Injury-induced decline of intrinsic regenerative ability revealed by quantitative proteomics

Functional annotation of injury-altered proteins

Next, we used Ingenuity Pathway Analysis (IPA) (Park et al., 2013) to assess which cellular functions might be related to these altered 62 proteins that are affected in the injury response. As shown in Fig. 2A and Table S3, most down-regulated proteins appear to be associated with cellular functions such as formation of cellular protrusions, organization of cytoskeleton and microtubule dynamics. Since these processes represent critical steps of axon regeneration, these changes might reflect the reduced intrinsic regenerative ability of injured RGCs.

Open in a separate window

Figure 2

Functional annotation and network analysis of injury-altered proteins in RGCs

(A) Protein ontology categories enriched in the proteins affected by injury. IPA analysis indicates the up-regulation (in orange), down-regulation (in blue) or non-predictable (in white) function variation after injury for each cluster. The size of each square is proportional to the number of altered proteins in this functional category. (B) Major hubs from the merged network analysis of injury-altered proteins. Factors and/or pathways of interest are added in this spider-web, based on their targets. (C) Table of major regulators ranked based on their numbers of total interactions (nodes) in the interaction web.

Putative key regulators of injury responses revealed by informatics analysis

We reasoned that such protein abundance alterations might reflect the outcomes of injury-response signaling events. Therefore, we subjected these 62 proteins to an IPA-based network analysis (Staab et al., 2007), in order to find major signaling events occurring in axotomized RGCs. This analysis revealed 9 signaling networks (Figure S2C, Table S4,) and relevant signaling hubs (Figure S2C). The top 12 signaling hubs were listed in Figures 2B and 2C.

Strikingly, this list includes several known regulators of neuronal survival and axon regeneration, such as TP53, Ca²⁺ (Ghosh-Roy et al., 2010, Chierzi et al., 2005; Ghosh-Roy et al., 2010; Sung et al., 2001; Bradke et al., 2012), MAPK (Chen et al., 2011; Hammarlund et al., 2009; Watkins et al., 2013), JAK (Smith et al., 2009; Sun et al., 2011) and the components of mTOR pathway, such as Rictor, Raptor and mTOR (Dupraz et al., 2013; Liu et al., 2010; Near et al., 1992; Park et al., 2008; Toth et al., 2006; Zhu et al., 2010), validating the use of this approach in analyzing neuronal injury responses. Interestingly, this list also includes several other molecules, such as c-myc, Huntingtin (HTT) and NFkB (Figure 2B and 2C), which might represent new candidates for critical regulators of the neuronal injury response.

c-myc expression is down-regulated in RGCs in response to injury

Our further analysis has been focused on c-myc, which was one of the most connective nodes of the regulator network (Figure 2B, 2C and table S4). c-myc was not among the 1409 proteins identified by proteomics analysis (Table S1 and S2), possibly due to its low protein abundance in adult RGCs. However, among 62 injury-altered proteins identified from our proteomics analysis, 10 proteins (highlighted in red in Table S4) have functional interactions with c-myc in the network. Importantly, similar to mTOR (Howell et al., 2013; Laplante and Sabatini, 2012; Shaw and Cantley, 2006), c-myc has been implicated as a master control gene of cellular metabolism (Dang, 2013), positioning it as an excellent candidate to improve RGCs survival and axon regeneration.

To verify the involvement of c-myc in the injury response, we examined the expression level of c-myc before and after injury. Using real time qPCR analysis of mRNA purified from samples subjected to identical treatments as the proteomics experiments (Figure 1), we observed that the expression of c-myc mRNA is decreased by 70% after injury (Figure 3A, B). This injury-induced c-myc down-regulation was also verified by immunohistochemistry with anti-c-myc antibodies on retinal sections from intact and injured mice (Figure 3C). These results together verified the injury-induced c-myc down-regulation in injured RGCs.

Open in a separate window

Figure 3

c-myc expression promotes neuronal survival and axon regeneration after optic nerve injury

(A) Timeline of the experimental procedures. (B) c-myc mRNA levels in intact and injured RGCs detected by real time q-PCR. mRNA samples were extracted from purified RGCs isolated from YFP⁺ retina at 3 days post injury or from intact control retinas. Results are normalized to GAPDH. **: p < 0.01, T-test. Error bars: S.E.M (C) Representative images of retinal sections stained with anti-c-myc (in red), anti-Tuj1 (in green) antibodies, and DAPI (in blue). Scale bar: 20 μm. (D) Scheme of experimental procedures to assess axonal regeneration. (E) Representative confocal images of CTB-labeled optic nerve sections from Rosamer-c-myc mice with intravitreal injection of AAV-PLAP (n = 5) or AAV-Cre (n =7). Red stars indicate the crush site. Scale Bar: 100um. (F) Quantification of the numbers of regenerative axons counted at different distances distal from the lesion. ***: p < 0.001, **: p < 0.01 and *: p < 0.05, T-Test. Error bars: S.E.M (G) Quantification of survived RGCs as detected by anti-Tuj-1 immunostaining in whole mount retinas from intact Rosamer-c-myc mice (n = 5), injured Rosamer-c-myc mice with injection of AAV2-PLAP (n = 5) or AAV2-Cre (n = 5). ***: p < 0.001, ANOVA with Bonferroni post test.

Over-expression of c-myc in RGCs promotes axon regeneration in vivo

We next assessed the effects of c-myc over-expression on neuronal responses by two different approaches. First, as we demonstrated previously, injection of adeno-associated virus serotype 2 (AAV2) vectors into the vitreous body results in highly selective and efficient transduction to RGCs (Park et al., 2008; Sun et al., 2011), thus we injected AAV2 expressing c-myc or a control placenta alkaline phosphatase (PLAP) into the vitreous body of adult mice. These mice were then subjected to a standard optic nerve injury procedure, followed by a Cholera Toxin B subunit (CTB) injection to analyze neuronal survival and axon regeneration (Park et al., 2008). At 2 weeks post-injury, we observed that 49% of RGCs survived in c-myc over-expressing mice, whereas only 25% survived in the mice injected with AAV2-PLAP (Figure S3A, B). Importantly, a significant axonal growth increase was observed in the optic nerves from c-myc over-expressing mice (Figure S3C, D), suggesting that c-myc over-expression could promote both neuronal survival and axon regeneration after injury.

As an independent approach to assess the effect of c-myc over-expression, we used an inducible c-myc transgenic mouse line, ROSAMER mice (Jager et al., 2004). In these mice, a cassette with the gene encoding a tamoxifen-inducible c-mycERT fusion protein (mycERT) down-stream of a lox-stop-lox sequence was inserted in the ubiquitously active ROSA26 locus. Therefore, mycERT transcription is Cre-dependent and the activity of the fusion protein is tamoxifen-inducible. We then tested the effect of a transient activation of c-myc prior to injury. After the intravitreal injection of AAV2-Cre or AAV2-PLAP, and subsequent tamoxifen induction, these mice were subjected to optic nerve injury (Figure 3D). As shown in Figure 3E–G, both robust neuronal survival and axon regeneration were observed in the mice injected with AAV2-Cre, but not with control AAV2-PLAP. The extent of the regenerative response was comparable to that observed in PTEN-deleted mice (Park et al., 2008). These phenotypes are more robust than what was observed after AAV2-c-myc expression (Figure S3), possibly due to a more efficient expression of c-myc in these transgenic mice. We also monitored the mTOR activity in RGCs under these treatments, three days after injury, by immunohistochemistry using antibodies against phosphorylated-ribosome protein S6 (P-S6), an established indicator of the mTOR activity (Guertin and Sabatini, 2007). As shown in Figure S4, RGCs with c-myc over-expression still exhibited significant injury-induced mTOR down-regulation, although the extent of down-regulation is less than that observed for the controls. These data further support the notion that c-myc over-expression is able to promote RGCs axons regeneration after optic nerve injury.

Delayed c-myc over-expression still promotes optic nerve regeneration

To mimic clinically relevant settings, we assessed the effects of expressing c-myc after a delayed post-injury period. We thus administered tamoxifen to ROSAMER mice starting from one day after optic nerve injury (Figure 4A). At two weeks post-injury, we observed that approximately 55% of RGCs survived, in comparison with 25% in controls (Figure 4D and 4E). In addition, significant numbers of regenerating axons were observed in the animals with delayed c-myc expression with many regenerating axons growing beyond the injury site and extending 1.5mm from the crush site (Figures 4B and 4C). As the apoptotic pathways might be activated at these axotomized RGCs (Figure 2), our results suggest that c-myc can rescue these injured neurons from apoptotic death and promote their axon regeneration.

Open in a separate window

Figure 4

Delayed c-myc expression promotes RGC survival and axon regeneration after optic nerve injury

(A) Timeline for delayed activation of c-myc after the optic nerve injury. (B) Representative confocal images of CTB-labeled optic nerve sections from control (AAV2-PLAP; n= 5) and Rosamer-c-myc (c-Myc) mice (n=7). Scale bar 100μm. (C) Quantification of regenerative axons counted at different distances distal from the lesion. T-Test. ***: p <0.001 and *: p <0.05. (D) Representative images of whole mount retinas, stained with anti-Tuj1, from intact, injured control (AAV2-PLAP; n= 5) and injured c-myc mice (AAV2-Cre, n=7). Scale bar: 20 um. Error bars: S.E.M (E) Quantification of Tuj1+ RGCs in different groups. AI: After Injury. BI: Before Injury. ANOVA with Bonferroni post test. ***: P <0,001.

c-myc is essential for regenerative phenotype of PTEN deletion

The experiments above implicated injury-induced c-myc down-regulation as an important mechanism reducing the intrinsic regenerative ability of adult RGCs. We then investigated whether c-myc is required for axon regeneration in adult RGCs induced by other manipulations. Since PTEN inhibition was previously shown to activate mTOR and promote neuronal survival and axon regeneration (Liu et al., 2010; Park et al., 2008; de Lima et al., 2012), we thus examined whether a genetic knockout of c-myc might affect the axon regeneration induced by shRNA-mediated PTEN silencing (Zukor et al., 2013). We co-injected AAV2-ShPTEN with AAV2-Cre or AAV2-PLAP into the vitreal body of the c-myc^f/f mice and performed optic nerve injury after 4 weeks. As expected, PTEN silencing by ShRNA resulted in significant axon regeneration as analyzed at 2 weeks post-injury (Figure S5A and B). The knockout of c-myc significantly reduced the numbers of regenerating axons (Figure S5A, B), but did not affect neuronal survival (Figure S5C, D) or the mTOR activity (Figure S5E). However, some residual regenerating axons were still observed after c-myc deletion (Figure S5A, B). It remains unclear whether this result reflects the differential involvement of c-myc in RGCs subtypes or the compensatory effects of other myc members.

Synergistic effects of c-myc over-expression and PTEN and SOCS3 deletion

Our systems biology analyses of the pathways involved in the neuronal response to injury strongly suggest that other major signaling hubs work in concert with c-myc (Figure 2B, C). To better understand their functional interactions, we analyzed the combinatorial effects in the optic nerve injury model by manipulating both c-myc and the mTOR pathway, which was also implicated by the network analysis (Figure 2). Thus we performed an intravitreal co-injection of AAV2-Cre together with AAV2-c-myc or AAV2-PLAP, in PTEN^f/f mice (Figure 5A). Two weeks after injury, we observed significant increase in neuronal survival and axon regeneration in the PTEN^f/f mice with c-myc expression with around 75% of RGCs alive (Figure 5B–D).

Open in a separate window

Figure 5

Synergistic effect of PTEN deletion with c-myc overexpression

(A) Timeline of experimental procedure. (B) Representative confocal images of the optic nerve sections from PTEN^f/f (infected with AAV2-Cre), Rosamer-myc (infected with AAV2-Cre and treated with Tamoxifen) and PTEN^f/f mice infected with AAV2-Cre and AAV2-c-myc. Axons are labeled with CTB. Red stars indicate the crush site. Scale bar: 100 μm. (C) Quantification of regenerative axons in the three groups presented in (B). ANOVA with Bonferroni post test. *p<0.05. Error bars: S.E.M. (D) Quantification of RGC survival as measured by Tuj1 staining ANOVA with Bonferroni post test. ***p-value <0,001 and *p<0.05.

As our previous studies showed robust axon regeneration after co-deletion of PTEN and SOCS3 in RGCs (Sun et al., 2011), we further examined the effects of this co-deletion in combination with c-myc over-expression (Figure 6A). Thus we performed an intravitreal co-injection of AAV2-Cre and AAV2-CNTF, together with AAV2-c-myc or AAV2-PLAP, in PTEN^f/f/SOCS3^f/f mice. As shown in Figure 6B, such a combinatorial treatment resulted in further enhanced axon regeneration. At 4 weeks post-injury, about 5 times more regenerating axons reached the proximal end of the chiasm when compared with PTEN/SOCS3 deletion alone (Figures 6B and C and Figure S6). In the mice with PTEN/SOCS3 co-deletion, the majority of regenerating axons ceased growing in the proximal end of the chiasm (Luo et al., 2013; Sun et al., 2011). However, with additional c-myc expression, while some axons grew ectopically into the contralateral optic nerve, many more axons crossed the chiasm and continued to grow in the optic tracts, with some ipsilaterally and more contralaterally (Figure 6B, E), After crossing the chiasm, many regenerating axons derail from the optic tract, similar to that shown by Luo et al., 2013. At 8 weeks post-injury, while most of regenerating axons disappeared in the mice with PTEN/SOCS3 deletion without c-myc expression, many regenerating axons remain in the mice with triple treatment (Figure S7). In these animals, however, the numbers of regenerating axons are reduced to some extent (Figure. S7), possibly due to the elimination of regenerating axons that have not formed proper synaptic connections. These results suggested that co-manipulation of these injury-response pathways could result in a synergistic effect on promoting neuronal survival and axon regeneration.

Open in a separate window

Figure 6

c-myc expression enhances optic nerve regeneration induced by co-deletion of PTEN and SOCS3 in RGCs

(A) Timeline of the experimental procedures. (B) Representative confocal images of CTB-labeled optic nerve whole-mounts from PTEN^f/f/SOCS3^f/f mice with intravitreal injection of AAV2-Cre, AAV2-CNTF and AAV2-c-myc (n=6) or AAV2-Cre, AAV2-CNTF and AAV2-PLAP (n=6). Optic nerve crush was performed on the right optic nerve head and regenerative axons were labeled with CTB-Alexa-555. Red stars indicate the crush site. Scale bar: 100 um. (C, D) Quantification of the numbers of regenerating axon at the proximal end of the chiasm (C) and regenerating axons that passed the chiasm, including those crossing the midline (contra) and navigating ipsilaterally (ipsi) (D). (E) Distribution of regenerating axons projecting in the contralateral optic nerve (gray), the contralateral optic tract (white) and ipsilateral optic tract (black) in the groups of PS (PTEN^f/f/SOCS3^f/f with AAV2-Cre, AAV2-CNTF and AAV2-PLAP) and PSM (PTEN^f/f/SOCS3^f/f with AAV2-Cre, AAV2-CNTF and AAV2-c-myc). T-test, ***: p < 0.001 and *: p < 0.05.

Axon regeneration quantification

Numbers of CTB labeled axons were estimated by counting the numbers of CTB-labeled axons at different distances distal from the crush site in at least 5 sections per animal as described previously (Park et al., 2008; Sun et al., 2011).

Immunohistochemistry

Sections or whole mount retinas were incubated over night at 4°C with primary antibodies: Tuj-1 (1/400 Covance); P-S6 (1/200-Cell Signaling); NeuN (1/100-Millipore) and c-myc (1/200-Abcam) diluted in PBS-3%BSA-0. 5% Triton-X100 as described in Nawabi et al 2010 (Nawabi et al., 2010). Then, tissues were incubated for 2 hours at room temperature with appropriate Alexa-fluor-conjugate secondary antibodies (Alexa-488; Alexa-594, Alexa-674, Invitrogen) diluted with the same blocking solution as the primary antibodies. Sections or whole mount retinas were mounted with DAPI-Fluoromont-G (Solulink).

RGC survival and phospho-S6 signal

Tuj1-positive cells in the ganglion layer were counted using a fluorescent microscope after immunostaining whole mount retinas with anti-Tuj-1 antibodies. 8 random fields per retina were enumerated. The average number per field was determined and the percentages of viable RGCs were obtained by comparing the values determined from the uninjured contralateral retinas. In the same condition, after phospho-S6 staining, the densities of phopsho-S6 positive RGCs were obtained by comparing the value from the uninjured contralateral retinas.

Whole mount optic nerve preparation

Optic nerves were prepared as described in (Dodt et al., 2007). Briefly optic nerves were post fixed overnight at 4°C in PFA4%, then dehydrated in ethanol (50%, 80% and 95% for 20min each) Optic nerves were incubated for 1h in ethanol 100% and then 2h in Hexan (Sigma). Finally samples were transferred in benzyl benzoate/benzyl alcohol (2:1 in volume-Sigma) clearing solution.

Imaging

Optic nerves from different experiments (sections and whole mount) were imaged using a spinning disc confocal microscope (Zeiss) with Velocity software (Perkin Elmer). 20x objective was used to image stacks of all the optic nerves with 20% of overlap between each image. The motorized stage enables the automation of the imaging procedure. All the images were then stitched together using the Velocity software and z-projected to maximal intensity for quantification. Whole mount retinas were imaged with Nikon 80i epifluorescence microscope with 20x objective. At least 8 random fields were imaged per retina.

Statistical analysis

One-way ANOVAs were performed using PRISM software and specific differences between groups were confirmed with Bonferrroni t-test. When two groups were compared, t-test was used.

FACS analysis

YFP-17 mice were subjected to optic nerve injury. 3 days after injury, animals were euthanized and injured or intact retinas were immediately dissected out for dissociation. Retinas were collected in 800μl HBSS supplemented with trypsin (5mg/ml) and DNase (0.2mg/ml) for 10min at 37°C. Retinas were washed 3 times with DMEM/10% FBS and dissociated mechanically using a P1000 pipetman. Cells were pelleted by centrifugation for 5min at 200g suspended in Neurobasal supplemented with L-glutamine and B27 plus DAPI (1mg/ml) and filtered through 40μm cell strainer (BD Falcon) before FACS sorting. FACS sorting was performed with a BD FACSAria IIu instrument. Prior to each sorting event, a purity test was performed to ensure that the specificity for the YFP signal was higher than 99%. Dissociated retinal cells were separated based on size (forward scatter) and surface characteristics (side scatter) as well as viability (DAPI staining). Aggregated cells were excluded based on the FSC-H vs FSC-A ratio. Retinal cells without YFP expression was used as negative controls to optimize the detection gate prior to each sorting. Sorted cells were immediately subjected to either proteins or RNA extraction.

Quantitative Proteomics analysis of RGC

Cells from each preparation were immediately lysed using Qproteome mammalian buffer and proteins were precipitated using chloroform/methanol. 100 ug protein from each sample was reduced with (tris 20 carboxyethyl) phosphine (TCEP) and alkylated with a 0.05 M iodoacetamide solution in the dark at room temperature (23 °C) for 20 minutes. A proteolytic digestion was carried out with trypsin at 37 °C for 16 hours. Labeling reagents from the TMT sixplex^™ Isobaric Label Reagent Set (PI-90064, Thermo Fischer Scientific) were added to the peptide samples and incubated at room temperature for one hour. The reaction was quenched by adding 8 ml of 5% hydroxylamine to the samples, incubated at room temperature for 1 hour and combined together. Peptides were eluted from the spin columns. After acidification, the peptides were desalted on an OASIS HLB column (Waters) and evaporated to dryness, prior to isoelectric focusing fractionation according to their isoelectric point. Resultant peptide fractions were dried before re-suspension in MS loading buffer (5% acetonitrile, 0.1% formic acid in H₂O).

Less than 10ug of each fraction was analyzed by nano-LC-MS/MS using a 60-minute gradient. Pulsed Q-Dissociation (PQD) on the LTQ-Orbitrap Classic was used to fragment the peptide precursor ions. Thermo RAW files were processed using an in-house informatics pipeline (Bigbang/Horizon) and searched with a Mascot (version 2.1) search engine against the IPI mouse protein database (IPI 3.68) using a target-decoy database to calculate the false discovery rate (FDR). TMT 6-plex modifications were set as dynamic modifications in the database search and as static modifications for all other data analysis. This allowed us to calculate labeling efficiencies -the labeling efficiency for our experiments was 98.73%. All PSMs used in this analysis had a MASCOT score above 32 and q value below 0.01. Only proteins identified with a high rate of confidence (1% FDR) were used for analysis. Functional analysis was performed using Ingenuity Pathways Analysis (IPA Ingenuity System). The extended methods are described in Supplemental Information.

We thank Katherine Zukor for reading the manuscript. This study was supported by a fellowship from William Randolf Hearst Foundation (SB), NIH grants EY021242 (ZH) and NS066973 (JS) and a grant from Dr. Miriam and Sheldon G. Adelson Medical Research Foundation (ZH). IDDRC and viral cores supported by the grants NIH P30 HD018655 and P30EY012196 were used for this study.