author/funder. it is made available under a cc-by-nc-nd 4 ... · 34 predictable changes in their...

1

Repeated evolution of circadian clock dysregulation in cavefish populations 1 2 Katya L. Mack1, James B. Jaggard2, Jenna L. Persons3, Courtney N. Passow4, Bethany A. 3 Stanhope2, Estephany Ferrufino2, Dai Tsuchiya3, Sarah E. Smith3, Brian D. Slaughter3, Johanna 4 Kowalko5, Nicolas Rohner3,6, Alex C. Keene2, Suzanne E. McGaugh4 5 6 1Biology, Stanford University, Stanford, CA, USA 7 2Department of Biological Sciences, Florida Atlantic University, Jupiter, FL, USA 8 3Stowers Institute for Medical Research, Kansas City, MO, USA 9 4Ecology, Evolution, and Behavior, University of Minnesota, Saint Paul, MN, USA 10 5Wilkes Honors College, Florida Atlantic University, Jupiter FL, USA 11 6Department of Molecular and Integrative Physiology, The University of Kansas Medical 12 Center, Kansas City, KS, USA 13

14 15

.CC-BY-NC-ND 4.0 International licenseauthor/funder. It is made available under aThe copyright holder for this preprint (which was not peer-reviewed) is the. https://doi.org/10.1101/2020.01.14.906628doi: bioRxiv preprint

https://doi.org/10.1101/2020.01.14.906628

http://creativecommons.org/licenses/by-nc-nd/4.0/

2

Abstract 16

Circadian rhythms are nearly ubiquitous throughout nature, suggesting they are critical for 17

survival in diverse environments. Organisms inhabiting environments with arrhythmic days, such 18

as caves, offer a unique opportunity to study the evolution of circadian rhythms in response to 19

changing ecological pressures. Here we demonstrate that the cave environment has led to the 20

repeated disruption of the biological clock across multiple populations of Mexican cavefish, with 21

the circadian transcriptome showing widespread reductions in rhythmicity and changes to the 22

timing of the activation/repression of genes in the core pacemaker. Then, we investigate the 23

function of two genes with decreased rhythmic expression in cavefish. Mutants of these genes 24

phenocopy reductions in sleep seen in multiple cave populations, suggesting a link between 25

circadian dysregulation and sleep reduction. Altogether, our results reveal that evolution in an 26

arrhythmic environment has resulted in dysregulation to the biological clock across multiple 27

populations by diverse molecular mechanisms. 28

29

30


https://doi.org/10.1101/2020.01.14.906628


3

Introduction 31

Circadian rhythms that maintain 24-hour oscillations in physiology and behavior are nearly 32

ubiquitous in nature1,2. Considered an adaptive mechanism for organisms to anticipate 33

predictable changes in their environment3,4, the biological clock coordinates diverse biological 34

processes, from the sleep-wake cycle and metabolism in animals to growth and photosynthesis in 35

plants5–7. In the majority of organisms, the biological clock is endogenous, but environmental 36

time-cues (“zeitgebers”) including light and temperature play a key role in synchronizing the 37

clock with an organism’s external environment8–11. Subjecting animals to light-dark cycles that 38

differ from that of a 24-hour day has profound impacts on organismal health, including reduced 39

performance, increased illness, and decreased longevity12–14. Consequently, systems where these 40

environmental zeitgebers have been lost provide a unique opportunity to examine how ecology 41

drives evolution in the biological clock, and characterize how clock evolution affects 42

downstream physiology and behavior15. Here, we examine daily expression changes across the 43

transcriptomes of cavefish populations that evolved under darkness and relatively low daily 44

temperature variation16,17. 45

46

The Mexican tetra, Astyanax mexicanus, exists as surface populations that live in rivers with 47

robust light and temperature rhythms, and at least 30 cave populations that live in perpetual 48

darkness with limited fluctuations in temperature or other environmental cues18. The existence of 49

multiple cave populations of independent origin and conspecific surface populations provides a 50

powerful comparative framework to study phenotypic evolution. Cave populations of A. 51

mexicanus have repeatedly evolved a suite of traits related to a cave-dwelling lifestyle, including 52

degenerate eyes19–21, reduced pigmentation22–25, and changes in metabolism and behavior26–35. 53

Circadian rhythms and sleep behavior are also substantially altered in cavefish27,36–38. While 54

cavefish largely maintain locomotor and physiological rhythms in light-dark conditions, many 55

populations show loss of these rhythms under constant darkness36–41. Cavefish populations have 56

also convergently evolved a drastic reduction in sleep compared to surface fish27. Alterations can 57

also be seen at the molecular level; an examination of clock genes (per1, per2, cry1a) in cave 58

and surface fish found changes in the expression level and timing of activation of these genes in 59

multiple cave populations.36 Consequently, the Mexican tetra provides a unique opportunity to 60


https://doi.org/10.1101/2020.01.14.906628


4

study the evolutionary response of the biological clock and circadian rhythms to the arrhythmic 61

cave environment across multiple cave populations. 62

63

In vertebrates, circadian rhythms are regulated by transcriptional feedback loops, where clock 64

proteins directly or indirectly regulate the expression of the genes from which they are 65

transcribed. The feedback loops of the circadian clock result in oscillations of gene expression of 66

~24 hours42. These oscillating transcripts make up the “circadian transcriptome” and are a 67

substantial source of rhythmic physiology and behavior43–45. The conserved nature of the 68

biological clock in vertebrates make comparative studies extremely powerful. Even across 69

evolutionary timescales, many clock genes and downstream regulators of circadian behavior 70

maintain similar functions.45 While many species of cave-dwelling organisms show evidence of 71

weakened or absent locomotor rhythms46–52, no global analysis of transcriptional changes 72

associated with clock evolution has been performed in any cave species, including A. mexicanus. 73

Comparing global transcriptional shifts over circadian time between cave and surface 74

populations can provide insight into naturally occurring dysregulation of the biological clock 75

across multiple cave populations. 76

77

Here we characterize and compare the circadian transcriptome of A. mexicanus surface fish with 78

that of three cave populations. We find widespread changes in the circadian transcriptome, with 79

cave populations showing convergent reductions and losses of transcriptional oscillations and 80

alterations in the phase of key circadian genes. Visualization of clock genes mRNAs of in the 81

midbrain and liver of surface and cave fish support the dysregulation observed in the RNAseq 82

data, and identify tissue-specific patterns unique to cave individuals. To functionally assess the 83

contribution of dysregulated genes to sleep and behavior, we investigate the function of two 84

genes linked to circadian regulation by creating CRISPR/cas9 mutants, and find that these genes 85

are involved in regulating sleep behavior in A. mexicanus. Our results suggest that circadian 86

rhythm, a highly conserved mechanism across most metazoans, is repeatedly disrupted at the 87

molecular level in cavefish, with potential consequences for organismal physiology and 88

behavior. 89

90

Results 91


https://doi.org/10.1101/2020.01.14.906628


5

Fewer genes show evidence of daily cycling in cavefish populations 92

To identify changes in rhythmic expression between cave and surface populations, we performed 93

RNAseq with total RNA from whole fish from three cave populations (Molino, Pachón, and 94

Tinaja) and one surface population (Río Choy) collected every 4-hours for one daily cycle under 95

constant darkness at 30 days post-fertilization (dpf)(see Methods). Fry were raised in a light-dark 96

cycle (14:10) and then transferred into total darkness 24 hours prior to the start of the sampling 97

period. An average of 14,197,772 reads were mapped per sample (File S1, Fig. S1-4). Filtering 98

of genes with low expression (< 100 total counts across all samples) resulted in 21,048 annotated 99

genes used for downstream analysis. Although the cave populations are not monophyletic53, 100

principle component analysis showed that the primary axis of differentiation among samples is 101

habitat (i.e., cave or surface; Fig. S5). 102

103

We used JTK_cycle54 to detect 24-hour oscillations in transcript abundance. We found that the 104

surface population had the greatest number of rhythmic transcripts (539), followed by Tinaja 105

(327), Pachón (88), and Molino (83), respectively (FDR < 0.05, see Table S1). Surprisingly few 106

genes (19) showed significant cycling across all populations (Fig. 1A,B)(File S1). In all 107

populations except Molino, we found significant overlap between rhythmic transcripts in A. 108

mexicanus and those identified in zebrafish55,56 (See Methods; at p < 0.05: surface, Pachón and 109

Tinaja, all p < 1 x 10-4; Molino p = 1, hypergeometric tests). Rhythmic transcripts in both the 110

surface and cave populations were enriched for the GO terms related to circadian rhythm, 111

including “regulation of circadian rhythm” (GO:0042752)(surface, q = 4.85 x 10-10; Tinaja, q = 112

4.16 x 10-8; Pachón, q = 2.33 x 10-6; Molino, q = 4.98 x 10-6). Rhythmic transcripts in the surface 113

population were also strongly enriched for “visual perception” (GO:0007601)(q = 4.80 x 10-12), 114

“sensory perception of light stimulus” (GO:0007602) (q = 9.28 x 10-12), and “phototransduction” 115

(GO:0007602)(q = 1.14 x 10-7), unlike cave populations (lists of enriched terms in File S1). 116

117

Nearly 22% (117) of genes found to be rhythmic in the surface population were arrhythmic (p-118

value > 0.5) in all three cave populations and 77% (416) were arrhythmic in at least one cave 119

population (Table S2, File S1)(see Methods), highlighting that the loss of rhythmicity has 120

evolved independently among independent origins of the cave phenotype. Conversely, no genes 121

that were rhythmic across all caves were arrhythmic in the surface population. Genes that were 122


https://doi.org/10.1101/2020.01.14.906628


6

rhythmic in the surface population but arrhythmic in cavefish populations include genes known 123

to be involved in the regulation of circadian rhythm (Table S3). For example, arntl2 is a gene 124

that encodes a transcriptional activator that forms a core component of the circadian clock and 125

shows conserved rhythmic activity in zebrafish55 and mice57. Despite exhibiting a conserved 126

function across vertebrates, this gene is arrhythmic in all three cave populations (surface, p = 127

0.0004, q=0.02; all caves, p > 0.5). Further, in comparing the p-values of genes with known roles 128

in circadian rhythm (61 genes, see Methods, File S1), we found that these genes were 129

significantly more rhythmic in their expression in the surface population than in cave populations 130

(Wilcoxon signed rank test, p < 0.002 in all cave-surface pairwise comparisons). 131

132

Cave and surface populations show differences in periodicity and amplitude of rhythmic 133

transcripts 134

To identify genes with changes in rhythmicity between cave and surface populations, we 135

calculated a differential rhythmicity score (SDR) for each gene for each population pair58. This 136

metric accounts for both changes in how rhythmic a transcript is, as defined by differences in the 137

JTK_cycle p-value for each gene between surface and cave populations, as well as differences in 138

the robustness of a transcript’s oscillation, as defined by differences in amplitude of gene 139

expression between the cave and surface populations. We found that 103 genes showed greater 140

rhythmicity in the surface fish for all surface-cave comparisons (e.g., surface-Pachón, surface-141

Molino, surface-Tinaja; File S1)(Fig. 1C). This set of genes was highly enriched for the pathway 142

“Circadian clock system,” with a 21-fold enrichment compared to the background set of genes 143

used in this analysis (q = 9.96 x 10-11), and was the only pathway significantly enriched after 144

false-discovery rate correction. Comparatively, relatively few genes showed significantly 145

improved rhythmicity in cave populations (Fig. 1C, genes below zero for differential periodicity 146

and differential amplitude z-score), and no genes showed increases in rhythmicity across 147

multiple caves. 148

149

This unbiased assessment revealed that genes with reductions in rhythmicity in cave populations 150

include several primary and accessory components of the core transcriptional clock (Fig. 151

2A)(Table 1). In the circadian clock’s primary feedback loop, members of the (bHLH)-PAS 152

family (e.g., CLOCKs, ARNTLs) heterodimerize and bind to E-box DNA response elements to 153


https://doi.org/10.1101/2020.01.14.906628


7

transcriptionally activate key clock proteins (e.g., PERs, CRYs). Negative feedback is then 154

achieved by CRY:PER heterodimers which inhibit the CLOCK:ARNTL complex. Another 155

regulatory loop is induced by the CLOCK:ARNTL complex activating the transcription of nr1d1 156

and rorc genes (e.g., rorca and rorcb), which in turn both positively and negatively regulate the 157

transcription of arntl (see Fig. 2B). Many of the genes that transcribe these activators and 158

repressors show reductions or loss of cycling in one or more cave populations, and a few show 159

reductions in rhythmicity in all three caves (e.g., rorca, rorcb, arntl2)(Table 1, Fig. 2). 160

Reductions in rhythmicity at core clock genes suggest an overall dampening of the core circadian 161

mechanism in cave populations compared to surface forms. 162

163

Genes with reduced rhythmicity in cavefish populations also include genes involved in 164

downstream regulation of oscillations in physiology and behavior (Table 1). Among the genes 165

which show reductions in rhythmicity across all three caves is arylalkylamine N-166

acetyltransferase 2 (aanat2), which is involved in melatonin synthesis in the pineal gland59–61. 167

Aanat2 is required for the circadian regulation of sleep in zebrafish, and mutant zebrafish sleep 168

dramatically less at night due to decreased bout length, but not decreased bout number61. 169

Interestingly, cave populations show a dramatic reduction in sleep bout length compared to 170

surface fish27. Like in zebrafish59,61, the expression of aanat2 in surface A. mexicanus shows 171

robust rhythmic behavior (q = 6.6 x 10-4) with peak expression during subjective night. In 172

contrast, cavefish do not show evidence of rhythmic transcription of aanat2 (Molino, p = 1.0; 173

Tinaja, p = 0.82; Pachón, p = 0.33) and expression does not increase in these populations during 174

subjective night (Fig. 2). 175

176

Another gene that shows reductions in rhythmicity across all three caves is camk1gb, a gene 177

involved in linking the pineal master clock with downstream physiology of the pineal gland. 178

Knock-downs of this gene in zebrafish significantly disrupt circadian rhythms of locomotor 179

activity62, similar to what is observed in Pachón cavefish39,40. Camk1gb also plays a role in 180

regulating aanat2; knockdowns of camk1gb reduce the amplitude of aanat2 expression rhythm 181

by half in zebrafish62. 182

183


https://doi.org/10.1101/2020.01.14.906628


8

Rhythmic transcripts in cave populations show a delay in phase compared to the surface 184

population 185

Next, we investigated whether the timing of the circadian clock differs between cave and surface 186

populations by comparing the peak expression time (i.e., phase) of cycling transcripts. Most 187

circadian genes show peak expression immediately before dawn or dusk63, and consistent with 188

this, surface and cave populations of A. mexicanus all showed a bimodal distribution of phases 189

(Fig. 3A). However, this distribution was shifted towards later in the day in fish from all three 190

cave populations compared to the surface population. Our results are consistent with the 191

individual gene analysis of Beale et al.36, which showed that the clock genes per1 and cry1a 192

display phase delays in fish from the Pachón and Chica caves relative to surface fish (Fig. S6). 193

We then compared the phase of A. mexicanus genes with phase estimates of orthologous core 194

clock genes in zebrafish sampled under dark conditions55. While zebrafish and A. mexicanus 195

diverged ~146 MYA64, the timing of peak expression of core clock genes was highly similar in 196

zebrafish and surface fish (median difference of 0.8 hours between orthologs, Table S4). 197

Comparatively, the phase of clock genes in the core and accessory loops are often more shifted 198

in cave populations relative to zebrafish phase than the surface population (median shifts of 4.7, 199

2.2, and 3.5 hours for Tinaja, Molino, and Pachón, respectively)(Table S4). 200

201

To further characterize differences in phase between surface and cave populations, we compared 202

the phase of all genes with evidence for rhythmic expression across cave-surface population 203

pairs (where p < 0.05 for a gene in the surface and the cave being compared). Consistent with the 204

phase shifts observed in core clock genes and accessory loops, for each cave-surface comparison, 205

we found that rhythmic transcripts showed significantly later peak expression in cave 206

populations (Wilcoxon signed rank test, all pairwise comparisons p < 2.2 x 10-16, Table S5)(Fig. 207

3B, phase of all genes in File S1). Rhythmic transcripts in Pachón showed the greatest average 208

shift, with a delay of 2.03 hours compared to surface fish (average shift, surface-Tinaja: 1.3 209

hours, surface-Molino: 0.48 hours). 210

211

Genes with circadian cis- elements show loss of cycling and changes in phase in cavefish 212

In light of the alterations we see in the expression of genes in the circadian transcriptome, we 213

next investigated circadian cis- elements in cavefish and surface fish. In vertebrates, circadian 214


https://doi.org/10.1101/2020.01.14.906628


9

oscillations in gene expression are a consequence of interlocking transcriptional feedback loops 215

of core clock proteins binding to a combination of known cis-elements (E-box, RRE, D-box) in 216

the promoters and enhancers of target genes55,65. The timing of activator and repressor binding to 217

these cis- elements results in the phase of oscillating transcripts65. 218

219

To better understand what drives oscillation patterns in A. mexicanus, we extracted promoter 220

sequences of genes with rhythmic transcription in the surface population and identified 221

transcription factor (TF) binding motifs in each sequence (Table S6). We consider rhythmic 222

genes with significant circadian binding motifs (i.e., E-box, RRE, D-box sequences) putative 223

targets of clock proteins in the circadian feedback loop. We then used a sliding window approach 224

to identify the specific time window when the phases of circadian TF targets are most enriched in 225

surface fish (see Methods). Consistent with observations in zebrafish55, the E-box, which is 226

bound by CLOCK-ARNTL complexes as part of the core feedback loop66, showed phase-227

specific enrichment within the early peak (CT 2), and RRE elements, which are bound by RORA 228

proteins and NR1D1/NR1D2 of the accessory loop67, were enriched within the later peak (CT 229

14-16)(Fig. S7)(see Methods). The D-box, which binds the repressor NFIL3, was also enriched 230

within the early peak (CT 0-6), as in zebrafish55. The correspondence between surface A. 231

mexicanus surface fish and zebrafish suggests that the timing of key regulatory cascades 232

mediating circadian oscillations in gene expression is largely conserved between these species. 233

234

We then searched for circadian cis- elements in cavefish sequences upstream of genes that have 235

lost rhythmicity in cave populations. A large proportion of the putative targets of core clock 236

elements identified in the surface fish (e.g., genes identified as having an E-box, RRE or D-box 237

motifs in their promoter sequence as described above) do not cycle in one or more cave 238

populations: >77% are arrhythmic in at least one cave population. However, for nearly all gene 239

for which we identified a proximal circadian motif in the surface fish, we also identified a 240

binding site in cavefish (see Table S7 and Supplemental Methods). The maintenance of clock 241

binding sites in cavefish populations suggests that the reduction in the number of rhythmic 242

transcripts in cavefish is likely to be a consequence of changes in the transcription of core clock 243

genes, rather than divergence at the target cis- binding sites at downstream genes. 244

245


https://doi.org/10.1101/2020.01.14.906628


10

We then compared the phase of putative clock targets in surface fish and cavefish. Consistent 246

with the phase shifts seen in the core clock (Table S4), putative clock targets also show shifts in 247

their phase in cavefish compared to surface fish (Table S8, Fig. S7). In the case of the E-box, in 248

cave populations the phase intervals with the highest proportion of these motifs occur later (CT 249

4, 6, 6-10 in Molino, Pachón, and Tinaja, respectively) than what is seen in the surface 250

population (CT 2) (Fig. S7). In contrast, for the RRE and D-box motif, there is no phase with 251

enrichment in cave populations. 252

253

Visualization of key circadian mRNAs reveals tissue-specific expression patterns in cave and 254

surface populations 255

Our whole-fry RNAseq data suggest that rhythms in the core clock are dampened, and that 256

rhythmic transcripts arephase shifted. Teleost circadian systems are highly decentralized and 257

autonomous core clock gene expression may be observed in many tissues68–70, and tissue-specific 258

differences may be obscured in whole organisms preparations71. 259

260

To address potential tissue-specific contributions to differences between cave and surface 261

populations, we used RNAscope® fluorescence in situ hybridization (FISH)72. We collected 262

brains and livers from individual 30dpf fry at 3 timepoints (CT0, CT8, and CT16)70 and 263

performed RNA FISH for key circadian mRNAs in the primary loop (per1 and arntl1a) and 264

regulatory loop (rorca and rorcb) (Advanced Cell Diagnostics, Hayward, CA, USA, see 265

Methods for details and controls). Brains were sectioned to allow visualization of the optic 266

tectum and periglomerular grey zone (Fig. S8). These are both sites showing significant clock 267

gene expression in zebrafish and these brain structures were consistently identifiable despite the 268

small size of the 30dpf midbrain73. 269

270

We find that temporal expression patterns of per1a and arntl1a mRNA in the midbrains of 271

surface fish is consistent with our RNAseq results and expression patterns in zebrafish55 (Fig. 4, 272

‘B’). Per1a expression peaks strongly at dawn (CT0) and rapidly diminishes. Arntl1a cycles in 273

anti-phase and shows high expression at dusk (CT16). In contrast, cavefish show a number of 274

population specific alterations to the temporal expression of per1a and arntl1a over the day (Fig. 275

4, S10, S11, S12). Per1a expression in cavefish shows broader temporal expression, consistent 276


https://doi.org/10.1101/2020.01.14.906628


11

with phase shift at this gene seen in whole fish (Fig. S6). Where per1a expression rapidly 277

diminishes in surface midbrain after CT0, expression of per1a persists through CT8 in cave 278

populations. Surprising, Tinaja per1a expression is at its lowest at CT0, when the surface 279

population per1a expression is greatest. Similarly, we see the persistence of arntl1a expression 280

in Tinaja and Molino outside of CT16, the primary window of expression in the surface 281

populations. Molino midbrains in particular showed high basal expression compared to the 282

surface, specifically at CT8 when per1a and arntl1a expression is nearly absent in surface fish 283

(Figs. S10, S12). 284

285

Temporal expression patterns for per1a and arntl1a in surface fish is also similar to zebrafish 286

(Fig. 4, ‘L’). Surface fish liver expression of per1a and arntl1a are largely in synchrony with the 287

expression of these genes in surface midbrain, however, expression of per1a and arntl1a in the 288

liver is also seen at the intermediate timepoint, CT8. Compared to surface, Pachón and Tinaja 289

livers show similar arntl1a expression but peak expression of per1a is delayed (occurring at CT8 290

instead of CT0). Per1a expression persists in Pachón and Tinaja at CT16, when per1a transcripts 291

are low in surface livers. Intriguingly, peak and trough expression in Pachón and Tinaja livers 292

appear to be out of sync with those in the midbrains (Fig. 4, ‘L’ vs. ‘B’). Per1a expression in 293

Tinaja liver is greatest at CT8, but is expressed highest in brains at CT16. In Pachón midbrains, 294

per1a expression is highest at CT0, but expression is highest in the livers at CT8 and CT16. In 295

contrast, Molino livers exhibit temporal expression of per1a that is synchronous with brains. 296

However, as with the midbrain, per1a has high basal expression in Molino compared to surface, 297

Pachón, and Tinaja. 298

299

Next, we examined the expression pattern and timing of rorca and rorcb, two gene members of 300

the regulatory loop that regulate the expression of arntl paralogs and circadian transcription67. 301

Our RNAseq results indicate that rorca and rorcb show peak expression in whole surface fish at 302

CT12 (Fig. 5, S13), similar to what has been observed in zebrafish (Table S4). Consistent with 303

this, in RNA FISH of surface fish livers and midbrains, rorca and rorcb expression was highest 304

midday (Fig. 5, S13). Temporal expression of rorca and rorcb was similar between livers and the 305

midbrain in cave populations. When compared to surface tissues, the expression of rorca and 306

rorcb is delayed in Pachón and much broader in Tinaja and Molino populations. Molino 307


https://doi.org/10.1101/2020.01.14.906628


12

midbrains have high expression of rorca and rorcb compared to all other populations. In 308

addition, rorcb expression is seen across timepoints in all caves, particularly in Molino, whereas 309

it is nearly absent in surface fish at CT0 and CT8 (Fig. 5). 310

311

Thus, as observed in the whole organism RNAseq analysis, tissue-specific expression of these 312

primary loop and regulatory loop genes confirm the phase shifts and dysregulation of cavefish 313

transcripts compared to surface fish, though, our RNAscope analysis suggests that disruptions to 314

the temporal regulation per1a and arntl1a expression may not manifest uniformly across tissues 315

in Pachón and Tinaja cavefish (Fig. 4). 316

317

Aanat2 and rorca regulate sleep behavior in surface fish 318

The circadian clock plays an important role in regulating the sleep-wake cycle70. We have found 319

that expression of the genes in the core pacemaker, as well as a circadian sleep regulator 320

(aanat2), are differentiated between cave and surface populations (see Fig. 2). Cavefish 321

populations have repeatedly evolved towards a reduction in sleep at night and overall, raising the 322

question of whether dysregulation of the biological clock is contributing to changes in sleep 323

behavior between cave and surface populations. However, the function of circadian genes in A. 324

mexicanus sleep regulation has not been investigated. 325

326

To explore the role of clock genes in sleep regulation in A. mexicanus, we created crispant 327

surface fish mosaic for mutations in aanat2 and rorca. Rorca and aanat2 were selected because 328

they show differences in expression between the surface and all cave populations, and because 329

aanat2 plays an important role sleep regulation in zebrafish61. Mutagenized 30 dpf fish and 330

wildtype (WT) sibling controls were phenotyped for sleep behavior. Raw locomotor behavior 331

was processed to quantify sleep duration and behavioral architecture (locomotor distance, 332

waking activity, sleep bout duration, and sleep bout number) as previously described38,74. For 333

each injected construct, three separate biological replicates were carried out to ensure the 334

phenotype was consistently reproducible. Genotyping confirmed mutagenesis (Figs. S14, S15). 335

336

Aanat2 crispants demonstrate the role of this gene in the regulation of sleep in A. mexicanus. 337

While there was no difference in total sleep duration between WT surface and aanat2 surface 338


https://doi.org/10.1101/2020.01.14.906628


13

fish crispants (unpaired t-test, p = 0.98, Fig. 6A), night sleep duration (minutes/hour) was 339

significantly reduced in aanat2 crispants (unpaired t-test, p = 0.0108, Fig. 6B,C). These results 340

are consistent with the function of this gene in zebrafish, where increased expression of this gene 341

during subjective night increases melatonin production, regulating nighttime sleep61. Zebrafish 342

aanat2 knockout mutants produce little or no melatonin and sleep almost half as much during the 343

night as control fish61. 344

345

Rorca crispants also show differences in sleep behavior compared to WT surface fish. Like 346

aanat2 crispants, rorca mutants sleep significantly less minutes per hour at night compared to 347

WT sibling fish (unpaired t-test, p < 0.0001, Fig. 6E,F); however, rorca crispants also sleep less 348

overall during a 24-hr period (unpaired t-test, p < 0.0001, Fig. 6D). While WT surface fish sleep 349

an average of 5.16 hours per 24-hr period, rorca surface mutants sleep just 2.26 hours per 24-hr 350

period. Rorca expression has not been previously associated with sleep regulation, but it is a 351

transcriptional regulator of the core clock gene arntl1a/b. The mammalian ortholog of arntl1a/b, 352

arntl1, has been implicated as a regulator in the sleep-wake cycle and deletions alter sleep 353

behavior in mammals75–77. Altogether, these results demonstrate that alterations to the biological 354

clock in surface fish can alter sleep behavior. Future investigations into the specific genetic 355

variants between surface and cavefish will help delimit the role of the biological clock in 356

cavefish sleep phenotypes. 357

358

Discussion 359

Circadian rhythms are nearly ubiquitous in eukaryotes, directing aspects of behavior, physiology, 360

and metabolism2. The biological clock is thought to provide a mechanism for organisms to 361

synchronize their physiology and behavior to predictable daily cycles3,4. However, we have a 362

limited understanding of how the biological clock is altered in the face of arrhythmic 363

environments, where organisms are isolated from regular environmental cues15,36,42,78,79. Caves 364

provide a natural laboratory for perturbations to circadian biology: isolated from light-dark 365

cycles and with little fluctuation in temperature, cave-dwelling organisms lack many of the 366

environmental stimuli found outside caves15,16,36. 367

368


https://doi.org/10.1101/2020.01.14.906628


14

Here we have utilized comparisons between conspecific Astyanax mexicanus cavefish and 369

surface fish to explore changes to the biological clock following adaptation to the cave 370

environment. Using RNAseq to survey gene expression changes over a 24-hr period, we 371

identified several repeated alterations in the biological clock in cavefish populations, including: 372

1) a reduction in the number of transcripts with evidence of daily cycling, 373

2) the changes in the oscillation patterns of genes in the primary circadian feedback loop (e.g., 374

arntl2, arntl1a/b, cry1ba, cry1bb) and accessory loops (rorca/b, bhlhe40/1), and 3) an overall 375

delay in phase of the genes in the circadian transcriptome, including core clock genes and their 376

putative targets. Population genetic analyses indicate that a number of genes in the circadian 377

transcriptome are also genetically differentiated between the surface and one or more cave 378

populations (see Supplemental Material and Methods). Visualization of mRNAs of core (arntl1a, 379

per1a) and accessory loop (rorca, rorcb) in the midbrain and the liver of cave and surface fish 380

support changes in the regulation of key clock genes and highlight tissue-specific changes in the 381

expression of core clock genes in two of the cavefish populations relative to surface fish. The 382

decoupling of rhythms between tissues in caves is an interesting biological feature that warrants 383

further investigation across more tissues in these populations. 384

385

In addition to changes in the core and accessory loops, we also found that an important regulator 386

of circadian physiology, aanat2, shows reductions in rhythmicity across all three cave 387

populations. Aanat2 encodes a protein that regulates melatonin production in the pineal gland, 388

and robust expression of this gene is considered a marker for circadian clock function59,61,80,81. 389

We found that cavefish populations, unlike surface fish and other teleosts82, did not show 390

increased aanat2 expression during the subjective night (Fig. 2), making dysregulation of the 391

circadian clock an interesting candidate for the repeated changes in sleep regulation in seen 392

across cavefish populations27. Investigating the function of aanat2 and rorca through the 393

creation of aanat2 and rorca crispants revealed that these genes play a role in sleep regulation in 394

A. mexicanus. Consequently, we speculate that dysregulation of the biological clock may have 395

behavioral consequences for cavefish, and in particular, may contribute to the reductions in sleep 396

seen across cavefish populations. 397

398


https://doi.org/10.1101/2020.01.14.906628


15

Beyond the regulation of circadian physiology, environmental light plays an essential role in cell 399

cycle control and activating DNA repair in teleosts83–86. Previous work in this system has found 400

that cave populations of A. mexicanus show higher levels of DNA repair activity and increased 401

expression of DNA repair genes in the dark36. As a common light signaling pathway controls 402

both DNA repair and clock entrainment, one purposed mechanism for this is that cavefish have 403

sustained upregulation of clock genes normally driven by light, akin to ‘perceiving’ sustained 404

light exposure36. While we found that DNA repair genes are more often upregulated in cave 405

populations (Table S10, see Supplemental Material and Methods), we did not find that genes 406

associated with light induction were more likely to be upregulated in cavefish compared to 407

surface fish under dark-dark conditions (Table S11). Nor did we find that light-activated genes in 408

the core circadian feedback loop are consistently upregulated in cave populations (see 409

Supplemental Material and Methods, Figure S16). Thus, more work is necessary to understand 410

the relationship between clock evolution and DNA repair in this system. 411

412

Altogether, these results suggest that highly conserved features of circadian biology have been 413

repeatedly disrupted across populations of cavefish. Further research on the molecular basis of 414

clock dysregulation will help delineate if clock dysfunction is a consequence of molecular 415

alterations to (1) the pathways that perceive, or synchronize the biological clock to, external 416

stimuli, or (2) alterations to the core transcriptional feedback loop, and whether this 417

dysregulation is a consequence of selection or drift after the movement into the cave 418

environment. 419

420

Methods 421

Sampling 422

Samples from each population were derived from the same mating clutch and raised in 14:10 423

light-dark cycle. All fish were exactly 30 days post fertilization at the start of the experiment. 424

Fish were kept in total darkness for 24 hours prior to sampling and throughout the duration of the 425

experiment. Six replicates were first sampled at 6am (Circadian Time 0) and then every four 426

hours throughout until 2am (corresponding to CT 20). During the experiment, fish were fed ad 427

libitum twice daily, in the morning and evening (exactly at 8am and 8pm). Individuals were 428

immediately flash frozen in liquid nitrogen. RNA was extracted from the whole organism with 429


https://doi.org/10.1101/2020.01.14.906628


16

extraction batch randomized for time of sampling and population. Samples were sequenced on 430

the Illumina HiSeq 2500 platform to produce 125-bp paired-end reads (File S1) using strand-431

specific library preparations. Samples were randomized across lanes relative to population and 432

time of sampling. Previous work showed no extraction batch or lane effect on these samples87. 433

434

Read mapping 435

Reads were cleaned of adapter contamination and low-quality bases with Trimmomatic 436

(v0.33)88. Cleaned reads were mapped to the A. mexicanus draft genome assembly v1.02 437

(GCA_000372685.1) with STAR89 and reads overlapping exonic regions were counted with 438

Stringtie (v1.3.3d)90 based on the A. mexicanus Ensembl v91 annotation. Genes with fewer than 439

100 reads across all samples were removed from the analysis. Reads were subsequently 440

normalized for library size and transformed with a variance stabilizing transformation with 441

DESeq291 for principle component analysis. 442

443

Identification of rhythmic transcripts 444

Rhythmic transcripts with 24-h periodicity were identified using the program JTK_cycle54, using 445

one 24-h cycle, with a spacing of 4 hours and 6 replicates per time point. JTK_cycle was also 446

used to estimate the amplitude and phase of each rhythmic transcript. We retained genes as 447

rhythmic at an False Discovery Rate (FDR) of 5%. Our overall result, that more genes are 448

cycling in the surface population that in cave populations (Table S1), was insensitive to specific 449

value of the FDR cut-off. We defined arrhythmic transcripts at p-value > 0.5, a conservative cut-450

off for arrhythmic expression that has been used in other studies58,92. 451

452

Differential rhythmicity analysis 453

To identify differential rhythmicity between pairs of populations, we calculated a differential 454

rhythmicity score (SDR) for genes with a JTK_cycle p-value < 1 in either population. As in 455

Kuintzle et al.58, SDR was defined as: 456

!"# = &' +&#√2

457

458


https://doi.org/10.1101/2020.01.14.906628


17

Where Zp is the Z-score for changes in periodicity between populations, where changes in 459

periodicity are defined as log(p -value of population 1) – log(p -value of population 2). ZR is the 460

Z-score computed for changes in amplitude between populations, where changes in amplitude 461

are defined as: log2(Amplitude of population 2 / Amplitude of population 1). Amplitude for each 462

gene was estimated with JTK_cycle. 463

464

We then computed a p-value for SDR values using a Gaussian distribution based on the fit to the 465

empirical distribution. We used R’s p.adjust to perform a Benjamini & Hochberg correction for 466

multiple testing (Table S9). 467

468

Promoter analysis of rhythmic genes 469

We defined putative promoters as the region 1-kb upstream to 200bp downstream from the 470

transcription start site (TSS). Population genetic data53 was used to identify variants with 471

putative promoter regions segregating between populations and create alternative reference 472

genomes for each population. Genes with a JTK_cycle FDR < 0.1 in the surface were used to 473

identify phase enrichment of circadian cis- elements in the surface population. FIMO of the 474

MEME suite93 was used to identify motifs in each promoter. FIMO was used to estimate p-475

values for motifs in each sequence; motifs with p-values < 0.0001 (the default cut-off) were 476

considered for downstream analyses. To identify phase-specific enrichment of binding motifs, 477

we tested sliding windows of time with Fisher’s exact tests for motifs in phase versus motifs out 478

of phase compared to the total set of in- and out- of phase genes. To compute changes in phase of 479

genes that are putative targets of circadian transcription factors between surface and cave 480

populations, we calculated the minimum distance between the two phases based on a 24-h 481

clock. Further details of this analysis are available in the Supplemental Methods. 482

483

Tissue preparation for RNA fluorescence in situ hybridization (FISH) 484

Pachón, Molino, Tinaja, and surface fish derived from the Río Choy population were raised in 485

densities of 20-25 individuals/3L tank in 14:10LD cycle at 23°C and 750-800μS/cm 486

conductivity, with feeding and water quality control as previously described94. 487

488


https://doi.org/10.1101/2020.01.14.906628


18

29dpf fish were placed in constant darkness in the dark phase of the day preceding tissue 489

collection. 30dpf fish were fed ad libitum at 8am and 8pm on the day of tissue collection. 490

Individual fish were euthanized in MS-222 at CT0, CT8, and CT16. Experiments were 491

performed in duplicate: two fish were euthanized and dissected at each time point from each 492

population. All efforts were made to reduce light exposure prior to fixation. Dissected brains and 493

livers were placed into a freshly prepared solution of 4% PFA in DEPC 1X PBS on ice. Samples 494

were transferred to room temperature after 3-5 minutes on ice and allowed to fix for 1 hour. 495

Samples were then placed at 4°C for 36 hours. After 36 hours, samples were rinsed briefly in 496

DEPC water and washed in DEPC 1X PBS three times for 15 minutes with constant shaking. A 497

graded EtOH wash was performed (30%- 50%- 70%- 100%) in DEPC treated water with 5 498

minutes between steps. 499

500

Tissue processing and paraffin embedding were performed with a PATHOS Delta hybrid tissue 501

processor (Milestone Medical Technologies, Inc, MI, USA)95. Paraffin sections were cut with 502

12μm thickness using a Leica RM2255 microtome (Leica Biosystems Inc. Buffalo Grove, IL, 503

USA) under RNase free conditions. Brains were sectioned coronally through the mesencephalon-504

diencephalon to allow visualization of the optic tectum and periventricular grey zone, sites 505

showing significant clock gene expression in zebrafish73 (Fig. S8). We were unable to 506

consistently section through the same region of the brain in order to compare expression in 507

hypothalamic nuclei due to the size of 30dpf brains73. Livers were sectioned longitudinally. 508

Sections were mounted on SureBond charged microscope slides (cat#SL6332-1, Avantik, 509

Springfield, NJ, USA). To allow for comparison between populations, all brains or livers from a 510

single probe set were processed at once (e.g., rorca-rorcb brains from all populations, and all 511

time points were processed at the same time). In addition, brains or livers from each time point 512

(CT0, CT8, CT16) were processed on a single slide to allow for a direct comparison between 513

time points within populations. 514

515

RNA FISH, imaging and analysis 516

RNAscope® probes were designed by ACDBio to target all known mRNA transcripts of genes 517

per1a (ENSAMXG00000019909), arntl1a (ENSAMXG00000011758), rorca 518

(ENSAMXG00000009363), and rorcb (ENSAMXG00000015029) (Advanced Cell Diagnostics, 519


https://doi.org/10.1101/2020.01.14.906628


19

Hayward, CA, USA) based on Astyanax_mexicanus-2.0, INSDC Assembly (GCA_000372685). 520

Probes were ordered for two-plexing (per1a-arntl1a and rorca-rorcb) to provide an internal 521

control in each sample for mRNA phase and quantity. Probes may be ordered at the ACD Online 522

Store using the following catalog numbers: per1a (590801), arntl1a (590831-C2), rorca 523

(590811), and rorcb (590821-C2). In situ hybridization of per1a, arntl1a, rorca, and rorcb 524

mRNA was performed on paraffin sections using RNAscope Multiplex Fluorescent Detection 525

Kit v2 (Advanced Cell Diagnostics, Hayward, CA, USA) according to the manufacturer’s 526

protocol. 527

528

Slides were stored in the dark at 4°C before imaging. Fluorescent images of sections were taken 529

with a Nikon Eclipse TI equipped with a Yokogawa CSU W1 spinning disk head and 530

Hamamatsu ORCA-Flash 4.O camera for high-resolution imaging using a Nikon 20x/0.75 Plan 531

Apo objective. Probes rorca and per1a were imaged with 561nm laser and collected with an 532

ET605/70m emission filter, and arntl1a and rorcb with 633nm laser and collected with an 533

ET700/75m emission filter. DAPI was excited at 405nm and collected with an ET455/50m 534

emission filter. Samples were identified automatically for imaging using a custom script as in 535

Guo et al.96, and each sample was imaged as a tile scan with 10% overlap and z-stack of depth of 536

18µm with 0.9 µm optical slice thickness. Microscope and camera settings during acquisition 537

were identical across all samples to allow for a direct comparison between timepoints and 538

populations. Image processing in Fiji97 was identical between populations and time points within 539

a tissue and probe set. Briefly, tiles were stitched into a complete image using Grid/Collection 540

Stitching98, maximum projected and contrast adjusted. To properly compare cave populations to 541

surface populations, we set the intensity ranges for each image to that of the surface sample 542

images (Fig. 4). The intensities of Molino and Tinaja brain images in Fig. S11 are adjusted for 543

oversaturation. Figs. 4 and 5 exclude the DAPI channel. Corresponding DAPI staining for these 544

figures can be found in Fig. S9. Images are representative of two fish collected from each 545

timepoint per population. For Figs. 4 and 5, maximum projected images are shown. 546

547

For quantification in Fig. S10, tiles were stitched into a complete image using Grid/Collection 548

Stitching98. Stitched images were sum projected and background subtracted. 400x400 anatomical 549

regions of interest were identified by visual inspection and average intensities were measured for 550


https://doi.org/10.1101/2020.01.14.906628


20

each channel. To allow for a comparison of relative expression per cell from multiple samples, 551

DAPI intensity was measured as a proxy for cell density, and FISH signal intensity was 552

normalized against DAPI intensity for each region. Technical replicates (two to three sections 553

from the same liver or brain) were averaged to make biological replicates represented as points 554

in Fig. S10. All groups have two biological replicates with the exception of rorca-rorcb Pachón 555

brains which have only one due to sample loss. Graphing and statistical analysis were performed 556

using GraphPad Prism software (GraphPad, Prism version 8.3.0, GraphPad Software, San Diego, 557

California USA). For each mRNA probe, cave population means were compared with the control 558

mean (Surface) within timepoints using 2-way ANOVA. Dunnett's test was used to correct for 559

multiple comparisons across populations and timepoints. Original data underlying this part of the 560

manuscript can be accessed from the Stowers Original Data Repository at 561

http://www.stowers.org/research/publications/libpb-1485. 562

563

CRISPR/Cas9 design and genotyping 564

Functional experiments were performed by generating crispant fish mosaic for mutations in 565

aanat2 and rorca. CRISPR gRNAs were designed using ChopChop v3 software99. gRNAs were 566

designed to target an exon in each of the genes and to produce a double stranded break close to a 567

restriction enzyme site for genotyping purposes. For aanat2, the gRNA was 5’-568

GGTGTGCCGCCGCTGCCGGA-3’ and for rorca the gRNA was 5’-569

gGAGAACGGTAACGGCGGGCA-3’ where the lower case g at the 5’ end was added to the 570

sequence for T7 transcription (restriction enzyme target sites are underlined). gRNAs were 571

synthesized as previously described100 with modifications101,102. Briefly, gRNA specific oligos 572

were synthesized (IDT) that contained the gRNA target site, T7 promoter, and an overlap 573

sequence: 574

575

aanat2 oligo A: 576

577

rorca oligo A: 578

5’- TAATACGACTCACTATAGGTGTGCCGCCGCTGCCGGAGTTTTAGAGCTAGAAATAGC-3’

5’-TAATACGACTCACTATAgGAGAACGGTAACGGCGGGCAGTTTTAGAGCTAGAAATAGC-3’


https://doi.org/10.1101/2020.01.14.906628


21

Each oligo A was annealed to the oligo B: 579

580

and these annealed oligos were amplified. gRNAs were transcribed using the T7 Megascript kit 581

(Ambion), as in Klaassen et al.103 and Stahl et al.102, and purified using a miRNeasy mini kit 582

(Qiagen). Nls-Cas9-nls104 mRNA was transcribed using the mMessage mMachine T3 kit (Life 583

Technologies) following the manufacturer’s instructions and purified using the RNeasy MinElute 584

kit (Qiagen) following manufacturer’s instructions. 150 pg Cas9 mRNA and 25 pg aanat2 gRNA 585

or 150 pg Cas9 mRNA and 25 pg rorca gRNA were injected into single-cell surface fish 586

embryos (2 nL/embryo were injected). 587

588

Genomic DNA from injected embryos and wild-type (uninjected) controls was extracted at 48 589

hours post-fertilization101,102 and used for genotyping by PCR to determine if mutagenesis was 590

achieved at the locus through gel electrophoresis (Figs. S14, S15). Briefly, gRNAs were 591

designed to disrupt specific restriction enzyme sites; samples were incubated with that restriction 592

enzyme to determine if the Cas9 cut. The undigested fragment acts as a control while the 593

digested fragment is used to genotype. Cases where we see a band where the undigested 594

fragment is indicates that the restriction enzyme was not able to cut due to the site being 595

disrupted because of the CRISPR/Cas9 + gRNA. Gene specific primers are given in File S1. 596

PCRs were performed with a 56 °C annealing temperature and a 1-minute extension time for 35 597

cycles. The resulting PCR product was split in half, and one half was restriction enzyme 598

digested. A DNA fragment containing the aanat2 target fragment was digested with BbvI (New 599

England Biolabs Inc.). The rorca DNA fragment was digested with Cac8I (New England 600

Biolabs). Both DNA fragments were digested at 37 °C for 1 hour. 601

602

For wildtype individuals, the PCR fragment was used for cloning. For crispant fish, restriction 603

enzyme digests were performed (as described above) and the uncleaved bands were gel purified 604

using a gel extraction kit (Qiagen). PCR products were cloned into the pGEM®-T Easy Vector 605

(Promega), and three clones per individual were picked, grown, and purified using QIAprep Spin 606

5’-GATCCGCACCGACTCGGTGCCACTTTTTCAAGTTGATAACGGACTAGCCTTATTTTAACTTGCTATTTCTAGCTCTAAAAC-3’


https://doi.org/10.1101/2020.01.14.906628


22

Miniprep Kit (Qiagen) and then sequenced by Sanger sequencing to confirm mutagenesis 607

(Eurofins Genomics). 608

609

Quantifying sleep behavior 610

30 dpf injected crispant fish and WT (uninjected) sibling controls were used for all behavioral 611

experiments. Fish were maintained on a 10-14 light-dark cycle throughout development as well 612

as for behavioral experiments. Fish were placed in 12-well tissue culture plates (Cellstar) 18-24 613

hours before behavioral recording to acclimate to the recording chamber. Fish were fed normal 614

brine shrimp meals before the start of the recording, which began at ZT0 (zeitgeber time) and 615

lasted 24 hours. Video recordings were processed in Ethovision XT (v13). Raw locomotor data 616

was processed with custom-written scripts to quantify sleep duration and behavioral architecture 617

such as locomotor distance, waking activity, sleep bout duration, and sleep bout number38,74. For 618

each injected construct, three separate biological replicates were carried out to ensure the 619

phenotype was consistently reproducible. Each biological replicate represented different clutches 620

of fish, injected on different days. Both wildtype and crispant individuals were assessed from 621

each clutch. 622

623

Data accessibility 624

All sequence data were deposited in the short-read archive (SRA) under the accession 625

PRJNA421208. Data underlying RNA FISH analysis can be found at 626

http://www.stowers.org/research/publications/libpb-1485. 627

628

Acknowledgements 629

We thank the University of Minnesota Genomics Center for their guidance and performing the 630

cDNA library preparations and Illumina HiSeq 2500 sequencing. The Minnesota 631

Supercomputing Institute (MSI) at the University of Minnesota provided resources that 632

contributed to the research results reported within this paper. Funding was supported by NIH 633

(1R01GM127872-01 to SEM and ACK) and NSF award IOS 165674 to ACK, and a US-Israel 634

BSF award to ACK. KLM was supported by a Ruth Kirschstein National Research Service 635

Award from NIH and a Stanford Center for Computational, Evolutionary and Human Genomics 636

(CEHG) postdoctoral fellowship. CNP was supported by Grand Challenges in Biology 637


https://doi.org/10.1101/2020.01.14.906628


23

Postdoctoral Program at the University of Minnesota College of Biological Sciences. 638

Experiments were approved by the Institutional Animal Care and Use Committee at Florida 639

Atlantic University (Protocols #A15-32 and #A18-38) and Stowers Institute for Medical 640

Research (institutional authorization 2019-084). 641

642 Author contributions 643 Conceptualization of experiments and analysis by SEM, ACK, NR, JK, KLM, JLP, JBJ; 644 Methodology by SEM, KLM, JBJ, BAS, EF, DT, SES, BDS, JK, NR, AK, CNP; Formal 645 analysis and investigation by KLM, JBJ, SEM, JLP, EF, DT, SES, BDS, JK; Writing – original 646 draft by KLM, SEM; Writing – review and editing by JBJ, JLP, CNP, BAS, EF, DT, SES, BDS, 647 JK, NR, ACK. 648 649 650 References 651 652 1. Dunlap, J. C. Molecular bases for circadian clocks. Cell 96, 271–290 (1999). 653

2. Bell-Pedersen, D. et al. Circadian rhythms from multiple oscillators: lessons from diverse 654

organisms. Nat Rev Genet 6, 544–556 (2005). 655

3. Yerushalmi, S. & Green, R. M. Evidence for the adaptive significance of circadian rhythms. 656

Ecology Letters 12, 970–981 (2009). 657

4. Vaze, K. M. & Sharma, V. K. On the adaptive significance of circadian clocks for their 658

owners. Chronobiology International 30, 413–433 (2013). 659

5. Dodd, A. N. et al. Plant circadian clocks increase photosynthesis, growth, survival, and 660

competitive advantage. Science 309, 630–633 (2005). 661

6. Fuller, P. M., Gooley, J. J. & Saper, C. B. Neurobiology of the sleep-wake cycle: sleep 662

architecture, circadian regulation, and regulatory feedback. J Biol Rhythms 21, 482–493 663

(2006). 664

7. Bass, J. & Takahashi, J. S. Circadian integration of metabolism and energetics. Science 330, 665

1349–1354 (2010). 666


https://doi.org/10.1101/2020.01.14.906628


24

8. Hunter-Ensor, M., Ousley, A. & Sehgal, A. Regulation of the Drosophila protein Timeless 667

suggests a mechanism for resetting the circadian clock by light. Cell 84, 677–685 (1996). 668

9. Hurd, M. W. & Cahill, G. M. Entraining signals initiate behavioral circadian rhythmicity in 669

larval zebrafish. J Biol Rhythms 17, 307–314 (2002). 670

10. Tamai, T. K., Carr, A. J. & Whitmore, D. Zebrafish circadian clocks: cells that see light. 671

Biochem Soc Trans 33, 962–966 (2005). 672

11. Ziv, L., Levkovitz, S., Toyama, R., Falcon, J. & Gothilf, Y. Functional development of the 673

zebrafish pineal gland: light-induced expression of period2 is required for onset of the 674

circadian clock. Journal of Neuroendocrinology 17, 314–320 (2005). 675

12. Libert, S., Bonkowski, M. S., Pointer, K., Pletcher, S. D. & Guarente, L. Deviation of innate 676

circadian period from 24 h reduces longevity in mice. Aging Cell 11, 794–800 (2012). 677

13. Pittendrigh, C. S. & Minis, D. H. Circadian systems: Longevity as a function of circadian 678

resonance in Drosophila melanogaster. PNAS 69, 1537–1539 (1972). 679

14. Scheer, F. A. J. L., Hilton, M. F., Mantzoros, C. S. & Shea, S. A. Adverse metabolic and 680

cardiovascular consequences of circadian misalignment. Proc Natl Acad Sci U S A 106, 681

4453–4458 (2009). 682

15. Beale, A. D., Whitmore, D. & Moran, D. Life in a dark biosphere: a review of circadian 683

physiology in “arrhythmic” environments. J Comp Physiol B 186, 947–968 (2016). 684

16. Poulson, T. L. & White, W. B. The cave environment. Science 165, 971–981 (1969). 685

17. Tabin, J. A. et al. Temperature preference of cave and surface populations of Astyanax 686

mexicanus. Developmental Biology 441, 338–344 (2018). 687

18. Keene, A., Yoshizawa, M. & McGaugh, S. E. Biology and Evolution of the Mexican 688

Cavefish. (Academic Press, 2015). 689


https://doi.org/10.1101/2020.01.14.906628


25

19. Yamamoto, Y., Stock, D. W. & Jeffery, W. R. Hedgehog signalling controls eye 690

degeneration in blind cavefish. Nature 431, 844–847 (2004). 691

20. Yamamoto, Y., Byerly, M. S., Jackman, W. R. & Jeffery, W. R. Pleiotropic functions of 692

embryonic sonic hedgehog expression link jaw and taste bud amplification with eye loss 693

during cavefish evolution. Developmental Biology 330, 200–211 (2009). 694

21. Krishnan, J. & Rohner, N. Cavefish and the basis for eye loss. Philosophical Transactions of 695

the Royal Society B: Biological Sciences 372, 20150487 (2017). 696

22. Jeffery, W. R. Regressive evolution of pigmentation in the cavefish Astyanax. Israel Journal 697

of Ecology & Evolution 52, 405–422 (2006). 698

23. Gross, J. B., Borowsky, R. & Tabin, C. J. A novel role for Mc1r in the parallel evolution of 699

depigmentation in independent populations of the cavefish Astyanax mexicanus. PLOS 700

Genetics 5, e1000326 (2009). 701

24. Protas, M. E. et al. Genetic analysis of cavefish reveals molecular convergence in the 702

evolution of albinism. Nat Genet 38, 107–111 (2006). 703

25. Stahl, B. A. & Gross, J. B. Alterations in Mc1r gene expression are associated with 704

regressive pigmentation in Astyanax cavefish. Dev Genes Evol 225, 367–375 (2015). 705

26. Yoshizawa, M., Gorički, Š., Soares, D. & Jeffery, W. R. Evolution of a behavioral shift 706

mediated by superficial neuromasts helps cavefish find food in darkness. Current Biology 20, 707

1631–1636 (2010). 708

27. Duboué, E. R., Keene, A. C. & Borowsky, R. L. Evolutionary convergence on sleep loss in 709

cavefish populations. Current Biology 21, 671–676 (2011). 710


https://doi.org/10.1101/2020.01.14.906628


26

28. Kowalko, J. E. et al. Convergence in feeding posture occurs through different genetic loci in 711

independently evolved cave populations of Astyanax mexicanus. PNAS 110, 16933–16938 712

(2013). 713

29. Kowalko, J. E. et al. Loss of schooling behavior in cavefish through sight-dependent and 714

sight-independent mechanisms. Current Biology 23, 1874–1883 (2013). 715

30. Bibliowicz, J. et al. Differences in chemosensory response between eyed and eyeless 716

Astyanax mexicanus of the Rio Subterráneo cave. EvoDevo 4, 25 (2013). 717

31. Elipot, Y., Hinaux, H., Callebert, J. & Rétaux, S. Evolutionary shift from fighting to foraging 718

in blind cavefish through changes in the serotonin network. Current Biology 23, 1–10 719

(2013). 720

32. Elipot, Y. et al. A mutation in the enzyme monoamine oxidase explains part of the Astyanax 721

cavefish behavioural syndrome. Nat Commun 5, 1–11 (2014). 722

33. Aspiras, A. C., Rohner, N., Martineau, B., Borowsky, R. L. & Tabin, C. J. Melanocortin 4 723

receptor mutations contribute to the adaptation of cavefish to nutrient-poor conditions. PNAS 724

112, 9668–9673 (2015). 725

34. Jaggard, J. et al. The lateral line confers evolutionarily derived sleep loss in the Mexican 726

cavefish. Journal of Experimental Biology 220, 284–293 (2017). 727

35. Jaggard, J. B. et al. Hypocretin underlies the evolution of sleep loss in the Mexican cavefish. 728

eLife 7, 22 (2018). 729

36. Beale, A. et al. Circadian rhythms in Mexican blind cavefish Astyanax mexicanus in the lab 730

and in the field. Nat Commun 4, 1–10 (2013). 731

37. Moran, D., Softley, R. & Warrant, E. J. Eyeless Mexican Cavefish save energy by 732

eliminating the circadian rhythm in metabolism. PLOS ONE 9, e107877 (2014). 733


https://doi.org/10.1101/2020.01.14.906628


27

38. Yoshizawa, M. et al. Distinct genetic architecture underlies the emergence of sleep loss and 734

prey-seeking behavior in the Mexican cavefish. BMC Biology 13, 15 (2015). 735

39. Erckens, W. & Martin, W. Exogenous and endogenous control of swimming activity in 736

Astyanax mexicanus (Characidae, Pisces) by direct light response and by circadian oscillator 737

I. analyses of the time-control systems of an epigean river population. Zeitschrift für 738

Naturforschung C 37, 1253–1265 (1982). 739

40. Carlson, B. M. & Gross, J. B. Characterization and comparison of activity profiles exhibited 740

by the cave and surface morphotypes of the blind Mexican tetra, Astyanax mexicanus. 741

Comparative Biochemistry and Physiology Part C: Toxicology & Pharmacology 208, 114–742

129 (2018). 743

41. Carlson, B. M., Klingler, I. B., Meyer, B. J. & Gross, J. B. Genetic analysis reveals candidate 744

genes for activity QTL in the blind Mexican tetra, Astyanax mexicanus. PeerJ 6, e5189 745

(2018). 746

42. Foulkes, N. S., Whitmore, D., Vallone, D. & Bertolucci, C. Chapter One - Studying the 747

evolution of the vertebrate circadian clock: The power of fish as comparative models. in 748

Advances in Genetics (ed. Foulkes, N. S.) vol. 95 1–30 (Academic Press, 2016). 749

43. Vatine, G., Vallone, D., Gothilf, Y. & Foulkes, N. S. It’s time to swim! Zebrafish and the 750

circadian clock. FEBS Letters 585, 1485–1494 (2011). 751

44. Takahashi, J. S. Transcriptional architecture of the mammalian circadian clock. Nature 752

Reviews Genetics 18, 164–179 (2017). 753

45. Kumar, V. & Sharma, A. Common features of circadian timekeeping in diverse organisms. 754

Current Opinion in Physiology 5, 58–67 (2018). 755


https://doi.org/10.1101/2020.01.14.906628


28

46. Park, O., Roberts, T. W. & Harris, S. J. Preliminary analysis of activity of the cave crayfish, 756

Cambarus pellucidus. The American Naturalist 75, 154–171 (1941). 757

47. Jegla, T. C. & Poulson, T. L. Evidence of circadian rhythms in a cave crayfish. Journal of 758

Experimental Zoology 168, 273–282 (1968). 759

48. Trajano, E. & Menna-Barreto, L. Free-running locomotor activity rhythms in cave-dwelling 760

catfishes, Trichomycterus sp., from Brazil (Teleostei, Siluriformes). Biological Rhythm 761

Research 27, 329–335 (1996). 762

49. Koilraj, A. J., Sharma, V. K., Marimuthu, G. & Chandrashekaran, M. K. Presence of 763

circadian rhythms in the locomotor activity of a cave-dwelling millipede Glyphiulus 764

Cavernicolus Sulu (cambalidae, Spirostreptida). Chronobiology International 17, 757–765 765

(2000). 766

50. Duboué, E. R. & Borowsky, R. L. Altered rest-activity patterns evolve via circadian 767

independent mechanisms in cave adapted Balitorid loaches. PLOS ONE 7, e30868 (2012). 768

51. Hervant, F., Mathieu, J. & Durand, J. P. Metabolism and circadian rhythms of the European 769

blind cave salamander Proteus anguinus and a facultative cave dweller, the Pyrenean newt. 770

78, 6 (2000). 771

52. Cavallari, N. et al. A blind circadian clock in cavefish reveals that opsins mediate peripheral 772

clock photoreception. PLOS Biology 9, e1001142 (2011). 773

53. Herman, A. et al. The role of gene flow in rapid and repeated evolution of cave-related traits 774

in Mexican tetra, Astyanax mexicanus. Molecular Ecology 27, 4397–4416 (2018). 775

54. Hughes, M. E., Hogenesch, J. B. & Kornacker, K. JTK_CYCLE: An efficient nonparametric 776

algorithm for detecting rhythmic components in genome-scale data sets. J Biol Rhythms 25, 777

372–380 (2010). 778


https://doi.org/10.1101/2020.01.14.906628


29

55. Li, Y., Li, G., Wang, H., Du, J. & Yan, J. Analysis of a gene regulatory cascade mediating 779

circadian rhythm in zebrafish. PLOS Computational Biology 9, e1002940 (2013). 780

56. Boyle, G. et al. Comparative analysis of vertebrate diurnal/circadian transcriptomes. PLOS 781

ONE 12, e0169923 (2017). 782

57. Shi, S. et al. Circadian clock gene Bmal1 is not essential; functional replacement with its 783

paralog, Bmal2. Current Biology 20, 316–321 (2010). 784

58. Kuintzle, R. C. et al. Circadian deep sequencing reveals stress-response genes that adopt 785

robust rhythmic expression during aging. Nat Commun 8, 1–10 (2017). 786

59. Gothilf, Y. et al. Zebrafish serotonin N-Acetyltransferase-2: marker for development of 787

pineal photoreceptors and circadian clock function. Endocrinology 140, 4895–4903 (1999). 788

60. Klein, D. C. Arylalkylamine N-Acetyltransferase: “the timezyme”. J. Biol. Chem. 282, 789

4233–4237 (2007). 790

61. Gandhi, A. V., Mosser, E. A., Oikonomou, G. & Prober, D. A. Melatonin is required for the 791

circadian regulation of sleep. Neuron 85, 1193–1199 (2015). 792

62. Tovin, A. et al. Systematic identification of rhythmic genes reveals camk1gb as a new 793

element in the circadian clockwork. PLOS Genetics 8, e1003116 (2012). 794

63. Doherty, C. J. & Kay, S. A. Circadian control of global gene expression patterns. Annual 795

Review of Genetics 44, 419–444 (2010). 796

64. Hughes, L. C. et al. Comprehensive phylogeny of ray-finned fishes (Actinopterygii) based on 797

transcriptomic and genomic data. Proc Natl Acad Sci U S A 115, 6249–6254 (2018). 798

65. Ueda, H. R. et al. System-level identification of transcriptional circuits underlying 799

mammalian circadian clocks. Nat Genet 37, 187–192 (2005). 800


https://doi.org/10.1101/2020.01.14.906628


30

66. Rey, G. et al. Genome-wide and phase-specific DNA-binding rhythms of BMAL1 control 801

circadian output functions in mouse liver. PLOS Biology 9, e1000595 (2011). 802

67. Preitner, N. et al. The orphan nuclear receptor REV-ERBα controls circadian transcription 803

within the positive limb of the mammalian circadian oscillator. Cell 110, 251–260 (2002). 804

68. Whitmore, D., Foulkes, N. S., Strähle, U. & Sassone-Corsi, P. Zebrafish Clock rhythmic 805

expression reveals independent peripheral circadian oscillators. Nat Neurosci 1, 701–707 806

(1998). 807

69. Whitmore, D., Foulkes, N. S. & Sassone-Corsi, P. Light acts directly on organs and cells in 808

culture to set the vertebrate circadian clock. Nature 404, 87–91 (2000). 809

70. Frøland Steindal, I. A. & Whitmore, D. Circadian clocks in fish—What have we learned so 810

far? Biology (Basel) 8, (2019). 811

71. Diessler, S. et al. A systems genetics resource and analysis of sleep regulation in the mouse. 812

PLOS Biology 16, e2005750 (2018). 813

72. Wang, F. et al. RNAscope. J Mol Diagn 14, 22–29 (2012). 814

73. Moore, H. A. & Whitmore, D. Circadian rhythmicity and light sensitivity of the zebrafish 815

brain. PLOS ONE 9, e86176 (2014). 816

74. Duboué, E. R., Borowsky, R. L. & Keene, A. C. β-Adrenergic signaling regulates 817

evolutionarily derived sleep loss in the Mexican Cavefish. BBE 80, 233–243 (2012). 818

75. Laposky, A. et al. Deletion of the mammalian circadian clock gene BMAL1/Mop3 alters 819

baseline sleep architecture and the response to sleep deprivation. Sleep 28, 395–409 (2005). 820

76. Ehlen, J. C. et al. Bmal1 function in skeletal muscle regulates sleep. eLife 6, e26557 (2017). 821

77. Qiu, P. et al. BMAL1 knockout macaque monkeys display reduced sleep and psychiatric 822

disorders. Natl Sci Rev 6, 87–100 (2019). 823


https://doi.org/10.1101/2020.01.14.906628


31

78. Bloch, G., Barnes, B. M., Gerkema, M. P. & Helm, B. Animal activity around the clock with 824

no overt circadian rhythms: patterns, mechanisms and adaptive value. Proceedings of the 825

Royal Society B: Biological Sciences 280, 20130019 (2013). 826

79. Zhao, H. et al. Modulation of DNA repair systems in blind cavefish during evolution in 827

constant darkness. Current Biology 28, 3229-3243.e4 (2018). 828

80. Ziv, L. & Gothilf, Y. Circadian time-keeping during early stages of development. PNAS 103, 829

4146–4151 (2006). 830

81. Ben-Moshe, Z. et al. The light-induced transcriptome of the zebrafish pineal gland reveals 831

complex regulation of the circadian clockwork by light. Nucleic Acids Res 42, 3750–3767 832

(2014). 833

82. Falcón, J. et al. Regulation of Arylalkylamine N-Acetyltransferase-2 (AANAT2, EC 834

2.3.1.87) in the fish pineal organ: evidence for a role of proteasomal proteolysis. 835

Endocrinology 142, 1804–1813 (2001). 836

83. Dekens, M. P. S. et al. Light regulates the cell cycle in zebrafish. Current Biology 13, 2051–837

2057 (2003). 838

84. Tamai, T. K., Vardhanabhuti, V., Foulkes, N. S. & Whitmore, D. Early embryonic light 839

detection improves survival. Current Biology 14, R104–R105 (2004). 840

85. Hirayama, J. et al. Common light signaling pathways controlling DNA repair and circadian 841

clock entrainment in zebrafish. Cell Cycle 8, 2794–2801 (2009). 842

86. Tamai, T. K., Young, L. C., Cox, C. A. & Whitmore, D. Light Acts on the Zebrafish 843

Circadian Clock to Suppress Rhythmic Mitosis and Cell Proliferation. J Biol Rhythms 27, 844

226–236 (2012). 845


https://doi.org/10.1101/2020.01.14.906628


32

87. Passow, C. N. et al. RNAlater and flash freezing storage methods nonrandomly influence 846

observed gene expression in RNAseq experiments. bioRxiv 379834 (2018) 847

doi:10.1101/379834. 848

88. Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina 849

sequence data. Bioinformatics 30, 2114–2120 (2014). 850

89. Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15–21 851

(2013). 852

90. Pertea, M., Kim, D., Pertea, G. M., Leek, J. T. & Salzberg, S. L. Transcript-level expression 853

analysis of RNA-seq experiments with HISAT, StringTie and Ballgown. Nature Protocols 854

11, 1650–1667 (2016). 855

91. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for 856

RNA-seq data with DESeq2. Genome Biology 15, 550 (2014). 857

92. Watanabe, T. et al. Genetic and molecular analysis of wild-derived arrhythmic mice. PLOS 858

ONE 4, e4301 (2009). 859

93. Bailey, T. L., Johnson, J., Grant, C. E. & Noble, W. S. The MEME Suite. Nucleic Acids Res 860

43, W39–W49 (2015). 861

94. Xiong, S., Krishnan, J., Peuß, R. & Rohner, N. Early adipogenesis contributes to excess fat 862

accumulation in cave populations of Astyanax mexicanus. Developmental Biology 441, 297–863

304 (2018). 864

95. Wang, Y. et al. Alkaline phosphatase-based chromogenic and fluorescence detection method 865

for basescopeTM In Situ hybridization. Journal of Histotechnology 0, 1–9 (2019). 866

96. Guo, L. et al. An adaptable chromosome preparation methodology for use in invertebrate 867

research organisms. BMC Biol 16, 25 (2018). 868


https://doi.org/10.1101/2020.01.14.906628


33

97. Schindelin, J. et al. Fiji: an open-source platform for biological-image analysis. Nat Methods 869

9, 676–682 (2012). 870

98. Preibisch, S., Saalfeld, S. & Tomancak, P. Globally optimal stitching of tiled 3D microscopic 871

image acquisitions. Bioinformatics 25, 1463–1465 (2009). 872

99. Labun, K. et al. CHOPCHOP v3: expanding the CRISPR web toolbox beyond genome 873

editing. Nucleic Acids Res 47, W171–W174 (2019). 874

100. Varshney, G. K. et al. High-throughput gene targeting and phenotyping in zebrafish using 875

CRISPR/Cas9. Genome Res. 25, 1030–1042 (2015). 876

101. Kowalko, J. E., Ma, L. & Jeffery, W. R. Genome editing in Astyanax mexicanus using 877

transcription activator-like effector nucleases (TALENs). JoVE (Journal of Visualized 878

Experiments) e54113 (2016) doi:10.3791/54113. 879

102. Stahl, B. A. et al. Manipulation of gene function in Mexican Cavefish. JoVE (Journal of 880

Visualized Experiments) e59093 (2019) doi:10.3791/59093. 881

103. Klaassen, H., Wang, Y., Adamski, K., Rohner, N. & Kowalko, J. E. CRISPR 882

mutagenesis confirms the role of oca2 in melanin pigmentation in Astyanax mexicanus. 883

Developmental Biology 441, 313–318 (2018). 884

104. Jao, L.-E., Wente, S. R. & Chen, W. Efficient multiplex biallelic zebrafish genome 885

editing using a CRISPR nuclease system. PNAS 110, 13904–13909 (2013). 886

105. Cermakian, N. & Sassone-Corsi, P. Multilevel regulation of the circadian clock. Nature 887

Reviews Molecular Cell Biology 1, 59–67 (2000). 888

106. Kobayashi, Y. et al. Molecular analysis of zebrafish photolyase/cryptochrome family: 889

two types of cryptochromes present in zebrafish. Genes to Cells 5, 725–738 (2000). 890


https://doi.org/10.1101/2020.01.14.906628


34

107. Takeda, Y., Jothi, R., Birault, V. & Jetten, A. M. RORγ directly regulates the circadian 891

expression of clock genes and downstream targets in vivo. Nucleic Acids Res 40, 8519–8535 892

(2012). 893

108. Honma, S. et al. Dec1 and Dec2 are regulators of the mammalian molecular clock. 894

Nature 419, 841–844 (2002). 895

109. Godinho, S. I. H. et al. The After-Hours mutant reveals a role for Fbxl3 in determining 896

mammalian circadian period. Science 316, 897–900 (2007). 897

110. Siepka, S. M. et al. Circadian mutant overtime reveals F-box protein FBXL3 regulation 898

of cryptochrome and period gene expression. Cell 129, 1011–1023 (2007). 899

111. Mitsui, S., Yamaguchi, S., Matsuo, T., Ishida, Y. & Okamura, H. Antagonistic role of 900

E4BP4 and PAR proteins in the circadian oscillatory mechanism. Genes Dev. 15, 995–1006 901

(2001). 902

112. Zhao, W.-N. et al. CIPC is a mammalian circadian clock protein without invertebrate 903

homologues. Nat Cell Biol 9, 268–275 (2007). 904

113. Appelbaum, L. et al. Circadian and homeostatic regulation of structural synaptic 905

plasticity in hypocretin neurons. Neuron 68, 87–98 (2010). 906

907


https://doi.org/10.1101/2020.01.14.906628


35

Table 1. Known circadian regulators with losses or reductions in rhythmicity in cave 908 populations1 909

Gene name Populations Predicted functions of transcribed proteins

Core loop

arntl1a Tinaja, Pachón Activator in core circadian feedback loop105

arntl1b Tinaja, Pachón Activator in core circadian feedback loop105

arntl2 Tinaja, Molino, Pachón Activator in core circadian feedback loop105

cry1aa Pachón Repressor in core circadian feedback loop106

cry1ba Tinaja, Pachón Repressor in core circadian feedback loop106

cry1bb Tinaja, Pachón Repressor in core circadian feedback loop106

cry4 Pachón Repressor in core circadian feedback loop106

Accessory pathways

rorca Molino, Tinaja, Pachón Regulator of core feedback loop107

rorcb Molino, Tinaja, Pachón Regulator of core feedback loop107

bhlhe40 Molino Regulator of core feedback loop108

bhlhe41 Tinaja, Pachón Regulator of core feedback loop108

fbxl3a Molino, Tinaja, Pachón Regulator of core feedback loop109,110

nfil3 Tinaja, Pachón Regulator of core feedback loop111

nfil3-5 Tinaja, Pachón Regulator of core feedback loop55

cipcb Pachón Regulator of core feedback loop112

Mediators of downstream physiology and behavior

aanat2 Molino, Tinaja, Pachón Controls daily changes in melatonin production61

camk1gb Molino, Tinaja, Pachón Involved in linking pineal master clock with peripheral circadian activity62

nptx2b Pachón, Molino Modulates circadian synaptic changes113 1Exhaustive list of genes with changes in rhythmicity can be found in File S1. 910 911


https://doi.org/10.1101/2020.01.14.906628


36

Figure 1. A. Overlap of genes with rhythmic expression between populations. B. Heatmap of 912 genes with rhythmic patterns in surface fish (Río Choy), ordered by gene phase, compared to 913 expression in cave populations. Each column represents gene expression at a single-time-point, 914 sampled every four hours from 0-20 hours. Redder boxes correspond to higher expression. C. 915 Identifying genes with changes in rhythmicity between cave and surface populations. Genes with 916 greater amplitude values have larger differences between their expression peak and trough, 917 where genes with greater periodicity show stronger cyclical oscillation patterns (see Methods). 918 Genes are colored based on their SDR q-values. Genes with positive values for both show 919 increased rhythmicity in the surface population, where genes with negative values show 920 increased rhythmicity in cave populations. 921 922

923 924

V2

V3

V4

V5

V6

V7

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

Pachon TinajaSurface

Gene

s rhy

thm

ic in

surfa

ce

0CT

28

0

40

20

8

19

1

2

446

126 3

4

267

25

Choy Tinaja

Molino Pachon

A B

C

−10

−5

0

5

10

−10 −5 0 5 10Z_Per

Z_AM

P

0.25

0.50

0.75

1.00p.adj.SDR

Molino

Surface

−5

0

5

10

−10 −5 0 5 10 15Z_Per

Z_AM

P

0.25

0.50

0.75

1.00p.adj.SDR

−5

0

5

10

−10 −5 0 5 10Z_Per

Z_AM

P

0.25

0.50

0.75

1.00p.adj.SDR

Differential amplitude z-score

Diffe

rent

ial p

erio

dicit

y z-

scor

e

FDR

Surface Surface

PachonTinaja

Diffe

rent

ial p

erio

dicit

y z-

scor

e

Diffe

rent

ial p

erio

dicit

y z-

scor

e

00 0

10 10 10

-10

-5-5

-10 -10 -1015 100 0 010

Differential amplitude z-score Differential amplitude z-score

V2

V3

V4

V5

V6

V7


20 V2

V3

V4

V5

V6

V7


V2

V3

V4

V5

V6

V7


Molino

0CT

20 0CT

20CT0 20

Surface


https://doi.org/10.1101/2020.01.14.906628


37

Figure 2. A. Key circadian genes with changes in rhythmicity in cave populations (see Fig. S17 925 for all core circadian genes with changes in rhythmicity). Colored lines represent a loess 926 regression of gene expression through time for each population. B. Simplified schematic of the 927 circadian feedback loops based on proposed interactions in zebrafish43,45,55. Grey boxes indicate 928 genes that are either arrhythmic or show significantly reduced rhythmicity between cave and 929 surface. White boxed genes do not show significant differences between cave and surface. 930 Highlighted in yellow is the core loop. Bright yellow circles represent regulating protein 931 complexes. Red lines indicate negative regulation, black lines indicate positive regulation. Genes 932 that are not boxed did not show evidence of rhythmic expression in any cave or surface 933 population. Dotted lines are for visual clarity. Genes without an annotated ortholog in cavefish 934 were not included in the schematic. 935

936 937

No

rma

lize

d e

xpre

ssio

n

cry1aa

cry1ba

cry1bb

cry4

cry5

bhlhe40

bhlhe41

arntl1a

arntl1b

arntl2

clock1a

clock1b

BMALCLOCK

CRY PER

per1a

per1b

per2

per3

rorca rorcb

nfil3

nfil3-5

nfil3-6

cipca

cipcb

tefa

tefb

dbpb

nr1d1

Arrhythmic or differentiated in all

caves

Arrhythmic or differentiated in 2 cave

populations

Arrhythmic or differentiated in 1 cave

population

cry1ab

A

B

Surface

Molino

Pachón

Tinaja

0

30

60

90

0 5 10 15 20

popCHOY

MOLINO

PACHON

TINAJA

arntl1a

0

200

400

600

0 5 10 15 20

nfil3

popCHOY

MOLINO

PACHON

TINAJA

nfil3

0

10

20

30

0 5 10 15 20

popCHOY

MOLINO

PACHON

TINAJA

0

20

40

60

80

0 5 10 15 20

popCHOY

MOLINO

PACHON

TINAJA

rorca

100

200

300

400

500

0 5 10 15 20

popCHOY

MOLINO

PACHON

TINAJA

cipcbarntl2

No

rma

lize

d e

xpre

ssio

n

Time

0

20

40

60

0 5 10 15 20

popCHOY

MOLINO

PACHON

TINAJA

cry1bb

0

25

50

75

0 5 10 15 20

popCHOY

MOLINO

PACHON

TINAJA

rorcb

0

100

200

300

400

500

0 5 10 15 20

popCHOY

MOLINO

PACHON

TINAJA

aanat2


https://doi.org/10.1101/2020.01.14.906628


38

Figure 3. A. Phase (e.g., peak expression time) distribution of rhythmic transcripts in each 938 population. B. Cycling genes on average show a delay in phase in cave populations relative to 939 their phase in the surface population. 940 941

942

Num

ber o

f gen

es

0

25

50

75

100

0 5 10 15 20V2

count

Surface100

75

50

25

00 2010

0

5

10

15

0 5 10 15 20V2

count

Molino15

10

5

00 2010

A

Phase time

0

50

100

150

0 5 10 15 20V2

count

Tinaja

150

50

100

0

0 2010

0

20

40

0 5 10 15 20V2

count

Pachón

40

20

0

0 2010

●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●●

●

●

●

●●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●●

●

●

●

●●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●●

●

●

●

●●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●

●●

●●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●

●●

●●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●

●●

●●

●

●

●

●●

●

●

●

●

●

−20

−10

0

10

20

Molino−choy Pachon−choy Tina−choyV2

V1

Dela

y in

cave

Dela

y in

surfa

ce

Molino-Surface

Pachón-Surface

Tinaja-Surface

0

10

-10

B


https://doi.org/10.1101/2020.01.14.906628


39

Figure 4. Temporal expression patterns of per1a and arntl1a in brain and liver tissue in 943 Astyanax mexicanus populations. In-situ staining of per1a (cyan) and arntl1a (magenta) using 944 RNAscope® in the midbrain (‘B’, top panels for each timepoint) and liver (‘L’, bottom panels 945 for each timepoint) of surface fish and cavefish (Pachón, Tinaja, Molino) at CT0, CT8, and 946 CT16. Each time point is a single fish sample with probes separated into two channels. Images 947 are representative sections of two fish collected per time point, per population. Scale bar is 948 25μM. 949

950

951 952 953


https://doi.org/10.1101/2020.01.14.906628


40

Figure 5. Temporal expression patterns of rorca and rorcb in midbrain and liver tissue in 954 Astyanax mexicanus populations. In-situ staining of rorca (cyan) and rorcb (magenta) using 955 RNAscope® in midbrain (‘B’, top panels for each timepoint) and liver (‘L’, bottom panels for 956 each timepoint) of Surface fish and cavefish (Pachón, Tinaja, Molino) at CT0, CT8, and CT16. 957 Each time point is a single fish sample. Images are representative sections of two fish collected 958 per time point, per population. Scale bar is 25μM. 959 960

961 962


https://doi.org/10.1101/2020.01.14.906628


41

Figure 6. Mutant aanat2 and rorca fish reveal a role for these genes in sleep behavior in A. 963 mexicanus. A. Total sleep is not significantly altered between control and aanat2 crispant surface 964 fish (unpaired t-test, p=0.99). B-C. Day sleep is not significantly altered between WT and 965 crispant aanat2 fish (unpaired t-test, p=0.55). Night sleep is significantly reduced in aanat2 966 crispants compared to WT controls (unpaired t-test, p<0.0001). C. Total sleep is significantly 967 reduced in crispant rorca fish compared to WT controls (unpaired t-test, p<0.0001). E-F. Day 968 sleep was not significantly altered between WT and rorca crispants (unpaired t-test, p=0.056). 969 Night sleep is significantly reduced in rorca crispants compared to WT controls (unpaired t-test, 970 p<0.0001). 971 972 973 974

975

N.S.

Surface WT

Surface WT Surface rorca

Surface rorca

A B C

Slee

p du

ratio

n (h

rs/2

4hrs

)Sl

eep

dura

tion

(hrs

/24h

rs)

Min

. sle

ep/h

our

Min

. sle

ep/h

our

Min

. sle

ep/h

our

Min

. sle

ep/h

our

D E F


https://doi.org/10.1101/2020.01.14.906628


author/funder. it is made available under a cc-by-nc-nd 4 ... · 34 predictable changes in their...

Documents