Tfap2a and 2b are genetically downstream of the Foxn4-Ptf1a pathway
To explore the molecular basis by which Ptf1a controls amacrine and horizontal cell development, we carried out RNA-seq analysis to identify genes differentially expressed in Ptf1a mutant retinas. RNA was extracted from Ptf1a
+/+ and Ptf1a
Cre/Cre retinas at E14.5 when amacrine and horizontal cells are being born and Ptf1a function is required. This analysis yielded 224 genes whose expression level is downregulated or upregulated by 2-fold or more in the mutant retina (Figure 1A, B; Additional file 1: Table S1). These include genes encoding transcription factors, G-protein coupled receptors, kinases and transporters, etc. (Figure 1C). Consistent with the crucial role of Ptf1a in retinal development, we found that the differentially expressed genes are enriched with GO (Gene Ontology) terms such as positive regulation of neurogenesis, nervous system development, tissue development, cellular component morphogenesis, response to extracellular stimulus, transcription factor activity, and so on (Figure 1D).
Among the genes differentially expressed in Ptf1a null mutant retinas, transcription factor genes constitute one of the largest functional groups (Figure 1C). These include Tfap2a and Tfap2b, which are downregulated by 4- and 13-fold, respectively (Figure 1E; Figure 2A). Similarly, our previous microarray data show that these two genes are downregulated by 4- and 9-fold, respectively, in the E14.5 Foxn4 null retina [29]. To confirm the RNA-seq data, we measured RNA levels of these two genes in E14.5 wild type and Ptf1a
Cre/Cre retinas by semi-quantitative RT-PCR, and found that there was a dramatic decrease in Tfap2a and 2b transcripts in the null retina compared to the control (Figure 2B). In addition, we examined Tfap2a and 2b protein expression levels by immunofluorescence using two antibodies, one of which cross-reacts with both proteins and the other is specific to Tfap2b. Either antibody barely detected any Tfap2a/2b-expressing cells in E16.5 and P0 Ptf1a null retinas despite plenty of them present in the control retina (Figure 2C-J).
During mouse retinal development, Tfap2b expression is barely detectable in E12.5 retinas but found in cells scattered within the central region of E13.5 retinas (Additional file 2: Figure S1A, B). From E14.5 to early postnatal stages, Tfap2b-expressing cells gradually concentrate into the presumptive inner nuclear layer (INL) (Additional file 2: Figure S1C, D; Figure 2G, I). In late postnatal and mature retinas, Tfap2b is expressed in numerous amacrine cells located within the inner half of the INL, in all horizontal cells residing at the outer plexiform layer, as well as in a subset of cells in the ganglion cell layer (GCL) (Additional file 2: Figure S1F). Moreover, Tfap2a and 2b are colocalized in most of these cells although the expression of Tfap2a is rather weak in horizontal cells (Additional file 2: Figure S1E-G). A similar spatiotemporal expression pattern was previously reported for Tfap2a during mouse retinal development [26]. However, unlike Tfap2a and 2b, our previous study indicates that Ptf1a expression is limited to retinal precursor cells as its expression is transient, present only in the outer neuroblastic layer throughout development, and has an onset time of E12.5 [15]. These results combined with the RNA-seq data thus suggest that Tfap2a and 2b may function genetically downstream of the Foxn4-Ptf1a pathway and have a role in amacrine and horizontal cell development.
Ptf1a induces Tfap2b expression and promotes the amacrine and horizontal cell fates
The drastic downregulation of Tfap2a and 2b expression in Ptf1a null retinas suggest that Ptf1a may act upstream of these two genes to activate their expression. We tested this possibility by overexpressing Ptf1a in the mouse retina using the pCIG expression vector carrying a GFP reporter [30-32]. pCIG-Ptf1a and pCIG plasmid DNA (Additional file 3: Figure S2A) was injected into the subretinal space of newborn mice and electroporated into the retina. At P12, we found that forced Ptf1a expression induced 2-fold more Tfap2b + cells in retinas transfected with the pCIG-Ptf1a plasmid than in the control retina (Figure 3G, H, Y). Given that Ptf1a is able to induce Tfap2a expression in the chick retina [18], it appears that the expression of both Tfap2a and 2b may be under positive regulation by Ptf1a.
We further analyzed the laminar position and morphology of GFP+ cells in transfected retinas at P12. In retinas transfected with pCIG-Ptf1a DNA, the fraction of GFP+ cells differentiated as photoreceptors in the ONL (outer nuclear layer) dropped from 81.8% in the control retina to 63.5% (Additional file 3: Figure S2B-D). The percentage of GFP+ cells located in the outer half of the INL (inner nuclear layer) also decreased from 10.0% in the control to 1.3% (Additional file 3: Figure S2B-D). In contrast, the proportion of GFP+ cells distributed within the inner half of the INL dramatically increased from 8.3% in the control to 35.2% (Additional file 3: Figure S2B-D). Thus, Ptf1a misexpression substantially changes the proportions of progeny distributed in different retinal cell layers.
The increased GFP+ cells in the INL of retinas transfected with the pCIG-Ptf1a plasmid displayed an amacrine cell morphology (Additional file 3: Figure S2C). Indeed, we found that ectopically expressed Ptf1a obviously increased the number of GFP+ cells immunoreactive for Pax6, GLYT1, calretinin, or TH (tyrosine hydroxylase), all proteins expressed in amacrine cells (Figure 3A-F, O, P). Quantification of colocalized cells revealed that forced Ptf1a expression dramatically increased the percentage of Pax6+ cells from 6.6% to 31.9%, GLYT1+ cells from 4.0% to 27.4%, calretinin + cells from 3.4% to 31.4%, and TH+ cells from 0.0% to 13.0% (Figure 3Y). Furthermore, smaller increase was observed in the number of GFP+ cells immunoreactive for GABA, Gad65, Gad67 and calbindin in retinas transfected with pCIG-Ptf1a DNA (Figure 3I-N, Q, R, Y). On the other hand, misexpressed Ptf1a decreased the percentage of GFP+ photoreceptor cells immunoreactive for recoverin from 81.8% to 63.5%, GFP+ bipolar cells immunoreactive for Chx10 from 6.7% to 1.1%, and Müller glial cells immunoreactive for GS (glutamine synthetase) from 3.3% to 0.0% (Figure 3S-Y). These data suggest that Ptf1a is able to promote the differentiation of all kinds of amacrine cells including glycinergic, GABAergic and dopaminergic neurons at the expense of photoreceptor, bipolar and Müller cells.
To determine the effect of misexpressed Ptf1a on development of horizontal and ganglion cells, which are born at embryonic stages, we used a replication-incompetent murine retroviral vector that carries a GFP reporter [33] to mediate Ptf1a overexpression. E13.5 retinal explants were infected with Ptf1a-GFP or Control-GFP viruses (Figure 4A), and the infected retinas were harvested after 4.5 days in culture to analyze horizontal and ganglion cells or collected after 12.5 days in culture for analysis of other cell types. We found that misexpressed Ptf1a increased Lim1+ horizontal cells by approximately 25-fold, decreased Brn3a + ganglion cells by 11-fold, and similarly reduced Brn3b + ganglion cells (Figure 4B, I-N). It also significantly increased Pax6+, GLYT1+ and Gad67+ amacrine cells but reduced Chx10+ bipolar and GS+ Müller cells (Figure 4B-H), similar to its effect in retinas transfected at P0 (Figure 3). Therefore, these data suggest that Ptf1a has the ability to not only promote the horizontal cell fate but also suppress the ganglion cell fate, in agreement with the finding in Ptf1a null retinas [15,16].
Tfap2b is expressed in amacrine and horizontal cells and promotes amacrine cell differentiation
As a downstream transcription factor, Tfap2b may mediate in part the function of Ptf1a in amacrine and horizontal cell development. To test this possibility, we first comprehensively characterized the types and subtypes of Tfap2b-expressing cells in the mouse retina by immunofluorescence using a battery of cell type- and subtype-specific markers. Consistent with it being expressed in amacrine cells, there is extensive colocalization between Tfap2b and Pax6, syntaxin, GABA, GAD67, GAD65, Nr4a2, ChAT (choline acetyltransferase), GLYT1, Ebf, calbindin, calretinin, or TH (Figure 5A-E, G-K; Additional file 4: Figure S3A, B). Tfap2b appears also to be completely colocalized with calbindin and Lim1 in horizontal cells (Figure 5 K, L; Additional file 4: Figure S3H). There is no expression of Tfap2b in Brn3a + and Brn3b + ganglion cells, Chx10+ bipolar cells, recoverin + photoreceptors, and Sox9+ Müller cells (Additional file 4: Figure S3C-H). Consistent with this, Tfap2b is co-expressed with Bhlhb5 only in a small set of GABAergic amacrine cells since Bhlhb5 is additionally expressed in bipolar cells (Figure 5F; Additional file 4: Figure S3H). Quantification of colocalized amacrine cells revealed that the proportions of Tfap2b-expressing cells in all Pax6+, syntaxin + and GLYT1+ populations are 54.7%, 64.7% and 73.9%, respectively (Additional file 4: Figure S3H), suggesting that Tfap2b is expressed in most but not all glycinergic amacrine cells. However, Tfap2b is expressed in 100% of GABA+, GAD67+, GAD65+, ChAT+, TH+, or Nr4a2+ cells, indicating that it may be expressed by all GABAergic amacrine cells including the starburst (marked by ChAT) and dopaminergic (marked by TH) subtypes.
Tfap2b and 2a are reported to be redundantly required for horizontal cell generation but appear to be dispensable for generating amacrine cells [26,28]. We investigated whether Tfap2b has the ability to promote amacrine cell differentiation by overexpressing it in retinas of newborn mice via electroporation. In P12 transfected retinas, similar to Ptf1a (Additional file 3: Figure S2), we found that overexpressed Tfap2b increased GFP+ cells within the inner half of the INL by 2-fold while significantly diminishing those in the ONL and slightly reducing those in the outer half of the INL (Additional file 5: Figure S4). There was an overt increase in the number of GFP+ cells immunoreactive for Pax6 and GLYT1 in retinas transfected with the pCIG-Tfap2b plasmid compared to the control (Figure 6A-D). Misexpressed Tfap2b increased the fraction of Pax6+ cells from 6.6% to 12.5% and GLYT1+ cells from 4.0% to 8.0% (Figure 6U). It also resulted in smaller but significant increase in the proportion of calretinin+, GABA+, Gad65+, Gad67+ and calbindin + amacrine cells while significantly reducing the fraction of recoverin + photoreceptors (Figure 6E-J, M, N, Q, R, U). However, unlike Ptf1a, misexpressed Tfap2b did not induce the differentiation of TH+ dopaminergic amacrine cells or significantly suppress the differentiation of bipolar and Müller cells (Figure 6 K, L, O, P, U; Figure 3). Thus, Tfap2b has the ability, albeit weaker than that of Ptf1a, to promote the differentiation of both glycinergic and GABAergic amacrine cells.
Tfap2a facilitates amacrine cell differentiation
As a homolog of Tfap2b, Tfap2a is expected to have a similar role during retinal cell development as Tfap2b. Indeed, we found that Tfap2a misexpressed in retinas of newborn mice significantly increased GFP+ cells distributed within the inner half of the INL and GCL whereas it significantly reduced those in the ONL (Additional file 6: Figure S5A-D). Similar to Tfap2b, it caused obvious increase in the number of GFP+ cells immunoreactive for Pax6 and GLYT1 in transfected retinas (Additional file 6: Figure S5E-H), and smaller increase in the number of GABA+ and Gad65+ GABAergic amacrine cells (Additional file 6: Figure S5I-L), but had no effect on the formation of TH+ dopaminergic neurons (Additional file 6: Figure S5M, N). Thus, Tfap2a may act similarly as Tfap2b to facilitate the differentiation of both glycinergic and GABAergic amacrine cells during retinogenesis.
Tfap2a and 2b are required for amacrine cell differentiation
To determine whether Tfap2a and 2b are necessary for amacrine cell differentiation, we sought to simultaneously knock down Tfap2a and 2b expression in retinal precursors. To this end, we screened for target sequences (oligonucleotides) in these two genes that are effective in knocking down the expression of Tfap2a or Tfap2b by shRNA-mediated interference. A Tfap2a shRNA (Tfap2ai5) expressed from the RNAi vector pU6 [34] was found to dramatically reduce GFP expression in HEK293 cells co-transfected with the pCIG-Tfap2a expression plasmid (containing a Tfap2a-IRES-GFP cassette) (Additional file 7: Figure S6A-D). Similarly, we identified a Tfap2b shRNA (Tfap2bi4) that was effective and specific in knocking down Tfap2b expression in cell culture (Additional file 7: Figure S6A, E-G).
To investigate whether simultaneous knockdown of both Tfap2a and 2b expression has any functional consequence, we co-electroporated Tfap2a and 2b shRNAs or pU6 plasmid with the pCIG vector into P0 mouse retinas and collected them at P12 for analysis. The proportions of GFP+ progeny distributed in different retinal cell layers were quantified. In retinas transfected with both shRNAs (Tfap2ai5 + bi4), compared to the control, the proportion of GFP+ cells distributed to the inner half of the INL decreased significantly from 8.9% to 3.3% (Additional file 8: Figure S7); whereas the ratio of GFP+ cells differentiated as photoreceptors in the ONL increased from 81.3% to 87.5% (Additional file 8: Figure S7). No change was seen in the proportion of GFP+ cells distributed to the outer half of the INL (Additional file 8: Figure S7). The significant reduction of cells in the INL is consistent with the idea that Tfap2a and 2b are redundantly required for proper differentiation of amacrine cells, which normally reside in the inner half of the INL.
Transfection of Tfap2a and 2b shRNAs led to more than 3-fold decrease in the fraction of GFP+ cells immunolabeled by the antibody cross-reacting with both Tfap2a and 2b (Figure 7S-U), demonstrating the effectiveness of the double knockdown strategy. Consistent with decreased GFP+ cells in the INL, it caused a significant reduction of amacrine cells immunostained by various molecular markers including Pax6, GLYT1, calretinin, GABA, GAD65, GAD67, and calbindin (Figure 7A-L, U). For instance, in retinas transfected with both shRNAs, the proportion of Pax6+, GLYT1+, GABA+ and GAD65+ cells decreased from 6.8%, 4.2%, 2.0% and 1.2% to 2.1%, 0.9%, 0.6% and 0.2%, respectively (Figure 7U). By contrast, misexpression of Tfap2a and 2b shRNAs increased the percentage of recoverin + photoreceptors from 81.9% to 87.1% and had no effect on the differentiation of Ch10+ bipolar cells and GS+ Müller cells (Figure 7 M-R, U). These results thus indicate that Tfap2a and 2b factors are not only sufficient but also necessary for promoting the differentiation of glycinergic and GABAergic amacrine cells during retinal development.