To obtain an expression pattern of egr1 during retinal development, in situ hybridization was conducted on wild-type (WT) embryos at 24, 28, 36, 40, 44, 48, 52, 60, 72, 84, 96 and 120 hpf (Figure 1 and data not shown). The first detectable signal of egr1 was found in the anterior-ventral retina at 40 hpf (Figure 1A). This is the approximate stage when the INL cells in the same region begin to withdraw from cell cycle [26]. Contrary to a previous report that egr1 expression in retina was detected at 40 hpf and disappeared by 48 hpf [24], the results in this study show that egr1 continued to express and spread to the dorsal retina. The signal was detected in the GCL in addition to the AC region at 52 hpf (Figure 1B), a stage when the retinal lamination is first established. The staining signal of egr1 became more intense by 72 hpf, and was mainly detected in the GCL and AC regions (Figure 1C & E). The late retinal expression of egr1 is further supported by the in situ hybridization data at zfin.org that show an intense retinal staining at the protruding-mouth stage (∼72 hpf) (ZDB-GENE-980526-320). In addition, a few cells in the HC and PR regions also began to express egr1 starting at this stage. The signal in these regions became more prominent, particularly in the peripheral outer retina, by 120 hpf (Figure 1D & F). The initial expression dynamics of egr1, which was primarily located in the proximity of the IPL during its first establishment, hints at the possibility of its role in guiding the differentiation of cells in this area. The later appearance of egr1 in PRs suggests that it may play a role in PR differentiation and/or function, a role that has been implicated by other investigations [27].

To determine egr1’s function in retinal development, Egr1 was knocked down by microinjection of MOs in developing zebrafish embryos and the resultant retinal phenotypes examined. A splice-blocking MO (egr1sMO) (Figure 2) and a translation-blocking MO (egr1tMO) were used. Since the retinal phenotypes of these two types of MO-knockdown embryos (morphants) were comparable, the results obtained by the injection of 4 ng of egr1sMO are presented below. This amount was chosen because of the following two reasons: first, the gross morphology of the embryos injected with 4 ng of 5-base mismatch control MO (5misCTL MO; the embryos injected with this MO will be referred to as controls hereafter) (Figure 2B) was indistinguishable from the uninjected embryos (Figure 2A) at 72 hpf; while the injection of the same amount of egr1sMO led to a reduction of eye and head size, as well as a general shortening of the trunk in most of the morphants (Figure 2C) (N = 134 out of 150, or 89.3%). The remaining morphants (N = 16 out of 150, or 10.7%) were unhealthy and excluded from subsequent characterizations. This morphological problem caused by Egr1 knockdown persisted to 120 hpf (Figure 2F), when the uninjected embryos (Figure 2D) and controls (Figure 2E) were still highly comparable to each other and healthy. Second, there was a substantial reduction of the amount of egr1 mRNA in the Egr1 morphants to 20–40% of the control level up to 120 hpf, as determined by quantitative PCR (qPCR) (Figure 2G). This suggests that Egr1 expression was substantially reduced in the morphants. Contrary to a previous report [25], Egr1 knockdown in this study did not induce a wide range of phenotypes. Since the injection volume in this study was calibrated and the resulting injected embryos were screened for the evenness of the fluorescence signal from the FITC-dextran tracer, it is believed that the phenotypes that were observed in this study are genuine.

To quantify the effect of Egr1 knockdown on eye size, the anterior-posterior length of the eyes was measured as described [28]. The results indicate that there was a reduction of eye size in the Egr1 morphants (mean () = 252.60 µm, standard deviation (s) = 14.37 µm, N = 19) compared with the controls ( = 301.36 µm, s = 7.05 µm, N = 10) (Mann-Whitney test, U = 0, p-value<0.001) at 72 hpf. After normalizing the eye length with the body length of the same embryo, the morphants still had a smaller eye/body ratio compared with the controls (Figure 2H; morphants:  = 0.0866, s = 0.0042; controls:  = 0.0899, s  =  0.0031; Mann-Whitney test, U = 49, p-value = 0.035). This reduction in eye/body ratio persisted to 120 hpf (Egr1 morphants:  = 0.0794, s = 0.0066, N = 21; controls:  = 0.0874, s = 0.0029, N = 16; Mann-Whitney test, U = 44, p-value<0.001). Together with a substantial reduction of egr1 mRNA at this stage, these results indicate that while there was a general reduction in the body size of the Egr1 morphants, there was also a specific reduction of the eye size in these embryos.

The retinal structure of the Egr1 morphants was abnormal compared with the controls at 72 hpf, a stage at which the retinas are mature enough to elicit visual activity [26]. First, the retinal lamination was not formed properly. The IPL and OPL that were highlighted by phalloidin were thinner and irregular in the Egr1 morphants compared with the controls (Figure 3A & B, arrows). The irregularity of the IPL was even more apparent in the sections stained by DAPI which highlighted the nuclei (Figure 3A’ & B’). In addition, some nuclei were mis-placed in the IPL and surrounded by the phalloidin signal (Figure 3A’’ & B’’, insets). Second, the nuclei of the INL cells were not stained as an intense apical sub-layer and a less intense basal sub-layer (Figure 3A’ & B’, asterisks). Moreover, cells in the ONL were less elongated (Figure 3B’, inset) compared with controls (Figure 3A’, inset), suggesting that PR differentiation might also be affected by Egr1 knockdown at this stage. The retinal lamination problem in the Egr1 morphants was largely resolved by 120 hpf (Figure 3D). In particular, the IPL and OPL in the Egr1 morphants were relatively normal compared with their counterparts at 72 hpf (Figure 3B) and were more comparable to the controls (Figure 3C). Nonetheless, the IPL remained thinner at this stage, as supported by measurements of the IPL thickness in the central retina (Egr1 morphants:  = 12.39 µm, s = 2.13 µm, N = 50; controls:  = 13.62 µm, s  =  2.51 µm, N = 33; two-tailed Student’s t-test, p-value  =  0.019) Consistent with the morphometric analysis, the overall size of the Egr1-morphant retinas was still smaller at 120 hpf. In addition, the phalloidin staining of the outer segments of PRs in the Egr1 morphants was still less intense compared with the controls (Figure 3C & D).

To further define the effect of Egr1 knockdown on retinal differentiation, immunostaining analysis of cells located in the INL was conducted with embryos collected at 72 and 120 hpf. Since egr1 begins to express in the AC region during retinogenesis (Figure 1), the analysis was first focused on ACs, an early retinal cell type that would be generated in this region. The markers used in the analysis include (1) anti-5E11 (5E11; Figure 4A–D), a pan-specific AC marker [29]; (2) anti-Parvalbumin (Parv; Figure 4E–H) [30] and (3) anti-GABA (GABA; Figure 4I–L) for GABAergic ACs [31]. The Parv marker labels a subset of GABA+ ACs (Figure S1); and (4) anti-Islet1 (Islet1; Figure 4M–P) for ACs which were shown to be cholinergic in mice [32]. This marker was also used to label a subset of ACs in zebrafish [33]. The results are also summarized in Table 1. The signal of 5E11 was substantially and moderately reduced in the Egr1-morphant retinas at 72 and 120 hpf respectively (Figure 4B & D); while the corresponding controls had extensive signal in the ACs and their projections into the IPL (Figure 4A & C). These observations suggest that ACs differentiation was compromised by Egr1 knockdown. This differentiation problem was also revealed by the analysis of three additional markers of AC subtypes. First, Parv+ ACs were mostly absent in the Egr1 morphants compared with the controls at both 72 and 120 hpf (Figure 4E–H). Second, GABA+ ACs were mostly absent in the Egr1 morphants at 72 hpf compared with the controls (Figure 4I & J); while the staining for these cells became more apparent by 120 hpf (Figure 4L), except for the intense staining on the basal IPL that was only observed in the controls (Figure 4K, arrow). Since the GABA staining in the normal retina overlapped substantially with Parv (Figure S1), this suggests that the suppressed GABA+ ACs might also be Parv+. Third, the number of Islet1+ ACs per retinal area was not reduced at both 72 hpf (Figure 4M & N and Table 2; Mann-Whitney Test, p-value  =  0.641) and 120 hpf (Figure 4O & P and Table 2; Mann-Whitney Test, p-value  =  0.705). Together, these results suggest that Egr1 knockdown specifically affected the differentiation of Parv+ and GABA+ ACs.

To determine the extent to which the differentiation problem of ACs was caused by a delay in development, immunostaining analysis of BCs and MCs, two late cell types in retinogenesis, was conducted. The markers used in this analysis include anti-PKCβ1 for BCs (PKCβ1; Figure 4Q–T) and anti-GS for MCs (GS; Figure 4U–X). While the staining of both PKCβ1+ BCs and GS+ MCs was suppressed in the Egr1 morphants (Figure 4R & V) compared with the controls (Figure 4Q & U) at 72 hpf, the expression of these markers was very comparable between the two groups at 120 hpf (Figure 4S & T; W & X). The suppression of MCs at 72 hpf but not at 120 hpf indicates that the differentiation defects at 72 hpf, including the malformation of the IPL (Figure 3B), were most likely caused by a delay in development induced by Egr1 knockdown. Nonetheless, the persistent of the suppression of Parv+ ACs (Figure 4H) and attenuation of GABA+ ACs (Figure 4L) amidst the recovery of MCs (Figure 4X) in the Egr1 morphants at 120 hpf strongly suggests that the suppression of these AC subtypes was a specific effect of Egr1 knockdown.

The lack of presynaptic projections from the INL cells would probably affect the differentiation of their postsynaptic partners in the GCL. In the meantime, egr1 also begins to express in the GCL at 52 hpf (Figure 1); hence, it was theorized that the differentiation of GCL would be affected by Egr1 knockdown. To investigate this possibility, immunostaining analysis of GCs and their dendritic projections into the IPL was conducted with anti-zn8 (zn8) at 72 and 120 hpf (Figure 5). The results are also summarized in Table 1. At 72 hpf, the differentiation of zn8+ GCs was compromised in the Egr1-morphant retinas. Specifically, the dendritic projection of the GCs into the IPL was almost absent (Figure 5B) when compared with the controls (Figure 5A). Intriguingly, the number of zn8+ GCs per retinal area was not different between the two groups (Table 2; Mann-Whitney Test, p-value  =  0.328), suggesting that the dendritic outgrowth of the zn8+ GCs was preferentially affected compared with the soma at this stage. By 120 hpf, the GCs in the Egr1-morphant retinas could extend dendritic projections into the IPL (Figure 5D), which was still thinner than the controls (Figure 5C). This is also supported by the IPL-thickness measurements as described above. Together, these observations indicate that the dendritic differentiation of GCs was substantially delayed in the Egr1 morphants. Nonetheless, the current experimental design did not discriminate whether this delay was a direct effect of Egr1 knockdown or a secondary effect induced by the differentiation defects of AC subtypes.

The formation of the outer retina was also affected in the Egr1 morphants, particularly at the earlier stage 72 hpf. The defects could be caused by an abnormal differentiation of HCs and PRs that are in proximity of the OPL, as well as the BCs and MCs as described above. Since it has been reported recently that the reduction in HC number is related to OPL formation [14], a similar analysis was conducted using Islet1 (Figure 4M–P; a magnified view is shown in Figure 6E–H) and anti-Prox1 (Prox1; Figure 4Y-AB; a magnified view is shown in Figure 6A–D) to investigate the extent to which HCs were reduced in the Egr1 morphants. The results are also summarized in Table 1. The number of Prox1+ HCs per retinal area in the Egr1 morphants was reduced at 72 hpf compared with the controls (Figure 6A & B) (Table 2; Mann-Whitney Test, p-value < 0.001). At 120 hpf, since the Prox1 staining became relatively faint (Figure 6C & D), a phenomenon that was also observed in another study [31], the morphologically distinct HCs with flattened nuclei and detectable Prox1 signal were counted and normalized by the retinal area. The results indicate that the number of Prox1+ HCs per unit retinal area was not different between the two groups at this stage (Table 2; Mann-Whitney Test, p-value  =  0.779). Interestingly, there were fewer Islet1+ HCs per retinal area in the Egr1 morphants compared with the controls at both 72 hpf (Figure 6E & F and Table 2; Mann-Whitney Test, p-value < 0.001) and 120 hpf (Figure 6G & H and Table 2; Mann-Whitney Test, p-value  =  0.003). Thus, these results indicate that at least the differentiation of Islet1+ HCs was specifically affected by Egr1 knockdown.

PR differentiation was investigated by immunostaining with anti-zpr1 (zpr1) for red-green double cones and anti-zpr3 (zpr3) for rods at 72 and 120 hpf. In the controls, zpr1+ and zpr3+ cells were detected in the whole ONL (Figure 7A & C) at 72 hpf, while these cells were primarily restricted to a small ventral region of the ONL in the Egr1 morphants at the same stage (Figure 7B & D). The results were then quantified by counting the number of sections with signal spanning from the ventral to a certain level of dorsal retina [14] (type 1: ≤ ¼, 2: ≤ ½, 3: ≤ ¾, 4  =  full retina). The results show that there was a difference in the counts between the controls and Egr1 morphants for zpr1 staining (control counts (type 1-4): 0, 0, 0, 14; Egr1-morphant counts: 10, 7, 10, 1; Mann-Whitney Test: U  =  7, p-value < 0.001) and zpr3 staining (control counts: 0, 0, 3, 13; Egr1-morphant counts: 20, 4, 2, 2; Mann-Whitney Test: U  =  22, p-value < 0.001). There was also a concomitant change in the opsin expression in different PR subtypes, including three opsins (opn1lw1: red, opn1sw2: blue and opn1sw1: uv) for three types of cone PRs and rhodopsin (rho) for rods (Figure S2). In addition, the expression of nr2e3, neurod and crx, three TFs that can specify PRs, was also investigated. The results show that the expression of nr2e3 was increased and more widespread in the Egr1 morphants (Figure S3B) compared with the controls (Figure S3A). For neurod and crx, their expression between the controls and Egr1 morphants was similar (Figure S3C-F). Nonetheless, the signal of the zpr1+ and zpr3+ PRs in the Egr1 morphants became much more comparable to the controls by 120 hpf (Figure 7E–H). Taken together, these experiments suggest that Egr1 knockdown altered the differentiation of all types of PRs at 72 hpf, but the differentiation of PRs was more comparable to the controls by 120 hpf. Thus, Egr1 knockdown delayed PRs differentiation.

Since egr1 is a TF, it is possible that it exerted its effect on the differentiation of ACs and HCs in the Egr1 morphants through transcriptional regulation of TFs that specify these cell types. To test this hypothesis, the expression of ptf1a that is transiently activated in all ACs and HCs precursors [31], [34] was studied in embryos collected at 52, 72 and 120 hpf by in situ hybridization (Figure 8). At 52 hpf, ptf1a was widely expressed in the developing neural retinas in both controls and Egr1 morphants (Figure 8A & B). By 72 hpf, the expression of ptf1a was restricted to the proliferative marginal zone (MZ) in the controls (Figure 8C), while there was still a noticeable ectopic expression in the INL of the Egr1 morphants (Figure 8D). The difference in the ptf1a expression pattern between the two groups diminished by 120 hpf, and the staining signal was detected in the MZs in both groups (Figure 8E–F).

Zebrafish AB line was maintained according to standard procedures [43]. Parental fish were bred for 15 minutes before embryo collection to ensure all embryos would be collected at a similar stage. Then, embryos were collected, raised at 28°C and staged as described [44]. For in situ hybridization, embryos were also treated with 0.003% PTU (Sigma) in E3 medium [45] between 12 and 23 hpf to prevent melanization. All protocols were approved by the Purdue Animal Care and Use Committee.

To knockdown Egr1 (NCBI Reference Sequence: NM_131248.1), either 3 ng of a translation-blocking MO (egr1tMO, sequence: AGCCATCTCTCTGGAGTGTGCTCGG) or 4 ng of a splice-blocking MO (egr1sMO, sequence: AAGAGGGATTTAGTGCTTACCTCCA) was injected into the yolk of embryos at one-cell stage as described [45]. Three nanograms of a standard control MO (stdCTLMO, sequence: CCTCTTACCTCAGTTACAATTTATA) was used as the control for egr1tMO, and 4 ng of a 5-base mismatch control MO (5misCTLMO, sequence: AACACGGATATAGTCCTTAGCTCCA) was used as the control for egr1sMO. All MOs were purchased from Gene Tools or Thermo Scientific (formerly Open Biosystems).

Total RNAs were extracted from 10 whole embryos at 48, 72, 96 and 120 hpf and reverse transcribed as described [46]. qPCR was performed using SYBR Green PCR Master Mix (Applied Biosystems) and run on an Applied Biosystems 7300 Real-Time PCR System as described [28]. Primers were designed and purchased from Integrated DNA Technologies (IDT). The mature spliced mRNA was amplified by egr1-F: 5’- AGTTTGATCACCTTGCTGGAG-3’ (located in exon 1) and egr1-R: 5’- AACGGCCTGTGTAAGATATGG-3’ (located in exon 2). β-actin was utilized as an internal control, and the primers for its amplification were (β-act-F: 5’-TGCTGTTTTCCCCTCCATTG-3’ and β-act-R: 5’-GTCCCATGCCAACCATCACT-3’).

In situ hybridization was conducted as described [13]. The riboprobes that were used in this study are as follows: early growth response 1 (egr1); cone-rod homeobox (crx); neurogenic differentiation (neurod); nuclear receptor subfamily 2 group E member 3 (nr2e3); pancreas specific transcription factor 1a (ptf1a); opsin 1 (cone pigments), short-wave-sensitive 1 (opn1sw1); opsin 1 (cone pigments), short-wave-sensitive 2 (opn1sw2); opsin 1 (cone pigments), long-wave-sensitive 1 (opn1lw1) and rhodopsin (rho).

All embryos were collected, fixed and stored according to a standard protocol [14], except for embryos used for GABA immunofluorescence, which were fixed in 4% paraformaldehyde (PFA) plus 0.1% glutaraldehyde. Ten-micrometer-thick transverse cryosections were collected and immunostaining conducted as described [14]. The antibodies used in this study and their dilutions are as follows: mouse anti-zn8 (1:500, ZIRC), mouse anti-Islet1 (1∶50, Developmental Studies Hybridoma Bank), mouse anti-parvalbumin (1∶500, Sigma P3088), rabbit anti-GABA (1∶500, Millipore AB131), mouse anti-5E11 (1∶10, [29]). mouse anti-Prox1 (1∶200, Millipore MAB5652), mouse anti-zpr1 (1∶200, ZIRC), mouse anti-zpr3 (1∶200, ZIRC), Alexa Fluor 488/555 goat anti-rabbit/mouse IgG (1∶1000, Invitrogen). 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. 100 ng/mL DAPI was used to counter stain cell nuclei.

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). For GCs, ACs and HCs immunostaining results, their cell counts were normalized by the corresponding retinal areas excluding the optic nerve region. For Islet1+ ACs counting, a line was drawn across the central IPL stained by phalloidin; then, the Islet1+ cells on the INL side were counted. It should be noted that if an Islet+ GC was substantially delaminated, it would be counted as a positive cell by this approach at 72 hpf. Nonetheless, the lack of a difference of Islet1+ ACs between the Egr1 morphant and controls at 120 hpf when the morphants formed a distinctive IPL (Figure 4P) indicates that the number of Islet1+ ACs was not affected by the knockdown at 120 hpf.

All standard descriptive statistics and data analyses were performed in SPSS 16.0. The analysis of data for two groups was conducted by Mann-Whitney test, except for IPL thickness analysis, which was conducted by two-tailed Student’s t-test. qPCR data were analyzed by the ΔΔCt method [47]. Standard error propagation was used to combine measurement errors of the variables. The qPCR results were reported in ratio of mature mRNA amount in the Egr1 morphants to that in the controls (2∧-ddCt) and the corresponding range in 2∧(-(ddCt ± ddCtErr)). The results are also plotted in Figure 2G. An alpha level of 0.05 was used for all statistical tests.