Ola-tmt-opsin1b mutants exhibit light-dependent altered avoidance responses

We started our investigation with medaka tmt-opsin1b, by generating several independent mutant alleles using transcription activator-like effector nuclease (TALEN) technology (Fig 1A). All 3 mutant alleles are predicted to result in an N-terminally truncated protein, prior to the first transmembrane domain (S1A–S1D and S1G Fig) and are thus likely functionally null mutants. We confirmed the absence of functional protein with a previously generated antibody against Ola-TMT-Opsin1b [14] (Fig 1B). We next tested for noticeable developmental alterations by detailed staging of the developing larvae according to morphological criteria [33]. We did not observe any difference between wild-type and homozygous mutant larvae (S2A–S2C Fig). We also crossed a strain expressing GFP under the control of an ath5 enhancer fragment in the developing retinal ganglion cells (RGCs) [34,35] with the tmt-opsin1b^−/− fish. The resulting heterozygous GFP⁺ carriers were incrossed and their larval offspring analyzed for any deviations in the onset of GPF expression (indicative of the timing of RGC differentiation), axonal outgrowth, and connectivity to the tectum (indicative of RGC and tectal development). After hatching (7–9 dpf), the fish were genotyped. As we again did not observe any differences between mutant and wild-type fish (n: wild-type = 19, heterozygous mutant = 27, homozygous mutant = 18; S2D–S2G Fig), we concluded that eye and tectal development is unaffected by the tmt-opsin1b mutation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Mutations in Ola-tmt-opsin1b cause light-dependent differences in an avoidance response test.

(A) Genomic locus of the Ola-tmt-opsin1b (abbreviated in figure as tmt1b) gene and corresponding mutant alterations. Transcription activator-like effector nuclease (TALEN) binding sites: black boxes; exons: gray boxes; inserted/substituted nucleotides: red; deleted nucleotides: dashes. (B) Confocal images of anti-Ola-TMT-Opsin1b (a-Ola-Tmt1b; green) and nuclear DAPI (blue) staining in wild-type and tmt-opsin1b (Δ-2/Δ-2) mutant coronal tectal slices. SPV: stratum periventriculare. Scale bars: 20 μm. (C) Schematic of the behavioral setup. (D) Size discrimination tuning curve of larvae reacting to moving dots presented in ascending size order. (E) Behavioral analysis as in (D), performed in a different laboratory with a different computer screen and surrounding (light intensity = 2.2 × 1,013 photons/m²/s). (F) Avoidance responses to the dots displayed in a pseudo-random order. (G) Avoidance responses in the ascending dot size paradigm with background light reduced to 35% of its initial intensity (35% light intensity = 0.77 × 1,013 photons/m²/s). (H) Total avoidance (area under the curve) under different light intensities. (I) Spectra of the emitted computer screen light used for the 100% (black) and the 35% (orange) light intensity. Data presentation: (D–G) mean (± SEM); (H) single bar: median; box: first and third quartile; whiskers: minimum and maximum; all assays done with larvae 9–12 days post-fertilization. a.u.: arbitrary units. Metadata: S1 Data. For further details on mutants, responses to contrast changes, spectra, and mutant versus wild-type development: S1–S3 Figs.

https://doi.org/10.1371/journal.pbio.3001012.g001

Given that Ola-TMT-Opsin1b is expressed in the medaka optic tectum and reticular formation [14] (Fig 1B and see below), we started our functional analysis with an assay that requires these structures. It has been shown that the responses of frogs and fish to approaching stimuli of different sizes specifically require the tectum and reticular formation [36–40]. We adapted an assay in which the response of individual animals to displayed moving dots of different sizes (mimicking a range of potential prey and predator stimuli) is recorded and analyzed (here referred to as “avoidance assay”; Fig 1C; S1 and S2 Videos). Such avoidance assays have been used for toads [36], larval Xenopus [37], mice [41], goldfish [42], and larval zebrafish [40]. Medaka larvae of different stages were collected in the morning (zeitgeber time [ZT] 2–5) and subjected to the avoidance assay. Newly hatched medaka larvae are immediately free-swimming and might best be compared to zebrafish larvae at about 7–8 dpf (constantly raised at 28°C), while later obvious developmental steps proceed similarly to zebrafish. Based on the responses of the larvae to the moving dots, we calculated an avoidance index (AI) for each dot size: AI = the total number of avoidances minus the total number of approaches, divided by the total number of dots presented. Each dot size was presented 6 times to each fish, and care was taken to use different mutant alleles in the same trial (S1H Fig).

After blind response scoring and subsequent genotyping, we observed that tmt-opsin1b mutant fish exhibited a significantly elevated avoidance response compared to their wild-type siblings, particularly to dot sizes 20 degrees (P = 0.0295, unpaired t test with Welch’s correction; t = 2.3; df = 26.4) and 30 degrees (P = 0.0155, unpaired t test; t = 2.5; df = 40), decreasing gradually towards bigger dots (Fig 1D and 1E; S1 Data). Shuffling dot size order resulted in an overall shift of the response curve, presumably due to habituation of the fish to the larger dots that appeared earlier (compare Fig 1D and 1F; S1 Data). Importantly, the elevated response of tmt-opsin1b mutants versus wild-type siblings was maintained.

Given that different lab environments resulted in slightly different response curves of the mutant fish (Fig 1D and 1E; S1 Data) and that TMT-Opsin1b should function as a light receptor in wild-type fish, we next tested if the observed elevated response in the avoidance assay would be altered by changes of light conditions. The difference in avoidance response between wild-type fish and tmt-opsin1b mutants at 100% light intensity vanishes when the light intensity is reduced (Fig 1G and 1H; P = 0.0318, unpaired t test; t = 2.2; df = 40, without changing the spectrum; Fig 1I; S1 Data). Specifically, when we lowered the light intensity of the white background projected by the computer screen to 35% of the original light intensity, the mutant AI curve and the wild-type AI curve (compare Fig 1E, 1G and 1H; S1 Data) became statistically indistinguishable (P = 0.888, unpaired t test; t = 0.14; df = 22). We also tested whether the observed changes in the behavioral responses observed between the different light intensity conditions might be due to contrast differences. For this, we analyzed the AI of Cab wild-type fish under different contrast conditions (S1I Fig; S7 Data) by testing the fish for their responses to dots of increasing luminance (keeping dot size constant at 20 degrees). Thirty-five percent lower light levels at the computer screen correspond to a Michelson contrast of 0.93, a contrast that does not cause any observable behavioral differences compared to full light levels (S1I Fig; S7 Data), making a change of contrast an unlikely explanation for the observed drop in the AI of tmt-opsin1b mutant fish at 35% ambient light intensity. These results suggest that tmt-opsin1b normally mediates constant responses, even under changing light conditions (for comparisons of the light spectra to natural light levels see S3 Fig).

Ola-tmt-opsin1b mutants exhibit strong age-dependent behavioral differences to wild-type in response to sudden light/dark changes

The avoidance assay requires the tectum for proper responses, but it also requires normally functioning eyes. Like probably all opsins, tmt-opsin1b is expressed in the eye, specifically in the amacrine layer [14]. We therefore next decided to use assays that do not necessarily require eyes. Animals typically respond to changes in ambient illumination with rapid changes in movement [43]. Changes in motor behavior in response to light in blinded and pinealectomized minnows [10], eels [44], and lamprey tails [45] provided the first evidence for extraretinal, extrapineal “photomotor” behavior in vertebrates. Furthermore, in zebrafish larvae, short periods of sudden darkness result in an increased overall activity, which has been interpreted as “light-seeking behavior” and termed “dark photokinesis” [13,46]. Zebrafish larvae lacking the eyes and pineal organ still react to a sudden loss of illumination with an elevated locomotor activity and an undirected light-seeking behavior [13]. After a few minutes of continued darkness, zebrafish will subsequently decrease their amount of swimming, resulting in less distance moved during the remaining darkness time [13,47].

We first evaluated the responses of free-swimming tmt-opsin1b mutant juveniles (20–22 dpf) to sudden light changes. Mutant versus wild-type medaka were subjected to 30 min each of white light and dark intervals (repeated 3 times; spectra: S3E and S4A Figs), while the movement of the larvae was tracked automatically and evaluated using Noldus EthoVision XT software. Both tmt-opsin1b mutant and wild-type fish changed their swimming distance depending on the light/dark condition (Fig 2A and 2B; S2 Data), and there was a significant difference between both during the dark phases (Fig 2A and 2B; S2 Data). As there are more than 40 opsins present in teleosts, we next wondered if this phenotype became stronger under light conditions that are more restricted to the maximal sensitivity of TMT-Opsins. TMT-Opsins are functional photoreceptors with a maximal sensitivity in the blue light range (about 460 nm [14,32,48,49]). Thus, we next modified the assay by using monochromatic blue light (for spectra see S3F and S4A Figs). Under these light conditions tmt-opsin1b mutant fish exhibited a clear increase of activity during the light and dark phases of the assay compared to wild-type (Fig 2C and 2D; S2 Data). Given this clear difference between white and blue light conditions, we next wondered if this response is specific to blue light. We assayed juvenile fish under lower intensity monochromatic blue, green, and red light at as similar photon numbers as possible. Under these conditions we find that juvenile wild-type and tmt-opsin1b^−/− mutant fish assayed under blue light, but not under green or red light, exhibited statistically significant differences (Fig 2E–2G, 2H, 2J and 2L; S2 Data; for spectra see Figs 2I, 2K, 2M and S3G–S3I). These results show that the response is specific and provides additional evidence for the blue light sensitivity of TMT-Opsin1b.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Mutations in Ola-tmt-opsin1b impact on activity levels upon light changes in a partly age-dependent manner.

(A–G, H, J, L, N, O) Average distance moved and corresponding statistical analysis during alternating 30 min of white light or monochromatic light/dark intervals from tmt-opsin1b (abbreviation in figure: tmt1b) homozygous mutant (red) and wild-type (black) juvenile fish (A–G, H, J, L) and larvae (N and O). Each data point (A, C, E–G, N) represents the mean (± SEM) distance moved for the preceding 10 s. Colored boxes on x-axis indicate light condition. (B, D, H, J, L, O) Locomotor activity measured as area under the curve (AUC) during the entire light versus dark intervals of the trial. ****P ≤ 0.0001; **P ≤ 0.01. (I, K, M) Plotted spectra of the different light conditions used in (E–G, H, J, L). The sum of photons is shown next to the respective spectrum. (P, Q) Cortisol levels measured as cortisol (μg/ml) over total protein levels (μg/ml) under baseline and mechanically disturbed (MD) conditions from larvae (P) and juvenile fish (Q). *P ≤ 0.05. Box plots: single bar: median; box: first and third quartile; whiskers: minimum and maximum; n.s.: no significance; a.u.: arbitrary units; dpf: days post-fertilization. Metadata: S2 Data. For further details on spectra and age-dependent differences: S3 and S4 Figs.

https://doi.org/10.1371/journal.pbio.3001012.g002

Many behavioral and brain imaging experiments with genetically altered and/or transgenic zebrafish (as a major functional vertebrate model system in neuroscience) are typically performed at early developmental stages when the fish have just started to swim freely (7 dpf) or even earlier (e.g., [50,51]). As detailed in the Introduction, also all analyses of nonvisual photoreceptor responses in teleosts have so far only been performed at stages not older than 7 dpf, and in most cases significantly younger [13,20,21,22]. However, it is also obvious from multiple studies that significant improvements of sensory systems and behavior occur during subsequent development (e.g., [52,53]).

We thus next tested how our results with juvenile fish compare to those at stages just after hatching. The result was strikingly different. While there was also clearly a difference compared to wild-type, instead of swimming more, mutant young larvae (9–12 dpf) swam less than their wild-type siblings (Fig 2N and 2O; S2 Data). This behavioral difference changed gradually during the days following hatching (S4B–S4I Fig; S8 Data), emphasizing that any molecular and cellular mechanism identified to control behavior has to be seen in connection with the developmental age of the tested organism.

Analyses of the “fast dark photokinesis” response rate (first 2 min after light cessation) revealed no difference in the amount of the response between tmt-opsin1b mutant and wild-type fish of either age, when the overall lowered or heightened mutant baseline level is accounted for (S4J and S4K Fig; S8 Data).

Finally, in order to test if the observed elevated locomotion levels are caused by generally altered stress or anxiety levels, we measured total cortisol levels. We observed no difference between mutant and wild-type siblings at young larval or juvenile stages (Fig 2P and 2Q; S2 Data), while our positive control, larvae exposed to mechanical disturbances, showed a significant increase of cortisol levels (Fig 2P; S2 Data). This strongly suggests that the difference in locomotion apparent between mutants and wild-type is indeed mediated by the acute light changes involving tmt-opsin1b.

Ola-tmt-opsin1b mutants exhibit altered, partly eye-independent, daytime activity levels

Given the relatively consistent changes over the entire period of the experiment, we next tested the free-swimming of tmt-opsin1b mutants across 2 consecutive days (blue light/dark, 16 h/8 h). tmt-opsin1b homozygous mutant larvae and juvenile fish swam significantly more during the 1.5-h period immediately after the lights went on (Fig 3A–3D; S3 Data). This difference was reduced or disappeared when the light conditions remained stable over the rest of the day (Fig 3A and 3C; S3 Data).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. tmt-opsin1b affects the amount of daytime rest, independently of eyes and the circadian core clock.

(A, C, G) Average distance moved during 8 h dark/16 h blue light periods for 2 consecutive days in tmt-opsin1b (tmt1b) mutant (red) and wild-type (black) larvae (A), juvenile fish (C), and enucleated juvenile fish (G). Each data point represents the mean (± SEM) distance moved for the preceding 1 min. Colored boxes along the x-axis represent light condition. (B, D, H) Locomotor activity (measured as area under the curve) during the first 30 min and 90 min after light is turned on during the 2-d trial. (B) *P_{(8–9.5 h)} = 0.0331; *P_{(32–32.5 h)} = 0.0247; *P_{(32–33.5 h)} = 0.0323. (D) *P_{(8–8.5 h)} = 0.0330; *P_{(8–9.5 h)} = 0.0299; ***P_{(32–32.5 h)} = 0.0002. (F) *P_{(32–32.5 h)} = 0.0192; **P_{(32–33.5 h)} = 0.0056. (E, F) Relative mRNA levels of per1b (E) and reverbβ (F) measured from whole larvae entrained to 16 h/8 h white light/dark cycles. Yellow and gray bars: light versus dark conditions during respective sampling. Gray arrows indicate the peaks and troughs of the curve. Box plots: single bar: median; box: first and third quartile; whiskers: minimum and maximum; n.s.: no significance; a.u.: arbitrary units; dpf: days post-fertilization; LD: light/dark; DD: constant dark; ZT: zeitgeber time; CT: circadian time. Metadata: S3 Data. For further details on spectra and individual, more time-zoomed movement plots: S3 and S5 Figs.

https://doi.org/10.1371/journal.pbio.3001012.g003

Light is an important entrainment cue for the circadian oscillator. We thus next analyzed whether the mutation in tmt-opsin1b impacts on the phase or period length of the medaka circadian core clock. We selected 2 representative core circadian clock genes, per1b and reverbβ [54], and tested their transcript oscillations in tmt-opsin1b mutant and wild-type larvae under light/dark and constant dark conditions. The timing of daily minima and maxima was indifferent between mutant and wild-type larvae, as was the overall cycling, clearly indicating that the tmt-opsin1b mutation has no impact on circadian phase or period length (Fig 3E and 3F; S3 Data).

Finally, we tested whether the increase in swimming upon sudden illumination would also occur in fish without eyes, thereby testing for a possible functional contribution of tmt-opsin1b outside the eye (and pineal organ, as tmt-opsin1b is not expressed in the pineal organ [14]). Fish were enucleated and left to recover for 1 week before the trial onset. Importantly, the differences between tmt-opsin1b fish and their wild-type counterparts were still observable: While the trend was the same during both days, it reached statistical significance only during the second day (Fig 3G and 3H; S3 Data). We thus conclude that tmt-opsin1b, at least in part, functions outside the eyes and modulates medaka behavior in response to different environmental light changes independently of the circadian clock.

Ola-tmt-opsin1b and Ola-tmt-opsin2 mutants exhibit additive and nonadditive responses to changes in environmental light

Teleosts such as zebrafish and medaka possess more than 40 opsins. Systematic analyses in zebrafish showed that most exhibit expression in the brain [15]. This raises the question if these opsins might just function redundantly or rather in a complex, non-redundant manner.

We purposely started to approach this question by investigating an additional opsin with similar characteristics as TMT-Opsin1b. Specifically, TMT-Opsin2 is an evolutionarily conserved ETO relative of TMT-Opsin1b, with highly similar spectral sensitivity and absorbance characteristics [14,32] and expression in adjacent, possibly partly overlapping, domains in the mid- and hindbrain [14] (Fig 4A–4E).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Individual and combined differential impact of Ola-tmt-opsin1b and Ola-tmt-opsin2 on daytime rest levels.

(A) Schematized lateral view of an adult medaka brain. Dashed lines: position of sections in (B, C) and (D, E). NDIL: nucleus diffuses of lobus inferioris (of hypothalamus). (B, C, D, E) In situ hybridization (ISH) of tmt-opsin1b (B, D) and tmt-opsin2 (C, E) on coronal sections from Cab wild-type fish. SPV: stratum periventriculare. Unlabeled scale bars: 100 μm. For detailed anatomical annotations see [14]. (F) Genomic locus of the Ola-tmt-opsin2 (abbreviation in figure: tmt2) gene and corresponding mutant. Transcription activator-like effector nuclease (TALEN) binding sites: black boxes; exons: blue boxes; inserted/substituted nucleotides: red; deleted nucleotides: dashes. (G, I) Profile of average distance moved during a 2-consecutive-day trial on a 16 h/8 h blue light/dark photoperiod of (G) 10–13 d post-fertilization (dpf) larvae and (I) 21–23 dpf juvenile tmt-opsin1b (tmt1b) and tmt-opsin2 (tmt2) mutant and wild-type fish. Each data point represents the mean (± SEM) distance moved for the preceding 1 min. Colored boxes along the x-axis represent light condition. (H, J) Locomotor activity (assessed by area under the curve) during the first 30 min or 90 min after light is turned on in the 2-d trial of (H) larvae and (J) juvenile fish. Arrows highlight the data from the double mutants, in comparison to the respective single mutants and wild-type fish. For all panels: *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P < 0.0001. Box plots: single bar: median; box: first and third quartile; whiskers: minimum and maximum; n.s.: no significance; a.u.: arbitrary units. Metadata: S4 Data. For further details on mutants and spectra, and individual, more time-zoomed movement plots: S1, S3, and S5 Figs.

https://doi.org/10.1371/journal.pbio.3001012.g004

We thus wondered, if mutating tmt-opsin2 would reveal redundant or synergistic functions with tmt-opsin1b. Using TALENs, we generated a large deletion in the tmt-opsin2 gene (Fig 4F), removing the first 2 transmembrane helices and resulting in a dysfunctional protein (S1E–S1G Fig). Larval and juvenile tmt-opsin2 mutant and wild-type sibling fish were assessed for their responses during the light/dark phases across 2 d, as described above.

Under these conditions, tmt-opsin2^−/− mutants did not display significant phenotypes, but, unexpectedly, tmt-opsin1b/tmt-opsin2 double homozygous mutants displayed phenotypes different from the tmt-opsin1b^−/− single mutants. In larval stages, adding the tmt-opsin2 mutant to the tmt-opsin1b mutant resulted in a complementation of the tmt-opsin1b phenotype during the beginning of the light phases (Figs 3A, 3B, 4G and 4H; S3 and S4 Data; for individual, high-temporal-resolution graphs see S5A–S5J Fig), but the combination of these mutations led to additive phenotypes in the same experimental test during juvenile stages (Figs 3C, 3D, 4I, and 4J; S3 and S4 Data; for individual, high-temporal-resolution graphs see S5A–S5J Fig; compare position of purple bar graphs—also indicated by arrows—to the corresponding wild-type and single mutant graphs in Fig 4H and 4J; S4 Data).

These data suggest 2 main conclusions. First, the fact that the tmt-opsin2 mutation can compensate for tmt-opsin1b loss shows that these photoreceptors do not function “just” redundantly. It suggests that light information received by the fish is fed into a complex light processing system, and the output for behavior is not simply the sum of the input of all possible light receptors. Second, the functional interaction of the different opsin-based photoreceptors has a clear age-dependent component.

Mutations in tmt-opsin1b and tmt-opsin2 induce transcriptional changes that impact on neuronal information transmission

We next wondered which molecular changes can be detected in the brain of the fish missing specific photoreceptors. It had previously been shown that differences in photoperiod cause changes on the transcript level in the rat brain, resulting in differences of neurotransmitter abundance [7]. We thus reasoned that quantitative RNA sequencing (RNAseq) might be a possible strategy to obtain unbiased insight into the changes that occur due to lack of tmt-opsin1b or tmt-opsin2 function.

We sampled 3 separate brain regions (eyes, forebrain, and mid-/hindbrain; Fig 5A) at ZT 2–3 under a 16 h/8 h white light/dark regime and sequenced stranded cDNA. It should be noted that for technical reasons part of the mid-/hindbrain sample includes a portion of the posterior hypothalamus, i.e., forebrain. The resulting sequences were mapped to the medaka genome (Ensembl version 96); reads mapping to annotated exons were quantified using edgeR (Bioconductor version 3.9) [55]. The resulting tables are included as S10 and S11 Data. When comparing tmt-opsin1b mutants and their wild-type siblings, one differentially regulated transcript caught our particular attention, the preprohormone sst1b (ENSORLG00000027736: also named sst3 or cort) in the mid-/hindbrain (Fig 5B; S5 Data), a member of the somatostatin/corticostatin family. We next independently confirmed the quantitative RNAseq results using quantitative PCR (qPCR) (Fig 5C, black versus red boxes for mid- and hindbrain; S5 Data). In this set of experiments, we separated the midbrain from the hindbrain tissue in order to obtain a more differentiated picture of the regulation. Given the differential effects of tmt-opsin1b/tmt-opsin2 double mutants on behavior, we next wondered how the presence of both mutations would affect sst1b regulation. Adding the tmt-opsin2 mutation compensated the downregulation of sst1b levels present in the tmt-opsin1b single mutant (Fig 5C, purple boxes compared with others for mid- and hindbrain; S5 Data). In an analogous approach, we identified the voltage-gated sodium channel subunit scn12aa as significantly regulated by the tmt-opsin2 mutation (Fig 5D; S5 Data; phylogenetic tree: S8 Fig). Again, while clearly visible in the single mutant (Fig 5E, gray versus blue boxes for mid- and hindbrain; S5 Data), the effect was compensated for in the tmt-opsin1b/tmt-opsin2 double mutants (Fig 5E, purple versus blue boxes for mid- and hindbrain; S5 Data).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Ola-tmt-opsin1b/2 double mutants are compensated for differential transcript changes present in Ola-tmt-opsin1b and Ola-tmt-opsin2 single mutants.

(A) Schematic drawing of the adult medaka brain in dorsal view. Dotted lines indicate the dissection boundaries (TE: telencephalon; P: pineal organ; OT: optic tectum; CC: corpus cerebelli). Note: part of the posterior hypothalamus is dissected with the midbrain, as it is attached to its ventral part and technically unfeasible to reliably separate. (B) Normalized transcript read counts (****adjusted P value < 0.0001) for sst1b. (C) mRNA transcript levels for sst1b determined by quantitative PCR in brain areas and eyes of tmt-opsin1b (tmt1b) mutant, tmt-opsin2 (tmt2) mutant, and corresponding wild-type and double-mutant fish. Decreased sst1b levels are unique to tmt-opsin1b^−/− mutants and are compensated by the tmt-opsin2^−/− mutant. *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; n.s.: no significance. (D) Normalized transcript read counts (****adjusted P value < 0.001) for scn12aa. (E) mRNA transcript levels for scn12aa in brain areas and eyes. Increased levels of scn12aa are unique to tmt-opsin2^−/− mutants, and compensated for when both tmt-opsin1b and tmt-opsin2 are jointly mutated. *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; n.s.: no significance. Box plots: single bar: median; box: first and third quartile; whiskers: minimum and maximum; each displayed data point corresponds to 1 individual biological replicate. Metadata: S5 Data, S10 Data, S11 Data.

https://doi.org/10.1371/journal.pbio.3001012.g005

These molecular data further support the notion that the loss of function of nonvisual opsins does not necessarily lead to a simple summation effect originating from the single mutants. Again, this strongly corroborates the notion of non-redundant, complex light information processing, which can modulate the neuronal function on the level of neuropeptides (tmt-opsin1b regulating sst1b) and voltage-gated channels (tmt-opsin2 regulating scn12aa).

The regulation of sst1b via tmt-opsin1b is non-cell-autonomous

In order to gain deeper insight into the connection between sst1b expression changes in the mid-/hindbrain, tmt-opsin1b, and responses to light, we next analyzed the spatial expression of sst1b. In the midbrain, sst1b is expressed in several highly specific clusters of cells, none of which overlap with tmt-opsin1b expression (compare Fig 6A and Fig 4B; for detailed anatomical annotation of sst1b expression see S6 Fig; for tmt-opsin1b expression see [14]). This is particularly obvious for the tectum, in which sst1b⁺ cells are consistently located more dorsally within the stratum periventriculare (SPV) than tmt-opsin1b⁺ cells, and close to or possibly overlapping with tmt-opsin2⁺ cells (compare Fig 6A with Fig 4B; anatomical annotation: [14], S6 Fig). Similar in the reticular formation of the hindbrain, likely all of the sst1b⁺ cells are separate from the tmt-opsin1b⁺ cells (Figs 6B and 4D; anatomical annotation: [14], S6 Fig). We thus conclude, that the changes of sst1b transcript levels are rather indirectly mediated by tmt-opsin1b⁺ cells upstream of sst1b-expressing cells.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. sst1b levels are reduced in specific cells in the tectum, posterior hypothalamus, and hindbrain in tmt-opsin1b^−/− medaka brains.

In situ hybridization for sst1b performed on coronal sections of midbrain (A) and hindbrain (B) from Cab wild-type fish. Magnification of boxed areas of the optic tectum are displayed below. SPV: stratum periventriculare. Note that sst1b expression identifies a distinct tectal population of dorsal cells and the ventral area of the hindbrain. (C, D) Comparison of sst1b expression in wild-type versus tmt-opsin1b mutant sibling coronal brain sections reveals specific reduction (denoted by arrows) in interneurons of the tectum and reticular formation, and in the posterior hypothalamus. Scale bars: 100 μm. Further details on anatomical annotations and co-localization of sst1b with gad2, as well as Scn12aa orthology relationships and localization: S6–S9 Figs.

https://doi.org/10.1371/journal.pbio.3001012.g006

We next analyzed where the regulation of sst1b transcripts occurs. We performed in situ hybridizations on wild-type and tmt-opsin1b mutant adult fish brains for sst1b, with specific attention to treating wild-type and mutant samples identically. We observed a specific reduction of sst1b⁺ transcript levels in the cells of the stratum periventriculare (SPV) layer in the optic tectum, in the dorsal periventricular hypothalamic zone surrounding the lateral recess, and in cells medial to the nucleus glomerulosus in the medial preglomerular nucleus (arrows, Fig 6C). We also observed a reduction of sst1b⁺ cells in the intermediate reticular formation of the hindbrain (arrows, Fig 6D). In order to independently verify the specificity of the effects, we cloned gad2, whose ortholog is co-expressed with the zebrafish sst1b ortholog in neurons, based on single-cell RNA (scRNA) analyses [56]. We confirmed co-expression of sst1b with gad2 in medaka tectal neurons (S7 Fig), and also did not find any visible alteration in the expression of gad2 in tmt-opsin1b mutant versus wild-type brains (S7 Fig). Taken together, these in situ expression analyses confirm a likely specific effect of tmt-opsin1b on sst1b transcripts. They further suggest an indirect regulation of sst1b transcripts by tmt-opsin1b. We additionally identified that the tectal sst1b⁺ cells are GABA-ergic (by the expression of gad2), and hence likely convey inhibitory input to the tectal circuitry.

We likewise further analyzed scn12aa brain expression. Medaka Scn12aa is the common ortholog of the 2 separate amniote Scn10a and Scn5a groups (S8 Fig; S9 Data). Expression analyses of both mRNA (by in situ hybridization; S9A and S9B Fig) and antibody staining (using a commercially available antibody with broad species reactivity generated against human Scn5a; S9C–S9H Fig) suggest a broad, possibly ubiquitous brain expression. A specific set of neurons regulated by tmt-opsin2 was not obviously detectable. It is hence plausible that the effect of tmt-opsin2 is present in multiple neuronal populations and not restricted to a few specific circuits.

Photoperiod regulates sst1b transcript levels via a tmt-opsin1b-dependent mechanism

We finally aimed to further investigate the possible light dependency of the observed transcript regulation. Given that sst1b exhibits the more specific expression, we focused on this transcript. Specifically, we tested whether changes in the illumination regime impact on sst1b transcript levels in a TMT-Opsin1b-dependent manner. We chose a “photoperiod”-type light regime for 3 reasons: First, any changes in transcription need sufficient time to occur; second, we wanted to test a light regime that has obvious natural relevance to medaka [30]; and, third, pre-pro-somatostatin expression is regulated in response to exposure to short- and long-day photoperiods in adult rats [7].

We exposed wild-type fish to 2 different white light regimes (Fig 7A and 7B). All fish were initially raised under a 16 h (light)/8 h (dark) white light regime (Fig 7B). One cohort remained exposed to this 16 h/8 h light/dark cycle (long day), while the other group was transferred to an 8 h/16 h light/dark cycle (short day). After 1 wk, eyes and brains were dissected at ZT 8 (long day) and ZT 4 (short day) (blue arrowhead, Fig 7B) and analyzed by qPCR.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. tmt-opsin1b sensitizes sst1b regulation to changes in photoperiod.

(A) Schematic representation of the 2 different photoperiods assigned to different cohorts of tmt-opsin1b wild-type and tmt-opsin1b mutant adult fish. Black boxes: dark period; white boxes: white light period. Red arrowheads: feeding time; blue arrowheads: time when fish were sacrificed and the brain dissections made. (B) Light spectrum of the light used in (A). (C) Relative mRNA expression of sst1b in tmt-opsin1b (tmt1b) wild-type and mutant larvae sampled during constant darkness after entraining with 16 h/8 h white light/dark cycle at 27°C. (D, E) sst1b mRNA transcript expression levels in tmt-opsin1b wild-type (D) and tmt-opsin1b homozygous mutant (E) relatives. ****P ≤ 0.0001, **P ≤ 0.01, *P ≤ 0.05. Box plots: single bar: median; box: first and third quartile; whiskers: minimum and maximum. Each data point represents 1 biological replicate. Metadata: S6 Data. For further details on spectra: S3 Fig.

https://doi.org/10.1371/journal.pbio.3001012.g007

As sst1b is circadianly regulated, with no phase difference between tmt-opsin1b mutants and wild-type fish (Fig 7C; S6 Data), we timed our brain part sampling to the middle of the respective light phases, an established sample timing strategy for photoperiod analyses in medaka fish [57].

A comparison of sst1b transcript levels at long- versus short-day light regimes revealed significantly lower transcript level for the short-day cohort, mimicking the tmt-opsin1b mutation (compare Fig 7D versus Fig 7E [S6 Data] and red versus black boxes in Fig 5B and 5C [S5 Data]). Comparing the transcript levels of sst1b under long-day versus short-day photoperiods in the tmt-opsin1b mutants revealed that this mutation leads to even stronger changes in sst1b transcript levels (Fig 7E; S6 Data). This is most apparent for the forebrain, where there are no changes between long- and short-day conditions in wild-type fish (Fig 7D; S6 Data), but a strong downregulation of sst1b under the short-day regime in the mutant (Fig 7E; S6 Data). This shows that the regulation of sst1b by light depends on tmt-opsin1b, but—similar to the behavioral changes in the avoidance assay—suggests that tmt-opsin1b normally prevents specific neurophysiology changes occurring in response to environmental light changes, a “stabilizing function” that is perturbed in tmt-opsin1b mutants.

Medaka fish rearing and ethics statement

All animal research and husbandry were conducted according to Austrian and European guidelines for animal research (fish maintenance and care approved under BMWFW-66.006/0012-WF/II/3b/2014, experiments approved under BMWFW-66.006/0003-WF/V/3b/2016, which is cross-checked by Geschäftsstelle der Kommission für Tierversuchsangelegenheiten gemäß § 36 TVG 2012 p. A. Veterinärmedizinische Universität Wien, A-1210 Wien, Austria, before being issued by the Bundesministerium für Bildung, Wissenschaft und Forschung). Medaka fish (O. latipes) strains were kept in a constant recirculating system at approximately 26–28°C in a 16 h light/8 h dark cycle and bred using standard protocols [33]. Collected embryos were kept at 28°C until hatching in Embryo Rearing Medium (ERM; 0.1% [w/v] NaCl, 0.003% [w/v] KCl, 0.004% [w/v] CaCl₂ × 2 H₂O, 0.016% [w/v] MgCl₂ × 6 H₂O, 0.017 mM HEPES, and 0.0001% [w/v] methylene blue). Mutant lines were generated in and outcrossed at least 4 times to the Cab wild-type background (tmt-opsin1b Δ+23 was outcrossed 13 times). Where available, different mutant alleles were used interchangeably, thereby reducing the probability of effects induced by off-target mutations. Mutant versus sibling wild-type batches and batches from heterozygous crosses were positioned in random, alternating positions relative to the respective light sources while grown prior to the assays. Parental fish grew in as similar fish room light conditions as possible.

Design and construction of tmt-TALENs

TALENs targeting the first exon of medaka tmt-opsin1b (Ensembl gene ENSORLG00000013181) and tmt-opsin2 (Ensembl gene ENSORLG00000012534) were designed using TAL Effector Nucleotide Targeter v2.0 (https://tale-nt.cac.cornell.edu/) with the following parameters: equal left and right repeat variable diresidue (RVD) lengths, spacer length of 15–20 bp, NN for G recognition, only T in the upstream base, and the presence of a unique restriction site in the spacer region. TAL effector modules were assembled and cloned into the array plasmid pFUS using the Golden Gate TALEN and TAL Effector Kit (Addgene, Cambridge, MA, US) according to validated procedure [64]. The RVD sequence of the left tmt-opsin1b-TALEN was HD NN NG NG HD HD HD NI NI HD NN HD NN NI NN HD HD NG, binding to the genomic sequence 5′-CGTTCCCAACGCGAGCCT-3′, and the RVD sequence of the right tmt-opsin1b-TALEN was NN HD NG HD HD NN HD NG NN HD NN NG HD HD HD HD NN NG, binding to the genomic sequence 5′-GCTCCGCTGCGTCCCCGT-3′ on the opposite strand, featuring a unique BssHII restriction site in the 17-bp spacer region. The RVD sequence of the left tmt-opsin2-TALEN was NN NG NG NG NG HD HD NN NN NG HD NI NN NI HD, binding to the genomic sequence 5′-GTTTTCCGGTCAGAC-3′, and the RVD sequence of the right tmt-opsin2-TALEN was NI NG HD NN HD NG HD NG NN NN NG NG NN NI, binding to the 5′-CATCGCTCTGGTTGA-3′ genomic sequence. The final pCS2+ backbone vectors, containing the homodimeric FokI nuclease domains, were used as previously [65]. All final TALEN pCS2+ expression vectors were sequence-verified using 5′-TTGGCGTCGGCAAACAGTGG-3′ forward and 5′-GGCGACGAGGTGGTCGTTGG-3′ reverse primers.

Generation of capped TALEN mRNA and microinjection

Full-length TALEN plasmids were digested with KpnI, gel purified (Gel Extraction Kit, Qiagen, Venlo, Netherlands), and in vitro transcripted (Sp6 mMESSAGE mMACHINE kit, Thermo Fisher Scientific, Waltham MA, US), followed by a purification using the RNeasy Mini Kit (Qiagen). The yield was estimated by NanoDrop (Thermo Fisher Scientific) and diluted to the final concentration for microinjection. Cab strain zygotes were microinjected with a mix containing 5 or 50 ng/μl of each transcribed TALEN mRNA and 0.6% tetramethylrhodamine (TRITC; Thermo Fisher Scientific) in nuclease-free water.

Genotyping of tmt-opsin1b- and tmt-opsin2-TALEN mutated fish

Genomic DNA from hatched larvae or caudal fin biopsies were retrieved with lysis buffer (0.1% SDS; 100 mM Tris/HCl [pH 8.5]; 5 mM EDTA; 200 mM NaCl; 0.01 mg/ml Proteinase K) at 60°C overnight, and 1 μl of a 1:20 dilution of the lysate was used in a PCR reaction (HotStarTaq, Qiagen). The sequence flanking the tmt-opsin1b TALEN binding sites was amplified using the forward 5′-GGGACTTTCTTTGCGCTTTA-3′ and the reverse 5′-CAGGTCAGAGCGGATCTCAT-3′ primers. Ten microliters of the reaction was directly used for restriction digest using 4 units of BssHII enzyme (New England Biolabs, Ipswich, MA, US) in a 20-μl reaction. Genotyping of the tmt-opsin2 locus was made with the forward 5′-CGGTGAGCGATGTGACTG-3′ and the reverse 5′-GGGAGATCTTTGTCCAGGTG-3′ primers. Mutation caused by TALENs was assessed by analyzing the band sizes of the restriction digest product on a 2% agarose gel. Undigested bands were gel extracted and subcloned into pJET2.1 using the Clone JET PCR Cloning Kit (Thermo Fisher Scientific) and sequence-verified.

Immunohistochemistry

Avoidance assay

The behavioral assay was essentially performed as described before [40]. A single medaka larva (9–12 dpf; unfed) or juvenile fish (20–22 dpf; fed) was placed in a custom-built acrylic glass chamber (18 × 14 × 305 mm) with a transparent bottom and opaque side walls and filled with 1× ERM (1 cm high). Five chambers lying parallel to each other were placed on a horizontal-facing computer screen, allowing for the simultaneous recording of 5 free-swimming animals in the same behavioral trial. Stimuli consisted of moving black dots (except when testing for contrast, where different abstract gray values were used) on a white background traveling at a constant speed of 13.5 mm/s in the same direction in every trial. Each behavior trial consisted of 7 blocks of 6 dots each, with each block characterized by a unique dot size, presented in a size-ascending or pseudo-random order. Stimuli were generated on the computer screen using custom-written programs (Ubuntu). Two consecutive blocks were presented with a 19-s interval in between during which no stimulus was displayed. The size of the dot was characterized by its diameter (in pixels) and was calculated as degrees of the larva visual field, as described before [40]. An industrial camera (acA1300-30gm GigE; Basler, Germany), positioned 40 cm above the computer screen, allowed visualization of the fish–dot interactions, recording videos at a frame rate of 33 frames per second for offline analysis. The fish were allowed to accommodate to the chamber for 5 min before the beginning of the trial, which is longer than in previous similar assays in zebrafish [40] and Xenopus [37], likely ensuring a robust standardization across different experiments. The experiments were performed during the natural light part of the fish light/dark cycle, between ZT 2 and ZT 5, and a single animal was used in only 1 behavior trial to avoid possible habituation (e.g., [67]) or learning. Each fish–dot interaction was scored either as an approach, when the fish swam toward a moving dot, or an avoidance, when the fish swam away from a moving dot [40]. A neutral interaction was scored when the dot entering the fish visual field caused no change in the initial swimming pattern of the animal or if the fish remained motionless. Approach behaviors consisted of swimming towards the dot, displaying a distinct attraction and predation behavior [67], whereas avoidance behaviors were faster than approaches, characterized by rapid swimming bouts away from the direction of the dot’s movement. All video analyses were made prior to genotyping of the fish to shield the identity of the subjects from the observer. An AI ([number avoidances − number approaches]/[total number of dots in a block]) was used as a quantitative behavioral readout. All behavioral assays were made at a constant RT of 26°C, with the room lights switched off and the behavioral setup covered by a black, light-impenetrable cover cloth, this way ensuring that the computer screen would be the only illumination source. The Michelson contrast [68] ([L_max − L_min]/[L_max + L_min]) was used to quantify the relative difference in luminance between the moving dot and the computer screen background. L_max and L_min are luminance maximum and minimum, respectively. Medaka are known to have highly acute vision, as demonstrated by previous studies on the optomotor response [53,69], showing excellent visual performance already from hatching [69].

Photokinesis assay

Individual larvae (9 to 12 dpf; unfed; kept at a 16 h/8 h light cycle at 28°C until hatching) or juvenile medaka fish (20 to 22 dpf; fed) were distributed randomly across a 6-well plate containing 10 ml of 1× ERM per well. A behavioral trial consisted of a behavioral paradigm that evaluates the animal swimming activity (i.e., distance moved) during alternating blocks of 30 min of light and darkness (3 h in total per trial) [70], at a constant temperature of 27°C. An initial acclimation phase of 5 min to darkness was used before the start of the trial. When assessing the swimming responses in a more naturalistic assay, a photoperiod of 16 h light/8 h dark was used during 2 consecutive days, at a constant temperature of 27°C. Behavioral assessment was made between ZT 3 and ZT 12. DanioVision (Noldus, The Netherlands) hardware was used to track the distance moved by the animals in each trial. Larval motion was tracked at 60 frames/s over the trial. In an alternate experiment, 1-min pulses of white light and darkness were presented to the larvae, during a total trial duration of 7 min (S2A Fig). Video data were posteriorly analyzed offline by the tracking software EthoVision XT (Noldus) to calculate the average distance moved by the animal every 10 s of the trial. When comparing the dark photokinesis between tmt-opsin1b mutant and wild-type siblings (S2M and S2N Fig), we normalized the different baseline levels by taking the average distance moved during the 5 min of preceding darkness in the trial as a normalization factor. Each animal was used only once per trial, thus avoiding possible habituation biases. An ILT950 spectrometer (International Light Technologies, US) was used to measure the spectra and intensity of the different light sources. A light-impenetrable cloth was placed over the behavior setup, thus assuring a lack of external light contamination.

Diel swimming test

Care was taken to start each trial at the precise light-to-dark transition, thus minimizing any deleterious effect that the shift to a new environment might impose on the natural behavior of the fish. Since scheduled food availability has an impact on the locomotor activity of fish [71], animals were kept unfed during the 2 consecutive days of the experiment, thus ensuring that the change in ambient light was the central feature impacting on swimming responses. The temperature was kept constant at 27°C. The setup, tracking, and off-line analysis were done as previously described for the “photokinesis assays,” with the exception that the activity of the animals was averaged every 1 min.

Enucleations

Enucleations were performed on 13-dpf wild-type and tmt-opsin1b mutant juvenile fish upon anaesthetization with 0.03% tricaine (MS-222) (Sigma, St. Louis, MO, US). After the surgery, fish were allowed to recover overnight in fresh rearing fish water at 26°C. At 1 wk post-surgery, the fish were used for behavior testing. Handling control fish were anaesthetized for the same duration and placed afterwards in fresh fish water overnight.

RNA extraction and sequencing

For RNA-seq experimentation, the eyes, forebrain, midbrain, and hindbrain (anatomical boundaries according to [72]) were dissected from age-matched wild-type and mutant tmt-opsin1b and tmt-opsin2 adult (>2 months old) medaka sibling fish. One metal bead (Peqlab Biotechnologie, Germany) and 350 μl of RLT buffer (Qiagen) including 1% β-mercaptoethanol was added to the frozen tissue parts, followed by homogenization for 3 min at 30 Hz using the Qiagen tissue lyser. Total RNA was extracted using the RNeasy Mini Kit (Qiagen) according to manufacturer’s protocol. Three independent biological replicates, each made of 3 individual fish (total of 9 fish per genotype) were used. For tmt-opsin1b, 1 fish per independent mutation was used. The quality of total RNA was checked using the Agilent RNA 6000 Nano Kit (Agilent, US) and then enriched for poly(A)+ RNA using the Dynabeads mRNA Purification Kit (Thermo Fisher Scientific). We used the SuperScript VILO cDNA Synthesis Kit (Thermo Fisher Scientific) to generate strand-specific cDNA that was further sequenced on the Illumina HiSeq 2500 platform by the Vienna Biocenter Core Facilities NGS Unit (http://www.vbcf.ac.at) as 100-base single-end reads, resulting in 15–61 million reads, on average, per biological replicate.

Differential gene expression

Sequences from each sample were mapped against the assembled chromosomes of the medaka genome (Ensembl version 96) using the read mapper NextGenMap [73]. After filtering for duplicates and low quality base, the stand-alone featureCounts tool [74] was used to count the mapped reads for each transcript in each sample. The Bioconductor package edgeR (version 3.9) was used to analyze read count data and to identify differential expression levels (Benjamini–Hochberg method). We classified genes as being differentially expressed between genotypes when the differences in expression level between wild-type and homozygous mutant fish were significant at a false discovery rate (FDR) of 5%.

qPCR

One metal bead (Peqlab Biotechnologie) and 700 μl of TRI Reagent (Sigma-Aldrich, St. Louis, MO, US) were added to the previously dissected and frozen eyes, forebrain, midbrain, and hindbrain, followed by homogenization for 2 min at 30 Hz using the Qiagen tissue lyser. Total RNA was extracted from the different brain parts and eyes using Direct-zol RNA MiniPrep (Zymo Research, US) according to the manufacturer’s protocol. Then 80–200 ng of total RNA was transcribed into cDNA using the QuantiTect Reverse Transcription Kit (Qiagen) with random hexamer primers. Each cDNA was further analyzed in duplicate in a 25-μl volume using a SYBR-green-containing mastermix in the Applied Biosystems (Thermo Fisher Scientific) mastercycler. Intron-spanning qPCR primers were designed with the Universal ProbeLibrary software from Roche (Table 1). Expression of each gene was normalized to beta-actin transcript levels, and fold changes were calculated.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Forward and reverse primer sequences used for quantitative PCR.

https://doi.org/10.1371/journal.pbio.3001012.t001

Cloning of medaka gad2 and scn12aa

A BLAST search using zebrafish gad2 identified predicted homologous sequences; homology was confirmed by reciprocal back-blast (for gad2).

cDNAs were cloned following standard protocols, using Phusion DNA Polymerase (NEB) and cloned in pJET1.2. Inserts were verified by sequencing.

Primers (5′→3′) were as follows: gad2, forward atggcatctcacgggttctggtctc, reverse ttatagatcatggccgaggcgttca; scn12aa, forward cgccatgcagagagaagattag, reverse gacactcggtggtgaagattac. GenBank accession numbers are MT267535, MT239386, and MT793601.

In situ hybridization (ISH) on adult brain sections

The generation of antisense digoxigenin (DIG)–labeled RNA probes and the ISH staining procedure on adult brain sections were performed as previously [14]. Briefly, adult fish (>2 months old) were anesthetized in fish water containing 0.2% tricaine and decapitated, and the brain was dissected, fixed in ice-cold 4% PFA overnight, and subsequently stored in 100% methanol for at least 1 wk prior to use. After rehydration in 1× PTW (PBS + 0.1% Tween-20), brains were digested with fresh 10 μg/ml Proteinase K (Merck, US) for 35 min, followed by a 4-h prehybridization step at 65°C with Hyb+ solution (50% formamide, 5× SSC [pH 6.0], 0.1% Tween-20, 0.5 mg/ml torula [yeast] RNA, 50 μg/ml heparin). Incubation with the specific DIG-RNA probe was made overnight at 65°C. Coronal whole-brain slices (100-μm thickness) were made using a vibrating blade microtome (Leica VT1000S, Leica Biosystems, Germany), blocked for >1 h in 10% sheep serum/1× PTW, followed by incubation with a 1:2,000 dilution of anti-DIG-alkaline-phosphatase-coupled antibody (Roche, Switzerland) in 10% sheep serum overnight. Detection of DIG probes was made in staining buffer (in 10% polyvinyl alcohol) supplemented with nitro-blue tetrazolium (NBT) and 5-bromo-4-chloro-3′-indolyphosphate (BCIP) (Roche). In addition, the gad2 probe was co-detected fluorescently as described in [75].

Cortisol assay

Phylogenetic analysis

Sequences were aligned using the MUSCLE alignment algorithm (https://www.ebi.ac.uk/Tools/msa/muscle/). The resulting alignment was used to generate a maximum likelihood (ML) tree by the IQ-TREE web server [77], visualized using iTOl (https://itol.embl.de/) and exported as an EPS file. Coloration and text formatting were performed in Adobe Illustrator CC 2017. The accession numbers can be found in the respective metadata tables (S9 Data).

Statistics

The D’Agostino and Pearson normality test was used to assess data distribution. The F-test for comparison of variances for normally distributed data was used to assess the variance of the standard deviation. When comparing 2 normally distributed groups of data with the same standard deviation, the unpaired t test was used. For comparing 2 groups of normally distributed data with different standard deviations, the unpaired t test with Welch correction was used. The Mann–Whitney test was used when comparing 2 non-normally distributed datasets. The software used was Prism version 8.

Data are presented as mean ± SEM. For all statistical tests used, a 2-tailed P value was chosen: *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001.