Subcellular localization of mRNAs is regulated by RNA–protein interactions. Here, we show that introduction of a reporter mRNA with the 3′UTR of β‐actin mRNA competes with endogenous mRNAs for binding to ZBP1 in adult sensory neurons. ZBP1 is needed for axonal localization of β‐actin mRNA, and introducing GFP with the 3′UTR of β‐actin mRNA depletes axons of endogenous β‐actin and GAP‐43 mRNAs and attenuates both in vitro and in vivo regrowth of severed axons. Consistent with limited levels of ZBP1 protein in adult neurons, mice heterozygous for the ZBP1 gene are haploinsufficient for axonal transport of β‐actin and GAP‐43 mRNAs and for regeneration of peripheral nerve. Exogenous ZBP1 can rescue the RNA transport deficits, but the axonal growth deficit is only rescued if the transported mRNAs are locally translated. These data support a direct role for ZBP1 in transport and translation of mRNA cargos in axonal regeneration in vitro and in vivo.
There is a Have you seen? (November 2011) associated with this Article.
Localized protein synthesis is a mechanism used by many, if not all polarized cells to modulate protein composition in subcellular domains. Locally generated proteins play a role in cellular migration, adherence, polarization, and metabolism (Holt and Bullock, 2009). For neurons, early ultrastructural studies suggested that only dendrites have protein synthetic capacity (Steward and Levy, 1982). However, it is now clear that neurons can also translate mRNAs in their axons (Donnelly et al, 2010). The axonally synthesized proteins appear to fulfil several different functions. For example, proteins synthesized in the axon, such as β‐actin, can be used for growth cone turning (Leung et al, 2006; Yao et al, 2006; Lin and Holt, 2008), and locally translated Par3, a protein involved in cell polarization, supports axon elongation (Hengst et al, 2009). Injury‐induced synthesis of Importin β1 and RanBP1 in axons allows these distal processes to communicate with their cell bodies (Hanz et al, 2003; Perlson et al, 2005; Yudin et al, 2008), and other locally synthesized proteins contribute to growth cone formation after axotomy (Verma et al, 2005).
Based on mRNA profiles of neuronal processes and fibroblasts’ pseudopodia, it is clear that neither all mRNAs are targeted for distal transport nor are only the most abundant mRNAs transported (Willis et al, 2007; Mili et al, 2008; Vogelaar et al, 2009; Andreassi et al, 2010). For example, β‐actin mRNA is transported into axons and dendrites, while the near equally abundant γ‐actin mRNA remains in the cell body (Bassell et al, 1998; Zheng et al, 2001; Tiruchinapalli et al, 2003). Delivery of mRNAs to subcellular sites requires mRNA–protein interactions. The 3′UTR of β‐actin mRNA contains a conserved ‘zipcode’ that is essential for its localization in cultured cells through interaction with the zipcode binding protein 1 (ZBP1; also known as the insulin‐like growth factor II mRNA binding protein 1 (IMP1)) (Kislauskis et al, 1993, 1994; Farina et al, 2003). Despite evidence for locally generated β‐actin's role in axonal guidance in vitro (Zhang et al, 2001; Yao et al, 2006; Sasaki et al, 2010), only a small fraction of its mRNA pool localizes to axons (Eng et al, 1999). While there is clearly enrichment of some neuronal mRNAs in neuronal processes (Poon et al, 2006; Andreassi et al, 2010), many transcripts behave similar to β‐actin showing only a small localizing fraction. The mechanisms underlying how much of a given mRNA localizes to axons and whether those low levels can influence axonal growth have not been investigated. Moreover, it is not clear if any RNA binding proteins (RBPs) are needed to drive axonal mRNA localization and influence axon growth in vivo. Here, we have used neuronal β‐actin mRNA as a tool to ask if limited availability of the trans‐acting protein machinery needed for its localization could account for the limited levels of this mRNA in axons. We show that ZBP1 is a critical determinant of how much β‐actin and GAP‐43 mRNAs localize to axons. This in turn contributes to the axons’ growth potential both in vitro and in vivo.
Exogenous 3′UTR of β‐actin decreases axonal levels of endogenous β‐actin mRNA
To determine if the machinery needed to transport β‐actin into sensory axons is limiting, we first used adenoviral (AV) preparations for expressing a myristoylated, destabilized GFP (dzGFPmyr) with the 3′UTR of β‐actin or γ‐actin (AV‐GFP–3′β‐actin and AV‐GFP–3′γ‐actin, respectively). Since the 3′UTR of γ‐actin has no localizing activity in dorsal root ganglion (DRG) neurons (Willis et al, 2007), the AV‐GFP–3′γ‐actin provided a rigorous control for any non‐specific effects of virus transduction. Fluorescent in situ hybridization (FISH) probes to the coding region of actin allowed us to distinguish endogenous β‐actin mRNA in neuronal processes from the GFP+β‐actin 3′UTR (GFP–3′β‐actin) mRNA (Supplementary Figure S1A). We refer to the processes of cultured DRG neurons as ‘axons’ throughout the remainder of the text since MAP2 mRNA and protein are excluded from these processes (Zheng et al, 2001). These DRG processes also show enrichment of an axonally localizing kinesin isoform by live cell imaging (Vuppalanchi et al, 2010) and microtubule polarity typical of axons (Baas et al, 1987).
The AV‐GFP–3′β‐actin transduced neurons showed a significant reduction in endogenous axonal β‐actin mRNA signals (Figure 1A and B). There was no corresponding difference in total endogenous β‐actin mRNA levels between AV‐transduced and control cultures by standard and quantitative RT–PCR; the mRNA levels from different AV‐derived GFP constructs were also comparable (Supplementary Figure S1B). β‐Actin is also expressed by the Schwann cells (Zheng et al, 2001) and this could mask any changes in β‐actin in the DRG neurons in these RT–PCR analyses. Thus, we used quantitative FISH so that we could assess the cell body actin mRNA levels; the AV‐GFP–3′β‐actin transduced neurons showed a slight increase in actin mRNA levels in the cell body but this did not reach statistical significance (Figure 1C). Thus, the exogenous β‐actin 3′UTR can compete with endogenous mRNAs for the localization machinery.
Depletion of axonal β‐actin mRNA decreases axonal outgrowth from adult sensory neurons
Antisense oligonucleotides to the zipcode region of β‐actin have been shown to prevent its localization, alter growth cone dynamics in neurons, and alter the polarity of migration in fibroblasts (Shestakova et al, 2001; Zhang et al, 2001). To determine if the depletion of endogenous β‐actin mRNA from the axons of DRG neurons alters their growth, we analysed the morphology of the AV‐transduced cultures. DRGs expressing GFP–3′β‐actin showed significantly fewer and shorter axons, less branched axons, and smaller growth cones than control or GFP+γ‐actin 3′UTR (GFP–3′γ‐actin) mRNA expressing DRGs (Figure 2A and B; Supplementary Table S1). Axon length also decreased as cultures were exposed to increasing amounts of AV‐GFP–3′β‐actin, but not to AV‐GFP–3′γ‐actin (Supplementary Figure S2). Thus, β‐actin's 3′UTR can function as a dominant‐negative agent by preventing endogenous β‐actin localization and attenuating axon outgrowth in these adult DRG neurons.
GFP–3′β‐actin transgene expression decreases axon outgrowth in vitro and nerve regeneration in vivo
To determine if introducing GFP–3′β‐actin mRNA into DRG neurons might also impair axonal RNA localization and regeneration in vivo, we examined DRG neurons from mice expressing GFP–3′β‐actin or GFP–3′γ‐actin transgenes under control of the neuronal‐specific Tα1 tubulin promoter (Tα1‐GFP–3′β‐actin and Tα1–GFP–3′γ‐actin, respectively). The Tα1‐GFP–3′β‐actin mice show axonal localization of GFP mRNA in sciatic nerves but the GFP mRNA does not localize to axons in the Tα1‐GFP–3′γ‐actin mice (Willis et al, 2011). We initially tested in vitro axonal growth after an in vivo conditioning sciatic nerve crush lesion in these mice. In contrast to the naive cultures used in Figure 2A and B, here we used injury‐conditioned DRG neurons that show a characteristically rapid and translation‐dependent axonal outgrowth during the first 24 h in vitro (Smith and Skene, 1997; Twiss et al, 2000). This translation‐dependent growth also includes robust intra‐axonal protein synthesis (Willis et al, 2005). The injury‐conditioned L4‐5 DRGs from the Tα1‐GFP–3′β‐actin mice showed decreased axon outgrowth at 1 day in vitro (DIV) compared with DRGs from the Tα1‐GFP–3′γ‐actin mice (Figure 2C and D). Moreover, DRGs from the four copy Tα1‐GFP–3′β‐actin mice were also significantly shorter than those from the two copy Tα1‐GFP–3′β‐actin mice, suggesting a dosage effect for the transgene. Note that naive DRGs cultured from these mice expressing GFP–3′β‐actin mRNA also showed reduced axonal growth over 3 DIV (see Figure 4B below).
To determine if expression of GFP–3′β‐actin mRNA might also alter in vivo axonal regeneration, we tested the ability of the transgenic mice to regenerate transected axons by grafting a wild‐type nerve segment into their common fibular nerve (English et al, 2005). The Tα1‐GFP–3′β‐actin mice showed significantly decreased regeneration into the nerve graft compared with the Tα1‐GFP–3′γ‐actin mice (Figure 3A). Since axonal regeneration is more robust after crush compared with transection injury, we asked if expression of GFP–3′β‐actin might also affect regeneration of crushed peripheral nerve. The Tα1‐GFP–3′β‐actin mice again showed significantly fewer axons at points distal from the crush site than the Tα1‐GFP–3′γ‐actin mice (Figure 3B and C). We saw similar decreased regeneration in the Tα1‐GFP–3′‐β‐actin mice when analysed by a threshold method with multiple axonal markers (Supplementary Figure S3A). To be certain that these differences were not due to lower levels of axonal proteins in the Tα1‐GFP–3′β‐actin mice, we determined the ratio of axon diameters to total diameters of the myelinated fibres (i.e., ‘G‐ratio’) in both transgenic lines. G‐ratios in the crushed nerves for the Tα1‐GFP–3′β‐actin mice were different than those for the Tα1‐GFP–3′γ‐actin mice (Figure 3D and E). Since the diameter and G‐ratio of sensory axons is overall less than motor axons, the right shift in the G‐ratio histogram for the Tα1‐GFP–3′β‐actin mice towards larger axon diameters is consistent with decreased regeneration in the sensory neurons that selectively express the GFP–3′β‐actin transgene. Finally, to determine if differences in axon retraction accounted for the decreased regeneration in the Tα1‐GFP–3′β‐actin mice, we directly compared retraction after severing axons in cultured DRGs. There were no significant differences in distances of axon retraction in DRGs from Tα1‐GFP–3′β‐actin and Tα1‐GFP–3′γ‐actin mice (Supplementary Figure S3B). We also compared neuron numbers in the L4‐5 DRGs at 0 and 14 days after crush injury with no differences between the Tα1‐GFP–3′β‐actin and Tα1‐GFP–3′γ‐actin mice (data not shown).
Considering that developmental abnormalities in the Tα1‐GFP–3′β‐actin mice could lead to the appearance of decreased nerve regeneration in the above studies, we analysed several parameters of sensory neuron development in these mice. Nerve morphology was assessed to determine if there is a selective loss of sensory fibres in the Tα1‐GFP–3′β‐actin mice. The G‐ratios and axon diameters showed no significant differences in uninjured Tα1‐GFP–3′β‐actin and Tα1‐GFP–3′γ‐actin sciatic nerves (Supplementary Table S2). Unmyelinated axons showed similar density in the Tα1‐GFP–3′β‐actin and Tα1‐GFP–3′γ‐actin mice (data not shown). The lines showed no significant differences in numbers of neurons in the L4‐5 DRGs and footpads showed similar densities of sensory fibres (Supplementary Table S2). Finally, the transgenic lines showed no statistical differences in temperature sensitivity or mechanosensation (data not shown). Thus, developmental differences between the Tα1‐GFP–3′β‐actin and Tα1‐GFP–3′γ‐actin mice do not account for the differences in their axon regeneration.
Limited levels of ZBP1 account for the dominant‐negative effect of GFP–3′β‐actin mRNA
Since the GFP–3′β‐actin mRNA could be competing for limiting quantities of ZBP1 or other components of the RNA transport machinery, we asked if increasing availability of ZBP1 might rescue the growth deficit in DRG neurons cultured from the Tα1‐GFP–3′β‐actin mice. Consistent with previous work in cortical neurons (Zhang et al, 2001), the ZBP1 fusion protein appeared granular compared with native mCherry, suggesting that it aggregates into ribonucleoprotein particles in these DRG neurons (Figure 4A). At 3 DIV for expression of transfected constructs, the ZBP1 transfection completely reversed the growth deficit in Tα1‐GFP–3′β‐actin DRGs, with axon lengths comparable to mock‐ and mCherry‐transfected Tα1‐GFP–3′γ‐actin DRGs (P=0.238; Figure 4B). Tα1‐GFP–3′β‐actin DRGs showed no rescue in axon length with expression of ZBP1 mutants lacking the KH domains needed for RNA binding or tyrosine phosphorylation site (Y396) needed for its translation derepressing activity (Figure 4B). ZBP1 represses translation of its mRNA cargo and, upon phosphorylation at Y396, it releases its cargo for translation (Huttelmaier et al, 2005). Thus, the ZBP1Y396F mutant transports mRNAs but cannot release the mRNAs for translation. Together, these data suggest that ZBP1 cargo mRNAs need both localization to and translation within axons to optimally support axon outgrowth. Interestingly, DRG neurons cultured from Tα1‐GFP–3′γ‐actin mice also showed a significant increase in their axon lengths when transfected with ZBP1 (Figure 4B), which suggests that availability of ZBP1 is limited in adult sensory neurons. Consistent with this, transfection of DRG neurons from adult rats with full‐length ZBP1 generated a clear increase in axonal outgrowth (Figure 4C).
The human ZBP1 orthologue appears to bind many different mRNA species in co‐immunoprecipitation assays (Jonson et al, 2007). To determine if the growth deficit in the GFP–3′β‐actin expressing neurons is solely from decreased axonal β‐actin translation, we reasoned that introducing the β‐actin coding sequence plus its 3′UTR should compete with GFP–3′β‐actin mRNA. Transfecting neurons from Tα1‐GFP–3′β‐actin mice with β‐actin coding sequence fused to mCherry±β‐actin 3′UTR did not rescue the axon outgrowth deficit despite that the encoded β‐actin–mCherry was clearly incorporated into the cytoskeleton (Supplementary Figure S4). Thus, loss of other mRNAs from the DRG axons in addition to β‐actin is responsible for the axon growth deficit in the GFP–3′β‐actin expressing neurons. This is consistent with recent observations that motor axons can develop and regenerate in complete absence of murine β‐actin (Cheever et al, 2011).
ZBP1 is needed for axonal localization of GAP‐43 mRNA
To determine if other mRNAs might show altered transport with GFP–3′β‐actin expression, we quantified axonal levels of several axonal mRNAs in DRG neurons transduced with AV‐GFP–3′β‐actin versus AV‐GFP–3′γ‐actin. Axonal levels of GAP‐43 and Neuritin/cpg15 mRNAs showed significant reduction in AV‐GFP–3′β‐actin compared with AV‐GFP–3′γ‐actin transduced cultures (Figure 5A). Axonal amphoterin mRNA levels were not affected by AV‐GFP–3′β‐actin transduction, while axonal calreticulin mRNA levels were increased by transduction with either AV preparation (Figure 5A). By RT–qPCR, transduction with AV‐GFP–3′β‐actin did not affect overall levels of any of these transcripts studied (Supplementary Figure S1B), but FISH analyses showed increased levels GAP‐43 and Neuritin/cpg15 mRNAs in the cell bodies of the AV‐GFP–3′β‐actin transduced cultures (Figure 5A). Cell body levels of calreticulin mRNA were increased with both AV preparations (Figure 5A), suggesting that the increases in axons reflect an overall increase in neuronal calreticulin mRNA, possibly as a stress response to the viral transduction.
We next asked if transport of these mRNAs into the peripheral nerve might be affected in the transgenic mice. After sciatic nerve crush to activate the Tα1 promoter, less GAP‐43 mRNA was detected in Tα1‐GFP–3′β‐actin compared with Tα1‐GFP–3′γ‐actin nerves, and GAP‐43 levels were also lower in the four copy versus two copy Tα1‐GFP–3′β‐actin nerves (Figure 5B; Supplementary Table S3). Levels of peripherin, which encodes a neuronal‐specific intermediate filament that localizes to DRG axons (Willis et al, 2005), were not affected and Neuritin/cpg15 levels were decreased but this did not reach significance. These data suggest that the exogenous 3′UTR of β‐actin may also compete in vivo with GAP‐43 mRNA. PCR products for actin showed a significant decrease in the Tα1‐GFP–3′β‐actin compared with Tα1‐GFP–3′γ‐actin nerve (Figure 5B; Supplementary Table S3).
To directly test whether limited availability of ZBP1 determines the axonal levels of β‐actin and GAP‐43 mRNAs, we transfected wild‐type rat DRG neurons with ZBP1–mCherry and assessed axonal mRNA levels. The ZBP–mCherry expression significantly increased axonal levels of both β‐actin and GAP‐43 mRNAs (Figure 5C). Similar was seen with transfection of wild‐type mouse DRGs with ZBP1 (see 7C below). We further asked if GFP–3′β‐actin displaces endogenous mRNAs from ZBP1 in vivo. Immunoprecipitation of ZBP1 from lysates of neonatal brain plus spinal cord showed displacement of β‐actin and GAP‐43 from ZBP1 immunocomplexes in the Tα1‐GFP–3′β‐actin but not in the Tα1‐GFP–3′γ‐actin tissues (Figure 5D). RT–qPCR for amphoterin, calreticulin, and neuritin/cpg15 gave high cycle threshold values (Ct⩾35); Ct values in this range are generally indicative of noise in the RT–qPCR assay and suggest that these transcripts were not present in these immunoprecipitates. The levels of these transcripts, as well as actin and GAP‐43, were equivalent in the input lysates from Tα1‐GFP–3′β‐actin and Tα1‐GFP–3′γ‐actin tissues, and ZBP1 was equally expressed in these tissues (data not shown).
ZBP1 levels limit regeneration capacity in the adult nervous system
Previous studies have shown that ZBP1 is expressed at much higher levels in embryonic compared with adult tissues (Nielsen et al, 1999). Consistent with this, ZBP1 mRNA in adult DRGs was at lower levels than seen in the neonatal tissues, and immunoblotting showed much higher ZBP1 protein levels in lysates from E12.5 compared with those from P20 or adult DRG plus spinal cord (Figure 6A and B).
Since these low levels of ZBP1 appeared to limit axonal growth in the adult DRG neurons, we took advantage of previously published ZBP1 knockout mice to determine if further loss of ZBP1 would similarly decrease axonal regeneration potential. Although multiple developmental abnormalities are seen in ZBP1−/− mice, ZBP1+/− mice are viable and reportedly normal (Hansen et al, 2004), but ability of ZBP1+/− mice to respond to injury has not been tested. By RT–qPCR for ZBP1, we see that the adult DRGs of ZBP1+/− mice have only 40±9% ZBP1 mRNA levels of ZBP+/+ littermates (data not shown). Similarly to the Tα1‐GFP–3′β‐actin mice, DRG neurons cultured from adult ZBP1+/− mice showed decreased axonal outgrowth (Figure 6C and D). Transfecting the ZBP1+/− DRGs with ZBP1–mCherry rescued the axonal growth deficit at 3 DIV, but the ZBP1Y396F–mCherry mutant did not (Figure 6E; Supplementary Figure S5). ZBP1–mCherry expression increased and ZBP1Y396F–mCherry decreased overall axon lengths in the wild‐type mouse neurons.
The ZBP1+/− mice also showed decreased axonal regeneration in vivo. With sciatic nerve crush, there was a significant decrease in axon extension beyond the injury site at 7 days post‐injury (Figure 6F). Similar was seen after common fibular nerve transection/grafting (Supplementary Figure S3C). To determine if these morphological changes also resulted in decreased functional regeneration, we measured evoked responses (M response) of gastrocnemius and tibialis anterior (TA) muscles to simulation with an electrode inserted into the proximal sciatic nerve. Prior to injury, the M responses of ZBP1+/− and wild‐type mice were comparable; however, the ZBP1+/− showed a clear decrease in M responses compared with wild type at 14 days after nerve crush (Figure 6G). Quantifying the M responses showed statistically decreased amplitude in ZBP1+/− compared with wild‐type mice (Figure 6H). Taken together, these studies indicate loss of a ZBP1 allele results in haploinsufficiency for axonal regeneration from adult neurons.
ZBP1+/− neurons are haploinsufficient for axonal mRNA transport
Since axonal growth is decreased in the ZBP1+/− neurons and this is rescued by transfection with ZBP1, we asked if axonal mRNA transport would also be affected with loss of a ZBP1 allele. We first examined RNA levels in the nerve transection/graft model. There were no significant differences in the levels actin, amphoterin, calreticulin, neuritin/cpg15, and peripherin mRNAs in the L4‐5 DRGs of ZBP1+/− versus wild‐type mice (Figure 7A; Supplementary Table S4). However, actin and GAP‐43 mRNA levels were reduced in the proximal nerve stump and the grafted nerve tissue from the ZBP1+/− versus ZBP+/+ littermates. Neuritin/cpg15 showed a significant decrease in the proximal nerve but not in the graft and amphoterin, calreticulin, and peripherin mRNA levels were not significantly different in the proximal nerve and graft of ZBP1+/− versus ZBP+/+ (Figure 7A; Supplementary Table S4). This suggests that axonal transport of β‐actin and GAP‐43 mRNAs decreases in vivo with decreases in ZBP1 levels. To further test this possibility, we evaluated these mRNAs in acellular allografts from mice expressing the GFP–3′β‐actin mRNA that competes with endogenous mRNAs for binding to ZBP1. Actin and GAP‐43 mRNAs also showed significantly decreased levels in allografts of Tα1‐GFP–3′β‐actin compared with Tα1‐GFP–3′γ‐actin mice (Figure 7B).
Using quantitative FISH, we analysed axonal levels of mRNAs in DRG neurons cultured from the ZBP1+/− versus ZBP1+/+ mice at 3 DIV after transfection with mCherry, ZBP1, or ZBP1Y396F. The mCherry‐transfected ZBP1+/− DRGs showed decreased axonal levels of actin and GAP‐43 mRNAs compared with DRGs from ZBP1+/+ littermates (Figure 7C; Supplementary Table S5). Transfecting with ZBP1–mCherry increased the axonal β‐actin and GAP‐43 mRNA levels in the ZBP1+/− DRGs and again increased these axonal mRNA in the wild‐type neurons. Though transfecting ZBP+/− DRGs with the ZBPY396F mutant did not rescue their axonal growth deficit (Figure 6D), this did restore the axonal β‐actin and GAP‐43 mRNAs to levels comparable to wild‐type neurons and increased the axonal levels of these mRNAs in wild‐type littermate DRGs (Figure 7C; Supplementary Table S5). These ZBP1 rescue studies distinguish mRNA transport and translational activities of ZBP1, and again emphasize that the translation of ZBP1 cargo mRNAs is needed for optimal axonal outgrowth.
Since axonal neuritin/cpg15 mRNA was not depleted to the same degree as β‐actin or GAP‐43 mRNAs with expression of GFP plus β‐actin 3′UTR (Figure 6A) and it did not show any significant change with ZBP1 depletion (Figure 7A and B), we reasoned that this mRNA might not be a direct target for ZBP1 binding. To test this possibility, we returned to the GFP–3′β‐actin expressing neurons as these showed the most consistent depletion of axonal neuritin/cpg15. As with the ZBP1+/− neurons, transfecting Tα1‐GFP–3′β‐actin DRGs with ZBP1Y396F rescued the axonal levels of β‐actin and GAP‐43 mRNAs (data not shown); however, transfecting ZBP1Y396F into the Tα1‐GFP–3′β‐actin DRGs did not rescue the axonal neuritin/cpg15 levels (Figure 7D). Since ZBP1Y396F does not restore axonal growth in the ZBP1+/− or GFP–3′β‐actin expressing neurons (Figures 6D and 4B), these data indicate that neuritin/cpg15 is not a ZBP1 cargo.
Our findings reveal an essential role for ZBP1–mRNA interactions in axonal regeneration, both in cultured neurons and in vivo. Previous work in non‐neuronal cells showed that ZBP1‐bound β‐actin mRNA is translationally inactive and requires ZBP1's phosphorylation at tyrosine 396 for derepressing this translational inhibition (Huttelmaier et al, 2005). In neurons, ZBP1 is locally phosphorylated by Src in growth cones, and overexpression of ZBP1Y396F impairs growth cone turning towards BDNF (Sasaki et al, 2010). Here, we show that wild‐type ZBP1, but not ZBP1Y396F, reverses the axon outgrowth deficits in GFP–3′β‐actin mRNA expressing and ZBP+/− neurons. However, the ZBP1Y396F mutant rescued deficits in axonal localization of both β‐actin and GAP‐43 mRNAs in neurons from the ZBP+/− mice and those expressing the GFP–3′β‐actin mRNA. Thus, both ZBP1–mRNA transport and localized translation of ZBP1–mRNA cargos contribute to axonal growth capacity. This is the first observation that clearly links function of a specific RBP to transport and translation of mRNA cargos for axonal outgrowth in vitro and in vivo.
These data also emphasize the need to develop means to functionally assess RNA localization mechanisms. Though it is not clear from our studies if GAP‐43 mRNA is bound directly by ZBP1 or requires additional proteins, there clearly is a functional interaction between this mRNA and ZBP1. Previous studies have shown that the Elav protein HuD binds to and stabilizes GAP‐43 mRNA (Perrone‐Bizzozero and Bolognani, 2002). Both GAP‐43 mRNA and HuD have been localized to axons and they colocalize (Smith et al, 2004; Willis et al, 2007), although a role for HuD in mRNA localization has not been established. However, HuD and IMP‐1 (ZBP1) have been shown to interact within a RNP complex associated with Tau mRNA (Atlas et al, 2004). Additional studies will be needed to define the composition of the GAP‐43 mRNA transport complex. It is intriguing that axonal depletion of GAP‐43 mRNA was never to the same extent as β‐actin mRNA, which may suggest additional proteins contribute to transport of GAP‐43 into axons or the affinity of ZBP1 for binding β‐actin and GAP‐43 may differ.
The UTR competition approach used here provides a direct test for co‐transport of mRNAs in living cells. Previous co‐immunoprecipitation analyses for IMP1 (ZBP1) detected neuritin/cpg15 mRNA as a ZBP1 target (Jonson et al, 2007). Here, the decrease in axonal levels of neuritin/cpg15 was restored by transfection with ZBP1 but not with ZBP1Y396F that has RNA binding activity but cannot be phosphorylated to efficiently release mRNAs for translation. Although we cannot exclude the possibility that the GFPmyr–3′β‐actin mRNA competes with endogenous mRNAs by binding to proteins beyond ZBP1, our data show that axonal localization of neuritin/cpg15 is not linked to the availability of ZBP1. These data indicate that, like calreticulin and amphoterin, other RBPs are undoubtedly used to transport neuritin/cpg15 mRNA into axons. This study emphasizes the need for studies to identify other RBPs that localize mRNAs to axons in vivo. Interestingly, two recent reports have shown that HuD binds to neuritin/cpg15 mRNA (Akten et al, 2011; Wang et al, 2011) and HuD is known to bind to GAP‐43 mRNA (Mobarak et al, 2000).
Although initially generated as a model to visualize sites of axonal translation in vivo (Willis et al, 2011), the transgenic mice carrying diffusion‐limited GFPmyr with 3′UTR of β‐actin proved to be an extremely useful resource for testing in vivo functional significance for axonal mRNA transport. This approach for localizing reporter mRNAs through 3′UTRs has previously been used to visualize protein synthesis in vivo in dendrites and developing axons (Brittis et al, 2002; Meyer‐Luehmann et al, 2009). However, our studies add a new facet by showing that increasing levels of the UTR elements can interfere with endogenous RNA trafficking in vivo. Such a squelching of mRNA transport has been demonstrated in vitro for UTR elements from Integrin α3 and COXIV mRNAs, but potential effects on the localization of other mRNAs or their function in axon growth was not reported (Adereth et al, 2005; Aschrafi et al, 2010). The ability of ZBP1 to rescue axonal localization of select mRNAs and axon growth defects provides proof for mRNAs utilizing a common RBP for localization.
A critical advance in the studies here is that the levels of an RBP can limit regeneration capacity. Although translation of localized mRNAs has been shown to play a role axon outgrowth, pathfinding, and growth cone formation in vitro (Holt and Bullock, 2009), this is the first clear evidence that transport of mRNAs from the neuronal cell body into distal axons contributes to regenerative axonal growth in vivo. By both morphological and electrophysiological measures, PNS axonal regeneration decreased when ZBP1 levels decreased. ZBP1 is expressed at high levels during embryonic development (Nielsen et al, 1999), which likely explains why measures of sensory development appear normal in the mice expressing GFP–3′β‐actin transgene. The decreased regeneration in the Tα1‐GFP–3′β‐actin and ZBP1+/− mice is undoubtedly the result of low expression of ZBP1 in adults. This raises the possibility that altering expression of ZBP1 in adult neurons may help to increase axonal regeneration by increasing the fraction of mRNAs delivered into the axonal compartment.
Materials and methods
All animal work was approved by Institutional Animal Care and Use Committees (IACUC) at Drexel University, Alfred I. duPont Hospital, Emory University, or Ohio State University.
PC12 cells were maintained as described (Twiss et al, 2000). Naive or 7 day injury‐conditioned DRGs from 4‐ to 12‐week‐old Sprague Dawley rats or mice were cultured as described with minor modifications (Twiss et al, 2000). The Tα1‐GFP–3′β‐actin and Tα1‐GFP–3′γ‐actin lines are on FVN background (Willis et al, 2011); ZBP1 gene‐trap mice are on C57Bl/6 background (Hansen et al, 2004). ZBP+/− lines were crossed with Thy1–YFP transgenic mice (Feng et al, 2000) to facilitate axon visualization. Surgery for injury conditioning is outlined below. In some experiments, ganglia were stored overnight at 4°C in Hibernate medium (BrainBits). Neurons were enriched by Percoll gradient (Hanz et al, 2003). All transfections were performed with the Amaxa Nucleofector (Lonza).
Plasmid and viral expression constructs
AV expressing GFPmyr with 3′UTRs of β‐ and γ‐actin mRNAs have been previously described (Willis et al, 2007). The Tg‐Tα1‐dzGFPmyr–3′β‐actin and Tg‐Tα1‐dzGFPmyr–3′γ‐actin transgene constructs have recently been described (Willis et al, 2011). Briefly, the transgenic construct consisted of a destabilized GFPmyr with 5′UTR of CamKIIα (Aakalu et al, 2001) and 3′UTRs of mouse β‐actin or γ‐actin mRNAs. Transgene expression was driven by the neuronal‐specific Tα1 tubulin promoter (Gloster et al, 1994; Bamji and Miller, 1996).
Fusion constructs for ZBP1, ZBP1Y396F, and ZBP1ΔKH1−4 have been described (Farina et al, 2003; Huttelmaier et al, 2005). For β‐actin–mCherry fusion protein, mCherry was amplified from mCherry–ZBP, cloned into pTOPO (Invitrogen), and then used to replace DsRed in pDsRed‐Actin (Clontech). β‐Actin 3′UTR was amplified and directly cloned into the BamH1 and Mfe1 sites directly downstream of mCherry–β‐actin. All cloned PCR products were sequenced prior to use.
Animal surgeries and procedures
For injury‐conditioning and morphological studies of regeneration, rodents were subjected to sciatic nerve crush at mid‐thigh as described (Smith and Skene, 1997). The contralateral (control) nerve was exposed but not manipulated. Nerve grafting procedure for common fibular nerve is published (English et al, 2005). For acellular grafts, the donor nerve segment was frozen in liquid nitrogen for 2 min and then placed in normal saline at 37°C for 2 min; this freeze‐thaw cycle was repeated twice prior to grafting (English et al, 2007). For the physiological studies of nerve regeneration, recordings were obtained prior to nerve injury (see below) and then the tibial branch of the sciatic nerve, which innervates gastrocnemius, was isolated and crushed 15 mm distal to the emergence of the sciatic nerve from the pelvis using No. 5 forceps. The common fibular branch of the sciatic nerve, which innervates the TA muscle, was left intact as an internal control.
To harvest tissues for immunofluorescence (IF), deeply anaesthetized mice were perfused transcardially with PBS followed by 4% paraformaldehyde. Tissues were post‐fixed with 3% paraformaldehyde for 2–4 h, and then cryoprotected at 4°C overnight in 30% sucrose. The tissues were processed for cryosectioning or used directly for whole mount IF.
For analyses of axon size, perfusion fixative consisted of 3% formaldehyde, 1.5% glutaraldehyde, and 2.5% sucrose in 100 mM cacodylate buffer (pH 7.4). Nerves were then immersion fixed for 60 min, post‐fixed in 1% OsO4, stained with uranyl acetate (Ted Pella), dehydrated through graded ethanol series, and embedded in Epon. In all, 1 μm sections were used for the morphometry of myelinated axons (i.e., G‐ratios, axon diameter). In all, 70 nm sections were contrasted with 1% uranyl acetate and lead citrate and viewed with a Hitachi H7650 electron microscope for unmyelinated axons.
Compound muscle action potentials (M responses) were recorded in response to sciatic nerve stimulation in ZBP1+/− and wild‐type littermates. All recordings were made in pentobarbital anaesthetized animals. Bipolar fine wire EMG electrodes were constructed from single‐stranded enamel‐coated stainless steel wire (California Fine Wire) and inserted into the gastrocnemius and TA muscles using a 25‐G hypodermic needle. A monopolar needle electrode was placed next to the sciatic nerve near the sciatic notch and used to deliver short (0.05 ms duration) cathodal stimulus pulses. An indifferent anodal electrode was placed into the adjacent lumbar epaxial musculature. Isolated constant voltage stimuli were applied at a rate of 0.3 Hz at increasing amplitudes ranging from subthreshold to supramaximal for evoking a twitch in innervated muscles. Direct muscle (M) responses evoked by this stimulation were acquired at a sampling rate of 10 kHz and recorded to disc. Activity was recorded for 20 ms prior to and 30 ms after each stimulation. All activity in these records was full wave rectified before analysis. Average voltages in the pre‐stimulus period were considered background activity and subtracted from each of the points sampled after stimulation to adjust for background. The average adjusted activity in a defined M response time window, generally 0.5–4.5 ms after stimulation, was determined and expressed in mV. The onset and duration of this window needed to be adjusted slightly over the course of muscle re‐innervation, as we have noted previously (English et al, 2007). All procedures were performed bilaterally and recordings were repeated on days 7 and 14 after nerve crush. On each day, the maximal M response amplitude was measured, and expressed as a proportion of the maximal M response amplitude recorded prior to transection. Averages of these relative amounts of recovery of the M response at different times were compared between the two mouse groups using an unpaired t‐test.
Thermal sensation and mechanosensation were assessed in uninjured mice using plantar infrared analgesia meter and Dynamic Plantar Aesthsiometer (Ugo Basile), respectively. After mice were habituated to the apparatus, withdrawal latency to heat source and thresholds to pressure were recorded ⩾5 times for each hind paw (n=15 mice).
Analyses of RNAs by RT–PCR
DRGs and sciatic nerves were placed in lysis buffer (RNAqueous kit, Ambion) and homogenized using a motorized homogenizer (Omni International). RNA was isolated from cultured cells with microRNAqueous kit (Ambion). RNA samples were treated with DNase prior to analysis. In all, 100 ng of total RNA was used as template for RT with iScript (Bio‐Rad). RT reactions were diluted five‐fold for PCR. Standard PCR (⩽30 cycles) was analysed by agarose gel electrophoresis and ethidium bromide staining. qPCR was performed as above with all analyses in quadruplicate. See Supplementary Table S7 for qPCR primers.
For analyses of RNA binding to ZBP1, brain+spinal cord tissue from P5 mice was homogenized at 4°C in 50 mM Tris–HCl, pH 7.5, 0.5 mM DTT, 150 mM NaCl, 1% NP‐40, 5 mM MgCl2, 10 U/ml SUPERase·In™ (Ambion), plus protease inhibitors. RNA was isolated from an aliquot of each lysate and processed for RT–qPCR to test for input RNA levels. Remainder of each lysate was pre‐cleared with protein A agarose (Roche), and then incubated for 3 h at 4°C with protein A agarose loaded with 4 μg of anti‐ZBP1 or IgG control (Jackson ImmunoRes). Immunocomplexes were collected and stringently washed. One third of the immunocomplex was used for SDS/PAGE and immunoblotting for ZBP1; the remainder of the immunocomplex was used for RNA extraction with Trizol (Gibco) and processed for RT–qPCR. For each ZBP1 immunoprecipitation, Ct values were normalized to the 10% input for the corresponding mRNA.
Unless otherwise indicated, all steps were performed at room temperature. For tissue sections, samples were immersion fixed in 4% paraformaldehyde for 2–4 h, cryoprotected overnight in 30% sucrose at 4°C, and then cryosectioned. Sections were warmed to room temperature, washed in PBS, and incubated in 20 mM glycine for 30 min followed by 0.25 M NaBH4 for 30 min. For cultures, coverslips were fixed in 4% paraformaldehyde for 20 min and then rinsed in PBS. IF was then preformed as described (Willis et al, 2005). The following antibodies were used: rabbit anti‐Peripherin (1:600; Millipore); chicken anti‐NFH (1:1000; Millipore); mouse anti‐NF (1:1500 each NFL, NFM and NFH; Sigma); rabbit anti‐CGRP (1:1000; Millipore/Chemicon); mouse anti‐GAP‐43 (1:100; Millipore); rabbit anti‐GAP43 (1:1000; Millipore); rabbit anti‐PGP9.5 (1:1500; Millipore/Chemicon); rabbit anti‐Nav1.8 (1:500; Millipore); Alexa546 goat anti‐chicken (1:1000; Invitrogen); and Alexa555 goat anti‐rabbit (1:1000; Invitrogen). In some experiments, 1:300 dilution of the FlouroMyelin Red (Invitrogen) was used per manufacturer's instructions.
In situ hybridization
FISH/IF to detect axonal mRNAs and proteins was performed with digoxigen‐labelled oligonucleotide probes as described (Willis et al, 2007). The probe sets are listed in Supplementary Table S7. Fluorescent signals were detected by epifluorescence. All images sets from individual experiments were matched for exposure, gain, and offset. ImageJ was used to quantify FISH signals (pixels/μm2) in the cell body or the distal 200 μm of the axon. Negative controls for these FISH analyses consisted of scrambled digoxigen‐labelled oligonucleotides; images for individual experiments were normalized to the negative control and reported as pixels/μm2 above this threshold level. For cell body analyses, DRG cultures were depleted of Schwann cells by Percoll gradient.
Quantitation of neuronal and axonal morphology
Morphology of the axons was analysed in immunostained cultures. Images of random neurons were collected and blinded for analyses. The mean axon length was obtained by measuring the longest axon per neuron using ImageJ. Axon branching and number of processes were manually scored. Growth cone area was calculated using ImageJ from merged NFH+peripherin and DIC images. For injury‐conditioned DRG cultures, morphology was analysed at 1 DIV. For naive DRG cultures, morphology was assessed at ⩾3 DIV to allow for expression of AV or plasmid constructs.
For analyses of crushed nerves, sections were immunostained for axonal proteins and then co‐stained with Fluoromyelin Red and DAPI. For ZBP1+/−/Thy1–YFP mice, YFP was used to identify axons. Matched exposure images were captured proximal and distal to the crush site. Montages of each nerve were constructed to encompass ±1.5 mm from injury site. PGP9.5‐ or YFP‐positive axons were counted in every fourth longitudinal section of the sciatic nerve using Neurolucida (MBF Biosciences) and total counts were calculated using the line transect method at −1.0, −0.5, 0, 0.5, and 1.0 mm from the centre of the crush site (Hill et al, 2001). Some images were evaluated using a ‘thresholding method’. For this, montage images for axonal markers were processed for thresholding (0–50) using ImageJ and mean pixel intensity per area was then measured in 100 μm bins. ‘Regeneration front’ was determined as bin relative to crush where signal thresholds were 50% of the proximal intact nerve.
For analysis of regeneration in nerve transection/graft experiments, whole mount tissues were used. For the transgenic mice, GAP‐43 immunostaining was used to detect axons; negative controls consisted of a GAP‐43‐stained naive nerve and grafts processed without addition of primary antibody. For the ZBP1+/−/Thy1–YFP mice, linear YFP profiles in the acellular graft were counted. Whole mount nerves were imaged confocally with Z‐stacks captured at 10 μm intervals. Montages of the nerve/graft were then analysed using Imaris X64 (Bitplane) at designated intervals from the proximal host–graft junction.
G‐ratio and axonal diameters were calculated from images of toluidine blue‐stained epon sections using Image Track software (http://www.ucalgary.ca/styslab/imagetrack). Average neurons/ganglion were calculated from every fourth section of L4/5 DRGs in NF+peripherin‐stained sections using Volocity software (Improvision). Density of CGRP‐positive nerve fibres in the footpads of transgenic mice was determined from transverse sections of hindlimb footpad over five sections per mouse.
Student's T‐tests were used to compare the means of two independent groups. When three or more independent groups were present, an one‐way ANOVA, followed by a Tukey's post hoc test was performed. To compare non‐parametric distributions, a Wilcoxon's rank‐sum test of frequency distributions was used. To compare differences between the linear distributions for axonal length and GFP intensity, an analysis of covariance (ANCOVA) was performed.
Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).
Conflict of Interest
The authors declare that they have no conflict of interest.
Erin Schuman provided the dzGFPmyr reporter. Jessica F Elder (Burke Inst.) and Jobayer Hossain (Nemours Biomed. Res.) provided guidance on statistical analyses. This work has been supported by funds from NIH (R21‐NS045880, R01‐NS049041, and R01‐NS041596 to JLT; K99‐NR010797 to DEW; R21‐NS060098 to GJB; R01‐NS057190 to AE; R01‐NS039472 and P30‐NS045758 to SOK; and P20‐RR15588 to CBK‐S), the Christopher and Dana Reeve Foundation (TB2‐0602), and the Adelson Medical Research Foundation. The Nemours Histotechnology and Biomolecular Core facilities provided technical support through funding from the Nemours Foundation and P20‐RR020173 COBRE grant.
Author contributions: CJD, DEW, and MX performed the analyses of RNA transport and axonal regeneration. CT and SOK performed behavioural analyses of transgenic mice. CJ and AE performed physiological analysis of mice. SY analysed axotomy responses. NCS and CBK‐S provided guidance on mouse cross‐breeding and analyses. JvM performed morphometric analysis of peripheral nerves. GJB and JLT oversaw experimental design and implementation.
- Copyright © 2011 European Molecular Biology Organization