p35 promotes the differentiation of amacrine cell subtype in the zebrafish retina under the regulation of egr1


  • Liyun Zhang,

    1. Department of Biological Sciences, Purdue University, West Lafayette, Indiana
    Search for more papers by this author
    • Liyun Zhang's present address is Department of Ophthalmology, University of Cincinnati, 3230 Eden Ave, Cincinnati, OH 45267.

  • Sylvia Bonilla,

    1. Department of Biological Sciences, Purdue University, West Lafayette, Indiana
    Search for more papers by this author
  • Yuqing Zhang,

    1. Department of Biological Sciences, Purdue University, West Lafayette, Indiana
    Search for more papers by this author
  • Yuk Fai Leung

    Corresponding author
    1. Department of Biological Sciences, Purdue University, West Lafayette, Indiana
    2. Department of Biochemistry and Molecular Biology, Indiana University School of Medicine Lafayette, West Lafayette, Indiana
    • Correspondence to: Yuk Fai Leung, Department of Biological Sciences, Purdue University, LILY 2-236, 915 W. State Street, West Lafayette, IN 47907. E-mail: yfleung@purdue.edu.

    Search for more papers by this author


Background: Early growth response 1 (egr1) is a transcription factor (TF) for controlling the differentiation of Parvalbumin (Parv) -expressing amacrine cells (ACs) in zebrafish. However, the downstream factors of this process have not been identified. The purpose of this study was to investigate the role of p35, a neuronal-specific activator of cyclin-dependent kinase 5 (Cdk5) and a known in vitro target of egr1, in the differentiation of these ACs. Results: In the p35-knockdown retinas, Parv+ but not islet1+ ACs were specifically reduced. This phenotype was highly similar to that in the Egr1-knockdown retinas. Furthermore, p35 expression was reduced in the Egr1-knockdown retinas, particularly in the AC region; while egr1 was only modestly reduced in this region in the p35-knockdown retinas. Conclusions: p35 likely acts downstream of egr1 to control the differentiation of Parv+ ACs. Developmental Dynamics 243:315–323, 2014. © 2013 Wiley Periodicals, Inc.


The retina is an extension of the central neural system (CNS) (Dowling, 2012). In vertebrate retinas, there are six types of neurons and one major type of glia that are organized into three cellular layers: the ganglion cell layer (GCL), inner nuclear layer (INL), and outer nuclear layer (ONL). The GCL is primarily composed of ganglion cells (GCs) and a small number of displaced amacrine cells (DACs). The INL is composed of amacrine cells (ACs), bipolar cells (BCs), and horizontal cells (HCs). In addition to these interneurons, Müller glial cells (MCs) are also found in the INL. The ONL is composed of cone and rod photoreceptors (PRs), the photosensitive neurons. During development, these retinal cells are sequentially generated from the same common progenitors in the retinal neuroepithelium (Livesey and Cepko, 2001). Generally, GCs are born first; then, they are followed by overlapping births of the other cell types with MCs being the last cell type to be formed. These cells will further migrate to the final location and differentiate, which is characterized by the expression of cell-type-specific proteins and extension of neurites. These neurites will finally establish synaptic connections in the inner plexiform layer (IPL) that connects the GCL and INL, as well as the outer plexiform layer (OPL) that connects the INL and ONL.

The formation of this intricate retinal structure is tightly regulated at the genetic level. A number of molecules have been reported to control retinal development through studies in zebrafish, mouse, and chick. These include sonic hedgehog a (shha) (Neumann and Nuesslein-Volhard, 2000), which mediates cell–cell signaling; N-cadherin, Sidekicks, Dscam, and Semaphorin, which mediate cell adhesion (Fuerst et al., 2012; Masai et al., 2003; Matsuoka et al., 2011; Yamagata et al., 2002; Yamagata and Sanes, 2008); MAGUK p55 subfamily member 5a (mpp5a), which controls cell polarity (Wei and Malicki, 2002); and SWI/SNF-related, matrix-associated, actin-dependent regulator of chromatin, subfamily a, member 4 (smarca4), which remodels chromatin and controls gene expression (Gregg et al., 2003; Link et al., 2000). To identify candidate genes regulated by smarca4 in the retinas, Zhang and colleagues microdissected the WT and smarca4 retinas and determined their differential gene expression by microarrays analysis (Zhang and Leung, 2010; Leung et al., 2008). The expression pattern and function of a number of the identified Smarca4-regulated genes were then subsequently validated by in situ hybridization (Hensley et al., 2011) and gene perturbation experiments (Leung et al., 2008; Zhang et al., 2012; Zhang et al., 2013) respectively. For example, we recently determined two transcription factors (TF) - irx7 and egr1 were essential for differentiation of INL cells (Zhang et al., 2012) and Parv+ and GABA+ ACs (Zhang et al., 2013), respectively. Nonetheless, it is unclear how these genes control retinal differentiation at the molecular level.

P35, a neuronal-specific activator of Cdk5 (Dhavan and Tsai, 2001; Ishiguro et al., 1994; Lew et al., 1994; Tsai et al., 1994), is an attractive candidate of Egr1 target for retinal development. There are several reasons to support this hypothesis. First, it was shown that egr1 mediated the induction of p35 in the PC12 cells treated with nerve growth factor (Harada et al., 2001). This in turn promoted the neurite outgrowth of these cells, a hallmark of terminal differentiation. Second, p35/Cdk5 was reported to be generally essential for neuronal differentiation in the CNS by regulating multiple important developmental events. These include cytoskeleton dynamics (He et al., 2008; Nikolic et al., 1998), cell adhesion (Kesavapany et al., 2001; Kwon et al., 2000), neurite outgrowth (Nikolic et al., 1996), cortical lamination (Chae et al., 1997; Ohshima et al., 1996), as well as maintenance of synapse and neuronal survival (Cheung et al., 2006; Cheung and Ip, 2004; Dhavan and Tsai, 2001). Third, p35 and Cdk5 were detected in developing retinas in rat (Hirooka et al., 1996) and zebrafish (Leung et al., 2008). The latter study also found that p35 was under-expressed in the smarca4 retinas. Fourth, an initial characterization of the p35-knockdown retinas in the same study showed that their structure was compromised, supporting that p35 might play a role in retinal differentiation (Leung et al., 2008). This p35 was determined as one of the two p35 isoforms in zebrafish subsequently (cyclin-dependent kinase 5, regulatory subunit 1b (cdk5r1b); ZFIN ID: ZDB-GENE-040718-220). The current study sought to investigate the possible role of this p35 in mediating Egr1-regulated AC differentiation (Zhang et al., 2012; Zhang et al., 2013).


The Expression Dynamics of p35 During Retinogenesis

To obtain an expression pattern of p35 during retinogenesis, in situ hybridization was conducted on wild-type (WT) embryos from 24–120 hpf (Fig. 1 and data not shown). P35 was initially detected in an anterior-ventral patch of the zebrafish retina by in situ hybridization at about 36 hr postfertilization (hpf) (Fig. 1A and A', arrows), a stage when GCs begin to differentiate (Schmitt and Dowling, 1996). Then, p35 expression spread to the dorsal GCL and also began to express in the basal part of the INL, which contains the differentiating ACs, at 52 hpf (Fig. 1B and B', arrows). This is also the stage that marks the initial establishment of the IPL (Schmitt and Dowling, 1996, 1999). By 60 hpf (Fig. 1C and C'), p35 expression became more intense in the GCL and basal INL, and an additional weaker signal was detected in the apical INL. The latter signal appeared to be more defined in the HC region by 72 hpf (Fig. 1D, D' and E), a stage when the retina is mature enough to elicit visual responses (Easter Jr. et al., 1996). In the meantime, the expression of p35 remained high in the GC and AC regions, and was largely excluded from the proliferative marginal zone (MZ). Starting at this stage, the expression pattern of p35 remained consistent (data not shown). The dynamic expression of p35 in the differentiating retina but not the proliferative cells suggests that p35's function might be essential for retinal differentiation and/or maintenance of differentiated cells. Moreover, the first detectable p35 staining in the AC region was at approximately 52 hpf, which is later than the commencement of egr1 expression in the same area at 40 hpf (Zhang et al., 2013). The spatial overlap and sequential expression of egr1 and p35 in the AC region is consistent with the possibility that p35 is the downstream target of egr1.

Figure 1.

The expression dynamics of p35 during zebrafish retinogenesis. A time-series whole-mount in situ hybridization was performed to detect the expression pattern of p35 in WT retina. The resulting eyes were dissected and imaged from the lateral and dorsal views. The signal of p35 was first detected in an anterior-ventral patch of the retina at around 36 hpf (A, A'; arrow). Then, it spread towards the dorsal side of the GCL and was mainly detected in this region at 52 hpf (B, B'; arrows). By 60 hpf, p35's staining was detected in the GCL and AC region (C, C'). Faint staining in the apical INL began to appear at this stage, and became more conspicuous at 72 hpf (D, D'). This signal originated from HCs, as revealed by cryosectioning of the embryos at the same stage (E). In addition, the staining in the GCL and AC region became stronger than the earlier stages. GC, ganglion cells; AC, amacrine cells; HC, horizontal cells; IPL, inner plexiform layer; OPL, outer plexiform layer. For A–D, anterior is to the left and dorsal is up; for A'–D', anterior is to the left and lateral is up; for E, lateral is to the left and dorsal is up. Scale bars = 50 μm.

P35 Knockdown Affected the Formation of the Retinal Plexiform Layers at 72 hpf

To investigate the role of p35 in retinal differentiation, its expression was knocked down by microinjection of two non-overlapping translation blocking MOs, p35 tMO1 and p35 tMO2. These MOs had been previously used for p35 knockdown in zebrafish (Leung et al., 2008). The phenotype of the two types of p35-knockdown embryos (morphants) was highly similar, and the results obtained from the p35tMO2 morphants (refer to as p35 morphants hereafter) are discussed below. The plexiform layers and nuclear layers in these morphants were compromised compared with the embryos injected with equal amount of standard control MO (stdCTLMO; referred to as controls hereafter). These problems were revealed by immunostaining of cryosectioned retinas that were stained with phalloidin and DAPI, which highlight the plexiform layers and nuclei, respectively (Fig. 2). The IPL appeared thinner and irregular, and the OPL was mildly reduced in the p35 morphants (Fig. 2B and D) compared with the controls (Fig. 2A and C). These defects indicate that p35 might mediate retinal differentiation.

Figure 2.

Retinal differentiation was affected by p35 knockdown at 72 hpf. The retinal histology of the p35 morphants (p35 tMO2) and controls (stdCTLMO) was analyzed at 72 hpf. At this stage, the phalloidin staining highlighted IPL and OPL with a sharp and distinct boundary in the controls (A), while these plexiform layers were thinner and irregular in the p35 morphants (B). A similar observation of the plexiform layer formation was also revealed by the DAPI nuclei staining on the same sections (C, D). IPL: inner plexiform layer; OPL: outer plexiform layer. Scale bar = 50 μm.

P35 Knockdown Compromised the Differentiation of ACs Similar to Egr1 Knockdown

Our previous study has shown that the differentiation of Parvalbumin-positive (Parv+) ACs was specifically compromised in the Egr1 morphants (Zhang et al., 2013). In addition, p35 is a putative downstream target of egr1 (Harada et al., 2001) and the current expression results indicate that p35 began to express in the same AC domain at a slightly later stage (Fig. 1) than egr1 (Zhang et al., 2013). Together, these observations led to the formulation of two hypotheses: first, the Egr1-regulated differentiation of ACs was mediated by p35; second, knocking down p35 would phenocopy Egr1-morphant retinas. To test these hypotheses, an immunostaining analysis of ACs was conducted in the p35 morphants at two stages: (1) at 72 hpf when the defects in AC differentiation could be detected in the Egr1 morphants and (2) at 120 hpf, a stage when all early- and late-born retinal neurons can be readily detected for the evaluation of potential developmental delay. In particular, different subtypes of ACs were analyzed by anti-Islet1 (Islet1; Fig. 3A–D) and anti-Parvalbumin (Parv; Fig. 3E–H) as in the previous report (Zhang et al., 2013). Islet1 is a marker that labels cholinergic ACs in mice (Elshatory et al., 2007) and a subset of ACs in zebrafish (Zhang et al., 2013), while Parv labels a subset of GABAergic ACs in zebrafish (Zhang et al., 2013).

Figure 3.

p35 knockdown compromised the differentiation of Parv+ but not Islet1+ ACs. Immunohistochemical analysis of ACs in the controls (stdCTLMO) and p35 morphants (p35tMO2) was performed with anti-Islet1 (Islet1; A–D) and anti-Parvalbumin (Parv; E–H) for cholinergic and a subset of GABAergic ACs, respectively, at 72 and 120 hpf. These markers are shown in green, while the plexiform layers are highlighted by phalloidin in red. A magnified view of the white box is shown at the bottom of each whole retinal section. For Islet1+ ACs, they were not aligned regularly on the boundary of the apical IPL at 72 hpf (A''', B'''; white arrows). In addition, their numbers were also reduced. By 120 hpf, these differentiation problems were largely resolved and the two conditions became comparable (C-C''' and D-D'''), indicating that the effect was caused by a developmental delay. Interestingly, Parv+ ACs were almost absent in the p35-morphant retinas compared with the controls at 72 hpf (E-E''' and F-F''') and was still substantially reduced at 120 hpf (G-G''' and H'-H'''). The differentiation defects of these ACs likely contributed to the thinning of the IPL at 72 hpf. In addition, the reduction of Parv signal at 120 hpf was quantified as described in the Experimental Procedures section and Table 1. I: An example of the four different staining grades. The white arrow in grade 3 indicates a small region with positive Parv projections detected in the IPL. For A–H: The cryosection of a whole retina is shown at the top and a magnified view of the white box in the central retina is shown at the bottom. For the whole retinal images, lateral is to the left and dorsal is up. Scale bars for the whole retinal section and magnified view were 50 and 25 μm, respectively.

At 72 hpf, both Islet1+ and Parv+ ACs in the p35-morphant retinas were compromised compared with the controls (Fig. 3A, B, E, F). This reduction of ACs and/or their dendritic processes likely contributed to the observed thinning of the IPL at 72 hpf (Fig. 2). Nonetheless, the analysis at 120 hpf strongly implicates that p35 knockdown only specifically reduced Parv+ ACs (Fig. 3G, H). At this stage, these ACs could extend clear neuronal projections to two sub-laminae of the IPL in the control retinas (Fig. 3G). However, this was not the case in the p35-morphant retinas. Many of these retinas did not have any signal (Fig. 3H) or only had small regions with projections in the two sub-laminae (Fig. 3I). In addition, the signal was generally weaker. To quantify this difference, the number of retinas with these two projections detected up to a certain fraction of the whole IPL was counted (Zhang et al., 2012; Zhang et al., 2013). The results show that there was a difference in the counts between the p35-morphant and control retinas (Table 1; Fisher's exact test, P = 0.01). In the meantime, the Islet1 signal in the p35-morphant retinas was very comparable to the controls (Fig. 3C, D). Specifically, there was no difference in the Islet1+ ACs per retinal area between the two groups (p35 morphants: inline image = 1215.3 mm−2, standard deviation s = 112.6 mm−2, N = 9; controls: inline image = 1,192.6 mm−2, s = 84.3 mm−2, N = 10; Mann-Whitney test: U = 44, P = 0.935). This result suggests that the reduction of Islet1+ ACs at 72 hpf was likely caused by developmental delay.

Table 1. The Counts of Different Staining Grades Obtained From the Immunostaining and In Situ Hybridization Analysesa
A. Parv analysis in p35 morphants
Control staining grade counts (grades 1–4)bMorphant staining grade counts (grades 1–4)Fisher's Exact Test PFigure (staining grade)
4,5,1,03,0,2,50.00733G (grade 1) and 3H (grade 4)
B. In Situ Hybridization Analysis in p35 and Egr1 Morphants
Gene probeMorphantControl staining grade counts (grades 1–3)bMorphant staining grade counts (grades 1–3)Fisher's Exact Test PFigure (staining grade)
  1. a

    (A) To quantify the effect on Parv immunostaining after p35 knockdown at 120 hpf (Fig. 3), the number of retinas with Parv+ ACs projection into two sub-laminae in the IPL detected up to a certain fraction of the whole IPL was counted: Grade 1 = full IPL; 2 ≤ ¾; 3 ≤ 1/2; 4 = no signal. (B) To determine the potential regulation of egr1 on p35, the expression of each gene was detected in the knockdown embryos of the other by in situ hybridization as shown in Figure 5. The alteration of the resulting staining patterns was categorized and counted. For p35 staining: grade 1 = strong signal reduction in general, faint and no staining in ACs; grade 2 = intermediate signal reduction in general and a prominent signal reduction in ACs; and grade 3 = strong signal detected in GCs, ACs, and HCs. For egr1 staining: grade 1 = strong signal reduction or no staining, grade 2 = intermediate signal reduction with more reduction in GCs than ACs; and grade 3 = strong signal detected in GCs and ACs. Finally, the difference in the count distribution between the controls and the morphants was determined by Fisher's exact test. The corresponding figures and staining grade showed are also indicated in the table.

  2. b

    The control MO for p35 knockdowns was a standard control MO (stdCTLMO), while the control MO for Egr1 knockdown was a 5-base mismatch MO (5misCTLMO).

egr1p350, 0, 92, 9, 01.19e-055C (grade 3) and 5D (grade 2)
p35Egr10, 0, 107, 1, 02.285e-055A (grade 3) and 5B (grade 1)

To confirm this possibility, immunostaining of the early-born GCs and three late-born BCs, rods, and MCs was conducted with anti-zn8 (zn8), anti-PKCβ1 (PKC), anti-zpr3 (zpr3), and anti-glutamine synthetase (GS), respectively (Fig. 4). The differentiation of early-born GCs in the p35 morphants was largely normal (Fig. 4A–D). For instance, the number of GC per retinal area was not altered at 72 hpf (data not shown). Even though there was a mild effect on the dendritic projection into the IPL (Fig. 4A, B), the staining became highly comparable between the p35 morphants and controls by 120 hpf (Fig. 4C, D). For BCs, rods, and MCs, even though their differentiation was compromised at 72 hpf in the p35 morphants (data not shown), they appeared largely normal by 120 hpf (Fig. 4E–J).

Figure 4.

The differentiation of GCs, BCs, rods, and MCs appeared normal in the p35-morphant retinas. A–D: Immunohistochemical analysis of GCs in the controls (stdCTLMO) and p35 morphants (p35tMO2) was performed with anti-zn8 (zn8) at 72 and 120 hpf. The zn8 signal is shown in green, while the plexiform layers are highlighted by phalloidin in red. At 72 hpf, the number of GCs, as counted by the DAPI nuclei with positive zn8, was not different between the control and p35-morphant retinas (A, B). This result suggests that the formation of GCs was not affected by p35 knockdown. Even though the dendritic projections of GC appeared compromised in the p35-morphant retinas at this stage, the overall zn8 staining pattern between the control and p35-morphant retinas was comparable at 120 hpf (C, D). These observations suggest that the GC differentiation was normal by this stage. E–J: Immunohistochemical analysis of BCs, rods, and MCs in the controls and p35 morphants was performed with anti-PKC βI (PKC), anti-zpr3 (zpr3), anti-GS (GS) at 72 hpf (data not shown) and 120 hpf. The p35 knockdown substantially reduced the staining signal of these markers in the p35-morphant retinas compared with the controls at 72 hpf. However, the staining became very comparable between the two conditions at 120 hpf. In these images, lateral is to left and dorsal is up. IPL: inner plexiform layer; OPL: outer plexiform layer. Scale bars = 50 μm.

Together, these immunostaining data strongly implicate that the differentiation of Parv+ ACs was specifically caused by the p35 knockdown. The other cell types might be slightly affected by the knockdown, but the data indicate that this was largely caused by a developmental delay. Interestingly, this specific differentiation problem in the Parv+ and not Islet1+ ACs is remarkably similar to that in the Egr1-morphant retinas (Zhang et al., 2013). Together with the spatial overlap in the expression of egr1 and p35 in the presumptive AC region, these observations in turn imply that p35 may act downstream of egr1, and that their function is important for the differentiation of Parv+ ACs.

Egr1 Positively Regulated p35 Expression, Particularly in the AC Region

To detect the potential relationship between p35 and egr1, p35 expression was determined in the Egr1 morphants and the corresponding controls by in situ hybridization (Fig. 5A,B) and quantitative reverse-transcription PCR (qPCR) at 72 hpf. In the Egr1 morphants, the staining of p35 was weaker in general and the signal in the AC region was strongly reduced (Fig. 5B–B”, arrows), while in the controls, the expression of p35 was always detected in GCL and AC region at 72 hpf (Fig. 5A–A”, arrows). The number of embryos with a specific type of staining with and without knockdown (see Table 1 for the definition of the specific staining grades) was counted and analyzed by Fisher's exact test. The results show that there was a change in the p35 staining in the Egr1 morphants compared with the controls (Table 1; P < 2.29e-05). In addition, the general reduction of p35's expression in the Egr1-morphant eyes was also confirmed by qPCR analysis (Egr1KD vs. Control = 0.62, range = 0.59–0.66; Welch Two Sample t-test, P = 0.00074). Hence, p35 expression in the retina, particularly in the AC region, was likely dependent on egr1.

Figure 5.

p35 expression in the AC region of the Egr1-morphant retinas was severely reduced, while egr1 expression in the p35-morphant retinas was modestly attenuated. A,B: Whole-mount in situ hybridization of p35 was performed with the control- (5misCTLMO) and Egr1-morphant (egr1sMO) embryos at 72 hpf. The p35 signal was generally reduced in the Egr1-morphant retinas (B, B') compared with the controls (A, A'), and it was significantly reduced in the AC region (arrows in A', B'). The latter was particularly obvious on the sectioned retinas (A”, B”). C,D: Whole-mount in situ hybridization of egr1 was performed with the control- (stdCTLMO) and p35-morphant (p35 tMO2) embryos. In the controls, egr1 was strongly expressed in the GCL and AC region at 72 hpf (C–C”; arrows), while in the p35 morphants, egr1 signal was reduced in the GCL and AC region (D–D”; arrows). Nonetheless, the overall signal reduction in the AC region in the whole-mount retinas was relatively less drastic than that in the GCL (C', D'). For the lateral view of the dissected eyes in A–D, anterior is to the left and dorsal is up; for the dorsal view in A'–D', anterior is to the left and lateral is up; while for the cross-section in A”–D”, dorsal is to the left and lateral is up. GC, ganglion cell; AC, amacrine cell; IPL, inner plexiform layer; OPL, outer plexiform layer. Scale bars = 50 μm.

Intriguingly, egr1 expression was also found to be modestly reduced by p35 knockdown at 72 hpf (Fig. 5C,D). First, the overall egr1 expression in the p35 morphants (Fig. 5D–D”) was slightly reduced compared with the controls (Fig. 5C–C”) (Table 1; Fisher's exact test, P = 1.19e-05). This observation was confirmed by qPCR analysis of samples collected at the same stage (p35KD vs. Control = 0.59, range = 0.47–0.72; Welch Two Sample t-test, P = 0.00018). Second, there was a possible regional difference in the alteration of egr1 expression in the p35-morphant retinas. The staining in the GCL was substantially reduced, while the staining in the AC region was only moderately attenuated (Fig. 5C', D').

Since p35 began to express in the GCL at 36 hpf (Fig. 1) and is well before the initial expression of egr1 at 40 hpf (Zhang et al., 2013), these expression results indicate that the initial induction of p35 in the GCL was likely to be independent of egr1. Furthermore, p35 might also maintain the expression of egr1 once it was induced. However, in the AC region, p35 expression was substantially influenced by egr1 but not vice versa. Together, these results indicate that the nature of interaction between these two genes was probably different in GCs and ACs.


P35 Expression Is Essential for Proper Retinal Development and Differentiation of Parv+ ACs

Cdk5/p35 has long been shown to play an important role in neural differentiation (Dhavan and Tsai, 2001). Even though it was previously found that these genes expressed dynamically during rat retinogenesis (Hirooka et al., 1996), their roles in early retinal development remain uncharacterized. The current study has identified several key defects in the p35-morphant retinas in zebrafish. First, the knockdown of p35 compromised retinal differentiation (Fig. 2). This observation is consistent with a previous initial investigation of p35 knockdown (Leung et al., 2008). Second, p35 knockdown specifically affected the differentiation of GABAergic Parv+ ACs. Third, the differentiation of cholinergic Islet1+ ACs (Elshatory et al., 2007) was delayed but appeared normal at a later stage (Fig. 3). This effect on AC-subtype differentiation is highly similar to that in the Egr1 knockdown (Zhang et al., 2012). The similarity of the phenotypes in these two knockdown models, as well as the sequential expression dynamics of egr1 and p35 in the AC region, have strongly implicated that p35 might be a downstream effector of egr1 that controlled the differentiation of Parv+ ACs. This is also consistent with the known regulatory relationship that egr1 activates p35 in vitro (Harada et al., 2001). Nonetheless, the current experimental design was not designed to address whether this regulatory relationship was direct or indirect in the zebrafish retina. This can be addressed by p35 promoter element analysis and chromatin immunoprecipitation with Egr1, as well as the over-expression of p35 by parvalbumin promoter in the Egr1 morphants.

It should also be noted that there are two alternative possibilities that may explain the defects in Parv+ ACs differentiation in the p35 morphants. The first one is an overall developmental delay. This possibility is unlikely, as several late stage markers for BCs, rods and MCs expressed normally in the p35 morphants (Fig. 4).The second possibility is that p35 knockdown affected the differentiation of GCs, which in turn compromised ACs differentiation. This possibility is also unlikely, as the GC number and differentiation appeared normal in the p35-morphant retinas (Fig. 4). In addition, two studies of the atoh7 mutants in mouse (Brown et al., 2001) and fish (Kay et al., 2001) showed that other retinal cell types including ACs differentiated fairly normally despite that GCs failed to form. Therefore, the defect in Parv+ AC differentiation in the p35 morphants was likely a direct effect of p35 knockdown.

Recently, two reports studied the birthdays of ACs and found that GABAergic ACs were formed earlier in development than glycinergic ACs (Cherry et al., 2009; Voinescu et al., 2009). The study conducted by Voinescu and colleagues (2009) also found a strong correlation between the soma position of the AC subtypes to their birthdays. Specifically, the earlier-born ACs were found in the basal region of the retina. Parv+ ACs somata were often found next to the IPL (Fig. 3). Thus, it is likely that Parv+ ACs are among the early-born GABAergic ACs and our study indicates that their differentiation might be mediated by the Egr1-p35 pathway.

The Regulatory Relationship Between p35 and egr1 Is Likely Different in the GCL

The interaction between egr1 and p35 might be different in the GCL. First, the expression dynamics of egr1 and p35 in the GCL do not support the idea that p35 was activated by egr1 in this region. While p35 began to express in the ventral GCL at 36 hpf and expressed considerably in the GCL by 52 hpf (Fig. 1), the earliest detectable expression of egr1 in the GCL was at approximately 40 hpf (Zhang et al., 2013). Therefore, the initial induction of p35 in GCs is likely independent on egr1. Also, the expression of p35 was mildly reduced in the GCL when Egr1 was knocked down, unlike the situation in the AC region (Fig. 5B). These observations indicate that there might be other factors that initiated p35 expression in the GCL and that egr1 might play a maintenance role of p35 expression in this region.

Other Cdk5 Activators May Play Different Roles in Retinal Differentiation

P35 is one of the two Cdk5 activators in the central nervous system. Another paralogue that can activate Cdk5 is P39 (Dhavan and Tsai, 2001). In zebrafish genome, the genes that encode p35 and p39 have undergone duplication and resulted in at least four possible Cdk5 activators. It is possible that these genes have diverged in their function during evolution. This study has revealed a specific effect of one of the p35 isoforms on the differentiation of Parv+ ACs. Since we previously showed that p39 mediated retinal differentiation under the control of Smarca4 (Leung et al., 2008) and our unpublished observations indicate that p39 has a different expression pattern in the retina compared with p35, thus, we anticipate that different Cdk5 activators may play different roles in retinal development and differentiation.


Zebrafish Maintenance and Embryo Collection

Zebrafish AB line was maintained according to standard procedures (Westerfield, 2000). Embryos were collected after breeding the parental fish in pairs for 15 min, maintained in E3 medium at 28°C and staged as described (Kimmel et al., 1995). For embryos that were collected for in situ hybridization, 0.003% PTU (Sigma, St. Louis, MO) in E3 medium was applied between 12 and 23 hpf to inhibit melanization (Li et al., 2012). All experimental protocols were approved by the Purdue Animal Care and Use Committee.

Morpholinos (MOs) Injection

To knock down Cdk5r1b (p35) (NCBI Reference Sequence: NM_001002515.1), two non-overlapping translation-blocking MOs (6 ng p35tMO1: TGAAAGTGATAGCACGGTTCCCATG or 2 ng p35tMO2: AGCCCACTGCTTCCTGAGACTAG) (Leung et al., 2008) were used. They were individually injected into the zebrafish embryos at the one-cell stage (Nusslein-Volhard and Dahm, 2002). The corresponding amount of a standard control MO (stdCTLMO) (CCTCTTACCTCAGTTACAATTTATA) was injected into the embryos to control for other non-specific effects induced by microinjection. Egr1 was knocked down with a splice-blocking MO as previously reported (Zhang et al., 2013). All MOs were purchased from Gene Tools (Philomath, OR) or Thermo Scientific (Waltham, MA; formerly Open Biosystems).


All embryos used for immunohistochemistry were collected and fixed in 4% paraformaldehyde (PFA) according to an established protocol (Godinho et al., 2005). Ten-micrometer-thick transverse cryosections were collected and immunostaining conducted as described. The antibodies used in this study and their dilutions are as follows: mouse anti-zn8 (1:500, ZIRC, Eugene, OR), mouse anti-Islet1 (1:50, Developmental Studies Hybridoma Bank, Iowa City, IA), mouse anti-Parvalbumin (1:500, Sigma P3088), rabbit anti-PKC βI (1:300, Santa Cruz Biotechnology, Santa Cruz, CA), mouse anti-zpr3 (1:200, ZIRC, Eugene, OR), mouse anti-GS (1:500, Millipore, Billerica, MA), and Alexa Fluor 488/555 goat anti-rabbit/mouse IgG (1:1000, Invitrogen, Carlsbad, CA). Alexa Fluor 633 phalloidin (1:50, Invitrogen) was included in the first antibody mixture to stain for F-actin, which would highlight the plexiform layers. DAPI (100 ng/mL) was used to counterstain cell nuclei.

In Situ Hybridization

In situ hybridization was conducted as described (Hensley et al., 2011). The riboprobes used in this study are as follows: p35 (this study) and egr1 (Zhang et al., 2013). To prepare a probe for p35, a fragment of the p35 gene was amplified from a cDNA library prepared as described (Leung and Dowling, 2005) with the following primers (p35-1F: 5′-TGGGAACCGTGCTATCACTT-3′; p35-1R: 5′- GAGGATCCGAGTTGATCTGC-3′). The resulting PCR fragment was cloned into the pGEM-T easy vector (Promega, Madison, WI) for propagation. The riboprobes were synthesized according to standard procedures (Nusslein-Volhard and Dahm, 2002).

Quantitative Reverse-Transcription PCR (qPCR)

Total RNAs were extracted from 30 eyes collected from 15 3-dpf larvae and reverse transcribed as described (Leung and Dowling, 2005). Four types of eye samples were collected: Egr1-morphant, 5misCTL (control of Egr1-morphant), p35-morphant, and StdCTL (control of p35-morphant). qPCR was performed using SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA) and run on an Applied Biosystems 7300 Real-Time PCR System as described by the manufacturer. The following primer pairs were used in this study to amplify p35: p35-F: 5′-GCGGCTCTTGTTTTGCTAAAG and p35-R: GCTTCCTGAGACTAGTTCAATCTG. The primers that were used to amplify egr1 were previously reported (Zhang et al., 2013). β-actin was utilized as an internal control, and the primers for its amplification were also previously reported (Zhang et al., 2013). All primers were designed and purchased from Integrated DNA Technologies (IDT, Coralville, IA). Two biological replications were conducted and there were three technical replications in each case.

Image Acquisition and Data Analysis

Bright-field and fluorescent images were acquired by a SPOT-RT3™ colour slider camera (Diagnostic Instruments) mounted on an Olympus BX51 fluorescence compound microscope or SZX16 stereomicroscope. Features of the samples in the images were extracted by i-Solution (IMT i-Solution). The counts of the zn8+ GCs and Islet1+ ACs were normalized by the corresponding retinal areas excluding the optic nerve region. The staining of Parv+ ACs at 120 hpf was categorized by the extent of the neuronal projections to two sub-laminae of the IPL: (type 1 = full IPL; 2 ≤ ¾; 3 ≤ 1/2; 4 = no signal) (Zhang et al., 2012; Zhang et al., 2013). For in situ hybridization, a three-level grading scheme was adopted to quantify the staining results (see Table 1 for the definition of the grades for each gene).

Statistical Analysis

All standard descriptive statistics and data analyses were performed in SPSS 16.0. The analysis of the immunostaining and in situ hybridization data for two sample groups was conducted by Mann-Whitney test or Fisher's exact test. qPCR data was analyzed by the ΔΔCt method (Livak and Schmittgen, 2001) and standard error propagation method was utilized to combine the errors. The qPCR results were reported in ratio of mRNA in the Egr1- or p35-morphant eyes to that in the controls (2^-ddCt) and the corresponding range as 2^-(ddCt ± ddCtErr). Welch two-sample t-test was used to compare the dCt values between the morphants and controls. An alpha level of 0.05 was used for all statistical tests.


The authors thank Daniel Szeto, Hung-Tat Leung, and members from the Leung lab for helpful discussions. This study was supported by Pediatric Ophthalmology Research Grants from Knights Templar Eye Foundation to L.Z. and Y.Z., a Charles D. Kelman, M.D. Scholar award from the International Retinal Research Foundation to L.Z., and awards from Hope for Vision and Showalter Research Trust to Y.F.L.