Prdm13 Regulates Subtype Specification of Retinal Amacrine Interneurons and Modulates Visual Sensitivity

Abstract

Amacrine interneurons, which are highly diversified in morphological, neurochemical, and physiological features, play crucial roles in visual information processing in the retina. However, the specification mechanisms and functions in vision for each amacrine subtype are not well understood. We found that the Prdm13 transcriptional regulator is specifically expressed in developing and mature amacrine cells in the mouse retina. Most Prdm13-positive amacrine cells are Calbindin- and Calretinin-positive GABAergic or glycinergic neurons. Absence of Prdm13 significantly reduces GABAergic and glycinergic amacrines, resulting in a specific defect of the S2/S3 border neurite bundle in the inner plexiform layer. Forced expression of Prdm13 distinctively induces GABAergic and glycinergic amacrine cells but not cholinergic amacrine cells, whereas Ptf1a, an upstream transcriptional regulator of Prdm13, induces all of these subtypes. Moreover, Prdm13-deficient mice showed abnormally elevated spatial, temporal, and contrast sensitivities in vision. Together, these results show that Prdm13 regulates development of a subset of amacrine cells, which newly defines an amacrine subtype to negatively modulate visual sensitivities. Our current study provides new insights into mechanisms of the diversification of amacrine cells and their function in vision.

Introduction

The vertebrate retina is a part of the CNS and responsible for the first step of vision. Amacrine cells are retinal interneurons that modulate visual information before transmission to ganglion cells enabling elaborate visual functions (Gollisch and Meister, 2010; Masland, 2012; Balasubramanian and Gan, 2014). Similarly to interneurons in other CNS regions such as the cortex, hippocampus, and spinal cord (Klausberger and Somogyi, 2008; Cossart, 2011), amacrine cells are classified into many different subtypes on the basis of morphological, neurochemical, and physiological profiles. Previous morphological analyses of amacrine cells revealed the existence of ∼30 subtypes that are distinctive in arborization size and projection location for neurites (MacNeil and Masland, 1998; MacNeil et al., 1999). Whereas ∼85% of amacrine cells use GABA or glycine as a neurotransmitter (Grünert and Wässle, 1993; Menger et al., 1998; Kay et al., 2011), there are also amacrine cells releasing glutamate, acetylcholine, and dopamine, among others (Iuvone, 1986; Kolb, 1995; Haverkamp and Wässle, 2000). Moreover, some amacrine cells release neuromodulators, including neuropeptide Y (NPY), nitric oxide (NO), and substance P (Haverkamp and Wässle, 2000).

Transcription factors are known to play essential roles in amacrine cell development and diversification. In previous reports, Foxn4 and Rorβ1, which are expressed in retinal progenitor cells, regulate amacrine and horizontal cell fate (Li et al., 2004; Liu et al., 2013). Ptf1a, a common downstream factor of Foxn4 and Rorβ1, is also essential for amacrine and horizontal cell fate specification (Fujitani et al., 2006). Loss of the Barhl2 produces a reduction of GABAergic and glycinergic amacrine cells (Mo et al., 2004; Ding et al., 2009). Islet1 homeoprotein deficiency results in a reduction of cholinergic amacrine cells (Elshatory et al., 2007b). Bhlhb5 deficiency leads to reduction of GABAergic amacrine cells (Feng et al., 2006), and Neurod6 overexpression induces the nGnG amacrine cells, which are neither GABAergic nor glycinergic (Kay et al., 2011). Nr4a2 (Nurr1) is known to specify a subset of GABAergic amacrine cells, including TH-positive amacrine cells (Jiang and Xiang, 2009). However, considering that there are >30 different subtypes of amacrine cells in the rodent retina, there still remains much to be clarified for our understanding of the specification mechanism of each amacrine cell subtype and its function in vision.

We previously reported that rod and cone cell fates were converted to those of amacrine-like cells in Otx2 conditional knock-out (CKO) mouse retinas (Nishida et al., 2003; Sato et al., 2007). We hypothesized that transcripts from various genes important for amacrine cell development were relatively upregulated in the Otx2 CKO retina compared with those of the wild-type (WT) retina (Omori et al., 2011). We found that Prdm13 is a highly upregulated gene in Otx2 CKO retinas. In the current study, we identified Prdm13 (PR domain containing 13) as a regulator of amacrine subtype specification in the mouse retina. We found that the majority of Prdm13-positive amacrine cells express calcium-binding proteins, Calbindin and Calretinin (also known as CALB1 and CALB2 respectively, herein called CALBs) in the mouse retina. Prdm13-null (Prdm13^−/−) retinas showed a specific reduction in the number of CALBs-positive GABAergic or glycinergic amacrine cells. Interestingly, Prdm13^−/− retinas exhibited a specific defect of the S2/S3 border neurite bundle labeled by anti-CALBs antibodies in the inner plexiform layer (IPL). Finally, in optokinetic responses (OKRs), we observed elevated spatial, temporal, and contrast sensitivities in Prdm13^−/− mice. Together, our current study reveals an essential role of Prdm13 in amacrine subtype specification in the retina and formation of proper visual acuity.

Materials and Methods

Animal care.

All procedures conformed to the Association for Research in Vision and Ophthalmology statement for the Use of Animals in Ophthalmic and Vision Research, and these procedures were approved by the Institutional Safety Committee on Recombinant DNA Experiments (approval ID 3380-3) and Animal Experimental Committees of the Institute for Protein Research (approval ID 24-05-1), Osaka University, and were performed in compliance with the institutional guidelines. Mice were housed in a temperature-controlled room at 22°C with a 12 h light/dark cycle. Fresh water and rodent diet were available at all times. All animal experiments were performed on mice of either sex.

In situ hybridization.

In situ hybridization was performed as described previously (Sanuki et al., 2011). Mouse embryos and eye cups were fixed by 4% PFA in PBS overnight on ice. Digoxigenin-labeled riboprobes for mouse Prdm13 and Nr4a2 were generated by in vitro transcription using 11-digoxigenin UTPs (Roche). Prdm13 and Nr4a2 cDNA fragments were obtained by RT-PCR. Primer sequences used to synthesize radiolabeled probes are as follows: for the Prdm13 in situ hybridization probe, forward, 5′-GAATTCATGTACATCTGCTGGTACTGCTGGAGGA-3′ and reverse, 5′-CTCGAGTTAGGAGTCGTGCTCGCCACGACC-3′ and for the Nr4a2 cDNA fragment, forward, 5′-CTCGAGATGCCTTGTGTTCAGGCGCAGTAT-3′ and reverse, 5′-GATATCTTAGAAAGGTAAGGTGTCCAGGAA-3′.

Northern blot analysis.

Northern blot analysis was performed as described previously (Sanuki et al., 2011). Total RNAs were extracted from the mouse retina at P0, P6, P9, P14, and P21. A total of 10 μg of total RNA was electrophoresed on a 1.0% agarose formaldehyde gel and transferred to a nylon membrane (Pall). The fragment (nucleotides 556–2265 in NM_001080771.1) of the Prdm13 cDNA obtained by PCR using the full-length Prdm13 cDNA was used to synthesize radiolabeled probes. The cDNA probe was labeled with ³²P-dCTP using the Rediprime II random prime labeling system (GE Healthcare).

Immunohistochemistry.

Immunohistochemistry was performed as described previously (Muranishi et al., 2011). Mouse embryos and eye cups were fixed by 4% PFA in PBS for 30 min at room temperature or on ice. The tissues were then rinsed in PBS, cryoprotected with 30% sucrose in PBS, embedded in TissueTec OCT compound 4583 (Sakura), frozen, and sectioned. Frozen 16 μm sections on slides were dried for 30 min at room temperature, rehydrated in PBS for 5 min, incubated with blocking buffer (4% normal donkey serum, and 0.1% Triton X-100 in PBS) for 1 h, and then with primary antibodies overnight at 4°C. Slides were washed with PBS three times for 10 min each time and incubated with secondary antibodies for 2 h at room temperature. For immunostaining of the whole retina, each retina was gently peeled off from the sclera, rinsed in PBS, and fixed with 4% PFA (w/v) in PBS for 1.5 h. The retinas were permeabilized by incubation in 0.1% Triton X-100 in PBS for 30 min. The samples were blocked with 4% donkey serum in 0.1% Triton X-100 in PBS for 1 h. The retinas were then incubated with primary antibodies at 4°C overnight. The samples were washed with PBS three times 10 min each time and incubated with a fluorescent dye-conjugated secondary antibody overnight at 4°C. The specimens were observed under a laser confocal microscope (LSM700, Carl Zeiss). Primary antibodies used in this study are described as follows: anti-Rhodopsin (Sigma, O4886, mouse, 1:10,000), anti-S-opsin (Santa Cruz Biotechnology, sc14363, goat, 1:500), anti-Calbindin (Sigma, PC253L, rabbit, 1:1000), anti-Calretinin (Millipore, AB5054, rabbit, 1:1000), anti-Chx10 (Koike et al., 2007, rabbit, 1:200), anti-Pax6 (DSHB, mouse, 1:100), anti-Brn3a (Millipore, MAB1585, mouse, 1:100), anti-Brn3b (Santa Cruz Biotechnology, sc-6026, goat, 1:100), anti-S100β (Sigma, S-2532, mouse, 1:100), anti-Gad65/67 (Millipore, AB1511, rabbit, 1:1000), anti-Gad65 (Sigma, G-4913, rabbit, 1:1000), anti-Glyt1 (Millipore, AB1770, goat, 1:1000 or Zafra et al., 1995, rabbit, 1:500), anti-TH (Millipore, MAB5280, mouse, 1:1000), anti-AP-2α (Novus Biological, NBP1-95386, rabbit, 1:1000), anti-AP-2γ (Sigma, A3108, mouse, 1:1000), anti-Islet1 (DSHB, 39.4D5, mouse, 1:1000), anti-BrdU (Becton Dickinson, 347580, mouse, 1:100), anti-GFP (Nacalai, 04404-26, rat, 1:1000), anti-GFP (Clontech, 632460, rabbit, 1:1000), anti-Syntaxin (Sigma, S-0664, mouse, 1:1000), anti-Vglut3 (Millipore, AB5421, guinea pig, 1:1000), anti-Calsenilin (Millipore, 05-756, mouse, 1:100), anti-active Caspase-3 (Promega, G748A, rabbit, 1:500), anti-phospho-Histone H3 (Millipore, 06-570, rabbit, 1:100), anti-Ki67 (BD Pharmingen, 556003, mouse, 1:500), anti-Lhx2 (Santa Cruz Biotechnology, sc-19342, goat, 1:200), and anti-Prdm13 (current study, guinea pig, 1:1000).

S-phase labeling.

The S-phase labeling experiment was performed as described previously (Muranishi et al., 2011). Timed pregnant mice (ICR, E15.5) were injected intraperitoneally with BrdU (50 μg/g body weight). Embryos were harvested 1 h after BrdU injection and fixed with 4% PFA for 30 min on ice. For BrdU immunostaining, sections were pretreated with 2N HCl for 15 min at 37°C before blocking.

Generation of Prdm13 flox mouse.

We subcloned an ∼10 kb Prdm13 genomic fragment, inserted one loxP site into intron 1 and another loxP site into intron 3, cloned it into a modified pPNT vector to make a targeting construct, and transfected the linearized targeting construct into the JM8A3 ES cells (Pettitt et al., 2009). The culture, electroporation, and selection of ES cells were performed as described previously (Muranishi et al., 2011). ES cells that were heterozygous for the targeted gene disruption were microinjected into ICR (Charles River) blastocysts to obtain chimeric mice. We obtained Prdm13 conventional knock-out (KO) mice by crossing a Prdm13 flox mouse line with the CAG-Cre transgenic mouse line, which expresses Cre recombinase driven under the CAG promoter (Sakai and Miyazaki, 1997). We also obtained retinal progenitor CKO mice of Prdm13 by crossing the Prdm13 flox mouse line with the Dkk3-Cre transgenic mouse line expressing Cre in retinal progenitors (Sato et al., 2007). The rd8 mutation carried in the C57BL/6N strain (Mattapallil et al., 2012) was removed by outcrossing the mice with the C57BL/6J strain before the mice were used for experiments. Primer sequences for genotyping are as follows: for genotyping of Prdm13 flox mouse, forward, 5′-AGCCTAGAGCTCATAAGCGAATGG-3′ and reverse, 5′-ATAGGACAGGAAGGACCCATGAAG-3′; genotyping of Prdm13-null mouse, forward, 5′-CACCTCAGTCTTTGCCTTCCTTGCAA-3′, reverse 1, 5′-CTACAACTGCGACTCCAACGCATGAT-3′ and reverse 2, 5′-CAGAGAAAGAGTACCCTTGTGCCT-3′.

Western blot analysis.

P14 mouse retinas were lysed in a SDS-sample buffer. Samples were resolved by SDS-PAGE and transferred to a PVDF membrane. The membrane was blocked with blocking buffer (5% skim milk (w/v), and 0.1% Tween 20 in TBS) and was incubated with an anti-Prdm13 antibody that we raised. The membrane was washed with 0.1% Tween 20 in TBS and incubated with an HRP-conjugated goat antibody against guinea pig IgG (1:10,000, 706-035-148, Jackson ImmunoResearch Laboratories). For the secondary immunoreaction, the PVDF membrane was incubated with WB Stripping Solution (Nacalai) to remove antibodies for the first reaction, and blocked again with 5% skim milk (w/v) in TBS. Further immunoblots were performed using a mouse antibody against β-actin (1:500, A5316, Sigma). After washing, the membrane was incubated with an HRP-conjugated goat antibody against mouse antibody IgG (1:10,000, 81-6520, Zymed). β-actin was used as a loading control.

Toluidine blue staining.

Retinal sections were rinsed with PBS then stained with 0.1% toluidine blue in PBS for 1 min. After washing with PBS three times for 1 min each time, slides were coverslipped and immediately observed under the microscope.

Plasmid constructs.

Full-length cDNA fragments of mouse Prdm13 (GenBank accession no. NM_001080771.1) and Ptf1a (NM_018809.2) were amplified by PCR using retinal cDNAs from E17 mouse embryos and cloned into the pCRII-TOPO vector (pCRII-TOPO-mPrdm13, Invitrogen). Full-length Prdm13 and Ptf1a were digested with EcoRI and XhoI and subcloned into the pCIG vector (Matsuda and Cepko, 2004). The pCIG vector contains an IRES-EGFP sequence to express EGFP in a bicistronic manner. Prdm13 mutants, named PR domain, zinc finger, and zinc finger mutant, were amplified by PCR using pCRII-TOPO-mPrdm13 as a template, cloned into pCRII-TOPO (Invitrogen), and ligated into the pCIG vector digested with EcoRI and XhoI. All cDNA clones obtained by PCR in the current study were validated by sequencing. Primer sequences are as follows: for a full-length Prdm13, forward, 5′-GAATTCATGCCCGCCCACGTCACTCCGCGGA-3′ and reverse, 5′-CTCGAGTTAGGAGTCGTGCTCGCCACGACC-3′; Prdm13 PR-domain, forward, 5′-GAATTCATGCCCGCCCACGTCACTCCGCGGA-3′ and reverse, 5′-CTCGAGTCAGTAGCGCTCCTCCCCTTTCTCATCA-3′; Prdm13 zinc finger domain, forward, 5′-GAATTCATGTACATCTGCTGGTACTGCTGGAGGA-3′ and reverse, 5′-CTCGAGTTAGGAGTCGTGCTCGCCACGACC-3′; Prdm13 zinc finger mutant, forward 1, 5′-TAACAGCCTCAAGGCACACCTGCGGTTCGCCTGCGTCCTC-3′ and reverse 1, 5′-GGGTAGCGAAAGGTCCTCCAGCAGTACCAGGCGATGTAGC-3′, forward 2, 5′-CGCGCAAGTACGGGCTCAAGATCGCCATGCGGACGCACAC-3′ and reverse 2, 5′-AGTACAGCTTGCCACAGTAGAGGGCCAAGTGGCCTGTCTT-3′, forward 3, 5′-GCGACCCCAGCAATCTCAACAAGGCCATCCGACTCCACGC-3′ and reverse 3, 5′-CGAAGGGCCGCAGGCACACACTTTGGCCTTGAGCGGCTTGTA-3′, forward 4, 5′-CCAGCCACCTTGCCGGCGAACTCGCAGCGGTAGGGCGTATT-3′ and reverse 4, 5′-TGCGCCGCCGAGACCTGGAGCGTGCCGTCAAGTCCCGCCA-3′; a full-length Ptf1a, forward, 5′-GAATTCATGGACGCCGTACTCCTGGAGCACTT-3′ and reverse, 5′-CTCGAGTCAGGAACACAAACTCAAAGGGTGG-3′.

In vivo electroporation.

In vivo electroporation was performed on the P0 mouse retina as described previously (Matsuda and Cepko, 2004). Plasmids in 0.4 μl of PBS at a concentration of 5 μg/μl were injected into P0 mouse retinas followed by in vivo electroporation. The electroporated retinas were harvested at P6 and were immunostained using antibodies against GFP, Rhodopsin, Chx10, and Pax6.

ERG recording.

ERGs were recorded with a white LED luminescent electrode placed on the cornea (PuREC). One-month-old mice were dark adapted overnight and anesthetized with an intraperitoneal injection of 100 mg/kg ketamine and 10 mg/kg xylazine, and then the mice were placed on a heating pad. The mice were stimulated with an LED flash. Four levels of stimulus intensities ranging from −4.0 to 1.0 log cd-s/m² were used for the scotopic ERG recordings, and four levels of stimuli ranging from −0.5 to 1.0 log cd-s/m² were used for the photopic ERG recordings. After mice were light adapted for 10 min, the photopic ERGs were recorded on a rod-suppressing white background of 1.3 log cd-s/m². Eight responses were averaged for scotopic recordings (−4.0 and −3.0 log cd-s/m²). Sixteen responses were averaged for photopic recordings. Oscillatory potentials were extracted from the brightest flash scotopic and photopic ERGs by bandpass filtering between 50 and 170 Hz (Heynen et al., 1985; Wachtmeister, 1998).

OKR recording.

OKR recording was performed as previously described in detail (Sugita et al., 2013). Data were collected from WT mice and Prdm13^−/− mice weighing 17–26 g. Each mouse was previously implanted with a head holder, which allowed the head to be fixed in the stereotaxic position during the experiments. Eye movement of the right eye of each mouse was measured using a video-based eye tracker system (Geteye, Matsuura-Denko-sha). The visual stimuli were moving sinusoidal gratings with one of five spatial frequencies selected randomly from a lookup table: 0.0313, 0.0625, 0.125, 0.25, and 0.5 cycle/deg in a given trial. The temporal frequency was selected from 0.1875, 0.375, 0.75, 1.5, 3, 6, 12, or 24 Hz. To compare the contrast sensitivity, we used contrast of 1%, 2%, 4%, 8%, 16%, 32%, 64%, and 96% (sinusoidal grating; mean luminance, 100 cd/m²). We set the spatial frequency and temporal frequency to 0.125 cycle/deg and 1.5 Hz, respectively. We calculated the mean eye velocity for initial phase.

Antibody production.

A cDNA encoding a C-terminal portion of mouse Prdm13 (685–754 amino acids) was amplified by PCR and subcloned into pGEX4T-1 (GE Healthcare). The fusion protein was expressed in Escherichia coli strain BL21 and purified with glutathione Sepharose 4B (GE Healthcare) according to the manufacturer's instructions. An antiserum against Prdm13 was obtained by immunizing guinea pigs with the purified GST-Prdm13. Antibodies against GST in the antiserum were removed by incubation with the GST-coupled CNBr-activated Sepharose4B (GE Healthcare). Then, antibodies against C-terminal Prdm13 were affinity purified with GST-Prdm13-coupled CNBr-activated Sepharose4B column.

Statistical analysis.

Data are presented as mean ± SEM in ERG analysis and ± SD in the other analyses. Statistical significance was calculated with Student's t test, and p < 0.05 was taken to be statistically significant.

Results

Prdm13 is expressed specifically in a subset of amacrine cells

To examine the expression pattern of Prdm13 in the developing mouse retina, we performed in situ hybridization using retinas at different developmental stages from embryonic day 11.5 (E11.5) to postnatal day 21 (P21). Prdm13 expression began at E12.5 in the mouse retina and was observed in the neuroblastic layer (NBL) through the embryonic stages (Fig. 1A). Prdm13-positive cells were gradually stacked in the inner nuclear layer (INL) from P6 to P21 and were confined to the innermost INL where amacrine cells are localized. This result suggests that Prdm13 is predominantly expressed in amacrine cells. Northern blot analysis using retinal RNAs showed that Prdm13 expression is gradually reduced from P9 but is maintained into adulthood (Fig. 1B). To further investigate which retinal cell types express Prdm13, we raised an antibody against mouse Prdm13 and performed coimmunostaining with an anti-Pax6 antibody (a pan-amacrine cell marker in the INL) or an anti-Brn3a/b antibody (a ganglion cell marker) on the mouse retina at P14 when all the retinal cell types have completed differentiation (Cepko et al., 1996). Consistent with the observation in in situ hybridization (Fig. 1A), Prdm13 signals were detected predominantly in the INL and all of the Prdm13⁺ cells were Pax6-positive (Fig. 1C). Prdm13⁺ cells accounted for 32.0 ± 2.5% of the Pax6⁺ cell population in the INL at P14 (Fig. 1C), and this ratio was similar with that at P9 when the Prdm13 expression level was much higher (31.7 ± 1.0%; Fig. 1D). We also observed that Prdm13 is expressed in a small population of cells in the ganglion cell layer (GCL) where ganglion cells and displaced amacrine cells are localized (Fig. 1C,E, arrowheads). Because Prdm13⁺ cells in the GCL did not express Brn3a/b (Fig. 1E), these cells are very likely to be displaced amacrine cells. These results showed that Prdm13 is expressed specifically in a subset of amacrine cells. To further characterize Prdm13⁺ cells, we immunostained P14 mouse retinas with antibodies against Gad65, Glyt1, AP-2α, AP-2γ (also known as Tcfap2a and 2c, respectively), Islet1, Calbindin, and Calretinin (Fig. 1F–M). Gad65 and Glyt1 are markers for GABAergic and glycinergic amacrine cells, respectively. We observed coexpression of Prdm13 with Gad65 or Glyt1 (13.5 ± 2.0% and 87.1 ± 1.4%, respectively; Fig. 1F,G,N). AP-2α and AP-2γ are transcription factors expressed in GABAergic and glycinergic amacrine cells and associated with amacrine cell development (Bassett et al., 2012). We found that subsets of Prdm13⁺ amacrine cells were AP-2α⁺ or AP-2γ⁺ cells (41.2 ± 3.7% and 47.1 ± 1.7%, respectively; Fig. 1H,I,N). Islet1 is a marker for cholinergic amacrine cells, called starburst amacrine cells, and involved in direction selectivity (Elshatory et al., 2007a). We observed that none of the Prdm13⁺ amacrine cell signals was colocalized with Islet1 signals (Fig. 1J), indicating that Prdm13 is not expressed in starburst amacrine cells. CALBs label horizontal cells, a type of ganglion cell and a subset of amacrine cells, including starburst amacrine cells and other amacrine subtypes containing GABA, NO, and/or substance P in the mouse retina (Haverkamp and Wässle, 2000; Mojumder et al., 2008). In rabbits and primates, AII amacrine cells, which transmit inputs from rod photoreceptor cells to cone bipolar cells, express CALBs, but in rodents AII amacrine cells do not express CALBs (Wässle et al., 1995; Volgyi et al., 1997; Gábriel and Witkovsky, 1998).

Figure 1.

Prdm13 is specifically expressed in a subset of amacrine cells in the mouse retina. A, In situ hybridization of Prdm13 using mouse retinas at developing (E11.5-P9) and mature (P21) stages. B, Northern blots of Prdm13 transcripts using developing (P1–P14) and mature (P21) mouse retinal total RNAs. The arrow indicates an ∼3.2 kb Prdm13 full-length mRNA. Bottom, Ethidium bromide (EtBr) staining of the RNAs. 18S and 28S rRNAs are indicated. C–M, Prdm13 localization in the mouse retina. Mouse retinas at P14 and P9 were immunostained with antibodies against Prdm13 (magenta) and pan-amacrine, ganglion cell, or amacrine subtype markers (green): Pax6 (a pan-amacrine cell marker, C, D), Brn3a/b (ganglion cell markers, E), Gad65 (F), Glyt1 (G), AP-2α (H), AP-2γ (I), Islet1 (J), Calbindin (K), Calretinin (L), and Calbindin plus Calretinin (CALBs, M). Arrowheads indicate Prdm13⁺ displaced amacrine cells and Prdm13⁺/Gad65⁺ amacrine cells (C, E, F, respectively). Insets, INL and GCL areas with high magnification (C, F, G, H, I, and E, respectively). N, The percentages of each amacrine subtype marker⁺ cells of Prdm13⁺ cells in the INL. Error bars indicate SD from the means of 3 or 4 different mice. O, Labeling of S-phase proliferating cells with an anti-BrdU antibody (green) in mouse retinas at E15.5. Prdm13 and BrdU double-positive cells were rarely observed (arrowhead). Nuclei were stained by Hoechst (blue signals). Scale bars, 50 μm.

Calbindin⁺ or Calretinin⁺ amacrine cells made up 33.5 ± 3.7% and 65.1 ± 3.9% of Prdm13⁺ amacrine cells, respectively (Fig. 1K,L,N). Total CALBs⁺ amacrine cells were 72.2 ± 5.0% of Prdm13⁺ amacrine cells (Fig. 1M,N). We therefore found that the major population of Prdm13⁺ amacrine cells expresses CALBs. Finally, we examined Prdm13 expression in S-phase proliferating cells in the E15.5 retina. Almost none of the Prdm13⁺ cells colocalized with BrdU⁺ cells (S-phase proliferating cells in this experiment) (Fig. 1O), suggesting that Prdm13 is expressed specifically in postmitotic cells. Together, these results showed that almost all of the Prdm13⁺ amacrine cells are GABAergic or glycinergic, and that the majority of Prdm13⁺ amacrine cells are CALBs⁺, except for starburst amacrine cells.

Targeted deletion of Prdm13 leads to a specific reduction of amacrine cells

To investigate the in vivo functions of Prdm13 in the retina, we generated Prdm13-null mutant mice by targeted gene disruption (Fig. 2A–C). We confirmed the loss of Prdm13 expression in the Prdm13^−/− retinas by immunostaining and Western blot analysis with an anti-Prdm13 antibody using retinas from WT and Prdm13^−/− mice at P14 (Fig. 2D,E). The Prdm13^−/− mice were viable and fertile, and did not show any prominent appearance abnormalities. We first examined WT and Prdm13^−/− retinas at P14 by immunohistochemistry using an anti-Syntaxin antibody, a pan-amacrine marker. We found that amacrine cell numbers in the INL notably decreased in the Prdm13^−/− retina (Fig. 2F). We next examined overall retinal morphology and layer thickness by toluidine blue staining in WT and Prdm13^−/− retinas at P14. Prdm13^−/− retinas exhibited normal layer thickness of the outer nuclear layer (ONL) and GCL, but a thinner INL compared with that in the WT retina (Fig. 2G). We measured layer thickness of WT and Prdm13^−/− retinas (Fig. 2H). Whereas the ONL thickness was unchanged, the INL+IPL and INL thicknesses were significantly thinner in the Prdm13^−/− retina (Fig. 2H). We then immunostained major cell types in WT and Prdm13^−/− retinas at P14 (Fig. 2I–O). In Prdm13^−/− retinas, the amacrine cell number decreased to approximately half of that in WT retinas (Fig. 2O).

Figure 2.

Loss of Prdm13 leads to a specific reduction of amacrine cells. A, Schematic representation of the Prdm13 wild-type allele, Prdm13 recombinant allele, Flp recombinant allele, and Cre recombinant allele. Black boxes represent exons. Yellow arrowheads represent loxP sites. Blue pentagons represent FRT sites. Green and red arrows indicate primer sets to detect the Flp recombinant allele and Cre recombinant allele, respectively. Removal of the floxed region by Cre-mediated recombination is predicted to result in a translational frame shift and complete loss of Prdm13 function. B, PCR products of 98 and 200 bp were amplified from the WT and flox allele, respectively. C, PCR products of 463 and 262 bp were amplified from the WT and KO allele, respectively. D, E, Expression of the Prdm13 protein in WT and Prdm13^−/− mouse retinas at P14. WT and Prdm13^−/− retinas at P14 were immunostained with an anti-Prdm13 antibody (green) (D). The Prdm13 signals in the INL disappeared in the Prdm13^−/− retina. Western blots of Prdm13 protein in WT, Prdm13^+/−, and Prdm13^−/− mouse retinas at P14 (E). β-actin was used as a loading control. F, WT and Prdm13^−/− retinas at P14 were immunostained with an antibody against Syntaxin (green), a pan-amacrine cell marker. Insets, INL area with high magnification (F). G, H, Toluidine blue staining of WT and Prdm13^−/− retinas at P14. Thicknesses of ONL, INL, INL+IPL, and IPL were measured (H). I–O, Immunostaining of WT and Prdm13^−/− retinas at P14 with antibodies against major cell type-specific markers (green): Rhodopsin (a rod photoreceptor marker, I), S-opsin (an S-cone photoreceptor marker, J), Calbindin (a horizontal cell marker, K), Chx10 (a pan-bipolar cell marker, L), S100β (Müller glia cell marker, M), and Brn3b (a ganglion cell marker, N). Nuclei were stained by Hoechst (blue signals). The numbers of major cell types were counted in P14 WT and Prdm13^−/− retinas (O). Error bars indicate SD from the means of 3 or 4 different mice. **p < 0.01. Scale bars, 50 μm.

We further examined the retinal architecture of WT and Prdm13^−/− retinas at 5 months of age by toluidine blue staining (Fig. 3A,B). We observed that the thicknesses of the INL+IPL, INL, and IPL were significantly less than those in Prdm13^−/− mice at 5 months on a similar level with those of in Prdm13^−/− retinas at P14, except for the IPL. The ONL thickness was unaltered between WT and Prdm13^−/− retinas. We examined whether apoptosis increases in the developing Prdm13^−/− retinas compared with WT retinas at P3 and P14 by active Caspase-3 immunostaining (Fig. 3C,D). We detected no significant increase of active Caspase-3 signals in the Prdm13^−/− retina. These results suggest that amacrine cell reduction did not significantly affect the maintenance of other cell types.

Figure 3.

Histological analysis of developing and mature retinas in Prdm13^−/− mice. A, B, Toluidine blue staining of the WT and Prdm13^−/− retinas at 5 months. Thicknesses of ONL, INL+IPL, INL, and IPL were measured (B). C, The numbers of apoptotic cells were counted in WT and Prdm13^−/− retinas at P3 and P14. D, AC3 (active Caspase-3, green) is a marker for apoptotic cells. Arrowheads indicate AC3⁺ cells. Nuclei were stained with Hoechst (blue signals). Scale bars, 50 μm. Error bars indicate the SD from the means of 3 different mice. *p < 0.05. **p < 0.01.

GABAergic and glycinergic amacrine cells are reduced in Prdm13^−/− retinas

To investigate what subtypes of amacrine cells decreased in the Prdm13^−/− retina, we analyzed cell numbers of eight subtypes of amacrine cells in WT and Prdm13^−/− retinas at P14 by immunostaining. We immunolabeled GABAergic and glycinergic amacrine cells using anti-Gad65/67 and anti-Glyt1 antibodies, respectively. We found that cell numbers of both GABAergic and glycinergic amacrine cells decreased in the Prdm13^−/− retina (Fig. 4A,B,K). Consistent with a reduction of GABAergic and glycinergic amacrine cells in the Prdm13^−/− retina, both AP-2α⁺ and AP-2γ⁺ amacrine cells significantly decreased in the Prdm13^−/− retina (Fig. 4C,D,K). Islet1- and Calbindin-double positive amacrine cell populations represent cholinergic amacrine cells, including starburst amacrine cells. We did not observe any changes in cell number of cholinergic amacrine cells by counting Islet1⁺ and Calbindin⁺ amacrine cells (Fig. 4E,K). We immunostained dopaminergic and glutamatergic amacrine cells using anti-TH and anti-Vglut3 antibodies, respectively. We observed that cell numbers of dopaminergic and glutamatergic amacrine cells were unaltered in the Prdm13^−/− retina compared with those in the WT retina (Fig. 4F–I,L). Furthermore, we performed in situ hybridization on WT and Prdm13^−/− retinas at P9 using the Nr4a2 probes. Whereas the numbers of Nr4a2⁺ amacrine cells in the INL were unchanged, the numbers of Nr4a2⁺ displaced amacrine cells in the GCL significantly decreased (Fig. 4J,M). Nr4a2⁺ displaced amacrine cells are GABAergic, not TH-positive, amacrine cells (Jiang and Xiang, 2009). In summary, the Prdm13^−/− retina displayed a significant reduction in cell numbers of GABAergic and glycinergic amacrine cells of the eight subtypes we examined.

Figure 4.

Decrease of GABAergic and glycinergic amacrine cells in the Prdm13^−/− retina. A–I, Sections or whole-mounts from WT and Prdm13^−/− retinas at P14 were immunostained with antibodies against amacrine cell markers (green): Gad65/67 for GABAergic amacrine cells (A), Glyt1 for glycinergic amacrine cells (B), AP-2α and AP2γ for each different amacrine cell population including GABAergic and glycinergic cells (C, D), Islet1 (green) and Calbindin (magenta) for cholinergic amacrine cells (starburst amacrine cells) (E), TH (green) for dopaminergic amacrine cells (F, G), and Vglut3 (green) for glutamatergic amacrine cells (H, I). J, In situ hybridization of Nr4a2 in WT and Prdm13^−/− retinas at P9. Arrowheads indicate Nr4a2⁺ cells in the GCL. K–M, The numbers of eight subtypes of amacrine cells in WT and Prdm13^−/− retinas at P14 were counted. Nuclei were stained with Hoechst (blue signals). Insets, INL area with high magnification. Error bars indicate SD from the means of 3 or 4 different mice. **p < 0.01. *p < 0.05. Scale bars, 50 μm.

Prdm13 potently induces amacrine cell differentiation

To test whether Prdm13 can induce amacrine cell differentiation, we overexpressed Prdm13 in the developing mouse retina by in vivo electroporation. We generated a pCIG-Prdm13 plasmid that simultaneously expresses full-length Prdm13, driven by the CAG promoter, and EGFP in a bicistronic manner mediated by the IRES sequence (Matsuda and Cepko, 2004). The pCIG plasmid without the cDNA insert was used as a control. These plasmids were in vivo electroporated into mouse retinas at P0 when rods, bipolar cells, and amacrine cells are the neurons being generated in addition to Müller glial cells. At this stage, the proliferating retinal progenitors basically express electroporated genes. We harvested the retinas at P6 and analyzed by immunostaining using an anti-GFP antibody together with retinal neuronal type marker antibodies (an anti-Rhodopsin antibody for rods, an anti-Chx10 antibody for bipolar cells, and an anti-Pax6 antibody for pan-amacrine cells; Fig. 5A). We quantified percentages of retinal cell marker-positive cells in the EGFP-positive cells (Fig. 5B). In pCIG-transfected retinas, EGFP⁺ cells differentiated into rods, bipolar cells, or amacrine cells at a ratio of 23%–32% (Fig. 5A,B). In retinas overexpressing Prdm13, 86.1 ± 6.6% of EGFP⁺ cells were Pax6⁺, whereas Rho and Chx10 were expressed in a very small population (2.0%–5.2%; Fig. 5A,B). This result showed that Prdm13 can potently induce amacrine cell differentiation. The Prdm13 protein contains a PR-domain similar to the SET domain related to histone methyltransferase activity in the N terminus and four zinc finger domains in the C terminus. Then to determine which domain of the Prdm13 protein is responsible for this activity, we generated constructs expressing a PR domain of Prdm13 (PR domain), four zinc finger domains of Prdm13 (zinc finger domain), and a full length of Prdm13 with mutations leading to inactivation of the four zinc finger domains (zinc finger mutant) (Fig. 5C) (Chang et al., 2013). The PR domain-containing retinas showed a similar cell composition with that of control retinas (Fig. 5D,E). The zinc finger domain-transfected retinas exhibited a potent induction of amacrine cells at a similar level to that by the full length; however, this effect was not observed in the zinc finger mutant retinas (Fig. 5D,E). Together, we showed that Prdm13 can potently induce amacrine cell differentiation, and the four zinc finger domains of Prdm13 are responsible for this effect.

Figure 5.

Prdm13 potently induces differentiation of amacrine cells. A, The pCIG (empty vector) or pCIG-Prdm13 plasmids were electroporated into P0 mouse retinas. The retinas were harvested at P6, sectioned, and immunostained with antibodies against GFP (green) and three retinal cell type-specific markers (magenta): Rhodopsin (Rho, a rod photoreceptor marker), Chx10 (a pan-bipolar cell marker), or Pax6 (a pan-amacrine cell marker). The electroporated cells express EGFP mediated by the IRES sequence. Insets, INL area with high magnification. B, The numbers of retinal cell type-specific marker-positive cells of EGFP-positive cells in pCIG or pCIG-Prdm13 electroporated retinas were counted. C, Schematic representation of a full length of Prdm13 and its mutants used for in vivo electroporation into P0 mouse retinas. D, E, Retinas electroporated with pCIG, pCIG-Prdm13, or pCIG-Prdm13-mutants (pCIG-PR domain, pCIG-zinc finger domain, and pCIG-zinc finger mutant) were immunostained with antibodies against EGFP (green) and cell-type specific markers, including Pax6, Rhodopsin (Rho), or Chx10. Immunostaining images of electroporated retinas with an anti-EGFP antibody (green) and an anti-Pax6 antibody (magenta) were displayed (D). The numbers of retinal cell type-specific marker-positive cells of EGFP-positive cells were counted (E). Error bars indicate SD from the means of 3–5 different mice. **p < 0.01. Scale bars, 50 μm.

Prdm13 induces more specific amacrine subtypes than Ptf1a does

Ptf1a, a bHLH transcription factor, acts directly upstream of Prdm13 (Chang et al., 2013). The targeted disruption of Ptf1a in the mouse retina results in a complete absence of horizontal cells and a severe reduction in cell numbers of amacrine cells, indicating that the Ptf1a^−/− retina displays more severe phenotypes than the Prdm13^−/− retina. In addition, we confirmed that Prdm13 expression was totally lost in the Ptf1a^−/− (Kawaguchi et al., 2002) retina at E17.5 (Fig. 6A) (Chang et al., 2013). To examine whether Ptf1a and Prdm13 can induce amacrine cell differentiation, we performed in vivo electroporation of pCIG-Prdm13 or pCIG-Ptf1a into P0 retinas and analyzed induced amacrine subtypes by immunohistochemistry at P14. We first observed that in vivo electroporated Ptf1a strongly promoted amacrine cell differentiation (93.6 ± 8.3% of EGFP⁺ cells; Fig. 6B,C). We immunostained control and Prdm13- or Ptf1a-electroporated retinas using an anti-GFP antibody together with an anti-AP-2α antibody, an anti-AP-2γ antibody, which mainly marks GABAergic and glycinergic amacrine cells, or anti-Islet1/anti-CALBs antibodies, which mark cholinergic amacrine cells at P14. Whereas 48.5 ± 3.2% of Prdm13-electroporated cells were positive for AP-2α, 17.0 ± 2.4% of Ptf1a-electroporated cells were AP-2α⁺ (Fig. 6D,G). AP-2γ⁺ amacrine cells were induced at a similar level in both Prdm13- or Ptf1a-electroporated retinas (pCIG-Prdm13: 23.6 ± 2.0% and pCIG-Ptf1a: 25.5 ± 1.9%; Fig. 6E,G). In contrast to relatively weak AP-2α induction, Ptf1a stimulated the differentiation of Islet1⁺ and CALBs⁺ cholinergic amacrine cells much more strongly than Prdm13 did (pCIG-Prdm13: 9.6 ± 1.9%, pCIG-Ptf1a: 35.0 ± 4.7%, p < 0.01; Fig. 6F,G). These data suggested that Prdm13 preferentially induced GABAergic and glycinergic amacrine cells, but Ptf1a induced GABAergic, glycinergic, and cholinergic amacrine cells.

Figure 6.

Amacrine cell induction by Prdm13 and Ptf1a. A, WT and Ptf1a^−/− retinas at E17.5 were immunostained with antibodies against Prdm13 (green, A). Nuclei were stained with Hoechst (blue signals). B, C, Retinas electroporated with pCIG or pCIG-Ptf1a constructs at P0 were harvested at P6 and immunostained with antibodies against EGFP (green) and Pax6 (magenta). The numbers of Pax6⁺ cells (magenta) of EGFP⁺ cells (green) in the retina electroporated with pCIG or pCIG-Ptf1a were counted (C). Insets, INL area with high magnification (B). D–G, Retinas electroporated with pCIG, pCIG-Prdm13, or pCIG-Ptf1a plasmid were immunostained with antibodies against EGFP (green) and amacrine cell markers: AP-2α (D, magenta) and AP-2γ (E, magenta) for GABAergic and glycinergic amacrine cells, or Islet1 (magenta) and CALBs (blue) for cholinergic amacrine cells (F). Arrowheads indicate EGFP⁺/Islet1⁺/CALBs⁺ cells (F). The numbers of AP-2α⁺ (left bars), AP-2γ⁺ (middle bars), or Islet1⁺/CALBs⁺ (right bars) cells of EGFP⁺ cells in the retinas electroporated with each plasmid were counted (G). Insets, INL area with high magnification (D–F). Error bars indicate SD from the means of 3 different mice. **p < 0.01. H, I, WT and Ptf1a^−/− retinas at E17.5 were immunostained with antibodies against AP-2α (green, H) and Islet1 (green) plus Brn3b (magenta, I). Insets, Inner NBL area with high magnification (I). Nuclei were stained with Hoechst (blue signals, H). J, The number of Islet1⁺/Brn3b⁻ cells of Islet1⁺ cells was counted. Brn3b is a marker for ganglion cells, and Islet1 marks cholinergic amacrine cells and ganglion cells at E17.5. Islet1⁺/Brn3b⁻ cells (indicated by arrowheads) are mainly cholinergic amacrine cells. Scale bars, 50 μm. Error bars indicate SD from the means of 3 different mice. **p < 0.01.

It was reported that a very small population of amacrine cells remains in the Ptf1a^−/− retina (Fujitani et al., 2006; Nakhai et al., 2007). To confirm that loss of Ptf1a affects AP-2α⁺ and cholinergic amacrine cells in the retina, we immunostained WT and Ptf1a^−/− retinas at E17.5 with anti-AP-2α and anti-Islet1/Brn3b antibodies. We observed a complete loss of AP-2α⁺ amacrine cells and a dramatic reduction of Islet1⁺/Brn3b⁻ cells, which mark mainly cholinergic amacrine cells at E17.5, in the Ptf1a^−/− retina (Fig. 6H–J), suggesting that Ptf1a is required for the specification of both AP-2α⁺ GABAergic/glycinergic and Islet1⁺ cholinergic amacrine cells. Together, it suggested that Ptf1a regulates the subtype specification in a broader range than Prdm13 does.

Prdm13^−/− retinas display a defect of the S2/S3 border in the IPL

Amacrine cells project into the IPL where their processes densely stratify. Immunostaining of CALBs visualizes three parallel borders, dense neurite bundles of CALBs⁺ amacrine cells that contact ganglion cell dendrites and separate the IPL into the sublayers S1, S2, S3, S4, and S5 (Wässle, 2004). To examine IPL formation in the Prdm13^−/− retina, we immunostained WT and Prdm13^−/− retinas at P14 with antibodies against CALBs (Fig. 7A,B). We observed a deficiency in the border between the S2 and S3 sublayers in the IPL in the Prdm13^−/− retina (Fig. 7A,B, a layer marked with orange color bars). We also observed a significant decrease in the cell numbers of CALBs⁺ amacrine cells by ∼30% compared with the WT retina (Fig. 7C). Furthermore, to confirm that the S2/S3 border bundle in the IPL is deficient in the Prdm13^−/− retina, we coimmunostained WT and Prdm13^−/− retinas with an anti-Calretinin antibody to label the three borders and an anti-Calsenilin or an anti-Vglut3 antibody to specifically label the S2 and S3 sublayers. Although calsenilin and Vglut3 signals were seen on the S2 and S3 sublayers in both WT and Prdm13^−/− retinas, the S2/S3 border bundle labeled with the anti-Calretinin antibody disappeared in the Prdm13^−/− retina (Fig. 7D–I). These results showed that a loss of Prdm13 leads to a specific reduction of ∼30% of the CALBs⁺ amacrine cells that project to the S2/S3 border in the INL.

Figure 7.

Loss of Prdm13 abolishes the S2/S3 border bundle in the IPL. A–C, WT and Prdm13^−/− retinas at P14 were immunostained with an anti-Calbindin (A) or an anti-Calretinin (B) antibody. Calbindin- or Calretinin-positive cells localized to the outer edge of the INL disappeared in the Prdm13^−/− retina (arrowheads). The numbers of Calbindin⁺ or Calretinin⁺ amacrine cells in the INL in WT and Prdm13^−/− retinas were counted (C). D–I, WT and Prdm13^−/− retinas at P14 were immunostained with antibodies against Calretinin (green), which visualizes three sublayers in the IPL and Calsenilin (E, F, magenta) or Vglut3 (H, I, magenta), both of which label S2 and S3 layers. Error bars indicate SD from the means of 3 different mice. **p < 0.01. Scale bars, 50 μm.

Prdm13^−/− mice exhibit elevated visual sensitivities in OKRs

To test visual function in the Prdm13^−/− mice, we first performed ERG recordings under scotopic and photopic conditions with several light intensities in WT and Prdm13^−/− mice at 2 months (Fig. 8). Both WT and Prdm13^−/− mice displayed typical a- and b-waveforms and the same levels of amplitudes in both scotopic and photopic conditions (Fig. 8A–F). These results were consistent with no observable change in photoreceptor cells and bipolar cells in the Prdm13^−/− retina (Fig. 2I,J,L,O). In addition, we analyzed oscillatory potential waves (OPs), which are several feedback wavelets and are thought to be derived from amacrine cells (Wachtmeister, 1998). Prdm13^−/− mice showed normal waveforms and amplitudes of OPs in both scotopic and photopic conditions (Fig. 8G–J). We did not detect any differences between WT and Prdm13^−/− mice in the ERG analysis.

Figure 8.

ERG analysis of Prdm13^−/− mice. A–F, ERGs were recorded from WT (n = 5) and Prdm13^−/− (n = 6) mice at 1 month. Representative scotopic ERG waves in WT and Prdm13^−/− mice are presented (A). The amplitudes of the scotopic ERG a-wave (B) and the b-wave (C) are shown as a function of the stimulus intensity. Representative photopic ERG waves in WT and Prdm13^−/− mice are presented (D). The amplitude of the photopic ERG a-wave (E) and the b-wave (F) in WT and Prdm13^−/− mice are shown as a function of the stimulus intensity. G–J, OPs are extracted from scotopic ERGs (1.0 log cd-s m⁻²) (G, H) and photopic ERGs (1.0 log cd-s m⁻²) (I, J) in WT and Prdm13^−/− mice. Representative waves of scotopic OPs (G) and photopic OPs (I) in WT and Prdm13^−/− mice are displayed. Amplitudes of the scotopic OPs (OP1-OP5) (H) and of the photopic OPs (OP1-OP3) (J) were measured in WT and Prdm13^−/− mice. Error bars indicate SEM.

To investigate the effect of loss of Prdm13 on visual function, we next measured OKRs in WT and Prdm13^−/− mice. OKR is a reflexive eye movement observed in responding to moving objects (Distler et al., 1999; Stahl, 2004; Büttner and Kremmyda, 2007) (Fig. 9A,B) that can be divided into initial and late phase based on time range during eye movement (Gellman et al., 1990; Tabata et al., 2010). Initial OKR is observed in a very short period within 500 ms after visually sensing a moving object, whereas late OKR (also known as optokinetic nystagmus) is a series of eye movements alternating slow tracking and quick resetting to initial eye position within a longer time period (30 s in this study). We quantitatively measured OKRs in WT and Prdm13^−/− mice by plotting mean amplitudes for each visual stimulus in the coordinate system of spatial and temporal frequencies (Fig. 9C,E,G,I). We also visualized quantified OKR responses of WT and Prdm13^−/− mice as a heat-map (Fig. 9D,F,H,J). In the initial OKRs, Prdm13^−/− mice showed OKR responses to a broader range of SF and TF than WT mice (Fig. 9C–F; SF: WT 0.06–0.25 cycles/deg, Prdm13^−/− 0.03–0.25 cycles/deg; TF: WT: 0.75–12 Hz, Prdm13^−/− 0.375–12 Hz). Consistent with the initial OKRs, in the late OKRs, Prdm13^−/− mice displayed OKR responses to a broader range of SF and TF compared with WT mice (Fig. 9G–J; SF: WT 0.12–0.25 cycles/deg, Prdm13^−/− 0.03–0.25 cycles/deg; TF: WT: 0.75–3 Hz, Prdm13^−/− 0.187–12 Hz). Moreover, in both initial and late OKRs, the eye speeds at the optimal response were significantly higher in Prdm13^−/− mice (Fig. 9K,L). These data suggest that Prdm13^−/− mice show significantly elevated sensitivities to visual stimuli. We furthermore examined contrast sensitivity in WT and Prdm13^−/− mice. We exposed mice to eight contrast levels (1%–96%) of visual stimuli constant in SF (0.125 cycles/deg) and TF (1.5 Hz), to which both WT and Prdm13^−/− mice showed the same levels of OKR amplitudes. Prdm13^−/− mice exhibited significantly higher responses to 8%, 16%, and 96% contrast levels than WT mice (Fig. 9M), suggesting that Prdm13^−/− mice can detect significantly lower contrast objects compared with WT mice (Fig. 9M). Finally, to examine whether abnormalities in the retina in Prdm13^−/− mice are responsible for the OKR alterations, we analyzed OKRs in Prdm13 CKO mice obtained by crossing Prdm13 flox mouse with Dkk3-Cre transgenic mice, which express Cre predominantly in embryonic retinal progenitor cells (Fig. 2A,B) (Sato et al., 2007). Before OKR analysis, we confirmed that Prdm13 CKO showed the same histological phenotypes as Prdm13^−/− mice (Fig. 10A–F). In initial OKRs, Prdm13 CKO mice showed higher responses to visual stimuli with a broader range of SF and TF than those observed in WT mice (Fig. 10G–J), showing a similar tendency as the OKRs in Prdm13^−/− mice. These results suggest that loss of Prdm13⁺ amacrine cells leads to abnormally elevated visual sensitivities, although we cannot completely exclude the possibility that some abnormalities outside the retina affect OKRs.

Figure 9.

Prdm13^−/− mice exhibit elevated visual sensitivities in OKRs. A, B, Schematic representation of OKR testing apparatus and procedures. A mouse is placed on a platform, and moving gratings are displayed on the computer screens surrounding the mouse (A). The visual stimulus was presented on three monitors that surround the mouse (19-inch LCD, refresh rate; 75 Hz, size: 270° × 65.7°). The right eye of the mouse was monitored with a CCD camera to trace the eye tracking. At the beginning of each trial, a stationary visual pattern was presented and then moved counterclockwise or clockwise at a constant speed for 0.5 s (initial OKR) and 30 s (late OKR). After a defined period, the pattern was removed to become a uniform gray (B). C–F, Amplitudes of the initial OKRs in WT (n = 5) and Prdm13^−/− (n = 6) mice at 2 months were measured. Amplitudes of the initial OKRs represented by diameter of the circles are plotted in the coordinate system of spatial and temporal frequencies (SF and TF, respectively). Filled symbols represent statistically significant (t test, p < 0.05) responses (C,E). Heat-map plots of the best-fit Gaussian functions for the initial OKRs in WT (D) and Prdm13^−/− (F) mice are displayed. G–J, Amplitudes of the late OKRs in WT (n = 8) and Prdm13^−/− (n = 7) mice at 2 months were measured. Filled symbols represent statistically significant responses (G, I). Heat-map plots of the best-fit Gaussian functions for the late OKRs in WT (H) and Prdm13^−/− (J) mice are shown. K, L, Eye speeds at the optimal stimulus in initial OKRs in WT (n = 5) and Prdm13^−/− (n = 6) mice (K) and in late OKRs in WT (n = 4) and Prdm13^−/− (n = 4) mice (L) are shown. M, Contrast sensitivities of WT (black, n = 5) and Prdm13^−/− (red, n = 5) mice are plotted. **p < 0.01. *p < 0.05. Error bars indicate SD.

Figure 10.

Histological and OKR analyses of Prdm13 CKO mice. A–F, Control and Prdm13 CKO retinas at P14 were immunostained with antibodies (green) against Prdm13 (A), Pax6 (B), Gad65/67 (C), Glyt1 (D), Calretinin (E), and TH (F). Insets, Inner INL and IPL area with high magnification (B–D and E, respectively). White bars represent the S1/S2 and S3/S4 border bundles. Orange bars represent the S2/S3 border bundle in the IPL (E). Nuclei were stained with Hoechst (blue signals). Scale bars, 50 μm. G–J, Amplitudes of the initial OKRs were measured in control and Prdm13 CKO mice at 2 months. Filled circles represent statistically significant (p < 0.05) responses (G, I). Amplitudes of the initial OKRs in control (G) and Prdm13 CKO (I) mice are represented by diameter of the circles are plotted in the coordinate system of SF and TF. Heat-map plots of the best-fit Gaussian functions for the initial OKRs in control (H) and Prdm13 CKO (J) mice are presented. G, H, Control mice (n = 3). I, J, Prdm13 CKO mice (n = 3).

Discussion

In the current study, we found that Prdm13 is specifically expressed in developing and mature amacrine cells in the mouse retina. We characterized molecular features of the Prdm13⁺ amacrine cells, and observed that the majority of the Prdm13⁺ amacrine cells express CALBs, which label a subset of amacrine cells, including starburst amacrine cells and other amacrine subtypes containing GABA, NO, and/or substance P (Haverkamp and Ws̈sle, 2000). However, we did not detect coexpression of Prdm13 with Islet1 (Fig. 1J). Instead, we detected coexpression of Prdm13 with Gad65 or Glyt1, which represents GABAergic or glycinergic amacrine cells respectively (Fig. 1F,G). Although we observed a significant reduction of GABAergic and glycinergic amacrine cells in the Prdm13^−/− retina, cell numbers of amacrine cells containing other neurotransmitters, such as glutamate, acetylcholine, and dopamine, were not affected in the Prdm13^−/− retina (Fig. 4). These results indicate that Prdm13⁺ amacrine cells consist of mainly GABAergic or glycinergic cells, and a small population of other subtypes, except starburst amacrine cells. CALBs label the three strata in the IPL (Fig. 7), with the outermost (S1/S2) and innermost (S3/S4) strata reported to be projections of starburst amacrine cells (Yoshida et al., 2001). Consistent with this observation, loss of Prdm13 did not affect starburst amacrine cells (Islet1⁺/Calbindin⁺ cells in the INL) but caused a loss of the CALBs⁺ S2/S3 border bundle in the IPL (Figs. 4E,K, 7), showing that the S2/S3 border bundle is projected from Prdm13⁺ amacrine cells. Based on these observations, we propose that Prdm13 newly defines a subtype of amacrine cells, which are GABAergic or glycinergic and specifically project to the S2/S3 border in the IPL.

In the spinal cord, Ptf1a directly acts upstream of Prdm13, and both of these factors promote the generation of GABAergic neurons (Chang et al., 2013). However, it had remained unclear whether Prdm13 specifies all of the Ptf1a-lineage neurons or whether Prdm13 generates specific subtypes of Ptf1a-lineage neurons. In the current study, we obtained two experimental observations supporting the latter hypothesis. First, loss of Ptf1a in the mouse retina leads to a complete loss of horizontal cells and a vast majority of amacrine cells, specifically almost a complete loss of GABAergic and glycinergic amacrine cells and a severe loss of CALBs⁺ amacrine cells (Fujitani et al., 2006; Nakhai et al., 2007) (Fig. 6A,H,I,J). In contrast, Prdm13^−/− mice showed normal cell numbers of horizontal cells but a partial loss of GABAergic/glycinergic amacrine cells. Second, whereas forced expression of Ptf1a promoted the specification of both cholinergic amacrine cells (Islet1⁺/Calbindin⁻ cells) and GABAergic/glycinergic amacrine cells (AP-2α⁺ and AP-2γ⁺ cells), forced expression of Prdm13 preferentially promoted generation of GABAergic/glycinergic amacrine cells. These observations demonstrate that Prdm13 is related to the specification of more limited amacrine subtypes than Ptf1a and suggest that Ptf1a upregulates Prdm13 to subdivide Ptf1a-lineage amacrine cells. At the same time, our observations also imply the existence of other downstream factor(s) that diversify Ptf1a-lineage retinal neurons (Fig. 11). In the current study, we could not determine whether reduction of amacrine cells in the Prdm13^−/− retina is due to cell fate conversion from amacrine cells to other cell types or simply due to cell death of Prdm13⁺ amacrine cells. We did not observe compensatory increases of other retinal cell types in Prdm13^−/− retinas as far as we examined (Figs. 2I–O, 4), suggesting that cell fate conversion may not occur in Prdm13^−/− amacrine cells. However, there remains the possibility that the population of Prdm13⁺ amacrine cells is too small to detect compensatory increases in cell numbers of other retinal cell types and to trace Prdm13⁺ amacrine cells during development. Future analysis, such as lineage tracing of Prdm13⁺ amacrine cells in Prdm13^−/− retina using, for example, a reporter mouse line expressing Cre recombinase under the Prdm13 promoter will be necessary to clarify this point.

Figure 11.

A hypothetical model for Prdm13 function in amacrine cell development. Foxn4 and Rorβ1 are expressed in retinal progenitor cells and control the commitment to amacrine and horizontal cell lineages. Ptf1a is upregulated by both Foxn4 and Rorβ1 and consolidates differentiation of amacrine and horizontal cells. Prdm13 is downstream of Ptf1a and is responsible for subtype specification of GABAergic, glycinergic, and CALBs⁺ amacrine cells. Ptf1a might activate other downstream factor(s) (factor X(s)) regulating a different subset of amacrine cells from those induced by Prdm13. Our current study suggests that the population of amacrine cells, whose development is regulated by Prdm13, is distinct from cholinergic amacrine cells whose development is controlled by Islet1. Prdm13⁺ amacrine cells projecting to the S2/S3 border in the IPL modulate visual sensitivity to spatiotemporal frequency and contrast, whereas islet1⁺ cholinergic starburst amacrine cells is essential for direction selectivity in vision.

In the chick spinal cord and the Xenopus neural tube during development, Prdm13 was suggested to promote the specification of GABAergic and glycinergic neurons by repressing Tlx3, which is involved in the generation of glutamatergic neurons (Chang et al., 2013; Hanotel et al., 2014). In contrast, we did not observe an increase of glutamatergic amacrine cells labeled with Vglut3 in the Prdm13^−/− retina (Fig. 4H,I,L). Moreover, our previous microarray data from embryonic and postnatal retinas showed that expression levels of the Tlx family genes (Tlx1–3) are very low or almost undetectable in the mouse retina (Omori et al., 2011). Therefore, we suppose that the mechanisms of amacrine subtype specification mediated by Prdm13 are different from those regulating the specification of inhibitory neurons in the spinal cord and neural tube. Our in vivo electroporation experiments in the mouse retina revealed that the zinc finger domain, not the PR domain of Prdm13, is responsible for specification of the amacrine subtypes, consistent with the previous result in the chick spinal cord. A zinc finger domain can mediate DNA-binding and/or protein-protein interaction. Mash1/Ascl1, a proneural bHLH gene, was reported to interact with Prdm13 (Chang et al., 2013). Another proneural bHLH factor, Neurog2 transcriptional activity is suggested to be repressed by Prdm13 (Hanotel et al., 2014). In addition, Prdm8, which exhibits the highest homology to Prdm13 among all of the Prdm family proteins, forms a repressor complex together with bHLHb5 (Ross et al., 2012). These suggest that Prdm13 is likely to form a repressor complex through protein–protein interaction with a bHLH factor. Thus, identification of this bHLH factor will be an important clue in elucidating a detailed molecular mechanism underlying amacrine specification mediated by Prdm13.

In OKRs, we observed higher temporal, spatial, and contrast sensitivities in Prdm13^−/− mice compared with those in WT mice (Figs. 9, 10G–J). The ablation of starburst amacrine cells almost completely abolished OKRs due to inability to detect the direction of moving objects (Yoshida et al., 2001). However, our histological analyses demonstrated that starburst amacrine cells were unaffected in the Prdm13^−/− retina (Fig. 4E,K), suggesting that starburst amacrine cells do not contribute to the OKR phenotypes observed in Prdm13^−/− mice (Fig. 9). A previous electrophysiological study suggested that ganglion cells are modulated by NPY-releasing amacrine cells to respond to visual stimuli with low spatial frequency. When this amacrine subtype was ablated, ganglion cell responses to low spatial frequency significantly diminished (Sinclair et al., 2004). Interestingly, Prdm13^−/− mice showed augmented OKRs to a range of spatial frequencies (Fig. 9). Together, Prdm13⁺ amacrine cells projecting to the CALBs⁺ S2/S3 border bundle in the IPL play important roles in visual processing in the retina (Fig. 11).

To our knowledge, the Prdm13^−/− mouse is the first mutant mouse in which temporal, spatial, and contrast sensitivities in OKRs are shifted significantly higher than those in WT mouse. What amacrine cell population and neural circuits in the retina are associated with the visual phenotypes in Prdm13^−/− mice? It was reported that Type 2 amacrine cells project to the S2/S3 border bundle in the IPL, contain GABA as a neurotransmitter, and express CALBs (Knop et al., 2011). Because the CALBs⁺ S2/S3 border bundle totally disappears in the Prdm13^−/− retina, we speculate that Type 2 amacrine cells are mainly Prdm13-positive. Type 2 amacrine cells exhibit ON-OFF light responses by transient depolarization to the onset and the offset of a light stimulus and form synapses with Vglut3⁺ amacrine cells (GACs) containing glutamate and glycine (Knop et al., 2011). Recently, GACs were found to project to both ON type and ON-OFF type direction selective ganglion cells (DSGCs) that are responsible for direction selectivity and play an essential role in OKRs. GACs mediate feedforward and crossover excitations of both types of DSGCs by releasing glutamate (Lee et al., 2014). Immunohistochemical analysis showed that GACs receive projections from Type 2 amacrine cells (Knop et al., 2011). If Type 2 amacrine cells are Prdm13⁺, Prdm13⁺ amacrine cells may negatively affect DSGC activities through GACs, suggesting a possibility that Prdm13 loss may lead to elevated sensitivities in OKRs through modulation of DSGCs. Future studies to clarify functional interactions among Prdm13⁺ amacrine cells, GACs and DSGCs, as well as a possible relationship between Prdm13⁺ amacrine cells and Type 2 amacrine cells will be needed to exactly understand the physiological roles and functional mechanisms of Prdm13⁺ amacrine cells in the retinal circuit.

Although amacrine interneurons, which are mainly inhibitory neurons, are thought to be crucial for visual information processing in the retina, their complexity makes our understanding of the exact function and mechanism of each amacrine subtype extremely difficult. The current study may shed new light on a functional role of the specific amacrine subtype specifically projecting to the S2/S3 border bundle. We found that visual sensitivities measured by OKR increased without Prdm13 in mice; however, we suspect that there is a deficit in some specific visual function that we cannot assess by OKR or ERG as a trade-off for elevated temporal, spatial, and contrast sensitivities. Future electrophysiological analyses on Prdm13⁺ amacrine cells at the single-cell level as well as a circuit level will be required to clarify the mechanisms underlying the observed OKR phenotypes in the Prdm13^−/− mice.