Abstract
purpose. The authors investigated the effect of brain-derived neurotrophic factor (BDNF) administration on the expression of Ca2+-binding proteins in the developing bdnf −/− mouse retina.
methods. Intraocular injections of BDNF (0.5 μg) were applied on postnatal day (P) 11 bdnf −/− mice, and their effects were evaluated on P14. Neurons expressing Ca2+-binding protein were studied by immunohistochemistry for PKC-α, recoverin, calbindin-D28K, calretinin, and parvalbumin.
results. Cell density and immunostaining intensity for Ca2+-binding proteins in horizontal, bipolar, amacrine, and ganglion cells were lower in the retinas of bdnf −/− mice than of wild-type mice. Mutant retinas treated with BDNF showed a 35% to 40% increase in the number of calbindin-positive horizontal and amacrine cells. Increases of 30% and 50%, respectively, were also observed for calretinin- and parvalbumin-positive cells in the inner nuclear layer after BDNF treatment. The retinas of bdnf −/− mice showed recoverin expression only in scattered bipolar cells; however, recoverin-positive bipolar cells were readily detectable after BDNF injection in mutants (80% increase). The number of parvalbumin-positive ganglion cells after BDNF treatment reached 100% of control values. Expression of calretinin and calbindin was also upregulated in the ganglion cell layers of BDNF-treated mutants.
conclusions. The expression of Ca2+-binding proteins is reduced in the mutant retina. This neurochemical phenotype can be reverted, at least partially, by providing exogenous BDNF during the second week of postnatal development.
Brain-derived neurotrophic factor (BDNF) and other neurotrophins have been implicated in the phenotypic specification and morphologic development of central neurons. Analysis of
bdnf −/− mice revealed striking abnormalities in forebrain GABAergic interneurons
1 and a marked reduction of neuropeptide Y and parvalbumin expression in the cortex. Moreover, if levels of BDNF are elevated, neuropeptide Y production is increased and the developmental upregulation of parvalbumin expression is accelerated.
2 3
BDNF was also shown to influence neuronal phenotype expression in the retina. Antisense oligonucleotides directed against the BDNF receptor TrkB decrease parvalbumin expression in AII amacrine cells.
4 In addition, parvalbumin expression is reduced in the retinas of
trkB −/− mice,
5 and treatment with BDNF rescues parvalbumin expression in retinal cultures (Pinzon-Duarte G, et al.
IOVS 2001;42:ARVO Abstract S372).
6 7 At least three other classes of amacrine cells—those using dopamine, nitric oxide, and vasoactive intestinal peptide as neurotransmitters—are influenced by BDNF.
8 9 10 In these cells, the levels of neurotransmitter or its biosynthetic enzyme are reduced in
bdnf −/− mice and upregulated after injection of BDNF in wild-type (
wt) animals.
BDNF can also modulate the morphologic development of retinal neurons such as dopaminergic amacrine cells and retinal ganglion cells (RGCs). Furthermore, the functional development of RGCs and the transmission of photoreceptor signals to bipolar cells are impaired in the retinas of
bdnf −/− and
trkB −/− mice.
5 11
The goal of this work was to determine the role of BDNF during the development of specific neurons in the mouse inner retina. Because Ca
2+-binding proteins are a useful tool for identifying distinct cell subtypes in the retina,
12 we systematically studied their expression and the expression of other markers that allow selective labeling of cell populations in the retinas of
bdnf −/− mice.
Moreover, to distinguish between the effects of cell death or general developmental retardation and a specific requirement for endogenous BDNF to complete phenotype expression, “rescue” experiments were performed in which exogenous BDNF was injected intraocularly in bdnf −/− mice during postnatal development.
bdnf −/− mice were anesthetized and received intraocular injections on postnatal day (P) 11. Human recombinant BDNF (0.5 μg; Regeneron Pharmaceuticals, Inc., Tarrytown, NY) was dissolved in 0.5 μL of 0.1% bovine serum albumin (BSA) in phosphate-buffered saline (PBS) and was injected with a fine-glass microelectrode through the sclera into the vitreous body at the level of the temporal peripheral retina (for the protocol, see Cellerino et al.
8 and Cellerino et al.
9 ). Vehicle solution-injected eyes of mutant and
wt mice were taken as controls. Experiments were performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and conformed to German law.
The line of
bdnf −/− mice used in the present study has been described elsewhere.
13 Homozygous mutants were obtained by crossing heterozygous mice, and the animals were genotyped by performing PCR.
P0 to P14 mice were used. After decapitation, the eyes were enucleated, and the anterior parts and lenses were removed; posterior eyecups with the retinas in place were fixed.
To prepare plastic sections, retinas were dehydrated in ascending ethanol concentrations, incubated in acetone, and embedded (Technovit 8100; Heraeus Kulzer GmbH, Wehrheim, Germany). Radial sections (4 μm) were cut and stained with aqueous 1% methylene blue/1% azure solution.
To prepare cryostat sections, the fixed retinas were cryoprotected, embedded in cryomatrix (Tissue Tek; Leica, Bensheim, Germany), and sectioned. Eyes were taken in the morning. Radial 10-μm sections were stored at −20°C.
Radial retinal sections were washed three times in PBS and incubated for 1 hour in a solution of PBS with 20% normal goat serum (NGS), 1% BSA, and 0.1% Triton X-100. Antibodies against the following Ca2+-binding proteins were used: calbindin-D-28K, clone CL-300 (mouse mAb, 1:200; Swant, Bellinzona, Switzerland), calretinin (rabbit pAb, 1:300; Chemicon, Temecula, CA), parvalbumin (rabbit pAb 1:5000; Swant), protein gene product 9.5 (PGP 9.5; rabbit pAb, 1:1000; Chemicon), protein kinase C (PKC), clone MC5 (mouse mAb, 1:100; Amersham, Buckinghamshire, England), recoverin (rabbit pAb, 1:500; gift from K.-W. Koch, Jülich, Germany), and TrkB (rabbit pAb, 1:100; Transduction Laboratories, Lexington, KY) at 4°C. Standard streptavidin-biotin method (DAKO Diagnostica, Glostrup, Denmark) was used, and fluorescent immunostaining was detected with Cy3-labeled goat anti-mouse antibody (1:500) (Rockland, Gilbertsville, PA). Apoptosis was determined by terminal transferase dUTP nick-end labeling (TUNEL; Roche Diagnostics GmbH, Mannheim, Germany).
All data shown here represent the mean and standard error of mean results from at least four separate experiments. For each retina, three nonconsecutive sections were selected; in each slice, two fields located in the central part of the retinal section were analyzed. The rectangular fields had a basal line of 0.34053-mm length that was positioned such that it was parallel to the ganglion cell layer (GCL). For better comparability, cell numbers were then normalized to 1-mm length. All measurements were made on coded preparations by one researcher unaware of their experimental history.
Differences between controls, mutants, and BDNF-injected mutants were tested statistically with the two-tailed, unpaired t-test and a Welch correction (Prism 4 for Windows; Microsoft, Redmond, WA).
Analysis of
bdnf −/− retinal cross-sections during the first 2 weeks of postnatal development showed no reduction in thickness for any layer
(Fig. 1) . It can be concluded that the survival of retinal neurons does not depend on endogenous BDNF. However, the disappearance of numerically sparse populations, such as horizontal cells
14 and recoverin-positive bipolar cells, cannot be detected by measuring the thickness of the inner nuclear layer (INL). It is, therefore, necessary to demonstrate that these cells are still physically present in the mutant retinas.
Horizontal Cells.
Retinal sections were labeled with antibodies against calbindin. Calbindin labels all horizontal cells in rats and mice.
12 15 Horizontal cells showed weaker staining, and their processes appeared less elaborated in the
bdnf −/− retina
(Fig. 2) . Quantitative analysis revealed a reduction of approximately 50% in the number of horizontal cells
(Fig. 3) .
Protein gene product 9.5 (PGP 9.5) is an immunohistochemical marker that primarily labels RGCs and horizontal cells.
16 A reduction of horizontal anti-PGP 9.5–stained cells was found
(Fig. 2) . In addition, the plexus of horizontal cell processes appeared less dense, regardless of whether it was labeled using antibodies against PGP 9.5 or calbindin.
Calbindin-immunoreactive horizontal cells in the mouse express low levels of parvalbumin.
15 Immunoreactivity to parvalbumin was not visibly reduced in horizontal cells of
bdnf −/− mice, despite an apparent downregulation of parvalbumin expression in amacrine and ganglion cells
(Fig. 4) . However, the plexus of parvalbumin-positive processes in the outer plexiform layer (OPL) appeared less dense.
To investigate whether BDNF might act on horizontal cells directly, we performed double-labeling experiments using calbindin and TrkB and found a clear localization of TrkB on the horizontal cell processes
(Fig. 2) .
Amacrine Cells.
Subpopulations of amacrine cells in the mouse retina have been found to express calbindin, calretinin, or parvalbumin.
12 Mouse retinas were labeled with parvalbumin antibodies and displayed a staining pattern similar to that reported by Haverkamp and Wässle
12 (Fig. 4)but strangely dissimilar to that obtained by Rohrer et al.
5 In the INL of the
wt retina, a variety of moderately labeled amacrine cells was detected. Parvalbumin labeling was strongly reduced in the
bdnf −/− retinas, and hardly any labeled cells could be detected in the INL
(Fig. 4) . Calbindin expression by amacrine cells in the INL was also reduced in the retinas of
bdnf −/− mice, with counts revealing a reduction to only 35% of control values
(Fig. 3) .
Calretinin antibodies label a heterogeneous population of amacrine cells with different levels of intensity in the
wt retina.
12 In addition, three clearly demarcated bands are visible in the INL, with the innermost and outermost layers corresponding to the processes of cholinergic amacrine cells.
12 17 The three bands of calretinin immunoreactivity were not visibly reduced in thickness or intensity in the
bdnf −/− retina, in contrast to the strong reduction in the number of moderately labeled amacrine cells. As a result, the strongly labeled cells in the INL and the three bands of labeling in the inner plexiform layer (IPL) corresponding to their processes were almost the only structures immunostained for calretinin in the retinas from the mutant mice
(Fig. 4) . These cells correspond, at least in part, to the cholinergic amacrine cells,
17 and our data fit well with the hypothesis that the development of the cholinergic network is not influenced by BDNF.
8 However, cholinergic neurons express calbindin,
12 and the immunoreactivity to calbindin of the two cholinergic bands in the INL is reduced in
bdnf −/− mice
(Fig. 2) . Much like calbindin and parvalbumin in the horizontal cells, it seemed that different Ca
2+-binding proteins were also differentially regulated by BDNF in cholinergic amacrine cells.
Retinal Ganglion Cells.
The RGCs of mice express calretinin, parvalbumin, and calbindin.
12 Moderate levels of calretinin are expressed in the soma of many (possibly all) RGCs and large amounts in their axons in the nerve fiber layer.
12 Displaced amacrine cells, primarily cholinergic, also express calretinin.
17
The number of calretinin-expressing cells in the GCL is reduced in
bdnf −/− mice. The number of labeled cells in these retinas was only 60% of the values in
wt mice
(Fig. 3) , and their staining intensity was visibly reduced
(Fig. 4) . This effect was very likely the result of reduced expression of calretinin in the RGCs, because, as discussed, the expression of calretinin in cholinergic amacrine cells is apparently not affected.
Parvalbumin in the mouse retina is highly expressed in a subpopulation of large-bodied RGCs
12 whose morphology is reminiscent of the well-characterized α retinal RGC of mammals.
18 Their dendrites could be clearly seen to ramify into the INL
(Fig. 4) . These cells were reduced in the
bdnf −/− retina, falling to only 40% of control values
(Fig. 3) . In addition, cells moderately labeled for parvalbumin are also present in the GCL.
12 Changes in the density of these cells were not quantified. Finally, immunoreactivity for calbindin in RGCs was apparently also reduced
(Fig. 2) .
Bipolar Cells.
In mammalian retinas, two types of bipolar cells could be distinguished: rod and cone bipolar cells. Rod bipolar cells represent a homogeneous population and can be labeled with antibodies against PKCα.
19 Rod bipolar cells appeared largely normal in
bdnf −/− mice, but careful examination revealed atrophy of their axonal endings
(Fig. 5) . Although rigorous quantification of these effects is difficult to provide, it was possible for a naive observer to reliably distinguish between slides from mutant and control retinas in a blinded experiment. The same phenotype is probably present in
trkB −/− mice as well because it can be recognized by careful examination of
Figure 5in Rohrer et al.
5
Cone bipolar cells constitute a heterogeneous set of cells composed of approximately 10 different classes.
20 In the mouse retina, three types of cone bipolar cells can be labeled by recoverin antibody.
12 Expression of recoverin in the photoreceptors of mutant mice appeared normal, but, in contrast to the mild neurochemical phenotype expressed by the rod bipolar cells, hardly any recoverin-positive bipolar cells could be detected in the mutant retinas
(Figs. 3 6) . This cone bipolar phenotype replicates what has already been described in
trkB −/− mice.
5
Exogenous BDNF Rescue of the Neurochemical Phenotype of Bipolar, Horizontal, Amacrine, and Ganglion Cells
The downregulation of Ca
2+-binding proteins in the retinas of
bdnf −/− mice could simply be secondary to a general developmental retardation rather than a direct consequence of BDNF absence.
bdnf −/− mice are, in fact, much smaller than their littermates and experience a plethora of peripheral syndromes.
21
To distinguish between these possibilities, retinas of
wt and mutant animals at P10 were analyzed. Major morphologic changes were not seen in the retinas at this early age, but a reduction of parvalbumin and calbindin expression was already evident
(Fig. 7) . Next, we investigated whether the neurochemical changes described in the previous paragraph could be reverted. Exogenous BDNF was injected in P11
bdnf −/− mice when developmental neuronal death was already completed in the retina,
22 and retinas were collected at P14. Recoverin expression in
bdnf −/− retinas was observed only in scattered bipolar cells. In contrast, recoverin-positive bipolar cells were easily detectable in the BDNF-treated retinas (80% increase;
Figs. 3 6 ).
After BDNF injection,
bdnf −/− mice also showed an upregulation of calbindin immunoreactivity, not only in horizontal cells but also in amacrine cells
(Fig. 2) . The number of labeled horizontal cells increased by 35%, reaching 80% of the values found in
wt retinas
(Fig. 3) .
Amacrine cells labeled by parvalbumin, calbindin, and calretinin antibodies showed a partial reversal of phenotypic impairment
(Figs. 2 4)that was also demonstrated by quantitative analysis. The small number of parvalbumin-positive amacrine cells in the mutant retinas doubled after BDNF treatment
(Fig. 3) . The number of calbindin- and calretinin-positive amacrine cells also increased by 40% and 30%, respectively
(Figs. 3 4) .
BDNF also reversed phenotypic changes in the RGCs. The density of parvalbumin-positive RGCs reached 100% of control values. The expression of calretinin and calbindin was also upregulated in the GCL of BDNF-treated mutant mice
(Figs. 2 3 4) .
We conclude that the neurochemical phenotypic changes in horizontal, bipolar, amacrine, and ganglion cells observed in bdnf −/− mice can be reverted, at least partially, by providing exogenous BDNF as late as the second week of postnatal life.
The expression of Ca2+-binding proteins is reduced in horizontal, bipolar, amacrine, and ganglion cells in the retinas of bdnf −/− mice. This neurochemical phenotype can be reverted by treating the bdnf −/− retinas with exogenous BDNF during the second week of postnatal development.
The central nervous system of transgenic mice carrying targeted deletions of neurotrophin genes has been repeatedly investigated. A lack of BDNF during development causes functional impairments,
5 11 13 23 24 morphologic and structural abnormalities,
1 8 24 25 and reduced expression of several phenotypic markers.
1 8 9 26 Although clear effects are observed at a functional and structural level, naturally occurring cell death does not seem to be increased in the
bdnf −/− mice.
1 8 25 27 The present study provides a detailed immunocytochemical analysis of Ca
2+-binding protein expression on different cell types in the retinas of
bdnf −/− mice. Ca
2+-binding proteins represent useful phenotypic markers for various classes of retinal neurons in the mouse,
12 and the regulation of their expression by BDNF has been described in the rat retina.
4 6 7
Calbindin expression in horizontal cells was reduced in the retinas of bdnf −/− mice. The plexus of horizontal cell processes also appeared less dense. This could reflect morphologic changes in the horizontal cells, reduced immunoreactivity that renders the processes less visible by calbindin immunohistochemistry, or both. PGP 9.5 immunoreactivity also visualized an attenuated plexus of horizontal cell processes. The consistency of the observations made using these two immunohistochemical markers strongly suggests that horizontal cells are indeed affected in their morphologic development. In addition, parvalbumin labeling intensity was not visibly reduced in the horizontal cells of bdnf −/− mice, but the plexus of parvalbumin-positive processes in the OPL appeared less dense.
Horizontal cells express the BDNF receptor TrkB
5 (and present paper). It is, therefore, likely that BDNF affects horizontal cells directly. Effects of BDNF on horizontal cells were also observed in organotypic cultures.
7
We found that the expression of calretinin in cholinergic amacrine cells was not visibly reduced in
bdnf −/− mice and that the plexus of cholinergic processes (as labeled by calretinin antibodies) was not affected. These data confirm that BDNF is not essential for their morphologic maturation.
8 However, labeling for calbindin in the processes of cholinergic neurons was reduced in
bdnf −/− mice, indicating that different Ca
2+-binding proteins are also differentially regulated by BDNF in cholinergic amacrine cells.
Other subtypes of amacrine cells express calretinin, calbindin, or parvalbumin in the mouse retina.
12 These cells require endogenous BDNF for the expression of Ca
2+-binding proteins, immunoreactivity of which is clearly reduced in the retinas of
bdnf −/− mice.
RGCs in the mouse express Ca
2+-binding proteins: α-like cells are labeled with antibodies against parvalbumin, whereas RGCs are also labeled with antibodies against calretinin and calbindin, though with different patterns.
12 The expression of Ca
2+-binding protein in the RGCs is reduced in
bdnf −/− mice, particularly in the α-like cells expressing parvalbumin.
RGCs were the first central neurons for which a responsiveness to BDNF was described.
28 BDNF can increase the survival of RGCs and can influence their morphologic development.
29 It has been shown that endogenous BDNF influences the development of RGC axons, dendrites, and electrical properties.
11 25 30 31 All these effects are likely the result of direct action of BDNF on RGCs because RGCs express the BDNF receptor
32 33 34 and respond to BDNF in purified cultures.
28 35 Expression of Ca
2+-binding proteins is probably one of the aspects of RGC biology physiologically and directly regulated by BDNF.
In mammalian retinas, two types of bipolar cells can be distinguished: rod and cone bipolar cells. Rod bipolar cells represent a homogeneous population and can be labeled with antibodies against PKCα.
19 Cone bipolar cells constitute a heterogeneous set of cells comprising approximately 10 different classes
20 ; in the mouse retina, three types of cone bipolar cells can be labeled by recoverin antibody.
12
Synaptic transmission between photoreceptors and bipolar cells in
trkB −/− mice is deficient.
5 Flash-evoked electroretinography in these mice shows exclusively the electrical response of the photoreceptors with no detectable electrical response in the bipolar cells, though rod bipolar cells have normal postsynaptic receptors.
5 Our data reveal that rod bipolar cells have a normal gross anatomy; however, their axon endings appear to be atrophied. This effect could be attributed to real morphologic changes in the synapses or to a redistribution of PKCα. Given that bipolar cell terminals are subject to activity-dependent plasticity and undergo daily changes in correspondence to light/dark shifts,
36 the phenotype could originate in a lack of functional input to the bipolar cells. This effect could also be attributed to a trophic action of BDNF on bipolar cells because BDNF can support the survival of bipolar cells in vitro.
7 37
Cone bipolar cells showed a more easily visible phenotype. Expression of their phenotypic marker, recoverin, was almost completely abolished, as already described in
trkB −/− mice.
5
Detailed analysis of the retinas of bdnf −/− mice revealed that expression of different Ca2+-binding proteins is regulated independently within the same cell type. Expression of calbindin, but not of parvalbumin, is reduced in horizontal cells, whereas expression of calbindin, but not of calretinin, is reduced in cholinergic amacrine cells.
Expression of Ca
2+-binding proteins is reduced in the retinas of BDNF-deficient mice. Cell loss in small neuron populations, such as horizontal cells and recoverin-positive bipolar cells, could go unnoticed. On the other hand, in dense cell populations such as the calretinin- and calbindin-positive amacrine cells, a cell density diminution of approximately 40% to 50% would cause an evident alteration of the INL. Interestingly, however, the retinas of
bdnf −/− mice do not show a reduction in the thickness of any retinal layer
5 25 or any other major change. We performed TUNEL staining at P1, P10, and P14 in
wt and mutant mice and at P14 in
bdnf −/− + BDNF. We did not find any difference in the quantity of TUNEL-stained cells in any stage between the groups (data not shown). This, in addition to the partially restored expression of Ca
2+-binding proteins in the
bdnf −/− retina by the administration of BDNF, demonstrated the physical presence of these cells. Based on our analysis of
bdnf −/− mice, we conclude that BDNF is required for the phenotypic maturation and expression of the Ca
2+-binding proteins. Downregulation of Ca
2+-binding protein expression can be reverted, at least in part, in all cell types if exogenous BDNF is provided as late as during the second postnatal week.