Astrocyte-derived CNTF switches mature RGCs to a regenerative state following inflammatory stimulation

Abstract

Retinal ganglion cells (RGCs) normally fail to regenerate injured axons and undergo apoptosis soon after injury. We have recently shown that lens injury (LI) or intravitreally applied zymosan allow RGCs to survive axotomy and regenerate axons in the injured optic nerve. Activated macrophages and oncomodulin have been suggested to be the principal mediators of this phenomenon. However, several lines of evidence show that macrophage-derived factors alone cannot account for all the beneficial effects of intraocular inflammation. We show here that LI or zymosan induce upregulation of ciliary neurotrophic factor (CNTF) in retinal astrocytes and release CNTF independent of macrophages and activate the transcription factor signal transducers and activators of transcription 3 (STAT3) in RGCs. Levels of CNTF expressed in retinal glia and STAT3 activation in RGC were correlated with the time course of RGCs switching to an active regenerative state. Intravitreal injections of antibodies against CNTF or a Janus-kinase inhibitor compromised the beneficial effects of LI, whereas an antiserum against oncomodulin was ineffective. Like the action of CNTF, the effects of LI were potentiated by drugs that increase intracellular cAMP levels, resulting in strong axon regeneration in vivo. These data indicate that astrocyte-derived CNTF is a major contributor to the neuroprotective and axon-growth-promoting effects of LI and zymosan. These findings could lead to the development of a therapeutic principle for promoting axon regeneration in the CNS as a whole.

ciliary neurotrophic factor, lens injury, axonal regeneration, oncomodulin, retinal ganglion cell

Introduction

Neurons of the central nervous system (CNS) do not normally regenerate injured axons. The failure of axonal regeneration is commonly attributed to inhibitory factors associated with myelin and/or the glial scar that is formed at the lesion site and an insufficient intrinsic ability of adult neurons to regrow axons (McKerracher et al., 1994; Mukhopadhyay et al., 1994; Asher et al., 2000; Chen et al., 2000; GrandPre et al., 2000; Morgenstern et al., 2002; Wang et al., 2002; Domeniconi and Filbin, 2005). Retinal ganglion cells (RGCs) fail to regenerate axons and start to undergo apoptosis soon after intraorbital optic nerve crush (ONC) (Berkelaar et al., 1994; Fischer et al., 2000). Research has recently shown that inducing a lens injury (LI) or intravitreally injecting the yeast wall extract zymosan causes a switch of RGCs to a robust regenerative state (Fischer et al., 2000; Leon et al., 2000). This state enables RGCs to regrow axons at higher growth rates, to extend lengthy axons into a peripheral nerve graft and even into the inhibitory environment of the injured optic nerve (Fischer et al., 2000, 2001; Yin et al., 2003). LI or injections of zymosan activate several retinal cell types such as Müller cells and astrocytes (Leon et al., 2000; Pernet and Di Polo, 2006; Hauk et al., 2007). Nevertheless, most attention has focussed on the activation and invasion of macrophages into the inner eye. It has been suggested that activated macrophages are the source of the factors mediating these effects (Leon et al., 2000; Yin et al., 2003), with oncomodulin being the principal mediating factor (Yin et al., 2006). There is general agreement that oncomodulin alone cannot account for all the effects of LI or zymosan because it is not neuroprotective and its axon-growth-promoting effects heavily depend on increased levels of intracellular cAMP in RGCs (Yin et al., 2006). In addition, intravitreal injections of activated macrophages have been reported to be insufficient to induce GAP-43 upregulation or axon regeneration of RGCs when applied intravitreally (Leon et al., 2000). These data indicate that other factors and cells must be also involved in this process. We have recently reported that the switch of RGCs to a regenerative state after LI is closely correlated with the activation of retinal astrocytes and Müller cells (Hauk et al., 2007), and therefore speculated whether these glia might be sources of an alternative factor that contributes to the effects of intraocular inflammation.

We show here that LI or zymosan injections strongly induce the expression and release of the neuroprotective and axon-growth-promoting cytokine ciliary neurotrophic factor (CNTF) from retinal astrocytes, and that its major downstream pathway, the Janus-kinase/signal transducers and activators of transcription 3 (JAK/STAT3) pathway, is strongly activated in regenerating RGCs and that the LI-induced switch to a regenerative state is dependent on both CNTF and JAK activity.

Experimental procedures

Optic nerve crush, lens injury and intravitreal administration

Surgical procedures were approved by the local authorities (Regierungspräsidium Tübingen). Adult female Sprague-Dawley rats weighing 220–250 g were anaesthetized by intraperitoneal injections of ketamine (60–80 mg/kg) and xylazine (10–15 mg/kg), and a 1.0–1.5 cm incision was made in the skin above the right orbit. The optic nerve was surgically exposed under an operating microscope, the dural sheath was longitudinally opened and the nerve was either cut or crushed 1 mm behind the eye for 10 s using a jeweller's forceps, avoiding injury of the central retinal artery. Optic nerve crush (ONC) was verified by the appearance of a translucent region at the lesion site. The vascular integrity of the retina was verified by fundoscopic examination. LI was induced by a retrolenticular approach, puncturing the lens capsule with the tip of a microcapillary tube. LI was supported by intravitreal injections of 15 µl phosphate-buffered saline (PBS) after retrieving the same volume from the anterior chamber of the eye. To study axon regeneration in the crushed optic nerve the following groups (n = 5) were prepared: ONC + intravitreal injection of bovine serum albumin (BSA), ONC + intravitreal injections of CNTF, ONC + LI and ONC + LI + intravitreal injection of dibutyryl-cAMP. Injections were repeated after 3 days.

To functionally assess the regenerative state of RGCs we used dissociated retinal cultures or retinal explants. To prepare retinas in vivo, animals (n = 4 for each group) received a complete optic nerve cut and intravitreal injections of (BSA) (1.5 µg/10 µl), same amounts of recombinant CNTF (Serotec), 10 µl (10 mM) of dibutyryl-cAMP (Sigma) solution, or combinations of each. Injections were repeated after 3 days. Other groups (n = 4 for each group) received ONC, LI + intravitreal injections of BSA (1.5 µg/10 µl), 5 µl of an anti-oncomodulin-antiserum (Swant, Switzerland, order number OM3, lot no.: 1§1, Swant distributes only this batch), suitable to neutralize oncomodulin and to be used for functional assays (Yin et al., 2006) in 10 µl with PBS, 2.5 µg of an anti-CNTF antibody (AAR21, Serotec, suitable for functional assays) diluted to 10 µl PBS, 2.5 µg of an anti-parvalbumin antibody (Santa-Cruz) in 10 µl of PBS or 10 µl (17 mM) of the Janus-kinase-2 (JAK) inhibitor AG490 (Calbiochem). To maintain intraocular concentrations high injections were repeated after 2 and 4 days. Retinas were isolated and prepared for cell-culture experiments 5 days after ONC. The suitability of the CNTF-antibody to neutralize the CNTF effects was verified in cell culture (data not shown).

Retinal cell cultures

Tissue culture plates (4-well plates; Nunc, Wiesbaden, Germany) were coated with poly-d-lysine (0.1 mg/ml, molecular weight <300 000 Da; Sigma), rinsed with distilled water and then air-dried. Wells were then coated with laminin (20 µg/ml) (Sigma). To prepare retinal cultures, pretreated animals were killed by an overdose of chloralhydrate solution (7%). Retinas were rapidly dissected from the eyecups and incubated at 37°C for 30 min in a digestion solution containing papain (16.4 U/ml, Worthington; Katarinen, Germany) and l-cysteine (0.3 µg/ml, Sigma) in Dulbecco's Modified Eagle medium (DMEM) (Invitrogen). Retinas were then rinsed with DMEM and triturated in 2 ml DMEM containing B27-supplement (Gibco) (1 : 50) and penicillin/streptomycin (0.2 mg/ml, Biochrom). Dissociated cells were then passed through a cell strainer (40 µm, Falcon). The cell suspension of one retina was adjusted with medium to a volume of 5 ml. Five-hundred microlitres of cell suspension were added into each well and arranged in a pseudo-randomized manner on the plates so that the investigator would not be aware of their identity. After 24 h in culture, cells were fixed with a paraformaldehyde solution (4%) and methanol (Sigma), and prepared for immunocytochemical staining with a βIII-tubulin-antibody (TUJ-1) (Babco, Richmond, CA; 1 : 2000). Eighty pictures per well were taken under a fluorescent microscope (200×). Axon lengths were determined using the AxioVision-software (Zeiss), and at least 100 RGCs were used per well. Values are given as the means ± SEM of four replicate wells. Furthermore, the number of RGCs per well was quantified to test for potential neurotoxic or neuroprotective effects after each treatment. Statistical significance was analysed by a two-tailed t-test (assuming equal variances). Each experiment was repeated in total at least three times.

Retinal tissue cultures

Five days after ONC and LI or sham intraocular surgery, rats were killed and their retinas dissected (n = 4 animals per group). Some groups received intravitreal injections of AG490 or PBS simultaneously to surgery. Injections were repeated after 2 and 4 days. Isolated retinas were cut into eight radial pieces and cultured in DMEM containing B27-supplement (Gibco) (1 : 50) and penicillin/streptomycin (0.2 mg/ml, Biochrom) in laminin-poly-d-lysine coated petriperm dishes (Greiner bio-one) (Bahr et al., 1988). After 2 days, explants were fixed with a paraformaldehyde solution (4%) and methanol. Axons were then stained with a monoclonal antibody against βIII-tubulin (TUJ-1) (Babco, Richmond, CA; 1 : 2000). The number of axons extending ≥50 µm from each explant was counted using a fluorescent microscope (200×, Axiovert, Carl Zeiss, Germany) and a calibrated ocular micrometer. The results from individual explants were averaged within each experimental group and differences between the groups were evaluated using a two-tailed t-test (assuming equal variances).

Separation of retina and vitreous body and RNA isolation

Retinas together with vitreous bodies were isolated and quickly flat-mounted on a nitrocellulose filter and then transferred to a Whatman filter. The vitreous body sticking on top of the retina was carefully separated from the retina by means of two jeweller's forceps as one piece. After LI lens fragments were visible in the isolated vitreous bodies, verifying that cells and small particles were retained in the isolated vitreous bodies. Vitreous bodies were transferred to different tubes either for protein preparation (20 mM Tris/HCl pH 7.5, 10 mM KCl, 250 mM sucrose, 10 mM NaF, 1 mM DTT, 0.1 mM Na₃VO₄, 1% TritonX100, 0.1% SDS) with 1/100 protease inhibitor (Calbiochem, California, USA) or lysis buffer for RNA extraction using the RNeasy-kit (Qiagen, Hilden, Germany) according to the manufacturer's protocol.

Western blot assays

Rats were killed and their eyeballs enucleated and dissected. Retinas were collected in lysis buffer (20 mM Tris/HCl pH 7.5, 10 mM KCl, 250 mM sucrose, 10 mM NaF, 1 mM DTT, 0.1 mM Na₃VO₄, 1% TritonX100, 0.1% SDS) with 1/100 protease inhibitor (Calbiochem, California, USA). Retinas were homogenized and centrifuged at 5000 rpm for 10 min. The supernatants were analysed by Western blot assay. Separation of proteins was performed by 12% sodium-dodecyl-sulphate–polyacrylamide gel electrophoresis (SDS–PAGE), according to standard protocols (Bio-Rad, Hercules, USA). After SDS–PAGE, proteins were transferred to nitrocellulose membranes (Amersham, UK). The blots were blocked in 5% dried milk in Tris-buffered saline-Tween-20 (TBS-T) and processed for immunostaining with a monoclonal primary antibody against growth-associated protein 43 (GAP-43, dilution 1 : 1000; a gift from Dr Benowitz, Children's Hospital, Harvard Medical School, Boston), an antiserum against rat phospho-STAT3 (Tyr705) (Cell Signaling, 1 : 1000), an antiserum against rat STAT3 (Cell Signaling, 1 : 1000), a monoclonal antibody against rat β-actin (Sigma) (1 : 7500) or a polyclonal antibody against rat CNTF (Serotec, 1 : 5000) at 4°C overnight. Bound antibodies were visualized with anti-sheep, anti-rabbit or anti-mouse immunoglobulin G (IgG) secondary antibodies conjugated with horse-radish peroxidase diluted at 1 : 80 000 (all Sigma, St Louis, USA). The antigen–antibody complexes were detected by enhanced chemoluminescence (ECL, Amersham, Buckinghamshire, UK).

Immunohistochemistry

Animals were anaesthetized and perfused through the heart with cold saline followed by PBS containing 4% paraformaldehyde. Eyes attached to optic nerve segments were separated from connective tissue, post-fixed for several hours, transferred to 30% sucrose overnight (4°C) and embedded in Tissue-Tek (Sakura). Frozen sections were longitudinally cut on a cryostat, thaw-mounted onto coated glass slides (Superfrost plus, Fisher, Pittsburgh, PA) and stored at −80°C until further use. Sheep anti-GAP-43 was used at a dilution of 1 : 1000. Monoclonal antibodies against βIII-tubulin (1 : 2000, TUJ-1, Babco, Richmond, CA, USA) and against glial fibrillary acid protein (1 : 50, Santa Cruz) and the polyclonal antibodies against rat phospho-STAT3 (Tyr705) (1 : 500, Cell Signaling), rat total STAT3 (1 : 500, Cell Signaling) and rat CNTF (1 : 10000; Serotec) were used. Secondary antibodies included anti-mouse IgG, anti-rabbit IgG and anti-sheep IgG antibodies conjugated to Alexa Fluor 488 and Alexa Fluor 594 (1 : 10000; Molecular Probes). To stain nuclei, sections were incubated in a solution containing 4′,6-diamidino-2-phenylindol (DAPI) for 1 min. Fluorescent sections were covered using Mowiol and analysed under a microscope.

Axon regeneration: quantitation

Regeneration was quantified as described previously (Leon et al., 2000). In brief, under 400× magnification, we counted the number of GAP-43-positive axons extending ≥250 µm and ≥500 µm from the injury site in eight sections per case, normalized these values to the cross-sectional width of the optic nerve, and used these data to calculate the total numbers of regenerating axons in each animal (Leon et al., 2000). The significances of inter-group differences were evaluated using a two-tailed t-test (assuming equal variances).

Reverse-transcription polymerase chain reaction (RT-PCR)

Total RNA was isolated from retinas or cells in the vitreous body including macrophages and lens fibre cells from the damaged lens after LI, peritoneal macrophages or an immortalized retinal Müller cell line (rMC-1; generously provided by Dr Sarthy, Northwestern University, Chicago, USA) using the RNeasy-kit (Qiagen) according to the manufacturer's protocol. Superscript II (Invitrogen) was used to generate cDNA. The following primers were used for the polymerase chain reaction (PCR): glyeraldehyde-3-phoshate dehydrogenase (GAPDH), forward 5′-GGCCAAGGTCATCCATGACAACTT-3′, reverse 5′-CCTGCTTCACCACCTTCTTGATGT-3′; CD68, forward 5′-GGTACTGCTTGTAGCCCAAGGA-3′, reverse 5′-GGTTCAATACAGAGAGGCCCCA-3′; CNTF, forward 5′-TTTCGCAGAGCAAACACC-3′, reverse 5′-AACTGTGGCAGGCATCC-3′. To compare samples in the linear range of amplification, RT-PCR was stopped and loaded on agarose gels after 25, 30, 35 and 40 cycles. To verify the accuracy of measured expression differences with CNTF and the GAPDH primers, 1/2 and 1/5 of the cDNA starting material of retinas treated with ONC + LI was used.

Results

LI and intraocular zymosan injections induce CNTF expression in retinal astrocytes

Using RT-PCR, Western blot and immunohistochemistry analyses we detected low levels of CNTF in naive retinas (Fig. 1A, B, O and P), slightly increased levels at 5 days after ONC (Fig. 1C, D, O and P), and strongly increased expression levels after zymosan injection, LI either alone or in combination with ONC (Fig. 1E–L, O and P). Immunohistochemical staining detected most CNTF in retinal astrocytes (Fig. 1E–L). Lower amounts of CNTF were also detected in Müller cells. Activated macrophages that had infiltrated the inner eye after LI or zymosan injections were CNTF-negative (Fig. 2D).

Fig. 1

LI and intravitreal injection of zymosan induce CNTF expression in retinal astrocytes: Immunohistochemical staining of the retina of: (A, B) untreated controls; (C, D) 5 days after optic nerve crush (ONC); (E, F) 5 days after lens injury (LI) alone; (G, H) 5 days after ONC + LI; (I, J) 5 days after ONC + zymosan injection using an anti-CNTF (red) and anti-βIII tubulin antibody (green), respectively. (K, L) Retina 5 days after LI + ONC doubled-stained with an anti-GFAP antibody indicating the colocalization of CNTF- (red) and GFAP-positive astrocytes and Müller cells (green). (GCL = ganglion cell layer; INL = inner nuclear layer) (M, N) High-power magnification of the ganglion cell layer exposed to LI 5 days previously. CNTF (red) is also detected on the surface of βIII-tubulin-positive RGCs (green), indicated by arrows. (O) RT-PCR. RNA was extracted from retinal/vitreal samples 5 days after ONC, LI, ONC + LI, and from untreated controls (con). CNTF expression levels in retinas were slightly increased after ONC and strongly increased after LI, either alone or in combination with ONC. Müller-cell-line derived RNA (MC) served as a positive control. RT-PCR of the housekeeping gene GAPDH verified that same amounts of RNA were used for each PCR. One-fifth (1/5) and one-half (1/2) of the cDNA material of LI + ONC treated retina was used to verify that the band intensities were correlated with expression levels of GAPDH and CNTF. (P) Western blot. Retinal/vitreal lysates of samples as used for RT-PCR (O) confirm results on the RNA level. Recombinant CNTF (rCNTF) served as a positive control. Bands of β-actin verified that the same amounts of protein were loaded per lane. (Scale bars: A–L: 50 µm; M, N: 10 µm).

Fig. 2

LI- or zymosan-induced CNTF upregulation is not dependent on the presence of activated macrophages or other peripherally circulating cells (A) Western blot. After LI, or no LI but intravitreal injection of PBS (con) retinas and vitreous bodies were carefully separated. CNTF expression was strongly increased in the retina 5 days following LI. CNTF was detectable in the vitreous body of LI-treated eyes but not of PBS-treated controls, suggesting that CNTF is released from retinal astrocytes into the vitreous body. Peritoneal macrophages (MP) did not express CNTF protein. Recombinant CNTF (rCNTF) served as positive control.(B) RT-PCR. Retinas and vitreous bodies were isolated 5 days after various treatments in vivo. CNTF upregulation in retinas was measured after LI treatment. RT-PCR of cells derived from the vitreous body (vit) that was carefully separated from the retina after LI treatment showed a strong signal for CD68 (macrophages) but not for CNTF, indicating that vitreal cells did not contribute to the CNTF upregulation. Also peritoneal macrophages (MP) strongly expressed CD68 but not CNTF, whereas cultured Müller cells (MC) expressed CNTF but not CD68. (C) Western blot. Untreated retinas were isolated and cultured for 2 days in the presence of increasing concentrations of BSA, lens proteins (derived from the water soluble fraction of lens homogenate) or zymosan. In contrast to BSA, exposure of retinas to lens proteins or zymosan increased retinal CNTF expression in a dose-dependent manner. (D) Immunohistochemistry. Astrocytes in the fibre layer were positively stained with an anti-CNTF antibody (red). Activated macrophages that infiltrated the inner eye after LI (indicated by arrows) and that were stained with an ED-1 antibody (green) were CNTF-negative. (GCL = ganglion cell layer, INL = inner nuclear layer; scale bar = 50 µm).

High-power magnification showed some CNTF-staining also on the surface of RGCs, suggesting that the cytokine was released from activated glial cells (Fig. 1M and N). To confirm CNTF release from retinal cells, we carefully separated the vitreous body from the retinas of eyes treated with or without LI but PBS injection instead. In contrast to PBS-treated controls, eyes subjected to LI showed detectable CNTF protein levels in the vitreous body (Fig. 2A) but no CNTF RNA (Fig. 2B). The vitreous bodies of LI-treated eyes were CD68- (Fig. 2B) and β-actin positive (Fig. 2A), indicating the presence of activated macrophages. These data suggest that CNTF was most likely released from astrocytes and was not derived from cells located in the vitreous body.

Lens proteins and zymosan stimulate CNTF expression in astrocytes in a macrophage-independent manner

LI and zymosan induce a strong invasion of activated macrophages into the eye (Leon et al., 2000; Yin et al., 2003; Pernet and Di Polo, 2006; Hauk et al., 2007). We therefore speculated whether the induction of CNTF in retinal astrocytes might be indirectly mediated by activated macrophages. To test this possibility, we isolated and cultured untreated retinas for 2 days in the presence of increasing concentrations of lens proteins, zymosan or BSA in the culture medium. In contrast to BSA, addition of same concentrations of lens proteins or zymosan increased the level of retinal CNTF expression in a dose dependent manner (Fig. 2C), indicating that the induction of CNTF expression in retinas occurs also ex vivo and is independent of the presence of activated macrophages or other peripherally circulating cells.

JAK/STAT3 pathway is activated in RGCs following LI

CNTF elicits its primary biological actions via several pathways including the Janus-kinase/signal transducers and activators of transcription 3 (JAK/STAT3) pathway, the phosphatidylinositol 3-kinase (PI3K/Akt) pathway, and the extracellular signal-regulated kinase (ERK)/mitogen activated protein kinase (MAPK/ERK) pathway (Park et al., 2004; Miao et al., 2006). To investigate whether LI activates one of these pathways in RGCs, we stained retinas with an antibody specific to the phosphorylated form of the transcription factor STAT3 (pSTAT3) 5 days after various treatments. As shown by immunohistochemistry (Fig. 3A–D, G and H) and confirmed by Western blot analysis (Fig. 3E), naive retinas showed no or very low levels of the activated transcription factor. Levels slightly increased after ONC and intravitreal injections of saline (Fig. 3B). Amounts of pSTAT3 strongly increased in retinas after exposure to LI alone, or LI + ONC (Fig. 3C and D). As shown by immunohistochemistry, the greatest pSTAT3 levels were detected in the nuclei of RGCs following LI (Fig. 3D–H; Supplementary Fig. 1). Levels were also increased in the nuclei of cells of the inner nuclear layer (Fig. 3D; Supplementary Fig. 1). Total retinal STAT3 was also higher expressed after ONC alone and was further increased after LI or the combination of both treatments (Fig. 3F). However, total STAT3 expression levels were unchanged in RGCs after ONC, LI or ONC + LI compared with untreated controls. The increase in total STAT3 level was due to upregulation of the protein mainly in cells of the fibre layer (astrocytes) as shown by immunohistochemical staining (Fig. 3I–P; Supplementary Fig. 2).

Fig. 3

LI activates JAK/STAT3 pathway in RGCs. Immunohistochemical staining of retina of (A) untreated control; (B) 5 days after optic nerve crush (ONC); (C) 5 days after lens injury (LI); (D) 5 days after ONC + LI using a specific antibody against phospho-STAT3. (G, H) Phospho-STAT3 levels were strongly increased in RGC nuclei after LI alone or LI + ONC, as shown by double-staining with an anti-βIII-tubulin antibody (G) and DAPI staining (H). (E, F) Western blot analysis of retinal lysates 5 days after ONC, LI, LI + ONC, or no previous treatment (con) using a specific antibody against phospho-STAT3 (E) or an antibody against total STAT3 (F). (I–P) Immunohistochemical staining of retinas 5 days after ONC (J, N), LI (K, O), ONC + LI (L, P) or no previous treatment (I, M) using an antibody against total STAT3 (red) (I–L) and double-staining with an anti-βIII-tubulin antibody staining RGCs (green) (M–P). Total STAT3 levels remained unaffected in RGCs after ONC, LI, ONC and LI.

To investigate whether the time course of STAT3 activation was correlated with CNTF expression in astrocytes and the regenerative state of RGCs we isolated retinas 0, 1, 2, 3, 5, 7, 10 and 14 days after ONC + LI treatment and measured retinal CNTF expression, pSTAT3 and GAP-43-levels by Western blot assays (Fig. 4F). Retinal CNTF levels increased as soon as one day after LI, reached a plateau after 3 days and remained increased during the total observation period of 14 days (Fig. 4F). The activation of STAT3 in RGCs was correlated with retinal CNTF expression. Phospho-STAT3 levels were detectably increased as soon as 1 day after LI, or LI + ONC treatments and reached a maximum levels after 2 and 3 days, respectively (Fig. 4F and Supplementary Fig. 1). The amount of retinal pSTAT3 decreased afterwards but remained increased even 14 days after LI and ONC compared with untreated controls. This reduction is likely to be due to induction of the expression of the protein suppressor of cytokine signalling-3 (SOCS-3), which has negative feedback effects on STAT3 activation (Starr et al., 1997; Fischer et al., 2004c; Miao et al., 2006) and which we have previously shown to be upregulated in RGCs after LI (Fischer et al., 2004b). The activation of STAT3 was also correlated with RGCs’ regenerative state: GAP-43 protein levels in retinas measurably increased 3 days after ONC + LI and reached a plateau after 5 days. The beginning of the upregulation of GAP-43 expression coincided with the time point when pSTAT3 levels were highest and most RGCs entered a regenerative state, as shown by the number of regenerating axons from retinal explants (Fig. 4A–E). Thus, CNTF expression and STAT3 activation occurred approximately 1 day before GAP-43 upregulation and at the same time at which most RGCs enter a regenerative state.

Fig. 4

Correlation of lens injury (LI) induced CNTF expression, activation of the JAK/STAT3 pathway and regenerative state of RGCs. (A–D) Retinal explants from mature rats after 48 h in culture and subsequent βIII-tubulin-staining. Retinas were subjected to ONC + LI 3 days (A), 7 days (B), 10 days (C) and 14 days (D) beforehand. Scale bar: 100 µm. (E) Quantitation of results: average number of outgrowing axons per explant after 48 h in culture. Data present results from normal retinas (0) or from retinas subjected to ONC + LI 1, 2, 3, 5, 7, 10, 14 days beforehand, as indicated. Most RGCs start to go into a regenerative state 3 days after ONC + LI. Numbers of outgrowing axons steadily increase over the time period of 14 days, indicating that surviving RGCs remain in an active growth state 2 weeks after ONC and LI. (F) Western blot analysis. Lysates of retinas treated as described in A–E were prepared and tested for CNTF, pSTAT3 and GAP-43 expression. β-actin expression verified that the same amount of protein was loaded per lane.

GAP-43 expression remained increased 14 days after ONC + LI, although at this time point ∼60% of RGCs had already undergone apoptosis (Fig. 4F) (Fischer et al., 2000; Leon et al., 2000; Hauk et al., 2007). These data imply that the RGCs that survived ONC increased GAP-43 expression even further. We therefore tested whether measured GAP-43 levels were correlated with the ability of RGCs to extend axons from retinal explants. In agreement with the results of Fischer et al. (2004b), axon outgrowth from retinal explants was markedly increased between 3 + 5 days after ONC + LI, and increased further when retinas were explanted 10 and 14 days after surgery (Fig. 4A–E). These data indicate that surviving RGCs remain in a robust regenerative state 2 weeks after ONC + LI.

Effects of LI are CNTF- and JAK-dependent

To test whether CNTF and subsequent JAK activation are functionally involved in mediating LI effects, we intravitreally applied a bioactive CNTF antibody or the JAK2 inhibitor AG490, which has recently been shown to be sufficient to compromise the neuroprotective and axon-growth-promoting effects of CNTF (Park et al., 2004; Teng and Tang, 2006). In preliminary experiments we confirmed that the CNTF antibody that we used for the following experiments or AG490 reduced CNTF-mediated axonal outgrowth of RGCs in culture (data not shown). Owing to the fact that CNTF is continuously released from astrocytes after LI, the anatomical proximity of astrocytes and RGCs, and the short half-life time of intravitreally applied antibodies we predicted that a single intravitreal injection of an anti-CNTF antibody would not be sufficient to neutralize enough CNTF to show measurable effects. Thus, to maintain high antibody concentrations, we intravitreally applied the anti-CNTF antibody, or an anti-parvalbumin antibody as a control at the same time as LI and again after 2 and 4 days. After 5 days, we isolated retinas and prepared dissociated retinal cell cultures. Additional groups of animals underwent the same treatment but received injections of an anti-oncomodulin antiserum that has recently been used to successfully eliminate oncomodulin from a macrophage-cell-line-conditioned medium (Yin et al., 2006) or same amounts of BSA. After 24 h cells were fixed and RGCs were stained with an anti-βIII-tubulin antibody. To determine the regenerative state of these RGCs and to test for potential effects of each treatment on neuronal survival we measured the average length of regenerating axons after various treatments and number of RGCs/well. Treatments did not affect the survival rate as determined by the number of RGCs/well (data not shown). However axon regeneration was affected. The average lengths of regrowing axons of each group after prior treatment in vivo are summarized in Fig. 5H. Controls that underwent ONC and received an intravitreal injection of BSA showed very little axon outgrowth after 24 h in culture (Fig. 5A, G and H), whereas intravitreal injections of CNTF or LI caused strong outgrowth (Fig. 5B, C and G). LI + intravitreal BSA treatment or LI + intravitreal injections of an anti-parvalbumin antibody resulted in similar strong axon outgrowth as LI alone (Fig. 5G and H). Intravitreal injections of the anti-oncomodulin antiserum did not reduce the effects of LI at all, suggesting that oncomodulin is not the principal mediator of this phenomenon and that LI effects are triggered by an alternative mechanism. In contrast, treatment with the anti-CNTF antibody or the JAK inhibitor significantly and reproducibly reduced the effects of LI treatment. These data suggest that CNTF and the activation of its main downstream pathway are essentially involved in mediating the effects of LI. As verified by immunohistochemistry analysis, AG490 treatment blocked the LI-induced STAT3 phosphorylation in RGCs (Fig. 5I–L), whereas the increased pSTAT3 levels in the nuclei of cells in the inner nuclear layer were not measurably affected. This is probably because intravitreally applied AG490 did not penetrate deeply enough into the retina and therefore did not reach sufficiently high concentrations in the inner nuclear layer. AG490 treatment in vivo did not reduce the survival of RGCs 5 days after ONC and LI, as determined by quantifying RGCs/eye section (data not shown). However it also abrogated the effects of LI when the regenerative state was assessed using retinal explants (Fig. 5M–O) and therefore confirmed the results measured in dissociated cell cultures.

Fig. 5

Axon-growth-promoting effects of LI are CNTF and JAK dependent. (A–F) Dissociated retinal cell cultures immunostained with an antibody against βIII-tubulin, showing regenerating RGCs after 24 h in culture. Eyes had received various treatments 5 days previously. The optic nerves of all animals were intraorbitally cut and eyes received two intravitreal injections at the same time as LI and ONC and again after 3 days according to the following protocol: (A) BSA injections; (B) CNTF injections; (C) LI; (D) dibutyryl-cAMP (db-cAMP) injections; (E) CNTF and db-cAMP injections; (F) LI and db-cAMP injections. (G) Quantitation of axonal outgrowth of A–F normalized to values received from the group treated with ONC + LI. (H) Quantitation of axon outgrowth of dissociated retinal cultures after various treatments in vivo: Animals were subjected to optic nerve cut, LI and three intravitreal injections of the following agents at the same time as LI and again 2 and 4 days after surgery: BSA, rabbit anti-rat parvalbumin antibody (a-Par), rabbit anti-rat oncomodulin antiserum (α-OM), rabbit anti-rat CNTF antibody (α-CNTF) and AG490 (JAKi). Controls received optic nerve cut and repeated injections of BSA without LI. Injections of the anti-CNTF antibody or JAK-inhibitor compromised the LI effects on axon outgrowth, whereas anti-parvalbumin or anti-oncomodulin antiserum showed no effects. **P < 0.001 compared with LI and BSA treatment; ^††P< 0.001 compared with LI and anti-parvalbumin treatment (α-Par). (I–L) Staining of retinas with antibodies against phospho-STAT3 (I, J) and βIII-tubulin (K, L) 5 days after LI treatment and intravitreal injections of vehicle (I, K) or AG490 (J, L). Intravitreal injections of AG490 abrogated LI-induced activation of the JAK/STAT3 pathway in RGCs whereas the activation in nuclei of the inner nuclear layer was unaffected. (M) Blocking LI-induced JAK/STAT3 activation by JAK-inhibitor treatment reduced the beneficial effects on axon regeneration from retinal explants, confirming the results from dissociated retinal cell cultures. **P < 0.001 compared with control (con); ^††P< 0.001 compared with LI-treatment. (N) Retinal explant 5 days after LI + ONC; (O) Explant 5 days after LI, ONC + intraocular injections of AG490. Scale bar: 100 µm.

Dibutyryl-cAMP treatment potentiates LI effects

Increases in intraocular cAMP levels potentiate CNTF-induced regeneration of RGC axons (Cui et al., 2003; Park et al., 2004). As our data indicate, CNTF is involved in mediating the beneficial effects of intraocular inflammation, thus we predict that a combinatorial treatment of LI and dibutyryl-cAMP would increase the effects of LI further. To test this hypothesis we performed ONC + LI, or alternatively injected recombinant CNTF, or injected BSA as a control. Other groups received a combinatorial treatment of LI, CNTF or BSA with intravitreal injections of dibutyryl-cAMP. All injections were repeated after 3 days. Retinal cell cultures were prepared after 5 days and the average axon length of regenerating RGCs was measured after 24 h in culture. The average axon length of RGCs of each group is shown in Fig. 5G. Intravitreal injections of BSA resulted in weak axon regeneration (Fig. 5A and G), whereas additional injections of dibutyryl-cAMP with BSA increased axon outgrowth significantly (Fig. 5D and G). LI or CNTF treatment strongly stimulated axon outgrowth of RGCs to the same extent (Fig. 5B, C and G), showing that, at sufficiently high concentrations, intravitreally applied CNTF is as potent as LI. The effects of CNTF and LI were significantly increased when animals received additional dibutyryl-cAMP injections (Fig. 5E, F and G). In combination with dibutyryl-cAMP, LI induced a stronger outgrowth than CNTF, indicating that other cAMP-dependent factors in addition to CNTF might contribute to the LI effects.

We then tested whether repeated injections of CNTF were sufficient to promote axon regeneration into the crushed optic nerve. We intravitreally applied either BSA, as a control, or CNTF at the same time as the animals underwent ONC and repeated the injections after 3 days. Owing to the risk of damaging the lens we did not perform additional injections. Two weeks after surgery we quantified the number of GAP-43 positive axons growing ≥0.25 and ≥0.5 mm beyond the lesion site of the optic nerve. GAP-43 immunoreactivity in axons extending beyond the injury site provides evidence that these are regenerating rather than spared axons, because the latter do not show detectable levels of GAP-43 (Schaden et al., 1994; Berry et al., 1996). The numbers of regenerating axons after each treatment are shown in Fig. 6E and the survival rates (RGCs/retinal section) in Fig. 6F. Two injections of CNTF were sufficient for regeneration of axons into the distal optic nerve, compared with controls that received BSA injections and were ∼75% (≥250 µm) and 56% (≥500 µm) of the LI effects. CNTF rescued ∼30% of the numbers of axotomized RGCs rescued by LI. As dibutyryl-cAMP injections potentiated the LI effects on axon outgrowth when evaluated on a growth-permissive substrate in culture, we investigated whether a combinatorial treatment of dibutyryl-cAMP and LI would result in more axon regeneration into the lesioned optic nerve than LI alone. Dibutyryl-cAMP was applied at the same time as ONC and LI, and injections were repeated after 3 and 7 days. Controls were treated accordingly but received injections of PBS instead. Combinatorial treatment with dibutyryl-cAMP and LI increased the numbers of axons regenerating ≥0.25 and ≥0.5 mm beyond the lesion site by a factor of 2.2 compared with animals treated with LI alone (Fig. 6E). The survival rate of axotomized RGCs was increased by ∼22%, as shown by the number of surviving RGCs per eye section (Fig. 6F).

Fig. 6

Axon regeneration and RGC survival in vivo after repeated intravitreal CNTF application and potentiation of LI effects by increasing intraocular cAMP levels. Longitudinal sections through the optic nerve showing GAP-43-positive axons distal to the injury site (asterisk) 2 weeks after ONC and two intravitreal injections of a solution containing (A) BSA; (B) LI + intravitreal injections of PBS simultaneously to ONC and again after 3 and 7 days; (C) intravitreal injections of recombinant CNTF at the same time as ONC and again after 3 days; (D) LI and repeated intravitreal injections of dibutyryl-cAMP (db-cAMP) simultaneously to ONC + LI and again after 3 and 7 days. Scale bar = 100 µm. (E) Quantitation of regeneration (number of axons growing ≥250 and ≥500 µm beyond the injury site per optic nerve) 2 weeks after surgery. (F) Quantitation of surviving RGCs (βIII-tubulin-positive RGCs per retinal cross section) 2 weeks after ONC. **P< 0.001 compared with BSA-treated controls; ^††P< 0.001 compared with LI-treated control. Total = number of RGCs in naive retina.

Discussion

The main new findings of this study are: (i) LI and intraocular zymosan treatment induce CNTF expression in retinal astrocytes independently of peripherally circulating cells; (ii) CNTF release and activation of the JAK are essentially involved in mediating the axon growth promoting effects of intraocular inflammation; (iii) LI effects are potentiated by drugs that increase intracellular cAMP and (iv) surviving RGCs remain in an active regenerative state for at least 2 weeks after LI + ONC. From these results we conclude that continuously released glial-derived CNTF is a key-factor in mediating the beneficial effects of intraocular inflammation on axotomized RGCs.

Inflammatory reactions in the inner eye enable RGCs to survive axotomy and to regenerate lengthy axons beyond the lesion site of the injured optic nerve (Fischer et al., 2000, 2001, 2004a, b; Leon et al., 2000; Yin et al., 2003; Lorber et al., 2005; Pernet and Di Polo, 2006). Activated macrophages have been proposed to be the main mediators of these effects (Leon et al., 2000; Yin et al., 2003) and oncomodulin was suggested to be the principle factor triggering this phenomenon (Yin et al., 2006). However, oncomodulin or macrophages cannot account for all the beneficial effects of LI or zymosan. Oncomodulin is not neuroprotective and its reported axon-growth-promoting effects in vivo and in vitro are dependent on the application of drugs elevating intracellular cAMP (Yin et al., 2006). The mechanism by which this dependence is overcome under inflammatory conditions is not yet clear (Filbin, 2006). It has also been shown that an intravitreal injection of activated macrophages alone was not sufficient to induce an upregulation of GAP-43 or axon regeneration of RGCs (Leon et al., 2000). In agreement with this, we have recently shown that the switch of RGCs to an active regenerative state also occurs in the presence of very low numbers of activated macrophages (Hauk et al., 2007), and several studies have shown that lens-derived factors contribute to the axon growth promoting effects in a macrophage-independent manner (Lorber et al., 2002, 2005; Stupp et al., 2005). Together, these data suggest that other factors, which are not likely to be derived from macrophages, also contribute to the beneficial effects of intraocular inflammation. Consistent with this idea, in the present study we show that intravitreally applied anti-oncomodulin antiserum, which has previously been successfully used to eliminate the axon-growth-promoting effects of a macrophage-cell-line-conditioned medium (Yin et al., 2006), does not compromise LI effects, providing further evidence that oncomodulin is not essentially involved in switching RGCs to a regenerative state.

We have recently shown that the regenerative state of RGCs is closely correlated with activation of retinal astrocytes and Müller cells in vivo (Hauk et al., 2007). These glial cells have been demonstrated to express CNTF (Lee et al., 1997; Ju et al., 1999; Chun et al., 2000; Sarup et al., 2004), a factor that protects axotomized RGCs from apoptosis, promotes regeneration of their axons and whose receptors are expressed by RGCs (Mey and Thanos, 1993; Meyer-Franke et al., 1995; Lee et al., 1997; Cui et al., 1999; Ju et al., 1999; Cui and Harvey, 2000; Cui et al., 2003; Sarup et al., 2004; Leaver et al., 2006). Although the neuroprotective and axon growth promoting effects of CNTF are potentiated by drugs that increase levels of intracellular cAMP, CNTF effects are, contrary to those of oncomodulin, not necessarily dependent on these drugs. We demonstrate here that LI or an intravitreal injection of zymosan induce CNTF expression in retinal glial cells. This induction also occurs ex vivo in the absence of activated macrophages or other peripheral cells circulating in the blood, for example, when lens proteins or zymosan were added to the medium of cultured retinas. That CNTF induction is independent of activated macrophages is further supported by the fact that CNTF upregulation occurs as rapidly as 1 day after LI, whereas macrophages first appear in the vitreous body 2–3 days after LI (Leon et al., 2000; Hauk et al., 2007). Although the mechanism underlying CNTF induction in retinal glia remains to be determined, it is likely the result of a direct interaction of lental components or zymosan with these cells. Recent publications have shown that astrocytes express the zymosan receptor Toll-like receptor-2, suggesting that they might be sensitive towards the yeast wall extract and that CNTF expression might be induced by this mechanism (Gantner et al., 2003). A direct induction of CNTF expression in astrocytes and its subsequent release might also explain the reports that lens-derived factors promote axon growth from retinal explants or dissociated cultures when added to the culture medium (Lorber et al., 2002, 2005; Stupp et al., 2005; Liedtke et al., 2007). These effects could be indirectly, at least partially, mediated by CNTF released from activated astrocytes. Nevertheless, lens-derived factors might also exert these effects through an alternative, so far unknown mechanism, since the addition of an antibody against glycoprotein 130 (gp130), which is an essential part of the CNTF-receptor complex, did reportedly not measurably abrogated the axon-growth-promoting effects of lens-derived factors in dissociated retinal cultures (Lorber et al., 2002).

We show here that intravitreal injections of a bioactive anti-CNTF antibody compromised the axon-growth-promoting effects of LI, whereas injections of the same amount of BSA, an anti-parvalbumin antibody or even an anti-oncomodulin antiserum did not significantly alter these effects. These data indicate that CNTF is essentially involved in the mechanism by which RGCs switch to a regenerative state. An earlier study also used anti-CNTF antibodies to test the contribution of this cytokine in this context, but found no reduction of axon regeneration in the injured optic nerve after LI (Leon et al., 2000). However, in that study, the antibody was just applied by one single injection into the vitreous body, and axon regeneration was measured 3 weeks after ONC. Based on our finding that CNTF is continuously expressed and released from astrocytes, it is conceivable that one single injection of anti-CNTF antibody neutralizes only a small percentage of the CNTF mainly that is released during the first few days after LI. We have previously shown that LI performed a few days after ONC causes even stronger neuroprotection and axon regeneration than a LI that is performed at the same time of ONC (Fischer et al., 2000; Yin et al., 2003), demonstrating that the success of axon regeneration is not critically dependent on the presence of beneficial factors within the first days after ONC. Thus, a single injection of anti-CNTF antibodies and an evaluation of regeneration after long time periods are likely to be insufficient to show a measurable reduction in LI effects. In the present study, we maintained high intravitreal antibody concentrations by repeated applications and measured the effects after 5 days to circumvent these types of problems. This might explain the different results in the studies.

Leon et al. (2000) also reported that intravitreally applied CNTF at the time of ONC was ineffective in stimulating RGCs to regenerate axons into the mature optic nerve. This result might be due to the short half-life time of CNTF in the vitreous and to the fact that it was not continuously delivered. A sustained delivery of CNTF from astrocytes that maintains high local concentrations is likely to be more effective. As shown here, a repeated injection of CNTF 3 days after optic nerve injury is sufficient to yield regeneration into the optic nerve and is as potent as LI when the regenerative state was evaluated 5 days after surgery. Even greater regeneration can be achieved by a continuous release of CNTF from virally transfected RGCs. This treatment enabled axons to regenerate over long distances into the injured optic nerve, and some even reached the optic chiasm (Leaver et al., 2006). This gene therapeutic approach also strongly stimulated axon regeneration into a peripheral nerve graft and protected 25% of RGCs from apoptosis several weeks after injury. Thus, the effects of continuously delivered CNTF should be considered at least as potent as LI.

The involvement of CNTF in mediating the effects of LI or zymosan is further supported by the correlations between CNTF expression and the regenerative state of RGCs, the activation of its major downstream pathway (JAK/STAT3) in RGCs and the fact that dibutyryl-cAMP potentiates the effects of CNTF and likewise of LI. However, it should be noted that several other factors are also potentiated or become active by intracellular cAMP elevation (Meyer-Franke et al., 1998; Cui et al., 2003; Yin et al., 2006). A major role of CNTF is also in line with a previous report that additional injections of this cytokine did not increase LI effects, whereas additional BDNF treatment resulted in stronger neuroprotection (Pernet and Di Polo, 2006).

CNTF lacks a classical consensus secretory signal sequence (Stockli et al., 1989). The mechanism by which CNTF is released is still not clearly understood. Nevertheless, cultured astrocytes have been shown to release CNTF, particularly after inflammatory stimulation (Kamiguchi et al., 1995). In addition corneal endothelium releases CNTF under conditions of stress (Koh, 2002). Consistent with these reports, we detected significant amounts of CNTF in the vitreous body after LI treatment. As the lens does not express CNTF (Fischer et al., 2000) and because intraocular macrophages or other cells in the vitreous body do not either as shown by RT-PCR and immunostaining the detected CNTF was likely derived from activated astrocytes. This means that the local CNTF concentration on the RGCs’ membrane is likely to be much higher than in the vitreous body owing to the close anatomical association of retinal astrocytes with RGCs.

Oncomodulin mediates its axon-growth-promoting effects in culture in a JAK-independent manner (Yin et al., 2006), whereas CNTF generally mediates its beneficial effects mainly through the JAK/STAT3 pathway (Heinrich et al., 2003; Teng and Tang, 2006). In agreement with an upregulation of retinal CNTF, STAT3 was activated as soon as 1 day after LI in RGCs, reaching a maximum after 2–3 days. Although pSTAT3 levels slightly decreased afterwards, likely due to the expression of the negative feedback regulator SOCS-3, the transcription factor remained activated over the total observation time of 14 days. However, activation of JAK in RGCs is essential for the full effect of LI-stimulated regeneration on RGCs, because its inhibition by intravitreal injections of AG490 compromised these beneficial effects. An activation of STAT3 after LI is also in line with results of our previous microarray study (Fischer et al., 2004b). Among the 16 000 genes that showed the strongest upregulation in RGCs 4 days after LI + ONC, were SOCS-3 and the transcription factor cebp-δ. The expression of both genes is controlled by STAT3 (Starr et al., 1997; Yamada et al., 1997, 1998; Teng and Tang, 2006). Based on our finding that CNTF is a key mediator of LI and zymosan effects and based on the report that next to the inhibition of the JAK/STAT3-pathway also an inhibition of the PI3K/Akt-, and the MAPK/ERK-pathways abrogates the neuroprotective and axon-growth-promoting effects of CNTF on adult RGCs (Park et al., 2004) it is likely that these pathways are also essentially involved in mediating the beneficial effects of intraocular inflammation. The contribution of these pathways in this context is currently under investigation.

JAK/STAT3 pathway activation is also essential for axon regeneration of DRG neurons. Peripheral but not central nerve injuries activated STAT3 in DRG neurons, and AG490 treatment compromised neurite outgrowth and reduced GAP-43 upregulation, demonstrating that the JAK/STAT3 pathway is essential for this process (Qiu et al., 2005). Thus, as in the eye zymosan may also induce CNTF or other factors such as IL-6 that subsequently activate the JAK/STAT3 pathway and thereby stimulate axon regeneration of DRG neurons in vivo (Steinmetz et al., 2005; Cao et al., 2006).

The results of our study do not exclude an involvement of macrophage-derived or other factors as additional contributors to the neuroprotective and axon-growth-promoting effects of LI or zymosan. However, we conclude that continuously released CNTF from retinal astrocytes is a main mediator of LI- and zymosan-induced effects in vivo. Understanding the mechanisms underlying the induction of CNTF expression and continuous release from neural bystander cells like astrocytes avoiding other detrimental processes that are associated with inflammation are therefore likely to be more effective than LI or zymosan injections. Furthermore, combinatorial treatment of these factors mediating the effects of LI with drugs that increase intracellular cAMP levels and measures to block inhibitory pathways that are activated by myelin or components of the gliotic scar are likely to promote greater regeneration than has been achieved to date, and might also become meaningful for the treatment of human patients in the future.

Supplementary material

Supplementary material is available at Brain online.

Acknowledgements

This work was supported by the state of Baden-Württemberg. We thank Anastasia Andreadaki for technical support, Jieun Lee for helping establishing methods and providing critical comments on the manuscript, Daniela van den Ecker for helpful comments, Dr Sarthy (Northwestern University, Chicago, USA) for providing the immortalized retinal Müller cell line rMC-1.

Abbreviations:

Abbreviations:
 
• BSA

  bovine serum albumin

 
• cAMP

  cyclic adenosine monophosphate

 
• CNTF

  ciliary neurotrophic factor

 
• GAP-43

  growth-associated protein 43

 
• JAK

  Janus-kinase 2

 
• LI

  lens injury

 
• ONC

  optic nerve crush

 
• RGC

  retinal ganglion cell

 
• STAT3

  signal transducers and activators of transcription 3