Corrigendum: Electrical synapses for a pooling layer of the convolutional neural network in retinas.

[This corrects the article DOI: 10.3389/fncel.2023.1281786.].

Electrical synapses for a pooling layer of the convolutional neural network in retinas.

We have an example of a synergetic effect between neuroscience and connectome via artificial intelligence. The invention of Neocognitron, a machine learning algorithm, was inspired by the visual cortical circuitry for complex cells to be made by combinations of simple cells, which uses a hierarchical convolutional neural network (CNN). The CNN machine learning algorithm is powerful in classifying neuron borderlines on electron micrograph images for automatized connectomic analysis. CNN is also useful as a functional framework to analyze the neurocircuitry of the visual system. The visual system encodes visual patterns in the retina and decodes them in the corresponding cortical areas. The knowledge of evolutionarily chosen mechanisms in retinas may help the innovation of new algorithms. Since over a half-century ago, a classical style of serial section transmission electron microscopy has vastly contributed to cell biology. It is still useful to comprehensively analyze the small area of retinal neurocircuitry that is rich in natural intelligence of pattern recognition. I discuss the perspective of our study on the primary rod signal pathway in mouse and macaque retinas with special reference to electrical synapses. Photon detection under the scotopic condition needs absolute sensitivity but no intricate pattern recognition. This extreme case is regarded as the most simplified pattern recognition of the input with no autocorrelation. A comparative study of mouse and macaque retinas, where exists the 7-fold difference in linear size, may give us the underlying principle with quantitative verification of their adaptational designs of neurocircuitry.

Multiple Invagination Patterns and Synaptic Efficacy in Primate and Mouse Rod Synaptic Terminals.

Purpose: Optical retina images are scaled based on eye size, which results in a linear scale ratio of 10:1 for human versus mouse and 7:1 for macaque monkey versus mouse. We examined how this scale difference correlates with the structural configuration of synaptic wiring in the rod spherule (RS) between macaque and mouse retinas compared with human data. Methods: Rod bipolar cell (BC) dendrites and horizontal cell (HC) axonal processes, which invaginate the RS to form synaptic ribbon-associated triads, were examined by serial section transmission electron microscopy. Results: The number of rod BC invaginating dendrites ranged 1∼4 in the macaque RS but only 1∼2 in the mouse. Approximately 40% of those dendrites bifurcated into two central elements in the macaque, but 2% of those dendrites did in the mouse. Both factors gave rise to 10 invagination patterns of BC and HC neurites in the macaque RS but only two in the mouse. Five morphological parameters: the lengths of arciform densities and ribbons, the area of the BC-RS contact, and the surface areas of BC and HC invaginating neurites, were all independent of the invagination patterns in the macaque RS. However, those parameters were significantly greater in the macaque than in the mouse by ratios of 1.5∼1.8. Conclusions: The primate RS provides a more expansive BC-RS interface associated with the longer arciform density and more branched invaginating neurites of BCs and HCs than the mouse RS. The resulting greater synaptic contact area may contribute to more efficient signal transfer.

Helical Fasciculation of Bipolar and Horizontal Cell Neurites for Wiring With Photoreceptors in Macaque and Mouse Retinas.

Purpose: The three-dimensional configurations of rod and cone bipolar cell (BC) dendrites and horizontal cell (HC) processes outside rod and cone synaptic terminals have not been fully elucidated. We reveal how these neurites are mutually arranged to coordinate formation and maintenance of the postsynaptic complex of ribbon synapses in mouse and monkey retinas. Methods: Serial section transmission electron microscopy was utilized to reconstruct BC and HC neurites in macaque monkey and mouse, including metabotropic glutamate receptor 6 (mGluR6)-knockout mice. Results: Starting from sporadically distributed branching points, rod BC and HC neurites (B and H, respectively) took specific paths to rod spherules by gradually adjusting their mutual positions, which resulted in a closed alternating pattern of H‒B‒H‒B neurites at the rod spherule aperture. This order corresponded to the array of elements constituting the postsynaptic complex of ribbon synapses. We identified novel helical coils of HC processes surrounding the rod BC dendrite in both mouse and macaque retinas, and these structures occurred more frequently in mGluR6-knockout than wild-type mouse retinas. Horizontal cell processes also formed hook-like protrusions that encircled cone BC and HC neurites below the cone pedicles in the macaque retina. Conclusions: Bipolar and horizontal cell neurites take specific paths to adjust their mutual positions at the rod spherule aperture. Some HC processes are helically coiled around rod BC dendrites or form hook-like protrusions around cone BC dendrites and HC processes. Loss of mGluR6 signaling may be one factor promoting unbalanced neurite growth and compensatory neurite coiling.

Dynamic assembly of ribbon synapses and circuit maintenance in a vertebrate sensory system.

Ribbon synapses transmit information in sensory systems, but their development is not well understood. To test the hypothesis that ribbon assembly stabilizes nascent synapses, we performed simultaneous time-lapse imaging of fluorescently-tagged ribbons in retinal cone bipolar cells (BCs) and postsynaptic densities (PSD95-FP) of retinal ganglion cells (RGCs). Ribbons and PSD95-FP clusters were more stable when these components colocalized at synapses. However, synapse density on ON-alpha RGCs was unchanged in mice lacking ribbons (ribeye knockout). Wildtype BCs make both ribbon-containing and ribbon-free synapses with these GCs even at maturity. Ribbon assembly and cone BC-RGC synapse maintenance are thus regulated independently. Despite the absence of synaptic ribbons, RGCs continued to respond robustly to light stimuli, although quantitative examination of the responses revealed reduced frequency and contrast sensitivity.

[Hepatic and Pancreatic Metastases of a Rectal Gastrointestinal Stromal Tumor in a Patient with Long-Term Survival following Combined Therapy-ACase Report].

A 68-year-old woman underwent Miles' surgery with a diagnosis of a rectalgastrointestinalstromaltumor (GIST)in 2004. In 2005 and 2006, she developed liver metastases that were surgically removed, but once again in June 2006, she presented with liver metastasis, and imatinib therapy(400mg/day)was administered. In October 2016, she was diagnosed with progression of liver metastasis, and a tumor in the pancreatic body was identified on a CT scan. The patient was referred to our institution for treatment. We performed right hepatectomy and distalpancreatectomy in January 2017. Immunohistochemically, the recurrent tumor was positive for c-kit and CD34, and the diagnosis of GIST was confirmed. The pathological diagno- sis was a high-risk GIST showing 43mitoses per 50 high-power fields. Imatinib therapy(400mg/day)was administered after surgery. She is currently alive without recurrence.

Morphological Survey from Neurons to Circuits of the Mouse Retina.

The mouse retina has a layered structure that is composed of five classes of neurons supported by Müller glial and pigment epithelial cells. Recent studies have made progress in the classification of bipolar and ganglion cells, and also in the wiring of rod-driven signaling, color coding, and directional selectivity. Molecular biological techniques, such as genetic manipulation, transcriptomics, and fluorescence imaging, have contributed a lot to these advancements. The mouse retina has consistently been an important experimental system for both basic and clinical neurosciences.

Classification of Mouse Retinal Bipolar Cells: Type-Specific Connectivity with Special Reference to Rod-Driven AII Amacrine Pathways.

We confirmed the classification of 15 morphological types of mouse bipolar cells by serial section transmission electron microscopy and characterized each type by identifying chemical synapses and gap junctions at axon terminals. Although whether the previous type 5 cells consist of two or three types was uncertain, they are here clustered into three types based on the vertical distribution of axonal ribbons. Next, while two groups of rod bipolar (RB) cells, RB1, and RB2, were previously proposed, we clarify that a half of RB1 cells have the intermediate characteristics, suggesting that these two groups comprise a single RB type. After validation of bipolar cell types, we examined their relationship with amacrine cells then particularly with AII amacrine cells. We found a strong correlation between the number of amacrine cell synaptic contacts and the number of bipolar cell axonal ribbons. Formation of bipolar cell output at each ribbon synapse may be effectively regulated by a few nearby inhibitory inputs of amacrine cells which are chosen from among many amacrine cell types. We also found that almost all types of ON cone bipolar cells frequently have a minor group of midway ribbons along the axon passing through the OFF sublamina as well as a major group of terminal ribbons in the ON sublamina. AII amacrine cells are connected to five of six OFF bipolar cell types via conventional chemical synapses and seven of eight ON (cone) bipolar cell types via electrical synapses (gap junctions). However, the number of synapses is dependent on bipolar cell types. Type 2 cells have 69% of the total number of OFF bipolar chemical synaptic contacts with AII amacrine cells and type 6 cells have 46% of the total area of ON bipolar gap junctions with AII amacrine cells. Both type 2 and 6 cells gain the greatest access to AII amacrine cell signals also share those signals with other types of bipolar cells via networked gap junctions. These findings imply that the most sensitive scotopic signal may be conveyed to the center by ganglion cells that have the most numerous synapses with type 2 and 6 cells.

ON Bipolar Cells in Macaque Retina: Type-Specific Synaptic Connectivity with Special Reference to OFF Counterparts.

To date, 12 macaque bipolar cell types have been described. This list includes all morphology types first outlined by Polyak (1941) using the Golgi method in the primate retina and subsequently identified by other researchers using electron microscopy (EM) combined with the Golgi method, serial section transmission EM (SSTEM), and immunohistochemical imaging. We used SSTEM for the rod-dense perifoveal area of macaque retina, reconfirmed ON (cone) bipolar cells to be classified as invaginating midget bipolar (IMB), diffuse bipolar (DB)4, DB5, DB6, giant bipolar (GB), and blue bipolar (BB) types, and clarified their type-specific connectivity. DB4 cells made reciprocal synapses with a kind of ON-OFF lateral amacrine cell, similar to OFF DB2 cells. GB cells contacted rods and cones, similar to OFF DB3b cells. Retinal circuits formed by GB and DB3b cells are thought to substantiate the psychophysical finding of fast rod signals in mesopic vision. DB6 cell output synapses were directed to ON midget ganglion (MG) cells at 70% of ribbon contacts, similar to OFF DB1 cells that directed 60% of ribbon contacts to OFF MG cells. IMB cells contacted medium- or long-wavelength sensitive (M/L-) cones but not short-wavelength sensitive (S-) cones, while BB cells contacted S-cones but not M/L-cones. However, IMB and BB dendrites had similar morphological architectures, and a BB cell contacting a single S-cone resembled an IMB cell. Thus, both IMB and BB may be the ON bipolar counterparts of the OFF flat midget bipolar (FMB) type, likewise DB4 of DB2, DB5 of DB3a, DB6 of DB1, and GB of DB3b OFF bipolar type. The ON DB plus GB, and OFF DB cells predominantly contacted M/L-cones and their outputs were directed mainly to parasol ganglion (PG) cells but also moderately to MG cells. BB cells directed S-cone-driven outputs almost exclusively to small bistratified ganglion (SBG) cells. Some FMB cells predominantly contacted S-cones and their outputs were directed to OFF MG cells. Thus, two-step synaptic connections largely narrowed down the S-cone component to SBG and some OFF MG cells. The other OFF MG cells, ON MG cells, and ON and OFF PG cells constructed M/L-cone dominant pathways.

Corrigendum: (1) Some OFF Bipolar Cell Types Make Contact With Both Rods and Cones in Macaque and Mouse Retinas; (2) OFF Bipolar Cells in Macaque Retina: Type-specific Connectivity in the Outer and Inner Synaptic Layers.

[This corrects the article on p. 105 in vol. 8, PMID: 25309346.][This corrects the article on p. 122 in vol. 9, PMID: 26500507.].

OFF bipolar cells in macaque retina: type-specific connectivity in the outer and inner synaptic layers.

OFF bipolar cells in the macaque retina were recently classified into five types: flat midget bipolar (FMB) and diffuse bipolar (DB) 1, 2, 3a, and 3b. We examined all parallel pathways from cone photoreceptors via OFF bipolar cells to parasol and midget ganglion cells by serial section transmission electron microscopy. Basal contacts of OFF bipolar cells to cone pedicles were previously categorized as triad-associated (TA) and non-TA (NTA). The latter was further divided into two groups located in the middle and marginal areas of the pedicle at the present eccentricity of 15°. We then mapped the distributions of all three basal contacts of the five OFF bipolar cell types in the same area of cone pedicles. TA contacts were more numerous than NTA contacts in FMB (93%), DB1 (67%), and DB3a (81%) cells, but less in DB2 (30%) and DB3b (21%) cells. Cluster analysis of these contact parameters reconfirmed five distinct OFF bipolar cell types and showed these positional configurations of basal synapses to be cell type-specific. This architecture is thought to provide a spatial framework for the interstitial diffusion and local uptake of the neurotransmitter (glutamate) that spills over from ribbon synapses. All five OFF bipolar cell types formed ribbon-synaptic contacts to both parasol and midget ganglion cells. DB2 and 3a, DB1 and 3b, and FMB predominantly, moderately, and negligibly contacted parasol ganglion cells, respectively. FMB almost exclusively contacted midget ganglion cells, to which DB1 provided dominant output (58%), and DB2, 3a, and 3b provided between 3% and 10% of their output. Consequently, the cone signal sampling routes of a midget ganglion cell consisted of two substructures: the narrow (mainly 2-3 cones) FMB pathway and the wide (mainly 10 cones) DB pathway, where connection strength was four-fold greater in the FMB than DB pathway. The narrow and strong FMB pathway may confer the highest spatial resolution and sporadically may include blue cone signals.

Some OFF bipolar cell types make contact with both rods and cones in macaque and mouse retinas.

This study compared the types of OFF bipolar cells found in the macaque retina with those found in the mouse retina and determined whether these OFF bipolar cells make direct contacts with both rods and cones by serial section transmission electron microscopy. We performed scatter plots and cluster analysis of the morphological variables of their axon terminals such as the stratification level, the arbor thickness, the arbor area, and the number of ribbons. Five OFF bipolar cell types, including the recently discovered DB3b type, were identified in the macaque retina. The macaque OFF bipolar cell types FMB, DB1, DB2, DB3a, and DB3b corresponded to the mouse OFF bipolar cell types 2, 1, 4, 3a, and 3b, respectively. In addition to contacting rod bipolar cells, ~7% of rods in the macaque retina made basal contacts exclusively with one cell type, DB3b, whereas 18% of rods in the mouse retina made basal contacts with one or two of types, 3a, 3b, and 4. Approximately 3% of mouse rods were divergently connected to two OFF bipolar cells of different types, but macaque rods were solely connected to one OFF bipolar cell. Rod-rod gap junctions were localized at rod cell bodies and axons in the outer nuclear layer in both macaque and mouse retinas. The direct rod-OFF bipolar connection system is slightly more developed in the mouse retina than in the macaque retina, possibly as a fine-tuned adaptation to nocturnal conditions. This one-step direct synaptic pathway from rods to OFF bipolar cells may enhance the response speed to OFF light stimuli compared with more indirect pathways via rod-cone gap junctions (a two-step pathway) and via rod bipolar and AII amacrine cells (a three-step pathway).

Effects of mGluR6-deficiency on photoreceptor ribbon synapse formation: comparison of electron microscopic analysis of serial sections with random sections.

This study examined the effects of metabotropic glutamate receptor 6 (mGluR6) deficiency on ribbon synapse formation in rod spherules and cone pedicles using serial-section electron microscopy. In a wild-type (WT) mouse, only 3% of spherules had one invaginating bipolar dendrite (1B-type) and 97% of spherules were 2B-type. In contrast, in an mGluR6-knockout (KO) mouse, 29% of spherules were 1B-type and 71% of spherules were 2B-type. Spherules without bipolar invagination were not observed in either genotype. The single invaginating dendrites in 1B-type spherules were larger and the surface areas of synaptic ribbons were 23% smaller in the mGluR6-KO mouse than in the WT mouse. In cones, the number of invaginating bipolar dendrites decreased from 12 in the WT mouse to 9.5 in the mGluR6-KO mouse. This decrease correlated with a decrease in the number of cone synaptic ribbons from 10 in the WT mouse to 8 in the mGluR6-KO mouse. The mGluR6-KO phenotype showed negative effects on ribbon synapse formation. This negativity was similar to those in mGluR6-nob4, Gß3-KO, Gß5-KO, and RGS-7:RGS-11 double-KO mice, but the detailed manners and degrees of alterations appeared to vary depending on different missing components. Two published morphological assessments of the RGS-7:RGS-11 double-KO phenotype reported conflicting data; therefore, we tested the statistical techniques used in the two analyses. One statistical evaluation measure was effective in identifying a significant difference in structure between the mutant and WT phenotypes, whereas the other measure was ineffective. Conventional random section analysis using the effective measure provided sufficient data for a statistical test of the occurrence of structural changes. However, serial section analysis was required to determine the absolute numbers of ribbons and invaginating dendrites and to estimate structural parameters such as ribbon surface area.

Functional allocation of synaptic contacts in microcircuits from rods via rod bipolar to AII amacrine cells in the mouse retina.

Retinal microcircuits for night vision at the absolute threshold are required to relay a single-photon rod signal reliably to ganglion cells via rod bipolar (RB) cells and AII amacrine cells. To assess the noise reduction of intercellular signal transmission in this rod-specific pathway, we quantified its synaptic connectivity by 3D reconstruction of a series of electron micrographs. In most cases (94%), each rod made ribbon synaptic contacts onto two adjacent RB cells. Conversely, each RB cell was contacted by 25 rods. Each RB axon terminal contacted four or five AII amacrine cells via 53 ribbon synapses. Thus, the signal from one rod may be represented as 106 replicates at two RB axons. Moreover, the two adjacent RB cells contacted two to four AII amacrine cells in common, where the signals relayed by two RB cells were reunited. In more detail, over 50% of each RB output was directed predominantly to a single, preferred AII amacrine cell, although each RB cell also separately contacted another one to three AII amacrine cells. Most of the replicate signals at two RB axons were collected on a few AII amacrine cells via reunions, dominant connections, and electrical coupling by AII-AII gap junctions. Thus the original signal may be reliably represented by signal amplification with focal accumulation without gathering unnecessary noise from a wide surrounding area. This allocation of RB-AII synaptic contacts may serve as the structural basis for the physiological properties of the AII single-photon response that include high amplification, local adaptation, and regenerative acceleration.

Ectopic synaptic ribbons in dendrites of mouse retinal ON- and OFF-bipolar cells.

The ectopic distribution of synaptic ribbons in dendrites of mouse retinal bipolar cells was examined by using genetic ablation of metabotropic glutamate receptor subtype 6 (mGluR6), electron microscopy, and immunocytochemistry. Ectopic ribbons were observed in dendrites of rod and ON-cone bipolar cells in the mGluR6-deficient mouse but not in those of wild-type mice. The number of rod spherules facing the ectopic ribbons in mGluR6-deficient rod bipolar dendrites increased gradually during early growth and reached a plateau level of about 20% at 12 weeks. These ectopic ribbons were immunopositive for RIBEYE, a ribbon-specific protein, but the associated vesicles were immunonegative for synaptophysin, a synaptic-vesicle-specific protein. The presence of ectopic ribbons was correlated with an increase in the roundness of the invaginating dendrites of the rod bipolar cells. We further confirmed ectopic ribbons in dendrites of OFF-cone bipolar cells in wild-type retinas. Of the four types of OFF-cone bipolar cells (T1-T4), only the T2-type, which had a greater number of synaptic ribbons at the axon terminal and a thicker axon cylinder than the other types, had ectopic ribbons. Light-adapted experiments revealed that, in wild-type mice under enhanced-light adaptation (considered similar to the mGluR6-deficient state), the roundness in the invaginating dendrites and axon terminals of rod bipolar cells increased, but no ectopic ribbons were detected. Based on these findings and known mechanisms for neurotransmitter release and protein trafficking, the possible mechanisms underlying the ectopic ribbons are discussed on the basis of intracellular transport for the replenishment of synaptic proteins.

Different roles of ribbon-associated and ribbon-free active zones in retinal bipolar cells.

Synaptic ribbons with a halo of synaptic vesicles are seen at the active zones of sensory neurons that release transmitter tonically. Thus, ribbons are assumed to be a prerequisite for sustained exocytosis. By applying total internal reflection fluorescence microscopy to goldfish retinal bipolar cell terminals, we visualized Ca2+ entry sites, ribbons, and vesicle fusion events. Here we show that the main Ca2+ entry sites were located at ribbons, and that activation of the Ca2+ current induced immediate and delayed vesicle fusion events at ribbon-associated and ribbon-free 'hot spots', respectively. The activation of protein kinase C (PKC) specifically potentiated vesicle fusion at ribbon-free sites. Electron microscopy showed that PKC activation selectively increased the number of docked vesicles at ribbon-free sites, which faced neuronal processes with the postsynaptic density. Retinal bipolar cells have both ribbon-associated and ribbon-free active zones in their terminals and might send functionally distinct signals through ribbon-associated and ribbon-free synapses to postsynaptic neurons.

A novel connection between rods and ON cone bipolar cells revealed by ectopic metabotropic glutamate receptor 7 (mGluR7) in mGluR6-deficient mouse retinas.

Since the discovery of direct chemical synapses between rod photoreceptor and OFF cone bipolar cells in mouse retinas, whether the ON cone bipolar cell also receive direct chemical input from rod has been a pending question. In finding that metabotropic glutamate receptor 7 (mGluR7) was uniquely expressed in dendrites of ON cone bipolar cells in the mGluR6-deficient mouse retina, we used this ectopic mGluR7 immunoreactivity as a specific marker for the ON cone bipolar to search for its rod connection. Here, we show that a certain type of ON cone bipolar cell forms ribbon-associated synapses not only with cones, but also rods. This finding was verified in the wild-type mouse retina by three-dimensional reconstruction of bipolar cells from serial electron micrographs. These ON cone bipolars were further identified as corresponding to type 7 of mouse bipolar cell described by Ghosh et al. (2004) and also to the green fluorescent protein (GFP)-labeled type 7 bipolars in the alpha-gustducin-GFP transgenic mouse. Our findings suggest that, in mice, rod signals bifurcate into a third ON and OFF pathway in addition to the two known routes to cone bipolar cells: (1) via rod chemical synapse --> rod bipolar --> AII amacrine --> ON and OFF cone bipolar cells; (2) via rod-cone gap junction --> cone chemical synapse --> ON and OFF cone bipolar cells; and (3) via rod chemical synapse --> ON and OFF cone bipolar cells. This third novel pathway is thought to transmit fast and moderately light-sensitive rod signals, functioning to smooth out the intensity changes at the scotopic-mesopic interface.

Xylazine promotes axonal regeneration in the crushed optic nerve of adult rats.

Xylazine, an alpha-2 adrenergic agonist, activates the endogenous trophic factors and neuronal survival signaling. Here, we tested the regenerative effect of xylazine on damaged optic nerve axons in adult rats. After optic nerve crush, xylazine was intraperitoneally injected into three groups of rats: a single administration immediately after the crush, intermittent administration, and daily administration. On day 14, the regenerated axons were quantitatively evaluated by anterograde labeling. Everyday administration but neither single nor intermittent administration markedly increased the number of labeled axons beyond the crush site, with upregulation of growth-associated protein-43 in the ganglion cell layer and the regenerated axons. It was concluded that xylazine promotes axonal regeneration in damaged optic nerves of adult rats.