Light modulates Drosophila lifespan via perceptual systems independent of circadian rhythms

Constant darkness increases Drosophila lifespan independent of key aging-related behaviors

We first asked whether the complete loss of a light stimulus modulates lifespan in Drosophila. We therefore compared the lifespans of flies aged under constant darkness (DD) with those aged under conventional conditions comprised of repeated 12 hr: 12 hr light:dark (LD) cycles. Preliminary experiments revealed that incubator-to-incubator variability in temperature and humidity, even among units from the same manufacturer programmed to the same conditions, were sufficient to induce significant changes in lifespan. To avoid having such differences confound effects that might be caused by different light regimes, we constructed light compartmentalization structures in a single incubator within which light was maintained and between which temperature was measurably indistinguishable (e.g., over a 60 day period the mean temperature in the dark compartment averaged 25.34°C [SD = 0.35°C] while the temperature in the 12 hr: 12 hr LD cycle compartment was 25.26°C [SD = 0.27°C]). In addition, we used lights with a warm spectral profile similar to indoor lighting commonly used in the home (Supplementary Figure 1). When we aged flies under DD and a standard 12 hr: 12 hr LD cycle, as has since been reported [14, 17], we found that flies of both sexes were significantly longer-lived under constant darkness (Figure 1A, 1B); mean and maximum lifespan was increased up to 19% and 14%, respectively. This effect was consistent across experimental replicates and genetic strains, suggesting that it is robust and not a genotype-specific phenomenon (Figure 1C, 1D).

We next investigated behavioral changes that might indirectly slow aging in constant darkness. First, we measured food consumption to evaluate whether flies were behaviorally limiting their nutrient intake in the dark, thereby executing self-dietary restriction [38, 39]. We observed that total food intake, as measured over a 24 hr period by a modified version of the ConEx feeding assay [40], was not significantly different between female flies previously maintained for 14 days in DD vs. siblings maintained in a standard 12hr:12hr LD cycle (Figure 2A). Interestingly, males subjected to DD consumed modestly but significantly more food than their siblings that were exposed to light (Figure 2B). Second, we measured total activity, which does not directly affect lifespan but may impact caloric balance and long-term health [41]. Flies were first maintained for 14 days in either DD or LD conditions, after which time they were transferred to activity tubes, placed in Trikinetics Drosophila Activity Monitors, and measured for five days in their experimental light conditions. We found that flies aged in DD maintained similar overall levels of activity as did flies aged in a 12 hr: 12 hr LD environment (Figure 2C). Third, we examined fecundity as a measure of potential reproductive costs of extended lifespan [42]. Flies aged in DD for 14 days showed no differences in fecundity over a subsequent 7-day period compared to their sibling control flies aged in LD conditions (Figure 2D). Fourth, we asked whether DD affected the decline in circadian rhythms that normally occurs when flies are aged in standard LD conditions [43]. We observed that flies aged in DD for 21 days exhibited measures of rhythm strength and circadian periodicity that were statistically indistinguishable from their siblings maintained in LD conditions (Figure 2E, 2F). We concluded that the extended lifespan observed in flies maintained in DD is not due to diet-restriction, changes in locomotion, reduced reproduction, or improved circadian function.

Slowed aging in constant darkness is modulated, at least in part, through the perception of light

We next asked whether the perception of light is necessary and/or sufficient to modulate fly lifespan. To determine necessity, we took advantage of visually blind flies that lack eyes and photoreceptor cells. These flies express the proapoptotic gene hid under the control of the Glass Multimer Reporter (GMR) promoter element, which expresses in the photoreceptor cells and downstream neurons [44]. Flies carrying two copies of the GMR-hid transgene are completely eyeless [45, 46]. We found that DD did not increase the lifespans of male or female GMR-hid flies compared to siblings aged in standard 12 hr: 12 hr LD conditions (Figure 3A, 3B). To test sufficiency, we sought to mimic light perception while avoiding the potential damaging physical effects of light. We therefore decided to manipulate the activity of light-perceiving neurons and to measure this effect on lifespan in the absence of external light. We again targeted GMR-expressing neurons, as well as neurons that specifically express the blue light photoreceptor, Rh1, because of the documented effects of this wavelength on lifespan [14, 17]. Spatiotemporal activation was accomplished by employing the GAL-4/UAS system to express the temperature sensitive cation TrpA1 selectively in GMR and Rh1 neurons, respectively. The Drosophila TRPA1 channel promotes neuron depolarization only at elevated temperatures (>25°C) and allows for temporal control over cell activation [47]. All experimental flies were aged in DD, and to mimic conventional 12 hr: 12 hr LD conditions, we cycled temperature from 18°C to 29°C on a 12 hr: 12 hr period to activate targeted neurons. Oscillatory activation of all visual neurons with the GMR driver and UAS-TrpA1 proved to have sexual dimorphic effects; male flies were unaffected (Figure 3C) but female flies exhibited a shortened lifespan (Figure 3D). More restricted activation of Rh1-expressing neurons reduced lifespan in both males and females (Figure 3E, 3F). These results suggest that light modulates lifespan, at least in part, by visual light perception.

It has been reported that exposure to high amounts of visual light, specifically in the blue range, can directly induce cellular damage and reduce lifespan in the nematode, Caenorhabditis elegans [14], and in Drosophila [17]. To evaluate whether broad-scale light-induced damage is involved in the lifespan differences that we observed in our standard rearing conditions, we studied the effects of variable exposure time, intensity, light:dark transitions, and wavelength. First, we reasoned that if light energy itself was directly damaging, perhaps by inducing senescence or cell death in visual neurons, then exposure time would be negatively correlated with lifespan. We therefore compared the lifespans of flies aged under constant light (LL) to those aged in 12 hr: 12 hr LD conditions and in DD. While DD reliably extended male lifespan, we found that flies aged in LL were not shorter-lived than those aged in LD conditions (Figure 4A). The same result was observed with female flies (Supplementary Figure 2A). Moreover, we found no significant difference in lifespan between male flies exposed to a 12 hr: 12 hr LD schedule with dim light (300 lux) and those similarly exposed to 5× brighter light (1050 lux; Figure 4B), although in females the dim light treatment had a reduced effect on lifespan compared to bright light (Supplementary Figure 2B).

It is possible that lifespan is subject to threshold modulation in which a small amount of light triggers a maximum effect on lifespan or that transitions from light to dark (and vice versa) are important. To test these ideas, we aged flies in conditions where darkness was interrupted by two light pulses each day from 8 am–9 am and 7 pm–8 pm. We chose this design so that the light pulses would coincide with the first and last hours of light under our standard 12 hr: 12 hr LD cycle while also doubling the number of transitions that the flies experienced. Male flies aged in these conditions lived significantly longer than flies exposed to 12 hr: 12 hr LD cycles but shorter than flies aged in DD (Figure 4C). Female flies did not show the same trend (Supplementary Figure 2C). These data suggest that the number of LD transitions is not causal for changes in lifespan and that a straightforward damage model is unlikely to account for our observations. Furthermore, the possibilities remain that light shortens lifespan as a result of perception and/or that the threshold at which light induces damage is above the levels used in these experiments.

Cell autonomous, light-induced damage has been shown to be wavelength dependent. We therefore asked whether different wavelengths of high intensity light were capable of modulating lifespan in our conditions and whether such effects were dependent on perception. Flies exposed to high intensity monochromatic blue (470 nm), green (527 nm), and red (640 nm) light were all significantly shorter lived than flies kept in DD. Shorter wavelengths had larger effect (Figure 4D, Supplementary Figure 2D), which is consistent with previously published studies [14, 17]. Notably, however, eyeless GMR-hid flies exhibited a similar response to intense blue light as did control animals suggesting that this treatment influences lifespan independent of visual perception, likely through mechanisms that are distinct from the effects caused by normal levels of visible light (Supplementary Figure 3A–3D). Further, when flies were exposed to the normal levels of broad-spectrum light that were used in LD experiments, or when we activated blue light Rh1-neurons for 12 hrs a day, we found there to be no significant changes in stress response gene transcripts (Supplementary Figure 4A, 4B).

The effects of light perception on lifespan are independent of the molecular circadian clock and of daily time perception

Given that sensory perception of light is responsible, at least in part, for extended lifespan in constant darkness, we next asked whether this effect was modulated by mechanisms involved in specifying endogenous circadian rhythms, which are entrained by light patterns. The genes period (per) and timeless (tim) are essential components of the repressive limb of the molecular clock, and their loss leads to molecular arrhythmicity. Although these mutants are capable of masking, which is showing behavioral rhythms that correspond with light cycles without light anticipatory behavior, and exhibit similar activity patterns as wild-type flies, they will not entrain to the light cycle and are unable to predict the onset of light. We observed that male flies homozygous for a complete loss of function in the per allele (per⁰¹) exhibited a significant increase in lifespan under DD, as did male animals carrying a deletion in tim (tim⁰¹) (Figure 5A, 5B). To more thoroughly explore whether clock function mediates lifespan extension in the dark, we also tested the potential involvement of the positive limb of the clock by using flies that are mutant for the gene cycle (cyc⁰¹), which also results in behaviorally arrhythmic flies. Cyc⁰¹ did not abolish the lifespan increase caused by DD in males (Figure 5C). Finally, we tested whether the circadian-light sensor, cryptochrome (cry), was required for lifespan extension in constant darkness. Male flies homozygous for the null mutation, cry^b, displayed a significant lifespan extension of similar magnitude to control flies when kept in DD (Figure 5D). While the degree of lifespan changes caused by DD are variable depending on the sex and circadian mutant used, males and females generally show similar trends (Supplementary Figure 5A–5E). These results indicate that lifespan extension in constant darkness is independent of molecular circadian rhythms.

The effects of sensory perception on lifespan are often more pronounced when information provided by sensory systems is uncoupled with the experiences that they were designed to predict [48]. For example, flies that smell food during periods of food scarcity or that detect the opposite sex in the absence of mating opportunities are significantly short-lived [7, 49]. In the context of light and time perception, this situation might be represented by a discordance between the predicted pattern of light cycling provided by the molecular clock and the realization of actual environmental light patterns. Indeed, it is currently thought that such asynchrony reduces lifespan, reproduction, and metabolic health [50–53].

We therefore investigated how different forms of uncoupling between light schedules and the circadian clock impact Drosophila lifespan. We began by exploring the effects of repeated exposure to a shifting light cycle (Figure 6A). We chose a light schedule that mimicked human shift workers who travel four days a week or who work nights several days a week and then experience a different schedule on weekends. This was executed by exposing flies to a standard 12 hr: 12 hr light-dark schedule, with lights on from 9 am–9 pm, on Monday-Thursday, imposing a 6-hr phase delay on Friday (i.e., with lights on from 3 pm-3 am), and restoring the normal cycle by applying a 6 hr phase advance on Sunday. A similar schedule had been shown to be detrimental to mouse lifespan [54]. Unexpectedly, this shifting light paradigm had no meaningful effect on fly lifespan, with two different laboratory strains exhibiting a mean reduction in lifespan of ≈3.7% (Figure 6B, 6C).

We next tested different patterns of light oscillation, including oscillation rates that were equivalent to the flies' free running period as well as those that exhibited different degrees of discordance. To do this we expressed mutant variants of the doubletime kinase, which is responsible for the phosphorylation of PER and thus the amount of time it takes for the molecular clock to cycle [17, 55]. In this way, we created flies with endogenous periods of 18, 24, and 27 hours, which we refer to as short-day, normal-day, or long-day flies, respectively. Short-day flies expressed UAS-doubletime-short (UAS-DBT^S) in clock neurons (using Clk856-GAL4), which are the master circadian neurons whose output serves to synchronize all body clocks. Long-day flies expressed UAS-doubletime-long (UAS-DBT^L) in those same cells [56], and flies carrying Clk856-GAL4 but no UAS element served as the control. Flies from each free-running period were exposed to each of three different environmental light:dark conditions of 9 hr: 9 hr, 12 hr: 12 hr, and 13.5 hr: 13.5 hr hours in a factorial design (Figure 7A). This design was chosen to allow direct, within-strain comparisons among treatments in which one environmental light cycle was in line with its endogenous free running period (termed the control environmental condition for that genotype), and two environmental light cycles that were distinct from it. Our design also allowed us to determine whether the magnitude of the difference between environmental and endogenous periods correlate with lifespan effects (Figure 7A).

Before executing the lifespan experiments, we sought to establish that our genetic manipulations were effective in maintaining distinct free-running periods throughout lifespan and that our disparate environmental light cycles were effective in masking them. We found that short-day flies exhibited a mean period length when young of 17.8 hours (SD = 0.34) and that long-day flies showed a mean period length of 26.8 hours (SD = 0.61). As expected, normal-day Clk856-GAL4 animals, which did not express either altered version of DBT, exhibited a mean period of 23.9 hr (SD = 0.19; Figure 7B). We also found that each genotype effectively entrained to each of the different light cycles and had an activity period that was within 0.1 hr of the diurnal cycle (Figure 7B), even the most disparate. Short-day flies, for example, exhibited 27-hour behavioral rhythms (mean = 26.93 hr, SD = 0.03) when exposed to the 13.5:13.5 hr light:dark regime, and long-day flies expressed an 18 hour behavioral rhythm rhythms (mean = 17.98 hr, SD = 0.02) when exposed to the 9:9 hr regime. When transferred to constant darkness after aging for three weeks in each light environment, flies reverted to their expected genotype-specific free-running periods (Figure 7C), establishing that endogenous rhythms were retained until older ages independent of light condition and that extended durations at different periods did not differentially affect rhythmicity.

We then measured lifespan of each of the three genotypes in each of the three light conditions. We observed that short-day flies exhibited statistically indistinguishable lifespans in conditions of 9 hr: 9 hr, 12 hr: 12 hr, and 13.5 hr: 13.5 hr light:dark regimes (Figure 7D). Similar results were obtained for both long- and normal-day flies: neither genotype exhibited differences in lifespans when aged across the three light regimes (Figure 7E, 7F). In other words, neither the magnitude nor direction of misalignment between oscillations of environmental light and of the endogenous clock affected lifespan in our experiments.

To examine the hypothesis that perception of time, per se, modulates lifespan, we aged the short-, normal-, and long-day flies in constant darkness, which, for a given amount of chronological time, would result in each genotype experiencing a different number of subjective days (Figure 8A). We reasoned that short-day flies, with their 18 hr period, might therefore perceive a more rapid passage of time than would normal-day or long-day flies with their 24 hr and 27 hr periods, respectively, and that a comparison among them would not be confounded by entrainment. We observed that short-, normal-, and long-day flies exhibited similar lifespans in constant darkness (Figure 8B).

When taken altogether, our results from manipulations that were designed to mimic shift work, to study the effects of concordance between endogenous and environmental rhythms, and to examine the effects of perceived time all indicate that the extension of lifespan observed in constant darkness is independent of the molecular clock, that the relationship between circadian timekeeping and external light has little effect on patterns of fly aging, and that the length of life is independent of the number of subjective days.

Fly husbandry

All D. melanogaster used in this paper were reared using the same method. Experimental flies were age-synchronized using a 3-step procedure. First, mated females and males were placed on a grape juice agarose plate supplemented with live yeast paste for 18–22 hours. The eggs laid during this period were collected briefly in PBS, distributed in 32 ul aliquots into culture bottles containing a modified Caltech Medium (CT food) [65], and reared at 25°C with a standard 12 hr: 12 hr LD cycle. Second, flies that eclosed within a 24 hr window were collected into bottles and maintained on a 10% sucrose/yeast food (SY10) for 2–3 days at 25°C with a 12 hr: 12 hr LD cycle, unless otherwise noted. After the 2–3 day mating period, flies were sorted under light CO₂ anesthesia into single-sex groups of 20 or 25, unless noted otherwise, and placed into vials containing SY10 media, which was changed every 2–3 days for the duration of their lifespans.

Fly strains

Canton-S [64349], yw [1495], and w¹¹¹⁸ [3605] fly strains were obtained from Bloomington Drosophila Stock Center. We thank the Todd Lab at the University of Michigan for providing us with GMR:Hid, GMR-Gal4, and Rh1-Gal4 fly stocks. The Giebultowicz lab at Oregon State University generously shared per⁰¹, tim⁰¹, cyc⁰¹, and cry^B mutants with us. All Drosophila husbandry and experimentation was performed according to the standards accepted by the field. The Shafer lab was kind to give us Clk856-Gal4, UAS-DBT^L, and UAS-DBT^S lines. We also thank the Garrity lab for providing us with the UAS-TrpA1 line.

Media recipes

The modified CT food recipe is as follows: 1 L water, 10 g agar, 6 g cornmeal, 30 g sucrose, 55 g dextrose, 45 g yeast, 15 mL tegosept solution (20% tegosept in 90% ethanol), 3 mL propionic acid. Sugar-yeast 10% (SY10) food recipe is: 1 L water, 20 g agar, 100 g sucrose, 100 g yeast, 15 mL tegosept solution (20% tegosept in 90% ethanol), 3 mL propionic acid, and 4 mL antibiotic supplement (1% tetracycline and 2.5% kanamycin) in water.

Environmental control

Carefully controlled environmental parameters were key for the success of these experiments. As such, all experiments involving comparisons across environmental light cycles were carried out within the same Percival incubator. Individual light cycles were maintained within 3, light-tight, cabinet drawers installed in the incubator. The lights used for the monochromatic light experiments were LEDs sourced from Luxeon Star which had wavelengths of blue (470 nm), green (527 nm), and red (640 nm). They were run on an automatic 12 v timers to create a LD cycle. The lights used for all white light LD and DD experiments were DIODER LED strip lights (item model #: 201.194.18) with an advertised color temperature of 2700K controlled by a digital timer. To compensate for heat produced by the lights, the dark cabinet drawer also contained the same lights on the same schedule, but they were contained within a light-tight aluminum box. Individual cabinet drawer temperature, humidity, and barometric pressure were determined to be statistically indistinguishable in all 3 boxes. Unless noted, temperature was maintained at 25°C +/−0.5°C, and humidity was maintained at 70% with fluctuations of up to 20%. Light spectral profile was determined with a Sekonic C-800-U spectrometer.

Lifespan assays

Adult male and female flies (aged 2–3 days post-eclosion) were collected from controlled larval cultures and mating conditions (see above), sorted, and transferred into standard vials (20 or 25 flies per vial) containing SY10 media. The number of flies used per treatment group at the start of the lifespan is given in the figure legends. Actual number used in the analysis is noted in the figure legends. Cohort censuses were taken every 2–3 days, at which time flies were transferred to fresh SY10 media. Experiments were coordinated using DLife computer software [66]. For experiments in which flies were aged under dark conditions, transfers occurred under indirect, dim red light (5 lux).

Temperature-dependent neuronal manipulations

Temperature was used to activate GMR or RH1 expressing neurons in adult flies. Parental crosses for activation strains were GMR-Gal4 x UAS-TrpA1 and Rh1-Gal4 x UAS-TrpA1, while the control crosses were GMR-Gal4 x w¹¹¹⁸ and Rh1-Gal4 x w¹¹¹⁸. The UAS-TrpA1 line used was backcrossed to the w¹¹¹⁸ line used for 10 generations to minimize potential genetic background differences. For all crosses, eggs were collected and raised in 18°C 12 hr: 12 hr LD. This temperature was maintained until the beginning of the experiment to ensure there would be no neuronal activation during development. At the beginning of the lifespan experiments flies were transferred to a Percival incubator in which they were maintained in constant darkness under temperature oscillations of 12 hr: 12 hr, 18°C: 29°C. TRPA1 activation was designed to mimic daytime light perception, and therefore flies were transferred to new media during the 29°C period.

Activity measurements

Adult male flies (aged 2–3 days post-eclosion) were collected from controlled larval cultures and mating conditions (see above), sorted, and transferred into standard vials (20 flies per vial) containing SY10 media. They were subsequently maintained in either 12 hr: 12 hr light-dark (control) conditions or in constant darkness for 14 days. Flies were then sorted individually into 5 mm × 65 mm polycarbonate tubes (TriKinetics part # PPT5x65), with the same sugar-yeast media placed at one end of the tube. Activity tubes were then loaded into Drosophila Activity Monitors (TriKinetics part # DAM2), and monitors were then transferred back into their respective light condition. Recording began 24 hours after the flies were loaded into the activity tubes to allow for acclimation to experimental housing conditions. Data was collected using TriKinetics DAMfilescan, and they were subsequently summed into 30-minute bins. Flies that died during the recording period were excluded from the analysis.

Circadian period and rhythm calculation

Rhythm and period experiments were conducted using activity measurements as an output of circadian rhythms. Briefly, adult male flies (aged 2–3 days post-eclosion) were collected from controlled larval cultures and mating conditions (see above), sorted, and transferred into standard vials (20 flies per vial) containing 10% SY10 media. Flies were aged for 21 days under 12 hr: 12 hr light-dark (control) conditions or in constant darkness and then were sorted into 5 mm × 65 mm polycarbonate tubes (TriKinetics part # PPT5x65) with SY10 food at one end of the testing tube before being assigned to the Drosophila Activity Monitors (TriKinetics part # DAM2). All monitors were placed in the same incubator where they received two days of 12 hr: 12 hr LD then 7 days of DD. TriKinetics DAMfilescan was used for data processing. Actimetrics ClockLab Analysis 3 was used for data analysis. A chi-square periodogram analysis was used to determine period, and fast Fourier transformation analysis was used to calculate rhythmicity values. Flies that died during the recording period were excluded from the analysis.

Fecundity assay

Adult male and female flies (aged 2–3 days post-eclosion) were collected from controlled larval cultures and mating conditions (see above), sorted, and transferred into standard vials (5 flies of each sex/vial) containing SY10 media. They were subsequently maintained in either 12 hr: 12 hr light-dark (control) conditions or in constant darkness for 14 days, during which time they were transferred to new SY10 media every 2–3 days. After this 14-day acclimation period, fecundity was measured by maintaining each group of flies in their corresponding light/dark conditions and transferring flies to new media daily, after which the number of eggs laid in each vial was recorded.

Feeding assay

To determine the effects of visible light on feeding we performed a modified version of the ConEx blue feeding assay where only excreted blue is measured [40]. Adult male and female flies (aged 2–3 days post-eclosion) were collected from controlled larval cultures and mating conditions (see above), sorted, and transferred into standard vials (15 flies of each sex/vial) containing SY10 media. Flies were aged 21 days in their respective light environments. Afterward, flies were transferred to new empty vials, which were topped with plastic caps containing SY10 media with 1% FD&C Blue #1. Flies were allowed to feed on dyed food for 24 hours, after which they were frozen for subsequent analysis (see also, ref [40]). Briefly, flies were removed from each vial and the food cap was discarded. Milli-Q water (3 ml) was added to each vial, after which vials were covered (using parafilm) and vortexed. From this 200 uL of solution was transferred into a flat bottom 96 well plate, and absorbance at 630 nm was determined for each well using a BioTek Synergy 2 microplate reader and Gen5 software. Absorbance values were converted to micrograms of food consumed per fly by interpolating a standard curve.

Quantitative PCR

To determine if the effects of visible or perceived light induce oxidative or stress response genes we assayed a small panel of stress resistance genes in heads of wild type Canton-S flies that had been exposed to LD and DD conditions, and Rh1-Gal4 x UAS-TrpA1 and control Rh1-Gal4 x w¹¹¹⁸ flies that were aged under constant darkness under temperature oscillations of 12 hr: 12 hr, 18°C: 29°C. Both experiments were allowed to age for 6 weeks at which point flies were quickly frozen in liquid nitrogen and heads and bodies separated. RNA was extracted using Trizol (Invitrogen) from 3 samples per treatment, with each sample containing 45–50 fly heads. cDNA was synthesized using Superscript III first strand synthesis kit (Invitrogen). Quantitative real-time PCR was performed with Power SYBR green (ThermoFisher). All expression was normalized to the housekeeping gene rp49, and all reactions were performed in triplicate for technical replication.

The following primers were used:

Rp49 F: ATCGGTTACGGATCGAAAA

Rp49 R: GACAATCTCCTTGCGCTTCT

TotA F: GCTTCAGCGTTCCAAAAAGT

TotA R: CTCACGATCTTCGTCGGAAT

Dipt F: ATTGGACTGAATGGAGGATATGG

Dipt R: CGGAAATCTGTAGGTGTAGGT

FMO2 F: CGCAACCAGAAGAAAGCACA

FMO2 R: TGCTCCTGTACGTGTCCAAT

Statistics

Group and pairwise comparisons on survivorship data were performed using the DLife computer software, and the statistical software R [66]. P-values for survivorship data were obtained using log-rank tests. Pairwise comparisons for transcript expression, feeding, fecundity, circadian period, and rhythmicity, and activity data were evaluated using a two-sided independent-samples t-test. Group comparisons of circadian period data were evaluated with a one-way ANOVA. All non-lifespan statistical calculations were run in the statistical software OriginPro. For all box plots, box represents Standard Error of the Mean (SEM, centered on the mean), whiskers represent 10%/90%, and the horizontal line represents the median.

Data availability

All data are available upon request from the corresponding author.