Introduction

AII amacrine cells are interneurons critical for transmitting scotopic rod photoreceptor signals through the retina (Strettoi et al., 1992; Bloomfield and Dacheux, 2001). They additionally contribute to photopic vision, generating crossover inhibition between the ON and OFF pathways (Demb and Singer, 2012; Werblin, 2010). AII cells are the most numerous of amacrine cell types (Jeon et al., 1998; Keeley et al., 2014a; Strettoi and Masland, 1996), having narrow-field processes arborizing in the inner plexiform layer (IPL). Each AII cell gives rise to two sets of processes. One set, the arboreal dendrites, is distributed to the ON stratum of the IPL and receives glutamatergic input from rod bipolar terminals while forming gap junctional contacts with neighboring AII cells as well as with ON cone bipolar cell (CBC) terminals. The other set, the lobular terminals, is distributed to the OFF stratum, forming inhibitory glycinergic synapses with the terminals of OFF CBCs and the dendrites of OFF retinal ganglion cells (Famiglietti and Kolb, 1975; Hartveit and Veruki, 2012; Kolb and Famiglietti, 1974; Marc et al., 2014). The total number of AII cells varies considerably across different strains of mice (Kulesh et al., 2023), being distributed locally as a random, rather than regular, array (Keeley et al., 2020). Their two sets of processes interact with those of their homotypic neighbors in distinctive manners to ensure uniform coverage across the retinal surface in the presence of such irregularity in their patterning and variation in their density (Keeley et al., 2023). Although the genetic signature of AII cells has recently been revealed through single-cell transcriptomic profiling (Yan et al., 2020), little is known of the genetic determinants of this particular amacrine cell type.

AII amacrine cells have previously been shown to express the transcription factor Nfia, upregulating it during early postnatal development and maintaining its expression into maturity, suggesting that it may play a critical role in their development (Keeley and Reese, 2018). Nuclear factor one A (NFIA) is one of a family of three NFI transcription factors (Nfia, Nfib, and Nfix) that are all enriched in late-stage retinal progenitor cells, and elimination of all three results in a progenitor cell population that remains proliferative at the expense of generating later-born neurons and Müller glia (Clark et al., 2019). Whether these factors play a role in specifying cellular fate in the retina remains unclear, although the upregulation of these factors in specific populations of bipolar and amacrine cells may indicate such a role in nascent neurons leaving the cell cycle (Keeley and Reese, 2018; Shekhar et al., 2016; Yan et al., 2020).

To study the role of NFIA in the development of AII cells, Nfia was conditionally ablated from retinal progenitors. Such Nfia–conditional knockout (Nfia-CKO) retinas showed a massive reduction in the population of AII cells in maturity while leaving the architecture of the retina and its cellular composition relatively unperturbed, with two notable exceptions. This reduction in AII cell number was apparent before their morphologic differentiation, early in the first postnatal week, yet there was no evidence for a coincident increase in cell death; furthermore, the induced conditional elimination of Nfia postnatally did not reduce the number of AII cells, nor did it affect their differentiation. Together, these results implicate Nfia in the specification of the AII amacrine cell fate. The physiological consequences of their elimination were assessed via the electroretinogram (ERG), demonstrating that AII cells provide a major contribution to the generation of the oscillatory potentials (OPs). Finally, the loss of the AII cells resulted in perturbations to two bipolar cell populations with which they are synaptically connected. Type 2 cone bipolar cells (CBCs) showed reduced numbers associated with an increase in apoptosis, suggesting a dependency on the AII cells for their survival. Rod bipolar cells (RBCs) showed no change in number but exhibited an expansion of their terminals further into the IPL, indicating a constraining role for the AII amacrine cells on their stratification.

ResultsConditional elimination of Nfia using Rx-cre depletes the retina of NFIA labeling

Single-cell transcriptomic analyses of adult mouse retinal cells confirm Nfia to be expressed in a small number of amacrine cell types, being most abundant in AII amacrine cells (Yan et al., 2020). Horizontal cells and a few bipolar cell types also express it, particularly the type 5D CBC (Macosko et al., 2015; Shekhar et al., 2016). Additionally, Müller glia express Nfia, as well as astrocytes in the optic fiber layer (Macosko et al., 2015). A microarray database profiling expression across 13 different retinal cell types (including five types of amacrine cells) on postnatal day 7, in contrast, showed only one retinal cell type with notable Nfia expression, the AII amacrine cell (Kay et al., 2012), intimating an early developmental significance for it in this cell type.

As previously reported, antibodies to NFIA in the developing mouse retina show only migrating retinal astrocytes to be immunopositive on the day of birth (Keeley and Reese, 2018). By P5, faint NFIA expression is detectable in cells scattered across the emerging INL, whereas more brightly labeled cells are seen to be coalescing adjacent to the developing IPL in the future amacrine cell layer. By P10, a discrete stratum of intensely labeled NFIA+ amacrine cells abuts the IPL, being the AII amacrine cells. Additionally, other cell types in the INL become immunopositive, including other amacrine cells, horizontal cells, and some bipolar cells as well as Müller glia. By maturity, the horizontal cells are no longer detected, but Müller glia and a few cone bipolar cells remain NFIA+, as does the stratum of AII amacrine cells and a few other amacrine cells. The identity of the AII amacrine cells among the NFIA+ population is confirmed by their coexpressing DNER, this combinatorial labeling pattern having been shown to identify exclusively the entire population of AII cells (Keeley and Reese, 2018).

Rx-cre-expressing mice were crossed with mice bearing floxed alleles of Nfia to generate Nfia-CKO mice. Rx is normally activated during eye formation, thereby producing Cre recombinase in early retinal progenitors prenatally (Swindell et al., 2006). In such Nfia-CKO retinas, large portions of nasal retina are entirely NFIA immunonegative (Fig. 1A,B), whereas other regions retain sporadic NFIA labeling, though considerably reduced in density (Fig. 1C–E). Cross sections of retina confirm this loss of NFIA labeling from the entire INL (Fig. 1F,G), or its reduced density in regions (Fig. 1H), and this elimination is widespread from the earliest stages of normal NFIA expression, as the reduction is already conspicuous on P5 (Fig. 1I–K). Regions of incomplete elimination reflect mosaicism in Cre-mediated recombination, evident from the mosaic expression pattern of the Cre reporter (Fig. 2).

Figure 1.

Nfia-CKO retinas show extensive loss of NFIA labeling. A, B, Quadrants of retinal wholemounts show conspicuous loss of NFIA labeling in the Nfia-CKO retinas (B) relative to littermate control (A) retinas. C–E, At higher magnification, the loss in such depleted quadrants, consistently nasal retina, is confirmed to be complete (D, compare with littermate control in C), whereas other quadrants in which recombination was incomplete show a considerable decline of NFIA+ cell density (E). Note the large, bright, NFIA+ cells in control retina (C) and their scarcity in Nfia-CKO retina including such partially depleted regions (E), being the AII amacrine cells (Keeley and Reese, 2018). F–H, Retinal sections show the characteristic labeling present within the INL of littermate controls (F), including amacrine, bipolar, and Müller glial cells, and its complete absence (G) or reduced density (H) in the INL in the Nfia-CKO retinas. I–K, This loss of NFIA labeling is already detected at P5, shown here in wholemounts, when NFIA+ amacrine cells are normally first detected. Scale bars: A, B, 1 mm; C–K, 50 µm.

Figure 2.

Rx-cre transgene produces robust Cre-mediated recombination across the retina. Left and right retinas expressing the td-tomato cre-reporter, exhibiting patterns of Cre activity characteristic of the loss of NFIA labeling, the latter being entirely absent in the nasal retina and showing slight mosaicism elsewhere. Scale bar, 1 mm. D, dorsal; T, temporal; V, ventral; N, nasal.

Nfia-CKO mice lack AII amacrine cells

Nuclear PROX1 labeling reliably discriminates AII amacrine cells within the INL in the mouse retina by virtue of its heightened intensity in a population of large amacrine cells immediately adjacent to the IPL (Keeley and Reese, 2018; Perez de Sevilla Müller et al., 2017). Nfia-CKO retinas lack this population of PROX1+ cells abutting the IPL, yet leave intact the PROX1+ bipolar and horizontal cells and a few other weakly labeled amacrine cells that are not AII cells (Fig. 3A–D). This loss of AII amacrine cells was validated by using an independent genetic marker for this cell type, the Cdh1-gfp reporter (Firl et al., 2015; Gamlin et al., 2020), which was bred onto these Nfia-CKO mice. This cytoplasmic GFP labeling is restricted to the AII amacrine cell population, and these GFP+ cells were also found to be reduced in the Nfia-CKO retinas (Fig. 3E–H). Indeed, quantification of either these brightly labeled PROX1+ cells in the amacrine cell layer, or of the GFP-positive cells, show a total reduction of comparable magnitude by 78% (p < 0.001) and 84% (p = 0.008), respectively, averaged across the entire retina (Fig. 3I,J). In partially depleted regions in the Nfia-CKO retinas, the few remaining GFP+ cells are always NFIA+ (Fig. 3K), confirming that Cre-mediated recombination was incomplete in these regions (Fig. 2), rather than suggesting that some AII amacrine cells still exist without functional NFIA.

Figure 3.

AII amacrine cells are missing from the Nfia-CKO retina. A–D, PROX1+ AII amacrine cells are eliminated from the Nfia-CKO retina (B, D), shown in both wholemounts (A, B) and retinal sections (C, D). Note that other PROX1+ cells in the INL remain intact (D). E–H, GFP+ AII amacrine cells, labeled by the Cdh1-gfp reporter, are also reduced in the Nfia-CKO retina. I, J, The total populations of either PROX1+ AII amacrine cells (I) or GFP+ AII amacrine cells (J), derived from sampling across the entire retina, both undergo substantial reductions of ∼80%. n, Number of retinas sampled. K, The few remaining GFP+ cells in Nfia-CKO retinas remain NFIA+, confirming these AII amacrine cells did not undergo Cre-mediated recombination. L–O, The GLYT1+ amacrine cell population undergoes a partial depletion in the Nfia-CKO retina (M, O), apparent in both wholemounts (L, M) and retinal sections (N, O). P–S, Similarly, the density of CX36+ puncta, used by AII amacrine cells in their gap junctional connectivity in the ON stratum of the IPL, is partially reduced. C, D, G, H, N, O, Red arrows indicate the approximate retinal depth shown in the wholemount micrographs, R, S, Red brackets indicate the approximate range of the IPL used to form the maximum projection wholemount images in P and Q. Scale bars, 50 µm. Asterisks indicate statistically significant differences.

AII amacrine cells form two main output channels using different synaptic mechanisms; they form inhibitory glycinergic synapses onto OFF CBCs and ganglion cells, and electrical synapses with neighboring AII amacrine cells and ON CBCs (Famiglietti and Kolb, 1975; Hartveit and Veruki, 2012; Kolb and Famiglietti, 1974; Marc et al., 2014). Thus further confirmation that the population of AII amacrine cells is missing comes from examining the population of amacrine cells that is immunopositive for the glycine transporter GLYT1 as well as observing the pattern of CX36 puncta within the IPL. Although membranous GLYT1 labeling is not completely abolished (Fig. 3L–O), because of the presence of other glycinergic amacrine cell types, cell density is reduced, particularly at the border with the IPL (Fig. 3N,O). Additionally, the density of CX36 puncta is reduced in the ON stratum of the IPL in Nfia-CKO retinas (Fig. 3P–S), where AII amacrine cells normally form gap junctional contacts. Together, these results indicate that NFIA is required to establish a population of functional AII amacrine cells.

An analysis of the morphologies of those few remaining GFP+ AII amacrine cells in the Nfia-CKO retina (Fig. 3K) offers further evidence that AII amacrine cells are missing from these retinas. AII amacrine cells are known to interact with their homotypic neighbors as they differentiate their lobular terminals and arboreal dendrites, doing so in characteristically distinct manners to constrain outgrowth (Keeley et al., 2023). The lobular terminals normally tile the retina, and the areal size of their domains is directly related to local density. For instance, abrogating naturally occurring cell death, in the Bax-KO retina, leads to a 33% increase in the number of AII amacrine cells, whereas the areas of their lobular terminal fields decline proportionately to maintain a coverage factor of ∼1.0. The arboreal dendrites, in contrast, normally overlap rather than tile with their neighbors, altering their branching density without modulating their areal size in the Bax-KO retina (Keeley et al., 2023). For those few GFP+ AII amacrine cells remaining in these Nfia-CKO retinas (and remaining NFIA+, as noted above), they go on to differentiate their characteristic lobular terminals and arboreal dendrites, as expected (Fig. 4A,B), but field areas of both have increased significantly, lobular terminal area doubling (p < 0.001) and arboreal dendritic area tripling (p < 0.001) in size on average (Fig. 4C,D). Note the conspicuous variability in the field sizes in the Nfia-CKO retinas (Fig. 4C,D) because of the variable reductions in local AII amacrine cell density surrounding each injected cell (Fig. 4E,F), with even the arboreal dendrites now increasing in areal extent. Together, the foregoing results would indicate that AII amacrine cells are missing from the mature retina when Nfia is eliminated during early retinal development. A scRNAseq analysis of the Nfia-CKO retina may ultimately provide independent confirmatory evidence for this.

Figure 4.

Remaining AII amacrine cells in the Nfia-CKO retina expand their field areas when homotypic neighbors are depleted. A, B, GFP+ AII amacrine cells were injected with AF564 to quantify the areal domains of their lobular terminal fields (top) and their associated arboreal dendritic fields (bottom) in retinal wholemounts from Nfia-CKO (B) and littermate control (A) retinas. C, D, Both the lobular terminals fields (C) and the arboreal dendritic fields (D) are two to three times larger in the Nfia-CKO retinas, indicating a loss of normal homotypic constraints on their growth. n, Number of cells quantified. E, F, A single injected GFP+ cell from an Nfia-CKO (F) and a control retina (E), with labeled dendritic and lobular terminal fields pseudocolored independently, relative to neighboring GFP+ somata in each field. Scale bars, 50 µm. Asterisks indicate statistically significant differences.

AII amacrine cells are absent during early postnatal development

The early loss of Nfia, already detected by P5 (Fig. 1I–K), might prevent AII amacrine cells from completely maturing, in turn dying after they have commenced their differentiation. Nfia-CKO retinas were consequently examined at P5 and P10 for the presence of AII amacrine cells during the period of AII amacrine cell differentiation. The population of AII amacrine cells can be detected by GFP expression at P10 in littermate control retinas (but not at P5, when only a few cells are labeled), yet their density is already reduced in the Nfia-CKO retina by this age (Fig. 5A,B). The PROX1 population of AII amacrine cells is also readily detected at P10 in CTRL retinas, yet in the Nfia-CKO retinas, it is reduced (Fig. 5C,D), and the same is true for the population of GLYT1+ amacrine cells (Fig. 5E,F). At P5, when PROX1+ AII amacrine cells are normally first detected, they are seen to be depleted already in the Nfia-CKO retinas (Fig. 5G,H), when the depletion of the GLYT1+ population is also considerably reduced (Fig. 5I,J). Together, these results would suggest that AII amacrine cells are never produced during development, rather than initiating their differentiation and then subsequently undergoing cell death.

Figure 5.

AII amacrine cells are eliminated in the Nfia-CKO retina before they differentiate. A–F, At P10, GFP-labeled cells are depleted in the Nfia-CKO retina (A, B) as are the PROX1+ (C, D) and GLYT1+ cells (E, F). Note that GLYT1 labeling remains, although few of the prominent circular profiles found in the littermate control retinas (E), positioned at the IPL border, are present (F), being the AII population. G–J, At P5, Cdh1-gfp expression is not yet detected in AII amacrine cells (data not shown), but PROX1+ AII amacrine cells already show the depletion present at later ages (G, H). GLYT1+ labeling is also diminished at this stage (I, J). Scale bars, 50 µm.

Nfia-CKO retinas exhibit normal retinal architecture and areal extent

Adult Nfia-CKO retinas were labeled to identify various other retinal cell types to examine the cellular composition of the retina as well as its architecture and stratification. Antibodies to proteins that label various other retinal cell types confirm that the characteristic features of the retina are largely normal, including somal positioning across the depth of the retina as well as the stratification of their processes (Fig. 6A–X). GFAP labeling identifies only astrocytic processes in the inner retina, showing no upregulation in the Müller glial endfeet, indicating a lack of reactivity in the Nfia-CKO retinas (Fig. 6Y,Z). Retinal architecture is normal (Fig. 6AA,BB), showing no evidence of rosettes in the ONL (Clark et al., 2019), although the thickness of the INL and IPL is slightly diminished, consistent with the absence of this densest of all amacrine cell types. Additionally, retinal area does not change (Fig. 6CC). Of note, three OFF CBC types (types 2, 3b, and 4) that receive glycinergic input from the lobular terminals of AII amacrine cells (reciprocating glutamatergic synapses back onto those lobules; Graydon et al., 2018; Tsukamoto and Omi, 2017) retain their characteristic stratification patterns in the absence of the AII cells (Fig. 6M–R). The RBCs normally innervate the arboreal dendrites of the AII cells in the ON division of the IPL, where their stratification also appears comparable in the absence of the AII cells (Fig. 6S,T). Closer examination, however, reveals an abnormality in the distribution of their terminals.

Figure 6.

Retinal architecture is not compromised in the Nfia-CKO retina. A–Z, Immunolabeling for various retinal cell types in retinal sections from littermate Nfia-CTRL and CKO retinas showed them to have largely normal somal positioning and stratification of their processes, including the Müller glial cells (U–X). GFAP labeling reveals only astrocytic processes in the innermost retina, showing no upregulation in Müller glial endfeet (Y, Z). AA, BB, Hoechst labeling confirms normal retinal architecture. Scale bars, 25 µm. CC, Retinal area (taken from the retinas used to derive AII amacrine cell number in Fig. 3I) is not compromised in the Nfia-CKO retinas. n, Number of retinas quantified.

Figure 7.

The distribution of rod bipolar terminals is altered in the Nfia-CKO retina. A, B, GFP+ AII amacrine cells (green) extend lobular terminals (being the large puncta) in the OFF stratum of the IPL, and arboreal dendrites (being the finer puncta) in the ON stratum of the IPL (A). The latter are located vitreal to the inner CHAT stratum (red), where RBC terminals are normally positioned. C, D, PKC+ RBCs (green) form ectopic terminals (D, arrows) more proximally along their axons in the IPL in the absence of AII cells (B), evinced relative to the inner cholinergic dendritic plexus (red). E, F, SYT2+ type 2 CBCs, by comparison, exhibit no change in stratification, despite the loss of their synaptic partners, the lobular terminals of the AII amacrine cells. (Hoechst labeling of the nuclear layers is shown in blue in A–F). G, H, Three z-stack projections (3 µm thick) taken through the inner cholinergic (CHAT+) stratum (middle, magenta) or scleral (top) and vitreal (bottom) to it, showing the relative density of PKC+ lobules (green) at each depth. Note the thinner axonal shafts passing through the IPL (being largely the only PKC-labeled profiles in CTRL retina, top and middle). I, Quantification of PKC+ lobular densities positioned at these same three levels in the IPL. n, Number of retinas sampled. J–L, Single optical sections from retinal wholemounts labeled for PKC and CTBP2, revealing the presence of CTBP2+ synaptic ribbons in normally positioned lobules vitreal to the inner cholinergic stratum in both control (G) and Nfia-CKO (H) retinas, as well as within ectopic lobules scleral to the inner cholinergic stratum from the same Nfia-CKO retina (I). Positioning of G, H, J–L, relative to the inner cholinergic stratum, is color coded in reference to the schematic of the IPL presented above G, H. Scale bars, A–F, 50 µm; G, H, J–L, 10 µm. Asterisks indicate statistically significant differences.

RBCs exhibit ectopically positioned terminals along their axons

The population of RBCs in Nfia-CKO retinas exhibits the characteristic presence of their stratifying terminals in the deepest parts of the IPL, primarily in stratum S5, where they normally overlap with the arboreal dendrites of the AII amacrine cells (Fig. 7A,C). In the absence of the AII cells, however, ectopic RBC terminals are found to be positioned more proximally along the axon (closer to the soma), in a portion of the IPL that they normally avoid (Fig. 7B,D). That their positioning is abnormal is shown by their expanded distribution relative to the inner cholinergic plexus (Fig. 7D, arrows). Optical sections from wholemounted control retinas, taken at this depth (Fig. 7G, top), or at the depth of the inner cholinergic plexus itself (Fig. 7G, middle), exhibit PKC+ axons coursing through the IPL, but rarely are lobules present, being largely restricted within the IPL between that inner cholinergic stratum and the ganglion cell layer (Fig. 7G, bottom). Nfia-CKO retinas, in contrast, contain lobular terminals mispositioned to these depths (Fig. 7H, top and middle). Counts of their frequency, in each of these three portions of the IPL confirm a significant difference in the density of such ectopic lobules (Fig. 7I, top and middle; p = 0.01 and p < 0.001; Student's t test). Their normal targets, being those arboreal dendrites of the AII amacrine cells primarily positioned in S5 and now missing in the CKO retina (Fig. 7A,B) may therefore play a role in constraining the distribution of their terminals to this innermost portion of the IPL. These ectopic lobular terminals make up ∼18% of all RBC terminals in the Nfia-CKO retina; furthermore, they colocalize with the synaptic ribbon protein c-terminal binding protein 2 (CTBP2) as observed in normally positioned lobules (Fig. 7J–L), suggesting they make functional synapses. Note that these changes in RBC morphology are restricted to the IPL. Within the OPL, RBC dendritic arbors are comparable with control retinas as are the processes of horizontal cells, whereas the distributions of rod spherules, cone pedicles, and synaptic ribbon proteins also exhibit no signs of reactive changes (Fig. 8A–F).