Injury-Specific Expression of Activating Transcription Factor-3 in Retinal Ganglion Cells and Its Colocalized Expression with Phosphorylated c-Jun

Abstract

purpose. To ascribe activating transcription factor (ATF)-3 as a specifically induced transcription factor after ON injury and to describe its putative role as a modulator of c-Jun transactivation.

methods. The adult rat optic nerve was crushed intraorbitally, and expression profiles of ATF-3, ATF-2, and phosphorylated c-Jun (p-c-Jun) were examined by immunohistochemistry and ISH. Western blot analysis for ATF-3 and -2 were also performed. Furthermore, colocalized detection of c-Jun mRNA with ATF-2 or -3 was attempted with a combined method of simultaneous immunohistochemistry and in situ hybridization.

results. In response to optic nerve injury, substantial expression of ATF-3 as well as that of p-c-Jun was observed in the retinal ganglion cells, whereas no expression of ATF-3 was seen in other noninjured retinal cells. In contrast, ATF-2 was normally expressed abundantly in both retinal ganglion cells and displaced amacrine cells, but expression dropped in retinal ganglion cells after nerve injury. The expression profiles of ATF-2 and -3 after optic nerve injury were confirmed by Western blot analysis. A higher degree of colocalization was observed for ATF-3 and c-Jun than the modest codetection for ATF-2 and c-Jun.

conclusions. The transcription factor ATF-3 is specifically induced upon optic nerve injury and colocalizes with p-c-Jun in surviving ganglion cells. These findings suggest that both ATF-3 and c-Jun are crucial to trigger various transcriptional responses and may act synergistically during the survival phase of the optic nerve in the injury model.

In the central nervous system (CNS) of adult mammals, neuronal death is the main consequence of axonal injury. Retinal ganglion cells (RGCs) readily die after optic nerve (ON) injury, ^{1 2 3 4 5} and therefore this ON injury model is widely considered and used as an experimental CNS injury model. In addition, the mode of RGC death by ON injury is generally accepted as apoptosis. ^{3 6 7} Upon ON injury, a large repertoire of molecules is upregulated in RGCs. ⁸ Among these, transcription factors are believed to play crucial roles in sensing and responding to ON injury ^{8 9} to actively execute apoptosis. c-Jun is a well-studied transcription factor that is correlated to neuronal cell death in vitro ^{10 11} and in vivo. ^{12 13 14} The induced expression of c-Jun coincides with the RGC degeneration after ON axotomy ^{9 15 16} and ON crush. ^{8 17 18 19} c-Jun is a member of the activating protein (AP)-1 family and acts as homo- or heterodimers with other members of the Jun, Fos, or activating transcription factor (ATF) families. ²⁰ Phosphorylation of c-Jun occurs after nerve growth factor (NGF) withdrawal stimuli for apoptosis of sympathetic neurons ²¹ and after middle cerebral artery (MCA) occlusion ²² and is required for apoptosis in cerebellar granule neurons ²³ or kainate-induced neuronal apoptosis. ²⁴ Although the phosphorylation of c-Jun seems to be concomitant with neuronal injury, the functional relevance of c-Jun expression to neuronal death or survival is still controversial and has lead to the proposition that its existence or phosphorylation is not decisive per se but requires other cofactor proteins. ²² Because AP-1 transactivation varies among combinations of these transcription factors, ²⁵ it is probable that factors other than c-Jun might modulate gene regulation. c-Fos, a classical member of the AP-1 transcription complex, could not be detected in rat RGCs after ON injury ⁸ but is related to hyperactivity-induced cell death in photoreceptors. ²⁶ Another member, ATF-2, has been studied in this context but shows downregulation upon various neuronal stresses. ^{27 28 29 30} Therefore, in this study we investigated expression profiles and the compositions of AP-1 transcription factors, with special reference to a less-studied member, ATF-3/LRF-1, in RGCs after ON crush with the aim of understanding the AP-1 gene transactivation upon CNS nerve injury. 

Materials and Methods

Animal Procedures

All experiments were carried out on male Wistar rats weighing 200 g. For all surgical procedures, rats were anesthetized with pentobarbital (0.3 mg/kg i.p.). All animals were observed by indirect ophthalmoscopy before surgery to avoid fundus abnormalities. The upper eyelid and limbal conjunctiva were dissected, and the ON was exposed gently. The ON crush was performed intraorbitally at distance of approximately 2 to 3 mm posterior from the eye. The ON was crushed for 30 seconds with forceps (E1815A; Storz, St. Louis, MO). Because ischemia of the retina may also alter c-jun mRNA and c-Jun protein, ^{31 32} the fundus was observed with a contact lens during the ON crush to check for the absence of retinal blood flow occlusion. After ON crush, the conjunctiva was replaced, and the eyelid was sutured to prevent abnormal condition of the anterior segment of the eye. A sham operation was performed on the left eye but without a crush. All animal procedures were performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Histology Analysis

For section preparation, rats were killed with an overdose of pentobarbital followed by perfusion with 4% paraformaldehyde/0.1 M phosphate buffer (PB). The eyes were enucleated and postfixed overnight in the same solution at 4°C. The eyes were dehydrated, embedded in paraffin wax (TissuePrep; Fisher Scientific, Pittsburgh, PA), and 7-μm sections were cut and mounted onto 3-aminopropyltriethoxysilane–coated slides. The sections were stored dry until used for histologic analysis. Histologic results shown are representatives from six independent experiments (n = 6). 

To evaluate the pattern of RGC loss in this model, hematoxylin-eosin (H-E)-stained paraffin sections were examined. The cell number in the ganglion cell layer (GCL) was counted for each time point from six evenly spaced sagittal sections, all including the ON disc, from six independent animals (n = 6). The cells in the nerve fiber layer or inner plexiform layer were not counted to avoid inclusion of astrocytes or displaced ganglion cells. To distinguish RGCs from displaced amacrine cells (DACs), Thy-1 in situ hybridization (ISH) was performed as described below and the proportion of RGCs was calculated. RGCs immunoreactive for phosphorylated c-Jun (p-c-Jun), ATF-2, and -3 also were counted from six independent sections at each time point and shown as a proportion of the remaining RGCs (immunopositive cells/H-E–stained cells − DACs). Sections from the control eye were processed at the same time to normalize immunoreactivity. For statistical analysis a paired t-test was used. 

Immunohistochemistry

Sections were dewaxed, rinsed in PBS, and incubated in blocking solution containing 0.5% Triton X-100/3% bovine serum albumin/0.02% sodium azide in phosphate-buffered saline (PBS) for 30 minutes at room temperature. These pretreated sections were incubated with primary antibody (p-c-Jun [Ser63] II antibody, diluted 1/200; New England Biolabs, Beverly, MA; ATF-2 [c-19], diluted 1/1000; Santa Cruz Biotechnology, Santa Cruz, CA; or ATF-3 [c-19], diluted 1/800; Santa Cruz) overnight at 4°C. Then the sections were rinsed three times in PBS, incubated with secondary antibody (goat biotinylated anti-rabbit IgG diluted 1/400; Vector Laboratories, Inc, Burlingame, CA) for 2 hours at room temperature, rinsed three times in PBS, and incubated in avidin/biotin-peroxidase complex (Vector) in PBS for 1 hour at room temperature. They were rinsed in PBS and immersed in 0.05 M Tris-HCl (pH 7.6). Coloration was performed in Tris-HCl containing diaminobenzidine (DAB) and hydrogen peroxide. 

Flat-Mount Preparation

Incisions were made at the ora serrata to remove the anterior segment of the eye, lens, and vitreous body. Retinas were carefully detached from the scleral wall and postfixed in 4% paraformaldehyde in 0.1 M PB overnight at 4°C. Retinas were rinsed in PBS, permeabilized with 1% Triton X-100 in PBS for 1 hour at room temperature, and rinsed in PBS again without detergent. Primary antibodies against p-c-Jun and ATF-3, and the secondary antibody were diluted as above. Retinas were incubated with primary antibody overnight at 4°C, rinsed in PBS three times for 30 minutes, 1% Triton X-100 for 10 minutes, again in PBS for 30 minutes, and incubated in secondary antibody for 2 hours at room temperature. Tissues were then permeabilized, incubated with avidin-peroxidase, and stained with DAB and hydrogen peroxide. Flat-mount results were obtained from four independent experiments at each time point. 

Retinal Protein Extraction and Western Blot Analysis

Animals were killed as described above. The eyes were immediately enucleated, and retinas were dissected from the scleral wall. The total protein of the retina was prepared according to the method of Kenney and Kocsis. ³³ The extracted tissue was solubilized in 3% SDS buffer (1 mM orthovanadium, 0.19 μl/ml aprotinin, and 0.1μ g/ml PMSF), and boiled for 10 minutes. The lysates were added to the same volume of 0.3 M sucrose, homogenized, and centrifuged at 14,000 rpm for 15 minutes at 4°C. The lysates were stored at −80°C until use. 

For Western blot analysis, 50 μg total protein in SDS sample buffer was applied to each lane. The samples were electrophoresed in 10% SDS-polyacrylamide gels for ATF-2 and 15% gels for ATF-3. After blotting, PVDF membranes were washed in TBST (Tris-buffered saline containing 0.1% Tween-20), incubated with primary antibodies (diluted ATF-2: 1/3000, ATF-3: 1/2000) overnight at 4°C. The membranes were incubated with secondary antibody (donkey horseradish peroxidase–linked anti-rabbit Ig [Amersham], diluted 1/4000 in 5% skim milk TBST) for 1 hour at room temperature, and the ECL Western blot analysis system (Amersham) was used for detection. 

In Situ Hybridization

Rat cDNA fragments for Thy-1 (GenBank X03150, nt 112–477, 365 bp), c-jun (GenBank X17163, nt 450-1469, 1019 bp), ATF-2 (GenBank M65148, nt 175–962, 787 bp), and ATF-3 (GenBank M63282, nt 165–708, 543 bp) were amplified by RT-PCR and subcloned into pBluescript II KS + vector (Stratagene, La Jolla, CA). These templates were linearized and digoxigenin (DIG)-labeled cRNA probes were prepared by in vitro transcription using T7 or T3 RNA polymerase (Boehringer Mannheim). Forty nanograms of the probes was used per slide. 

All prehybridization procedures were performed in RNase-free conditions at room temperature. The sections were dewaxed, treated with proteinase K (10 μg/ml) for 5 minutes, washed in sodium PB, fixed in 4% paraformaldehyde/PB for 10 minutes, and washed with PB again. After treating with 0.2 M HCl for 10 minutes and washing in PB, acetylation was performed with 0.1 M triethanolamine/0.25% acetic anhydride for 10 minutes. Sections were then dehydrated in an ascending ethanol series, incubated in chloroform for 10 minutes, and dried. Hybridizations were carried out for approximately 12 hours at 58°C in hybridization buffer (50% deionized formamide, 20 mM Tris-HCl, pH 8.0, 5 mM EDTA, 0.3 M NaCl, 10 mM PB, 10% dextran sulfate, 0.2% sarcosyl, 1× Denhardt’s solution, 0.5 mg/ml yeast tRNA, and denatured 0.2 mg/ml salmon sperm DNA) for each probe. After hybridization, the slides were washed in 50% formamide/2× SSC for 30 minutes at 65°C, immersed in RNase buffer (0.5 M NaCl, 10 mM Tris-HCl, and 1 mM EDTA), treated with RNase buffer containing 0.1 mg/ml RNaseA for 30 minutes at 37°C, and immersed again in RNase buffer. They were then washed in 50% formamide/2× SSC for 30 minutes at 65°C, rinsed in 2× SSC for 10 minutes at 65°C, and 0.1× SSC for 10 minutes at room temperature. After equilibration in buffer 1 (100 mM Tris-HCl, pH 7.5, and 150 mM NaCl) for 5 minutes, blocking was performed with 1.5% blocking reagent (Boehringer Mannheim) in buffer 1 for 60 minutes at room temperature. Slides were incubated with alkaline phosphatase–conjugated Fab fragments against DIG (diluted 1/2000; Boehringer Mannheim) overnight at 4°C. For coloration, the slides were washed two times for 15 minutes in buffer 1, equilibrated in buffer 3 (100 mM Tris-HCl, pH 9.5, 100 mM NaCl , and 50 mM MgCl₂) for 5 minutes, and stained with NBT/BCIP (Boehringer Mannheim) in buffer 3 at room temperature for approximately 12 hours. The reaction was stopped with 10 mM Tris-HCl (pH 7.6)/1 mM EDTA, and the slides were mounted or subsequently processed for immunohistochemistry (IHC) for double staining as described above. 

Results

Profile of RGC Death after ON Crush

Intraorbital injury of the ON leads to substantial responses of RGCs, followed by a vast cell death. Because the primary aim of the study was to temporally link RGC loss with the expression of various transcription factors after ON crush, we started by counting H-E–stained cells in the GCL at various time points after the ON crush (Fig. 1) . In control retinas, the total number of GCL cells was 552 ± 32 cells/section (c/s; mean ± SD). However, the number of RGCs compared to the number of DACs, another major cell population within the GCL, could not be determined. ³⁴ DACs overlap in size and are indistinguishable from RGCs on H-E sections. Hence, the RGC number is only a proportion of the total cell number within the GCL. To estimate the proportion of RGCs compared to DACs within the GCL, we analyzed sections for Thy-1 expression, which has been shown to be RGC specific in intact retinas. ^{35 36 37} When ISH was performed on noninjured retinal sections, 58% of the cells in the GCL appeared to be Thy-1 positive (data not shown). Given that DACs are negative for Thy-1, we concluded that the cells within the GCL of the rat used in our experiment consists of 58% RGCs and 42% DACs when counted this way. In experimental retinas, apparent cell loss was not observed until 5 days after ON crush. However, a prominent sign of cell loss appeared 1 week after the crush injury (462 ± 25 c/s; *P < 0.05); thereafter the cell number rapidly decreased until 2 weeks after crush (356 ± 21 c/s). At 8 weeks, the latest time point analyzed, the GCL cell number was 252 ± 14 c/s. In summary, only 6.6% of the RGCs persisted after 8 weeks if we considered that the DAC number is unchanged throughout the process (also shown as a broken line baseline in Fig. 1 ). 

c-Jun Expression and Phosphorylation after ON Injury

Previous work showed that c-Jun is induced after nerve injury, including ON axotomy. This induction occurred essentially in all RGCs, and immunohistologic evidence demonstrated the protein localization to be nuclear as well as cytosolic. c-Jun is modified by phosphorylation of its N-terminal serine residues, which is believed to be a consequence of the augmented phosphorylating activity of upstream kinases, and results in altered trans-activity. To extend these previous observations, we made use of an antibody that specifically detects the phosphorylated form of c-Jun at serine 63, together with c-jun ISH to distinguish effects of its posttranscriptional modification from changes in its transcriptional activity. Using the p-c-Jun antibody, we could not detect any histologic immunoreactivity throughout the control retinas (Figs. 2A 2A′ ). This result is affirmative inasmuch as no signal for c-jun mRNA was detected using ISH under the same conditions (data not shown). However in the experimental retinas, exclusively nuclear immunoreactivity was observed in RGCs and the number of expressing cells and the intensity of the signals increased to attain its maximum level by 1 day after ON injury (Figs. 2B 2C 2B′ 2C′ ). The upregulation of these signals is probably a product of de novo synthesis of the protein, because ISH signals paralleled the induction observed (Figs. 3A 3B 3C) . Subsequently, expression of both the phosphorylated protein and mRNA tapered off until 8 weeks after operation, the latest time point analyzed (Figs. 2D 2E 2F 2D′ 2E′ 2F′ 3D 3E 3F ). It should be noted though, that even at 8 weeks when the majority of RGCs had died, some remaining cells still expressed c-jun mRNA (Fig. 3F) . 

ATF-2 Expression Declines in a Subset of RGCs

One CREB/ATF family member implicated in mediating cellular stress is ATF-2. This transcription factor has been shown to bind to the c-jun promoter as a c-Jun/ATF-2 dimer to actively transcribe c-jun. ³⁸ We examined the expression of this molecule at the cellular level in our model using IHC and ISH. In control retinas, ATF-2–positive cells were detected widely in the GCL and the inner nuclear layer (INL) (Fig. 4A ). The ATF-2 immunoreactivity was visible in the nuclei of the cells in both layers. In experimental retinas, the nuclei of the cells within the GCL and INL remained immunoreactive. However, we observed a clear variegation in the intensity of immunoreactivity, especially in the GCL (Fig. 4B) , when compared to the control retinas. The number of ATF-2–positive cells reached its lowest level 4 weeks after crush (Fig. 4C) . Although we did not quantify it, ATF-2 immunoreactivity in the INL seemed to be unchanged throughout the period studied (Figs. 4A 4B 4C) . ATF-2 mRNA expression is shown in Figures 4D 4E 4F . In control retinas, ATF-2 mRNA was expressed in cells within the GCL and INL (Fig. 4D) in good agreement with the IHC. In experimental retinas, ATF-2 mRNA expression became fainter in a subset of RGCs 3 days after crush (Fig. 4E) and also started to show variegation in the intensity of the staining. 

Specific Induction of ATF-3 in RGCs after ON Injury

We next examined the expression profile of ATF-3, a CREB/ATF family member known to be induced by some tissue injuries including seizures. ³⁹ In control retinas, no ATF-3–immunoreactive or mRNA-expressing RGCs were detected in any layer (data not shown). In experimental retinas, immunoreactivity was undetectable 12 hours after ON crush (Figs. 5A 5A′ ), but positive RGCs started to appear 1 day after crush (data not shown). The ATF-3 immunoreactivity was detected most likely in the nuclei of RGCs; as in an independent experiment using axotomized and fluorogold back-filled RGCs, immunoreactivity was solely colocalized with the fluorogold signals (data not shown). The number and intensity of ATF-3 expression in RGCs greatly increased afterward and reached its peak by 3 days after injury (Figs. 5B 5B′ ). The number of immunoreactive cells was decreased by half 1 week after crush (Figs. 5C 5C′ ). Although scattered ATF-3–positive RGCs were still detectable at 2 weeks (Figs. 5D 5D′ ), they eventually disappeared by 4 weeks after crush (data not shown). These changes in the ATF-3 protein level have been confirmed by ISH. The onset, peak timing, and disappearance of the transcript (Fig. 6) followed a very similar time course shown by its protein. This again signifies that the RGC-specific induction of this protein is due to a de novo synthesis of its transcript. 

Quantitative Protein Analysis for ATF-2 or ATF-3

To confirm and quantify changes in the expression levels of ATF-2 and -3, we performed quantitative protein analysis using Western blotting. Because the cell number in the GCL decreases after crush, it is difficult to find a suitable internal control for these assays. However, we ensured equal loading of total protein, which was confirmed by gel staining (data not shown). Figure 7A shows a retinal Western blot for ATF-3 at various time points. The detectable level of ATF-3 in control retinas remained unchanged at 6 hours after ON crush. ATF-3 expression reached a peak expression by 12 hours after the crush injury, and this continued until 3 days after crush, which was in line with histologic data. The ATF-3 protein migrated as a 21-kDa band, in good agreement with that observed in regenerating rat liver. ⁴⁰ In contrast, the ATF-2 expression level showed little change for 3 days after crush but decreased slightly in later stages (Fig. 7B) . As mentioned above, a consistent expression level of ATF-2 is present in cells other than RGCs and has hindered a comparative expression analysis in RGCs (see Figs. 4A 4B 4C ). 

Differential Expression of ATF-2 and -3 within the Diminishing RGCs

Changes in the calculated number of RGCs expressing each transcription factor at various time points are shown in Figure 8A for direct comparison (see Materials and Methods). ATF-3–immunopositive RGCs were undetectable at 12 hours after crush but first appeared at 1 day after crush (9 ± 3 c/s). After 3 days, the number of ATF-3–positive RGCs reached its peak (184 ± 8 c/s). Then ATF-3 immunoreactivity rapidly decreased until 3 weeks (6 ± 8 c/s) and disappeared 4 weeks after crush. ATF-2 immunoreactivity was observed in control GCL cells (462 ± 12 c/s). In experimental retinas, the number of immunoreactive RGCs rapidly decreased until 2 weeks after crush (181 ± 18 c/s). The rate of decrease became slower in later stages, and the number of ATF-2–positive RGCs reached 127 ± 14 c/s at 8 weeks. 

Because the number of the RGCs diminished over the period, the number of immunoreactive cells is also shown in Figure 8B as a proportion of the remaining RGCs, based on the calculated RGC number shown in Figure 1 (GCL cells − DACs). This clearly depicts the differential onset of induction between ATF-3 and p-c-Jun. Also only 55% of the RGCs expressed ATF-3 at 3 days after crush when no cell death was observed. One week after crush, when the total RGC number started to decrease significantly, ATF-3 was expressed in 35% of the RGCs. The ATF-3 expression then faded and disappeared at 4 weeks. 

Higher Colocalization of ATF-3 and c-Jun after ON Crush

Because ATF-2, -3, and c-Jun share a consensus binding sequence TGACTCA (TRE/AP-1 site) and can form homo- or heterodimers among themselves, ²⁰ we carried out simultaneous detection of these factors to gain an insight into the cellular context of the AP-1 partnership. Because the antibodies against ATF-2, -3, and p-c-Jun were all polyclonal rabbit IgG, we tried to double-stain them using both IHC and ISH techniques simultaneously and to examine the colocalization of each factor. A brown staining indicates expression of proteins detected by IHC, and blue indicates cytosolic mRNA detection by DIG-ISH (Fig. 9) . We examined these stainings to evaluate colocalization on sections 5 days after crush injury, when all factors were still expressed and just before the onset of RGC loss. Figure 9A shows DIG-ISH for c-jun and IHC for ATF-3. The majority of RGC nuclei were stained dark red-purple, which indicates coexpression at the cellular level. Although smaller in number, singly stained RGCs (c-jun or ATF-3 alone) also were observed. A high rate of colocalization was observed within the GCL, where up to 80% of the ATF-3–expressing cells also expressed c-jun mRNA. Figure 9B shows c-jun DIG-ISH and ATF-2 IHC. In contrast to ATF-3, less colocalization of ATF-2 and c-jun was evident (approximately 50% of the total ATF-2 immunoreactive RGCs), which was expected as most (if not all) DACs also expressed ATF-2. 

Discussion

Axonally injured neuronal cell responses vary upon the cellular context, exemplified by degeneration in the CNS and regeneration in the peripheral nervous system. The ON injury model is widely accepted to be an experimental CNS injury, inasmuch as RGCs will eventually die when injured intraorbitally, sometimes with an initial regenerative reaction. ^{3 41} We started by quantifying RGC loss in our ON crush model because the severity of the degenerative reaction is largely dependent on the crush time and/or operational procedures (e.g., distance from the eye to be crushed). ⁵ In our ON crush model, we did not observe any significant histologic change in the RGCs (by Nissl and H-E staining; data not shown) until 5 days after crush. However, cells rapidly died after this time, leaving only 41.5% of the initial RGC number at 2 weeks, with the majority having died by 8 weeks. This initial moratorium of cell loss and eventual disappearance of the RGCs are in good agreement with another quantitative report of intraorbital ON axotomy, ³ although in our study a slower rate of cell loss was observed. 

Given this predictable timing of the degenerative reaction in the ON injury system, it is tempting to speculate that molecular switch involving transcriptional regulation occurs during this time window. Involvement of transcription factor c-Jun has been implicated in neuronal cell death ^{10 11 18} or survival for the different cellular context (reviewed in Herdegen et al. ²⁹ ). c-Jun mRNA and protein are also expressed and upregulated after axotomy-induced peripheral neuronal injury, where very little cell death occurs. ^{12 13 14} Previous reports show that RGCs also express c-Jun upon ON axotomy ^{9 15 16} and after ON crush. ^{8 17 18} Although c-Jun is well documented to be expressed in RGCs after ON injury, its posttranslational modification by phosphorylation has not been shown in this system. c-Jun has phosphorylation sites at serine 63 and 73 in the transcriptional activation domain, ^{42 43} and phosphorylation of these sites has been shown to potentiate its ability to activate transcription. ^{42 43 44} The expression of phosphorylated c-Jun was also reported after NGF withdrawal stimuli for apoptosis of sympathetic neurons, ²¹ after MCA occlusion, ²² and is required for apoptosis in cerebellar granule neurons. ²³ In this context, we investigated the phosphorylation of c-Jun and found that it occurs very early after injury, coincident with its early transcriptional upregulation (Figs. 2 3) . By simultaneously comparing p-c-Jun–expressing and c-jun mRNA-bearing cells, we did not see any significant difference between them (data not shown). Our findings contradict those of Herdegen et al. ²² where they observed partial phosphorylation of c-Jun among the expressing cells after transection of different central nerve fiber tracts. However, this discrepancy can be due to the different system used in our study or possibly due to the different antibody (recognizing different phosphorylated residue) used in our study. The second possibility is plausible because the phosphorylation on serine 63 is a vast cellular reaction observed in the dorsal root ganglia upon sciatic nerve axotomy. ³³ In this context, it would be informative to review these findings using both antibodies. Recently, an elegant study that targeted a phosphorylation-mutant form of c-jun by a knock-in gene-targeting strategy showed that the phosphorylation of c-Jun is dispensable for mouse development but is nonredundant for kainate-induced neuronal apoptosis. ²⁴ This with other recent reports raises the possibility that phosphorylation of c-Jun per se might not be a major switch for c-Jun–dependent transactivation. ^{45 46 47 48} Our model also indicates that most of the RGCs that produced c-Jun have phosphorylated and have nuclear-translocated c-Jun, regardless of their cell fate. It seems that phosphorylation of c-Jun is not a decisive modification to AP-1 transactivation for cell death in this model. 

We next examined whether ATF-2 could exert such a modulatory role for AP-1 activity. In addition to previous observations, ^{29 30} we found that the diffuse decline of its expression in the GCL cells is further characterized by variegations in the expression level in each cell (Fig. 4) . Although we have been thwarted in our attempt to correlate this finding with RGC fate, a very recent comparative study of ON crush and axotomy ⁴⁹ showed that most of the retrogradely back-filled RGCs coexpressed c-Jun and high levels of ATF-2. It is suggested that ATF-2 favors cell survival when coexpressed with c-Jun. Our finding that the expression level of ATF-2 declines more gradually compared with their finding but that a subtotal RGC loss is still exerted suggests a further role for ATF-2 in cell fate regulation, for example, by its phosphorylation. ⁵⁰  

Because we failed to identify c-Jun as the sole determinant for cell death upon ON injury and could not find direct supportive data for ATF-2 as being decisive for cell fate, we screened other putative AP-1–binding partners that might be correlated with this phenomenon. Of these, ATF-3, a member of the ATF/CREB family, is a good candidate because it has been shown that its mRNA level greatly increases in many cellular stresses, including ischemia, tissue wounding, genotoxicity, and brain seizures. ^{39 51} To date, no evidence of an axonal injury–related induction has been reported. In our model, ATF-3 expression was undetectable before injury but clearly showed a specific upregulation in the RGCs after ON crush. We point out that although this expression pattern is reminiscent of the one for c-jun, in the case of ATF-3, the expression was observed in a smaller population of RGCs when compared to c-Jun and started later but extinguished earlier. Moreover, most of its temporal expression shut down before the start of cell death (Fig. 8) . An unexpected finding was that only a subset of the axotomized RGCs expressed ATF-3 at a given time. This may be explained in two ways: It is possible that only a subset of the RGCs express this molecule over the period, depending on RGC types with different cell diameters or projections. Another more plausible explanation is that ATF-3 is expressed more transiently than c-Jun, and the total number of expressing cells is underscored at a given time, because a diffuse expression of ATF-3 is observed in the affected tissues. ^{39 52} In addition, ATF-3 mRNA contains a degradation signal in its 3′ untranslated region. ⁴⁰  

Two questions arise from these findings of ATF-3. First, what could be the cause of its induction? In Hela cells, ATF-3 is induced by c-Jun NH₂-terminal kinase activation, and cotransfection of ATF-2- and c-Jun expression vectors activates the ATF-3 promoter, presumably by binding to the ATF/CRE or AP-1 sites. ⁵³ The successive detection of c-Jun phosphorylation and ATF-3 in the present ON crush study may thus indicate a c-Jun–related induction of ATF-3. Another appealing possibility is that calcium influx, which occurs after axonal damage ^{54 55 56} might mediate the induction of ATF-3. One report shows a specific induction of ATF-3 in neuroblastoma cells affected by calcium influx and cAMP elevation. ⁵⁷ A transient elevation of intracellular calcium levels shown under these circumstances would be consistent with the brief duration of ATF-3 expression observed in our model. Second, what would be the consequences of ATF-3 induction? This is more difficult to estimate because ATF-3 represses transcription when it homodimerizes but activates transcription with its putative partners when heterodimers are formed. ²⁵ Because our colocalization study indicates that ATF-3 and c-jun mRNA are coexpressed at a high rate after injury (Fig. 9) , heterodimer formation with c-Jun may occur after RGC injury. If ATF-3 works as a transcriptional activator, we can speculate a cell-supportive role by inducing survival factors because this molecule is transiently and selectively expressed during the initial regenerative period. 

In conclusion, our data strongly suggest that the transcriptional regulation by the promiscuously present c-jun is modulated by partner switches between ATF-3 and -2 during nerve injury. Future experiments would be warranted to determine whether c-jun coimmunoprecipitates with different cofactors after axonal injury and/or whether these complexes bind different AP-1 binding sites in different target genes activated at different times after injury. Interestingly, axotomy or crush-induced cell injury of RGC has a certain period of moratorium on cell death ( ^{Refs. 3 6 58,} and this study). Although the exact molecular mechanisms underlying this phenomenon still remains to be elucidated, it is plausible that a gene expression switch occurs in this time window, probably by altering cellular trans-activating properties. Regarding the rapid disappearance of ATF-3 just after this critical point (5 days after crush), we propose a model where c-Jun might temporarily change its partner from ATF-2 to -3 to temporally resist cell death but eventually die by swapping again to other transcription factor(s). 

 Supported by Ministry of Education, Science, Sports and Culture, Ministry of Health and Welfare, and Japan Society for the Promotion of Science (JSPS).

 Submitted for publication September 15, 1999; revised January 24, 2000; accepted March 6, 2000.

 Commercial relationships policy: N.

 

Figure 1.

 

Changes in RGC number after ON crush. Broken line has been inserted to show the proportion of the displaced amacrine cells (DACs) among the counted cells within the GCL. In sham-operated retinas, cells within the GCL were 552 ± 32 cells/section (c/s). Cell number did not change significantly until 5 days after ON crush (P > 0.05) but apparently decreased to 462 ± 25 c/s from 1 week after crush (*P < 0.05). At 8 weeks after operation, cell number dropped to 252 ± 14 c/s. Given that DAC number does not change significantly during this period, the calculated RGC number corresponds to 6.6% of the original number.

Figure 2.

 

P-c-Jun–immunopositive RGCs. Immunohistochemical results on sections are shown in (A) through (F) and flat-mounts in (A′) through (F′). No immunopositive cells are shown in control retina (A, A′). P-c-Jun immunoreactivities are shown at 12 hours (B, B′), 1 day (C, C′), 5 days (D, D′), 2 weeks (E, E′), and 4 weeks (F, F′) after ON crush. Original magnification, sections, ×200; flat-mount, ×100.

Figure 3.

 

The expression of c-jun mRNA shown using high-resolution DIG-ISH. RGCs expressing c-jun are shown at 6 hours (A), 12 hours (B), 1 day (C), 5 days (D), 2 weeks (E), and 8 weeks (F). Original magnification, ×200.

Figure 4.

 

ATF-2 expression after ON crush. ATF-2–immunopositive cells in the GCL (A through C) and mRNA expression (D through F). In control retinas, ATF-2–positive cells were detectable in the GCL and INL (A, D). Expression started to decrease at 3 days, with significant variation in the intensity among expressing cells (B, E). At 4 weeks after ON crush, sparse immunopositive cells were found in the GCL (mostly DACs) (C). mRNA expression also showed weak labeling in a subset of cells within the GCL at this time (F). Original magnification, ×200.

Figure 5.

 

ATF-3 immunoreactive RGCs after ON crush. Immunohistochemical results on sections are shown in (A) through (F) and flat-mounts in (A′) through (F′). ATF-3 immunoreactivity is shown at 12 hours (A, A′), 3 days (B, B′), 1 week (C, C′), and 4 weeks (D, D′) after crush. Original magnification, sections, ×200; flat-mounts, ×100.

Figure 6.

 

ATF-3 expression of digoxigenin-ISH. RGCs expressing ATF-3 mRNA shown at 1 day (A), 3 days (B), 1 week (C), and 4 weeks (D) after ON crush. Original magnification,× 200.

Figure 7.

 

Western blot of ATF-3 (A) and ATF-2 (B). Each lane was loaded with 50 μg of total protein extracted from normal retinas (n) or samples taken at various time points after crush. Numbers on the left indicate protein size in kDa.

Figure 8.

 

Changes in immunoreactive cell number (A) and calculated proportion of immunopositive RGC out of remaining RGCs (B) for c-Jun, ATF-2, and -3. (A) ATF-3–immunopositive RGCs (○) were not detectable 12 hours after ON crush, when p-c-Jun (▵) already started to show immunoreactivity. ATF-3 reached a later peak for its expression compared to p-c-Jun but also disappeared earlier. The number of ATF-2 (□) immunoreactive cells in the GCL started to decrease gradually after ON crush and continued to decrease until 8 weeks. (B) The peak proportional expression of ATF-3 reached 55% of the remaining RGCs at 3 days after crush and 98% for p-c-Jun at 1 day.

Figure 9.

 

Colocalization display using double labeling. (A) DIG-ISH for c-jun (blue cytosolic staining) and IHC for ATF-3 (brown nuclear staining) were simultaneously detected in RGCs. Colocalization of c-jun and ATF-3 is indicated by arrows. RGCs showing single staining for c-jun or ATF-3 also were observed. The ATF-3-only immunoreactive RGC was indicated by an arrowhead. To the right side of this RGC, blue staining is closely situated but represents a different cell cytosol of RGC that expresses c-jun. (B) DIG-ISH for c-jun and IHC of ATF-2. Colocalization of c-jun and ATF-2 is indicated by arrows. The ATF-2–only immunopositive RGCs are indicated by arrowheads. Original magnification, ×400.

The authors thank Nick Allen (Cambridge, UK) for his critical reading of the manuscript, and Koichi Noguchi for the helpful discussion.