GFAP Promoter Drives Müller Cell–Specific Expression in Transgenic Mice

Abstract

purpose. In an attempt to identify Müller cell–specific promoters and to better understand the gene regulatory mechanisms in retinal glial cells, the expression of the glial fibrillary acidic protein (GFAP) gene was studied in Müller cell cultures and in GFAP-enhanced green fluorescent protein (EGFP) transgenic mice.

methods. A transfection assay of GFAP-luciferase constructs carrying a series of nested deletions was performed in an established Müller cell line. For in vivo analysis, transgenic mice were generated by injecting a construct carrying a 2.5-kb, 5′ fragment of the mouse GFAP gene linked to the EGFP gene. Isolated retinas from transgenic mice were screened for GFP expression. Subsequently, the identity of the GFP-expressing cells was established by immunostaining cryostat sections of the retina with antibodies against Müller cell antigenic markers. Induction of the transgene and the endogenous GFAP gene was examined by injecting ciliary neurotrophic factor (CNTF) into the eye.

results. The DNA transfection data suggested that proximal 5′ sequences of the GFAP gene are sufficient to direct high-level reporter expression in Müller cell cultures. In transgenic mice, GFP fluorescence appeared in radially oriented processes that spanned almost the entire thickness of the retina. Immunostaining with antibodies to cellular retinaldehyde-binding protein (CRALBP) and glutamine synthetase showed that the GFP-expressing cells were Müller cells. GFP-expressing Müller cells were observed in the retinas of both albino and pigmented transgenic mice. In eyes injected with CNTF, both GFP and GFAP levels were highly elevated. These observations suggest that the 2.5-kb, 5′ GFAP sequence can direct inducible reporter gene expression in Müller cells. In addition to Müller cells, a few GFP-labeled astrocytes were present in the adult retina. In the developing retina, GFP-expressing astrocytes were first present at the optic nerve head, and as development progressed, the cells gradually moved toward the periphery of the retina and acquired their adult, stellate morphology.

conclusions. The present study shows that the 2.5-kb, 5′ flanking region of the mouse GFAP gene can be used to express GFP, and possibly other genes, specifically in Müller cells in the mouse retina. Furthermore, expression of the transgene can be upregulated by intravitreal injection of CNTF.

Gene regulation studies are necessary for identifying cell type–specific promoters, which are crucial for designing vectors for targeted delivery in gene therapy. In the retina, gene expression studies so far have primarily focused on photoreceptor-specific genes, ¹ and genes selectively expressed in other retinal cell types have not been well studied. To facilitate the development of Müller cell–specific promoters and to better understand the gene regulatory mechanisms in retinal glial cells, we have studied regulation of the glial fibrillary acidic protein (GFAP) gene in Müller cells. ²  

GFAP is a 54-kDa, type III intermediate filament protein that is the major constituent of glial filaments in astrocytes. ³ The GFAP gene has been isolated and characterized in mouse, rat, and human. ^{4 5 6 7 8} The regulatory sequences of the mouse GFAP gene have been extensively analyzed in astrocyte cultures. The studies show that the necessary sequences for astrocyte-specific expression are located within −256 bp from the transcription start site of the mouse GFAP gene, ⁹ and that the GFAP proximal promoter is composed of both a ubiquitous positive element (GFII), and two tissue-specific negative elements (GFI and GFIII). Another study has reported a strong positive element between nucleotides −1631 and −1479 and a negative element in the first intron of the mouse GFAP gene. ¹⁰ In the rat GFAP gene, two cell-specific negative regulatory elements have been found within the gene, ¹¹ and the human GFAP gene contains an additional transcription initiator site located downstream from the transcription start site, between nucleotides +10 and +40. ^{12 13 14} Taken together, the findings in these studies indicate that GFAP gene transcription is highly complex and that it is regulated by a combination of positive regulatory sequences and tissue-specific negative elements. 

GFAP regulation has also been examined in transgenic mice carrying mouse and human GFAP sequences fused to the lacZ gene. ^{15 16 17} In one study, the transgene carried the entire mouse GFAP gene along with the 5′ and 3′ flanking regions, whereas in another investigation, upstream sequences between −1919 and +92 were used. ^{15 16} The human GFAP transgene tested, carried a 2.2-kb, 5′ sequence. ¹⁷ In all cases, transgene (lacZ) expression was detected in astrocytes, and induction of gliosis led to increased transgene activity in astrocytes. However, not all astrocytes expressed the transgene, indicating heterogeneity in expression. 

In addition to astrocytes, other glial cells including nonmyelinating Schwann cells, Bergmann glia, radial glia, and Müller cells are known to express GFAP. ³ GFAP gene regulation is not as well studied in these cases. Müller cells in mammalian retina normally contain only a low level of GFAP or none at all. GFAP level, however, is strongly upregulated in Müller cells in response to photoreceptor degeneration, retinal detachment, ischemia, or trauma. ^{2 18 19 20 21} GFAP induction in Müller cells appears to be mainly regulated at the transcription level. ²² The cis elements and positively and negatively acting trans factors, and the complex regulatory networks that control Müller cell–specific GFAP expression, however, remain to be identified. 

Several investigators have examined reporter gene expression by Müller cells in GFAP transgenic mice, but the results are not in agreement. In transgenic mice carrying the ∼ 2-kb 5′ sequence of the human GFAP gene, the reporter lacZ was expressed in retinal astrocytes, but was not detected in Müller cells. ¹⁷ This finding is surprising because the transgenic mice were generated in the FVB/N strain that is known to harbor the rd mutation, which results in gliosis and GFAP expression in Müller cells. ²² Experiments with mouse GFAP-lacZ transgenic mice generated in the B6 strain, suggest that the 5′ and 3′ immediate, flanking regions and the intronic and exon sequences of the mouse GFAP gene are sufficient to direct β-gal expression in astrocytes, but not in Müller cells. ^{23 24}  

In another study, transgenic animals were generated in the FVB/N strain, by injecting the human GFAP, 5′ DNA fragment linked to the green fluorescent protein (GFP). In these animals, GFP was present in Müller cells. ²⁵ When the animals were crossed with C57Bl/6J mice (rd ⁺), however, retinas from the heterozygotes, expressed GFP in retinal astrocytes, but not in Müller cells. In a more recent study with human GFAP-EFGP transgenic mice generated in the FVB/N background, ²⁶ strong GFP fluorescence was observed in Müller cells, but the retinal astrocytes were reported to express no GFP. In conclusion, although several studies have examined reporter expression in retinal cells in GFAP-transgenic mice, the pattern of transgene expression in Müller cells appears to be variable. Some of the observed differences may be due to anomalous behavior of the human GFAP sequences in mice, the absence of appropriate regulatory elements in the transgene constructs, or the genetic background of the mice. In the present study, we examined regulation of GFP expression in CD1 and C57B1/6J transgenic mice carrying a 2.5-kb 5′ mouse GFAP-GFP gene. 

Methods

DNA Transfection Assays

All DNA manipulations were performed according to protocols described by Ausubel et al. ²⁷ A 2.5-kb GFAP fragment was cloned in the correct orientation into the HindIII site of a luciferase (luc) expression vector (pGL2 Basic; Promega, Madison, WI). A series of nested deletions were created from the 5′ end using a kit (Erase-a-Base; Promega). The vector was first linearized by cutting with unique restriction enzymes within the multiple cloning site to produce a 5′ overhang just upstream of the 2.5-kb GFAP fragment. The nested deletions were created by treating the vector with exonuclease III (exo III) followed by S1 nuclease (Promega). ²⁸ DNA (5 μg) in exo III buffer was treated with 300 U of exo III at 37°C. Incubations were performed for 1- to 5-minute durations to generate different size fragments. After completion of exo III digestion, S1 nuclease buffer and S1 nuclease (12 U/ μg DNA) were added, and the mixture was incubated at 23°C for 30 minutes. S1 nuclease stop buffer was added and samples were heated to 70°C for 10 minutes to inactivate S1 nuclease. The deleted GFAP linear vectors were recircularized with T4 DNA ligase and transformed into competent bacteria. The sizes of the resultant GFAP fragments were determined by DNA sequencing, and the vectors were used for transfection. 

Transfection assays were performed in the Müller cell line E6/7, which is known to express GFAP endogenously. ²⁹ In addition, C6 cells (astrocytic glioma; GFAP positive) and NIH 3T3 (mouse fibroblast, GFAP negative) were used as control cultures. All cell lines were routinely maintained in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum. 

The GFAP-luc constructs and the control plasmid pSVlac were transfected into cells (LipofectAmine kit; Promega), and cells were assayed for luciferase and β-galactosidase activities with a commercial assay kit (Tropix Inc., Medford, MA). Luciferase activities were normalized to β-gal levels, and all values were expressed as the average of three to six separate determinations ± SD. 

Transgenic Mouse Generation and Screening

The transgene construct used for microinjection was generated by inserting a 2.5-kb 5′ flanking fragment or a 0.3-kb 5′ fragment of the mouse GFAP gene into an expression vector containing the EGFP gene (Fig. 1) . The expression vector was obtained from Paul Overbeek (Baylor College of Medicine, Houston, TX). Transgenic mice were generated at the Northwestern University Transgenic Facility. Microinjection and all manipulations were performed as described in the literature. ^{30 31} Outbred CD1 (ICR) albino mice were used as the source of fertilized eggs, which were reimplanted into C57B1/6J X BALB/c surrogate mothers. GFAP-GFP transgenic founders were identified by PCR. 

For screening, tail samples or ear punches were digested with proteinase K at 55°C in a water bath for 1 to 2 hours. Samples were boiled for 5 minutes and centrifuged at 10,000 rpm for 5 minutes. For PCR, we designed a set of primers in which the sense primer was complementary to the 3′-region of the GFAP promoter (5′-AGG AAG TCA GGG GCA GAT TT-3′) and the anti-sense primer was complementary to a region of the EGFP sequence (5′-GGG TCT TGT AGT TGC GTC GT-3′). PCR was performed by using a programmable thermal controller (model PTC-100; MJ Research, Inc., Watertown, MA) with the following program: 5 minutes at 95°C, 1 cycle; 1 minute at 94°C, 1 minute at 57°C and 1 minute at 72°C, 40 cycles; and 5 minutes at 72°C, 1 cycle. DNA products (∼500 bp) were subjected to agarose gel electrophoresis (1%) and visualized under UV light. 

Transgenic lines were maintained as heterozygotes by backcrossing to nontransgenic CD-1 mice. Animals were maintained in a barrier facility ensuring isolation from pathogens. All animal procedures conformed to the provisions of the ARVO Statement for the Use of Animals in Ophthalmic and Animal Research. A total of 298 mice were used in these studies. 

Immunocytochemistry

After enucleation, eyecups were briefly fixed in 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4). One eye of each animal was used for preparing retinal wholemounts. The retina was gently isolated in one piece, flattened on a microscope slide, and examined under a microscope equipped with epifluorescence (Carl Zeiss Meditech, Thornwood, NY). The other retina was rinsed and incubated for several hours in a series of cryoprotective solutions containing increasing amounts of sucrose and glycerol. Finally, retinas were placed in optimal cutting temperature (OCT) compound and quickly frozen in liquid nitrogen. Frozen blocks were stored at −80°C for up to 4 weeks before sectioning. 

Frozen retinas were cut on a cryostat, and 20-μm-thick sections were collected on poly-l-lysine–coated slides. Sections were blocked for 1 hour in phosphate-buffered saline (PBS) containing 5% goat serum, 1% BSA, and 0.5% Triton X-100. For immunocytochemistry, we used rabbit anti-GFAP (Dako, Glostrup, Denmark), monoclonal anti-glutamine synthetase (BD Transduction Laboratories, Palo Alto, CA), and rabbit anti-CRALBP (a gift from John C. Saari, University of Washington, Seattle, WA). Rhodamine (TRITC)-conjugated goat anti-rabbit and goat anti-mouse antibodies (Jackson ImmunoResearch, West Grove, PA) were used as secondary antibodies. Antibody incubations were conducted in PBS containing 3% goat serum, 1% BSA, and 0.5% Triton X-100. Primary antibodies were incubated overnight at 4°C. Sections were washed three times for 10 minutes each in PBS and incubated with secondary antibodies for 1 hour at room temperature. Subsequently, sections were washed three times for 5 minutes each in PBS and mounted in TBS Shur/mount (Triangle Biomedical Sciences, Inc., Durham, NC). Digital images of immunostained retinal sections were acquired on a confocal microscope (LSM 510; Carl Zeiss Meditech, Thornwood, NY) with a krypton/argon laser by using 488-, 543-, and 633-nm lines and 40× oil-immersion objective. Images were later processed by graphics software (Photoshop; Adobe Systems, Mountain View, CA). Colocalization was evaluated by combining the two images in two different color channels. GFP data are represented in green; the immunostaining data in red and yellow represent the areas of colocalization. 

CNTF Treatment

Transgenic mice, 2 to 3 months old, were anesthetized with ketamine, and intravitreally injected with 0.5 μL of ciliary neurotrophic factor (CNTF) at 1 μg/μL, as described elsewhere. ³² The fellow eye was injected with 0.5 μL PBS and served as the control. After 2 days, retinas were dissected and processed for immunocytochemistry, as described previously. Retinas from transgenic animals injected with PBS or CNTF were also used for immunoblot analysis. 

Immunoblot Analysis

Retinas were homogenized in lysis buffer containing: 50 mM Tris HCl, 150 mM NaCl, 1% Triton X-100, 0.1% SDS, 0.5% sodium deoxycholate, 1.5 nM aprotinin, 0.01 μM E-64 protease inhibitor, 5 μM EDTA, 0.01 μM leupeptin, and 5 μM AEBSF-HCl (Calbiochem, San Diego, CA). Protein levels were quantified with a protein assay kit (Bio-Rad, Hercules, CA). Protein samples (40 μg each) were separated on a polyacrylamide gel (NuPAGE; Invitrogen, Carlsbad, CA) and transferred to a polyvinylidene difluoride (PVDF) membrane (Tropix) overnight at 4°C. After transfer, the membrane was washed with TBS for 5 minutes at room temperature and incubated in blocking buffer (5% nonfat dry milk in TBS) for 1 hour. After washing (three times for 5 minutes each) with TBS, the blot was incubated in primary antibody at 1:200 dilution. We used GFAP rabbit polyclonal antibody (Dako) and GFP mouse monoclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA). Proteins were visualized with the avidin biotin (ABC) staining system (Santa Cruz Biotechnology). 

Results

GFAP Regulation in Müller Cells In Vitro

To identify regulatory sequences required for Müller cell–specific GFAP expression, a series of nested deletion mutants were transfected into Müller cell cultures, and the cultures were assayed for luciferase expression. Results of the experiment are presented in Figure 2 . The data showed that proximal 5′ sequences (∼0.3 kb) of the GFAP gene were sufficient to direct high level reporter expression in Müller cell cultures. Moreover, distal sequences (>2.5 kb) appeared to contain negative elements that could suppress GFAP expression. Other intervening sequences showed either a positive or a negative effect. Finally, GFAP 5′ regulatory sequences behaved very similarly in Müller cell and astrocyte cultures. There was little luciferase expression in fibroblast cultures that were used as a negative control (data not shown). 

GFP Expression in Transgenic Mice

Transgenic mice were generated by microinjecting DNA constructs carrying either a 2.5-kb 5′-flanking fragment or a 0.3-kb 5′ fragment of the mouse GFAP gene into an expression vector containing the EGFP gene. We used the two different constructs, because the 0.3-kb fragment provides the minimal sequence necessary for GFAP-driven expression, whereas the 2.5-kb sequence carries positive and negative regulatory sequences. ⁹  

In the initial screening, three lines from 2.5-kb GFAP-GFP transgenic mice were found to express GFP in retinal cells. Among these, line 3747 had the strongest GFP fluorescence and was used in all the experiments. When retinas were examined in wholemounts, GFP-expressing cell bodies were widely distributed across the retina (Fig. 3A) . A retinal cross section examined with the confocal microscope showed radially oriented GFP⁺ processes stretching from the outer limiting membrane to the inner limiting membrane (Fig. 3B , OLM and ILM, respectively). Cell bodies of these processes were located in the middle of the INL. The GFP⁺ cells had the morphology of mammalian retinal Müller cells (Fig. 3C) . A few GFP-labeled cell bodies were also found in the ganglion cell layer (GCL) where the retinal astrocytes are located (Fig. 3B) . 

We performed immunocytochemical studies to establish the identity of the radially oriented, GFP-labeled cells, by using known Müller cell–specific antigenic markers. Results of the experiments are presented in Figures 3D 3E 3F 3G 3H 3I . In retinal sections stained with an antibody to cellular retinaldehyde-binding protein (CRALBP), a well-established, cell-specific marker for Müller cells in the adult retina, colocalization of GFP and CRALBP was observed in the radial processes and cell bodies. All GFP⁺ cells were CRALBP⁺. Many CRALBP⁺ cells, however, were GFP⁻ (Figs. 3D 3E 3F) . Immunostaining with an antibody to glutamine synthetase, another Müller cell marker, showed that all the GFP⁺ cells were also glutamine synthetase positive (Figs. 3G 3H 3I) . 

When retinas from transgenic mice were stained for GFAP, we frequently observed GFAP immunostaining in Müller cells (Figs. 3J 3K 3L) . This was a surprising finding because the animals had been maintained under ambient light conditions, and histologically the retina looked normal. To resolve the question of whether GFP expression in Müller cells is dependent on endogenous GFAP levels, we crossbred the CD1 transgenic mice with C57B1/6J mice. Results of GFAP-immunostaining studies with retinas from pigmented transgenic mice are shown in Figures 3M 3N 3O . In these retinas, strong GFP expression was present in many Müller cells (Fig. 3M) ; however, little GFAP was detected in the Müller cells. Instead, GFAP labeling was heavily distributed among astrocytic processes in the GCL and the ILM. These observations establish that the mouse GFAP promoter used in the present study is capable of driving reporter gene expression in Müller cells independent of endogenous GFAP gene activity. 

Modulation of Transgene Expression by CNTF

It is well known that retinal injury, detachment, ischemia, light damage, or photoreceptor degeneration leads to GFAP induction in Müller cells. ² An obvious question is whether GFP expression is also upregulated in Müller cells under these conditions, in the GFAP-GFP transgenic mice. Recently, we showed that intravitreal injection of CNTF leads to rapid induction of GFAP in Müller cells. ³² We have used this paradigm to examine whether expression of the GFP transgene is regulated the same way as that of the endogenous GFAP gene. 

CNTF was intravitreally injected into one eye of GFAP-GFP transgenic mice, and PBS was injected into the other eye. ³² Two days later, retinas were isolated and processed for GFAP immunocytochemistry. Results of the experiment are presented in Figures 4A 4B 4C . In the eyes injected with PBS, most of the GFAP was present at the ILM, and in a few radial processes in the IPL (Fig. 4B) . In the eye injected with CNTF, GFP fluorescence was strikingly enhanced throughout the retina (Fig. 4D) . There was also a large increase in the intensity and the number of GFAP-expressing radial processes in the retinal section (Fig. 4E) . Furthermore, the number of Müller cell processes coexpressing GFP and GFAP in the CNTF-injected eye (Fig. 4F) , far exceeded similar cells in the PBS-injected eye (Fig. 4C) . 

To further substantiate the immunocytochemical data, we examined GFP and GFAP levels by immunoblotting. Proteins were extracted from normal eyes (WT), GFAP-GFP transgenic eyes injected with PBS or transgenic eyes injected with CNTF. After separation by polyacrylamide gel electrophoresis and blotting, GFP and GFAP were detected by treating the blots with a monoclonal anti-GFP (Santa Cruz Biotechnology) or a polyclonal GFAP antibody (Dako). Results of the experiment are presented in Figures 4G (GFAP) and 4H (GFP). In blots treated with anti-GFP, there was no detectable GFP in the normal eye. In the transgenic eyes injected with PBS, however, there was a distinct GFP band. Furthermore, the level of GFP was highly elevated in CNTF-injected transgenic eyes. In the case of GFAP, there was a faint band in the normal (WT) eye, whereas transgenic eyes injected with PBS showed a strong GFAP band. In transgenic eyes injected with CNTF, there was a large increase in the GFAP content. Taken together, the immunocytochemical and immunoblot data clearly show that CNTF-treatment leads to induction of both the endogenous GFAP and the transgene GFP in the retina. 

In addition to 2.5-kb GFAP-GFP transgenic mice, we also generated transgenic mice carrying a 0.3-kb, 5′ GFAP-GFP construct. Unfortunately, among more than 20 transgenic lines screened, only 1 line exhibited strong GFP expression. Furthermore, histologic analysis showed that, in this line, GFP expression was confined to subpopulations of rod and cone bipolar cells (Kuzmanovic M, et al. IOVS 2002;43;ARVO E-Abstract 739). There was no GFP in Müller cells or astrocytes. We have not studied this transgenic line further. 

Developmental Expression of GFP

GFAP expression has been well studied in the developing mouse retina. ^{33 34} These studies have shown that there are few astrocytes in the retina before birth and that most astrocytes, at this developmental stage, reside in the optic nerve and the optic disc area. Soon after birth, however, there is rapid migration of astrocytes from the optic disc into the retina, and by P10, astrocytes can be found in the far peripheral retina. To examine whether the GFAP-EGFP transgene exhibits the same expression pattern as the endogenous GFAP gene, we examined GFP expression during retinal development. In these experiments, wholemounts and transverse sections of retina from postnatal day (P)2, P7, and P14 pups were examined for GFP. The results are presented in Figure 5 . At P2, there were a large number of GFP-expressing cells in the optic nerve head but not in the neural retina (Figs. 5A 5D 5G) . By P7, there were many GFP⁺ cells in the ILM (Figs. 5G 5E 5H) and by P14, GFP-expressing cells were visible from the central retina all the way to the far periphery (Figs. 5C 5F 5I) . In addition, by P14, many radially-oriented cells, which are probably Müller cells, began to express GFP (Fig. 5I) . Concurrently, there was a decrease in the number of GFP-expressing cells in the ILM, and after 2 months, GFP expression was restricted mostly to Müller cells. The GFAP-immunostaining pattern in the transgenic retinas was similar to that reported for nontransgenic mice earlier. ^{33 34}  

Discussion

The present study shows that the 2.5-kb 5′ flanking region of the mouse GFAP gene is capable of directing GFP gene expression in Müller cells. In the transgenic mice, the level of GFP expression was high enough to readily visualize Müller cells in retinal wholemounts and in retinal tissue sections. The experimental data also clearly show that transgene expression is not dependent on coexpression of the endogenous GFAP gene or the genetic background. Furthermore, expression of the GFP transgene is regulated the same way as that of the endogenous GFAP gene. 

It is important to note that not all Müller cells appeared to express GFP. Many CRALBP-, glutamine synthetase–, or GFAP-labeled cells were devoid of GFP fluorescence. Moreover, GFP-labeled Müller cells often appeared in patches rather than in uniform distribution across the retina. The reason behind this heterogeneity in GFP expression is not known, but a similar situation has been noted in reporter expression among astrocytes in GFAP transgenic mice. ^{15 16 17 25 26} It is possible that additional GFAP regulatory sequences may be needed for full GFP expression. Alternatively, the site of integration of the transgene or genetic background of the transgenic mice may be important factors. These questions remain to be investigated in the future. Nevertheless, the present study suggests that it should be possible to express exogenous proteins in Müller cells using the 2.5-kb 5′ mouse DNA sequence. The ability to express specific proteins in Müller cells should provide a useful tool to investigate questions related to Müller cell biology. 

The cell transfection data indicated that promoter proximal sequences were able to drive high-level reporter expression in a Müller cell line and that 5′ sequences longer than 2.5 kb had an inhibitory effect. The results from GFAP-GFP transgenic mice, however, showed that the 2.5-kb GFAP sequence is able to drive reporter expression in Müller cells and astrocytes. Moreover, as mentioned earlier, in a 0.3-kb GFAP-GFP transgenic line, we saw GFP expression only in bipolar cells, not in Müller cells or astrocytes. These observations suggest that expression of the mouse GFAP gene in cell culture differs substantially from that in the organism. A discrepancy in gene expression in cell cultures and in transgenic mice has also been reported in other studies. ^{35 36 37} Although the underlying cause of this difference is not known, absence of endogenous transcription factors in the Müller cell line may be a contributing factor. 

Our experimental findings are in general accord with results of other studies on GFAP transgenic mice showing that 5′ sequences of both human and mouse GFAP gene are able to direct reporter expression in Müller cells and astrocytes. Some of the observed discrepancies, however, might be due to anomalous behavior of the human GFAP gene in transgenic mice, the presence of negative regulatory elements in the mouse GFAP gene, or that some of the transgenic mice were generated in the FVB/N strain, which is known to harbor the rd mutation resulting in GFAP synthesis and gliosis in Müller cells. ²² Nolte et al. ²⁶ did not find GFP-expressing astrocytes in the adult retinas of GFAP-GFP transgenic mice. Our transgenic mice, however, contained a small number of GFP⁺ astrocytes in the retina. This difference could be due to the age at which the animals were examined, because GFP expression appears to be downregulated with age. Indeed, few GFP⁺ astrocytes have been reported in mice older than 6 months. A difference in the site of integration of the transgene may also be a contributing factor. 

The developmental expression of GFP closely parallels that of GFAP in retinal astrocytes. ^{33 34} GFP is found only in the optic nerve at birth, and in postnatal stages, GFP labeled-astrocytes appear progressively from the central to the peripheral retina. A surprising finding is that the number of GFP-expressing cells appeared to decline drastically at postnatal stages. The loss of GFP expression in mature retinal astrocytes, which has also been noted in other studies, ²⁶ suggests that the GFAP-GFP transgene is regulated differently than the endogenous GFAP gene in astrocytes. It is possible that additional GFAP regulatory sequences are needed to maintain GFP expression in the mature astrocytes. GFP expression in Müller cells first occurs around the second postnatal week, which suggests that transgene expression is initiated at different times in distinct glial cell types. 

It is clear from our experimental data that the expression from the GFAP-GFP transgene can occur independent of the transcriptional status of the endogenous GFAP gene. Because the GFAP gene is not normally active in Müller cells, the data suggest that GFAP gene expression is normally controlled by negative elements that reside in distal 5′ (>2.5 kb) intragenic or 3′ sequences. Furthermore, the results of the CNTF injection experiments show that transgene expression is regulated the same way as that of the endogenous GFAP gene. 

 Supported by National Eye Institute Grant R01 EY03523 and unrestricted funds from Research to Prevent Blindness, Inc.

 Submitted for publication December 11, 2002; revised March 17, 2003; accepted April 1, 2003.

 Disclosure: M. Kuzmanovic, None; V.J. Dudley, None; V.P. Sarthy, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Structure of the GFAP-GFP construct used for generating transgenic mice. A 2.5-kb 5′ fragment of the mouse GFAP gene (−2560 to +61) was cloned into a EGFP expression vector. The GFAP-GFP construct was excised by digestion with KpnI and SpeI and used for generating transgenic mice.

Figure 2.

 

Transfection assay of GFAP-luc constructs. GFAP-luc constructs were prepared by cloning nested deletions into the luciferase expression vector PGL2. Transfections were performed with a Müller cell line and a glioma line (C6), both of which express GFAP. Luciferase activities were normalized to β-gal levels. All values are expressed as the average of three to six separate determinations ± SD. The size of the DNA fragment is indicated on the abscissa.

Figure 3.

 

Morphology and immunocytochemical characterization of GFP-expressing cells in the transgenic retina. GFP-labeled cells were present in wholemounts (A) and transverse sections of 2-month-old retinas. The radially oriented GFP⁺ cells had the morphologic features of Müller cells (B, C). The other micrographs (D–O) show transverse sections of transgenic retina immunostained for Müller cell antigenic markers. The sections show GFP fluorescence (D, G, J, M, green), TRITC-labeling for the Müller cell antigen (red; E, CRALBP; H, glutamine synthetase; K and N, GFAP), and double labeling (F, I, L, O, yellow). Nonspecific labeling of the IPL with the GFAP antibody (L) resulted in some spurious double labeling.

Figure 4.

 

Induction of GFP and GFAP expression by CNTF. GFAP-GFP transgenic mice were intravitreally injected with PBS (A–C) or with CNTF (D–F), as described previously. ³² Retinas were processed for immunocytochemistry (A–F) or immunoblot analysis (G, H). The micrographs show GFP fluorescence (A, D, green), TRITC-labeling for GFAP (B, E, red), and double labeling (C, F, yellow) in retinal sections. Nonspecific labeling with the GFAP antibody was detected in the OPL (B, C, E, F). The immunoblots show GFAP (G) and GFP (H) levels in retinas from normal mice (WT), transgenic mice injected with PBS (TG), and transgenic mice injected with CNTF (TG+CNTF). (G, arrow) Band of approximately 55 kDa corresponds to the size of the GFAP monomer. (H, arrow) Major GFP band of approximately 67 kDa, corresponding to the GFP dimer that has been reported in immunoblots. ³⁸ The smaller bands are probably degradation products or monomers.

Figure 5.

 

GFP expressing cells in the developing mouse retina. The micrographs show wholemounts (A–F) or transverse sections of developing eyes from postnatal days (P)2, P7, and P14, examined with a confocal microscope. GFP-labeled cells emanated from the optic nerve in P2 retina (A, D). At P7, stellate cells were apparent in the retina (B, E), and by P14, GFP⁺ cells were present across the entire retina. In addition, there was a drastic reduction in the number of GFP-labeled cells in the retina. In a transverse section of P2 eye, GFP was strongly expressed in the optic nerve and the optic disc region (G). At P7, GFP-labeled cells lined the ILM region (H). At P14, radially oriented, GFP-expressing cells were present in the retina (I). ON, optic nerve; R, retina; ILM, inner limiting membrane.

The authors thank Gregory McGillem for help with confocal microscopy.