and β-tubulin that was used as a negative control for immuno-FRET (data not shown).
Induction of oxidative stress with arsenite causes proteins to aggregate in SGs to varying degrees with further distinctions between endogenous and exogenous proteins. Specifically, arsenite-induced stress caused rapid formation of cytoplasmic SGs with the presence of endogenous HuR, AUF1, KSRP and TIA-1 being readily detected. However, robust FRET signals in SGs were limited to the interactions between TIA-1/HuR and TIA-1/KSRP and to a lesser extent, TIA-1/AUF1, HuR/AUF1, KSRP/AUF1 and KSRP/HuR. Interestingly, and in contrast to their endogenous counterparts, the exogenous pair of TIA-1-YFP/AUF1-CFP does not demonstrate a readily detectable FRET signal.
The presence and interaction all four ARE-BPs in SGs may seem anomalous given that they have putatively different roles when bound to an ARE. When preceding experiments failed to detect AUF1 in SGs, we initially reasoned that this result was consistent with the biology of AUF1 generally being considered an mRNA destabilizing protein; hence, it might not be expected to be bound to an RNA being sequestered and stabilized within an SG. This reasoning could be extended to KSRP. However, in biochemical studies, we have shown that AUF1 and HuR are able to bind to each other and to co-reside on the β2
-adrenergic receptor ARE [15
]. Thus, the notion that a protein, based on putative function, would or would not localized to a compartment with a putative antithetical function does not appear to be valid. For this to be true, rapid movement of an mRNA molecule to SGs, upon the initiation of cellular stressor, would require a screening mechanism to be present upon 'entrance' to the SG such that it would remove all 'destabilizing' protein entities. This does not appear to be occurring as the formation of SG and hence removal of all non-essential mRNA from active translation in the cytoplasm during a stress response appears to take precedence.
Potentially adding to the uncertainty of ARE-BP localization and interactions is the discrepancy, upon cellular activation, between endogenous and exogenous proteins. As our studies show, this appears to be the case for KSRP. This particular result underscores one of the potential problems associated with transient transfection of exogenous proteins and the advantage, when possible, of visualizing endogenous proteins by immuno-FRET.
An unanticipated finding was the strong FRET signal between endogenous TIA-1 and Hedls, Figure . The FRET signal between TIA-1 and Hedls seen when the proteins are present in juxtaposed SGs and PBs respectively, is supported by the work of Anderson, Kedersha and colleagues, demonstrating that certain constituent RNA binding proteins and other compositional constituents of SGs and PBs can undergo dynamic interchange [38
]. Potentially underlying the interchange of SG and PB components is translationally silenced RNA and its associated RNP complexes, some of which appear to be present in both PBs and SGs.
Studies indicate that mRNAs cycle between PBs and SGs, depending on the availability of translation initiation or degradation machinery. A working model, called the mRNA cycle, was proposed by Balagopal and Parker [66
]. In this model, transcribed mRNAs initially undergo subsequent rounds of translational initiation, elongation and termination, producing polypeptides until a change in the cellular environment alters this cycle directing the mRNAs away from ribosomes and towards SGs and PBs. What is undoubtedly a complex and as yet unknown pathway is that determining whether a stalled mRNA will differentially populate PBs or SGs.
As RNAs are dynamic between PBs and SGs so then too are the proteins associated with them. The complete proteomic composition of PBs and SGs is still being elucidated, however, a number of proteins have been identified and localized to one or both of these cytoplasmic granules [67
]. Relevant to our work, TIA-1 (and TIAR) nucleates SGs but is not found in PBs; in contrast, Hedls and Dcp1 are found in PBs but not in SGs. More RNA binding proteins have been found to be common to both SGs and PBs than exclusively present in one or the other. This might indicate that the proteins common to SGs and PBs play a more general role in the maintenance/structure of an mRNA molecule while present in a particular granule and those RNA BPs specific to SGs or PBs have a role in the fate of the mRNA in the granule. Movement of mRNPs from polysomes to SGs or PBs may be attributed to different mRNP conformational states due to rearrangement, presence or exchange of ARE-BPs on specific RNAs so our finding that all ARE-BP studied are present in SGs may, in reality, have not been so unexpected.
A somewhat disappointing finding was the more or less general, diffuse effect of MAP kinase stimulation on the relocalization of RNA binding proteins, Figures and . As demonstrated previously, anisomycin causes a rapid shuttling of HuR from nucleus to cytoplasm, a result consistent with the observation that MAPK activation is widely recognized to cause stabilization of ARE containing mRNAs [55
]. In contrast, AUF1 and KSRP shuttle from nucleus to cytoplasm with a considerably longer kinetic. Unlike treatment with arsenite, MAPK activation does not cause SG formation. Instead, the cytosolic presence of HuR, AUF1, and KSRP are detectably increased but remain diffuse with a peri-nuclear to cell periphery decreasing expression gradient. As in the nucleus, each of these pairs of proteins exhibits FRET with the relative intensity decreasing along their concentration gradients. Consistent with arsenite studies, MAPK activation does not produce a visibly detectable interaction between Hedls and either KSRP or HuR.
There is also biochemical evidence in the literature that MAPK activation causes the relocalization of RNA and associated RNA binding proteins in a phosphorylation-dependent manner [13
]. For example, in the case of HuR, several kinases appear to regulate its intracellular localization. Kim, et. al. [74
], have described nuclear retention of HuR promoted by Cdk1 mediated phosphorylation (S202). HuR is also phosphorylated at a number of other residues: S88, S100, and T118 by checkpoint kinase 2 (Chk2) [23
], at S158 by PKCα, at S221 and S318 by PKCδ [25
] and at S242 by an as yet unidentified kinase [74
]. The location of the modified residues are either in or between two of the three RNA recognition motifs, (RRM 1 and 2), or within the nucleocytoplasmic shuttling sequence (205-237) of HuR [4
]. The effect of phosphorylation at S202 and S242 is nuclear retention whereas phosphorylation of S158, S221, and S318 leads to nuclear export of HuR. Phosphorylation at S88, S100 and T118 does not appear to affect the localization of HuR. In addition, phosphorylation at S88, T118 [23
], S158, S221 [70
] or S318 [25
] was shown to enhance the binding of HuR to target AREs with S100 phosphorylation showing the opposite affect [23
] and S221 phosphorylation no effect [25
] on HuR binding. HuR is also modified by methylation at R217 by CARM1, a change that was implicated in cytoplasmic shuttling of HuR in response to lipopolysaccharide [76
]. The phosphorylation of S202 lead to nuclear retention of HuR by allowing binding of 14-3-3θ to this phospho-form of this protein [74
In addition to HuR, phosphorylated KSRP [26
], TTP [77
], AUF1 [78
] and BRF1 [79
], have all been reported to bind a member of the 14-3-3 protein family in either the nucleus or cytoplasm. This may be a general method of alteration of ARE binding protein function and localization and is consistent with the cellular functionality of 14-3-3 proteins, in general [80
In a related experimental direction, we attempted to address a long-standing interest, that of colocalization of ARE-containing mRNAs and RNA binding proteins. To perform these experiments, in vitro
transcribed, capped, polyadenylated, chemically stabilized (2'F) fluorescently labeled (Cy3-UTP) RNA was transfected into DDT1-MF2 cells. Two outcomes are noted. First, that the 2'F/Cy3 RNA is remarkably stable demonstrating virtually no change in abundance at 24 hrs post transfection. Secondly, the transfection of RNA forms distinct cytoplasmic aggregates. This finding is entirely consistent with that of Barreau et al [81
], where aggregates of a similar description were noted. However, these authors conclude that the aggregates may be secondary to a stress response simply due to the presence of excess RNA. Our finding is also similar to that of Wilkie and Davis [82
], who described, in Drosophila
, the localization and trafficking of capped, fluorescently labeled run
Going beyond this observation, we used immuno-detection of the decapping proteins, Dcp1 and Hedls, to demonstrate that the aggregated RNA and the decapping enzymes were colocalized. In fact, we were able to observe robust FRET between Cy3-labeled β-AR mRNA and Dcp1. These results lead to a couple of conclusions. First, that interaction of labeled RNA and protein can be observed, a potentially highly useful proof of principal. Secondly, that transfection of in vitro transcribed, chemically stabilized RNA is probably trapped in PB due to its inability to undergo appropriate degradation. In contrast, FRET was not observed between labeled RNA and the ARE binding protein, HuR. This is perhaps not surprising given the absence of HuR in PBs.
At least two other papers have been published describing FRET between RNA and RNA binding proteins. Lorenz [83
] infers FRET by a reduction in fluorescence lifetime (FRET-FLIM) secondary to the interactions of both Venus-PTB and YFP-Raver1 proteins and generic SytoxOrange labeling of RNA. Huranova, et. al. [84
], describe FRET-FLIM between eCFP-tagged hnRNP H bound to an engineered high-affinity consensus binding site within the RNA and eYFP-tagged MS2 coat protein bound to its cognate stem-loop binding site. The advance to the field from the data presented here in, is the demonstration that a specific labeled RNA molecule can interact with a specific RNA binding protein and that their interaction can be detected within a specific intracellular compartment. Thus, alternative approaches, each with their relative strengths and weaknesses are increasingly becoming available.