Leukemia Inhibitory Factor Blocks Expression of Crx and Nrl Transcription Factors to Inhibit Photoreceptor Differentiation

Abstract

purpose. Activating ligands of gp130, including leukemia inhibitory factor (LIF), can block differentiation and function of retinal neurons. This study focused on determining whether LIF inhibits differentiation of photoreceptors by altering cell fate or by blocking the expression of essential transcription factors in vivo.

methods. Transgenic mice were generated that had lens-specific expression of the secreted human LIF protein. Retinal differentiation was assessed by histology and by gene expression analysis, with in situ hybridization, immunohistochemistry, and real-time qRT-PCR. Electroretinograms were used to assess retinal function.

results. LIF did not prevent or alter the timing of outer and inner nuclear layer separation, but it inhibited phototransduction gene expression in both rods and cones, thereby blocking functional maturation of photoreceptors. LIF also reduced the expression of Crx, Nrl, and Nr2e3, and upregulated the expression of transcription inhibitors Baf and Fiz1.

conclusions. LIF expression did not appear to alter photoreceptor cell fate specification, but it inhibited subsequent differentiation. These results suggest that gp130 ligands can inhibit photoreceptor functional differentiation by reducing Crx- and Nrl-dependent transcription.

Significant progress has been made toward understanding the transcriptional programs that regulate commitment and differentiation of retinal cells. Accumulating evidence suggests that retinal neural differentiation is a sequential process. ¹ In the initial phase, dividing progenitors become postmitotic and initiate expression of lineage-specific transcription factors. In a subsequent step, cells express additional transcription factors that drive the expression of genes required for maturation and function. The identification of mechanisms that control lineage commitment and functional differentiation is a particularly active area of research in the neuroretina. ^{2 3}  

In vertebrates, the retina contains seven neuronal cell types and one glial cell type organized in a well-ordered laminar structure. Approximately 70% of the cells in the retina are photoreceptors. These cells express proteins involved in phototransduction, including the visual pigment opsin. ^{4 5 6} There are two types of photoreceptors, rods, and cones. Rods and cones differ from each other in structure and function due to the expression of distinct proteins. Rod photoreceptors express rod opsin and are specialized for monochrome vision in dim light. Cones are specialized for use in relatively bright light. In mice there are two cone opsin genes, a medium-wavelength opsin (M-opsin) and a short-wavelength opsin (S-opsin) that absorb light in the green and far blue ends of the spectrum, respectively. ^{7 8} Most cones express both M- and S-opsin, but there are some cones that express only one or the other. ^{9 10}  

In development, cells in the photoreceptor lineage turn on expression of the homeobox transcription factor Otx2, which is essential but not sufficient for commitment to the photoreceptor lineage. ¹¹ Photoreceptor commitment also involves the expression of proneural basic helix-loop-helix (bHLH) transcription factors, including Mash1, NeuroD, and Math3. ¹² Together, these early transcription factors are responsible for inducing and committing neurons to the photoreceptor lineage. One of the target genes for Otx2 is the cone rod homeobox gene (Crx), which is induced in committed cells. ^{11 13} Mice lacking Crx fail to express proteins involved in photoreceptor function, but are still committed to the photoreceptor lineage. ¹⁴ Therefore, Crx is a marker for cell commitment and is essential for the expression of genes associated with maturation of photoreceptors. 

Crx may be able to regulate its own expression as its promoter contains potential Crx binding sites. ¹⁵ The upregulation of Crx may be a necessary step for the expression of both rod and cone photoreceptor genes. ¹⁶ Another transcription factor that is necessary for rod differentiation is the neural retina leucine zipper protein Nrl. ¹⁷ The expression of Nrl is necessary for the expression of an orphan-nuclear receptor, Nr2e3. ¹⁷ Mature rods express NeuroD, Crx, Nrl, and Nr2e3. These transcription factors cooperate to drive rod phototransduction gene expression as a multiprotein transcriptional complex. ¹⁸ It is not yet known what is necessary in addition to Crx for differentiation of S-cones, but, in the differentiation of M-cones, the expression of Crx and the thyroid hormone receptor beta 2 (Trβ2) is essential. ¹⁹  

Two related cytokines, ciliary neurotrophic factor (CNTF) and leukemia inhibitory factor (LIF), are members of the interleukin (IL)-6 family of cytokines. Members of this family do not share sequence homology, but are grouped together based on activation of a common signal-transducing receptor, gp130. ²⁰ To induce signal transduction, LIF binds to and dimerizes the LIF receptor β (LIFRβ) with the coreceptor gp130, whereas CNTF binds to and trimerizes CNTF receptor alpha (CNTFRα), LIFRβ, and gp130. ^{20 21} Because both ligands signal through gp130, there is significant overlap in the biological activity of CNTF and LIF. Activated gp130 signals through the JAK/STAT pathway, the mitogen activated protein kinase pathway (Erk1/2), and the phosphatidylinositol 3 kinase (PI3K) pathway. ^{22 23 24 25} The receptor components LIFRβ, gp130, and CNTFRα have all been localized to retinal cells, including ganglion cells, Müller glial cells, and photoreceptors. ^{26 27 28 29 30} In addition, the activation of the MAPK and STAT3 pathways has been shown to occur in differentiating progenitors, including those in the photoreceptor lineage, as well as in mature ganglion cells and mature Müller glial cells. ^{29 31 32}  

Several studies have shown that LIF and CNTF inhibit the function of photoreceptors by suppressing the expression of photoreceptor genes, including opsin. ^{33 34 35 36} In primary retinal cultures from rats, CNTF and LIF have been reported to decrease the number of cells expressing photoreceptor markers and to increase the number of cells expressing bipolar cell markers. ³⁵ Thus, activation of gp130 in development may switch the cell’s fate from photoreceptors to bipolar neurons. However, this effect was not observed in mouse primary cultures or in retinal explant cultures from rats. ^{35 36} The ability of CNTF or LIF to influence photoreceptor differentiation has been shown to be dependent on the stage of photoreceptor differentiation. ^{35 36} It is possible that photoreceptor or bipolar cell commitment had progressed too far in some cases to allow a switch in cell fates. 

The purpose of this study was to identify in vivo the point in differentiation that is susceptible to inhibition by ligands of gp130. To accomplish this, we generated transgenic mice that express human LIF in the ocular lens. Our transgenic mice initiate LIF expression from embryonic day (E)11 and continue to express LIF throughout development. ^{37 38} This approach exposes neuroblasts in the retina to LIF before decisions about the cell’s fate have been made. This early and sustained delivery of LIF has allowed us to study how gp130 stimulation affects cell fate and differentiation in the retina. Our data show that LIF does not alter cell fate in mice, but it can inhibit functional differentiation of photoreceptors. It does so by downregulating the expression of key transcription factors for both rods and cones. 

Materials and Methods

Generation and Screening of Transgenic Mice

All procedures were approved by the IACUC committees for Baylor College of Medicine, the University of Oklahoma Health Sciences Center, and the Dean A. McGee Eye Institute, and were conducted in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. We generated transgenic mice by microinjection of an αA-crystallin-LIF minigene (Fig. 1A) . A 600 bp (base pair) cDNA encoding the 180 amino-acid–secreted human LIF protein was cloned into the BamHI and HindIII sites downstream of a minimal 360 bp αA-crystallin promoter. ³⁹ SV40 sequences were included to provide an intron and polyadenylation signals. The minigene was released from the plasmid vector by NotI digestion and purified by agarose gel electrophoresis using a quick-spin gel extraction kit (Qiagen, Valencia, CA). The DNA fragment was eluted in 10 mM Tris-HCl (pH 7.4), 0.1 mM EDTA, and diluted to 2 ng/μL before microinjection into pronuclei of FVB/N embryos. To identify transgenic mice, PCR amplifications were performed on 1 μL of proteinase K digested tail DNA samples using a sense primer located in the αA-crystallin promoter (5′-GCATTCCAGCTGCTGACGGT-3′) and an anti-sense primer (5′-GTGACATGGGTGGCGTATGGC-3′) located in the human LIF cDNA. PCR reactions (30 μL) consisted of 10× NH₄ buffer (Bioline, London, UK), 1.5 mM MgCl₂, 0.05 mM each dNTP, 0.4 pM of each primer, and 0.1 units of Taq DNA polymerase (Bioline). Reactions were denatured at 95°C for 6 minutes and then subjected to 35 cycles of 95°C for 30 seconds, 60°C for 30 seconds, and 72°C for 1 minute. A 30-μL aliquot from each reaction was analyzed by gel electrophoresis in a 1% agarose gel for the presence of a 250-bp band. Founder mice were mated to FVB/N mice to establish transgenic lines. The FVB mouse strain is homozygous for the retinal degeneration mutation (rd1), which is caused by a mutation in the gene encoding rod-specific cGMP phosphodiesterase (Pde6b^rd1). ^{40 41 42} This autosomal recessive defect results in rapid rod cell death followed by slow loss of cones. To avoid this complication, LIF transgenic mice were mated to C57BL/6 mice. The F1 progeny of this mating are heterozygous for the rd1 mutation (rd1 ^+/−) and do not have photoreceptor degeneration. As there are subtle differences in the electroretinograms between rd1 ^+/− and rd1^+/+ mice, in this study we used mice that were all rd1 ^+/−. ⁴³ Therefore, the changes in retinal phenotype and function described in this study are due to the expression of LIF in the lens and not to a complication of the rd1 mutation. 

Tissue Preparation and Routine Histology

Eyes were enucleated, fixed in 10% neutral-buffered formalin for 24 hours, dehydrated, and embedded in paraffin. All histologic procedures were performed on 5-μm paraffin-embedded sections that had been deparaffinized in xylene and rehydrated in an ethanol series. Sections were stained with hematoxylin and eosin for routine histology. 

In Situ Hybridization

In situ hybridization was performed with ³⁵S-labeled riboprobes as described previously. ³⁷ A riboprobe specific for transgene expression was generated by in vitro transcription of a plasmid containing the SV40 region (Fig. 1A) . To detect Crx expression, a plasmid containing a 1.6-kb mouse Crx cDNA (obtained from Constance Cepko, Harvard Medical School, Boston, MA), was linearized with HindIII and transcribed with T3 RNA polymerase. To assay for Chx10 expression, a 1.4-kb cDNA fragment from the 3′ end of the human Chx10 cDNA (obtained from Rod McInnes, Hospital for Sick Children, Toronto, Ontario, Canada) was first subcloned (Bluescript; Strategene, La Jolla, CA), then linearized with BamHI, and transcribed with T3 RNA polymerase. The human cDNA sequences of Chx10 is 97% identical with the mouse sequence and is highly cross-reactive to the mouse Chx10 mRNA. ⁴⁴ After autoradiography, darkfield images were captured with a 40× objective and Nomarski optics on a microscope (Eclipse; Nikon, Tokyo, Japan) and a digital camera (Roper Scientific, Duluth, GA). Image-analysis software (Metamorph; Universal Imaging Corp., Downingtown, PA) was used to analyze the images. 

ELISA Assay

The concentration of LIF in retinas was measured by ELISA. Three eyes from 1-, 2-, and 4-week-old mice were dissected from each line and from nontransgenic mice and were processed separately for triplicate analysis. For dissections, eyes were enucleated and placed in ice-cold 1× PBS. Under a dissecting microscope, corneas and lenses were removed, and the retinas were dissected from the eyecup. Retinas were inspected visually under the microscope to verify that they were free of contaminating lens material. Retinas were placed in 1.5-μL tubes and homogenized in 100 μL of PBS. The concentration of LIF was measured by ELISA, by using an immunoassay that is specific for human LIF (Quantikine DLF00; R&D Systems, Minneapolis, MN). Protein concentrations were determined with the bicinchoninic acid (BCA) protein assay (Pierce, Rockford, IL), with bovine serum albumin used for the standard curve. 

Immunostaining

To detect opsin expression, we placed paraffin-embedded tissue sections on glass slides, rehydrated them in an ethanol series, and blocked them in 10% horse serum in PBS for 30 minutes at room temperature before incubation with primary antibody. A mouse anti-rhodopsin monoclonal antibody, RetP1 ⁴⁵ (obtained from Colin Barnstable, Yale University), was diluted 1:200 in 10% horse serum/PBS. Sections were incubated overnight at 4°C, washed three times in PBS, and incubated with a biotin-conjugated anti-mouse antibody diluted 1:200 in 10% horse serum and PBS. Peroxidase staining was performed with an avidin-biotin complex (ABC) kit (Vector Laboratories, Burlingame, CA) and with 3,3′-diaminobenzidine substrate (SK-4100; Vector Laboratories). Slides were coverslipped and imaged (Eclipse 800; Nikon). 

Analysis of mRNA Expression by Real-Time RT-PCR

Total RNA was extracted from nontransgenic and transgenic retinal tissues (Tri reagent; Sigma-Aldrich, St. Louis, MO) according to the manufacturer’s recommendations. To remove trace DNA contamination, samples were treated with DNase 1 (RNAqueous-4PCR Kit; Ambion, Austin, TX). Five micrograms of RNA was then used as a template for first-strand cDNA synthesis (Script cDNA synthesis kit; Bio-Rad, Hercules, CA). cDNA was diluted 1:10 and used in real-time experiments. Real-time PCR was performed with a supermix (iQ Green Supermix; Bio-Rad). Briefly, the PCR mixture contained 12.5 μL of the supermix, 1 μL of RT product, and 125 nM of the primers, in a total volume of 25 μL. PCR amplification was performed according to the temperature profile: 95°C for 3 minutes to denature the DNA templates, followed by 40 cycles of 95°C for 30 seconds, and annealing-extension at 60°C for 30 seconds. The amount of the PCR product was monitored by the fluorescence intensity of SYBR green on a thermal cycler (MyIQ; Bio-Rad). All reactions were performed in duplicate. For each primer pair, serial dilutions of nontransgenic P14 cDNA were used to assess linearity and efficiency. Controls for each reaction included non–reverse-transcribed cDNA. The β-actin gene was included in each assay as a cDNA loading control. Primer pairs were designed by computer (Oligo Analysis software; Integrated DNA Technologies, Inc., Coraville, IA). Details of the primers and amplification products are given in Table 1 . The designed primers shared 100% homology with the target sequence, but no significant homology with other sequences, by BLAST search (National Center for Biotechnology Information, Bethesda, MD). For each gene, the real-time PCR experiments were performed with three independent batches of cDNA for each genotype. The change (x-fold) in expression was calculated by the 2^−ΔΔCT method. ^{46 47} ANOVA statistical analysis was performed on computer (Excel; Microsoft, Cupertino, WA). 

Electroretinogram Analysis

To test the function of retinal neurons, ERGs were recorded in animals anesthetized with tribromoethanol (Avertin, Winthrop Laboratories, New York, NY) and maintained on an isothermal heating pad (37°C). A silver silver-chloride cotton-wick electrode was placed on the edge of the cornea of the right eye, and a reference electrode was placed on the tongue. The recorded signals were amplified and displayed on a storage oscilloscope. The stimuli were 50-ms flashes of 500-nm light (maximum intensity 1100 μW/cm²) which were attenuated by neutral-density filters and delivered to the eye through a 2-mm fiber optic bundle. Beginning with subthreshold stimuli, flashes increased in intensity until there was no increase in ERG amplitude (saturated response). The interval between flashes of moderate to high intensity was 4 to 5 minutes. At least two animals from each transgenic line (736 and 774) were analyzed for each age (P19 and P48). 

Morphologic Evaluation by Quantitative Histology

Retinal thickness was measured as described previously. ^{48 49} At postnatal day 14, eyes (n = 5) were removed from both wild-type and transgenic mice and fixed (20% isopropanol, 2% trichloroacetic acid, 4% paraformaldehyde, and 2% zinc chloride; Perfix; Fischer Scientific, Pittsburgh, PA), and embedded in paraffin. Sections 5 μm thick were cut and stained with hematoxylin-eosin. The sections were taken along the vertical meridian, to allow comparison of the superior and inferior hemispheres. Thicknesses of the total retinal (the distance from the inner limiting membrane to the outermost extent of the layer, including the photoreceptor segments), outer nuclear layer (ONL; the distance from the innermost to the outermost extent of the layer, excluding the photoreceptor segments), and inner nuclear layer (INL; the distance between the outer plexiform layer [OPL] and inner plexiform layer [IPL]) were measured at 14 points, starting at the optic nerve head and extending toward the superior and inferior ora serrata. Measurement points were made at 5-μm intervals. The mean of the 14 measurements collected was used as the estimate of ONL, INL, and total retinal thicknesses in each eye. 

Results

Dose-Dependent Effects of Lens-Specific LIF Expression

We generated five αA-crystallin-LIF transgenic lines (OVE736, -773, -774, -775, and -777). Eyes from each line (Fig. 1C 1D 1E 1F 1G)as well as from nontransgenic (Fig. 1B)animals were analyzed by histology at 3 weeks of age. Whereas all retinas had the normal three-layer lamination, the transgenic lines had a range of defects ranging from mild disorganization (Fig. 1C , 774) to an almost complete absence of photoreceptor outer and inner segment (Fig. 1G , 777). One severely affected line (Fig. 1F , 736), and one mildly affected line (Fig. 1C , 774) were chosen for more detailed analyses. 

To verify that transgene expression was restricted to the lens, we analyzed the 736 and 774 lines by in situ hybridization. In both lines, the transgene was found to be expressed in lens fiber cells, with no expression detected in the retina (Fig. 2) . The signal was more intense in the 736 line than in the 774 line, suggesting a higher level of expression. To determine whether LIF protein was able to accumulate in the retina, we cleanly dissected retinas from the lenses under a dissecting microscope. Retinal proteins were extracted, and LIF protein was measured by ELISA (R&D Systems). The amount of LIF protein present at all ages was 6 to 20- times higher in the 736 line than in the mildly affected 774 line (Table 2) . The ELISA was specific for human LIF and therefore we did not detect LIF in nontransgenic retinas (Table 2 , ND). Although the LIF transgene is expressed only in the lens, our data show that the protein is secreted and can accumulate in the retina. Higher expression was present in the more severely affected 736 line. This suggests that the effects on the retina are dose dependent. 

Effect of LIF on Photoreceptors

The first structural evidence for the specification of photoreceptors was the formation of the ONL. This occurred at the end of the first postnatal week (P5–P7) in both normal retinas and αA-LIF transgenic retinas (Fig. 3) . Photoreceptors also began opsin expression and the initiation of outer segment (OS) synthesis at P5 to P7 (Fig. 3A) . Rod photoreceptor differentiation was complete by P14, with the establishment of inner and OS with abundant rhodopsin (Fig. 3B) . In the high-expressing 736 line, opsin expression and OS synthesis was not detected at any age, indicating impaired photoreceptor differentiation (Figs. 3G 3H 3I) . In the low-expressing 774 line, initiation of opsin expression and OS synthesis was delayed until approximately P14 (Fig. 3E) . 

LIF Regulation of Rod and Cone Opsins

To determine whether rod and cone opsins were downregulated at the RNA level, we measured their expression by quantitative (q)RT-PCR (Fig. 4) . In the high-expressing 736 line, rhodopsin was expressed at 0.1%, S-opsin at 13%, and M-opsin at 2% of nontransgenic levels. In addition to opsins, we found that expression of transducin α, arrestin, and cGMP phosphodiesterase were reduced (data not shown). In the low-expressing 774 line, rhodopsin and M-opsin were expressed at a higher level than in the 736 line, but their expression level was significantly reduced (13% and 20%) versus normal. The S-opsin expression level was variable between different mice in the 774 line, and the apparent increase is not statistically significant from normal (P = 0.069). 

Absent or Delayed Photoreceptor Light Responsiveness

To test whether photoreceptors in transgenic mice are functional, we measured their response to light by ERG. We tested three mice for each line at P19, P28, and at 8 weeks of age. Figure 5shows representative ERGs for P19 and P28 mice. At each age, nontransgenic mice had normal ERG responses (Figs. 5A 5B) . None of the high-expressing 736 animals had a detectable ERG signal (Fig. 5C 5D) , suggesting a complete loss of photoreceptor function. None of the ERGs from 774 mice at P19 showed a light response (Fig. 5E) . However, the ERG signals in all older 774 animals contained a small a-wave and a prominent c-wave (Fig. 5F) , both of which indicate photoreceptor stimulation by light. The ERG signal in 774 animals did not have an intervening b-wave, suggesting that signal transmission through other retinal neurons was defective. 

Differentiation into Photoreceptors

Because retinas in the high-expressing 736 mice failed to express genes normally found in photoreceptors, it was possible that LIF directed cells away from a photoreceptor cell fate or kept them in the proliferating neuroblastic state. To investigate these possibilities, we analyzed the mice for expression of the transcription factors Chx10 and Crx by in situ hybridization. Chx10 is normally expressed in proliferating neuroblasts and differentiated bipolar neurons. ⁵⁰ The expression of Chx10 at P14 in both nontransgenic and 736 transgenic retinas was localized to cells at the outer half of the INL (Figs. 6C 6D , respectively). At P14, this is the normal location of differentiated bipolar cells. ⁵⁰ The absence of Chx10 in the ONL suggests that the cells located there are not proliferating neuroblast or bipolar cells. Crx is expressed only in retinal photoreceptors and pinealocytes and is essential for their differentiation. ^{13 14} In both nontransgenic and 736 transgenic retinas, Crx expression was found throughout the ONL (Figs. 6A 6B) . The presence of Crx transcripts in the ONL demonstrates that these cells are photoreceptors and not any other cell type. We observed Crx expression in a subset of cells in the INL, as reported previously. ⁵¹ The staining in the INL was similar in both nontransgenic and transgenic mice, demonstrating that the expression in the INL was not caused by LIF. 

Inhibition of Expression of Key Transcription Factors

We have shown that opsin expression is reduced at the RNA level, which may reflect decreased expression of the transcription factors that regulate maturation of photoreceptors. To explore this possibility, we used qRT-PCR to analyze the expression of a set of transcription factors that are known to be necessary for normal photoreceptor function (Fig. 7) . In the high-LIF–expressing line we observed no change in the expression of Otx2, and a modest reduction in the expression of Mash1 and NeuroD (Fig. 7A) . In addition, we found that both Crx and Trβ2 were expressed at 40% of the normal level (Fig. 7A) . Reduced expression of Crx and Trβ2 is the likely cause for reduced expression of S- and M-opsin. Rod opsin expression is regulated by the combined activity of Crx, Nrl, and Nr2e3. ^{18 52} In the high-LIF–expressing 736 mice, Nrl and Nr2e3 were down to 10% and 20% of their normal levels. The combined reduction in these transcription factors may lead to the reduction of opsin expression in rods. In the lower-expressing 774 line, Crx, Trβ2, Nr2e3, and Nrl were much less affected, which is consistent with the dose-dependent reduction in opsin expression (Fig. 7B) . 

In addition to regulating the RNA levels of Crx and Nrl, LIF may also regulate these factors at the protein level. Crx has been shown to interact with several proteins, including phosducin, phosducin-like proteins (PhLP1 and PhLOP1), ataxin-7, and barrier to autointegration factor (Baf), which can suppress the transcriptional activation activity of Crx. ^{53 54 55} Nrl has also been shown to interact with the zinc finger protein Fiz1, which can suppress the Nrl-mediated transcription. ⁵⁶ In the high-expressing line, Baf was expressed at 140% of normal, whereas Fiz1 was expressed at 200% (Fig. 7A) . The elevated expression of both Baf and Fiz1 would be likely to impact the ability of Crx and Nrl to drive transcription. In the low expressing line both genes were expressed at normal levels (Fig. 7B) . 

Morphologic Analysis of Retinal Thickness in LIF Transgenic Mice

To determine whether the reduced expression of Crx or Nrl was due to a reduced number of photoreceptors, we measured total retinal, INL, and ONL thicknesses in nontransgenic and transgenic mice at P14. The nuclear layer thickness is commonly used as a measure of the number of cells in the retina. ^{49 48} Our data show there was reduced thickness across the entire span of the retina from the optic nerve to the ora serrata in both the superior and inferior halves of the retina (Fig. 8A) . There was no difference in retinal thickness between the high- and low-expressing lines, demonstrating that the reduced cell number was not dose dependent. The average thickness data show that retinas from 774 and 736 mice had 66% of the normal retinal thickness (Fig. 8B) . Thinning was uniform throughout the retina since both the outer and INLs were reduced to the same extent (Figs. 8C 8D) . Therefore, the ratio of ONL to INL thickness was the same in both nontransgenic and transgenic retinas. The data indicate that while the number of cells in the retina was reduced, the proportion of cells that are in the ONL had not changed. The loss of photoreceptor cells was not dose dependent, and therefore the change in gene expression was not due simply to a change in the number of photoreceptors. 

Discussion

Our data show that LIF does not grossly alter determination of neuronal cell fate in the mouse retina, even if it is expressed early in development. Instead, the data show that LIF inhibits photoreceptor differentiation after lineage commitment. LIF appears to regulate differentiation by inhibiting the expression of transcription factors that are necessary for phototransduction gene expression in both rods and cones. Although the present study was focused on the mechanism by which photoreceptor differentiation is suppressed, we have also observed inhibition of differentiation of other retinal neurons (manuscript in preparation). We do not yet know the mechanism for this retinal thinning. We analyzed retinas from E18 mice as well as newborn mice and have found that, at all ages, transgenic retinas are thinner than nontransgenic retinas (not shown). TUNEL staining did not show a significant increase in cell death (data not shown). The mechanism for the decrease in retinal cells may be due to a subtle increase in cell death over a long period or a subtle decrease in the proliferation rates of progenitor cells. 

Determination of Cell Fate in the Retina

The αA-crystallin promoter in our transgenic mice initiates expression from E11, which is before cells in the retina commit to a specific cell lineage. Because of this, we can argue that LIF does not grossly alter cell fate in the mouse retina. If LIF caused retinal cells to remain in the neuroblastic state or if it caused excess cells to become bipolar cells, we would expect to see an increase in cells expressing Chx10 located in the ONL. ⁵⁰ Our data show that Chx10 expression is properly turned off in the photoreceptor layer. Targeted mutations in mice have shown that the combined activities of Otx2, Mash1, NeuroD, and Math3 are essential for determination of photoreceptor fate. ^{11 12} In their absence, Crx expression is not induced. In our LIF mice the expression levels of Otx2, Mash1, and NeuroD were either slightly reduced or normal (Fig. 7) . Crx expression is induced in cells located in the ONL that do not express Chx10. These data suggest that activation of gp130 does not block the early stages of photoreceptor or bipolar cell specification. In addition, the fact that LIF expression did not alter the ratio of ONL cells to INL cells indicates that there was not a massive switch in cell fates. 

Inhibition of Photoreceptor Differentiation

Our data show that LIF blocks the expression of phototransduction genes by mechanisms that probably involve reduced transcription. Crx is expressed in our transgenic retinas, but at 40% of the normal level. Similar results have been found in mouse explant cultures. ³¹ In both situations, gp130 activation does not block the induction of Crx but does inhibit its upregulation. It has been shown that Crx can transactivate its own promoter. ¹⁵ The absence of high-Crx expression in our mice is likely to be part of the mechanism for reduced phototransduction gene expression in rods and cones, OS morphogenesis, and synaptogenesis with bipolar neurons. Similar effects have been observed in mice without Crx. ⁵⁷ With high levels of LIF in the 736 line, we saw greatly reduced Nrl and Nr2e3 expression levels. Our results are strikingly different from mice without Nrl or Nr2e3, since we did not observe an increase in S-cones. ^{17 58} This is likely because LIF inhibits both Crx and Nrl expression, which would inhibit both rod and cone differentiation. Our data suggest that LIF traps both rods and cones in a differentiation state that occurs shortly after they are committed to the photoreceptor lineage, but before maturation takes place. In the low-expressing 774 line we observed normal to slightly increased S-opsin expression. An interesting observation was that there was less inhibition of Crx than Nrl in these mice. This finding suggests that the ratio of Crx to Nrl, not just the presence or absence of Nrl, is an important determinant in rod versus cone opsin expression, and is in agreement with recent in vitro studies showing that the ratio between Crx, Nrl, and Nr2e3 are important for expression from rod or cone opsin promoters. ⁵⁹  

The mechanisms by which LIF inhibits the expression of Crx, Nrl, and Nr2e3 are not yet known. After gp130 activation, STAT3 dimerizes and is imported into the nucleus where it binds DNA and regulates transcription. ⁶⁰ STAT3 signaling has been shown to be necessary for CNTF-mediated reduction of Crx and opsin expression. ³¹ Therefore, STAT3 is likely to play a role in the suppression of Crx and Nrl in the LIF transgenic mice. Based on what is known about STAT3, there are at least three mechanisms by which STAT3 could suppress Crx and Nrl. 

One of these mechanisms is that STAT3 could directly bind to the promoters of Crx and Nrl to suppress their expression. We identified two STAT3-binding sites in both Crx and Nrl proximal promoters using Web-based search tools (Patch Search, BioBase GmbH, Wolfenbüttel, Germany; www.gene-regulation.com). If STAT3 can bind these sites and act as an inhibitor in photoreceptors, it is possible that it could suppress their promoter activity. Although STAT3 leads to activation of transcription in some cells, in olfactory neurons activated STAT3 binds the promoters of differentiation-specific genes and suppresses their expression. ^{60 61}  

It is also possible that STAT3 activation acts indirectly on photoreceptor genes by competing for common transcriptional cofactors such as CBP/p300. Several photoreceptor transcription factors including NeuroD, Mash1, and Crx require CBP/p300 as a transcriptional cofactor. ^{62 63 64 65} CBP/p300 is also a cofactor for STAT3. ^{65 66} It is possible that active STAT3 sequesters a limiting pool of CBP/p300. A similar competition mechanism has been shown to exist for the bHLH transcription factor neurogenin in the developing brain. ⁶⁷ According to this mechanism, any transcriptional activity of NeuroD, Mash1, or Crx would be reduced because of the sequestration of CBP/p300 by active STAT3. 

The third possibility is that STAT3 could induce the expression of factors that suppress the expression or activity of Crx and Nrl. Yeast two-hybrid screens were used to identify Baf as a Crx-interacting protein that can suppress Crx-dependent transcription in cotransfection assays. ⁵⁵ Likewise, Fiz1 has been shown to interact with Nrl, and can suppress Nrl-dependent transcription. ⁵⁶ In LIF transgenic mice, both Baf and Fiz1 are expressed at higher levels than normal (Fig. 7) . Even though we observed a modest change in Baf expression, the effects on phototransduction genes may be highly significant. Because Baf and Fiz1 act as transcriptional repressors, the absolute value of either is less important than their expression relative to Crx and Nrl. The large reduction in Crx and Nrl expression coupled with the modest increase in Baf and Fiz1 would be likely to drive most of Crx and Nrl into inhibitory complexes, which would result in inhibition of Crx- and Nrl-dependent transcription. 

In this study, we showed that LIF inhibits differentiation by reducing or preventing the transcription of proteins that are necessary for phototransduction. Our data suggest that reduced transcription is caused by inhibiting the high-level expression and perhaps activity of both Crx and Nrl. Thus, it seems possible that the gp130 receptor may be an important regulator of photoreceptor maturation during development. Gp130 knockout mice die early in development and cannot be used to study the role of gp130 in retinal development. ⁶⁸ Therefore, testing the role of gp130 in normal photoreceptor development would necessitate tissue-specific inactivation of gp130. 

 Supported by National Institutes of Health Grant RR017703 (JDA); National Eye Institute Grants EY14206 (JDA), P30 EY012190 (JDA), EY10448 (PAO), EY10803 (PAO), and EY11348 (PAO); and a grant from Research to Prevent Blindness.

 Submitted for publication February 1, 2005; revised April 6, 2005; accepted April 9, 2005.

 Disclosure: D.R. Graham, None; P.A. Overbeek, None; J.D. Ash, 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.

 

Transgenic mice were generated with the αA-crystallin-LIF minigene (A), which consisted of a 360-bp αA-crystallin promoter (arrow) linked to 600-bp human LIF cDNA and SV40 sequences. The LIF cDNA encodes the 180-amino-acid LIF protein. The locations of the annealing sites for the primers SV40A and SV40B, which were used to identify transgenic mice, are shown. The retinal histologies of five transgenic lines (C–G) and a nontransgenic control (B) were analyzed at 21 days of age. Representative histologies are shown in order of severity, from least severe (774, C) to most severe (777, G). In the least severely affected line, photoreceptors had disorganized and shortened OS (os, C). The moderately affected lines, 775 and 773, showed significantly reduced photoreceptor OS (D, E, arrow). A severely affected line (736) had only remnants of photoreceptor IS (F, arrow) and significant disorganization of the INL and ganglion cell layer (GCL). In the most severe line (777, G), three distinct cell layers were observed. However, all layers were disorganized, and photoreceptors did not have visible inner or OS. Scale bar, 100 μm.

Table 1.

 

RT-PCR Parameters

Gene	Accession Number		Oligonucleotide Sequence (5′–3′)	Size of Product (bp)	PCR Efficiency (%)
NRL	L14935	F	GACCACACACACCTCTTCC	87	109.6
		R	CTCTCCTGTATAGCGCCATC
Nr2e3	NM_013708	F	GGTTGGGCCCAGCAACTTCT	90	102
		R	CGCCACAGACACAGGCATAG
Rhodopsin	M55171	F	GGAAGTCACCCGCATGG	87	98.3
		R	TGGGTGAAGATGTAGAAGGCCACA
CRX	NM_007770	F	GGACCCTCTGGACTACAAAG	93	100.9
		R	GAAAGAGTGATTCCGCTG
M-opsin	NM_008106	F	CTTCTTGGCTCCAGGTCCC	94	99
		R	GACCACAAGAATCATCCAGG
S-opsin	NM_007538	F	GTCTCTTCTAGCAAAGTTGGC	87	82.1
		R	GATGTAGATACTTAAATGTGGC
Trβ2	NM_009380	F	GTGAAGGCTGCAAGGGCTTCTTTA	111	95
		R	ACTGGTTGCGGGTGACTTTGTCTA
Otx2	NM_144841	F	CCAGGGTGCAGGTATGG	92	95
		R	GGCAGGCCTCACTTTGTT
Mash1	NM_008553	F	GACTTGAACTCTATGGCGGGTTCT	150	98
		R	TTCCAAAGTCCATTCCCAGGAGAG
NeuroD	U28068	F	GACGCTCTGCAAAGGTTTG	85	86.8
		R	CTCAGGCAAGAAAGTCCG
Baf	NM_011793	F	TGAGCAAGAGGCTGGAGG	90	100.3
		R	CTCGGAAGAGGTCTTCATCT
Fiz1	NM_011813	F	GGGCTTCAAGCATAGCTT	92	104.8
		R	GGAGTCTCGGAATCCCTTA
β-Actin	NM_007393	F	AGAGAGGTATCCTGACCCTGAAGT	105	90
		R	CACGCAGCTCATTGTAGAAGGTGT

Figure 2.

 

Transgene expression was analyzed by in situ hybridization. Bright-field images are shown for the severely affected 736 line (A) and the mildly affected 774 line (B). Eyes were assayed at birth (P0). Dark-field images show a more intense signal in the 736 line (C) than in the 774 line (D). In all the eyes, the iris and retinal pigmented epithelium did not express transgenic LIF, but appeared positive in the dark-field image due to pigment granules.

Table 2.

 

ELISA Transgene Expression

	1 Week	2 Week	4 Week
NTG	ND	ND	ND
736	91 ± 47	104 ± 54	113 ± 49
774	5.3 ± 2.3	16.2 ± 7.9	25 ± 7.5

 Data are expressed as picograms LIF/milligram total protein ± SD. ND, none detected.

Figure 3.

 

LIF expression blocks photoreceptor maturation and opsin expression. Paraffin-embedded sections were immunohistochemically stained with an anti-opsin antibody, RetP1. Brown staining indicates the presence of opsin protein. The melanin in the retinal pigmented epithelium is black. Sections were counterstained with hematoxylin to visualize cell nuclei. Retinas at P7 (A, D, G), P14 (B, E, H), and P28 (C, F, I) are shown. Retinas were from (A–C) nontransgenic mice, (D–F) 774 transgenic mice, and (G–I) 736 transgenic mice. No opsin protein was detected in the 736 retinas. Opsin expression was delayed in the 774 photoreceptors.

Figure 4.

 

Reduced expression of rod and cone opsins. Real-time PCR was performed on total retina mRNA from P14 nontransgenic and transgenic mice. Expression levels were normalized to that of β-actin. The percentages of expression in transgenic mice, relative to nontransgenic mice are graphed. Data are an average of results in samples from three independent RNA isolations, with each sample run in duplicate. Error bars, SE. *P < 0.05 by ANOVA.

Figure 5.

 

ERG analysis. Representative electroretinograms (ERG) are shown for nontransgenic (A, B), 736 transgenic (C, D), and 774 transgenic mice (E, F) at P19 (A, C, E) and at P28 (B, D, F). In nontransgenic animals, the initial negative deflection from baseline (bl), the a-wave, was produced by the photoreceptor polarization in response to light stimulation. Subsequent to the a-wave, the signal was passed by synaptic transmission from photoreceptors to the bipolar cells in the inner retina (the b-wave). After the light response, a slower positive potential in the RPE cells was detected as the c-wave. Line 736 transgenic animals did not have any detectable signals (C, D), indicating that the photoreceptors were not functional. Line 774 transgenic animals did not have an ERG signal at P19 (E), but had a prominent c-wave at P28 (F), suggesting that photoreceptors in these animals were responding to light.

Figure 6.

 

LIF does not prevent cells in the ONL from expressing photoreceptor genes. In situ hybridizations show the expression of Crx (A, B), and Chx10 (C, D) in P14 retinas from nontransgenic mice (A, C), and 736 transgenic mice (B, D). Dark-field images were obtained using Nomarski optics. The colored dark-field images were overlaid onto the bright-field images of the same hematoxylin and eosin–stained retinas. In nontransgenic P14 retinas, Crx RNA was detected in photoreceptors located in the ONL (A) and was similarly localized to the ONL in 736 retinas (B). Chx10 RNA was localized to cells at the outer edge of the INL in nontransgenic (C) and 736 transgenic retinas (D).

Figure 7.

 

Real-time qRT-PCR of essential photoreceptor transcription factors in normal and transgenic retinas. qRT-PCR was performed on total retina mRNA from P14 nontransgenic and transgenic mice in duplicate, with three different sets of cDNA from independent RNA isolations. Expression levels were normalized to that of β-actin. The levels of expression of each gene in transgenic mice relative to nontransgenic mice are shown. With high LIF expression (A), CRX and Trβ2 expressions were downregulated to 43% of normal. Nr2E3 and Nrl were down to 20% and 9%, respectively. The Crx interacting protein Baf was upregulated nearly 1.4-fold, whereas the Nrl interacting protein Fiz1 increased nearly twofold. With lower levels of LIF expression (B), all four genes were less affected. Error bars, SEM. *P < 0.05 by ANOVA.

Figure 8.

 

Quantitative analysis of retinal thickness along the vertical meridian of the eye in P14 mice, demonstrates a reduction of total retinal thickness along the entire span of the neural retina in both the superior and inferior quadrants (A). The average thickness of the whole retina (B) showed a 40% reduction in both 736 (n = 5) and 774 (n = 5) LIF transgenic mice in comparison to nontransgenic mice (n = 5) (B). Both lines had a corresponding 40% reduction in the thicknesses of the ONL and INL (C, D). Each bar indicates a mean of seven measurements taken for ONL, INL, and total retinal thicknesses, beginning from the optic nerve head. Data are the mean ± SEM.

The authors thank Larry Rapp and Fang Li for advice and assistance with the ERG and morphologic analysis of retinal thickness.