Enhancement of axonal regeneration by in vitro conditioning and its inhibition by cyclopentenone prostaglandins.

Axonal regeneration is enhanced by the prior `conditioning' of peripheral nerve lesions. Here we show that Xenopus dorsal root ganglia (DRG) with attached peripheral nerves (PN-DRG) can be conditioned in vitro, thereafter showing enhanced neurotrophin-induced axonal growth similar to preparations conditioned by axotomy in vivo. Actinomycin D inhibits axonal outgrowth from freshly dissected PN-DRG, but not from conditioned preparations. Synthesis of mRNAs that encode proteins necessary for axonal elongation might therefore occur during the conditioning period, a suggestion that was confirmed by oligonucleotide microarray analysis. Culturing PN-DRG in a compartmentalized system showed that inhibition of protein synthesis (but not RNA synthesis) in the distal nerve impaired the conditioning response, suggesting that changes in gene expression in cultured DRG depend on the synthesis and retrograde transport of protein(s) in peripheral nerves. The culture system was also used to demonstrate retrograde axonal transport of several proteins, including thioredoxin (Trx). Cyclopentenone prostaglandins, which react with Trx, blocked the in vitro conditioning effect, whereas inhibition of other signalling pathways thought to be involved in axonal regeneration did not. This suggests that Trx and/or other targets of these electrophilic prostaglandins regulate axonal regeneration. Consistent with this hypothesis, morpholino-induced suppression of Trx expression in dissociated DRG neurons was associated with reduced neurite outgrowth.

The mechanisms that lead to changes in gene expression in axotomised neurons are poorly understood but probably involve retrograde axonal transport of activating signalling molecules (reviewed by Hanz and Fainzilber, 2006;Zhou and Snider, 2006;Raivich and Makwana, 2007); their identification is of considerable importance. Signal transducer and activator of transcription 3 (STAT3) might act as a retrograde activating signal because it is rapidly phosphorylated in lesioned nerves and, subsequently, appears in the nuclei of both sensory and motor neurons (Haas et al., 1999;Sheu et al., 2000;Schwaiger et al., 2000;Lee et al., 2004). Inhibition of STAT3 phosphorylation with AG490 suppresses regenerative axonal growth of adult DRG neurons (Liu and Snider, 2001;Qiu et al., 2005). Jun is essential for the expression of certain RAGs and optimal axonal regeneration (Raivich et al., 2004), and retrograde axonal transport of its activator Jun terminal kinase (JNK) has been demonstrated (Lindwall and Kanje, 2005). In their phosphorylated form, the extracellular signal-regulated protein kinases 1 and 2 (Erk1/2) are also retrogradely transported and involved in initiation of axonal regeneration (Perlson et al., 2005).

Axonal regeneration is enhanced by the prior 'conditioning' of peripheral nerve lesions. Here we show that Xenopus dorsal root ganglia (DRG) with attached peripheral nerves (PN-DRG)
can be conditioned in vitro, thereafter showing enhanced neurotrophin-induced axonal growth similar to preparations conditioned by axotomy in vivo. Actinomycin D inhibits axonal outgrowth from freshly dissected PN-DRG, but not from conditioned preparations. Synthesis of mRNAs that encode proteins necessary for axonal elongation might therefore occur during the conditioning period, a suggestion that was confirmed by oligonucleotide microarray analysis. Culturing PN-DRG in a compartmentalized system showed that inhibition of protein synthesis (but not RNA synthesis) in the distal nerve impaired the conditioning response, suggesting that changes in gene expression in cultured DRG depend on the synthesis and retrograde transport of protein(s) in peripheral nerves. The culture system was also used to demonstrate retrograde axonal transport of several proteins, including thioredoxin (Trx). Cyclopentenone prostaglandins, which react with Trx, blocked the in vitro conditioning effect, whereas inhibition of other signalling pathways thought to be involved in axonal regeneration did not. This suggests that Trx and/or other targets of these electrophilic prostaglandins regulate axonal regeneration. Consistent with this hypothesis, morpholinoinduced suppression of Trx expression in dissociated DRG neurons was associated with reduced neurite outgrowth.
In the present investigation, we demonstrate that Xenopus dorsal root ganglia (DRG) with attached short peripheral nerves (PN-DRG) that were incubated for 3 days in serum-free medium (in vitro conditioned) show enhanced axonal growth in response to neurotrophins, similar to PN-DRG axotomized by sciatic nerve lesions (in vivo conditioned), even in the presence of actinomycin D (ActD), indicating prior synthesis of mRNAs necessary to support axonal regeneration. Using the in vitro conditioning model to investigate signalling mechanisms involved in the control of axonal regeneration, we show that this might depend (in part) on synthesis and retrograde transport of proteins that are encoded by mRNA within the peripheral nerve. By 2D gel electrophoresis (2-DE) and liquid chromatography tandem mass spectrometry (LC-MS/MS), several retrogradely transported proteins were identified, including the oxidoreductase thioredoxin (Trx). Pharmacological inhibition of Trx function by using cyclopentenone prostaglandins blocked the in vitro conditioning effect, whereas inhibition of other signalling pathways thought to be involved in axonal regeneration did not, suggesting that Trx and/or other redox-sensitive factors regulate RAG expression and axonal regeneration. Moreover, morpholino oligonucleotide (MO)-induced suppression of Trx levels in dissociated DRG neurons was associated with reduced neurite outgrowth.

De novo transcription is required for axonal growth following conditioning in vitro
Since neurite elongation rates of dissociated rat DRG neurons increase spontaneously during culture (Smith and Skene, 1997), which suggests the 'conditioning lesion effect' is mimicked in vitro, we investigated this possibility using Xenopus PN-DRG, which show excellent survival in culture (Tonge et al., 2004). In the absence of trophic factors, freshly dissected PN-DRG in collagen gels showed limited axonal outgrowth (Fig. 1A), but outgrowth distance was significantly increased (P<0.01) compared with in vitro conditioned preparations (Fig. 1B). Brain-derived neurotrophic factor (BDNF) markedly increased the lengths (by 91%; P<0.001) and apparent numbers of outgrowing axons from freshly dissected PN-DRG ( Fig. 1C) but even more so from in vitro conditioned preparations (Fig. 1D), with a significant (P<0.001) increase (by 157%) in axon length compared with freshly dissected preparations.
To investigate whether changes in axonal growth following in vitro conditioning were transcription dependent, effects of ActD on BDNF-induced axonal outgrowth from freshly dissected and in vitro conditioned PN-DRG were investigated. In the absence of neurotrophins, ActD significantly reduced axonal outgrowth distances from both freshly dissected and in vitro conditioned PN-DRG (by 64% and 68%; P<0.01 and P<0.001, respectively), compared with outgrowth from preparations cultured without ActD (Fig. 2). However, although ActD caused a marked (P<0.001) reduction in axonal outgrowth distance from freshly dissected preparations (by 76%; Fig. 1E) in the presence of BDNF, it had little effect on in vitro conditioned preparations, which showed abundant axonal outgrowth for several days (Fig. 1F). Similar increases in axonal outgrowth from in vitro conditioned PN-DRG occurred in response to nerve growth factor (NGF), neurotrophin 3 (NT3) and neurotrophin 4 (NT4), also persisting in the presence of ActD (data not shown). These results (summarized in Fig. 2A) suggest that, during the initial conditioning period in vitro, neurons synthesize mRNAs that are necessary and sufficient to sustain subsequent axonal growth in response to neurotrophins for several days. To test this hypothesis, PN-DRG were incubated free-floating for 1 day with α-amanitine (α-AM), an irreversible RNA polymerase II inhibitor, followed by 3 days without the drug. When these preparations were subsequently cultured on collagen gels, (with or without ActD), BDNF did not increase axonal outgrowth (Fig. 2B). By contrast, PN-DRG that had been in vitro conditioned for 3 days before α-AM exposure on day 4 showed vigorous BDNFinduced axonal outgrowth that was not significantly different from preparations incubated for 4 days without α-AM treatment or significantly reduced by ActD (Fig. 2B), indicating that mRNA required for axonal growth is synthesized during the first 3 days in culture.

Transcriptional dependence of axonal growth following axotomy in vivo
To determine whether in vivo conditioned DRG neurons also show transcription-independent axonal regeneration, PN-DRG were cultured in collagen gels 3 or 5 days after unilateral sciatic nerve section. The mean axonal outgrowth distances from prior axotomized (in vivo conditioned) and contralateral (control) PN-DRG are summarized in Fig. 3. In the absence of BDNF, control PN-DRG showed sparse axonal outgrowth after 3 days in culture (similar to those shown in Fig. 1A) but axonal outgrowth distance In vitro conditioned PN-DRG show enhanced BDNF-induced axonal outgrowth in the presence or absence of ActD. After 3 days in collagen gels, freshly dissected PN-DRG show limited axonal growth (A) that is markedly increased by BDNF (C) but not in the presence of ActD (E). In vitro conditioned preparations show slightly increased axonal outgrowth (B) that is greatly enhanced by BDNF (D), even in the presence of ActD (F). Note that not all axons are within the plane of focus. Scale bar: 200 μm. was slightly increased from 3-day in vivo conditioned PN-DRG and significantly so (P<0.02) from 5-day conditioned preparations (82% and 171%, respectively). In the presence of BDNF, dense axonal outgrowth occurred from control preparations (similar to Fig. 1C), with significant increases in axonal outgrowth distance (P<0.02) compared to preparations cultured without BDNF (420% and 213%, respectively). However, BDNF caused even greater axonal outgrowth from both 3 and 5 day in vivo conditioned preparations, with significantly longer axonal outgrowth than from control preparations (950% and 565%, respectively; P<0.02 and P<0.005, respectively).
In the presence of ActD, BDNF significantly increased the mean axonal outgrowth distances in 3-and 5-day in vivo conditioned preparations (349% and 75%, respectively; P<0.001 and P<0.01 respectively), compared with preparations cultured without BDNF and (Fig. 3). By contrast, BDNF did not significantly increase axonal outgrowth distance from control preparations of 3-day operated animals in the presence of ActD, although it did so in preparations from 5-day operated animals (60%; P<0.001), suggesting that peripheral nerve lesions have contralateral effects, as observed previously (Rotshenker and Tal, 1985 and references therein). These results (summarized in Fig. 3) show that ActD has relatively little effect on BDNF-induced axonal outgrowth of in vivo conditioned PN-DRG, supporting the hypothesis that either axotomy in vivo or culture in vitro cause changes in gene expression that is required for axonal regeneration.

Changes in gene expression in cultured DRG
In view of the above observations, we carried out an oligonucleotide microarray analysis to compare gene expression in freshly dissected and 3-day in vitro conditioned DRG. Results of this experiment showed extensive increase and decreases in gene expression (Table S1 in supplementary material) with 58 identified genes that were upregulated fivefold or more in 3-day in vitro conditioned compared with freshly dissected DRG (Table 1). To verify changes in gene expression, quantitative PCR (qPCR) was carried out using the cDNA samples from control and conditioned DRG that had been used in the array analysis, with primers recognizing six genes that had been selected for their relevance in axonal regeneration. Arginase and galectin expression increase during axonal regeneration (Lange et al., 2004;Horie et al., 2005). Transportin mediates nuclear import of proteins lacking conventional nuclear localization signals (Carson et al., 2006) and might therefore be involved in the transport of signalling molecules that regulate expression of RAGs -similar to the role of importins in rodents, (Hanz et al., 2003). Vg1 RNA binding protein (Vg1RBP) mediates the translocation of β-actin and cofilin mRNAs to growth cones where it is involved in local protein synthesis during growth cone turning Piper et al., 2006). Secreted frizzledrelated protein sequence protein 1 (SFRP) is involved in axonal growth of retinal ganglion cells (Rodriguez et al., 2005) and Xenopus Hen1 is a transcription factor involved in neural development (Bao et al., 2000). Therefore, both might control expression of genes required for axon elongation. The results, normalized to those of 18S RNA, show all six genes were upregulated at least twofold ( Fig. 4) consistent with extensive Mean axonal outgrowth distances from freshly dissected and in vitro conditioned PN-DRG in collagen gels after 3 days, in response to BDNF in the presence or absence of ActD. Conditioned PN-DRG show enhancement of both spontaneous and BNDF-induced axonal growth. ActD inhibits BDNF-induced axonal outgrowth from freshly dissected preparations but not from conditioned preparations. (B) Effects of prior incubation with α-AM on subsequent BDNFinduced axonal outgrowth in the presence or absence of ActD. Preparations were incubated (free-floating) for 4 days with α-AM present on day 1, day 4 or not at all, prior to culturing in collagen gels for 3 days with or without BDNF. Incubation with α-AM on day 1 strongly inhibits subsequent BDNFinduced axonal outgrowth in collagen gels, but incubation with α-AM on day 4 does not. changes in gene expression detected by microarray analysis in conditioned DRG.
Results of these experiments ( Fig. 5) showed that pre-incubation of PN-DRG with 50 μM AG490 inhibited subsequent BDNFinduced axonal outgrowth with and without ActD (by 80% and 57%, respectively; P<0.001 and P<0.005, respectively). However, at concentrations of AG490 as low as 2 μM also strongly inhibited BDNF-induced axonal outgrowth from both freshly dissected and in vitro conditioned preparations, similar to findings by Liu and Snider (Liu and Snider, 2001). These results indicate that JAK/STAT signalling is necessary for axonal regeneration even after the induction of RAGs, and/or that AG490 also affects cells in other ways. For example, in addition to inhibiting JAK2, AG490 has antioxidant properties (Gorina et al., 2007); so, results of experiments involving its use should be treated with caution. In the presence of the other inhibitors and in Ca 2+ -free RPMI medium, BDNF-induced axonal outgrowth was generally reduced as expected from other studies (reviewed by Zhou and Snider, 2006). However, prior incubation of the preparations with these inhibitors or culturing in Ca 2+ -free medium did not impair subsequent BDNF-induced axonal outgrowth, even in the presence of ActD (data not shown), suggesting the expression of RAGs was not inhibited.
Since our experiments did not confirm the involvement of hitherto-postulated signalling pathways that control the expression of genes required for axonal regeneration, we used the PN-DRG preparations to investigate its dependence on axonal protein synthesis because mRNA translation of β-importin and vimentin in lesioned axons is important for the retrograde transport of signalling Journal of Cell Science 121 (15)  Validation of microarray data using qPCR. Analysis of mRNAs isolated from freshly dissected and 3-day in vitro conditioned DRG using qPCR. The level of expression for each gene in cultured DRG is normalized to that in freshly dissected DRG in each case. Enhancement of axonal regeneration molecules that are involved in the initiation of axonal regeneration (Hanz et al., 2003;Perlson et al., 2005). Translation of many proteins is known to occur in axons (Giuditta et al., 2002;Piper and Holt, 2004;Willis et al., 2005) and may be essential for efficient axonal regeneration (Verma et al., 2005). To investigate the involvement of axonal protein synthesis for the induction of genes required for axonal regeneration, PN-DRG preparations were incubated in twocompartment culture dishes (Edbladh et al., 1994) allowing localized application of drugs to either DRG or the cut end of the attached peripheral nerve. After 3 days, the DRG and the proximal ~1-2 mm part of the peripheral nerve (which had not been exposed to the drug) were cultured in collagen gels with BDNF (with and without ActD). Results of these experiments (Fig. 6) showed that the inhibition of protein translation with cycloheximide (CHX) in the distal part of the nerve reduced subsequent BDNF-induced axonal growth from the trimmed nerves in presence of ActD by 63% (P<0.02) compared with preparations incubated without CHX. BDNF-induced axonal outgrowth distance in the absence of ActD was also reduced (by 42%) in preparations in which the peripheral nerve had been incubated with CHX, but the difference was not significant. By contrast to the inhibitory effects of CHX on subsequent BDNF-induced axonal growth, ActD applied to the distal end of the peripheral nerve in the two-chamber culture system did not inhibit subsequent axonal outgrowth (data not shown). These findings suggest that the synthesis of mRNA required for axonal regeneration, usually occurring within the first 3 days in culture, is partly dependent on protein synthesis in the distal part of the peripheral nerve.

Retrograde axonal transport of proteins
To identify retrogradely transported proteins synthesized in peripheral nerves, we used the two-chamber culture system to metabolically label proteins in the distal part of the nerve with [ 35 S]methionine-[ 35 S]cysteine followed by 2-DE and autoradiography to determine whether labelled proteins subsequently appeared in the DRG. In these experiments, the distal ends of the PN-DRG preparations were in the inner compartment, containing [ 35 S]methionine-[ 35 S]cysteine, whereas the DRG and proximal parts of peripheral nerves were in the outer compartments. As controls, additional PN-DRG were incubated (free-floating) in the outer compartments.
In three separate experiments, 2-DE gels of DRG proteins stained with silver showed several hundred spots (Fig. 7) of which onlỹ 35 were radioactive on each gel. 2-DE gels of the control DRG (from the outer compartments) showed similar patterns of (nonradioactive) silver-stained spots (data not shown) and addition to the inner compartment of vinblastine (100 μM), which blocks axonal transport in amphibian nerves (Hanson and Edstrom, 1977), also prevented appearance of radio-labelled proteins in the DRG, indicating that this depends on retrograde axonal transport rather than isotope leakage. This conclusion is also reinforced by the fact that on 2D-gel autoradiographs of DRG cultured directly in [ 35 S]methionine-[ 35 S]cysteine, virtually all silver stained spots were radioactive (data not shown), a pattern completely different to the highly selective group of radioactive spots observed following localized isotope application to the distal nerve.
To characterize the retrogradely transported proteins, 35 radioactive spots from a silver-stained gel were excised and subjected to trypsin digestion, LC-MS/MS and database searching, resulting in the identification of 32 proteins (Table 2). However, 13 spots contained more than one protein and in those cases it was uncertain which protein was radioactively labelled and therefore retrogradely transported. Several proteins in the radioactive spots, including cofilin, profilin, vimentin, actin, tubulin, UchL-1, enolase,   6. In vitro conditioning is impaired by inhibition of protein synthesis in the peripheral nerve. PN-DRG were incubated for 3 days in compartmentalized culture dishes, with the end of the peripheral nerve in the inner compartment (with or without CHX), prior to culturing the DRG and proximal part of the peripheral nerve (which had not been exposed to CHX) in collagen gels with BDNF (with or without ActD). Protein synthesis inhibition in the peripheral nerve significantly reduced subsequent BDNF-induced axonal outgrowth in the presence of ActD.
glyceraldehyde-3-phosphate dehydrogenase and aldolase, are translated from mRNAs in regenerating axons (Willis et al., 2005) and retrograde axonal transport of vimentin has been demonstrated (Perlson et al., 2005). To investigate the presence of mRNAs encoding other proteins than those identified in the spots, we harvested axons that had been grown in matrigel through 8-μm pores in a membrane barrier and were therefore free from contaminating migratory cells (Fig. 8). To confirm the absence of cellular contamination, others (see also Willis et al., 2005) have used the absence of γ-actin as a marker of axonal RNA purity as this is restricted to neuronal cell bodies (Bassell et al., 1998). However, several bands were consistently detected in the present samples by using primers for Xenopus γ-actin, making this approach unsuitable here. As a consequence, both axonal and cell-body samples were assayed for both genomic DNA as well as Trx mRNA because we identified the latter as a candidate axonal mRNA (see below). Assaying the genomic DNA samples by PCR for a 5Ј flanking sequence of the zinc-finger transcription factor GATA2 (Brewer et al., 1995) revealed a strong signal in the cell body samples, whereas levels were insignificant in the axonal preparation (Fig. 9). By contrast, reverse transcriptase (RT)-PCR revealed that Trx cDNA were abundant in both axonal and cell-body samples, confirming that any cellular contamination was either absent or negligible.
RT-PCR of cDNAs generated from mRNAs of the outgrowing axons in matrigel by using primers that recognize seven cDNAs that correspond to proteins identified in the radioactive spots (and, therefore, presumably retrogradely transported) showed that mRNAs for all seven proteins were present in the axons (Fig. 10). This suggests that most proteins in the radioactive spots on the 2D gels can be translated in axons and subsequently be transported to the soma.
The retrogradely transported proteins comprised several different classes, including enzymes and cytoskeletal molecules. However, apart from dynein and vimentin, which are known to be involved in retrograde axonal transport (Vallee et al., 2004;Perlson et al., 2005), none of the signalling molecules thought to be involved in initiating axonal regeneration -such as STAT3, JNK or Erk1/2 (reviewed by Hanz and Fainzilber, 2006) -was detected, although the possibility that such factors are first synthesized in the soma and then anterogradely transported along axons before returning by retrograde transport cannot be excluded. However, our finding that Trx might be retrogradely transported could be relevant because in addition to functioning as an anti-oxidant, it also regulates activity of various signalling molecules including apoptosis signalling kinase, JNK and PKC. were also seen in autoradiographs of gels from other experiments. The silverstained gel in B showns many more spots, of which only a small subset correspond to those on the autoradiograph (arrows). factors, including NF-κB, activator protein 1 (AP1) and p53 (reviewed by Yoshioka et al., 2006), that are involved in axonal regeneration (Di Giovanni et al., 2006;Raivich et al., 2004;Gallagher et al., 2007).

Cyclopentenone prostaglandins block the in vitro conditioning effect on axonal regeneration
To investigate possible involvement of Trx in axonal regeneration, we incubated PN-DRG for 1 day with 50 μM prostaglandin A1 (PGA 1 ), or with 5 μM 15-deoxy-⌬ 12,14 prostaglandin J 2 (15d-PGJ 2 ) that covalently binds Trx (Moos et al., 2003;Shibata et al., 2003), and observed significant inhibition of subsequent BDNF-induced axonal outgrowth in the presence (P<0.001 and P<0.01, respectively) or absence (P<0.001 and P<0.05, respectively) of ActD. By contrast, incubation of PN-DRG with these prostaglandins for 1 day after 3-day in vitro conditioning did not inhibit subsequent axonal outgrowth, even in the presence of ActD (Fig. 11). Moreover, in three separate experiments using the two- Proteins were excised from radioactive spots on 2D gels of DRGs following metabolic labelling in the distal part of the peripheral nerve in a 2-chamber culture system. Note that the retrogradely transported proteins (with the exception of vimentin) differ from those up-regulated in the cultured DRGs (listed in Table 1). pI, isoelectric point.
compartment culture system, application of PGA 1 or 15d-PGJ 2 to the distal part of the peripheral nerve of PN-DRG for 3 days reduced subsequent BDNF-induced axonal outgrowth in the presence of ActD from 594±47 μm (controls) to 312±17 μm and 406±31 μm (P<0.01 and P<0.05; respectively). These results are consistent with the hypothesis that Trx is involved in regulating expression of RAGs and axonal regeneration, although involvement of other protein(s) bound by these cyclopentenone prostaglandinsincluding the redox-sensitive transcription factors Jun, p53 and NF-κB (reviewed by Kim and Surh, 2006) that are known to be involved in axonal growth (Raivich et al., 2004;Di Giovanni et al., 2006;Gutierrez et al., 2005;Gallagher et al., 2007) -should also be considered.
To determine whether Trx synthesis is required for axonal regeneration, we electroporated FITC-labelled MOs recognizing Xenopus Trx and zebrafish embryonic β-globin (as control) into dissociated DRG neurons, and observed fluorescence -often localized to nuclei -in most neurons (Fig. 12). DRG neurons showed Trx-like immunoreactivity over both soma and neurites (Fig.  12) which in three separate experiments and after 2 days in vitro, was 44±4% lower in the cultures treated with Trx MOs than in those threated with β-globin MOs. In cultures from freshly dissected DRG, the mean neurite outgrowth distance from neurons electroporated with Trx MOs (148±28 μm) was significantly less (P<0.05) than from those electroporated with β-globin MOs (250±40 μm). By contrast, there was no significant difference between the mean distances of neurite outgrowth from neurons dissociated from 3-day in vitro conditioned DRG following electroporation with MOs using Trx or β-globin (501±28 μm or 543±10 μm, respectively).

Changes in gene expression after axotomy or culture
The present findings indicate that changes in gene expression following axotomy are essential for axonal regeneration because axonal growth in response to BDNF and other neurotrophins in Journal of Cell Science 121 (15) Fig. 9. Absence of cellular contamination of axonal RNA samples. Agarose gel stained with ethidium bromide. A strong signal for GATA2 genomic sequence is seen in the DRG samples only. By contrast, Trx can readily be detected in cDNA prepared from both DRG and axonal samples (+RT). The predicted sizes for GATA2 and Trx amplicons are 216 bp and 226 bp, respectively. No products were amplified when reverse transcriptase was omitted from the reactions (-). Lanes at the edges of the figure represent HpaII-digested Bluescript vector as a size marker. Fig. 10. Detection of axonal mRNAs by RT-PCR. Seven representative retrogradely transported axonal proteins that comprise elongation-initiation factor 4A (eIF4A), elongation-initiation factor 5A (eIF5A), Trx, vimentin 4 (Vim4), β-tubulin at 56 D (betaTub56D), calpactin 1 light chain (AnxA2), and dynein light chain 1 (Dlc1) were selected to determine whether their corresponding transcripts were present in axonal RNA extracts using primers described in Table 1. The predicted sizes of the amplified products of these cDNAs are 175,163,226,150,163,200 and 223 bp respectively. Although there is variability in the levels of each of the amplicons, expression can be seen in each case (+). No products were amplified when reverse transcriptase was omitted from the reactions (-). M, HpaII-digested Bluescript vector as a size marker. or not present prior to incubation in gels for 3 days with BDNF. Incubation with the prostaglandins on day 1 strongly inhibited subsequent BDNF-induced axonal growth (in the presence or absence of ActD), but not if added on day 4. Enhancement of axonal regeneration freshly dissected PN-DRG was almost abolished by the inhibition of mRNA synthesis, but not from in vivo or in vitro conditioned preparations. These observations suggest that mRNAs that encode proteins that are required for axonal growth are synthesized in response to axotomy or culture in vitro within 3 days, and are sufficiently stable to sustain regeneration thereafter for several days. Transcription-independent neurite outgrowth from in vivo and in vitro conditioned mammalian DRG neurons, and also NGF-primed PC12 cells has been reported previously (Smith and Skene, 1997;Twiss and Shooter, 1995).
To investigate changes in gene expression, we used oligonucleotide microarray analysis on RNA extracted from fresh and cultured DRG, and found extensive changes in gene expression. Of the 58 genes that show increases in expression of more than fivefold in cultured DRG (Table 1), at least seven, including arginase, galectin, cyclin b1, metallothionein and glutathione Stransferase are upregulated in mammalian sensory or sympathetic ganglia following axotomy (Costigan et al., 2002;Tanabe et al., 2003;Cameron et al., 2003;Boeshore et al., 2004;Nilsson and Kanje, 2005), which suggests their involvement in evolutionarily conserved programs of axonal regeneration and/or neuronal survival. Arginase has neuroprotective properties and its forced expression promotes neurite outgrowth on myelin, normally inhibitory to axonal regeneration (Cai et al., 2002). Expression of galectin is correlated with axonal regeneration, which it might influence by stimulating macrophages to release factor(s) that promote axonal growth and migration of Schwann cells (Horie et al., 2005). Both glutathione-S-transferase and metallothionein have neuroprotective functions (Townsend and Tew, 2003;Penkowa et al., 2005). In cultured DRG, upregulation of other genes that might be important for axonal regeneration include vimentin, which is involved in retrograde axonal transport of activating signals for initiation of axonal growth (Hanz et al., 2003;Perlson et al., 2005), and the cytoskeletal proteins paxillin and calponin (Huang et al., 2004;Plantier et al., 1999). Vg1RBP is also of interest because, in developing neurons, it is involved in the translocation of mRNAs to growth cones Piper et al., 2006) but might also be important during axonal regeneration.
It is unclear which signalling pathways are involved in the control of gene expression following axotomy. Extensive studies of GAP43, the best known RAG, have not identified any factor(s) that are directly responsible for its observed upregulation during regeneration. However, nerve-injury-induced expression of GAP43 and Tα1 tubulin (also known as Tuba1a) is absent in CAATenhancer-binding protein β (C/EBPβ)-knockout mice and moreover, Tα1 tubulin is a direct C/EBPβ target (Nadeau et al., 2005). Other RAGs whose expression might be regulated by C/EBPβ include arginase (Gray et al., 2005) and SPRR1A (Preverand et al., 2004). C/EBPβ is transcriptionally regulated by STAT3 (Niehof et al., 2001) and can be activated by phosphorylation through different pathways involving MAPK, PKC, CAM kinase and Erk1/2 (Ramji and Foka, 2002;Park et al., 2004). Since phosphorylated Erk1/2 and STAT3 are retrogradely transported in injured snail and mammalian nerves (Sung et al., 2001;Perlson et al., 2005;Lee et al., 2004) these could promote C/EBPβ activity in lesioned nerves, leading to expression of specific RAGs. However, in the present study, pharmacological inhibition of MAPK, PKC, CAM kinase and Erk1/2 pathways failed to prevent the in vitro conditioning effect on axonal regeneration and although this was blocked by inhibition of JAK/STAT signalling using AG490, its specificity is uncertain (Gorina et al., 2007).

Retrograde signalling
The mechanisms by which peripheral nerve lesions cause changes in gene expression in neurons might involve retrograde activating signals that are generated at the site of the axon lesion (reviewed Note that immunofluorescence of Trx is reduced in the Trx MO-treated neurons and also that in this latter case the extent of neurite outgrowth is significantly reduced in freshly dissociated (but not conditioned) neurons. Scale bars, 50 μm. Magnifications of A-F and I, G and J, H and K are identical.
by Hanz and Fainzilber, 2006). To investigate whether effects of in vitro conditioning on subsequent axonal growth depend on such signals, preparations were incubated in two-compartment culture dishes, enabling the aplication of drugs to either the DRG or the distal end of the peripheral nerve. Here we show that blockade of protein synthesis in the distal part of the nerve reduces subsequent BDNF-induced axonal growth in the presence of ActD. This suggests that the retrograde signal is generated, at least in part, by local translation from existing mRNAs because local application of ActD to this part of the nerve did not block the conditioning effect. These observations are consistent with previous findings (Hanz et al., 2003;Perlson et al., 2005) that demonstrate the importance of intra-axonal synthesis of β-importin and vimentin in retrograde signalling, and initiation of axonal regeneration. However, other proteins that might be carried by importin-dynein complexes were not identified in those investigations. Retrograde transport of ~30 proteins was demonstrated in cultured frog nerves by Edbladh et al. (Edbladh et al., 1994), and of >100 proteins in lesioned nerves of the mollusc Lymnia (Perlson et al., 2004), although the identities and possible roles of those proteins in axonal regeneration are largely unknown.
In this study, we have demonstrated the retrograde axonal transport of ~35 proteins, 32 of which were identified -including nine that are known to be synthesized within regenerating mammalian axons (Willis et al., 2005). However, we did not detect any of the signalling molecules that are thought to mediate the expression of RAGs following axotomy in vivo (reviewed by Hanz and Feinzilber, 2006). Nevertheless, the possible retrograde transport of Trx is of interest because it influences the activity of various transcription factors (reviewed by Yoshioka et al., 2006), including NF-κB, which is important for axonal growth in developing neurons (Gutierrez et al., 2005;Gallagher et al., 2007), and also p53 and the AP1 transcription factors Jun and ATF3, which are known to be involved in axonal regeneration wihtin mature neurons (Di Giovanni et al., 2006;Raivich et al., 2004;Seijffers et al., 2006;Seijffers et al., 2007). Trx might thus have a pivotal role in regulating the activities of several transcription factors that are important for axonal regeneration. It is interesting to note that, in PC12 cells, Trx is induced by NGF and is essential for neurite-like outgrowth (Bai et al., 2003).
To investigate the involvement of Trx, we used PGA 1 and 15d-PGJ 2 , which both react with Trx (Moos et al., 2003;Shibata et al., 2003), and found that incubation of freshly dissected PN-DRG with these prostaglandins on day 1 of a 4-day period of in vitro conditioning resulted in the inhibition of subsequent BDNF-induced axonal growth. Incubation on day 4, however, did not, which suggests that these electrophilic prostaglandins had blocked -during the first days in culture -synthesis of mRNA that is necessary for subsequent axonal growth. We have also shown that application of these prostaglandins to the distal part of nerves in the twocompartment-culture system reduced any subsequent BDNFinduced axonal outgrowth from PN-DRG in the presence of ActD. However, in addition to Trx, 15d-PGJ 2 also binds to several redoxsensitive transcription factors (reviewed by Kim and Surh, 2006) including NF-κB, Jun and p53, which are all known to be involved in axonal growth (Gutierrez et al., 2005;Gallagher et al., 2007;Raivich et al., 2004;Di Giovanni et al., 2006). PGA 1 reacts with both the Trx receptor (TrxR) and Trx (Moos et al., 2003), and other targets, which seem to be similar to those of 15d-PGJ 2 (see Gayarre et al., 2005). To determine directly whether Trx is involved in axonal regeneration we suppressed its synthesis using specific MOs and observed reduced neurite outgrowth from dissociated neurons of freshly dissected DRG but not from in vitro conditioned DRG.
In conclusion, this study demonstrates that de novo mRNA synthesis is essential for axonal regeneration and it can be induced by culture in vitro as well as axotomy in vivo. Our results also indicate that the TrxR-Trx system and possibly other redox-sensitive factors are involved in axonal regeneration, probably by regulating the expression of RAGs, although the possibility that the cyclopentenone prostaglandins inhibit axonal regeneration by other mechanisms is not excluded. However, whether Trx acts as a retrograde signal within lesioned axons during initiation of axonal regeneration or whether it is involved globally remains to be determined.

Materials and Methods
Reagents were purchased from Sigma (Poole, UK) unless otherwise stated. Xenopus laevis were supplied by Blades Biologicals Ltd (Edenbridge, UK), Aquatic Environment Services (Tonbridge, UK), Neil Hardy Aquatica Ltd (Carshalton, UK) or bred at King's College London and the University of Bristol (UK). Surgical procedures were licensed by the British Home Office and approved by the Animal Care and Use Committee of King's College London. Post-metamorphic animals (~1.5-2 cm body length) were anaesthetized by immersion in an aqueous solution (0.1% w/v) of tricaine methanesulphonate (MS 222) and killed by destruction of the brain. Lumbar and thoracic PN-DRG were removed and either incubated free floating for 3 days for in vitro conditioning or cultured in batches of four to five in shallow collagen gels or matrigel (Beckton Dickenson, Oxford, UK) as described previously (Tonge et al., 1998). For in vivo conditioning, animals were anaesthetized with MS 222 and the sciatic nerve in one leg was exposed and sectioned, followed by wound closure with medical adhesive (Histoacryl, Melsungen, Germany) 3 or 5 days prior to removal of the PN-DRG for culture. Neurotrophic factors: NGF (PeproTech EC, London, UK), BDNF (RandD Systems Europe Ltd, Abingdon, UK), NT3 and NT4 (Alomone Laboratories, Jerusalem, Israel) were used at 50 ng/ml. Actinomycin D (ActD), cycloheximide (CHX) and α-amanitine (α-AM; Merck Chemicals, Nottingham, UK), were used at concentrations of 5 μg/ml, 10 μg/ml and 50 μg/ml, respectively. ActD inhibits RNA polymerases I, II and III, whereas α-AM inhibits RNA polymerase II and, at the concentrations used in this study, reduces RNA synthesis in Xenopus skeletal muscle by ~90% and 70%, respectively (Brehm et al., 1987). PGA 1 or 15d-PGJ 2 (Axxora Ltd, Nottingham, UK), AG490, PD98059, UO126, SP600125, LY294002, k252a (Merck Chemicals, UK) and vinblastine were used as described in individual experiments.
In some experiments, the effects of drug application to either DRG or the distal 2-3 mm end of the peripheral nerve were studied using a two-compartment culture system (Edbadh et al., 1994), comprising a 35-mm plastic dish and inner cylindrical compartment with a small notch for the nerve, coated with silicone grease to prevent leakage between compartments. After 3 days, PN-DRG were removed, the peripheral nerve trimmed followed by culture in collagen gels.
To obtain isolated axons for RT-PCR, PN-DRG were cultured in matrigel on membranes with 8-μm pores (Nuclepore track-edged membranes; Thermo Fisher Scientific Inc, Loughborough, UK) glued to the bottom of 4-well culture dishes with a drop of matrigel and folded in half to provide a vertical barrier facing the cut ends of the peripheral nerves. In response to BDNF, axons extended through the pores (unaccompanied by cells) and into matrigel beyond the barrier, and were harvested after 10 days by removing the folded membrane (together with the PN-DRG), leaving the outgrown axons in the matrigel on the bottom of the culture dish.

MO-induced inhibition of Trx expression
To obtain isolated neurons, DRG were removed from 2-3 animals, trimmed and incubated in 0.125% collagenase type III (Worthington) for 4 hours at 24°C, followed by trituration in RPMI medium containing 10% bovine serum. Following centrifugation, cells were re-suspended in 100 μl of Rat Neuron Nucleofector TM solution (Amaxa Biosystems, Germany), diluted to 80% with water, containing 2 μl of 1 mM 3Ј-labelled carboxyfluorescein MOs followed by electroporation using program O-05 of a Nucleofector TM (Amaxa Biosystems). After electroporation, cells were cultured for 2 days in 4-well dishes (Nunc) pre-coated with 10 μg/ml laminin in PBS and containing 10 ng/ml of BDNF in RPMI medium. The MO sequences were 5Ј-ATGTG -TCGAACCATATTGGCGTTTG-3Ј (Xenopus Trx1) and 5Ј-ATCCAGCACCCA -TCTGAGCAGCAAC-3Ј (zebrafish embryonic β-globin). The β-globin MO was a kind gift of Adam Rodaway (King's College, London, UK). Additionally, some dissociated DRG neurons were cultured for 1 day without electroporation. Enhancement of axonal regeneration tubulin (LeDizet and Piperino, 1991) and axons visualized using peroxidaseconjugated avidin as described (Tonge et al., 1998). Stained preparations were mounted in Glycergel (DakoCytomation, High Wycombe, UK) and photographed using a microscope equipped with phase contrast (BH2; Olympus, Japan) and C5810 digital camera (Hamamatsu Photonics, Welwyn Garden City, UK). Dissociated DRG neurons were fixed after 1 or 2 days in culture with the same fixative for 30 minutes, blocked with 3% (w/v) bovine serum albumin (BSA) and 0.1% (v/v) Triton X-100 in PBS, prior to incubation for 2 hours with a rabbit polyclonal antibody to human Trx (Abcam), diluted 1:500. Following washing with PBS, preparations were incubated with TRITC-conjugated goat anti rabbit IgG (1:300 dilution) for 1 hr, washed again with PBS, mounted in Vectashield containing 4Ј,6-diamidino-2phenylindole (DAPI, Vector Laboratories, Burlingame, USA), viewed using an Eclipse TE200 fluorescence microscope and images captured using a DXM1200F digital camera (both from Nikon, Tokyo, Japan). Cross-reactivity of the antibody for Xenopus Trx was verified in one experiment by western blotting, showing recognition of a 12 kDa band in protein extracts from both mouse and Xenopus heart (data not shown). Moreover, in a further experiment, prior incubation of the antibody with recombinant human Trx reduced subsequent mean fluorescence labelling of dissociated Xenopus DRG neurons by 65% (36 neurons). Relative expression levels of Trx were assessed from images (captured using the same exposure time) by comparing the averaged immunofluorescence intensity of labelled neuronal somas against background along orthogonal axes using a PC version of NIH Image (Scion Image).

Outgrowth measurements
After staining, PN-DRG preparations were viewed under a Nikon TMS invertedphase-contrast microscope. Where large numbers of outgrowing axons were present (forming a halo), distances from the cut end of the peripheral nerve to the tips of the outgrowing axons were measured at five separate points, using an eye piece scale and graticule, and their mean calculated for each PN-DRG. Where axonal outgrowth was sparse, the lengths of the five longest axons from the cut end of the peripheral nerve of each preparation were measured and their mean calculated. In some experiments, axonal outgrowth was visualized by fluorescence microscopy following overnight incubation with 1 μg/ml Calcein-AM or Calcein Orange-Red (Invitrogen, Paisley, UK), membrane-permeable dyes which are hydrolysed by intracellular esterases in viable cells yielding green or orange fluorescent products, respectively. Some preparations were fixed, mounted in Vectashield containing DAPI, and viewed using the fluorescence microscope with the digital camera described above. In cultures of dissociated DRG neurons, neurite outgrowth was quantified from 15-20 digital images per well by measuring the longest neurite of neurons, labeled by Calcein, using Scion Image.

Oligonucleotide microarray analysis
Total RNA was extracted from two separate pools, each comprising about 60 freshly dissected (control) or 3-day in vitro conditioned DRG using Absolutely mini RNA kits (Stratagene, Cambridge, UK) according to the manufacturer's instructions. RNA quality was assessed on an Agilent 2100 bioanalyser using RNA LabChips according to the manufacturer's instructions. RNA from control and in vitro conditioned DRG was used to generate biotinylated cRNA according to Affymetrix protocols and hybridized with separate Affymetrix Xenopus Genechips, comprising >14,400 transcripts (Affymetrix UK High Wycombe, UK). The arrays were scanned and analysed at the Genomics Centre, King's College London, using a Hewlett Packard Gene Array Scanner and Affymetrix MicroArray Analysis Suite 5.0 software. Standard software settings were used to establish the presence or absence of transcripts, relative mRNA abundance, and also to determine the fold differences in transcript abundance in the RNA samples.

Quantitative PCR
To validate the microarray data, Quantitative PCR (qPCR) was carried out on cDNA that had been prepared by using random decamers from samples of RNA used in the microarray analysis, using avian myeloblastosis virus reverse transcriptase (AMV RT) according to the manufacturer's protocol (Promega, Southampton, UK). Control reactions, omitting AMV RT, were also performed. Relative expression levels of different transcripts were quantified (in triplicate) by qPCR using fluorescent SYBR Green 1 technology on an ABI PRISM 7000 HT sequence detection system/LightCycler (Applied Biosystems) and normalized to those of 18S RNA. Primers (supplementary material Table S2) were designed using Primer Express software (Applied Biosystems, Warrington, UK) with default settings, applied to Xenopus cDNA sequences derived from the NCBI web site http://www.ncbi.nlm.nih.gov. PCR reactions were performed on a MicroAmp Optical 96-well reaction plate and the Ct method was used for relative quantification as described (Brewer et al., 2005).

Non-quantitative PCR and reverse-transcription PCR
For analysis of the axonal RNA panel, total RNA was also extracted from isolated axon preparations in matrigel and used to generate cDNA as described above. Control reactions omitting AMV RT were also performed. cDNAs were used as templates for PCR reactions using 40 cycles and primers designed using Primer Express software (Applied Biosystems, UK) applied to Xenopus sequences derived from the NCBI web site. To confirm the purity of axonal RNA, individual preparations were divided into two pools for parallel RNA and DNA analyses. From one pool, total RNA was prepared from DRG and axon samples, using the Absolutely RNA T Microprep Kit (Stratagene). For conventional RT-PCR, cDNA preparation was performed as described by Weber et al. (Weber et al., 2000). To control for contamination with genomic DNA, parallel reactions were carried out, with and without AMV RT. PCR was carried out using GoTaq premix (Promega) according to the manufacturer's instructions.
For genomic DNA analysis, samples from the second pool were diluted in 0.5 ml 50 mM Tris (pH 7.5), 1 mM EDTA, 0.5% Tween with 200 μg/ml proteinase K and digested overnight at 56°C They were then extracted with phenol/chloroform, precipitated with isopropanol (tRNA used as carrier), and resuspended in Tris-EDTA. PCR (as described for RT-PCR, except the annealing temp was 58°C). The primer sequences for all PCR and RT-PCR reactions are shown in supplementary material Table S2.
Detection of retrogradely transported proteins by metabolic labelling and 2-DE PN-DRG preparations (eight to ten per experiment) were placed in the twocompartment culture system described above; to each inner compartment, containing 2-3 mm of the distal peripheral nerve, 0.32 MBq [ 35 S]methionine-[ 35 S]cysteine (GE Healthcare, Chalfont St Giles, UK) was added. After 2 days in culture the DRG were pooled and solubilized in DeSteak Rehydration Solution with 0.5% Pharmalyte (pH 3-pH 10) (GE Healthcare, UK) and separated by iso-electric focusing (IEF) using pH 3-pH 10 Immobiline DryStrips (GE Healthcare, UK) for the first dimension. After focusing, the strips were equilibrated in buffer containing 6 M urea, 50 mM Tris-HCl (pH 8.8), 30% v/v glycerol, 2% SDS, with 1% w/v DTT for 15 minutes, followed by 2.5% IAA in the same buffer for a further 15 minutes. In the second dimension, the IEF strips were run on 12% SDS-PAGE. Gels were then fixed with 50% methanol and 10% acetic acid, followed by transfer to 30% methanol 3% glycerol overnight. The next day, gels were dried and exposed to Xray films for up to 30 days. After the development of the films, gels were re-swelled and silver-stained using a protocol compatible with mass spectrometry (Yan et al., 2000). Autoradiographs and images of silver-stained gels were scanned at 360 pixels per inch resolution using an EPSON GT-7000 scanner and images compared manually.

Analysis of 2DE spots by LC-MS/MS
Following autoradiography and silver-staining, spots of interest were excised from gels and subjected to reduction, alkylation and digestion (with trypsin) prior to subsequent analysis by LC-MS/MS using published procedures (Hye et al., 2006;Gallois-Montbrun et al., 2007). Briefly, peptides were extracted from gel pieces by a series of acetonitrile and aqueous washes, lyophilized and re-suspended in 50 mM ammonium bicarbonate before analysis by LC-MS/MS. Chromatographic separations were performed using Ultimate LC or Ultimate 3000 systems (Dionex, Camberley, UK). Peptides were ionised by electrospray ionization using Z-spray sources fitted to a Q-Tof micro (Waters Corporation, Manchester, UK) or a QTRAP 4000 (Applied Biosystems, UK). Both instruments were set to run in automated switching mode, selecting precursor ions based on their intensity, for sequencing by collision-induced fragmentation. MS/MS analyses were conducted using collision energy profiles that were chosen based on the m/z and the charge state of the peptide. Mass spectral data were searched against the Swiss Prot (Uniprot release 52.40, NCBI non-redundant (release 157) and Xenopus laevis (in-house) databases using Mascot software (v2.2, Matrix Science, London, UK). Data searching was performed using specific aminoacid-modification parameters (e.g. variable cysteine carbamidomethylation and methionine oxidation). Assumptions included peptide-mass tolerance of 1 Da, fragmentation tolerance 0f 0.5 Da and a maximum of two missed cleavages. MS/MSsequence information was obtained for all peptides included in the results (MOWSE algorithm score of 95% confidence) based on matching evidence of MS/MS fragmentation to protein sequences in the databases. Where reported matches are below 95% confidence limits, each peptide was visually verified before inclusion. The percentages of sequences covered are given in Table 2.

Statistical analysis
Experiments were repeated at least three times unless otherwise stated. Differences between means were evaluated throughout using Student's t-test and considered significant at P<0.05. This work was supported by the BBSRC.