The Developmental Progression of Eight Opsin Spectral Signals Recorded from the Zebrafish Retinal Cone Layer Is Altered by the Timing and Cell Type Expression of Thyroxin Receptor β2 (trβ2) Gain-Of-Function Transgenes

Abstract Zebrafish retinal cone signals shift in spectral shape through larval, juvenile, and adult development as expression patterns of eight cone-opsin genes change. An algorithm extracting signal amplitudes for the component cone spectral types is developed and tested on two thyroxin receptor β2 (trβ2) gain-of-function lines crx:mYFP-2A-trβ2 and gnat2:mYFP-2A-trβ2, allowing correlation between opsin signaling and opsin immunoreactivity in lines with different developmental timing and cell-type expression of this red-opsin-promoting transgene. Both adult transgenics became complete, or nearly complete, “red-cone dichromats,” with disproportionately large long-wavelength-sensitive (LWS)1 opsin amplitudes as compared with controls, where LWS1 and LWS2 amplitudes were about equal, and significant signals from SWS1, SWS2, and Rh2 opsins were detected. But in transgenic larvae and juveniles of both lines it was LWS2 amplitudes that increased, with LWS1 cone signals rarely encountered. In gnat2:mYFP-2A-trβ2 embryos at 5 d postfertilization (dpf), red-opsin immunoreactive cone density doubled, but red-opsin amplitudes (LWS2) increased <10%, and green-opsin, blue-opsin, and UV-opsin signals were unchanged, despite co-expressed red opsins, and the finding that an sws1 UV-opsin reporter gene was shut down by the gnat2:mYFP-2A-trβ2 transgene. By contrast both LWS2 red-cone amplitudes and the density of red-cone immunoreactivity more than doubled in 5-dpf crx:mYFP-2A-trβ2 embryos, while UV-cone amplitudes were reduced 90%. Embryonic cones with trβ2 gain-of-function transgenes were morphologically distinct from control red, blue or UV cones, with wider inner segments and shorter axons than red cones, suggesting cone spectral specification, opsin immunoreactivity and shape are influenced by the abundance and developmental timing of trβ2 expression.


Introduction
Spectral patterns in the zebrafish cone ERG shift with development (Saszik et al., 1999;Nelson et al., 2019). These shifts are determined in part by factors regulating opsin expression. We develop an algorithm to extract the electrical contributions of each of the eight zebrafish cone opsins from massed ERG cone signals and use this tool to examine alterations in cone-signal development brought about by perturbations in expression of the regulatory factor thyroid hormone receptor b 2 (trb 2). The process examines the correlation of opsin expression with opsin signals during perturbations in the levels of a key transcription factor for cone development. Thyroid hormone receptor b 2, a splice variant of the thrb gene, is selectively expressed in vertebrate retinal cones (Ng et al., 2001). When deleted, cones expressing opsins in the long-wavelength-sensitive (LWS) subfamily of the opsin molecular phylogenetic tree (Terakita, 2005) fail to be produced. This includes the MWS cones of mouse (Ng et al., 2001;Roberts et al., 2006), human MWS and LWS cone function (Weiss et al., 2012), and both LWS1 and LWS2 red-cone signals of zebrafish (Deveau et al., 2020). The LWS cone subfamily senses the longest wavelengths a species detects, with opsin spectral peaks ranging from 511 nm in mouse (Jacobs et al., 1991) to 625 nm in goldfish (Marks, 1965).
Two gain-of-function transgenics, crx:mYFP-2A-trb 2 (crx:trb 2) and gnat2:mYFP-2A-trb 2;mpv17À/À (gnat2: trb 2; Suzuki et al., 2013), are used to perturb both the developmental timing and the cellular locus of trb 2 expression. The first (crx:trb 2) is active by day 2 in larval development, in embryonic retinal progenitor cells (Shen and Raymond, 2004;Suzuki et al., 2013). The second (gnat2:trb 2), trb 2 is active only in differentiated cone cells. The gnat2 gene is the promoter of the cone transducin a subunit (Brockerhoff et al., 2003;Kennedy et al., 2007). In the embryonic larvae of each transgenic Suzuki et al. (2013) found an excess of red-opsin immunoreactive cones. In crx:trb 2 the larval densities of green-opsin, blue-opsin, and ultraviolet (UV) opsin immunoreactive cones were reduced, whereas in gnat2:trb 2 excess red-opsin immunoreactivity found a home in cones expressing other opsins, forming mixed-opsin cones. These transgenic mixedopsin cones were thought to model rodent cones that natively express both UV and MWS opsins (Applebury et al., 2000). We here confirm the immunoreactive patterns of these transgenics raised under the same conditions used for electrophysiological recordings of spectral signals to determine whether altered opsin patterns have a parallel in cone spectral signals. The physiological consequences of altered trb 2 expression are unknown, and might result in either retinal disease, or a spectrally unique visual system.
Although trb 2 might be thought of as a binary ON/OFF switch for LWS-cone development, it clearly has other actions, and the activity level and cell-type expression of trb 2 may influence the action. Dilution of trb 2 in adult zebrafish trb 2 1/À animals made red cones less dense and depressed long-wavelength sensitivity (Deveau et al., 2020). In embryonic and juvenile zebrafish, while unbound trb 2 sufficed for red-cone development, binding of exogenous thyroid hormone (TH) shifted larval cones expressing LWS2 opsin to expression of LWS1 opsin, and athyroidism switched them back (Mackin et al., 2019). TH depressed the expression of both SWS1-(UVopsin) and SWS2-(blue-opsin) message (Mackin et al., 2019). TH and trb 2 appear further to be involved in the preferential expression of several of the tandem quadruplicate Rh2 cone opsins in zebrafish (Mackin et al., 2019). It appears likely that TH and trb 2 are upstream regulators in the specification of all cone types, though only for red cone neuronal specification is trb 2 essential (Deveau et al., 2020). Therefore, the developmental impact on electrical signaling in all eight zebrafish cone types is investigated.
Spectral physiology was collected from larvae on days 5, 6, and 7 postfertilization (dpf), on juveniles (12 dpf), and a from both male and female adults (8-18 months). Larvae lack gender. For 5-7 dpf, larvae were group incubated at 28°C in 3.5-inch Petri dishes atop a heating pad on a 14/ 10 h light/dark cycle. Larval medium contained 60 mg/l sea salt, 75 ml/l 1.5% methylene blue (Sigma-Aldrich catalog #03978). Five-to 7-dpf larvae were not fed. The methylene blue was omitted for live confocal microscopy because of its fluorescence. Eight-to 12-d larvae were group housed in system nursery tanks (520-650 mV water, 28°C, pH 7.5-7.7) on the same light/dark cycle and fed both Larval AP100 (Pentair Aquatic Eco-Systems) and live rotifers (Brachionus plicatilis; Reed Mariculture).
Some larvae were raised to adulthood (8-18 months) both for spectral studies and for retention as breeders. These were group housed using 1.5-or 3-l recirculating tanks shelved in stand-alone, recirculating, Aquatic Habitats benchtop systems (Pentair Aquatic Eco-Systems, Apopka, Florida) on the facility 14/10 h light/dark cycle (Zeitgeber 0 = 8:00 A.M.). Adults were fed pulverized TetraMin flakes (Tetra GMBH) and live rotifers. The protocols for breeding and experimentation were approved by the National Institute of Neurologic Disorders and Stroke/National Institute on Deafness and Other Communication Disorders/National Center for Complementary and Integrative Health IACUC (ASPs 1307, 1227).

Preparation and perfusion of isolated eyes for physiology
Larvae were captured with disposable pipettes and placed on a glass lantern slide. After removing excess water, larvae were adsorbed onto a chip of black  (WT). The heterozygotes are recognized by both pupil (p) and heart (h) fluorescence. The yolk sac (y) is autofluorescent. B, The cornea of a 5-d transgenic eye isolated from a crx:mYFP-2A-trb 2 larva is penetrated with a patch electrode for ERG recordings. The eye is ,0.5-mm diameter. C, The gnat2:mYFP-2A-trb 2;mpv17À/À gain-of-function phenotype is studied on a roy orbison (mpv17À/À) background strain. The darkly pigmented, nonreflective iris of this control strain aids in visualizing the dim transgenic fluorescence of the pupil (p). The yoke (y) is autofluorescent. D-F, Live confocal imaging of retinas in 6-dpf larvae. D, WT red (red) and UV (green) cone morphology visualized with trb 2:tdTomato and sws1:GFP fluorescent reporter transgenes. E, The mYFP construct in the crx:mYFP-2A-trb 2 transgene causes cones (C) and bipolar cells (BC) to fluoresce. F, The mYFP construct in the transgene gnat2:mYFP-2A-trb 2 marks only cone cells. nitrocellulose filter paper (Millipore, 0.45 mm pore, catalog #HABP02500, MilliporeSigma), and decapitated (without anesthetic) using a long (37 mm) insect pin (Carolina Biological Supply). Using a binocular microscope (MZ12-5; Leica Microsystems), a longitudinal, dorsal-ventral cut through the head proptosed and isolated larval eyes, which were positioned facing up, taking care not to touch the eye directly. In the recording chamber, larval eyes mounted on the nitrocellulose chip were perfused at 0.1 ml/min with Minimal Essential Medium (MEM; ThermoFisher Scientific catalog #11090-099, equilibrated with 95% O 2 and 5% CO 2 ) using a syringe pump (New Era 500L; Braintree Scientific) and a 28-guage microfil syringe needle as applicator (MF28G67; World Precision Instruments). The chamber was an inverted lid for a 35-mm culture dish (ThermoFisher Scientific), with a disk of 41-mmmesh nylon filter (Millipore) covering the bottom to wick away perfusate. The perfusion applicator was positioned on the nylon mesh; 20 mM L-aspartate (Sigma-Aldrich), added to the MEM perfusate, blocked postsynaptic, glutamatergic, photoreceptor mechanisms leaving only photoreceptor signals (Sillman et al., 1969). Aspartate medium blocks cone synaptic transmission through saturation of glutamatergic receptors of three types: ionotropic cation channels, metabotropic-mediated cation channels, and glutamatetransporter-mediated anionic channels Dowling, 1995, 1996;Connaughton and Nelson, 2000). Patch electrodes (3-mm tip) were inserted trans-corneally ( Fig. 1B) to record the isolated cone ERG signals (cone PIII; Wong et al., 2004).
Adult eyecups were prepared from eyes removed from fish decapitated with a fresh, single-edged razor. Corneas and lenses were removed from the isolated eyes mounted upright on a 5-to 10-mm square of black nitrocellulose paper, and the preparation was placed in a recording chamber (as above). The perfusion applicator was placed directly above the retina, oxygenating the vitreal surface with MEM containing 10 mM L-aspartate at 0.3 ml/min. Microelectrodes broken to 300 mm tip diameter placed in the eyecup recorded cone-PIII ERG signals (Nelson and Singla, 2009).

Live imaging of larval retinas
Transgenic larvae were raised at 28°C in 300 mM phenylthiourea (PTU; Sigma-Aldrich) to prevent melanin formation in the pigment epithelium and allow imaging of the retina in vivo. At 6 dpf, each larva was anesthetized with Tricaine Methanesulfonate (MS222, Sigma-Aldrich) and mounted individually in 1.5% agarose (Sigma-Aldrich type VII-A) on an eight-chamber slide with the right eye against the cover glass floor of the chamber. Eyes were imaged in zstacks on Zeiss 880 confocal microscope at either 25Â or 40Â magnification at 1024 Â 1024-pixel resolution. Cone morphometrics were measured for 5 or 6 fish in each transgenic line in Fiji (ImageJ) on the optical slice that offered the longest stretch of resolved cones. These measurements were analyzed for differences using a one-way ANOVA and Tukey's post hoc test. Fluorescent reporters identified the morphology of wild-type (WT) red and UV cones in eyes from trb 2:tdTomato;sws1:GFP larvae (Fig. 1D) and of transgenic gnat:mYFP-2A-trb 2 cones (Fig. 1F). In crx: mYFP-2A-trb 2, reporter fluorescence appeared in both cones and bipolar cells (Fig. 1E) as previously seen in antibody staining for Crx in zebrafish (Shen and Raymond, 2004).

Spectral stimulus protocol
Larval eyes and adult eyecups were stimulated with nine wavelengths ranging from 330 to 650 nm (20-nm half-width interference filters, 40-nm increments, Chroma Technology). Seven intensities were presented at each wavelength (UV compliant neutral density filters, 0.5 log unit increments covering 3 log units, Andover Corporation). A calibrated photodiode (Newport Corporation) was used to determine stimulus irradiance in quantal units. This was placed in the plane of the cornea for larval eyes or the plane of the retina for adult eyecups. All spectral model calculations are based on absolute, wavelength-specific photodiode calibrations of quanta delivered to the eye. The light source was a 150-W OFR Xenon arc with two optical channels gated by Uniblitz shutters (Vincent Associates). The stimulus channel passed through three Sutter filter wheels, through a UV-visible compliant liquid light guide (Sutter Instruments), through the epifluorescence port of the BX51WI upright microscope (Olympus-Life Science Solutions), and through either a 10Â UPlanFLN/0.3 microscope objective (larvae) or a 4Â UPlanSApo/0.16 objective (adults). The second optical channel passed through hand-inserted filters and an infrared (IR) compliant liquid light guide (Newport Corporation) providing infrared (IR) side illumination for visualization and "dark" backgrounds, or through a 627-nm interference filter for red adapting backgrounds.
To record a spectral dataset, the stimulating objective was positioned over the eye with a translation stage (MT-800; Sutter Instrument). Microelectrodes were inserted into eyes or eyecups with a micro-positioner (Sutter Instrument, MPC-385). ERG signals from the microelectrode were amplified by 10,000 (World Precision Instruments, DAM80, 0.1-Hz to 1.0-kHz bandpass), and digitized (2000 Hz) with an Axon instruments 1440A (Molecular Devices) using Clampex 10 software. Setting the Clampex averaging feature to retain all the elements of an average, the 280 ERGs within a single spectral dataset were captured in a single file.

Analysis of ERG signals
Datasets were imported into Origin analysis software and processed using Origin LabTalk scripts (Origin, various versions; Originlab Corporation). The four replicatewaveforms at each of the 70 wavelength and irradiance combinations were averaged and boxcar filtered (17 ms, one 60-Hz line-frequency cycle). Peak-to-trough amplitudes were extracted during the 850-ms following stimulus onset, an interval including both the hyperpolarizing trough and repolarizing peak of the aspartate-isolated cone-PIII response. There was no b-wave component in this signal (Nelson and Singla, 2009). Each amplitude was associated in Origin with the wavelengths and irradiances of the stimuli, providing 70 wavelength, irradiance, and amplitude data points for each spectral dataset. Datasets with unstable response amplitudes over the collection period were rejected. For each genetic variant, multiple datasets from ;10 eyes were normalized to the maximal response of each dataset to form a cumulative dataset, which was fit to a spectral algorithm. The normalization weighted the individual datasets making up the cumulative dataset equally. Combining datasets from many eyes separated the trends in genetic alterations of spectral properties from the variations among individual eyes. Nonlinear fits of models to spectral datasets used the Levenberg-Marquardt iteration algorithm provided by Origin. The algorithm finds the best spectral model and extracts amplitudes of various cone signals together with standard errors of estimate (SEs) from the cumulative cone-PIII response.
The above spectral protocol does not record zebrafish rod signals as the retinas are not dark adapted, saturating stimuli are used, and the interstimulus intervals are short (Bilotta et al., 2001). There are four rods surrounding each UV cone in adult zebrafish (Fadool, 2003) but less than one rod for every four UV cones in 5-dpf larvae (Alvarez-Delfin et al., 2009). Rod b-waves amount to ,10% of maximal b-wave amplitude in larval zebrafish and can only be discerned as a slower peaking threshold response after 2 h of dark adaptation (Venkatraman et al., 2020). Slow-peaking (;400 ms) rod b-waves are recorded in dark adapted adults by avoiding suprathreshold stimuli and with interstimulus intervals of 15 s or greater (Bilotta et al., 2001).

Statistical analysis
We use t tests, F tests, and ANOVAs to compare results from different treatments or measurements. To determine the most appropriate spectral models, F tests on residual variances for different spectral models were employed. For statistical tests, GraphPad Prism software (RRID:SCR_ 002798), web-based calculators, and statistical functions addressable within Originlab Labtalk software were used.

Finding the cone combinations best representing spectral datasets
Eight cone opsins are expressed in zebrafish retina (Chinen et al., 2003). The algorithm to identify the opsin signals generating the ERG spectral shape is based on the axiom that the tiny radial photocurrents of individual cones sum linearly in extracellular space to produce a net extracellular current which, by traveling through extracellular resistivity, generates an ERG photovoltage. We assume individual cone photocurrents relate to irradiance through Hill functions of exponent 1.0 and semi-saturation irradiance s (Baylor and Fuortes, 1970) and that s varies with wavelength in inverse proportion to each opsin absorbance. This scheme is represented in Figure 2A (Eq. 1). V is the net summed photovoltage in the cone ERG (cone PIII) which depends on I the stimulus irradiance in quanta and wl the wavelength. The Vm i values are nonlinear fit values, the maximal or saturation voltages for each of the i cone types. The semi-saturation irradiance for the i th cone is k i, as evaluated at the i th cone absorbance maximum. The k i values differ among cone types relative to each other. The relative k i values, expressed as log(k i ) relative to UV-cone sensitivity, are literature values (Nelson et al., 2019) listed in Figure 2C. The UV-cone value for log(k) is fit by the algorithm. A(wlmax i , wl) is absorbance as a function of wavelength (wl) for the i th cone, whose wavelength maximum is wlmax i . The maximum wavelengths are literature values (Nelson et al., 2019). The Dartnall nomogram (Dartnall, 1953) is used as the absorbance function A(wlmax i , wl). This approximation has the convenience of making opsin absorbance across wavelengths a function of a single variable, wlmax i . The Dartnall nomogram posits that opsin absorbance shape is constant when plotted on a reciprocal wavelength axis. The template nomogram functions derive from suction electrode recordings of cones in Danio aequipinnatus (Palacios et al., 1996), a relative of zebrafish (Danio rerio). These templates are represented as 8 th order polynomials. We use a single template polynomial (Hughes et al., 1998) for all red, green, and blue cones listed in Figure 2C, but a separate polynomial for UV cones (Palacios et al., 1996). The resulting absorbance shapes appear in Figure 2B. Altogether there are nine parameters fit by the algorithm: eight Vm i values, and a single k i value (k UV ).
The spectral algorithm ( Fig. 2A) sums signals from eight cone types (Fig. 2B,C). These include a UV cone, two blue cones (B1, B2), three green cones (G1, G3, G4) and two red cones (R1, R2). The gene equivalencies of this nomenclature and the functionally measured wavelength peaks appear in Figure 2C. The spectral algorithm chooses among 28-1 or 255 unique cone combinations that might best represent a cumulative ERG spectral dataset. Each cone combination is called a model. The best model is chosen based on four constraints: (1) the model iteration must converge; (2) all model Vm i values must be significantly greater than zero (t test, p 0.05); (3) all Vm i values must be ,2.0, so as not to greatly exceed the largest amplitudes in the cumulative datasets; and (4) the value of r 2 for the fit must be larger than that of any other model, as determined by the F test for residual variance. If equivalent models are found (F test, p ! 0.95), they will be noted. A similar modeling algorithm has been used to determine the cone combinations impinging on larval ganglion cell impulse discharges (Connaughton and Nelson, 2021). . Aspartate-isolated cone signals from zebrafish eyes are a summation of signals from eight cone types, each distinguished by different maximal amplitudes (Vm i ), semi-saturation irradiances (k i ), and opsin peak absorbances (wlmax i ). There are 255 unique combinations of eight cones, each combination is a candidate to best model cone spectral responses in the ERG. B, Spectral shapes of opsin absorbances [A(wlmax i , wl)] are generated from 8th-order template polynomials (Palacios et al., 1996;Hughes et al., 1998) using Dartnall nomogram translations along the wavelength axis to represent opsins of different peak wavelengths (Dartnall, 1953). C, Parameters for each of the i cone types are numbered in order from short to long wavelengths. SWS, short-wavelength-sensitive opsins; RH2, rhodopsin-like green-cone opsins; LWS, longwavelength-sensitive opsins.

Results
Cone distributions in the larval retinas of wild-type and crx:trb2 transgenics Suzuki et al. (2013) developed the crx:mYFP-2A-trb 2 gain-of function transgenic as a rescue line to restore redcones during a morpholino blockade of native trb 2 that abolished their formation. In zebrafish the crx gene promotor becomes activate at the retinal progenitor stage (;2 dpf; Shen and Raymond, 2004). The gain-of-function crx:trb 2 transgene replaces the missing retinal trb 2, but not in the same cell types or at the same developmental stage. In the experiments of Suzuki et al. (2013), it was nonetheless effective. Red-opsin immunoreactive cones were restored with supernormal density, but there was suppression of green-opsin, blue-opsin, and UV-opsin immunoreactive cones (Suzuki et al., 2013).
In Figure 3A,B, the experiment of Suzuki et al. (2013) is repeated on crx:mYFP-2A-trb 2 larvae without morpholino suppression of native trb 2. In wild-type retinas, immunoreactive mosaics of both UV cones (Fig. 3Ai) and red cones ( Fig. 3Aiii) stain with opsin antibodies. Superposition of both mosaics (Fig. 3Aii) show no opsin overlaps, with each cone type expressing a single opsin. The situation is similar for blue cones, which exist in a mosaic pattern separate from red cones (Fig. 3Aiv,v,vi).
Cone spectral signals from wild-type and crx:trb2 retinas in larvae and adults The spectral pattern of aspartate-isolated cone signals from larval crx:mYFP-2A-trb 2 eyes parallels the altered cone densities. With short and UV wavelength stimulation (410 and 330 nm; Fig. 3C), an isolated WT eye responds with substantial signals for three of the brightest stimulus irradiances (green, dark yellow or red traces) but at these same wavelengths, signals from a crx:trb 2 eye (Fig. 3D) are evident at only the two brightest irradiances (dark yellow and red traces) and are of lesser amplitude than WT. Maximal amplitudes were evoked at 490 nm in the WT eye ( Fig. 3C), while amplitudes in the crx:trb 2 eye plateaued at longer wavelengths (490, 570, and 650 nm; Fig. 3D). The loss of amplitudes at 490 nm ( Fig. 3D) is not a rod signal loss, as no change in the time to peak of threshold signals is apparent.
Adult spectral differences in crx:mYFP-2A-trb 2 and WT eyecups were even more pronounced than those in larval eyes. In the WT traces, the maximal amplitude occurred at 490 nm ( Fig. 4A), whereas the maximal amplitude from a crx:trb 2 eyecup was seen at 650 nm ( Fig.  4B). Amplitudes at 410 and 330 nm (Fig. 4B) were greatly reduced as compared with the WT control ( Fig. 4A). At these short wavelengths WT signals were more sensitive than crx:trb 2 signals, with the dimmer stimuli ( Fig. 4A,B, gray and blue traces) evoking deflections in WT, but only overlapping baselines in crx:trb 2 eyecups. The slowerpeaking, mid-spectral threshold signals of rods are not evoked in either WT or crx:trb 2.
"PIII ON trough" is the time interval from stimulus onset to the minimum in the vitreal negative phase of the cone PIII signal. In larvae, PIII-ON-trough latencies for the crx: mYFP-2A-trb 2 waveforms were similar to WT, but in adults the latency to the PIII ON trough was shorter (Fig.  4D). Response peak and trough timings were measured on the mean waveform for all 280 responses in a spectral dataset. In this average response, noise is minimized and interferes least with determination of extrema timing. For larval eyes the latency to the cone PIII trough for 29 crx: trb 2 datasets (12 eyes) was 135 6 6 ms from stimulus onset (mean and SE). The trough time for 28 WT datasets (10 eyes) was 134 6 5 ms. The WT and transgenic trough times were not significantly different (t (55) = 0.218, p = 0.828) and the trough-time variances were also similar (F (1.80,28,27) = 0.935, p = 0.133; Fig. 4D). In 16 adult crx:trb 2 eyecups the mean latency to PIII ON troughs was 111 6 3 ms (21 datasets). In 14 WT eyecups the PIII trough latency was 124 6 5 ms (20 datasets). The adult crx:trb 2 transgenics were somewhat quicker . Opsin distributions and spectral responses in embryonic wild-type and crx:mYFP-2A-trb 2 larval eyes. Ai, UV opsin (SWS1) immunoreactive cones in a WT retina. Aiii, Avi, Red opsin (LWS1, LWS2) immunoreactive cones in WT retinas. Aiv, Blueopsin (SWS2) immunoreactive cones in a WT retina. Aii, UV and red opsins are expressed in separate cones in a WT retina. Av, Red and blue opsins are expressed in separate cones in a WT retina. Bi, UV-opsin immunoreactive cones in a crx:trb 2 retina. Biii, Bvi, Red-opsin immunoreactive cones in crx:trb 2 retinas. Biv, Blue-opsin immunoreactive cones in a crx:trb 2 retina. There are fewer UV and blue cones in crx:trb 2 retinas than in WT retinas. Bii, One crx:trb 2 cone is immunoreactive for both UV and red opsins (arrowhead). Bv, Arrowhead points to a crx:trb 2 cone immunoreactive for both red and blue opsins. C, Cone signals from a WT larval eye respond to all stimulus wavelengths with largest amplitudes at 490 nm. D, A larval crx:trb 2 retina responds with maximal amplitudes at wavelengths 490, 570, and 650 nm but is less responsive than WT for 330, 410, and 490 nm (arrows), wavelengths that stimulate blue and UV cones. A, B, 5-dpf larvae. C, D, 6-dpf larvae. Perfusion medium contains 20 mM aspartate to isolate photoreceptor signals in the ERG. Five of nine stimulus-protocol wavelengths are illustrated. The stimulus irradiances [units of log(quanta·mm À2 ·s À1 )] appear in legends to the right of stacked irradiance-response traces.
"PIII OFF peak" (Fig. 4E) is the time interval from stimulus offset to the maximum of the upward course in the PIII Figure 4. Adult cone spectral signals in WT and crx:mYFP-2A-trb 2 retinas. A, In WT eyecups, all wavelengths and most irradiances evoke signals, with a maximal response at 490 nm. B, Cone signals from crx:trb 2 eyecups are maximal at 650 nm, where they are relatively greater than WT (up arrow). Response amplitudes to wavelengths that stimulate blue and UV cones (330, 410, and 330 nm) are relatively less than WT (down arrows). In A and B, adults are 8-18 months. The perfusion medium contains 10 mM aspartate to isolate photoreceptor signals. The nested responses are irradiance-response series at each wavelength, with irradiances given in legends to the right of the traces in units of log(quanta·mm À2 ·s À1 ). C-E, Properties of ERG cone-PIII waveforms in larval and adult, WT and crx:trb 2 retinas. C, Distributions of maximal trough-to-peak amplitudes in larval and adult spectral datasets. D, Cone-signal latency from stimulus onset to the minimum of the ON trough of the cone signal are measured on the mean responses of each 280-stimulus spectral dataset. E, Cone-signal latency to the OFF peak in the mean responses of spectral datasets. OFF latencies are measured from stimulus offset. Asterisks (n.s., not significant) are probabilities that WT and crx:trb 2 distributions differ in larvae or adults (GraphPad Prism convention, statistics given in text). OFF signal. The larval PIII-OFF-peak latencies for crx: mYFP-2A-trb 2 (360 6 36 ms) did not differ significantly from WT controls (313 6 17 ms, t (55) = 1.171, p = 0.247), although the variance in crx:trb 2 timing was significantly greater (F (4.68,28,27) = 0.999, p = 1.3 Â 10 À4 ). In adults the crx:mYFP-2A-trb 2 OFF peak (262 6 14 ms), like the ON trough, occurred significantly sooner than in WT (329 6 20 ms, t (39) = 2.728, p = 0.0095). The variances of adult OFF-peak latencies were similar in crx:trb 2 and WT adults (F (1.99,19,20) = 0.932, p = 0.135; Fig. 4E). For WT controls, larval and adult PIII OFF peak timing was not significantly different (t (46) = 0.600, p = 0.551). For crx:trb 2 transgenics PIII-OFF-peak times for adults were somewhat quicker than for larvae (t (48) = 2.234, p = 0.030). Taken together, the latency distributions (Fig. 4D,E) show little evidence of waveform abnormalities associated with errors in cone phototransduction or retinal disease. The quicker waveforms of crx:trb 2 adults may result from different weighting of contributing cone types.
In crx:trb2 larvae, red-cone signals increase but UV-cone and blue-cone signals diminish To determine the larval cone contributions affected by the crx:mYFP-2A-trb 2 transgene, 255 models comprising all combinations of eight cone spectral types, were fit to crx:trb 2 and WT cumulative spectral datasets. For the 6dpf WT and crx:trb 2 larvae, the cumulative datasets included 1858 response amplitudes (28 datasets) and 1860 amplitudes (29 datasets), respectively. Optimal models differed (WT, #111; crx:trb 2, #202). The residual variance of the WT model (#111) as fit to the crx:trb 2 cumulative dataset differed from the residual variance of the crx:trb 2 dataset fit to its own optimal model 202 (F (9.16,1821,1825) = 1, p = 0), indicating that the WT model is not an equivalently good representation of crx:trb 2 data.
In Figure 5A,B, amplitudes are plotted against irradiance for four of the nine test wavelengths. Continuous curves are calculated from the optimal models for WT and crx:trb 2 datasets. These curves are generated from a global model fit to all wavelength data, not just the illustrated data. The adherence of irradiance-response curves to datapoints for individual wavelengths is an index of the ability of the global models to represent the cumulative spectral data. The distribution of points and curves is more compressed along the irradiance axis for crx:trb 2 larvae (Fig. 5B) than for WT controls (Fig. 5A). The points and curve at 370 nm (magenta) lie to the left of the 490nm points and curve (green) for WT spectral signals ( Fig.  5A) but to the right of the 490-nm curve and points for crx: trb 2 larvae (Fig. 5B), suggesting a loss of sensitivity for 370 nM-sensing cones (UV cones). The log of semi-saturation irradiance calculated for the R1 and R2 cones was 4.56 6 0.018 for WT larvae and 4.53 6 0.027 for crx:trb 2 (Fig. 5A,B). These semi-saturation irradiances did not significantly differ (t (3681) = 1.858, p = 0.063).
Modeled spectral sensitivities represent the impact not only of Vm i but also k i values and include Hill-function response compressions, which flatten cone spectral functions for bright stimulation (Eq.1; Fig. 2A). This is the fullest model representation, but less interpretable in terms of individual cone contributions. The spectral curves of Figure 5D are modeled response amplitudes for three levels of constant quantal stimulation across the spectrum. The strong UV signal in WT larval eyes (Fig. 5A) appears as an ultraviolet spectral peak (;370 nm) in Figure 5D, regardless of stimulus brightness, despite the larger saturation amplitudes of the R2 (Vr556) cones, which, in WT controls, manifest as a long-wavelength bulge more prominent with greater constant-quantal irradiances, which better stimulate the higher semi-saturation characteristics of R2 physiology. The failure to find UV cone signals in the crx:trb 2 larval transgenics precludes an ultraviolet spectral peak at any quantal-irradiance and allows red-cone signals to create spectral peaks (540-550 nm). The residual blue and green signals of crx:trb 2 cooperate to create a 444-nm peak at the dimmest level of stimulation (Fig. 5D).
short-wavelength (blue and green) cones, absent in the transgenic. As judged by residual variance no other models were indistinguishable from the illustrated ones for either crx:trb 2 or WT adults (F tests for all other models, p , 0.95).
Larval gnat2:trb2 cone types and spectral signals The gnat2:mYFP-2A-trb 2;mpv17 transgenic was developed to test whether red cones could be restored to a population of mid and short wavelength cones already differentiated under morpholino blockade of the native trb 2 gene, which prevented red-cone development (Suzuki et al., 2013). The gnat2 locus codes for cone transducin a, a gene product only expressed in differentiated and functional cone cells. It is expressed by four dpf. Suzuki et al. (2013) found the gnat2:trb 2 transgene induced a supranormal 5-dpf density of red-opsin immunoreactive cones in the absence of native trb 2, but unlike the crx:mYFP-2A-trb 2 rescue, green-opsin (Rh2), blue-opsin (SWS2), and UV-opsin (SWS1) immunoreactive cone densities were normal (Suzuki et al., 2013).
Larval and adult cone spectral signals from control and gnat2:trb2 retinas Despite greater red-cone density and numerous mixed opsin cones (Fig. 7B), the spectral patterns of larval, aspartate-isolated cone signals (cone-PIII) from in vitro gnat2:mYFP-2A-trb 2;mpv17À/À eyes are similar to the mpv17À/À controls. In the trace recordings of Figure 7C, an mpv17À/À control larval eye produces substantial signals at all wavelengths, with maximal amplitudes evoked by the 490-nm stimulus. A gnat2:trb 2 larval eye gives a similar amplitude pattern for the test wavelengths and irradiances (Fig. 7D), with maximal amplitudes at 490 nm.
In an adult mpv17À/À control eyecup (Fig. 8A), the spectral pattern of cone-PIII ERG waveforms shows a with a broad range of responsiveness across wavelengths, and the 490-nm stimuli yielding an amplitude peak. An adult gnat2:mYFP-2A-trb 2;mpv17À/À eyecup also shows a broadly responsive spectral pattern (Fig.  8B), but greater amplitudes are evoked at long wavelengths (570 and 650 nm; Fig. 8B, arrows), suggesting that the underlying input signals from red, green, blue, and UV cone types have undergone a late developmental transformation under the influence of gnat2:trb 2.
In gnat2:trb2 larvae the transgene increases R2-cone signals The larval cone PIII spectral responses of gnat2:mYFP-2A-trb 2;mpv17À/À are similar to the mpv17À/À control strain (Fig. 7C,D). Signal loss in the UV and signal gain at long wavelengths, as seen crx:mYFP-2A-trb 2 larvae (Fig.  3C,D), are not as evident. Modeling suggests subtle changes. The optimal model for gnat2:trb 2 is #79, and optimal control model is #77. Applying the control model 77 to the gnat2:trb 2 dataset gives a greater residual variance than model 79 (F (1.21,1506,1511) = 0.999, p = 0.00021), indicating that cone inputs generating the gnat2:trb 2 Figure 8. Adult cone PIII spectral properties in mpv17À/À and gnat2:mYFP-2A-trb 2;mpv17À/À retinas. A, Cone signals in an mpv17À/À control eyecup are evoked by all stimulus wavelengths and all but the dimmest irradiances (black traces). The largest amplitudes occur at 490 nm. B, Cone signals from a gnat2:trb 2 eyecup are stimulated by all wavelengths and irradiances. Greatest amplitudes occur at 650, 570, and 490 nm, with a larger amplitude spectral pattern at 650 and 570 nm than in the control (up arrows). A, B, Adults are 8-18 months. Eyecups are perfused with medium containing 10 mM aspartate to isolate retinal cone signals. Stimulus irradiances [log(quanta·mm À2 ·s À1 )] appear in the legends to the right of irradiance-response waveform stacks. C, Voltage distributions of maximal trough-to-peak amplitudes found in spectral datasets. D, Cone-signal latencies from stimulus onset to the minimum in the ON trough. E, Cone-signal latency from stimulus offset to the OFF peak. C, D, E, The n.s. labels on all distributions indicate that that mpv17À/À controls did not significantly differ from gnat2:trb 2 transgenics in waveform characteristics (t test and p-values given in text). Peak amplitudes (dataset Vmax), trough (PIII ON trough), and peak (PIII OFF peak) latencies are measured on the mean waveforms from each 70-stimulus spectral dataset. cumulative dataset differ from control. In Figure 9A,B, amplitudes are plotted against irradiance for four of the nine test wavelengths and compared with continuous curves calculated from best-fitting models. These are fit to 1945 points (15 eyes, 28 spectral datasets) for mpv17À/À controls and 1540 points (17 eyes, 22 spectral datasets) for gnat2:trb 2. The distribution of points and curves at 570, 490 and 370 nm are more overlapping along the irradiance axis for gnat2:trb 2 larvae than are the same wavelength points and curves for the control strain. Points and curves at 370 and 490 nm (magenta, green) lay to the left of the 570-nm points and curve (yellow) for mpv17À/À controls but coincide with the 570-nm curve for gnat2: trb 2, suggesting less sensitivity from short-wavelengthsensitive and mid-wavelength-sensitive cones. The logs of semi-saturation irradiances for the red (R1, R2) cones were 4.50 6 0.02 for mpv17À/À larvae and 4.51 6 0.02 for gnat2:mYFP-2A-trb 2;mpv17À/À (Fig. 9A,B), nearly identical (t (3422) = 0.251, p = 0.802).
Of the 255 models fit to the 6-dpf larval gnat2:trb 2 combined dataset none were deemed to fit equally as well as model 79 based on residual variance, that is by an F test with p ! 0.95. Two were indistinguishable from the best-fit model 77 for the mpv17À/À controls. Model 103 (F (0.9992,1939,1940) = 0.493, p = 0.986), employed an additional 5 th cone substituting G4 for G1 and adding B1. Model 101 (F (0.9972,1940,1940) = 0.475, p = 0.95) substituted G4 for G1. All three indistinguishable control models agreed on the presence of UV, B2, and R2 cone signals in mpv17À/À controls.
Adult gnat2:trb2 retinas generate large R1 red-cone signals To determine the cone-type composition of adult gnat2: trb 2 retinal cone signals we searched for the best models to represent the gnat2:trb 2 cumulative spectral dataset. Trough-to-peak amplitudes for all eyes and datasets (mpv17À/À, 13 eyes, 23 datasets, 1610 responses; gnat2: trb 2, 11 eyes, 20 datasets, 1400 responses) were fit to the 255 combinations of eight cones. Model 219 fit best for mpv17À/À control eyecups, and model 202, for gnat2: trb 2 (Fig. 10A,B). When the adult gnat2:trb 2 cumulative dataset was fit to the control model 202, the residual variance was significantly greater than with the optimal gnat2: trb 2 model, indicating the cones signals contributing to the gnat2:trb 2 gain-of-function dataset differed significantly from the control (F (1.942,1370,1374) = 1, p = 0).

The larval dual opsin UV cone functions as a UV cone in trb2 gain-of-function transgenics
Many cones in gnat2:mYFP-2A-trb 2;mpv17À/À larvae express mixed opsins, overlaying a red opsin on the native expression of a shorter-wavelength opsin (Fig. 7B). A few dual-opsin cones also occur in crx:mYFP-2A-trb 2 larvae (Fig. 3B). The spectral physiology and molecular development of such cones is of interest. The wide spectral separation of red and UV opsins makes the two opsins in dual-opsin UV cones amenable to separate stimulation. One proposal is that zebrafish transgenic dual-opsin cones are analogous to the stable mws-sws "dual physiology" cone configuration seen in rodents (Jacobs et al., 1991;Applebury et al., 2000;Nikonov et al., 2005;Suzuki et al., 2013). Another proposal is that the mixed-opsin cones in gnat2:trb 2 zebrafish are transitional states, later to develop into red cones. The tendency to lose UV-opsin signals in trb 2 gain-of-function transgenics suggests an adverse action on UV opsins, and a comparable model in mouse development is that trb 2 changes differentiated S cones to M cones (Swaroop et al., 2010).
For the mpv17À/À control, all UV-opsin immunoreactive cones ( Fig. 11A-C, green) express the sws1:nfsBmCherry UV-opsin gene reporter as a magenta inner segment halo surrounding the narrower green immunofluorescence of the UV-cone outer segment (Fig. 11A). UV-opsin and redopsin immunoreactivities are segregated, being expressed in separate cone cells (Fig. 11C), the native larval opsin expression pattern (Allison et al., 2010). In the larval gnat2: trb 2 double transgenic not all UV-opsin immunoreactive cones show the magenta halo of the sws1:nfsBmCherry reporter gene. In some, the sws1 reporter gene generates no fluorescence (Fig. 11E, red and green arrowheads), suggesting the native sws1 gene locus is inactive, and UV opsin, while still present, is no longer being synthesized. Only legacy SWS1 immunoreactivity remains. In Figure  11G, patterns of gnat2:trb 2 immunoreactivity for UV opsin and red opsin are compared. The arrowheads point to cone cells where both UV and red opsin are co-expressed. A comparison of Figure 11F with E reveals that UV-opsin immunoreactive cones with inactive sws1 reporter genes are the same cones that are double immunoreactive for both UV and red opsins. The expression of red opsin in a UV cone by the gain-of-function gnat2:trb 2 transgene appears incompatible with continued expression of UVopsin, suggesting that the overabundance trb 2 in a dual UV-red opsin cone either directly or indirectly blocks the sws1 gene. Therefore, co-expression of UV-opsin and redopsin immunoreactivity cannot continue indefinitely, as in rodents, but would be limited by the catabolism of previously expressed, but not renewed, UV-opsin.
What was unexpected was that the 370-nm irradiance functions for the gnat2:trb 2 gain-of-function larvae, with coexpression of red opsin in some UV cones, were unaffected by the red background (Fig. 11K). The red opsin in the mixed-opsin UV cones, which would be activated by the 627-nm background, does not desensitize 370-nm UV signals. For 370-nm stimuli, Hill fits give semi-saturation irradiances of 4.42 log(quanta·mm -2 ·s À1 ) on the IR background and 4.44 log(quanta·mm -2 ·s À1 ) on the 627-nm background, values not significantly different (t (205) = 0.300, p = 0.765). But the same red background significantly desensitizes redopsin signals from gnat2:trb 2 red cones, as seen with 650nm stimuli (Fig. 11K). The Hill semi-saturation changed from 6.26 log(quanta·mm À2 ·s À1 ) on the IR background to 6.55 log (quanta·mm 2 ·s À1 ) on the 627-nm background, a significant sensitivity loss (t (415) = 5.01, p = 1.5 Â 10 À12 ), which demonstrates the effectiveness of this background for red opsins (Fig. 11K). The failure of long-wavelength backgrounds to affect UV signals from mixed-opsin cones has also been observed in ERG spectra of rodents (Jacobs et al., 1991). In zebrafish the spectral physiology of UV cones appears, at least early in development at 5 dpf, not to be affected by the introduction of transgenic red opsins into many UV-cone members, or by inactivation of the UV-opsin gene.
Thyroxin receptor b2 gain-of-function transgenes alter cone morphology Thyroid hormone receptor b 2 is required in zebrafish for both the development of red cones and the expression of red opsins (Deveau et al., 2020). Morphologically, adult red cones are the principal members of zebrafish double cones (Engstrom, 1960;Raymond et al., 1993). Double cones failed to develop in trb 2À/À mutants (Deveau et al., 2020). The impact of trb 2 in determining the course of cone morphologic development is seen in transverse optical sections of individual transgenic cones from in vivo 6dpf larvae (Fig. 12A,C). These are higher magnifications taken from larval retina confocal image stacks such as seen in Figure 1D-F. In the Figure 12 examples, the fluorescent-reporter shapes of trb 2 gain-of-function cones are distinct from the control morphologies of red, blue, and UV cones, the latter marked by the reporter fluorescent Figure 12. Thyroxin-receptor-b 2 gain-of-function transgenes alter cone morphology. A, Width of cone types identified by transgene markers, is measured at the greatest extent of the inner segment. B, The trb 2 gain-of function cones are significantly wider than red-cones. C, The cone axon length is measured from the base of the inner segment to the apex of the cone pedicle. D, The trb 2 gain-of-function axon lengths are significantly shorter than red-cone axon lengths. B, D, Asterisks indicate significant differences (GraphPad convention, ANOVA and Tukey post hoc p-values given in text). A, C, Images are of 6-dpf in vivo larval fluorescent cones from confocal stacks. Control UV and red cones were imaged in sws1:GFP;trb 2:tdTomato larvae; blue cones, in sws2:GFP larvae; fluorescent trb 2 gain-of-function cones, in crx:mYFP-2A-trb 2 and gnat2:mYFP-2A-trb 2;mpv17À/À larvae. Larvae were anesthetized with MS222 and embedded in agarose after raising to 6 dpf in 300 mM PTU to block melanin formation in the pigment epithelium.
On morphometrics, the mYFP-marked trb 2 gain-offunction cones have not attained red-cone shape at 6 dpf, but nonetheless enhance red cone signal amplitudes. Gain-of-function trb 2 morphologically alters larval cones and appears to shift the metrics toward a larval blue-cone pattern.
Impact of excess thyroid hormone receptor b2 on development of zebrafish spectral signals During zebrafish maturation, patterns of cone opsin mRNA expression change through interactions with a "developmental factor" (Takechi and Kawamura, 2005). Green-cone and red-cone spectral peaks shift from shorter to longer wavelengths by adulthood (Nelson et al., 2019) as different members of gene-duplicated green (Rh2) and red (lws) opsin groups are sequentially expressed. In the present control data, the R2 opsin signal (Vr556) is largest in embryonic (5, 6 dpf) and juvenile (12 dpf) ages but the R1 opsin signal (Vr575) attains equal amplitude status in adults, while, over the same developmental course, the UV cone signal (Vu358) diminishes (Fig. 13A,B, WT and mpv17À/À controls, gray bars). Introduction of gain-of-function trb 2 shifts native developmental patterns of opsin expression. In the adults of both crx:trb 2 and gnat2:trb 2 gain-of-function transgenics, UV-cone (Vu358), blue-cone (Vb415, Vb440), and green-cone (Vg460, Vg480, Vg500) signals are reduced or extinguished by adulthood. In embryos and juveniles, mainly R2 signals (Vr556) increase but in adults, R1 signals (Vr575) increase (Fig. 13A,B).

Discussion
Modeling of the massed cone signals from zebrafish retinas yielded estimates of amplitude contributions from eight spectrally distinct cone types during embryonic, juvenile, and adult developmental stages. The technique provided a window on the impact of transgene-induced overabundance in the red-cone transcription factor trb 2 for the balance among electrical signals from red-cone opsins and all other opsin types. Two trb 2 gain-of-function transgenics were studied. In crx:mYFP-2A-trb 2, the crx promoter introduced trb 2 into retinal progenitor cells, whether ultimately fated to become cone cells or other types, such as bipolar cells (Shen and Raymond, 2004). In gnat2:mYFP-2A-trb 2;mpv17À/À, the gnat2 promoter increased trb 2 levels only in differentiated cone types, including green, blue, and UV cones where it is not native, as well as adding an extra dose to the red-opsin cones where it is normally expressed (Ng et al., 2001;Suzuki et al., 2013). Neither transgene caused major alterations in the amplitudes or kinetics of massed cone signals as isolated from the ERG by blockade of cone synapses. In crx: mYFP-2A-trb 2 larvae, response amplitudes were larger than WT controls, as was the variance in amplitudes. In the adults, peak times of both onset and offset waveform elements were significantly faster. Neither alteration suggested a major net influence on phototransduction. In gnat2:mYFP-2A-trb 2;mpv17À/À, there were no significant changes either in amplitudes or in onset and offset kinetics. But in both transgenics, significant changes were found in the relative contributions from different cone types to retinal spectral responses.
Zebrafish UV (SWS1) cones are the molecular phylogenetic relatives of mammalian S-cones (Terakita, 2005). In crx:trb 2 transgenics the amplitude of UV cone signals was greatly decreased or eliminated at all developmental stages, suggesting the diminished densities of UV opsin immunoreactive cones noted herein and by Suzuki et al. (2013), led directly to decreased UV-cone signal strength. With the progression of developmental stages in teleosts, the role of UV cones typically diminishes (Cheng et al., 2006;Carleton, 2009;Nelson et al., 2019). Juvenile trout lose UV sensitivity as they mature. The process is thyroid hormone sensitive (Browman and Hawryshyn, 1994) and correlates with regional loss of SWS1 UV-opsin immunoreactivity and UV cone morphologies. Based on regional maturation in thyroid hormone and thyroid hormone receptor levels within retinal quadrants, together with experimental treatments with thyroid hormone (T4), Raine and Hawryshyn (2009) proposed that regional loss of UVcone signaling was caused by regional increases in both in T4 and trb . In zebrafish, similarly, T4 reduced UV-opsin transcript levels (Mackin et al., 2019). Present results add that increased levels of trb 2 receptor itself reduce the signal amplitude and numbers of UV cones and that trb 2 is a potential candidate regulating their density. How this might occur in the absence of a transgene is less clear, as trb 2 is not normally expressed by UV cones (Suzuki et al., 2013).
Opsin signals in gnat2:trb2 transgenics For gnat2:trb 2 5-d embryos, introduction of trb 2 into functional cones of all spectral types doubled the numbers of red-opsin immunoreactive cones without changing the densities of cones with other opsin immunoreactivity, a result first noted by Suzuki et al. (2013) and repeated here. But in the present cone ERG analysis, red-cone signal amplitude increased ,10%. In this counterexample, the densities of red-opsin immunoreactive cones and signal strength were not proportional. Densities of green-opsin, blue-opsin, and UV-opsin immunoreactive cones, and the distribution of opsin signals among them, were unchanged. As discovered by Suzuki et al. (2013) and confirmed here, much of the increase in density for red-opsin immunoreactive cones is accounted for by co-expression. Suzuki et al. (2013) found red-green and red-UV immunoreactive cones. To this we add cones immunoreactive for both red and blue opsins. Red opsins induced in differentiated gnat2:trb 2 green, blue, and UV cones evidently do not immediately result in a greater red-cone electrical signal at the 5-d larval stage. By adulthood the spectral signals gnat2:trb 2 do come to resemble those of adult crx:trb 2 red opsin dichromats, with largest amplitudes originating from LWS1 opsins and green blue and UV amplitudes reduced or suppressed. This suggests the introduction of trb 2 into differentiated cones quickly expresses red opsins but is much slower to generate electrical signals than is the case with crx:trb 2, which introduces trb 2 into cone progenitors.
In embryonic gnat2:trb 2 transgenics the amplitude of UV cone signals was not affected by co-expression of red opsins. To examine red-UV immunoreactive cones further we examined sws1 reporter gene expression and the sensitivity of UV cones to adaptation by red backgrounds. In controls, sws1 reporter activity and UV-opsin immunoreactivity co-localized, but in gnat2:trb 2 transgenics UV-opsin immunoreactive cones that co-expressed red opsin lost sws1 reporter activity, suggesting that, even without transcriptional resupply of UV opsin, long-lived previously synthesized UV-opsin survived in the cone disks and functioned for some time after synthesis was suppressed. We suggest that the co-expressed red opsins remain electrically dormant during this initial period, as red backgrounds which desensitize redcone signals failed to desensitize UV-cone signals in 5-dpf gnat2:trb 2. In the longer term UV-opsin would be lost to disk shedding (O'day and Young, 1978). This may be a path by which UV/red mixed opsin cones are gradually lost, and/ or converted to red cones, similar to an important model in mouse M-cone embryogenesis from primordial UV cones (Swaroop et al., 2010).

Suppression of green and blue cones
Overproduction of trb 2 in zebrafish gain-of-function transgenics reduced or eliminated green-cone (Rh-2) and blue-cone (SWS2) signals. Vb415 (B1, SWS2) and Vg460 (G1, Rh2-1) signals were significant in crx:trb 2 embryos and juveniles but lost in adults. G1-cone signals might be either increased or decreased compared with controls in gnat2:trb 2 embryos or juveniles but both were reduced in adults. In adult zebrafish trb 2À/À mutants (Deveau et al., 2020) green-cone signals increased significantly in amplitude. These observations suggest a late-stage inhibitory effect of trb 2 on green and blue cones. Knock-out of the zebrafish homeobox transcription factor six7 eliminates green-cone Rh2 transcripts in adults and six6 knock-outs adversely affected blue-cone SWS2 transcript (Ogawa et al., 2015(Ogawa et al., , 2019. An association of these transcription factors with blue and green cones was made by single-cell sequencing and machine learning methods (Ogawa and Corbo, 2021). Speculatively there is an inhibitory action of trb 2 on these homeobox genes.

Changes in cone morphology
Larval cone morphologies in both trb 2 gain-of-function transgenics were altered. In gnat2:mYFP-2A-trb 2, where the transgene was expressed in all cone types, larval cones expressing the mYFP fluorescent transgene reporter did not look like red cones, resembled none of the control cone-type morphologies, but had a transgenic shape closest in morphometrics to, but visually distinguishable from, the control larval blue cones. There appears to be an early alteration of native morphologies induced by activity of the gain-of-function transgene. In crx:mYFP-2A-trb 2 the mYFP transgene reporter was more sparsely expressed in the cone layer, nonetheless revealing altered cone morphology with inner segment widths and axon lengths similar to gnat2:mYFP-2A-trb 2 cones. The basis of these trb 2-induced shape changes is not known, but despite altered shape, the cones are robustly functional. Clearly trb 2 is importantly involved in large swaths of cone development, including the formation of adult double cones, with characteristic Arrestin 3a antigenicity (Deveau et al., 2020). Thyroid hormone receptor b 2 pathways are not yet fully elaborated, but studies of zebrafish mutants and gain-of function transgenics expand the inventory of physiological, morphologic, and genetic targets and provide insight into further roles in development.