Nfe2l3 promotes neuroprotection and long-distance axon regeneration after injury in vivo

Abstract

Nuclear factor erythroid 2 like (Nfe2l) gene family members 1-3 mediate cellular response to oxidative stress, including in the central nervous system (CNS). However, neuronal functions of Nfe2l3 are unknown. Here, we comparatively evaluated expression of Nfe2l1, Nfe2l2, and Nfe2l3 in singe cell RNA-seq (scRNA-seq)-profiled cortical and retinal ganglion cell (RGC) CNS projection neurons, investigated whether Nfe2l3 regulates neuroprotection and axon regeneration after CNS injury in vivo, and characterized a gene network associated with Nfe2l3 in neurons. We showed that, Nfe2l3 expression transiently peaks in developing immature cortical and RGC projection neurons, but is nearly abolished in adult neurons and is not upregulated after injury. Furthermore, within the retina, Nfe2l3 is enriched in RGCs, primarily neonatally, and not upregulated in injured RGCs, whereas Nfe2l1 and Nfe2l2 are expressed robustly in other retinal cell types as well and are upregulated in injured RGCs. We also found that, expressing Nfe2l3 in injured RGCs through localized intralocular viral vector delivery promotes neuroprotection and long-distance axon regeneration after optic nerve injury in vivo. Moreover, Nfe2l3 provided a similar extent of neuroprotection and axon regeneration as viral vector-targeting of Pten and Klf9, which are prominent regulators of neuroprotection and long-distance axon regeneration. Finally, we bioinformatically characterized a gene network associated with Nfe2l3 in neurons, which predicted the association of Nfe2l3 with established mechanisms of neuroprotection and axon regeneration. Thus, Nfe2l3 is a novel neuroprotection and axon regeneration-promoting factor with a therapeutic potential for treating CNS injury and disease.

INTRODUCTION

Nfe2l3 (also known as Nrf3) belongs to the Cap’n’collar (Cnc) family of proteins, along with Nfe2l1 and Nfe2l2, which regulate several physiological and pathophysiological processes, including cellular response to oxidative stress^1,2. Nfe2l3 itself is involved in antioxidant response, lipid metabolism, proliferation, and oligodendrocyte differentiation^1-6. Multiple factors upstream and downstream of Nfe2l3 also have been identified^2,5,7. Nfe2l3 knockout (KO) mice develop without overt abnormalities, presumably due to redundancy with Nfe2l1 or Nfe2l2 members of the same protein family⁸. However, Nfe2l3 KO mice challenged with pulmonary injury lose weight⁹, and when challenged with carcinogens produce lymphoblasts¹⁰. In the adult central nervous system (CNS), based on single cell transcriptomics, Nfe2l3 is expressed primarily in oligodendrocytes, but not expressed meaningfully in neurons¹¹. Nevertheless, at least embryonically and neonatally, Nfe2l3 was associated with the development of cortical layer 5 pyramidal near-projecting and corticothalamic projection neurons, respectively^12,13, in which its expression declined during maturation and was silenced by adult age¹⁴. Because in the CNS neuronal intrinsic axon growth capacity declines during developmental maturation^15,16, it is possible that Nfe2l3, whose expression declines during maturation in subsets of cortical projection neurons, may play a role in developmental axon growth of certain types of neurons. Although Nfe2l3 KO mice appear to develop normally (which could be due to compensatory redundancy with Nfe2l1 and Nfe2l2), Nfe2l3 KO mice exhibit phenotypes when challenged^9,10. Thus, we hypothesized that the Nfe2l3 may exhibit axon growth phenotype when challenged by injury. Because oxidative stress is one of the earliest pathological intracellular events in RGCs after axonal optic nerve crush (ONC) injury¹⁷, whereas Nfe2l3 is associated with antioxidant response^1-3, we selected ONC as a challenge for investigating Nfe2l3’s potential neuronal functions in vivo. First, we asked whether Nfe2l3 expression is associated with development of other cortical and retinal ganglion cell (RGC) CNS projection neurons. Then, we tested whether expressing Nfe2l3 in RGCs would elicit neuroprotection and/or promote axon regeneration after challenge by ONC injury, and compared the effects to targeting of prominent regulators of long-distance axon regeneration, Pten and Klf9^18-21. Finally, we characterized gene network associated with Nfe2l3 in neonatal RGCs (before its expression declines during maturation), which provided insights into the potential molecular mechanisms of Nfe2l3 functions in neurons.

RESULTS

Using previously published scRNA-seq datasets of various subdivisions of cortex^11,22,23 (from which we selected cortical projection neurons based on markers, as described below), we analyzed expression of Nfe2l1, Nfe2l2, and Nfe2l3 in cortical projection neurons from different stages of development (embryonic days 14-15²², postnatal day 0²³, and adult¹¹), integrated for visualization of cell clusters using a UMAP (Fig. 1A). Cortical projection neurons were selected based on co-expression of neuronal markers (Rbfox3/NeuN²⁴ and Gap43^25,26) (Fig. 1B-C), expression of glutamatergic (Vglut1²⁷ or Vglut2²⁷) and absence of interneuron (Sst²⁸ and Pvalb²⁸) markers (Fig. 1D-F), and co-expression of cortical projection neuron markers (Uchl1²⁹ and L1cam³⁰) (Fig. 1G-H). Adult cortical layers 5/6 projection neuron markers (Foxp2²⁸, Ctip2/Bcl11b²⁷, Fezf2²⁷) also were enriched in subsets of neurons (Fig. 1I-K). We then found that, whereas Nfe2l1 was expressed robustly in all stages of cortical neuron development, Nfe2l2 was nearly not expressed, bordering noise (Fig. 1L-M). Nfe2l3 expression in cortical neurons, however, peaked only neonatally but declined by adult age (Fig. 1N), which is consistent with Nfe2l3 being previously associated with neonatal development of cortical layer 5 pyramidal near-projecting and corticothalamic projection neurons and then being downregulated by adult age^{12,13 14}.

Figure 1. Expression of Nfe2l1, Nfe2l2, and Nfe2l3 genes in scRNA-seq-profiled cortical projection neurons at different developmental stages.

(A) UMAP of cortical projection neurons, showing cell clusters corresponding to 14-E15 (embryonic), P0 (postnatal), and adult stages of development, as marked (see Methods for details). Immature (embryonic and postnatal) cortical projection neurons are outlined with a dashed-line oval.

(B-N) UMAPs showing gene expression of: neuronal markers Rbfox3/NeuN and Gap43 (B-C), glutamatergic neuron markers Vglut1 and Vglut2 (D), negative for interneuron markers Sst and Pvalb (E-F), cortical projection neuron markers Uchl1 and L1cam (G-H), adult cortical layers 5/6 projection neuron markers, which were enriched in subsets of neurons, Foxp2, Ctip2/Bcl11b²⁷, and Fezf2 (I-K), and Nfe2l1, Nfe2l2, and Nfe2l3 genes (L-N), in clusters of cortical projection neurons from different stages of development (clusters’ developmental ages are indicated in panel A). Nfe2l3 expression in immature cortical neurons overlapped with expression of adult layer 5 projection neuron markers Bcl11b and Fezf2, as indicated by a dashed red oval line outlining respective cluster in panels J-K and N. Color-coded scale bars, normalized gene expression (NE).

Next, we analyzed expression of Nfe2l1, Nfe2l2, and Nfe2l3 in scRNA-seq-profiled retinal cell types from different stages of development (using previously published datasets^31-33). The differentiated retinal cells (excluding undifferentiated progenitor cells) from various ages retinas (embryonic day 16, postnatal day 0, adult, and ONC-injured) were integrated for visualization of cell type clusters using a UMAP (Fig. 2A). We found that only Nfe2l3 was enriched specifically in the RGCs, whereas Nfe2l1 and Nfe2l2 were enriched in non-RGC retinal cell types but also expressed in RGCs (Fig. 2B-D). Next, we analyzed average expression of Nfe2l1, Nfe2l2, and Nfe2l3 in scRNA-seq-profiled RGCs from different stages of development and after injury (2 weeks post-ONC). In contrast to higher expression of Nfe2l1 and Nfe2l2 in the adult RGCs, which was further upregulated after injury, expression of Nfe2l3 peaked in early development, declined during maturation, and after injury decreased to levels bordering noise (Fig. 2E-G). Further analysis did not find Nfe2l1, Nfe2l2, or Nfe2l3 to be RGC subtype-specific, although there was modest subtype-to-subtype variability in expression within uninjured and within injured adult RGC subtypes (Fig. 2H-J). We also characterized Nfe2l1, Nfe2l2, and Nfe2l3 protein coding transcript isoforms expressed in RGCs, using bulk-RNA-seq of RGCs purified from neonatal (postnatal day 5; see Methods) retinas (around the time when Nfe2l3 expression transiently peaks during RGC development). We identified two Nfe2l1 protein coding transcripts that are robustly expressed in RGCs, one protein coding Nfe2l2 transcript that was moderately expressed in RGCs, and two protein coding Nfe2l3 transcripts that were also moderately expressed in RGCs (Fig. 2K-M). These data show that, in contrast to Nfe2l1 and Nfe2l2, Nfe2l3 is enriched in RGCs within the retina, transiently expressed in early development, downregulated during maturation, and is nearly silenced after injury.

Figure 2. Expression of Nfe2l1, Nfe2l2, and Nfe2l3 genes in scRNA-seq-profiled retinal cell types and in RGC projection neurons at different developmental stages.

(A) UMAP of cell type clusters from E16, P0, adult uninjured, and adult injured (2 weeks post-ONC) retinas, as marked. RGC cluster is outlined by a dashed line. Because except for RGC most retinal cell types are not yet born in the embryonic retina, adult retina dataset was used for other retinal cell types and included proportionately fewer adult RGCs. Thus, most RGCs in the UMAP are from E16 and P0 datasets (see Methods for details).

(B-D) UMAPs showing gene expression of Nfe2l1 (B), Nfe2l2 (C) and Nfe2l3 (D) in clusters of retinal cell types (per annotation in panel A). RGC cluster (comprised mostly from neonatal RGCs, as detailed in A) is outlined by a dashed line. Color-coded scale bars, normalized gene expression (NE).

(E-G) Line plots of Nfe2l1 (E), Nfe2l2 (F) and Nfe2l3 (G) gene expression in scRNA-seq-profiled RGCs at different ages/conditions, as marked. Mean ± SEM of NE values are shown (see Methods for details).

(H-J) Barplots showing subtype-specific expression of Nfe2l1 (H), Nfe2l2 (I), and Nfe2l3 (J) genes in uninjured and injured (2 weeks post-ONC) scRNA-seq-profiled adult RGCs. Asterisk indicates RGC clusters which did not survive 2 weeks after ONC. Mean ± SEM of NE values are shown.

(K-M) Purified postnatal (P5) RGC bulk-RNA-seq reads alignment to gene loci of Nfe2l1 (K), Nfe2l2 (L), and Nfe2l3 (M), along with specific transcripts isoforms assembled by Cufflinks, respectively, and annotated with Ensembl transcript IDs (shown on the left side of the panels). Visualization with IGV viewer.

Because we found that Nfe2l3 is downregulated in RGCs during maturation and further after ONC injury, its antioxidant properties³ would not be available to protect RGCs from ONC-induced oxidative stress¹⁷. Therefore, we tested whether experimentally expressing Nfe2l3 in ONC-challenged RGCs would promote their survival and axon regeneration. In order to address this question, we selected from the transcript-specific bulk-RNA-seq-profiled RGC dataset an open reading frame (ORF) of the more highly expressed Nfe2l3 transcript (Fig. 2M) for functional testing in an axon regeneration assay (see Methods for details). To validate AAV2 vectors (which selectively transduce RGCs within the retina) expression in injured RGCs, a Myc-reporter was added to the N-terminus of Nfe2l3 ORF. For negative control, mCherry-expressing AAV2 vector was used. For positive control, established axon regeneration-promoting AAV2 vectors, expressing shRNAs to knockdown (KD) expression of Pten or Klf9, respectively, and co-expressing an mCherry reporter, were used^19-21,34,35. The viruses were injected intravitreally in adult mice, and 2 weeks later ONC injury was performed. AAV2 pre-treatment achieved approximately 30% transduction efficiency (see Methods for details). Immunostaining the retinas from uninjured and injured conditions with antibodies for Nfe2l3 and βIII-Tubulin or Myc-reporter confirmed downregulation of Nfe2l3 after ONC in injured RGCs, relative to already modest expression in uninjured RGCs. However, Nfe2l3 upregulation was robust in the injured Nfe2l3 AAV2-treated RGCs (Fig. 3A-C). Also, mCherry reporter-labeled injured RGCs confirmed transduction and expression of the established AAV2 vectors, used here for negative (scrambled shRNA) and positive (anti-Pten and anti-Klf9 shRNAs) controls (Fig. 3D-E).

Figure 3. Validation of Nfe2l3 expression and AAV2 vectors transduction in RGCs.

(A) Representative confocal images of the retinal flatmounts’ ganglion cell layer horizontal sections from adult (10 weeks old) mice co-immunostained for Nfe2l3 and βIII-Tubulin (Tuj1; an RGC marker within the retina), and counterstained with DAPI (to label nuclei), show modest Nfe2l3 signal in uninjured Tuj1+/DAPI+ RGCs.

(B-F) Mice (8 weeks old) were pre-treated with AAV2 vectors (which selectively transduce the RGCs within the retina) expressing mCherry (control), myc-tagged Nfe2l3, anti-Pten shRNA, or anti-Klf9 shRNA. ONC injury was performed 2 weeks later. Animals were sacrificed for histological analysis 2 weeks after ONC. Representative confocal images of the retinal flatmounts’ horizontal sections co-immunostained for Myc reporter and Nfe2l3 or Tuj1, and counterstained with DAPI, show marginal Nfe2l3 signal in the injured control mCherry+/DAPI+ RGCs (B), which also became punctate relative to the diffuse RGC soma labeling in the uninjured condition (shown in A). However, Nfe2l3 signal in the injured Nfe2l3-treated Myc+/DAPI+ RGCs is robust (C). A subset of the Tuj1+/DAPI+ RGCs are also mCherry reporter+, in the AAV2 mCherry (control), anti-Pten shRNA, and anti-Klf9 shRNA-treated conditions, as marked (D-F). OE = Overexpression; KD = Knockdown. Confocal microscopy, 63x Oil objective; Scale bar, 20 μm.

Then, we tested whether expression of Nfe2l3 in injured RGCs, which otherwise do not express it (except for bordering noise levels), would be sufficient to promote RGC survival and axon regeneration after being challenged by ONC injury, and how would possible neuroprotection compare to the effects of established axon regeneration-promoting treatments, which knockdown expression of Pten or Klf9. We used an established 2-weeks after ONC axon regeneration assay^18,21,34-36, in which the optic nerve is injured by crush in order to sever all RGC axons, and RGC survival and axon regeneration are assayed at 2 weeks after injury. AAV2 vectors expressing Nfe2l3, anti-Pten shRNA, anti-Klf9 shRNA, or mCherry control, were injected intravitreally in adult mice, and 2 weeks later ONC was performed. To visualize the regenerating axons or their absence, axonal tracer CTB was injected 1 day prior to sacrifice at 2 weeks after ONC. The number of regenerating axons was quantified in longitudinal sections of the optic nerve (no spared axons were detected in either group), and RGC survival was quantified in retinal flatmounts (see Methods for details; experimental timeline in Fig. 4A). We found that Nfe2l3 expression promoted RGC survival by approximately 20%, which is a similar extent of neuroprotection as with anti-Pten shRNA and anti-Klf9 shRNA AAV2 treatments by 2 weeks post-ONC, compared to RGC survival in the injured control group (Fig. 4B-D). Although given that over 80% of RGCs die by 2 weeks after ONC^37-39 a 20% increase in survival is a modest gain, as only about 4% less of RGCs died, if experimental neuroprotection persists permanently this would be meaningful, because without a treatment all the RGCs eventually die. Remarkably, relative to only minor axonal sprouting past the injury site detected in control animals (as expected), Nfe2l3 expression promoted long-distance axon regeneration approximately 3 mm past the injury site, which is comparable to the extent of axon regeneration achieved by anti-Pten shRNA and anti-Klf9 shRNA AAV2 treatments (Fig. 5A-D).

Figure 4. Nfe2l3 promotes RGC survival after optic nerve injury.

(A) Experimental timeline: 8 weeks old mice were pre-treated with AAV2 vectors expressing Nfe2l3, anti-Pten shRNA, anti-Klf9 shRNA, or mCherry control. ONC injury was performed 2 weeks later. Animals were sacrificed for histological analysis 2 weeks after ONC. Axonal tracer CTB was injected intravitreally prior to sacrifice.

(B) Representative images of the retinal flatmounts immunostained for an RGC marker βIII-Tubulin (Tuj1 antibody) at 2 weeks after ONC, pre-treated with AAV2 expressing Nfe2l3, anti-Pten shRNA, anti-Klf9 shRNA, or mCherry control, as marked. OE = Overexpression; KD = Knockdown. Scale bar, 20 μm.

(C-D) Quantitation of RGC survival in retinal flatmounts immunostained for an RGC marker βIII-Tubulin (Tuj1 antibody) at 2 weeks after ONC, pre-treated with AAV2 expressing Nfe2l3, anti-Pten shRNA, anti-Klf9 shRNA, or mCherry control, as marked (C). Data analyzed using ANOVA, overall F = 1017.9, p < 0.001, with p-values of pairwise comparisons determined by posthoc LSD. Significant difference (p < 0.05) indicated by an asterisk * (D). Mean ± SEM shown, n = 4 retinas per group.

Figure 5. Nfe2l3 promotes robust axon regeneration after optic nerve injury.

(A-B) Representative images of the optic nerve longitudinal sections with CTB-labeled axons at 2 weeks after ONC from the animals pre-treated with expressing Nfe2l3, anti-Pten shRNA, anti-Klf9 shRNA, or mCherry control, as marked (A). Insets: Representative images of the optic nerve regions proximal and distal to the injury site are magnified for better visualization of the axons or their absence (B). The edges of the tissue were optically trimmed (i.e., cropped-out) due to artefactual autofluorescence that is common at tissue edges (see Methods). OE = Overexpression; KD = Knockdown. Scale bars, 500 μm (main panels), 50 μm (insets).

(C-D) Quantitation of CTB-labeled regenerating axons at 2 weeks after ONC, at increasing distances from the injury site, after pre-treatment with AAV2 expressing Nfe2l3, anti-Pten shRNA, anti-Klf9 shRNA, or mCherry control, as marked (C). Data analyzed using repeated measures ANOVA, sphericity assumed, overall F = 13.2, p < 0.01, with p-values of pairwise comparisons determined by posthoc LSD shown in the inset table, and significant differences (p < 0.05) indicated by an asterisk * (D). Mean ± SEM shown; n = 4 optic nerves per group.

To gain insight into the mechanisms through which Nfe2l3 could promote neuroprotection and axon regeneration, we selected the genes that are associated with Nfe2l3 based on gene-ontology (GO) and are expressed in neonatal RGCs (before Nfe2l3 expression declines during maturation). Next, we identified GO biological processes (GO:BP) and GO molecular functions (GO:MF), which were enriched amongst these Nfe2l3-associated genes expressed in Nfe2l3+ neonatal RGCs relative to all genes expressed in neonatal RGCs. The top enriched GO: BP and GO:MF were both related to RNA-associated processes involved in gene expression/translation (Fig. 6A). We then analyzed the relationship between the genes comprising these top enriched gene expression/translation-associated GO:BP and GO:MF, and presented them as a gene network plot (Fig. 6B). Amongst the identified genes were numerous ribosomal proteins, including Rpl7 and Rpl7A, which we recently showed promote axon regeneration⁴⁰. Other identified axon regeneration-prompting genes included Klf7, Sox11, 14-3-3 adaptors (Ywha-x genes), and Atp5g1^16,41-44. Axon regeneration-suppressing genes, Set and Gsk3b, were also identified in the gene network plot^45,46. Gsk3b is developmentally downregulated in RGCs but is upregulated after ONC (Fig. 6C). Thus, although Gsk3b-mediated degradation of Nfe2l1-3 proteins is established^47-49, experimentally-upregulated Nfe2l3 protein may compensate and thereby promote neuroprotection and axon regeneration. The most prominent Pten KO-activated mTOR pathway of axon regeneration¹⁸ was not detected by the gene network analysis, but a recent report demonstrated that Nfe2l3 directly activates the mTOR pathway through mTorc1⁵⁰. Mtor gene, which encodes mTorc1, is substantially downregulated but not completely silenced in injured RGCs (Fig. 6D). Thus, mTorc1 may also be associated with Nfe2l3’s effect on neuroprotection and axon regeneration^51,52. To explore this possibility, we tested whether phosphorylated ribosomal protein S6 (p-S6), a downstream marker of the mTOR pathway activation¹⁸, is enriched in the surviving Nfe2l3-treated, compared to control, RGCs. Using phospho-S6-specific antibody, we quantified Nfe2l3 and mCherry-treated surviving RGCs that were also p-S6 immunopositive, but did not find a significant difference between the percent of Myc+/p-S6+ relative to all Myc+ RGCs (in the experimental condition) and the percent of mCherry+/p-S6+ relative to all mCherry+ RGCs (in the control condition) (Fig. 7A-E). These data suggest that, Nfe2l3 promotes RGC survival and axon regeneration through a different pathway downstream of mTOR or through a different mechanism. Taken together, these bioinformatic analyses predicted the association of Nfe2l3 with several established mechanisms of neuroprotection and axon regeneration, whereas the extent of their cooperation with, and dependency on, each other in regulating neuroprotection and axon regeneration is an important direction for future research.

Figure 6. Nfe2l3-associated gene ontology enriched in Nfe2l3-expressing RGCs.

(A) Barplots showing gene ontology (GO) annotations for biological process (BP; left) and molecular function (MF; right), which were enriched within the Nfe2l3-associated genes meaningfully expressed (>= 1 NE) in Nfe2l3+ neonatal RGCs relative to all genes expressed (>= 0.1 NE) in all neonatal RGCs (FC > 0.25-fold, p < 0.05), ordered by adjusted p-values (shown on x-axis in −log10; determined by the R packages gprofiler2 and enrichplot), with terms in the upper rows corresponding to the most significant enrichment (indicated by the lower adjusted p-value; up to top 5 shown, and only 4 were significant for MF). Color-coded scale bar indicates GO terms’ fold-change (FC) enrichment. See methods for more details.

(B) Gene-Concept Network Plot of genes associated with the top GO:BP and GO:MF terms by p-value significance. GO terms shown as gray circles (i.e., nodes), with circle size (per scale on the side) indicating the number of genes within that node. Color-coded scale bar indicates each gene’s normalized expression (NE) in neonatal RGCs (see Methods for details). Nfe2l3 and Gsk3b genes are outlined by dashed line boxes.

(C-D) Line plots of Gsk3b (C) and Mtor (Mtorc1) (D) gene expression in scRNA-seq-profiled RGCs at different ages/conditions, as marked. Mean ± SEM of normalized expression (NE) values are shown (see Methods for details).

Figure 7. Experimental upregulation of Nfe2l3 does not significantly alter levels of p-S6 in injured RGCs.

(A-B) Representative confocal images of the retinal flatmounts’ ganglion cell layer horizontal sections 2 weeks after ONC injury from animals pre-treated with control AAV2 expressing mCherry reporter, immunostained for p-S6, and counterstained with DAPI (A). Inset from A shown enlarged for better visualization of subsets of mCherry+/p-S6+ and mCherry+/p-S6- RGCs (B).

(C-D) Representative confocal images of the retinal flatmounts’ ganglion cell layer horizontal sections 2 weeks after ONC injury from animals pre-treated with AAV2 expressing myc-tagged Nfe2l3, immunostained for p-S6, and counterstained with DAPI (C). Inset from C shown enlarged for better visualization of subsets of Myc+/p-S6+ and Myc+/p-S6- RGCs (D). OE = Overexpression. Scale bars, 25 μm (main panels), 6.5 μm (insets).

(E) Quantification of the percent of p-S6+/mCherry+ of the total mCherry+ RGCs per retina, at 2 weeks after ONC, compared to the percent of p-S6+/Myc+ of the total Myc+ RGCs per retina, did not show a significant difference between the mCherry control or Nfe2l3 transduced conditions (mean ± SEM shown, n = 3 retinas per group; *p = 0.86 by independent samples t-test, 2-tailed). N.S. = Not significant.

DISCUSSION

The Nfe2l1, Nfe2l2, and Nfe2l3 family of proteins regulate various cellular processes, including response to oxidative stress^1-7,53. Nfe2l1 and Nfe2l2 were also shown to play roles in neurodegeneration and neuroprotection^54-57. Although in the adult CNS Nfe2l3 is expressed primarily in oligodendrocytes¹¹, transient expression of Nfe2l3 during early development of a subset of cortical projection neurons^12-14 suggested that it may also be involved in neuronal development. Indeed, we showed that Nfe2l3 expression also transiently peaks in subsets of developing immature cortical and RGC projection neurons. Furthermore, we showed that within the retina, only Nfe2l3 is enriched in RGC neurons, whereas Nfe2l1 and Nfe2l2 are expressed robustly in other retinal cell types as well. One of the features of immature CNS neurons is axon growth capacity, which declines during maturation^15,16. However, potential roles of Nfe2l3 in axon growth or neuroprotection were unknown. Although Nfe2l3 KO mice develop normally⁸, they exhibit phenotypes when challenged^9,10. Because Nfe2l3 is involved in antioxidant response^1-3, whereas axonal ONC injury induces oxidative stress in RGCs¹⁷, we challenged RGCs by ONC injury in order to test whether Nfe2l3 regulates neuroprotection and axon regeneration in vivo. We found that experimental expression of Nfe2l3 in injured RGCs, through locally-delivered AAV2 gene therapy vector, promoted neuroprotection and long-distance axon regeneration. Remarkably, the extent of neuroprotection and axon regeneration was similar to the effect of AAV2-targeting Pten and Klf9 genes, which are potent regulators of neuroprotection and long-distance axon regeneration^18-21 (although conditional genetic knockout of Pten in RGCs leads to more axon regeneration than AAV2-mediated knockdown of Pten, we used the same viral vector delivery method for all conditions in order to enable more appropriate comparison of the treatments per se). However, because cellular context is different between immature neonatal and mature adult RGCs, it remains to be investigated whether Nfe2l3 also plays a role in developmental axon growth of a subset of immature neurons which express Nfe2l3 neonatally. Thus, Nfe2l3 is an important neuroprotection and axon regeneration-promoting factor, and as such holds promise for development of a novel therapeutic approach to treat CNS injury and disease, whereas its potential role in developing neurons needs to be investigated in future studies.

We also utilized bioinformatics-based predictions to provide insights into the molecular mechanisms through which Nfe2l3 could elicit neuroprotection and promote axon regeneration. Analysis of the genes which are co-expressed, and also associated based on gene ontology, with Nfe2l3 in neonatal RGCs (before its expression declined during maturation), identified enriched RNA-associated biological processes involved in gene expression/translation. The identified genes included ribosomal proteins, such as axon regeneration-promoting Rpl7 and Rpl7a⁴⁰, as well as axon regeneration-prompting Klf7, Sox11, 14-3-3 adaptors (Ywha-x genes), and Atp5g1^16,41-44. Axon regeneration-suppressing Set and Gsk3b were also amongst the identified genes^45,46. Gsk3b inhibits axon regeneration and its KO is neuroprotective^46,58,59. Gsk3b is upregulated in injured RGCs, and although Gsk3b-mediated degradation of Nfe2l1-3 proteins is established^47-49, experimental upregulation of the Nfe2l3 protein may compensate and thereby promote neuroprotection and axon regeneration. Another implication of this hypothesis is that, Gsk3b KO may be promoting neuroprotection and axon regeneration due to rescue of the antioxidant response mediated by Nfe2l1 and Nfe2l2 proteins. Although Nfe2l2 also upregulates Klf9 suppressor of axon regeneration^20,21,60, overexpressing Nfe2l2 in RGCs at least promoted survival after ONC injury⁵⁷, while knocking-out Nfe2l2 hindered Pten KO-promoted neuroprotection⁶¹. Pharmacological activation of Nfe2l1 is also neuroprotective in an animal model of neurodegeneration⁵⁶, whereas Nfe2l1 KO contributes to neuordegenration^54,55, although no axon regeneration in vivo through Nfe2l1 was reported despite it being the most widely studied member of this protein family. However, we showed that Nfe2l3 promotes both neuroprotection and axon regeneration, and is therefore the most promising for therapeutic purposes. In addition, a recent report demonstrated that Nfe2l3 directly activates the mTOR pathway through mTorc1⁵⁰, and Gsk3b also was reported to regulate mTOR pathway^62,63, raising a possibility that Nfe2l3-promoted neuroprotection and axon regeneration may involve the mTOR pathway, which is the most widely studied mechanism of axon regeneration¹⁸. Although we did not find that the mTOR pathway downstream marker pS6 was enriched by the Nfe2l3 treatment amongst the surviving RGCs, other mTOR downstream effectors may still be involved. Future studies need to test whether Pten KO or Gsk3b KO would enhance Nfe2l3-promoted neuroprotection and axon regeneration.

Furthermore, even if Nfe2l3 is involved in the Gsk3b and/or Pten pathways, it may still be a more potent regulator of these pathways in relation to neuroprotection and axon regeneration, or it may also recruit another relevant known or novel pathway. These possibilities need to be investigated in future studies, along with the identification of the effector genes regulated by Nfe2l3, and its binding partner proteins, specifically in the CNS projection neurons. Testing in comparison to, and combination with, other established regulators of neuroprotection and axon regeneration (e.g., AAV2 CNTF^64,65) also will be important. Although Nfe2l3-promoted gain in RGC survival by 2 weeks after ONC was modest, if it persists permanently this would be meaningful, because without a treatment all the RGCs eventually die^37-39. Thus, future studies over long-term are needed, using a variety of RGC survival markers⁶⁶, and comparing to uninjured RGCs by assessing only the transduced RGCs (e.g., Myc+/Tuj1+ cells), in order to further characterize Nfe2l3’s neuroprotective potential. In addition, as Nfe2l3 is associated with several established regulators of neuroprotection and axon regeneration, the extent of their cooperation with, and dependency on, each other in promoting neuroprotection and axon regeneration also needs to be investigated in future studies.

METHODS

Animal use, surgeries, intraocular injections.

All animal studies were performed at the University of Connecticut Health Center with approval of the Institutional Animal Care and Use Committee and of the Institutional Biosafety Committee, and performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Visual Research. Mice were housed in the animal facility with a 12-h light/12-h dark cycle (lights on from 7:00 AM to 7:00 PM) and a maximum of five adult mice per cage. The study used wild-type 129S1/SvImJ mice (The Jackson Laboratory). Optic nerve surgeries and intravitreal injections, were carried out on mice of both sexes 8-12 weeks of age (average body weight 20-26 g) under general anesthesia, as described previously^21,35,36,67. The viruses included AAV2 vectors expressing Nfe2l3 (ORF of ENSMUST00000005103), anti-Pten shRNAs (target sequences: 5′-GCAGAAACAAAAGGAGATATCA-3′, 5′-GATGATGTTTGAAACTATTCCA-3′, 5′-GTAGAGTTCTTCCACAAACAGA-3′, and 5′-GATGAAGATCAGCATTCACAAA-3′), anti-Klf9 shRNAs (target sequences: 5′-GGAGGCGCTGCCGGTTACGTA-3′, 5′-TGGCTGCCCAGTGTCTGGTTT-3′, 5′-CGGGGGACACCTGGAAGGATT-3′, and 5′-GCAAATAAATGCTTTTGGTAC-3′), and mCherry alone for control (titers ~1 × 10¹² GC/mL; VectorBuilder, Inc.). Nfe2l3 ORF and mCherry ORF had an N-terminally fused myc-tag reporter. Vectors expressing anti-Pten or anti-Klf9 shRNAs, also co-expressed an mCherry reporter. Viruses (2 μl per eye) were injected intravitreally, avoiding injury to the lens, in 8-week-old mice, which were randomly assigned to experimental or control conditions, 2 weeks prior to ONC surgery. This lead time allowed for sufficient transduction and expression of the transgenes in RGCs at the time of ONC. Transduction efficiency was approximately 30%, which is comparable to prior reports that also used AAV2 to target the RGCs^{21,34-36,40,45}. Cholera toxin subunit B (CTB) conjugated to Alexa Fluor 488 dye (C34775, ThermoFisher Scientific) was injected (1% in 3 μl PBS) intravitreally one day prior to sacrifice, at 2 weeks after ONC, in order to visualize the regenerating axons.

Tissue processing and immunostaining.

Standard histological procedures were used, as described previously^{21,35,36,67,68}. Briefly, anesthetized mice were transcardially perfused with isotonic saline followed by 4% paraformaldehyde (PFA), the eyes and optic nerves were dissected, postfixed 2 hours, the retinas were dissected-out, and optic nerves were transferred to 30% sucrose overnight at 4 °C. The optic nerves were then embedded in OCT Tissue Tek Medium (Sakura Finetek), frozen, cryosectioned longitudinally at 14 μm, and then mounted for imaging on coated glass slides. For analyzing RGC survival, resected (into PBS at 4 °C) free-floating retinas were immunostained in 24-well plate wells and, after making 4 symmetrical slits, flat-mounted on coated glass slides for imaging. For analyzing Nfe2l3 transgene expression in AAV2-transduced RGCs, the flattened retinas were embedded in OCT Tissue Tek Medium (Sakura Finetek), frozen, and cryosectioned at 14 μm horizontality (capturing the ganglion cell layer of the whole retina), and then immunostained and mounted on coated glass slides for imaging using a confocal microscope (see below). For immunostaining, the tissues were blocked with appropriate sera, incubated overnight at 4 °C with primary antibodies, βIII-Tubulin (1:500; rabbit polyclonal, Abcam, Ab18207), Myc (1:400; mouse monoclonal, SC-40 SCBT), Nfe2l3 (1:100; rabbit polyclonal, Biorbyt, Orb312344), p-S6 (1:200; rabbit, CTS, 2211S), and counterstained with DAPI (1:5000; Thermo Fisher Scientific) to label nuclei, then washed 3 times, incubated with appropriate fluorescent dye-conjugated secondary antibodies (1:500; IgG H+L or IgG 2a Alexa Fluor, Thermo Fisher Scientific) overnight at 4 °C, washed 3 times again, and mounted for imaging. Nfe2l3 immunostaining required optimization, which included the addition of 2% Triton X-100 to the blocking and primary antibody buffer solutions.

Quantification of regenerated axons, RGC survival, and p-S6+ RGCs.

To visualize the regenerating axons or their absence after treating with the AAV2 vectors expressing Nfe2l3, anti-Pten shRNA, anti-Klf9 shRNA, or mCherry control, axonal tracer (Alexa Fluor 488-conjugated CTB 1% in 3 μl PBS) was intravitreally injected one day before animals were euthanized 2 weeks following ONC. Longitudinal sections of the optic nerve were examined for possible axon sparing³⁵. No spared axons were found in control, and no evidence of axon sparing was found in experimental conditions (i.e., at 2 weeks after injury, no axons were found at the most distal from the injury region of the optic nerve). Regenerated axons (defined as continuous fibers, which are absent in controls and are discernible from background puncta and artefactual structures) were counted manually using a fluorescent microscope (40x/1.2 C-Apochromat W; AxioObserver.Z1, Zeiss) in at least 4 longitudinal sections per optic nerve at 0.5 mm, 1 mm, 1.5 mm, 2 mm, and 3 mm distances from the injury site (identified by the abrupt disruption of the densely packed axons near the optic nerve head, as marked by a rhombus in Fig. 5A), and these values were used to estimate the total number of regenerating axons per nerve, as described^{21,35,36,40,67}. For representative images, serial fields of view along the longitudinal optic nerve tissue section were imaged as above; z-stacks with 5 planes at 0.5 μm intervals were deconvoluted, merged, and stitched (ZEN software, Zeiss). Then, processed images of 3 tissue sections from the same optic nerve were superimposed over each other and merged using Photoshop CS6 (Adobe), shown as representative images. RGC survival was quantified in retinal flatmounts as described^{21,35,36,40,67,68}, by immunostaining with an antibody to βIII-Tubulin (neuronal marker Tuj1), taking advantage of the selective expression (within the retina) of βIII-tubulin in RGCs. ImageJ software Cell Counter Plugin was used to count βIII-Tubulin positive cells from images acquired (using a fluorescent microscope, 20x LD; Zeiss, AxioObserver.Z1, Zeiss) at 1 mm and at 2 mm from the optic nerve head in four directions, then averaged to estimate overall RGC survival per mm² of the retina. In order to determine whether p-S6 marker was enriched amongst the surviving RGCs in the experimentally-treated condition, p-S6+/Myc+ cells were quantified (using ImageJ software Cell Counter Plugin) as a percent of total Myc+ RGCs in the experimental condition, and p-S6+/mCherry+ cells were quantified as a percent of total mCherry+ RGCs in the control condition, from images acquired using a confocal microscope (63x Oil; LSM800, Zeiss) at 3 randomly sampled regions per retina, which were then averaged (at least 3 retinas per condition were analyzed). Representative images of RGCs in the horizontality cryosectioned flattened retinas’ ganglion cell layer immunostained for Nfe2l3, Tuj1, Myc, and p-S6 (where applicable) were imaged, along with mCherry reporter (where applicable), using a confocal microscope (63x Oil; LSM800, Zeiss). Investigators performing the surgeries and quantifications were masked to the group identity by another researcher until the end of the experiment.

Statistical analyses.

All tissue processing, quantification, and data analysis were done masked throughout the study. Sample sizes were based on accepted standards in the literature and our prior experiences. Sample size (n) represents total number of biological replicates in each condition. All experiments included appropriate controls. No cases were excluded in our data analysis, although a few animals that developed a cataract in the injured eye were excluded from the study, and their tissues were not processed. The data are presented as means ± SEM, and was analyzed (as specified in the applicable Figure legends) by ANOVA with or without Repeated Measures, and a posthoc LSD test (SPSS). All differences were considered significant at p < 0.05.

RGC scRNA-seq and bulk-RNA-seq datasets.

Normalized matrices and cell metadata (e.g., type assignment) for embryonic E14, embryonic E15, postnatal P0, and adult cortical neurons, are available from the Gene Expression Omnibus (GEO) accession numbers GSE143949, GSE123335, and GSE116470^11,22,23. Seurat v4.3.0^69,70 was used to merge and normalize the datasets mapped to the mm10 genome. Projection cortical neurons that passed scRNA-seq QCs were identified based on co-expression of neuronal markers, Rbfox3 (NeuN)²⁴ and Gap43^25,26, along with cortical projection neuron markers, L1cam³⁰ and Uchl1²⁹. Normalized matrices and cell metadata (e.g., type assignment) for embryonic E13, embryonic E14, embryonic E16, postnatal P0, adult uninjured, and adult injured (2 weeks post-crush) RGCs, as well as for adult uninjured and injured (2 weeks post-crush) retinas, are available from the GEO accession numbers GSE185671³¹, GSE137400³², and GSE199317³³. Seurat v4.3.0^69,70 was used to merge and normalize the datasets mapped to the mm10 genome. Datasets were batch adjusted and integrated using Harmony⁷¹. Cell type identity of cells that passed scRNA-seq QCs was confirmed based on established retinal cell type markers; retinal progenitor cells were excluded. RGC cell identity was confirmed based on co-expression of pan-RGC markers (Rbpms, Slc17a6, Sncg, and Tubb3). Barplots (in Fig. 2H-J) with normalized expression (mean ± standard error of the mean) for RGC subtypes were generated using the R packages Seurat and ggplot2⁶⁹. Bulk-RNA-seq samples from postnatal day 5 RGCs were purified by immunopanning for Thy1 (after immunopanning depletion of macrophages). RNA with RNA Integrity Number (RIN) ≥ 9 (by Bioanalyzer 2100 using the Nano 6000 kit, Agilent) was extracted using Direct-zol RNA MiniPrep kit (R2050, Zymo Research). The RNA was processed and cDNA libraries were prepared using RiboZero and ScriptSeq v2 kit (Epicentre) which depletes rRNAs. Paired reads were sequenced in a DNA-strand-specific manner, 100 bp from each end, on HiSeq 2000 Sequencer (Illumina), passed QC filters, mapped to the mm10 genome and transcriptome by Hisat2, and analyzed by Cufflinks/CuffDiff, as we previously described^72,73. Transcript isoforms expressed > 0.1 FPKM were retained and visualized using IGV browser⁷⁴.

Data availability.

The scRNA-seq processed datasets, normalized matrices, and integrated data-frame we generated for this study are available through the NCBI GEO accession number GSE253840. P5 RGC bulk-RNA-seq raw reads and processed data for alternative transcripts isoforms from the Nfe2l(1-3)-family genes loci analyzed in this study are available through the NCBI GEO under accession number GSE253840.

Gene-concept network and expression plots.

Genes expressed moderately or above (>= 1 normalized expression) in Nfe2l3-positive (i.e., expressing Nfe2l3 >= 1 normalized expression) RGCs were analyzed using the R package gprofiler2, using all genes expressed in RGCs as background (normalized expression >= 0.01). False discovery rate (FDR) was used for multiple testing correction, and a minimum p-value of 0.05 was set as cutoff for significance of GO terms enrichment. GO: BP/GO:MF terms were rank ordered by decreasing significance (FDR adjusted p-value). Fold enrichment was calculated as the DEG ratio divided by the background ratio⁷⁵. The top GO:BP and top GO:MF pathways (p-value < 0.05) containing Nfe2l3 were plotted in a Gene-Concept Network Plot using the clusterProfiler and enrichplot R packages⁷⁶.

Highlights.

• Novel role for neuronal Nfe2l3 in promoting neuroprotection and long-distance axonal regeneration after CNS injury in vivo.

• Viral vector expression of Nfe2l3 in the adult retinal ganglion cells provides a similar extent of neuroprotection and axon regeneration as viral vector shRNA knockdown of Pten or Klf9.

• Nfe2l3 is not meaningfully expressed in the adult retinal ganglion cells before or after optic nerve injury.