What the salamander eye has been telling the vision scientist’s brain

5 Abstract 6 Salamanders have been habitual residents of research laboratories for more than a century, 7 and their history in science is tightly interwoven with vision research. Nevertheless, many 8 vision scientists – even those working with salamanders – may be unaware of how much our 9 knowledge about vision, and particularly the retina, has been shaped by studying 10 salamanders. In this review, we take a tour through the salamander history in vision science, 11 highlighting the main contributions of salamanders to our understanding of the vertebrate 12 retina. We further point out specificities of the salamander visual system and discuss the 13 perspectives of this animal system for future vision

1 Introduction 19 "Why salamander?" If you are a neuroscientist working with the salamander visual system, 20 this may well be the most common question that you hear after presenting your work at an 21 outside talk. And if you are not, you may have asked this question yourself when coming 22 across one of the surprisingly many works in visual neuroscience built on investigating these 23 animals. How indeed have these cold-blooded, egg-laying amphibians, which spend a great 24 deal of their lives in water and are distant from us humans by more than 300 million years 25 of separated evolution (San Mauro, 2010) come to be a model for studying the early visual 26 system? 27 It is this question that we focus on in this review. We take a historical tour that 28 highlights key contributions that salamanders have brought to our understanding of the 29 early visual system. These contributions have been successively built upon each other and 30 have often used two key properties of the salamander nervous system: particularly large 31 neurons and extraordinary robustness to experimental manipulations. We find general 32 concepts about the visual system that have emerged from work on the salamander, as well 33 as peculiarities that are of interest for comparative and ethological studies. Finally, we ask 34 what future role the salamander plays for vision research. Throughout this tour, our focus 35 will be on the retina, the neural network at the back of the eyeball where the first stages of 36 visual processing in vertebrates occur. This is where the salamander has had an outstanding 37 influence on the field of vision science. Altogether, the body of work on the salamander 38 visual system is truly immense, and we necessarily had to leave out many important works; 39 there is no pretension of completeness of this overview. 40 2 The order of salamanders 41 Salamanders, together with newts, form the amphibian order Urodela. The other two 42 amphibian orders are Anura (frogs and toads) and Apoda (the limbless and mostly blind 43 caecilians). All amphibians can be considered evolutionarily early vertebrates. Relatively 44 soon after the first tetrapod vertebrates started treading dry land, amphibians separated 45 from what would become reptiles, birds, and mammals. These latter groups experienced 46 radical changes in body plan (Radinsky, 1987) that allowed more complex patterns of 47 locomotion and the occupation of new ecological niches. Concomitantly, brain areas 48 enlarged, differentiated and gave rise to new structures such as the cortex. Many 49 amphibians, on the other hand, did not undergo such drastic changes. Urodeles, in 50 particular, seem to have kept close to their original lifestyle and are thus considered to 51 occupy an intermediate step in evolution, with brains lacking a cortex and displaying an 52 anatomy that may resemble those of the first land-dwellers (Herrick, 1948). 53 Salamander brains are relatively simple (Kingsbury, 1895;Herrick, 1948) even when 54 compared to those of other amphibians or lampreys and hagfishes, suggesting a certain 55 phylogenetic simplification (Roth et al., 1997). For instance, the salamander tectum shows 56 little lamination and only 30,000-90,000 cells, compared to the 800,000 in the tectum of 57 anurans (Roth et al., 1997). Nevertheless, salamanders can see -and process what they see 58 -well enough to help them flee, feed, and procreate (Roth, 1987). Both larvae and adults 59 are carnivorous and need to hunt. Some species, like the tongue-projecting salamanders 60 (genus Bolitoglossa), have been shown to depend on vision for determining the distance to 61 prey quickly and precisely . Others, like the tiger salamander 62 (Ambystoma tigrinum), which despite its name prefers to sit and wait for its prey, rely on 63 vision for deciding when to strike (Lindquist and Bachmann, 1982). 64 65 Salamanders comprise more than 700 species (Frost, 2020) and are overall very diverse. 66 While it is commonly thought that salamanders start their life as larvae in water until 67 metamorphosing into a terrestrial adult form, this view is incorrect for two thirds of 68 salamander species (Elinson and del Pino, 2012). In the lungless salamander family 69 (Plethodontidae), the most speciose, animals hatch directly from eggs into a terrestrial form 70 (Wake and Hanken, 1996). Other species, like Necturus maculosus (mudpuppy) and 71

Diversity of species
Ambystoma mexicanum (axolotl), display neoteny: individuals can reach sexual maturity in 72 their larval forms and may never metamorphose (Vlaeminck-Guillem et al., 2004). 73 With so many species of salamanders, it is no wonder that vision has been studied in 74 many of them. And although findings are often treated as coming from a single type of 75 animal (and we here may do the same for expediency when the context is clear), it is 76 important to note that there really is no "the salamander" as a species in vision research. 77 Yet, three species have contributed dearly to our understanding of vision and thus have a 78 special place in this tour. They are the three darlings of salamander retinal research: 79 Necturus maculosus (mudpuppy) and two closely related species of mole salamanders, 80 Ambystoma tigrinum (tiger salamander) and Ambystoma mexicanum (axolotl). Their retinas 81 display the same, characteristic structure (Figure 1), with fewer and larger cell bodies as  82  compared to mammalian retinas, which has proved a boon for retina research. Knowing  83  about the characteristics and idiosyncrasies of these species provides an essential context  84 for studying their visual systems. 85 Figure 1: Cross-section of the retina for three salamander species. From top to bottom in each cross-section: outer nuclear layer (ONL), outer plexiform layer (OPL), inner nuclear layer (INL), inner plexiform layer (IPL), and ganglion cell layer (GCL). Thin lines indicate the borders between layers. Mudpuppy, tiger salamander, and axolotl retinas are structurally alike with large cells and thin plexiform layers. Mudpuppy and tiger salamander retina cross-sections are autoradiographic and adapted respectively from (Pourch et al., 1984) and (Yang and Wu, 1997) with permission (Elsevier). Axolotl retina cross-section from light microscopy adapted from (Custer, 1973)  become sexually mature at about 5 years of age, and are often found in the region of the 89 Great Lakes in North America (Bishop, 1994). 90 The anatomy of the mudpuppy brain was described in detail more than a hundred 91 years ago by Kingsbury (1895). At the time, the mudpuppy brain was already considered to 92 strike a good balance in size for anatomical investigations at both microscopic and 93 macroscopic levels, a property that was exploited later on for detailed anatomical 94 descriptions of rods and cones (Howard, 1908;Brown et al., 1963) and for recordings of the 95 retinal output (Bortoff, 1964;Werblin and Dowling, 1969). 96 2.1.2 Ambystoma tigrinum (tiger salamander)

97
The genus "Ambystoma" has been plagued with controversy, starting with its name 98 (Willoughby, 1935). This genus was first proposed in the early 19 th century by Tschudi (1838)  99 to refer to North-American mole salamanders. Believing the name to be a misspelling (Scott,100 1916; Lyon, 1916), some authors took the liberty to rename the genus as "Amblystoma". 101 The case was only settled after a vote by the International Commission on Zoological 102 Nomenclature in 1963 (International Commission on Zoological Nomenclature, 1963; Smith,  103 1969). 104 Tiger salamanders (Ambystoma tigrinum) were once considered to be a single species 105 extending over most of North America but are now best divided into several subspecies, 106 each with a specific geographic range (Shaffer and McKnight, 1996). All subspecies of 107 Ambystoma tigrinum, as well as multiple closely related species from Mexico (including the 108 axolotl, see below), form the tiger salamander species complex (Shaffer and McKnight, 109 1996). Some subspecies are facultative paedomorphs while others must metamorphose to 110 reach maturity (Routman, 1993). Tiger salamanders are the largest mole salamanders, and  111 adults in the wild can be more than 20 cm long (Bishop, 1994). Their brains have been 112 studied in detail as early as the 1940s (Herrick, 1948), and over recent years, tiger 113 salamanders have received much attention in vision research and thus have become 114 something like the standard salamander system in the field. 115

116
Historically, the name axolotl referred to the larval stage of ambystomatid salamanders 117 regardless of species (Shufeldt, 1885). Nowadays, it is reserved for a single species, 118 Ambystoma mexicanum, originally coming from an area near Lake Xochimilco in Mexico 119 (Smith, 1969;Farkas and Monaghan, 2015). In the wild, axolotls are facultative 120 paedomorphs and are known to metamorphose if needed (Smith, 1969). In laboratories, 121 likely because of artificial selection, axolotls remain in their larval forms (Figure 2, top) 122 unless hormonally induced to metamorphose (Smith, 1969;Vlaeminck-Guillem et al., 2004). 123 The larvae can look strikingly like those of tiger salamanders. 124 Axolotls have a long history as a laboratory animal (Reiß et al., 2015). At the beginning of the 20 th century in the United States, Humphrey started a colony 131 with many well-characterized axolotl mutants (Humphrey, 1975). This included strikingly 132 white axolotls, with reduced skin pigmentation but pigmented eyes. However, the absence 133 of a true axolotl albino and the discovery of a tiger salamander albino in the wild led 134 Humphrey to create a hybrid of a white axolotl with this albino tiger salamander 135 (Humphrey, 1967 An interesting mutant is the eyeless axolotl. First observed as a spontaneous 141 occurrence in a stock (Humphrey, 1969), these mutants lack eyes due to a developmental 142 defect (Harris, 1983). Yet, transplanting eyes from a regular axolotl to an eyeless one at an 143 early developmental stage can recover visual object localization and the optokinetic reflex 144 as well as normal vision-driven skin pigmentation (Epp, 1972;Hibbard and Ornberg, 1976). 145 Indeed, nerve fibers from the transplanted eye manage to find their usual target areas, 146 though through unusual paths that can differ from animal to animal (Hibbard and Ornberg,147 1976; Harris, 1983). 148

149
There are probably two aspects that explain why salamanders early on became such a well-150 studied system in vision research. On the one hand, their nervous system appears to be 151 particularly robust to handling and manipulations (Stone and Ussher, 1927;Stone et al., 152 1937;Sperry, 1943aSperry, , 1943b, allowing experiments and functional studies that might be 153 more difficult in other animal systems. 154 On the other hand, and perhaps most importantly, salamanders turn out to have 155 notably large cells. Their photoreceptors, for example, can have outer segment diameters of 156 10-13 µm (Mariani, 1986), considerably larger than the 1-2 µm of typical mouse 157 photoreceptors (Carter-Dawson and Lavail, 1979) ( Figure 3A). But large cells can be found 158 throughout the salamander's nervous system and indeed throughout their entire body 159 (Gregory, 2001 Figure 2: Samples from the salamander tour in vision science. The timeline shows selected contributions of the salamander to vision research and can be roughly divided into a period centered on neuroanatomy (orange region), a period with focus on cellular neuroscience and neurochemistry (green), and a period with major contributions to systems and computational neuroscience (blue). The image on top shows an axolotl (photo kindly provided by Norma Kühn). The fuzzy-looking appendages at the neck are external gills, a typical feature of aquatic salamanders.

Structural studies 173
There is a long history of using salamanders for investigating the retina (Figure 2), which 174 considerably helped advance our basic understanding of the retina's structure and function. 175 Even some of the very first studies of retinal organization were already performed with 176 salamanders (Hulke, 1867;Landolt, 1871). The large cells of the mudpuppy allowed general 177 descriptions of rods and cones (Howard, 1908) as well as a count of all cells in a single retina 178 (Palmer, 1912), leading later to one of the first and most detailed structural 179 characterizations of photoreceptors (Brown et al., 1963). The salamander retina also 180 contributed to revealing electrical gap junctions in the retina, which had been proposed to 181 explain signal spread between neighboring cones in electrophysiological experiments 182 (Baylor et al., 1971). Observing junctions in electron microscopic examinations of the axolotl 183 (Custer, 1973) and tiger salamander (Lasansky, 1973) retina then provided structural 184 evidence for electrical connections between photoreceptors as well as between horizontal 185 cells. 186

Synapses and signal transmission
187 Salamanders were also present as the first electrophysiological investigations of the retina 188 were performed. Already Hartline -in his seminal studies of single optic nerve fibers, which 189 led to his eventual Nobel prize -recorded from the mudpuppy, though his amphibian work 190 mostly focused on frogs (Hartline, 1938). For the next few decades, the mudpuppy retina -191 thanks to the large cells that allowed intracellular recordings (Bortoff and Norton, 1965) -192 was one of the most widely studied early vision systems, used to show the match of 193 morphology and physiology for the different retinal cell classes (Werblin and Dowling,194 1969), to characterize light and dark adaptation (Werblin, 1971; Grabowski et al., 1972), to 195 reveal the different kinetics of rods and cones to flashes of light (Lasansky and Marchiafava,196 1974), and to elucidate the role of amacrine cells in lateral inhibition (Werblin, 1972). 197 The mudpuppy retina also played an essential role in dissecting the ON and OFF 198 pathways in the retina. The possibility to record intracellularly from all retinal cell types in 199 the mudpuppy retina in chloride-free solutions (Miller, 2008)  The previous section has highlighted the use of the salamander retina as a beneficial system 292 for studying general features of the retina. Yet, interesting insights also come from 293 differences to other animals, and investigating the salamander retina has certainly provided 294 a rich set of specifics and idiosyncrasies that distinguish it from mammals or other 295 vertebrates. Some of these we discuss in this section. of photopigments since those first appeared around 500 million years ago (Bowmaker,299 2008). Regarding salamanders, photoreceptors have been most thoroughly described in 300 tiger salamanders, which have six types ( Figure 3B), comprising two rods and four cones 301 (Mariani, 1986;Sherry et al., 1998). In total, rods and cones are almost equally numbered in 302 the larval tiger salamander retina, with cones slightly outnumbering rods near the center 303 and vice versa in the periphery (Zhang et al., 2004). Among the rods, the vast majority is 304 tuned to medium wavelengths (M-rod), with highest sensitivity for green light. The other 305 rod type only comprises a few percent of the rods and is smaller and tuned to short 306 wavelengths (S-rod). The presence of two rods is common in amphibians (Hárosi, 1975). 307 Because the rods were first distinguished (in frogs) based on their apparent color under a 308 microscope, the M-and S-rods are also (perhaps confusingly) referred to as "red" (green-309 absorbing) and "green" (blue-absorbing) rods (Denton and Wyllie, 1955). 310 including UV cones (Deutschlander and Phillips, 1995). Mudpuppies, on the other hand, 317 exhibit a simpler layout with potentially only one rod and two cone types (Fain and Dowling,318 1973; Hárosi, 1975). Despite the rich set of photoreceptor types in some salamanders, little 319 is known about whether these animals have color vision, except that one species 320 (Salamandra salamandra) appears to use differences in color to guide behavior (Przyrembel 321 et al., 1995;Tempel and Himstedt, 1979). 322 Interestingly, S-cones and S-rods in the tiger salamander share the same opsin, but S-323 rods have more pigment, which may explain their higher sensitivity to flashes (Ma et al.,324 2001) (see Figure 3B). Furthermore, UV-and single L-cones, as well as the accessory 325 member of the double cones, express more than one opsin. Besides their primary opsins 326 that determine their peak sensitivity, UV-cones express low amounts of S-and L-opsins, 327 while the single L-cones and the accessory member of the double L-cones express UV-and 328 S-opsins. The exact pigment ratios in L-cones may differ from cell to cell, but UV-and S-329 pigments can comprise up to a third of all pigments in some L-cones ( Rods are coupled to neighboring photoreceptors. In the axolotl and tiger salamander, 335 there is evidence for gap junctions from rods to other rods and cones (Custer, 1973;336 Mariani, 1986) but no direct connections have yet been found between cones. Each rod is 337 typically coupled electrically to four other rods and four cones (Attwell et al., 1984). Some 338 rods are so strongly coupled to cones that they change their spectral sensitivity with 339 changes in background illumination . 340 properties. In general, salamander OFF bipolar cells are observed to be about 30 ms faster in 346 their response kinetics than ON bipolar cells (Burkhardt, 2011). Curiously, it has been 347

Signal transmission from photoreceptors to bipolar cells
reported that one bipolar cell type, which stratifies in two layers of the inner plexiform 348 layer, may possess both ON-type and OFF-type response properties, perhaps depending on 349 light levels (Wu et al., 2000). 350 It has been realized early on for the salamander retina that rods as well as cones 351 make direct synapses to multiple types of bipolar cells (Lasansky, 1973). For both ON and 352 OFF bipolar cells, rod-dominated as well as cone-dominated types can be found (Hensley et  353 al., 1993), with rod-dominated bipolar cells stratifying preferentially at the two edges of the 354 inner plexiform layer and cone-dominated bipolar cells more centrally. It is worth noting 355 that the interconnectedness of rod and cone signals at the level of bipolar cells originally 356 appeared to be a striking difference from the mammalian retina, which contains distinct rod 357 and cone bipolar cells. Meanwhile, however, evidence has been accumulating that, at least 358 in mouse, the rod bipolar cell and some cone bipolar cells also receive input from those 359 photoreceptors that are not part of their name (Soucy et al., 1998;Hack et al., 1999;360 Behrens et al., 2016), making this distinction between salamander and mammalian retina 361 more gradual than absolute. 362 Morphologically, an interesting feature of salamander bipolar cells is the occurrence 363 of a Landolt club (Landolt, 1871), a protrusion of the cell, potentially rich in mitochondria 364 and extending towards the photoreceptor cell bodies similar to dendrites but without 365 synaptic contacts (Hendrickson, 1966). Landolt clubs are observed in most, if not all, bipolar 366 cells in amphibians (Lasansky, 1973), as well as in some other non-mammalian species. 367

Inhibitory interactions 368
The information flow through the retina from photoreceptors via bipolar cells to ganglion 369 cells is modulated by inhibitory signals from horizontal and amacrine cell (Roska et al., 370 2000). Horizontal cells come in two types in the tiger salamander (Lasansky and Vallerga, 371 1975;Zhang et al., 2006). One of the two types has two distinct regions of neurite 372 branching, coupled by a thin axon, providing in total three potentially distinct horizontal cell 373 processing entities, with differences in relative rod versus cone inputs, receptive field sizes, 374 and gap junction coupling (Zhang et al., 2006). 375 Amacrine cells in the tiger salamander release the conventional inhibitory 376 neurotransmitters GABA and glycine as well as the neuromodulators dopamine and 377 serotonin (Li et al., 1990;Watt et al., 1988;Yazulla, 1988a, 1988b). In addition, 378 some amacrine cells appear to be cholinergic (Deng et al., 2001;Zhang and Wu, 2001), 379 which, in the mammalian retina, is usually associated with the circuit of direction-selective 380 ganglion cells, though a similar function of cholinergic salamander cells has not yet been 381 shown. 382 Somewhat of a controversy exists about whether amacrine cells in the salamander 383 follow the same relation of neurotransmitter to size as observed in the mammalian retina, 384 where GABAergic amacrine cells are mostly large, wide-field or medium-field cells and 385 glycinergic ones mostly narrow-field (Werblin, 2011). Studies in retinal slices of the tiger 386 salamander indicated longer interaction distances for glycinergic as compared to GABAergic 387 amacrine cells ) and mostly wide-field characteristics of 388 glycinergic cells (Yang et al., 1991), suggesting that the neurotransmitter-to-size relation 389 may be opposite to that in mammals (Wässle and Boycott, 1991). However, later analyses of 390 amacrine cells in whole-mount preparations found mostly wide-field GABAergic cells and 391 narrow-field glycinergic cells (Deng et al., 2001), in accordance with the mammalian system. 392 Thus far, this question remains unresolved. 393 Bipolar cells mostly express GABAC receptors at their synaptic terminals (Lukasiewicz 394 et al., 1994). Here, the release of glutamate can be modulated by GABAergic amacrine cells 395 (Roska et al., 1998), which may enhance the temporal contrast at the terminals (Dong and 396 ). There also have been observations of glycine receptors at the dendrites of 397 bipolar cells (Yang and Yazulla, 1988a), though they don't appear to contribute to the 398 receptive field surround (Hare and Owen, 1996). These glycine receptors may be the target 399 of glycinergic interplexiform cells, which have been shown to affect the dendrites of bipolar 400 cells (Maple and Wu, 1998), perhaps to regulate the gain of signal transmission between 401 photoreceptors and bipolar cells (Jiang et al., 2014). 402 Interplexiform cells form a class of retinal neurons that receive input at the inner 403 retina, resembling amacrine cells, but stratify at the outer plexiform layer and are thought 404 to provide feedback across the synaptic layers (Dowling, 1987). At least three morphological 405 types of interplexiform cells have been described in the tiger salamander (Maguire et al., 406 1990). They are all spiking cells, receive ON as well as OFF sustained excitation from bipolar 407 cells, and release GABA or glycine (Yang and Yazulla, 1988a). Dopaminergic interplexiform 408 cells, which have been found in other animals such as frog (Witkovsky et al., 1994), appear 409 to be absent or extremely rare in the salamander retina (Watt et al., 1988). studies), when additional response characteristics were considered, such as adaptation 458 (Kastner and Baccus, 2011) or direction selectivity (Kühn and Gollisch, 2016). It thus remains 459 to be seen whether enhanced classification methods might provide a refined separation of 460 recorded ganglion cells into perhaps a larger number of types with tiling receptive fields. 461 A functional class of ganglion cells of widespread interest is the class of direction-462 selective cells. These cells respond to a specific direction of visual motion, but are 463 suppressed by the opposite direction (Wei, 2018). Yet, for the salamander, investigations of 464 direction selectivity were conspicuously absent, despite early examples in the mudpuppy 465 (Werblin, 1970;Karwoskj and Burkhardt, 1976;Tuttle, 1977) and tiger salamander retina 466 (Pan and Slaughter, 1991) areas are the optic tectum, the thalamus, the pretectum, the basal optic nucleus, and the 477 hypothalamus (Grüsser-Cornehls and Himstedt, 1976). The anatomical layout of the optic 478 tracts that connect the retina to these areas and of the brain regions involved in visual 479 processing are described in detail elsewhere (Jakway and Riss, 1972; Grüsser-Cornehls and 480 Himstedt, 1976;Roth, 1987 preference that matches the salamander's prey capture response, which is preferentially 488 triggered by horizontally elongated shapes moving along the horizontal direction, at least at 489 low velocities (Luthardt and Roth, 1979). Recordings in different salamander species, 490 however, found a variety of shape tunings in individual neurons that generally did not match 491 the behavioral preference (Himstedt and Roth, 1980;Roth, 1982), suggesting a more 492 complex representation of prey stimuli in the tectum (an der Heiden and Roth, 1987). Later 493 recordings in the red-legged salamander (Plethodon shermani) with prey-like stimuli 494 indicated that processing in tectal neurons involves feedback from other brain areas and 495 integration of visual information over ranges much larger than classical receptive fields 496 (Schuelert and Dicke, 2005). 497 The ability to test visual behavior through prey-like stimuli also helped establish the 498 importance of ordered connectivity of nerve fibers with their downstream targets. At first, 499 observations that salamanders (and other amphibians) could recover vision after eyes had 500 been excised and grafted back into the eye socket (Stone and Ussher, 1927; Stone et al.,  501 1937) had been taken as evidence that neural plasticity in central areas upon regeneration 502 of the optic nerve was so potent as to make specific connectivity unnecessary. However, 503 Roger Sperry -another eventual neuroscience Nobel laureate who appreciated the 504 robustness and simplicity of salamanders -then showed that rotating the eyes of newts 505 either while keeping the optic nerve intact (Sperry, 1943a) or during grafting after 506 enucleation (Sperry, 1943b) led to inverted vision. Animals turned away from prey stimuli 507 and displayed an inverted optokinetic reflex. These effects remained over several months, 508 indicating a lack of plasticity. Thus, Sperry concluded that orderly, retinotopic connectivity is 509 essential and that this may be (re-)established by (chemical) signals that are carried by the 510 nerve fibers, which became known as the chemoaffinity hypothesis (Meyer, 1998). 511 The stereotypic, reflex-like visuomotor responses (Arbib, 1987)  There is an abundance of salamander species living in diverse ecological niches, some with 526 significant terrestrial life. These species had millions of years to specialize their visual system 527 for these niches , perhaps developing differences in their retinas. For 528 example, already in 1897, Slonaker mentioned two salamander species (Salamandra atra, 529 Triturus cristatus) that presented a higher density of visual cells in central areas of their 530 retinas, suggestive of an area centralis (Hulke, 1867;Slonaker, 1897). Surveys of other 531 species found no area centralis (Linke et al., 1986;Roth, 1987), and further reports of such 532 specialized regions appear to be lacking in the literature. However, evidence has surfaced of 533 a weak spatial inhomogeneity in the tiger salamander retina, e.g., in the density of 534 photoreceptors and certain amacrine cells (Zhang et al., 2004). Comparisons across species 535 of such aspects may help us understand how visual systems are adapted to particular 536 environments. 537 A drastic change in salamander lifestyle comes with the metamorphosis of the aquatic 538 larvae to terrestrial adults. How the visual system adjusts to its new environment is a 539 fascinating question, about which surprisingly little is known. In the retina, the morphology 540 of the inner plexiform layer and the sensitivity of bipolar cells are apparently unaffected 541 (Wong-Riley, 1974; Burkhardt et al., 2006). On the other hand, S-cones in the tiger 542 salamander degenerate and are replaced by additional S-rods after metamorphosis (Chen et  543 al., 2008) -possibly as an adaptation to darker environments on land. This exemplifies that 544 the switch from aquatic to terrestrial life provides an intriguing opportunity to study how 545 the visual system adapts to its environmental challenges. 546

547
The lack of standard lines in amphibians has been a longstanding issue, with most 548 specimens captured in the wild (Gibbs et al., 1971;Nace, 1976). Even for axolotls, despite 549 their tradition as laboratory animals (Reiß et al., 2015) and well-described genetic 550 background of inbred strains (Humphrey, 1975;Shaffer, 1993;Voss et al., 2009), there are 551 no clear, standardized lines available, which could affect reproducibility of scientific findings 552 across laboratories. Thus, it is custom that researchers report the supplier of their animals. 553 Over the past decade, mice have developed into arguably the primary model system 554 for vision research, owing to the rich genetic toolkit now available for them. Yet, other 555 animal systems may be catching up, and among salamanders, axolotls appear to be in the 556 best position to compete. While slow reproduction had been an issue in the past, optimized 557 protocols have ensured that transgenic axolotls can be more easily obtained (Khattak et al., 558 2014). Recently, the complete axolotl genome was assembled (Nowoshilow et al., 2018). 559 And the interest in limb regeneration (Tanaka, 2016)  research. This benefit may not be as significant nowadays. Nevertheless, the sheer 565 knowledge accumulated about the physiology and morphology of the salamander retina 566 now provides an expedient background for further explorations of the system. Given the 567 ease of use, the opportunity of comparisons across species as well as across 568 metamorphosis, and the anticipated possibility of transgenic salamanders, we expect 569 salamanders to have, after their long and fruitful past, also a prosperous future in vision 570 research. 571 The future investigations should also contribute to a more general understanding of 572 early visual processing across species (Baden et al., 2019). Their comparatively simple 573 nervous system and the link to stereotypic visual behaviors make salamanders a particularly 574 appealing system for comparison with the current standard model systems of mice and 575 primates in order to study which features of early visual processing generalize across 576 vertebrate species and what the scope of species-specific specializations may be. Thus, a 577 better understanding of visual processing in salamanders will likely be conducive to a more 578 general theory of vision than one that is based on only few select model species. 579 As a system for studying the early visual system, the salamander has had a fascinating 580 tour over the last hundred years. It started with the discovery that the large cells of the 581 salamander's neural system provide excellent access for experimental investigations. And 582 the rest -as they say -is history. A history that has greatly influenced the fields of 583 neuroanatomy, neurochemistry, neurophysiology, as well as computational neuroscience 584 and should continue leaving its mark. 585 -Projektnummer 154113120 -SFB 889, project C1). 590