Abstract
The development of the visual system is a complex and multistep process characterized by the precise wiring of retinal ganglion cell (RGC) axon terminals with their corresponding neurons in the visual nuclei of the brain. Upon reaching primary image-forming nuclei (IFN), such as the superior colliculus and the lateral geniculate nucleus, RGC axons undergo extensive arborization that refines over the first few postnatal weeks. The molecular mechanisms driving this activity-dependent remodeling process, which is influenced by waves of spontaneous activity in the developing retina, are still not well understood. In this study, by manipulating the activity of RGCs in mice from either sex and analyzing their transcriptomic profiles before eye-opening, we identified the Type I membrane protein synaptotagmin 13 (Syt13) as involved in spontaneous activity-dependent remodeling. Using these mice, we also explored the impact of spontaneous retinal activity on the development of other RGC recipient targets such as nonimage-forming (NIF) nuclei and demonstrated that proper frequency and duration of retinal waves occurring prior to visual experience are essential for shaping the connectivity of the NIF circuit. Together, these findings contribute to a deeper understanding of the molecular and physiological mechanisms governing activity-dependent axon refinement during the assembly of the visual circuit.
Significance Statement
Waves of correlated activity spontaneously triggered in the retina during perinatal stages play a crucial role in establishing topographic maps and eye-specific segregation in image-forming brain nuclei, contributing to proper adult visual function. Here, we analyze visual nuclei that lack topography and other typical characteristics, referred to as nonimage-forming nuclei, in mice with altered retinal waves and found that retinal waves significantly influence the assembly of these circuits. Furthermore, by analyzing the transcriptomic profiles of retinal ganglion cells from mice with modified retinal activity, we identified novel players implicated in spontaneous activity-dependent refinement. This research provides valuable insights into the molecular and physiological mechanisms that govern the precise wiring of the visual circuitry.
Introduction
Retinal ganglion cells (RGCs) are the cells in the eye that collect visual information from other neurons in the retina and send it to different nuclei in the brain including the superior colliculus (SC) and the lateral geniculate nucleus (LGN) of the thalamus. These nuclei, also known as image-forming nuclei (IFN), are both essential for sight processing. While the SC directs visually guided behaviors, the dLGN relays visual information to the cortex for a deeper conscious visual perception (Stryker and Schiller, 1975). Neighboring RGCs projecting to neighboring cells in the dLGN and the SC form a continuous topographic map that reflects the image projected in the retina (Rakic, 1976; Godement et al., 1984; Chalupa and Snider, 1998). Thus, in the IFN, the spatial relationship among the RGCs in the retina is maintained as an orderly representation (map) of the visual space. In most mammals, in addition to forming topographic maps in the dLGN and SC, RGC axons project in an eye-specific manner. In mice, axons from a subpopulation of RGC located in the ventrotemporal retina project to the SC and dLGN in the same hemisphere (ipsilateral RGCs), while the rest of RGC axons cross the brain midline and project to the opposite side (contralateral RGCs). Ipsi- and contralateral terminals first overlap at early postnatal stages in the dLGN but eventually segregate to finally rearrange into complementary nonoverlapping domains in the mature visual system (Godement et al., 1984).
Both the formation of a raw retinotopic map and the targeting of ipsi- and contralateral visual terminals at the SC and the thalamus are initially governed by guidance and adhesion molecules (Nakamoto et al., 2019; Su et al., 2021). In rodents, spontaneous waves of correlated activity triggered by starburst amacrine cells and transmitted to RGCs refine and prepare the network to receive environmental stimuli when eyes open 2 weeks after birth (Galli and Maffei, 1988; Wong et al., 1993; Demas et al., 2003; McLaughlin et al., 2003; Pak et al., 2004; Torborg et al., 2005; Torborg and Feller, 2005).
Pharmacological studies altering interretinal correlation result in severe segregation defects (Penn et al., 1998; Rossi et al., 2001; Huberman et al., 2002). For instance, monocular injections of forskolin or CPT-cAMP increase wave frequency in the treated eye, which has a greater facility to perform LTP and thus stabilize synaptic buttons, resulting in an increase of its territory at the expense of that corresponding to the opposite retina (Stellwagen and Shatz, 2002). However, segregation was not impaired when wave frequency was increased binocularly since intraretinal activity correlation was maintained (Stellwagen and Shatz, 2002).
Several genetic mouse models demonstrated that the impairment of cholinergic waves in mice resulted in the disruption of eye-specific segregation (Xu et al., 2011; Burbridge et al., 2014). The use of optogenetic methods in different mouse lines later confirmed that retinal waves are present and propagate throughout the entire visual system before eye-opening. This patterned activity encompassed the visual field, relied on cholinergic neurotransmission, preferentially initiated in the binocular retina, and exhibited spatiotemporal correlations between the two hemispheres (Ackman et al., 2012). All of these experimental models have provided in-depth insight into the functioning of spontaneous activity for the refinement of retinal projections prior to the emergence of stimulus-dependent activity. However, the molecular mechanisms that shape the visual circuit by transducing retinal spontaneous activity in axon remodeling at the IFN remain largely unknown.
In addition to innervating the IFN, RGC axons project to NIFN such as the olivary pretectal nucleus (OPN), the nucleus of the optic tract (NOT), or the suprachiasmatic nucleus (SCN). The OPN and NOT mediate the pupillary reflex (Young and Lund, 1994) and saccadic movements (Kato et al., 1986) for image stabilization while the SCN regulates entrainment of the circadian clock, hormone rhythms, and sleep cycles (Hattar et al., 2003; Yonehara et al., 2009; Noseda and Burstein, 2011; Dhande et al., 2013). Correlated retinal activity encodes positional information essential for the remodeling of RGC axons at the IFN, but the impact of spontaneous retinal activity on the arborization of RGC terminals in the NIFN has not been investigated. NIFN receive inputs from the intrinsically photosensitive retinal ganglion cells (ipRGCs), a particular type of RGC that responds directly to light and expresses melanopsin photopigment (OPN4; Berson et al., 2002; Hattar et al., 2002; Do and Yau, 2010; Schmidt and Kofuji, 2011). ipRGCs can be activated by retinal waves, and they carry out modulatory effects even in the absence of light (Renna et al., 2011). As in other visual targets, eye-specific segregation in the OPN is independent of light (Tiriac et al., 2018). However, unlike other RGC recipient nuclei, NIFN support vision indirectly and do not have an apparent need to transmit spatial information. Therefore, the organization of RGC axons in topographic maps and eye-specific projection patterns may not be as evident or might be even absent in NIFN.
Here, by generating transgenic mouse lines with altered retinal wave frequency and analyzing the transcriptomic profile of their RGCs, we identified molecular signatures triggered by spontaneous activity involved in the refinement or remodeling of retinal axons. Moreover, the analysis of RGC terminals in different visual nuclei of these mutant mice revealed that proper frequency and/or duration of perinatal retinal waves is essential for the establishment of accurate connectivity in nonimage-forming visual circuits.
Materials and Methods
Mice
The Kir2.1eYFPflx-Stop line was generated in our laboratory by cloning the human Kir2.1 cDNA (Benjumeda et al., 2013) in a pEYFP-N1 plasmid adding the eYFP reporter protein at the C-terminal of Kir2.1. This fusion protein was subcloned in the pCAG-SE plasmid downstream of the CAG promoter and a loxP-flanked STOP cassette. The function and plasma membrane localization of the chimeric channel were confirmed by electrophysiology. The linearized and purified construct was injected in mice oocytes (CBATEG). Twelve founders were obtained and crossed with Pou4f2-cre mice (Pou4f2tm1(cre)Bnt/J; RRID:IMSR_JAX:030357) to select the founder with the highest Kir2.1eYFP expression in RGCs. Slc6a4-cre line belongs to the GENSAT Project at Rockefeller University (B6.FVB(Cg)-Tg(Slc6a4-cre)ET33Gsat/Mmucd; RRID:MMRRC_031028-UCD). Rosa26Stop-TdTm (B6.Cg-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/J; RRID:IMSR_JAX:007914) and GCaMP6f mice (B6;129S-Gt(ROSA)26Sortm95.1(CAG-GCaMP6f)Hze/J; RRID:IMSR_JAX:024105) were acquired from the Jackson Laboratory. Animals of both sexes were used equally in all experiments, except for in utero electroporations, where females were used for pregnancy. Animals were housed in a timed-pregnancy breeding colony at the Instituto de Neurociencias de Alicante, Spain. Conditions and procedures were approved by the IN Animal Care and Use Committee and met European (2013/63/UE) and Spanish regulations (RD 53/2013).
Electrophysiology
Retinas extracted from P4 or P11 Pou4f2-Kir2.1; Rosa26TdTm and control (Pou4f2; Rosa26TdTm) mice were placed in artificial cerebrospinal fluid solution (ACSF) at 4°C, containing the following (in mM): 126 NaCl, 26 NaHCO3, 10 glucose, 5 MgCl2, 3 KCl, 1.25 NaH2PO4, and 1 CaCl2. pH of the ACSF solution was tampered with using carbogen (5% CO2 and 95% oxygen), and the resulting osmolarity was ∼310 mOsmol/kg. Further, retinas were subjected to enzymatic digestion with collagenase/dispase diluted in ACSF (1:3) for 15–20 min at 35°C with carbogen. Collagenase/dispase mix was prepared as follows: 155 mM NaCl, 1.5 mM K2PO4, 10 mM HEPES, 5 mM glucose, 900 uni/ml collagenase type XI (Sigma-Aldrich, D9542), and 5.5 uni/ml dispase (BD Biosciences, 10103578.001). Recordings were performed using an Olympus BX50WI upright confocal microscope and constant perfusion with ACSF solution at ∼34°C. The temperature was controlled by a feedback Peltier device CL-100 (Warner Instruments). Neurons were visualized through a 40× water immersion objective. To identify TdTm labeling, the field was exposed to a 554 nm light with a polychrome V monochromator (Thermo Fisher Scientific), and the fluorescence emission was captured on a cooled CCDge camera (sensicam, PCO AG). Neurons were current-clamped, and once in the whole configuration, resting membrane potential was measured. Recordings were performed using a MultiClamp 700B amplifier, pCLAMP 10 software and a Digidata 1440A, and pCLAMP 10 software (Molecular Devices). Borosilicate electrodes were pulled using a puller P-97 (Sutter Instrument) with resistance ranging between 4 and 8 MΩ and filled with an intracellular solution containing the following (in mM): 115 K-gluconate, 25 KCl, 9 NaCl, 10 HEPES, 0.2 EGTA, 1 MgCl2, 3 K2-ATP, and 1 Na-GTP adjusted to pH 7.2 with KOH (280–290 mOsmol/kg). Data were sampled at a frequency of 20 KHz and low-passed filtered at 10 KHz. Statistical analyses were performed with SigmaStat 4.0 software (Systat Software), and Origin Pro8 software (OriginLab) was used for graphing.
Retinal wave recordings
Whole-mount retinas from P4 Pou4f2-Kir2.1; GCaMP6f and control (Pou4f2-GCaMP6f) mice were dissected in oxygenated (5% CO2 and 95% oxygen) ACSF at room temperature (RT), containing the following (in mM): 119 NaCl, 5 KCl, 1.3 MgSO4 7H2O, 1 NaH2PO4, 26 NaHCO3, 11 glucose, and 2.4 CaCl2. Retinas were mounted over a Whatman filter paper (Whatman, WHA150446) with the ganglion cell layer side up and transferred turned upside down to a cell culture insert (Millipore, PICM0RG50) in a MatTek glass bottom dish (MatTek, P35G-1.5-14-C) with 1 ml of ACSF. Then, the dishes with the retinas were placed in a warmed (32°C) and gassed recording chamber of an inverted Leica Thunder Imager microscope. After 20 min in the chamber, time-lapse spontaneous calcium recordings were obtained with a 5× dry objective (5×/0.12 dry WD = 14 mm) after exciting the retinas at 475 nm. Images were acquired with a DFC 9000 GTC sCMOS camera for 10 min using a time interval of 300 ms, an exposure time of 200 ms, an LED intensity of 50%, and 4 × 4 binning.
Image analysis of retinal wave recordings
The movies with a resolution of 512 × 512 pixels were read with Fiji/ImageJ software (Schindelin et al., 2012). The mean value of the fluorescence as a function of time was obtained with the Fiji function Plot Z-axis Profile and saved as .csv files for further analysis with Python. The fluorescence peaks corresponding to the presence of calcium waves were detected with the SciPy function scipy.signal.find_peaks as described by Virtanen et al. (2020). For the analysis of regions of interest (ROIs), we overlaid 3 × 3 pixel areas (corresponding to 15.6 × 15.6 μm) onto the retina image. These ROIs were separated by 2 pixels (corresponding to 10.4 μm). The average fluorescence intensity value was computed for all frames, and the resulting signal was detrended and normalized. Additionally, a global intensity signal was obtained by averaging the signals from all ROIs. Wave initiation sites were detected and manually marked in Fiji using the ROI Manager. The data from these marked regions were saved in compressed zip files and later analyzed with Python (3.10) and Numpy, and the results were visualized with Matplotlib. Wave durations and intertransient intervals (ITI) were measured using Fiji and then analyzed statistically using R software. When counting waves during the first minute, only those that swept an area of ≥5% of the retina were considered for further analysis. To compare average measured values between the control and Pou4f2-Kir2.1; GCaMP6f groups, the normality of the data was first assessed. If the data were normally distributed, a t test was used to compare group means. Otherwise, a nonparametric Wilcoxon test was performed using R software. Box plots were used to visually compare the measured values between the control and Pou4f2-Kir2.1; GCaMP6f retinas. Each data point in the box plots represents a measurement from a different animal. The level of significance is presented in the plots, using asterisks, i.e., *p < 0.05, **p < 0.01, ***p < 0.005, and ****p < 0.001.
Anterograde labeling of retinal projections
Postnatal mice were anesthetized in ice or with isofluorane depending on the age. Whole-eye anterograde labeling was performed using a Nanoject II Auto-Nanoliter Injector (Drummond, 3-000-204) injecting twice on opposite points with 0.5 µl of CTB subunit conjugated with Alexa Fluor 488 or 647 (Invitrogen, C34775/C22843/C34778) as described in Rebsam et al. (2009). Two days later, mice were perfused transcardially with 4% paraformaldehyde (PFA) in 0.1 M phosphate-buffered saline (PBS). Brains were incubated overnight in the PFA solution at 4°C and stored in PBS. Fixed tissue was embedded in 3% agarose diluted in PBS and sectioned in a vibratome. Sections were incubated in blocking solution (0.2% gelatin, 5% FBS, 0.005–0.01% Triton X-100) at RT for 1 h and mounted to be analyzed.
Immunofluorescence and in situ hybridization
Sectioned PFA fixed tissue was incubated in blocking solution (0.2% gelatin, 5% FBS, 0.005–0.01% Triton X-100) at RT for 1 h, followed by incubation overnight with the primary antibodies. The following primary antibodies were used at the specified concentrations: rabbit anti-Dsred Pab (Takara Bio, catalog #632496, RRID:AB_10013483; 1:500), rabbit polyclonal Kir2.1 (Abcam, catalog #ab65796, RRID:AB_1140953; 1:500), chicken anti-GFP (Aves Labs, catalog #GFP-1020, RRID:AB_10000240; 1:1,000), and rabbit polyclonal anti-melanopsin (Advanced Targeting Systems, catalog #AB-N38, RRID:AB_1608077; 1:2,500). After three washes in PBS, samples were incubated with the secondary antibodies diluted 1:1,000 in blocking solution at RT for 2 h. For in utero electroporation experiments, the protocol was the same, but the blocking solution used was 5% horse serum and 0.25% Triton X-100 in PBS. For whole-mount retinas, the same protocol was used for sections, but Triton X-100 concentration was increased to 1%, and samples were fixated in methanol. This process includes washes with 25–50–80–100% methanol/H2O for 20 min at RT, bleaching in 3% H2O2 in methanol for 1 h at RT, two washes in methanol, rehydration in 80–50–25% methanol/PBS 20 min each at RT, and blocking in 5% BSA PBST (3% Tween 20) for 1 h at RT.
In situ hybridization for mRNA Syt13 was performed according to the reported methods (Murcia-Belmonte et al., 2019). The mouse probe (688 bp) was amplified from mouse cDNA using the following primers: Forward 5′-GAAACACCAGGCTCAGAAGC-3′; Reverse 5′-ACTGAGGGTACCTGGCACAT-3′.
Brain clarification
Clarification was performed following the iDISCO protocol (Renier et al., 2014). Briefly, samples were dehydrated in methanol/H2O series (20–40–60–80–100–100%) for 1 h at RT. Then samples were incubated 3 h, with shaking, in 66% DCM/methanol at RT, and washed in 100% DCM to wash the methanol. Finally, brains were incubated and stored in a glass tube totally filled with dibenzyl ether at RT. Brains showing total filling of contralateral projections labeled with CTB-Alexa 546 and ipsilateral with CTB-Alexa 647 were selected. For 3D quantifications, dLGN stacks were previously transformed to IsoData binary masks in ImageJ (IsoData algorithm) to standardize rendering in Imaris (Bitplane).
Image acquisition and analysis
Images from tissue sections were captured using an Olympus Fluoview FV1200 confocal microscope and FV10-ASW software (Olympus). Acquisitions from clarified brains were made using a light sheet microscope (LaVision Ultramicroscope II) and LaVision BioTec Inspector Pro (LaVision). 3D rendering and processing were carried out in Imaris 9.1.2 (Bitplane). Images were processed with Fiji/ImageJ software (Schindelin et al., 2012) in order to denoise, enhance, threshold, colocalize, or measure areas and distances, depending on the type of quantification. In all cases, background fluorescence was subtracted from sections using a rolling ball filter, and grayscale was renormalized so that the range of grayscale values was from 0 to 256. Data obtained from the image analysis were loaded into R 3.6.0 to perform mathematical calculations and generate some of the graphs. Statistical analysis and the rest of the graphs were carried out using GraphPad Prism version 6.00 (GraphPad Software). Figures and graphs were edited with Adobe Illustrator CS6 (Adobe). Error bars indicate ± SEM (**p < 0.01, ***p < 0.001, Student's unpaired t test).
Eye-specific segregation analysis
Binocular colocalization was computed overlapping the “IsoData” binary mask of the Z-stack for each channel. We selected the two (OPN, SCN) or three (dLGN) most central 70 µm sections from the nucleus. The result was normalized by the total area of the nucleus. R-distribution was carried out as described by Torborg and Feller (2004) computing the logarithm of the intensity ratio
Suprachiasmatic nucleus density and proximity analysis
All 60 µm coronal sections of the SCN were analyzed, aggregated by sample, and averaged by group. Since retinal projections to the SCN project bilaterally, on each side of the nucleus, we can find terminals labeled with the same CTB-Alexa conjugate coming from opposite retinas. Therefore, TdTm pixels colocalizing with both Alexa fluorophores were discarded because of the inability to assign them to a specific retinal input. For the ipsilateral/contralateral proximity analysis, regions with high CTB-Alexa accumulation were selected as an approximation to synaptic boutons. Ipsilateral boutons are those labeled by the corresponding CTB-Alexa conjugate and tdTm. Contralateral boutons are those positive for the other CTB-Alexa conjugate and lacking tdTm signal. In the computation of the density maps, terminals were measured by summing the binary mask for each Z-plane along all sections of the SCN, combining and averaging the results for each source retina based on the corresponding CTB-Alexa conjugate (Alexa 488 or 647). Samples were normalized for nucleus size and signal intensity prior to aggregating by group to generate a density map for each SCN side per condition.
RNA-seq and analysis
Retinas were dissected out from P4 and P8 Pou4f2-cre; Rosa26TdTm (control RGCs) and Pou4f2-Kir2.1; Rosa26TdTm (Kir2.1 RGCs) mice. Retinas for each genotype were isolated and enzymatically dissociated in a mixture of collagenase/trypsin and 1% BSA for 20 min at 37°C followed by mechanical dissociation. A 40 μM cell strainer was used for aggregate removal after dissociation, and TdTm-expressing RGCs were isolated by fluorescent-activated cell sorting (BD FACS Aria III, BD Biosciences). Total RNA extraction was performed with the Arcturus Picopure RNA Isolation Kit (Thermo Fisher Scientific). Three independent samples containing total RNA from 9 to 11 pooled retinas from at least three different litters/condition were sequenced according to the manufacturer instructions in a HiSeq Sequencing v4 Chemistry (Illumina). Briefly, RNA-seq reads were mapped to the mouse genome (Mus_musculus.GRCm.38.83) using STAR (v2.5.0c; Dobin et al., 2013). Quality control of the raw data was performed with FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Library sizes were between 17 and 23 million single reads. SAMtools (v1.3.1) was used to handle BAM files (Li et al., 2009). To retrieve differentially expressed genes (DEG), mapped reads were counted with HTSeq v0.6.1 (Anders et al., 2015) with the following parameters: -s reverse -i gene_id and with a gtf file reference for GRCm38.83. Samples were adjusted for library size and normalized with the variance stabilizing transformation (VST) in the R statistical environment using DESeq2 (v1.10.0) pipeline (Love et al., 2014). When performing differential expression analysis between groups, we applied the embedded IndependentFiltering procedure to exclude genes that were not expressed at appreciable levels in most of the samples. A heatmap of this distance matrix gives us an overview of similarities and dissimilarities between samples with a clustering based on the distances between the rows/columns of the distance matrix (FigS4). The PCA is generated using the most variable genes detected in the entire dataset. DESeq2 default values are set to use only the 500 most variable genes. This number is often applied when the transcription of protein-coding genes is analyzed (FigS4). Gene expression values were represented using the normalization techniques provided by each algorithm: Reads per Kilobase of Mapped reads (RPKM) and Relative Log Expression (rlog from DESeq2). Analysis and preprocessing of data were performed with custom scripts using R (https://cran.r-project.org/; v3.4.3) statistical computing and graphics and Bioconductor v3.2 (BiocInstaller 1.20.3; Huber et al., 2015). Genes were considered differentially expressed at Benjamini–Hochberg (BH) adjusted, padj < 0.1 and abs|Log2FC| > 1. Significantly upregulated and downregulated genes were visualized with IGV (v2.3.72; Thorvaldsdóttir et al., 2013). We performed an enrichment analysis for Gene Ontology using the platform Panther (Mi et al., 2019) Fisher's exact test and the padj correction were used, obtaining the top terms using the filters by ratio enrichment > 2, number of GO family group genes between 3 and 2,000, and number of enrichment genes >3.
In utero electroporation experiments
In utero electroporation of E13.5 retinas was performed using plasmids encoding short-harpin RNAs against Syt3 (shRNA Syt13) or control shRNA (scramble shRNA), along with GFP-plasmids, as previously described (García-Frigola et al 2007; Morenilla-Palao et al., 2020). Silencer siRNA Expression Vector was purchased in Ambion. The Syt13 RNAi target sequence used to downregulate Syt13 was 5′-GGCTGAGTTATTTGTGACA-3′. P9 and P14 electroporated mice were perfused transcardially with 4% PFA in 0.1 M PBS, and brains were incubated overnight in PFA at 4°C and then stored in PBS. Fixed brains were embedded in 4% agarose diluted in PBS and cut into 100 µm sagittal sections with a vibratome. The immunofluorescence of sections of the SC was done as above.
Images of the SC were captured with an Olympus FluoView FV1200 confocal microscope and FV10-ASW software (Olympus; z-step size of 3 µm, z-stack of 18 µm for P9 mice, and z-stack of 39 µm for P14 mice).
Analysis of axon terminals occupancy in the SC of electroporated mice
Analysis was performed using Fiji/ImageJ software (Schindelin et al., 2012). For each animal, we selected the three most central sections of the SC, and for each section, we obtained the maximum intensity projection of the GFP channel to do the analysis. In each section, we drew three regions of interest of the same size (ROI; 103 µm × 198 µm) in the center of the area marked by the electroporation. The image was thresholded, and the area fraction of GFP positive pixels was calculated. In addition, we drew another ROI (95 µm × 128 µm) at the entrance of RGC axons to the SC, and we did the same process. The area fraction of the three first ROIs was normalized to the area fraction of the axons ROI, and we calculated the mean normalized area fraction per section and finally the mean normalized area fraction per mouse. For the statistical analysis, we assessed the normality and homogeneity of variances of the data, and an unpaired nonparametric Mann–Whitney test was performed. The level of significance is presented in the plots, using asterisks, by the usual convention, i.e., *p < 0.05, **p < 0.01.
Results
Generation and characterization of mice ectopically expressing Kir2.1 in RGCs
To investigate the activity-dependent genetic program that underlies axonal remodeling at the visual targets and a potential function of spontaneous retinal waves in NIFN, we sought to create a mouse line with disrupted embryonic retinal activity. To achieve this, we first developed a transgenic line wherein the rectifying potassium channel hKir2.1 was fused to the fluorescence protein eYFP and contained a CAG promoter and a stop sequence to regulate hKir2.1 expression in a Cre-dependent manner (Kir2.1eYFPflxStop). These mice were crossed with a Pou4f2-cre line enabling gene expression specifically in RGCs (Fuerst et al., 2012), resulting in double mutant mice (Pou4f2-cre;Kir2. 1eYFPflxStop, renamed Pou4f2-Kir2.1 from now on; Fig. 1A). The analysis of retinal sections from newborn Pou4f2-Kir2.1 mice confirmed hKir2.1eYFP expression in RGCs (Fig. 1B,C). Then, to visualize RGCs ectopically expressing Kir2.1 and perform patch-clamp electrophysiological recordings on these cells, we bred Pou4f2-Kir2.1 mice with a reporter line (Rosa26TdTomato (Tm)). As expected, the resting membrane potential of RGCs recorded from postnatal Pou4f2-Kir2.1; Rosa26TdTm mice (−50.7 ± 1.4 mV) was significantly lower than in neurons from control mice (Pou4f2-Rosa26TdTm; −45.5 ± 1.9 mV; Fig. 1D).
Pou4f2-Kir2.1 mice ectopically express Kir2.1 in retinal ganglion cells. A, The Pou4f2-cre mice were bred with a conditional reporter line that contains a CAG-[Stop] cassette followed by the human Kir2.1 cDNA sequence fused to the eYFP coding sequence. B, Immunostaining for Kir2.1 and eYFP in retinal sections from adult Pou4f2-Kir2.1 and control littermate mice. C, Immunostaining for Kir2.1 combined with calbindin or Isl2 in P9 Pou4f2-Kir2.1 mouse retinas. D, Averaged resting membrane potential in control cells and cells expressing Kir2.1. The dots represent tdTomato cells from the retinas of three different mice. Whiskers extend to the min and max values (**p < 0.01, Student's unpaired t test).
Next, we aimed to evaluate the impact of Kir2.1 ectopic expression in retinal waves by allowing their visualization through genetically encoded calcium indicators. To this end, Pou4f2-Kir2.1 mice were crossed with a Cre-dependent GCaMP6f calcium indicator reporter mouse line (Rosa26-stoploxP-GCaMP6f) obtaining Pou4f2-Kir2.1; GCaMP6f mice (Fig. 2A,B). We recorded semi-intact whole-mount retinal preparations from P4 Pou4f2-Kir2.1; GCaMP6f mice and counterpart controls (Pou4f2; GCaMP6f mice). We observed correlated retinal waves of activity in Pou4f2-Kir2.1; GCaMP6f, but they were altered compared with control retinas. The distribution of their origin was similar in both genotypes, but their number was significantly reduced in the mutant retinas, and the duration was longer compared with the control retinas (Fig. 2C–H, Movies 1, 2).
The frequency and duration of retinal waves are altered in the Pou4f2-Kir2.1 mice. A, Pou4f2-cre mice were bred with the conditional Kir2.1 and the reporter Rosa26-GCaMP6f lines to obtain Pou4f2; GCaMP6f mice expressing Kir2.1 and GCaMP6f in retinal ganglion cells. B, Representative image of a flattened retinal preparation from a P4 Pou4f2;GCaMP6f mouse. Retinal waves can be visualized in real time by dynamic GFP fluorescence. Scale bar, 300 µm. C, The top images represent the average of ΔF/F traces signifying calcium transients from all ROIs, in control and Pou4f2-Kir2.1 retinas. The middle images show raster plots depicting calcium transients for all ROIs with ΔF/F signals >50%. Images at the bottom show the percentage of ROIs that were active at any given time during the next 10 min after plating. D, Representation of individual waves in Pou4f2-Kir2.1 and control retinas in the first minute after plating. Each color is a different wave. E, The top left panel delineates a representative retina for each genotype. The dots represent the location where a particular wave originated (initiation site). The histograms at the top right indicate the distribution of initiation sites along the vertical dimension of the retina, and the histograms at the bottom left show the same in the horizontal dimension. The histograms on the bottom right show the distribution of inter-transient intervals and the text indicates the mean value in each case. All histograms also show overlayed a kernel density estimate providing a smoother representation of the distributions. F, Mean amplitude of the global fluorescence signal in Pou4f1-Kir2.1 and control mice. G, Number of waves detected as fluorescence peaks in 10 min recordings. H, Duration of the waves. I, Mean of inter-transient intervals (ITIs). J, Number of waves in the first minute of recording. K, Density of initiation sites measured as number of sites per square millimeters. Two-tailed unpaired t test (*p < 0.05; ***p < 0.005). Results show means ± SEM.
Next, we analyzed the projection pattern of ipsi- and contralateral RGC terminals at the dLGN and the SC of P9 Pou4f2-Kir2.1 and control (Kir2.1eYFPflxStop) mice by injecting cholera toxin subunit B conjugated with far-red (CTB-Alexa 647) or green (CTB-Alexa 488) fluorophores into each eye at P9 (Fig. 3A). Two days later, when axons are already refined but eyes are still not open, anterograde axonal tracing was monitored in both iDISCO-clarified brains and coronal sections. As expected, and in agreement with previous reports, segregation between ipsilateral and contralateral terminals was almost finished in the dLGN of control mice, and a well-defined ipsilateral territory with minimal presence of contralateral terminals was clearly observed (Godement et al., 1984; Upton et al., 1999; Pham et al., 2001; Fig. 3B,D,D’). In contrast, contralateral terminals invaded the ipsilateral region in the dLGN of Pou4f2-Kir2.1 mice (Fig. 3C,E,E’,G,J), resulting in an increased unsegregated area (Fig. 3F,G). The area occupied by ipsilateral terminals was larger, but this was accompanied by a reduction of the area innervated mainly by ipsilateral RGCs (Fig. 3H,I,J). The analysis of RGC axons at the SC also showed an aberrant refinement of ipsilateral RGC arborizations in Pou4f2-Kir2.1 mice compared with their control littermates (Extended Data Fig. 1-1). However, Pou4f2 is also expressed in the SC, and therefore, we cannot rule out that axon refinement defects in this nucleus result from activity alterations in the collicular cells. Despite this caveat in the SC, the lack of refinement in the thalamus confirms that a correct patterning of spontaneous retinal waves is essential for eye-specific refinement at the IFN and validate the Kir2.1eYFPflxStop mouse line as a model to investigate the impact of correlated spontaneous activity in other visual nuclei where Pou4f2 is not expressed.
Figure 1-1
Pou4f2-Kir2.1 mice show eye specific segregation defects in the SC. (A) Scheme illustrating the experimental approach consisting in intraocular injections of CTB fused to different Alexa fluorophores for each eye and analysis of coronal sections at the level of the SC (yellow). (B) Sagital sections from P11 Pou4f2-Kir2.1 and control mice showing contralateral and ipsilateral projections at the level of the SC. Right panels show only the ipsilateral projections. (C) Quantification of the area occupied by ipsilateral terminals in the Stratum Griseum Superficialis (SGS). Each biological replicate (red dots) represents mean area values from three central sagittal sections of one SC/animal. Two-tailed unpaired t test (***p < 0.001). Results show means ± SEM. Download Figure 1-1, TIF file.
Pou4f2-Kir2.1 mice show eye specific segregation defects in the dLGN. (A) Retinal projections were labeled through intraocular injections of cholera toxin subunit B conjugated to different Alexa fluorophores (CTB-Alexa647 and CTB-Alexa488) for each eye. (B–C) 3D reconstructions of the dLGN of P11 Pou4f2-Kir2.1 and control littermate mice injected with CTB-Alexa647 and CTB-Alexa488. Overlapping of ipsilateral and contralateral terminals is visualized in white. (D–E) Coronal sections through the dLGN of P11 Pou4f2-Kir2.1 and control littermate mice injected with CTB-Alexa647 and CTB-Alexa488. Overlapping of ipsilateral and contralateral terminals is visualized in white. (D'–E') Contralateral terminals shown in D and E. (D''–E'') Ipsilateral terminals shown in D and E. (F, G) R-distribution (log[ipsilateral/contralateral]) representing contralateral dominant area in red, unsegregated in green and ipsilateral dominant in blue. (H) Percentage of the volume containing ipsilateral and contralateral terminals relative to the total volume occupied by RGC terminals at the dLGN. (I) Percentage of the dLGN area occupied by both ipsilateral and contralateral retinal terminals (unsegregated area) relative to the total area occupied by RGC terminals. (J) Percentage of the dLGN area occupied by ipsilateral projections related to the total area occupied by RGC terminals. (K) Percentage of the dLGN area mostly occupied by ipsilateral terminals. Each biological replicate (red dots) represents an animal. Two-tailed unpaired t test (***p < 0.001). Results show means ± SEM. (L) Histograms averaging by group of mice (left) or with individual samples (right) showing the values of the R distribution.
Transcriptional changes governing spontaneous-dependent refinement in the visual targets
Together with previous reports (Bito et al., 1997; West et al., 2001), our results indicate that retinal waves and the associated calcium influx into the cytoplasm influence the expression of genes that integrate the machinery required for axonal remodeling and synapse formation. However, the molecular programs that act downstream of spontaneous activity to mediate RGC axon remodeling at the target nuclei have not been characterized. Thus, to identify molecular mechanisms underlying this activity-dependent process, we took advantage of the Kir2.1 conditional mice and compared the transcriptional profile of RGCs from Kir2.1-expressing and control mice. For this, we isolated the retinas of Pou4f2-Kir2.1; Rosa26TdTm and control Pou4f2; Rosa26TdTm mice at two time points: P4 when most RGC axon arbors are remodeling at the visual nuclei and P8 when a large number of terminals have already refined (Fig. 4A). Tomato labeled RGCs from Pou4f2-Kir2.1 or control mice were purified by fluorescence-activated cell sorting (FACS), and their transcriptome profile was determined by high-throughput sequencing (RNA-seq). Principal component analysis (PCA) confirmed the differential grouping of samples from each genotype at both stages (Figs. 2–4).
Identification of activity-dependent transcriptional programs underlying axon remodeling during axon refinement. A, Scheme of the experimental approach. RGCs from P4 Pou4f2-Kir2.1 and control littermate mice were purified by FACS, and their total RNA was extracted to compare their transcriptomic profiles. B, Heatmaps of gene transcriptional profiles of RGCs from Pou4f2-Kir2.1 and control mice highlighting common and specific DEGs at P4 and P8. C, Volcano plots showing the significance value distribution after DEGs analysis in Pou4f2-Kir2.1 samples referred to controls at both P4 and P8 (padj < 0.1); |LFC|>1 = 217 genes. D, Heatmaps of upregulated and downregulated DEGs in P4, P8, or common to both stages comparing Pou4f2-Kir2.1 with control retinas. E, Biological processes GO enrichment analysis of DEGs from genes with altered expression in P4, P8, or both in Pou4f2-Kir2.1 mice compared with controls. padj < 0.1. F–H, Gene profiles and graphs representing transcripts expression obtained in Pou4f2-Kir2.1 and control mice corresponding to three DEGs, Adcy1, Sema6a, and Syt13 at both stages.
Figure 4-1
(A) Pearson correlation matrix between normalized RNA-seq samples clustered by Euclidian dendrogram. (B) Principal component analysis of RNA-seq samples at P4 and P8. Download Figure 4-1, TIF file.
Table 4-1
Sheet 1. Differential expression analysis of WT vs mutants samples a P4 and P8) Sheet 2. Statistically significant DEGs (Padj <0,1) Sheet 3. Gene Ontology analysis of DEGs WT vs mutant samples at P4 and P8 Sheet 4. Differential expression analysis of WT samples at P4 and P8 (Statistically significant DEGs (Padj <0,1) Sheet 5. Quality controls in the alignment of the raw data samples during the analysis of the data. Download Table 4-1, XLSX file.
The analysis of transcriptional profiles from P4 and P8 Pou4f2-Kir2.1; Rosa26TdTm mice and their respective controls (Fig. 4B) revealed distinct misregulated transcriptional programs at each stage. The number of differentially expressed genes (DEG) between mutant and control retinas was lower at P4 (468) than at P8 (1,533). At P4, 232 and 236 were upregulated and downregulated, and at P8, 895 and 638 were downregulated and upregulated, respectively (Fig. 4B,C; Extended Data Table 4-1). The profile differences found at the two stages likely reflect the switch from cholinergic to glutamatergic-dependent axonal remodeling known to occur during this time window. In a Gene Ontology (GO) analysis for biological processes, terms such as synapse formation and clustering of AMPA receptors (including genes such as Chrdl1, Nlgn1, Dlg4, and Shank3 encoding for proteins essential for synaptogenesis) were differentially expressed between the Kir2.1 mutants and the controls at P4 (Fig. 4D,E; Extended Data Table 4-1). However, at P8, we retrieved terms generally associated with mature synapse stages such as synaptic assembly and maintenance, regulation of NMDA and AMPA receptor activity, or transmission of spontaneous activity. Among the DEGs found at this stage, we found Cacng2/4/7/8, members of Type I transmembrane AMPA receptors regulatory proteins, Shisa8, involved in the regulation of AMPA receptors and Nrxn1, important for the formation and maintenance of synaptic junctions via their calcium-dependent interactions with Neuroligin family members (Fig. 4D,E; Extended Data Table 4-1).
Interestingly, among the genes exclusively downregulated at P4, we found Adcy1 that encodes for adenylate cyclase 1, an enzyme involved in the refinement of visual axons at the dLGN and the SC (Nicol et al., 2006b; Fig. 4F). In the set of DEGs specifically downregulated at P8, we found Sema3B, known to play an important function in neuroplasticity regulation right after eye-opening (Carulli et al., 2021; Fig. 4G). Among the DEGs highly downregulated at both stages, we identified Syt13 that encodes for synaptotagmin 13, a Type I membrane protein involved in vesicular trafficking, exocytosis, and secretion (Fukuda and Mikoshiba, 2001; von Poser and Südhof, 2001; Fig. 4H).
Syt13 was downregulated at both stages, and we aimed to manipulate the expression of this gene in vivo as a proof of concept to validate our screening. First, we analyzed the expression pattern of Syt13 mRNA by in situ hybridization in retinal sections from postnatal mice and confirmed that Syt13 transcripts are highly and specifically expressed in P4 RGCs and less expressed after eye-opening (Fig. 5A,B). Then, to investigate a potential function of Syt13 in axon refinement, we electroporated short-hairpin RNAs against Syt13 (shRNA Syt13) or control shRNA (scramble shRNA) together with plasmids encoding for the green fluorescence protein (GFP) into embryonic retinas (E13.5) as described by Garcia-Frigola et al., (2007) and analyzed the trajectories of the targeted RGCs at different time points (Fig. 5C). At perinatal stages, axons from RGCs targeted with shRNA Syt13 or control shRNA plasmids both grow normally along the visual pathway and reached the dLGN and SC in a similar manner (Fig. 5D,E,G). However, axons from shRNA Syt13 targeted RGCs showed significant differences with control-electroporated mice when projecting into the SC. Although labeled axons reached their corresponding topographical area at the SC in both conditions, axons electroporated with shRNA Syt13 were significantly less elaborated, covering less area than control axons at P9 (Fig. 5E,F). This phenotype was also very evident at eye-opening (P14; Fig. 5G,H). These results suggest a role for Syt13 in axon RGC remodeling before eye-opening and reinforce our unbiased screening.
Syt13 is involved in axonal remodeling during the formation of the visual circuit. A, B, Representative image of a retinal section from P4 and P15 wild-type mice showing in situ hybridization for Syt13. GCl, ganglion cell layer; INl, inner layer; ONl, outer layer. C, Schematic representation of the experimental approach. D, 3D view of axons from RGCs electroporated with control shRNA or Syt13shRNA show similar phenotypes as they pass the thalamus. E, Representative examples of sagittal sections through the superior colliculus (SC) of P9 mice electroporated with plasmids encoding for GFP and control shRNA or Syt13 shRNA and counterstained with DAPI. The right panels (E’, E’’) show high-magnification images from the squared areas. F, Area covered by GFP signal from axon terminals of RGCs electroporated with GFP plus Syt13shRNA or control shRNAs at the SC (a.u., arbitrary units). G, Representative examples of sagittal sections through the superior colliculus (SC) of P14 mice electroporated with plasmids encoding for GFP and control shRNA or Syt13 shRNA and counterstained with DAPI. The right panels (G’, G’’) show high-magnification images from the squared areas. H, Area covered by GFP signal from axon terminals of RGCs electroporated with GFP plus Syt13shRNA or control shRNAs at the SC (a.u., arbitrary units). The red dots represent the normalized mean area covered by GFP axons obtained from averaging results of three consecutive sagittal sections from the medial part of the SC (unpaired nonparametric Mann–Whitney test, **p-value < 0.01). Whiskers extend to the min and max values.
Altered frequency of retinal waves disrupts axon terminal refinement at the NIFN
Retinal activity is known to encode spatial information among neighboring RGCs. Since NIFN are not as functionally related to spatial information and sight as IFN, we wondered to what extent, if at all, the alteration of spontaneous retinal waves affects the refinement of visual terminals in these nuclei. To address this, we analyzed axonal terminals at two NIFN, the olivary pretectal nucleus (OPN) and the suprachiasmatic nucleus (SCN).
The OPN is the largest of the seven nuclei that integrate the pretectal nuclei located in a region anterior to the SC. It is involved in mediating behavioral responses to acute changes in ambient light and the optokinetic reflex (Gamlin, 2006). We analyzed serial coronal sections from Pou4f2-Kir2.1 and control mice previously injected with CTB-Alexa 647 or CTB-Alexa 488 into each eye (Fig. 6A). A larger number of ipsi- and contralateral terminals overlapped in space in the OPN of Pou4f2-Kir2.1 mice compared with control littermates (Fig. 6B–F). These results demonstrate segregation defects in this NIFN as a consequence of disturbed retinal spontaneous activity.
Altered frequency of retinal waves disturbs eye-specific segregation in the OPN. A, Retinal projections were labeled through intraocular injections of cholera toxin subunit B fused to different Alexa fluorophores (CTB-Alexa 647 and CTB-Alexa 488) for each eye. B–E’’, 3D reconstructions (left panels) and coronal sections (right panels) through the OPN of P11 Pou4f2-Kir2.1 and control mice injected with CTB-Alexa 647 and CTB-Alexa 488. The overlapping of ipsilateral and contralateral terminals is visualized in white. F, Percentage of the area containing ipsilateral and contralateral terminals (white) relative to the total area occupied by RGC terminals. Each biological replicate (red dots) represents an animal. Two-tailed unpaired t test (***p < 0.001). Results show means ± SEM.
The SCN is located in the diencephalon just behind the optic chiasm. This NIFN nucleus, considered the circadian pacemaker, is responsible for controlling the circadian rhythm and is innervated only by intrinsically photosensitive RGCs that express melanopsin (OPN4), distribute all over the retina (Sekaran et al., 2003; Prigge et al., 2016), and can be activated by the retinal waves (Renna et al., 2011). Most Pou4f2-RGCs do not project to the SCN (Chen et al., 2011). Thus, to investigate the function of retinal spontaneous activity in these nuclei, we used a second Cre line (Slc6a4/Et33-Cre) that drives gene expression to a subset of RGCs that mainly project to the ipsilateral IFN (García-Frigola and Herrera, 2010) but also innervate the SCN (Su et al., 2021).
To visualize Slc6a4-positive RGCs and their terminals in the SCN nuclei, we bred the Slc6a4/Et33-Cre line with the Rosa26TdTm reporter line. While intrinsically photosensitive retinal ganglion cells make up approximately 2% of the total RGC population (Berson et al., 2002; Hattar et al., 2002; Do and Yau, 2010; McNeill et al., 2011; Schmidt et al., 2011), there was a twofold enrichment of OPN4-positive cells (∼4%) within the Slc6a4-RGCs population, and 10% of the OPN4 cells were positive for Slc6a4 (Fig. 7A). Next, we analyzed the distance between axon terminals in the SCN coming from the ipsilateral or contralateral eye in the Slc6a4-Kir2.1; Rosa26TdTm and control mice. Far from being segregated, we noticed that after injecting CTB-Alexa 488 and CTB-Alexa 647 into the eyes of Slc6a4-TdTm mice, ∼90% of the Slc6a4 terminals coming from one eye (TdTm/CTB-Alexa 488) were located only 5 μm apart from axons arising in the opposite eye (CTB-Alexa 647; Fig. 7B–D). This result confirms previous studies showing that melanopsin RGC terminals do not undergo eye-specific segregation in the SCN and different subtypes of SCN cells indeed receive input from both retinas (Fernandez et al., 2016).
Altered frequency of retinal waves leads to defects in the refinement of visual axons in the SCN. A, A’, Whole-mount retina from P11 Slc6a4-Rosa26TdTm mice stained with αOPN4 and αtdTomato antibodies. The yellow arrowheads point to double-stained cells. B, Injections of CTB fused to Alexa-488 or Alexa-647 Slc6a4-Rosa26TdTm mice and analysis of coronal sections at the level of the SCN (yellow) 2 d later (P11). Retinal projections from each eye bilaterally innervate the SCN. C, Representation of retinal terminals from control mice displaying ipsilateral terminals colored according to their distance to the nearest contralateral neighbor. Contralateral terminals are shown in gray. Representative image of ipsilateral (C’) and contralateral (C’’) terminals used to create the mask shown in C. D, Percentage of ipsilateral terminals closer than 5 µm to a contralateral. E, Area occupied by TdTm terminals in control and Slc6a4-Kir2.1 mice at both sides of the SCN expressed as fold-change. F, Density maps depicting projections from Slc6a4-RGCs in the suprachiasmatic nucleus (SCN) of both Slc6a4-Kir2.1 and control mice. These maps were generated by averaging data from all replicates for each condition. The percentages at the top of the maps indicate the proportion of axonal projections on each side of the SCN (ipsilateral or contralateral) with respect to the combined total of both sides. G, Ratio of projections covering the ventrolateral quadrant of the SCN for each replica on both sides of the SCN. Each biological replicate (red dots) represents an animal. Two-tailed unpaired t test (***p < 0.001). Results show means ± SEM.
Figure 7-1
Slc6a4-Kir2.1 mice show impaired eye-specific segregation in the dLGN and SC. (A) Averaged resting membrane potential in control cells and cells expressing hKir2.1. Dots represent TdTomato positive cells from retinas from three different mice. Whiskers extend to the min and max values (*p < 0.05, ** p < 0.01, Student’s unpaired t-test) (B, C) 3D reconstructions and (D, E) coronal sections through the dLGN of P11 Slc6a4-Kir2.1 mice and control littermate mice injected with CTB-Alexa647 and CTB-Alexa488. Overlaping of ipsilateral and contralateral terminals is visualized in white. (D’-E’) R-distribution (log[ipsilateral/contralateral]) representing contralateral dominant area in red, unsegregated in green and ipsilateral dominant in blue. (F) Percentage of the area containing ipsilateral and contralateral terminals relative to the total area occupied by RGC terminals at the dLGN. (G) Percentage of the area with a majority of ipsilateral terminals relative to the total area occupied by RGC terminals. (H) Percentage of the area occupied by ipsilateral projections with respect to the total area occupied by RGC terminals. (I) Sagital sections from P11 mice at the level of the SC from Slc6a4-Kir2.1 and control mice in contralateral and ipsilateral projections. Ipsilateral projections are also shown in B&W at the right panel. (J) Quantification of the area occupied by ipsilateral terminals in the Stratum Griseum Superficialis (SGS). Each biological replicate (red dots) represents mean values from three central coronal (dLGN) or sagittal (SC) sections from one nucleus/animal. (K) 3D reconstruction and coronal sections through the OPN of P11 Slc6a4-Kir2.1 and control mice injected with CTB-Alexa647 and CTB-Alexa488. Overlapping of ipsilateral and contralateral terminals is visualized in white. (L) Percentage of the area containing ipsilateral and contralateral terminals (white) relative to the total area occupied by RGC terminals. Each biological replicate (red dots) represents an animal. Two-tailed unpaired t test (***p < 0.001). Results show means ± SEM. Download Figure 7-1, TIF file.
Representative time-lapse of calcium waves in a flattened whole-mount retina from a Pou4f2; GCaMP6f P4 mouse. The movie represents 2 min 30 s of recording at 1.8 times the speed. Scale bar, 300 µm. [View online]
Representative time-lapse of calcium waves in a flattened whole-mount retina from a Pou4f2; Kir2.1eYFPflxStop; GCaMP6f P4 mouse. The movie represents 2 min 30 s of recording at 1.8 times the speed. Scale bar, 300 µm. [View online]
Then, to investigate the influence of spontaneous waves in the wiring of visual input at the SCN, we injected Alexa dyes into the eyes of Slc6a4-Kir2.1; Rosa26TdTm (Slc6a4-Kir2.1) and control littermate mice (Slc6a4-Rosa26TdTm). First, we confirmed that the resting membrane potential of RGCs recorded from postnatal Slc6a4-Kir2.1 (−55.1 ± 1.5 mV) was significantly lower than in control mice (Slc6a4-Rosa26TdTm; −46.4 ± 2.6 mV; Figs. 3–7). Also, the RGC axon terminals of Slc6a4-Kir2.1 mice do not refine properly at the LGN and the SC indicating that retinal activity is also perturbed in these mice (Figs. 3–7). The segregation of ipsi- and contralateral terminals at the OPN was unaffected in Slc6a4-Kir2.1 mice and similar to controls (Figs. 3–7). However, we found significant differences in the SCN. While Slc6a4 terminals (TdTm/CTB-Alexa 488) covered similar areas at the SCN of Slc6a4-Kir2.1 and control mice (Fig. 7E), density map analysis demonstrated that arborizations were significantly reduced in the Slc6a4-Kir2.1 mice compared with the controls in the ventrolateral area of this nucleus (Fig. 7F,G). These results suggested that, despite eye-specific segregation and topographic mapping are not evident features of the SCN, spontaneous retinal activity before eye-opening influences the arborization of RGCs terminals also in this NIFN.
Discussion
Spontaneous neural activity is essential in a multitude of processes in the developing nervous system. Spontaneous calcium transients have been detected in the neocortex and thalamus in prenatal stages and may modulate neurogenesis, wiring, and neurotransmitter identity (Borodinsky et al., 2004; Corlew et al., 2004; Bonetti and Surace, 2010; Antón-Bolaños et al., 2019). In the visual system, correlated spontaneous retinal activity is known to be essential for the fine-tuning of retinotopic maps and eye-specific segregation in the IFN, but it was unclear to what extent the correct connectivity of RGC axons at NIFN also depends on retinal waves. In this study, by generating several mouse lines with disturbed retinal waves before eye-opening and analyzing the refinement of RGCs axon terminals in the SCN and the OPN, we found that proper frequency of retinal waves in pre–eye-opening stages is essential for the establishment of a correct connectivity also in NIFN. In addition, we retrieved transcriptional programs triggered by retinal activity mediating the assembly of the visual circuit.
Ectopic expression of Kir2.1 in RGCs alters the frequency and/or duration of retinal waves
The rectifying potassium channel Kir2.1 is a known suppressor of intrinsic excitability, which has a crucial role in establishing the connectivity of several circuits. Overexpression of Kir2.1 has been used to electrically silence neurons in both in vitro and in vivo experiments (Burrone et al., 2002; Benjumeda et al., 2013). Moreover, ectopic expression of Kir2.1 in transgenic mice has been reported to block neural activity in the olfactory system (C. R. Yu et al., 2004) and may lead to different effects in vivo, depending on the developmental stage of the network (Burrone et al., 2002; Moreno-Juan et al., 2017; Antón-Bolaños et al., 2019). We observed that recombinant RGCs in the Kir2.1 mice exhibit a more negative resting potential than RGCs in the control mice. This translated into a lower frequency and extended duration of retinal waves, which is sufficient to alter the refinement of retinal projections at the different visual targets.
Transcriptional programs involved in activity-dependent axonal remodeling
Pou4f2-Kir2.1 mice show significant defects in fine-scale/local refinement of axonal arborizations at the different visual nuclei, confirming that the mechanisms involved in the remodeling of visual axons are influenced by retinal waves. By comparing the transcriptomic profile of Pou4f2-Kir2.1 and control retinas at P4 and P8, we identified genes (DEGs) potentially regulated by retinal activity during postnatal stages before eye-opening. Interestingly, these DEGs seem to reflect the transcriptional changes associated with the transition of RGCs switching from a scenario dominated by axon branching dynamics to an assembled circuit carrying out synaptic maturation. Adcy1 was already described as important for retinal axon refinement (Nicol et al., 2006a), and we found this gene among the DEG at P4. Nlgn1, which encodes for a protein important for the recruitment and clustering of other proteins during synaptogenesis, is also differentially expressed at this stage according to our screen. At a later stage (P8), when connections are more mature, we found neurexin 1 downregulated in the Kir2.1 mutant retinas. Neurexins and neuroligins form complexes at the synapses during synaptogenesis, and our results suggest that the expression of these genes that encode proteins essential for synapse formation depends, at least partially, on proper patterns of retinal waves.
Interestingly, we also found a number of DEGs related to the cell cycle. It has been recently reported that beyond its role in DNA damage response, the serine/threonine kinase ATR plays a function in regulating neuronal activity by modulating presynaptic firing. ATR interacts with synaptotagmins in synaptosomes, and we found synaptotagmin13, one of the less studied members of the synaptotagmin family (Wolfes and Dean, 2020), altered in the Kir2.1 mice. This could explain, at least in part, the “double-strand break repair via break-induced replication” GO term retrieved in our analysis.
Unlike other members of the Syt family, Syt13 does not sense calcium (Bhalla et al., 2008), but we observed that Syt13 downregulation leads to altered axonal arborizations at the SC. These results revealed a novel role for this poorly known member of the Syt family. The loss-of-function results from electroporation appear contradictory to the screening data. Kir2.1 retinal ganglion cells (RGCs) show excessive arborization in the visual nuclei. Syt13 expression is reduced in Kir2.1 mice, suggesting that Syt13 downregulation would produce similar phenotypes to Kir2.1 mutants (excess overlap and arborization, as observed in adenylate cyclase 1 mutants). However, the opposite phenotype was observed instead. These results could be partially explained if Syt13 is involved in arborization in a spontaneous activity-independent manner. However, further experiments are necessary to precisely determine Syt13’s function.
Particularities of retinal activity-dependent refinement in NIFN
A large number of studies have demonstrated that ipsilateral and contralateral axon terminals initially overlap in the dLGN and are later expelled from the territory that corresponds to the other eye, competing with the contrary retinal input in terms of activity correlation. We observed eye-specific segregation defects in the IFN of both Pou4f2-Kir2.1 and Slc6a4-Kir2.1 mice. As in the IFN, when activity is bilaterally altered in RGCs (Pou4f2-Kir2.1 mice), the terminals do not refine properly in the OPN, similar to what occurs in the IFNs. Therefore, despite the anatomical and functional differences between NIFNs and IFNs, RGC terminals seem to follow similar refinement rules in NIFNs. We do not detect refinement defects in the OPN of Slc6a4-Kir2.1 mice, but this could be due to the sequential innervation of contralateral and ipsilateral terminals and/or the fact that ipsilateral axons reaching this nucleus are very few and their arborization quite sparse, making it difficult to detect a potential defect.
The scenario in the SCN is different since eye-specific segregation does not occur in this nucleus. RGC axons project bilaterally (Fernandez et al., 2016), and arborizations coming from one eye are likely not abundant enough to compete with those coming from the other eye. The SCN was thought to be just a regulator of the circadian clock. However, retinal input to the SCN encodes for much more complex visual information than that necessary to just perform this function (Mouland et al., 2017; Stinchcombe et al., 2017; Y-Q. Yu et al., 2017), suggesting a potential role for these nuclei in spatial vision as well. Our experiments indicating that the correct frequency of retinal waves is required for the proper connectivity of this nucleus support this hypothesis. Future experiments in conditional mice with altered retinal activity in all melanopsin RGCs should better define the role of spontaneous activity in the refinement of retinal input to the SCN as well as a potential impact on mice behavior.
Data Availability
The RNA-seq datasets generated in this study are deposited at the NCBI Gene Expression Omnibus (GEO) with accession code GSE193498. Go to https://www.ncbi.nlm.nih. Enter token ynirkwgshryzlkt into the box. All other included data in this study are available from the corresponding authors upon reasonable request. Analysis scripts are available on reasonable requests.
Footnotes
The laboratory of E.H. is funded with the following grants: PID2022-138245NB-I00 from the National Grant Research Program, PROMETEO Program (PROMETEO/2020/007) from Generalitat Valenciana, and HR21-00824 from La Caixa Foundation. The laboratory of A.G. is funded by PID2022-140961OB-I00 from the National Grant Research Program and GVA PROMETEO/2021/031. We also acknowledge the financial support received from the “Severo Ochoa” Program for Centers of Excellence in R&D (CEX2021-001165-S).
↵*S.N. and C.M-P. contributed equally to this work.
The authors declare no competing financial interests.
- Correspondence should be addressed to Eloísa Herrera at e.herrera{at}umh.es.
