Remodeling of the Cone Photoreceptor Mosaic During Metamorphosis of Flounder (Pseudopleuronectes Americanus)

Original Paper

Brain Behav Evol 2006;68:241–254 Received: October 26, 2005 DOI: 10.1159/000094705 Returned for revision: December 20, 2005 Accepted after revision: May 26, 2006 Published online: July 20, 2006

Remodeling of the Cone Photoreceptor Mosaic during Metamorphosis of Flounder (Pseudopleuronectes americanus)

a b a, c Kim L. Hoke Barbara I. Evans Russell D. Fernald

a b Program in Neurosciences, Stanford University School of Medicine, Stanford, Calif., School of Biological Sciences, c Lake Superior State University, Sault Sainte Marie, Mich. , Department of Biological Sciences, Stanford University, Stanford, Calif. , USA

Key Words Introduction Retina Opsins Differentiation Cell division Development Patterning Cone Photoreceptor A major issue facing developmental biologists is how Mosaic Teleost undifferentiated cells give rise to adult patterns with multiple different cell types. Retinal development has been a focus for studies on cell differentiation because photore- Abstract ceptors form precise arrays or mosaics. These regularly The retinal cone mosaic of the winter flounder, Pseudopleu- spaced arrangements of cells have predictable cell sub- ronectes americanus, is extensively remodeled during meta- types relative to neighboring cell types. Photoreceptor morphosis when its visual system shifts from monochromat- phenotypes within a mosaic can be distinguished by ic to trichromatic. Here we describe the reorganization and morphology and by expression of the light-sensitive op- re-specification of existing cone subtypes in which larval sin photopigments. Two primary hypotheses have cones alter their spatial arrangement, morphology, and op- emerged to explain photoreceptor mosaic patterning in sin expression to determine whether mechanisms control- vertebrates: one postulating that a series of sequential in- ling cell birth, mosaic position, and opsin selection are coor- ductive events specify cell fate based on preexisting mo- dinated or independent. We labeled dividing cells with saic position as found in Drosophila photoreceptor devel- tritiated (3 H) thymidine prior to mosaic remodeling to deter- opment [Raymond et al., 1995; Stenkamp and Cameron, mine whether existing cone photoreceptors change pheno- 2002; Raymond and Barthel, 2004]; and a second hypoth- type. We also used in situ hybridization to identify mosaic esis proposing that differential adhesion between cell type and opsin expression in transitional retinas to under- subtypes establishes position within the mosaic based on stand the sequence of transformation. Our data indicate that previously determined cell fate [Galli-Resta, 2001; Mo- in the winter flounder retina the choice of new opsin species chizuki, 2002; Reese and Galli-Resta, 2002; Tohya et al., and the cellular rearrangement of the mosaic proceed inde- 2003]. These two hypotheses generate distinct predic- pendently. The production of the precise cone mosaic ar- tions as to the order of mosaic formation and cell differ- rangement is not due to a stereotyped series of sequential entiation. The rapidity of retinal development in most cellular inductions, but rather might be the product of a set retinal model systems makes distinguishing the timing of distinct, flexible processes that rely on plasticity in cell of developmental events difficult. Studies of retinal devel- phenotype. Copyright © 2006 S. Karger AG, Basel opment in these model systems are complemented by

© 2006 S. Karger AG, Basel Kim Hoke University of Texas at Austin Fax +41 61 306 12 34 1 University Station C0930 E-Mail [email protected] Accessible online at: Austin, TX 78712 (USA) www.karger.com www.karger.com/bbe E-Mail [email protected] or by changes in existing cone morphology, opsin chromophore, or opsin subtype [Evans and Fernald, 1993; Ev- ans et al., 1993; Archer et al., 1995; Hope et al., 1998; Shand et al., 1999, 2002; Novales Flamarique, 2000; Zhang et al., 2000; Haacke et al., 2001; Chinen et al., 2003; Cheng and Novales Flamarique, 2004; Mader and Cam- eron, 2004; Takechi and Kawamura, 2005]; (2) cell classes may be lost by cell death [Bowmaker and Kunz, 1987; Beaudet et al., 1993; Novales Flamarique and Hawryshyn, 1996; Novales Flamarique, 2000; Deutschlander et al., 2001; Allison et al., 2003], or (3) cells might move in relation to one another to form different arrangements [Ev- ans and Fernald, 1993; Shand et al., 1999; Haacke et al., 2001; Helvik et al., 2001]. In the winter flounder, Pseudo- pleuronectes americanus, photoreceptors change the peak Fig. 1. We categorized individual cone photoreceptors as members of either hexagonal or square mosaic regions by analyzing wavelength sensitivity and physical arrangement when their nearest neighbors within a unit A1 – A 3 combined with the larvae settle into deeper water, providing an excellent identity of adjacent mosaic units B1, B2. In these schematic illus- model for examining the relationship between photore- trations, circles represent single cone photoreceptors. A1 A com- ceptor cell birth, opsin expression, and physical posi- plete unit of the hexagonal mosaic characteristic of premetamor- tion. phic flounder retina has a central cone (drawn in gray) surrounded by six uniformly spaced single cones. A2 A complete square P. americanus progresses from a pelagic bilaterally mosaic unit characteristic of postmetamorphic flounder retina symmetric fish to a benthic flatfish approximately two consists of a central single cone (drawn in gray) surrounded by months after hatching. During this metamorphosis, im- four pairs of regularly spaced double cones. Note that we refer to portant changes in photoreceptor phenotypes occur as this as a complete square mosaic unit for simplicity, although the fish transition from a monochromatic to trichromatic vi- ratio of double to single cones is actually 2: 1 in the retina. A3 The gray circle indicates half of a double cone in a square mosaic, sual system [Evans and Fernald, 1993]. Premetamorphic which is positioned very closely to one other cone (drawn in black larval retinas contain one photoreceptor type arranged in above the gray circle) and more distantly from other neighbors. a hexagonal array (fig. 1). Photoreceptors in larvae are B1 Two adjacent hexagonal mosaic units are shown as found in middle-wavelength-sensitive cones ( max 519 nm) [Evans pre-transitional flounder retina, with the two central cones et al., 1993] that express photopigment RH2 opsin [Ma- marked by the gray circles. B2 Two adjacent square mosaic units, characteristic of postmetamorphic flounder retina, differ from der and Cameron, 2004]. In contrast, retinas of post- the adjacent hexagonal units in B1 in that the central cones of the metamorphic flounder contain rod photoreceptors as two mosaics, drawn in gray, are not nearest neighbors. well as single and double cones with new opsin types expressed [Evans and Fernald, 1993]. Rods appear during metamorphosis and are inserted among cones without an obvious pattern, as is typical for teleosts retinas. The work in vertebrates in which protracted periods of retinal cones in a postmetamorphic retina form a square mosaic, patterning allow greater resolution of the temporal rela- which is a repeating array of double cones surrounding tionships in photoreceptor differentiation. one single cone (fig. 1). The single cones absorb blue light Species that move to a new habitat as part of their life ( max 457 nm) [Evans et al., 1993] and express SWS2 op- history typically encounter changes in their visual envi- sin mRNA [Mader and Cameron, 2004]. One partner in ronment and remodel their retinas to accommodate such the double cone is maximally sensitive to green light new visual conditions [Beaudet and Hawryshyn, 1999; ( max 531 nm) [Evans et al., 1993] and expresses RH2 op- Evans, 2004]. Because vision can be affected by photore- sin gene [Mader and Cameron, 2004], whereas the other ceptor arrangement and spectral sensitivities, remodel- member of the pair can be maximally sensitive to either ing these features might improve the match between vi- the same ( max 531 nm) or longer ( max 547 nm) wave- sual function and habitat, as described extensively in a lengths [Evans et al., 1993] and express LWS opsin mRNA variety of teleost fish species. Teleost retinal remodeling [Mader and Cameron, 2004]. Why the pre- and post- includes several possible developmental changes: (1) new metamorphic cones that contain RH2 opsin have differ- photoreceptor classes may be added by new cell addition ent spectral sensitivities is unknown [Mader and Cam-

242 Brain Behav Evol 2006;68:241–254 Hoke /Evans /Fernald

eron, 2004]. Neither microspectrophotometric study wild-caught postmetamorphic flounder (MBL). For in situ hy- found evidence for mixing opsin types nor for changes in bridization experiments, flounder were reared in captivity (a gift from M. Costello and L. Lush, Huntsman Marine Science Centre, chromophore usage [Evans et al., 1993; Mader and Cam- St. Andrews, New Brunswick). eron, 2004]. Furthermore, no other RH2 opsin sequences We studied animals at several developmental stages to inves- or alternative splice forms were identified by PCR from tigate changes in retinal structure during metamorphosis. The larval and postmetamorphic tissue [Mader and Camer- developmental stage in winter flounder is most accurately pre- on, 2004; this paper and unpubl. results using four degen- dicted by animal size rather than true chronological age [Wil- liams, 1902; Chambers and Leggett, 1987]. We used lens diameter erate primer pairs]. Thus, during metamorphosis the to reflect animal size because it is straightforward to measure and hexagonal array of single cones becomes a square array provides a reliable estimate of body size, and hence developmental of single and double cones with new opsin genes ex- stage, in two independent analyses (data not shown). We mea- pressed, each with distinct absorbance peaks. sured the largest lens diameter in each series of retinal sections How does this dramatic retinal transformation occur? using previously calibrated microscope images (Image 1.5, Wayne Rasband, NIH) and checked this estimate by multiplying the Postmetamorphic mosaics may be formed from existing number of serial sections that included the lens by the section larval cones or from new cones. Mosaic patterning could thickness, which gave the same result (not shown). rely on mechanisms such as opsin expression directing mosaic position, mosaic position determining opsin sub- Cone Mosaic Patterns and Morphology type, or independent regulation of opsin expression and We assessed cone morphology to determine the location of square mosaics and double cones within transitional retina. Wild- mosaic patterning. Based on light and electron micro- caught winter flounder from the Woods Hole, Mass. area were scopic analysis in metamorphic retina of another flatfish, kept at the Environmental Systems Lab at WHOI until gravid. Shand et al. [1999] found that individual cones fuse to Gametes were taken and eggs fertilized, then eggs developed in form postmetamorphic double cones, but no studies have hatching trays in running seawater (7–10 ° C) on a 14 h light:10 h linked new cell addition, mosaic arrangement, and pho- dark cycle. After hatching, larvae were transferred to black plastic rearing tanks (10–20 l) and fed protozoa, rotifers, copepods or topigment expression during metamorphosis. Here we brine shrimp. We observed several cohorts of larvae through the analyze the retinas of flounder as the cone mosaic in their stages of hatching until metamorphosis was complete and fixed retina is transformed during metamorphosis from a hex- the animals for 2–3 days in Bouin’s fluid [Humason, 1979] prior agonal array with a single cone type to a square array with to tissue processing (lens diameters 68–150 m, 1–120 dph, n = multiple cone types. For brevity we will refer to these as 102 of all stages; n = 20 transitional retina, 120–140 m lens diameters). ‘transitional retinas’. We labeled dividing cells with triti- Fixed flounder were dehydrated through an ethanol series, 3 ated ( H) thymidine prior to mosaic remodeling to deter- embedded in resin (Immunobed, Polysciences, Warminster, Pa., mine whether existing cone photoreceptors change phe- USA) and sectioned on a microtome at 2–3 m through the dor- notype. We examined morphology and used in situ hy- soventral axis. The lack of skeletal calcification in flounder larvae bridization to identify mosaic type and opsin expression allowed sections to be made through the entire animal, simplify- ing tissue orientation, which can be complex due to eye migration in transitional retinas to understand the sequence of at metamorphosis. We made coronal sections from the nasal to transformation and determine whether mechanisms the temporal pole of the eye. Depending on where the parallel sec- controlling cell birth, mosaic position, and opsin selec- tions cut through the eyecup ( fig. 2 ), tissue sections provided ei- tion are coordinated or independent. ther tangential views in which cone mosaic patterns (i.e., hexagonal or square mosaic arrangements) were visible, or, more centrally, radial views in which cone morphology (presence of single or double cones) was assessed. In flounder eyes used to analyze Materials and Methods mosaic patterns and for alkaline phosphatase in situ hybridization (described below), we determined nasotemporal and dorso- Animals ventral position of tissue sections by charting the progressive ap- Treatment of animals was in compliance with protocols appearance of both landmarks within the eye (e.g., optic nerve) as proved at the institutions providing the animals (University of well as external structures such as the brain or nose. We took pho- Oregon, Eugene, Oreg.; Stanford University, Stanford, Calif.; Ma- tomicrographs with a Nikon Eclipse E600 Microscope (Nikon, rine Biological Laboratory (MBL), Woods Hole, Mass., USA; In- Melville, N.Y., USA) and Micropublisher 5.0 RTV camera (QIm- stitute for Marine Bioscience, Halifax, Nova Scotia, Canada; Ma- aging, Burnaby, B.C., Canada) and calibrated them (QCapture- rine Science Center, New Brunswick, Canada). Flounder used for Pro, QImaging). autoradiography and morphology were reared at Woods Hole Oceanographic Institute (WHOI). For cloning opsin cDNA, we Autoradiography used tissue from larval flounder reared in captivity (8–15 days To discover whether cones present in larvae are retained after post hatch (dph); a gift of S. Douglas, Institute for Marine Biosci- metamorphosis as members of double cones in square mosaics, ences, National Research Council, Halifax, Nova Scotia) and from we labeled dividing cells prior to the metamorphic transition and

Remodeling of the Cone Photoreceptor Brain Behav Evol 2006;68:241–254 243 Mosaic 244 Brain Behav Evol 2006;68:241–254 Hoke /Evans /Fernald

assessed the fate of these pre-transition cones following meta- NTB-2, VWR, Westchester, Pa., USA) and exposing the emulsion morphosis. We labeled proliferating cells by treating larvae with to radioactive emissions from the incorporated 3 H-thymidine for tritiated thymidine, which is incorporated into the DNA of mi- several weeks in the dark at 4 ° C. Slides were developed (Kodak totically active cells. We dehydrated 3 H-thymidine (1 g/l, 1 mCi/ D-19 developer, VWR), fixed (Kodak Rapid fix, VWR), and ml; ICN Pharmaceuticals, Costa Mesa, Calif., USA) and then re- stained with cresyl violet, a non-specific cellular stain. hydrated with an equal volume of seawater. Larvae were im- We viewed retinal tissue sections at high magnif ication (Nikon mersed in the 3 H-thymidine labeled seawater for 1 h, and then Eclipse E600 Microscope and Micropublisher 5.0 RTV camera) given two 30-min rinses in non-radioactive thymidine (1 g/l in and identified cells dividing at the time of radiolabeling by the seawater, Sigma, St. Louis, Mo., USA) to clear any residual isotope. presence of black silver grains in the emulsion above the cell body. We sampled two animals immediately in each experiment, trans- We considered a cell labeled if it had at least 5 times the number ferred the remainder to a separate container of fresh seawater, and of silver grains over the nucleus as compared with background reared them through metamorphosis (n = 11 labeled pre-transi- areas. Retinal tissue taken from animals sampled immediately tion and reared until transition). following radiolabeling showed the number and location of cells We processed resin-embedded sections taken from radioac- undergoing mitosis at the time of 3 H-thymidine treatment. Reti- tively labeled tissue (prepared as described above) for autoradiog- nal tissue from animals that survived for extended periods fol- raphy by dipping the slides in photographic emulsion (Kodak lowing 3 H-thymidine incorporation had cells that may or may not have retained the label. For example, cells that stopped dividing and differentiated shortly after 3 H-thymidine incorporation would have retained their label, whereas cells that underwent several mitotic divisions would have diluted the radiolabel to unob- servable levels because the label was divided among all daughter Fig. 2. Square cone mosaic arrangements occur near the periph- cells. Thus radioactively labeled cells observed at long survival eral margin in metamorphic flounder (lens diameter 139 m) as times represented the subset of dividing cells that differentiated the photoreceptor array changes from hexagonal cone mosaics shortly after 3 H-thymidine exposure. typical of premetamorphic animals to square mosaics characteristic of postmetamorphic animals. Schematic illustration of eyes Winter Flounder Opsin cDNA Sequences indicating location and planes of sections pictured in B–E . Because flounder opsin cDNA sequences were not available A Planes of section near the edge of the eye produce tangential when we began these experiments, we isolated opsin cDNA se- views of the photoreceptor mosaic arrangements (B–D ), whereas quences in larval and postmetamorphic flounder. We later found more centrally located planes in the same eye show the photore- these to be 99% identical to published sequences in regions of ceptor layer in radial section (E ). B Low magnification view of a overlap (GenBank accession numbers AY631037, AY631038, section near the nasal margin showing brain in relation to the AY631039), and we follow the published nomenclature through- retina. D = Dorsal; V = ventral. Scale bar = 25 m. C High-mag- out [Mader and Cameron, 2004]. We identified additional nucle- nification photomicrograph of the section in B at the level of the otide sequences from the 3 untranslated region for LWS (Gen- cone photoreceptors. The dorsal portion (top) has cones arranged Bank accession number DQ225171). in a hexagonal mosaic as in premetamorphic flounder (fig. 1; A1), To isolate opsin cDNA sequences, we performed degenerate whereas the ventral portion (bottom) has cones arranged in a reverse transcription polymerase chain reaction (RT-PCR) on square mosaic as in the postmetamorphic retina (fig. 1; A2). The mRNA extracted from retina of postmetamorphic flounder and dashed line indicates a transitional zone between square and hex- from whole larvae. Larvae were flash-frozen whole in liquid ni- agonal mosaics. Scale bar = 10 m. D Section through the tempo- trogen then thawed and stored for subsequent mRNA extraction ral margin of the same retina shows cones arranged in a square (RNAlater; Ambion, Austin, Tex., USA). We anesthetized post- mosaic (top), whereas more ventrally the cones are in a hexagonal metamorphic flounder in 0.3% MS-222 (Sigma) prior to rapid mosaic (bottom). The dashed line indicates transitional zone be- cervical transection. Eyes were removed, lenses dissected out, and tween square and hexagonal mosaics. Scale bar = 10 m. E This complete eyecups frozen in liquid nitrogen for mRNA isolation. radial section shows a double cone (arrow) close to dorsal margin Larval and postmetamorphic tissue was disrupted in liquid nitro- (asterisk). Other double cones are visible in this region, however, gen in a mortar and pestle. We then added lysis solution and ex- more central photoreceptors in the right half of image are single tracted total RNA (MicroPoly(A)Pure kit; Ambion). cones (arrowhead indicates one example). Scale bar = 10 m. We designed PCR primers by first aligning fish cone opsin se- F Higher magnification view of square cone mosaic region on left quences [Clustal W; Chenna et al., 2003 and BlockMaker], then of D . Top shows photomicrograph of cones in square mosaic for- designing degenerate primers [CODEHOP; Rose et al., 1998; mation, with individual cones marked on the image below, and http://blocks.fhcrc.org/blocks/codehop.html]. We synthesized two square mosaic units outlined. Scale bar = 10 m. G Higher single-stranded cDNA (SUPERSCRIPT First-Strand Synthesis magnification view of double cone region in E . Double cone la- System for RT-PCR; Invitrogen, Carlsbad, Calif., USA) and per- beled in E is again indicated by black arrow. Note the nucleus and formed PCR (upper primers 5 TCACCGCCCAGCACAARA- inner segment of labeled cone are thicker compared to the single ARYTNMG or GGCATGCAGTGCTCCTGYGGNCCNGA with cones (H ) and have a dorsoventral division marking paired cones. lower primer 5 TGCTTGTTCATCAGCACGTAGATNAYN Scale bar = 10 m. H Higher magnification view of single cone GGRTT). We amplified RT-PCR reactions in a thermal cycler region in E . Single cone labeled in E is indicated by white arrow- (PTC-100; MJ Research, Boston, Mass., USA; 35 cycles of 94 ° C head. Cone morphology is slender and regular compared to mix 30 s denaturation, 55 ° C 1.5 min annealing, and 72 ° C 3 min ex- of double and single cones in H. Scale bar = 10 m. tension). PCR products were purified (GENECLEAN; Qbiogene,

Remodeling of the Cone Photoreceptor Brain Behav Evol 2006;68:241–254 245 Mosaic Vista, Calif.), cloned into a vector (TOPO TA Cloning Kit for se- pH 9.5, 50 m M MgCl2 , 0.1 M NaCl, 0.1% Tween 20), and incubated quencing; Invitrogen), and then sequenced (Biotech Core, Moun- in BM Purple (Roche) for 4–24 h. When sufficient purple color tain View, Calif., USA). developed in the outer retina to be visible under a dissecting We cloned and sequenced the 3 untranslated region of all scope, we coverslipped slides in Fluoromount-G (Southern Bio- cone opsins to produce in situ hybridization probes to less con- technologies Associates, Birmingham, Ala., USA). served regions of the opsin gene to avoid cross-hybridization be- We v i e w e d t i s s u e u n d e r 4 0 0 ! or 1,000 ! total magnification tween opsin types. We used single-stranded cDNA templates in using brightfield illumination (Axioskop, Zeiss) and captured PCR amplifications with one specific opsin primer designed from digital images using a SPOT digital camera and SPOT Advanced our degenerate PCR sequences and an oligo(dT) adapter using the software (Diagnostic Instruments, Sterling Heights, Mich., USA). conditions described above. Amplified products were purified Preliminary experiments confirmed expression of RH2, SWS2, (GENECLEAN) and either sequenced directly or cloned for sub- and LWS opsin mRNA in postmetamorphic flounder in the same sequent sequencing (TOPO TA cloning). Note that PCR amplifi- pattern as proposed by Mader and Cameron [2004]. cations from larval and postmetamorphic tissue spanning the complete RH2 opsin sequence yielded no differences in coding or In situ Hybridization: Double-Label Fluorescent Visualization untranslated regions between these stages of fish in which cone To localize RH2, LWS, and SWS2 opsin mRNAs simultane- absorption spectra differ [Evans et al., 1993; Mader and Cameron, ously, we used double in situ hybridization on transitional retinal 2004]. Opsin probes used for in situ hybridization ranged in tissue sections cut in varying planes to allow tangential views of length from 300–600 bp. the orderly array of photoreceptors. To increase the number of locations in the eye in which photoreceptor mosaics were sec- In situ Hybridization: Alkaline Phosphatase Visualization tioned tangentially, we rotated the eyecup several times during We performed in situ hybridization on retinal tissue from sectioning such that portions of each eye were cut coronally and flounder in metamorphic transition to describe the general pat- others longitudinally. This sectioning method was biased toward tern of individual opsin expression throughout metamorphosing the nasal and central retina and away from the less accessible tem- retinas. We generated specific RNA probes for in situ hybridiza- poral pole. In contrast to the other tissue sections used for mor- tion for all opsin sequences from TOPO TA templates purified phology, autoradiography, and alkaline phosphatase in situ hy- using Qiaprep Spin Miniprep Kit (Qiagen, Chattsworth, Calif., bridization (described above), we could not reconstruct the dor- USA). Plasmids were linearized using restriction enzymes, puri- soventral or nasotemporal position of each mosaic type within fied with GENECLEAN, then used as template in MAXIscript In these retinas due to the varying section planes. vitro Transcription Kit (Ambion) with either digoxigenin- or flu- After sectioning, slides were heated at 40 ° C or dried in a vac- orescein-labeled UTP (Roche Molecular Biochemicals (Roche), uum chamber for 1 h and subsequently processed. We digested Indianapolis, Ind., USA). RNA probes were precipitated prior to tissue in Proteinase K, re-fixed, rinsed, and dehydrated as above. partial alkaline hydrolysis in 60 m M sodium carbonate and Slides were hybridized overnight at 55 ° C as above with pooled

40 m M sodium bicarbonate pH 10.2 at 60 ° C. We targeted an aver- digoxigenin- (antisense LWS and SWS2) and fluorescein-labeled age size of 200 bp probe fragments for greater access to the target (antisense RH2) probes diluted to 0.25 or 1 ng/ l in hybridization RNA in the tissue. We again precipitated probes and measured buffer. We removed nonspecific hybridization in decreasing SSC yield by comparison with RNA mass standards, then diluted concentrations to a final concentration of 0.1 ! SSC at 55 ° C for probes in 1! hybridization buffer (Sigma) for long-term stor- 20 min. Between steps in the fluorescent detection protocol, we age. rinsed slides in PBS with 0.3% Tween-20. We blocked nonspe- We fixed whole flounder (lens diameters 132–143 m, 42– cific antibody binding for 1 h in 1% blocking buffer in PBS (PBSB;

56 dph) in formalin for 24 h prior to storage in saline, placed tissue NEN, Boston, Mass.) and incubated overnight at 4 ° C or for 3–4 h in 30% sucrose at 4 ° C overnight for cryoprotection, and sectioned at room temperature in sheep anti-digoxigenin antibody conju- embedded tissue (Tissue-Tek O.C.T. Compound; Sakura Finetek, gated with horseradish peroxidase (HRP) (Roche; 1: 250 diluted

Torrance, Calif., USA) in a cryostat (Microm, Zeiss, Thornwood, in PBSB). We applied biotinyl tyramide diluted 1: 50 in Amplifica- N.Y., USA) at 10–16 m. After sectioning, slides were heated at tion Diluent (TSA kit, NEN) for 5–7 min and incubated in avidin-

40 ° C or dried in a vacuum chamber for 1 h and processed imme- conjugated Alexa546 (1: 1,000 in PBS; Molecular Probes, Eugene, diately as follows. After rehydration in 3.7% formaldehyde, we Oreg., USA) for 1.5 h at room temperature. We quenched peroxi- digested tissue in Proteinase K (Ambion; 10 g/ml) for 10 min at dase remaining from the initial probe detection (15 min, 0.3% room temperature. Slides were re-fixed in 3.7% formaldehyde in hydrogen peroxide in PBS) and blocked free biotins remaining PBS, rinsed in 2 ! SSC, then dipped in 70% ethanol and allowed after incubation in the avidin conjugate using Avidin-Biotin to dry. Slides were hybridized overnight at 55 ° C with probes di- blocking kit (Vector), then applied HRP-conjugated anti-fluores- luted to 1 ng/ l in hybridization buffer. We washed slides in de- cein antibody (Roche, 1: 250 in PBSB) for 3–4 h at room tempera- creasing SSC concentrations to a final concentration of 0.1 ! SSC ture. Biotinyl-tyramide treatment (TSA kit, NEN) was followed at 55 ° C for 20 min. Nonspecific antibody binding was blocked by avidin-conjugated fluorescein (1: 1,000 in PBS, Vector) and (5% heat inactivated normal goat serum (Pel-Freez, Rogers, Ariz., coverslipping in Fluoromount-G. USA), 2% Boehringer Blocking Reagent (Roche), 0.1% Tween-20 The density and small size of the cones required confocal mi- in Maleic Acid Buffer (MAB)) for 1 h at room temperature. We croscopy to minimize overlap of signals from cells in different then applied sheep alkaline phosphatase-conjugated anti-digoxi- planes of focus. We used 1- m optical sections to analyze fluo- genin antibody (Roche: 1: 1,000 in blocking solution) overnight at rescently labeled sections using a confocal microscope (Zeiss

4 ° C or at room temperature for 3–4 h. Slides were washed in MAB LSM510, Thornwood, N.Y., USA or Molecular Dynamics Multi- with 0.3% Tween-20, rinsed in AP development buffer (0.1 M Tris Probe 2010 Amersham Biosciences, Sunnyvale, Calif., USA).

246 Brain Behav Evol 2006;68:241–254 Hoke /Evans /Fernald

Preliminary experiments verified that the double label proto- Partial No full square mosaic units ( fig. 1 , A2) col did not produce any cross-reaction between digoxigenin and hexagonal One or two hexagonal mosaic units ( fig. 1 , A1) fluorescein detection in postmetamorphic flounder retinas. Pairs of closely associated cones Omission of digoxigenin-labeled probes, anti-digoxigenin anti- (putative double cones; fig. 1, A3) body, the initial biotinyl tyramide step, or biotinylated Alexa546 Partial One or two full square mosaics ( fig. 1, A2) eliminated fluorescence in the red channel without influencing square Many pairs of closely associated cones fluorescein label. Conversely, omitting components important for (putative double cones; fig. 1, A3) fluorescein detection preserved Alexa546 localization while re- moving green fluorescence. In addition, we performed pairwise Square Two adjacent full square mosaic units ( fig. 1, B2) combinations of in situ hybridization probes to identify co-local- No hexagonal mosaic units ( fig. 1, A1) ization of any opsins in cones of transitional or postmetamorphic fish. RH2 and LWS opsin probes were never co-localized in the These classifications did not vary with confocal plane. In four same cell with each other or with SWS2 opsin expression (not cases we independently determined mosaic type in two or three shown). different confocal planes through the same regions of a single eye, and in every case the duplicates yielded identical classifica- Localization of Opsin Expression within Cone Photoreceptor tions. Mosaics Because retinas from animals in metamorphic transition have some regions with hexagonal mosaics and some with square mo- R e s u l t s saics, we classified the mosaic type using tangential sections of transitional retinas and identified opsins expressed within each mosaic type based on confocal sections of retinas processed with Locating Square Mosaic Initiation in Transitional double label fluorescent in situ hybridization. Images containing Retina separate fluorescent signals were converted to false color (Adobe To discover how re-patterning in the retina proceeds, Photoshop 5, Adobe Systems., Inc., San Jose, Calif., USA) and we classified mosaic type in tangential and radial views merged to combine information on positions of cones expressing opsins typical of postmetamorphic flounder (pooled LWS and of retinas from transitional animals to compare the first SWS2 probes) and opsins expressed in both pre- and postmeta- regions of the retina that become organized into square morphic animals (RH2) [Mader and Cameron, 2004]. We con- mosaics with the first regions to show opsin expression verted images to grayscale for mosaic classification, and marked characteristic of a postmetamorphic retina (described the locations of individual cells with circles. We then categorized below). We used tangential views through cone inner and mosaic type by analyzing the circles alone without the original photomicrograph using criteria described below. After mosaic outer segments to assign mosaics as either square or hex- classification, we compared the mosaic pattern with the original agonal, and radial views through the same eye to confirm color image to determine where the opsins LWS/SWS2 (post- single or double cone morphology. We examined transi- metamorphic) and RH2 (pre- and postmetamorphic) were ex- tional retinas and in all cases found both regions of the pressed within the mosaic. retina with hexagonal mosaics and areas of square mo- We based mosaic classifications on the distinct spatial differences between square and hexagonal mosaics as shown in fig- saic containing both double and single cones (fig. 2). Sin- ure 1. In an ideal hexagonal mosaic, all cone cells are equidistant gle cones arrayed in hexagonal mosaics were visible in the from one another, and each cone can be considered as a central central retina, whereas square mosaics and correspond- cone surrounded by six equally spaced neighbors (fig. 1, A1; one ing double cones occurred near the margins of the eyes. hexagonal ‘mosaic unit’). In a square mosaic on the other hand, Figure 2 includes photomicrographs of square mosaics cone cells have one of two possible positions: cones can be central, surrounded by four pairs of double cones ( fig. 1, A2; one square near the nasal and temporal margins (square mosaics mosaic unit), or they can be part of a double cone, in contact with present at dorsal and ventral margins are not shown). its twin and more distant from the central cone or neighboring Premetamorphic fish (lens diameters less than 115 m) double cones ( fig. 1, A3; one double cone). had pure hexagonal mosaics. Based on these differences in cone neighbor positions, we classified mosaics in transitional flounder retinal tissue processed for double in situ hybridization into four categories: hex- Cell Addition in Transitional Retina agonal, partial hexagonal, partial square, and square. Distortions To determine the origin of cells in the square mosaic in sectioning caused an imperfect recreation of mosaics, so clas- in transitional eyes, we identified dividing cells using tri- sification was based on both the single mosaic components ( fig. 1 , tiated thymidine before retinal transition (68–120 m A1–A3) and on the presence of neighboring mosaic units (hex- lens diameters) and examined mosaic structure in cohort agonal, fig. 1, B1, and square, fig. 1, B2) using the following criteria: siblings either within one day or up to 84 days after label. Just after uptake of label, cells containing 3 H-thymidine Hexagonal Two adjacent hexagonal mosaic units ( fig. 1, B1) were restricted to the periphery of the retina, where pre- No double cones ( fig. 1 , A3) cursor cells divide in the marginal germinal zone ( fig. 3 A).

Remodeling of the Cone Photoreceptor Brain Behav Evol 2006;68:241–254 247 Mosaic Fig. 3. Photomicrographs taken from a cohort of flounder ex- prior to retinal transformation. The image in B is focused on posed to 3 H-thymidine that survived for either 1 day (A), 11 days the photoreceptors whereas that in C shows the overlying silver (B–D ), or 57 days (F– H ) demonstrate that the square cone mosaic grains indicating the 3 H-thymidine label. D , E Higher magnifica- formation includes photoreceptors added before retinal transfor- tion image of box in B . For clarity, an example of a double cone is mation. Panel A is a composite image (Image-Pro Express Ver- outlined in E in an otherwise identical image. F For animals sur- sion5) combining 4 focal planes; all other panels are unmanipu- viving 57 days, postmitotic cells with 3 H-thymidine label are seen lated. Scale bars = 10 m. A Newly divided cells in the retinas of in a band of cells (white arrowhead) central to the dividing mar- flounders that survived for 1 day are restricted to the peripheral ginal cells (asterisk). Black arrow marks the location of a double margin of the retina, where the only cells labeled with 3 H-thymi- cone central to the band of 3 H-thymidine labeling, an example of dine are found (asterisk). The central retina has only single cones a double cone formed from cells that differentiated prior to retinal organized into hexagonal mosaics. We found no portions of the transformation. G, H Higher magnification photomicrograph of retina that had begun the transformation associated with meta- the retinal section shown in F. The micrograph in G shows the morphosis. B , C For flounder surviving 11 days, newly differenti- photoreceptors whereas that in H is focused on the silver grains of ated cells labeled with 3 H-thymidine (between dashed and solid the 3 H-thymidine label. Double cones (black arrow marks one lines) are visible adjacent to the peripheral margin (asterisk). Ret- example) can be seen in retina that is central to the 3 H-thymidine inal transformation has just begun, with a small number of double label in the outer nuclear layer (white arrowhead). Gray arrow- cones visible central to the zone of 3 H-thymidine labeling (out- heads indicate the location of the 3 H-thymidine band in the inner lined by box) that differentiated prior to labeling, and therefore nuclear layer and ganglion cell layer.

248 Brain Behav Evol 2006;68:241–254 Hoke /Evans /Fernald

Note that no double cones were present in these pre-transition eyes. By 57 days after label, a band of 3 H-thymidine labeling was visible at a distance of100 m from the margin, comprised of cells that differentiated within a few cell divisions after the 3 H-thymidine labeling (fig. 3B). Given that new cell addition occurs at the margin, cells peripheral to this band were born after labeling, whereas more central cells were born prior to labeling. We found unlabeled double cones central to this labeled band ( fig. 3 C, D) and thus concluded that these double cones likely differentiated as single cones prior to retinal transformation and subsequently altered their morphology and mosaic position as the retina reorganized.

Localization of Opsin Gene Expression in Transitional Retina To determine whether onset of opsin genes expressed in postmetamorphic flounder (SWS2 and LWS) occurs with the same spatial pattern as the square mosaic formation, we localized RH2, LWS, and SWS2 opsin mRNA in transitional flounder retina using alkaline phosphatase in situ hybridization. We examined both eyes from two fish at three stages to generate a series of images showing the distribution of opsin mRNA expression throughout the transitional eye. Figure 4 shows representative adjacent sections from the youngest transitional fish. Fish at all ages showed the same pattern. RH2 opsin mRNA was present throughout the photoreceptor layer. In contrast, LWS and SWS2 opsin expression was more restricted. LWS and SWS2 opsin mRNA was expressed in two patches in the retina, one in the ventronasal region ( fig. 4 ) and one in a more dorsal and temporal position (not shown). Note that LWS and SWS2 are absent from the temporal half in the ventral retina ( fig. 4 ) including the region near the peripheral margin.

Double in situ Hybridization to Localize Opsin Gene Expression within the Cone Mosaic To determine whether opsin expression changes before or after mosaic rearrangement, we classified mosaic type in tangential sections through transitional retinas

Fig. 4. These representative sections through the retina of a tran- tom illustrates extent of the ventronasal region of LWS and SWS2 sitional flounder show SWS2 and LWS opsin expression in similar opsin expression as well as a second patch of LWS and SWS2 opsin locations in the ventronasal retina. RH2, SWS2, and LWS opsin expression in the dorsotemporal retina (not shown), both of which mRNAs were localized in the photoreceptor layer using alkaline were visible in all ages of metamorphic flounder examined. Sec- phosphatase in situ hybridization. The ventronasal patch of retina tions through more regions dorsal or ventral to these patches con- with LWS and SWS2 opsin expression is indicated by arrows. Note tained only RH2 opsin expression (not shown). ON = Optic nerve that the temporal retina, including the region near the peripheral head; T = temporal; N = nasal. Scale bar = 100 m. margin, lacks LWS and SWS2 opsin expression. Schematic at bot-

Remodeling of the Cone Photoreceptor Brain Behav Evol 2006;68:241–254 249 Mosaic Fig. 5. Postmetamorphic opsin mRNA is expressed within larval- expressing LWS or SWS2 opsin are in regions of RH2-expressing like mosaic regions of retinal tissue from transitional flounder. cells that form a hexagonal (A , B ) or partial hexagonal (C ) mosaic. A1, B1, and C1 are confocal micrographs of tangential retinal sec- A3 and B3 show the outlines of three adjacent hexagonal units in tions showing opsin mRNA expression. Red indicates expression each section (units outlined in blue, turquoise, or purple) required of LWS or SWS2 opsin mRNA (pooled probes), and green label for categorization of these regions as hexagonal mosaics. In C3, indicates RH2 opsin mRNA localization. A2, B2, and C2 show one hexagonal mosaic unit is outlined in blue, but note the paired schematics used to classify the mosaic types present in the confo- cones outlined in green. The occasional paired cones in this re- cal micrographs. The brightly labeled red cells in A1, B1, and C1 gion led to classification of the mosaic in C as partial hexagonal are indicated by red dots for clarity; opsin subtype expression was rather than hexagonal. not considered in mosaic classification. These isolated red cells

( fig. 1 ) and identified the opsin genes expressed in each the alkaline phosphatase expression patterns described mosaic type. RH2 opsin-expressing cones are found in above, we expected that LWS and SWS2 opsin gene ex- pre- and postmetamorphic flounder, but LWS and SWS2 pression would be found in overlapping regions. There- opsin genes are expressed only in postmetamorphic fore, we combined probes for LWS and SWS2 opsin flounder retina [Mader and Cameron, 2004]. Based on mRNAs and considered the combined expression indica-

250 Brain Behav Evol 2006;68:241–254 Hoke /Evans /Fernald

tive of postmetamorphic opsin localization. Each transi- al., 2000; Helvik et al., 2001; Cheng and Novales Fla- tional retina had both regions that were classified as marique, 2004; Cornish et al., 2004; Takechi and square or partially square (similar to postmetamorphic Kawamura, 2005]. Cell spacing in mosaics might proceed mosaics) as well as regions that were hexagonal or par- via tangential cell movement, perhaps employing differ- tially hexagonal (similar to premetamorphic retina). We ential adhesion between neighboring cells based on cell predicted that if opsin expression switches prior to mo- type to produce square or hexagonal mosaics [Mochizu- saic formation, we would find hexagonal arrays with sin- ki, 2002; Reese and Galli-Resta, 2002; Tohya et al., 2003]. gle cones expressing LWS and SWS2 opsin genes. Unincorporated or misplaced cells might be induced ei- We compared mosaic type with opsin expression in 16 ther to switch fate again or be eliminated by cell death, planes of section through 6 transitional eyes to determine although we did not test these possibilities. We argue be- whether opsin type changes before or after mosaic reor- low that our results do not favor an alternative model in ganization. Square or partial square mosaics had regular which nearest neighbor interactions determine cone cell spacing of cones expressing LWS or SWS2 opsin mRNA fate [Raymond and Barthel, 2004]. (n = 8, not shown). Partial hexagonal mosaics had regularly spaced (n = 2; not shown) or scattered cones (n = 1; Newly Formed Square Mosaics Incorporate Existing fig. 5 C) with LWS or SWS2 opsin mRNA. One hexagonal Cone Photoreceptors mosaic contained only RH2-expressing cells (not shown) Transformation of the cone mosaic from a hexagonal whereas others contained a single LWS- or SWS2-labeled mosaic with one cone type in premetamorphic flounder cone in an otherwise RH2-containing region (n = 4; to a square mosaic with three different cone types in post- fig. 5A, B). metamorphic flounder could in principle occur by two mechanisms: reorganization of existing cones that switch phenotypes, or insertion of new cones throughout the Discussion retina. Phenotype switching would entail coordination of both morphological changes (formation of double cones, We examined cone photoreceptor cell birth, mosaic alteration in photoreceptor shape) and molecular chang- arrangement, and opsin mRNA expression during reti- es (shifts in opsin gene expression). Individual photore- nal transformation in the metamorphic winter flounder ceptors alter opsin content throughout development in to understand how cell birth and differentiation are co- other fish [Hope et al., 1998; Cheng and Novales Fla- ordinated to produce precise cone photoreceptor mosaics marique, 2004], although we do not demonstrate opsin characteristic of adults. We found that cone mosaic trans- switching directly in this paper. Previous research in an- formation was initiated near the peripheral margins of other flatfish favored the hypothesis that larval cones re- the retina, and these square mosaics included cones pro- organize to form the square mosaic arrangement: chang- duced prior to the onset of mosaic transformation. We es in cone morphology, including appearance of chains also showed that expression of opsin mRNA types pres- of cones connected by subsurface cisternae, precede ent in postmetamorphic but not premetamorphic retinas square mosaic formation in the sea bream [Shand et al., arose in two distinct patches in the retina. The expression 1999]. Our data complement this evidence suggesting of new opsin types occurred, in part, within cone mosaic that double cones form by fusion of existing single cones. regions that had not yet transformed to the square mo- We labeled dividing cells using 3H-thymidine prior to saic characteristic of postmetamorphic retinas. We con- retinal transformation then examined retinas after trans- clude that retinal changes during metamorphosis are information using 3 H-thymidine presence to identify cells dependently regulated; existing cone photoreceptors re- that underwent terminal division shortly after labeling. organize into new mosaic arrangements and switch their These 3H-thymidine-labeled cells form a ring around the opsin expression with distinctly different spatiotemporal retina (a band in radial section) indicating the position of patterns ( fig. 6 ). the peripheral margin, where cell division occurs, at the We present our data in the framework of an emerging time of label. In radial sections through the retina, cones model of photoreceptor mosaic formation in which cone peripheral to this band are younger than the labeled cells cell fate determines both initial opsin protein content and as they divided after the incorporation of 3 H-thymidine, cone morphology as cells differentiate, but these features whereas cones central to this band are older because they are plastic and may be changed [Szel et al., 1994; Archer likely stopped dividing before exposure to 3H-thymidine et al., 1995; Hope et al., 1998; Shand et al., 1999; Zhang et label. The unlabeled double cones we found central to the

Remodeling of the Cone Photoreceptor Brain Behav Evol 2006;68:241–254 251 Mosaic Fig. 6. Schematic illustration summarizing the patterns of mo- already transformed from hexagonal to square mosaics. In this saic reorganization and opsin changes expected under two hy- case, the dark band indicating predicted LWS/SWS2 opsin local- potheses: (1) mosaic change regulates changes in opsin expression ization should occur as a subset of the light band indicating square or (2) independent regulation of mosaic type and opsin expres- mosaic extent. B If mosaic position and opsin selection are inde- sion. The second hypothesis was supported by our data. In both pendently regulated, we would expect that LWS and SWS2 opsin panels, the observed restriction of square mosaics to the periph- expression (shown here as two dark patches) might have no rela- eral part of the eye is shown in light gray. Predicted extent of ex- tionship to the regions of the retina with cones arranged in square pression of LWS and SWS2 opsin mRNA characteristic of square mosaics (light gray). Our experiments localizing opsins within mosaics in postmetamorphic flounder is shown in dark gray. A If serial radial sections indicated two patches of expression of LWS mosaic position dictates opsin expression, we would expect ex- and SWS2 opsin mRNA as pictured here. pression of LWS and SWS2 opsin mRNA only in regions that have

3H-thymidine band in these postmetamorphic fish had differentiate in situ as double cones containing LWS and most likely differentiated first as single cones prior to ret- RH2 opsins or as single cones expressing SWS2 opsin inal transformation, and later changed from single cones mRNA. We do see 3H-thymidine-positive cells in the out- in a hexagonal mosaic to double cones in a square mo- er nuclear layer, but these centrally dividing cells are re- saic. stricted to the ventral retina and have the morphology of Our data are most consistent with a reorganization of rods not cones [B. Evans and R. Fernald, unpubl. observ.], existing cones during metamorphosis of flatfish to pro- consistent with the addition of rods in the central retina duce the postmetamorphic square mosaic, congruent of in adult teleosts [Johns and Fernald, 1981]. We con- with the conclusions of Shand et al. [1999] in another clude that the precise square mosaic pattern in the post- metamorphosing fish species and the generally accepted metamorphic flounder results from reorganization of ex- mechanism for retinal remodeling. There are two less isting larval cones during metamorphosis, a reorganiza- likely alternative explanations for our data. First, it is pos- tion which necessitates changes in opsin expression, sible that cells born in the periphery during retinal trans- morphological changes such as the fusion of single cones formation move towards the central retina where they to produce double cones, and short-distance migrations express SWS2 or LWS opsin mRNA and become double to align disorganized cells. cones. This explanation is unlikely because a large number of cones would have to migrate long distances and Photopigment Expression and Cell Position Are because we would not expect the 3H-thymidine-labeled Independently Regulated cells to form the compact band we observe if these cells Are the reorganization of the cone mosaic and the had moved through this ring. Second, new cells could be changes in opsin expression during metamorphosis co- born throughout the retina during metamorphosis and ordinated or are these events independently regulated?

252 Brain Behav Evol 2006;68:241–254 Hoke /Evans /Fernald

We present two lines of evidence consistent with the hy- whether earlier markers of early cone fate might be pres- pothesis that cone photoreceptor opsin expression and ent. Thus, we conclude that cone cell fate could play a role cone mosaic arrangement occur independently during in mosaic formation, but that opsin expression and mo- retinal transition. First, the earliest cones expressing LWS saic position likely have separate regulatory influences. and SWS2 opsin mRNA occur in two patches in the ven- Overall, our data suggest that the precise square mo- tronasal and dorsotemporal retina in all transitional ani- saic of cones in postmetamorphic flounder arises without mals. In contrast, square mosaic formation occurred in a a rigid sequential series of cell-cell inductions such as that continuous ring close to the retinal margin throughout responsible for Drosophila ommatidia formation as pro- the retina in flounder of equivalent developmental stages posed by previous researchers [e.g., Stenkamp and Cam- based on lens diameter. These distinct spatial patterns eron, 2002; Raymond and Barthel, 2004]. Not only do we suggest that the onset of postmetamorphic cone opsin support the conclusion that cells shift opsin content and expression and square mosaic formation occur indepen- alter cone type during metamorphic reorganization of dently (fig. 6). Second, we show LWS/SWS2 opsin expres- the cone mosaic, but we also provide evidence that cell sion in regions of the retina containing cones arrayed in position within the mosaic is not the cue guiding opsin hexagonal patterns, a mosaic type which in early pre- selection. Shand et al. [1999] similarly conclude that cone metamorphic flounder only contains RH2 opsin mRNA morphological development proceeds independently [Evans et al., 1993; Mader and Cameron, 2004]. This ob- from cone mosaic formation in the black bream, and Hel- servation precludes the possibility that square mosaic vik et al. [2001] show that opsins restricted to double formation is a necessary prerequisite to LWS and SWS2 cones in postmetamorphic halibut appear in single cones opsin expression, and thus does not support the idea that of larvae. Vertebrate cone mosaic formation, in contrast cone mosaic position dictates opsin protein expression. to the series of fixed near-neighbor inductions in fly om- Conversely, the absence of LWS and SWS2 opsin expres- matidial development, might require individual cones to sion in the ventrotemporal retina (fig. 4), or in the far alter gene expression while jostling for position in the ventral or dorsal retina (not shown) suggests that the crystalline array of receptors on the retina. square cone mosaics in these regions express only RH2 opsin mRNA, indicating that LWS and SWS2 opsin selection are unlikely to direct mosaic organization. Because Acknowledgements our tangential sectioning was biased toward the central and nasal portions of the retina and confocal analyses We thank Mark Costello, Lynn Lush, Susan Douglas, Scott Gallager, Laura McInnery, Michael Cummings, Seung Choi and biased toward regions with LWS or SWS2 expression, we Karen Beck for help with experiments, and Molly Cummings, did not directly demonstrate square mosaics expressing Anna Greenwood, Rachel Henderson, Kathleen Lynch, Rex Ro- only RH2 opsin mRNA. Moreover, the absence of detect- bison, and Ryan Taylor for comments on the manuscript. Sup- able in situ hybridization signal using LWS or SWS2 ported by a Grass Foundation Fellowship to K.L.H., NSF MRI- probes in a region does not prove the absence of low levels 0116086 to B.I.E., and NIH EY 05051 to R.D.F. of opsin proteins in this region, and we did not examine

References

Allison WT, Dann SG, Helvik JV, Bradley C, Beaudet L, Browman HI, Hawryshyn CW (1993) Cheng CL, Novales Flamarique I (2004) New Moyer HD, Hawryshyn CW (2003) Ontoge- Optic nerve response and retinal structure mechanism for modulating colour vision.

ny of ultraviolet-sensitive cones in the retina in rainbow trout of different sizes. Vision Res Nature 428: 279.

of the rainbow trout (Oncorhynchus mykiss) . 33: 1739–1746. Chenna R, Sugawara H, Koike T, Lopez R, Gib-

J Comp Neurol 461: 294–306. Bowmaker JK, Kunz YW (1987) Ultraviolet pho- son TJ, Higgins DG, Thompson JD (2003) Archer SN, Hope AJ, Partridge JC (1995) The toreceptors, tetrachromatic colour vision, Multiple sequence alignment with the Clus-

molecular basis for the green-blue sensitivity and retinal mosaics in the brown trout tal series of programs. Nucleic Acids Res 31:

shift in the rod visual pigments of the Euro- (Salmo trutta): age dependent changes. Vi- 3497–3500.

pean eel. Proc R Soc Lond B 262: 289–295. sion Res 27: 2101–2108. Chinen A, Hamaoka T, Yamada Y, Kawamura S Beaudet L, Hawryshyn CW (1999) Ecological as- Chambers RC, Leggett WC (1987) Size and age (2003) Gene duplication and spectral diver- pects of vertebrate visual ontogeny. In: Adap- at metamorphosis in marine fishes: an anal- sification of cone visual pigments of zebra-

tive Mechanisms in the Ecology of Vision ysis of laboratory-reared winter flounder fish. Genetics 163: 663–675. (Archer SN, Djamgoz MBA, Loew ER, Par- (Pseudopleuronectes americanus) with a re- tridge JC, Vallerga S, eds), pp 413–437. Dor- view of variation in other species. Can J Fish

drecht: Kluwer. Aqua Sci 44: 1936–1947.

Remodeling of the Cone Photoreceptor Brain Behav Evol 2006;68:241–254 253 Mosaic Cornish EE, Xiao M, Yang Z, Provis JM, Hen- Hope AJ, Partridge JC, Hayes PK (1998) Switch Rose TM, Schultz ER, Henikoff JG, Pietrokovski drickson AE (2004) The role of opsin expres- in rod opsin gene expression in the European S, McCallum CM, Henikoff S (1998) Con- sion and apoptosis in determination of cone eel, Anguilla anguilla (L.). Proc R Soc Lond sensus-degenerate hybrid oligonucleotide

types in human retina. Exp Eye Res 78: 1143– B 265: 869–874. primers for amplification of distantly related

1154. Humason GL (1979) Animal Tissue Techniques, sequences. Nucleic Acids Res 26: 1628–1635. Deutschlander ME, Greaves DK, Haimberger 4 Edition. San Francisco: W. H. Freeman & Shand J, Archer MA, Collin SP (1999) Ontoge- TJ, Hawryshyn CW (2001) Functional map- Co. netic changes in the retinal photoreceptor ping of ultraviolet photosensitivity during Johns PR, Fernald RD (1981) Genesis of rods in mosaic in a fish, the black bream, Acan-

metamorphic transitions in a salmonid fish, teleost fish retina. Nature 293: 141–142. thopagrus butcheri. J Comp Neurol 412: 203–

Oncorhynchus mykiss. J Exp Biol 204: 2401– Mader MM, Cameron DA (2004) Photoreceptor 217. 2413. differentiation during retinal development, Shand J, Hart NS, Thomas N, Partridge JC (2002) Evans BI (2004) A fish’s eye view of habitat growth, and regeneration in a metamorphic Developmental changes in the cone visual

change. In: The Senses of Fishes: Adapta- vertebrate. J Neurosci 24: 11463–11472. pigments of black bream Acanthopagrus

tions for the Reception of Natural Stimuli Mochizuki A (2002) Pattern formation of the butcheri. J Exp Biol 205: 3661–3667. (von der Emde G, Mogdans J, Kapoor BG, cone mosaic in the zebrafish retina: a cell re- Stenkamp DL, Cameron DA (2002) Cellular pat-

eds), pp 1–30. New Delhi: Narosa Publishing arrangement model. J Theor Biol 215: 345– tern formation in the retina: retinal regen-

House. 361. eration as a model system. Mol Vis 8: 280– Evans BI, Fernald RD (1993) Retinal transfor- Novales Flamarique I (2000) The ontogeny of ul- 293. mation at metamorphosis in the winter traviolet sensitivity, cone disappearance, Szel A, van Veen T, Rohlich P (1994) Retinal cone

flounder (Pseudopleuronectes americanus) . and regeneration in the sockeye salmon On- differentiation. Nature 370: 336.

Vis Neurosci 10: 1055–1064. corhynchus nerka . J Exp Biol 203: 1161–1172. Takechi M, Kawamura S (2005) Temporal and Evans BI, Harosi FI, Fernald RD (1993) Photore- Novales Flamarique I, Hawryshyn CW (1996) spatial changes in the expression pattern of ceptor spectral absorbance in larval and Retinal development and visual sensitivity of multiple red and green subtype opsin genes adult winter flounder (Pseudopleuronectes young pacific sockeye salmon (Oncorhyn- during zebrafish development. J Exp Biol

americanus). Vis Neurosci 10: 1065–1071. chus nerka). J Exp Biol 199: 869–882. 208: 1337–1345. Galli-Resta L (2001) Assembling the vertebrate Raymond PA, Barthel LK (2004) A moving wave Tohya S, Mochizuki A, Iwasa Y (2003) Differ- retina: global patterning from short-range patterns the cone photoreceptor mosaic ar- ence in the retinal cone mosaic pattern be-

cellular interactions. Neuroreport 12:A103– ray in the zebrafish retina. Int J Dev Biol 48: tween zebrafish and medaka: cell-rearrange-

106. 935–945. ment model. J Theor Biol 221: 289–300. Haacke C, Hess M, Melzer RR, Gebhart H, Smo- Raymond PA, Barthel LK, Curran GA (1995) De- Williams SR (1902) Changes accompanying the la U (2001) Fine structure and development velopmental patterning of rod and cone pho- migration of the eye and observations on the of the retina of the grenadier anchovy Coilia toreceptors in embryonic zebrafish. J Comp tractus opticus and tectum opticum in Pseu-

nasus (Engraulididae, Clupeiformes). J Mor- Neurol 359: 537–550. dopleuronectes americanus. Bull Mus Comp

phol 248: 41–55. Reese BE, Galli-Resta L (2002) The role of tan- Zool Harvard 40: 1–57. Helvik JV, Drivenes O, Harboe T, Seo HC (2001) gential dispersion in retinal mosaic forma- Zhang H, Futami K, Horie N, Okamura A, Utoh

Topography of different photoreceptor cell tion. Prog Retin Eye Res 21: 153–168. T (2000) Molecular cloning of fresh water types in the larval retina of Atlantic halibut and deep-sea rod opsin genes from Japanese

(Hippoglossus hippoglossus). J Exp Biol 204: eel Anguilla japonica and expressional anal-

2553–2559. yses during sexual maturation. FEBS Lett

469: 39–43.

254 Brain Behav Evol 2006;68:241–254 Hoke /Evans /Fernald