Adipose tissue mitochondrial respiration in Atlantic 1 salmon: implications for sex-dependent life-history 2 variation 3

38 Adipose tissue is essential for energy homeostasis, with mitochondria having a central role in its 39 function. Mitochondria-mediated adipose tissue dysfunction has been linked to several 40 metabolic disorders in humans but surprisingly little is known about whether variation in adipose 41 tissue processes, such as mitochondrial function, is maintained in wild animal populations, how 42 it is related to reproductive success, and its evolutionary significance. Early sexual maturation 43 (age-at-maturity) in Atlantic salmon ( Salmo salar ) is promoted by higher adiposity and has a 44 strong genetic association with the vestigial like 3 ( vgll3 ) locus. Vgll3 is also linked to the control 45 of adipose tissue growth and pubertal timing in mammals. This makes Atlantic salmon ideal for 46 studying whether functional variation in adipose tissue processes is linked to reproductive 47 timing. Here, using 16 Atlantic salmon individuals of different sexes reared in common-garden 48 conditions, we conducted a proof-of-principle study and tested whether mitochondrial respiration 49 in adipose tissue, mitochondrial DNA (mtDNA) amount, and adipocyte size are associated with 50 sex and vgll3 genotype. Mitochondrial respiration was quantified from freshly collected white 51 adipose tissue using high-resolution respirometry, and mtDNA amounts were quantified relative 52 to nuclear DNA amounts. We found differences in mtDNA amount between sexes, plausibly 53 caused by a confounding effect of maturation status, but without corresponding differences in 54 mitochondrial respiration. Mitochondrial respiration, leak respiration, and coupling capacity (P/E 55 ratio) were marginally lower in immature females carrying the vgll3 early maturation genotype 56 compared to those with the late maturation genotype, although larger sample sizes are required 57 to validate this difference. Our results generate new hypotheses on how coupling capacity of 58 oxidative phosphorylation could be linked with the timing of maturation via adiposity as well as 59 pave the way to study the mechanistic relationships between life-history variation and 60 mitochondrial bioenergetics in wild populations. (Succ) : 1 CI / CI&CII, indicating the proportion respiration from CI & CII respiration; efficiency (CoupEff) : (CI&CII - Leak Omy) / CI&CII, indicating the proportion of synthesis -linked respiration from CI & CII respiration; Coupling control ratio (L/P): Leak CI / CI&CII, indicating the proportion of CI-linked proton leak from CI & CII respiration; and Coupling capacity (P/E) : CI&CII / max CI&CII, indicating the proportion of phosphorylating, coupled CI & CII mediated respiration from non-phosphorylating, uncoupled CI & CII mediated respiration.

complexes I-IV, ubiquinone (Q) and cytochrome c, and finally to the electron acceptor oxygen 83 ( Fig. 1a). Simultaneously, complexes I, III, and IV pump protons into the mitochondrial 84 intermembrane space, generating a proton gradient that drives the phosphorylation of ADP to 85 ATP by the ATP synthase enzyme. The efficiency of mitochondrial respiration is determined by 86 the proportion of electrons in OXPHOS that are coupled to generating ATP (coupled 87 respiration), versus proton leakage and electron slippage across the inner membrane 88 (uncoupled respiration) (Fig 1a). Intact tissues exhibit both coupled and uncoupled OXPHOS, 89 whereby highly efficient mitochondria are tightly coupled (9). 90 91 Energy metabolism is central to diversity within species; variation in the organisation and 92 efficiency of mitochondria among individuals is likely under selection and adaptive (10, 11). 93 Within species, how variation in mitochondrial function could lead to variation in other traits can 94 be better understood by measuring the different stages and efficiency of mitochondrial 95 respiration (see e.g., (10). For example, the ability to increase ATP synthesis, which requires 96 sufficient reserve capacity in the electron transport pathway, allows organisms to respond via 97 mitochondria to changes in energetic demand and environmental stressors (12, 13). Likewise, 98 although proton leak reduces the efficiency of mitochondria, it may restrict the production of 99 reactive oxygen species and may thereby limit oxidative stress (14). Variation in mitochondrial 100 processes may thus allow individuals to respond differently to energetic demand and stressors. 101 Despite the central role of mitochondria in the control of adipose tissue function, little is known 102 of how adipose tissue mitochondrial activity relates to growth and body condition and affects 103 life-history traits, such as the timing of sexual maturation (15). The fact that obesity in humans is 104 associated with a significant decline in adipose tissue mitochondrial respiration (6, 16) also 105 makes it appealing to study adipose tissue mitochondria in relation to life-history decisions in 106 other species. 107 108 Critical time points of an organism's life cycle constitute life-history traits, the variation in which 109 is largely shaped by energy allocation differences and maintained within species via 110 evolutionary trade-offs (17). In Atlantic salmon (Salmo salar), earlier maturation shortens 111 generation time and increases the survival probability prior to reproduction compared to delayed 112 maturation, but this comes at the expense of smaller size at maturity, which is associated with 113 lower fecundity (18). The Atlantic salmon is an emerging wild model species to study the 114 energetic basis of life-history adaptations for two main reasons (15,19). First, faster 115 accumulation of adipose tissue, quantified as a high condition factor, is closely related to earlier 116 sexual maturation in salmon, and higher fat content has been found in juvenile males compared 117 to females before male early maturation occurs (20-22). Second, a single genomic region 118 explains a substantial amount of variation in age-at-maturity (23). The strongest candidate gene 119 in this region, vgll3 (24), is likely important for adipose tissue growth and function; in mice, 120 expression of vgll3 was negatively correlated with adipose tissue mass and body weight (25), 121 and in humans, variation in the VGLL3 locus is associated with the timing of puberty (26, 27). 122 Finally, vgll3 early maturation genotype is linked to higher body condition in salmon (20). 123 Therefore, adipose tissue processes could provide a functional basis of variation in age-at-124 maturity, and subsequently, potential evolutionary constraints (15,19). 125

126
In this study, we first measure mitochondrial quantity and respiration as well as adipocyte 127 morphology in the visceral adipose tissue of Atlantic salmonas fishes are not known to have 128 brown adipose tissue, the focus is on white adipose tissue (28). We then investigate sex-and 129 vgll3 genotype -dependent differences in mitochondrial respiration to obtain preliminary insights 130 into whether mitochondrial function is a potential cellular pathway underlying life-history 131 variation in salmon. Samples of adipose tissue were collected from fish reared as part of another study during 137 August-September 2020 and were approximately 2 years 8 months post hatch (average mass 138 1kg (Table 2)). Details of fish rearing, feeding, and temperatures are shown in (29) until Feb 139 2020 (in Feb-Aug 2020, conditions largely followed those in 2019). We collected samples from 140 Neva population individuals from the high temperature treatment. The mean (±SD) temperature 141 during the sampling period was 11.7 ± 0.8°C. 142

143
The fish were fasted for ~57 hours prior to sampling. One day prior to sampling, the individuals, 144 which had been previously tagged with passive integrated transponders (PIT-tags) and 145 identified for vgll3 genotype and sex, were selected for sampling, anaesthetised with buffered 146 tricaine methanosulfonate (MS-222, 0.125g/L, sodium bicarbonate buffered), and measured for 147 body mass (to the nearest 0.1g) and fork length (to the nearest mm). After measurement, the 148 fish were placed in a floating cage held inside the rearing tank until sampling the following day. 149 The sampling was balanced in terms of sex and vgll3 genotypes across four sampling days 150 (from in total four tanks), on which four individuals (one female and one male of both EE and LL 151 genotypes, referring to homozygous early and late maturation genotypes, respectively) from 152 within the same rearing tank were captured by netting each day and euthanized with an 153 overdose of MS-222 (0.250g/L, sodium bicarbonate buffered). Fish were sampled from a 154 different rearing tank between 8:40 and 11:40 AM on each day of sampling; three tanks had 155 been reared with normal feed used in Atlantic salmon aquaculture (control), and one tank with a 156 low-fat feed diet (details in (29)). Visceral adipose tissue samples were collected using sterilized  We calculated mitochondrial respiration coefficients from tissue mass -normalised parameters 221 shown in Table 1  Next, quantitative real-time PCR (qPCR) was performed using a double-stranded DNA-binding 257 dye as a reporter (HOT FIREPol EvaGreen qPCR Supermix, Solis Biodyne), in a 384-plate 258 format using a Biorad CFX384 C1000 thermal cycler. We targeted two mitochondrial (16s and 259 cytb), and two nuclear (app, EF1a) genomic regions using a double-stranded DNA-binding dye 260 (eva green) as a reporter. Primers specific to these genomic regions were designed with Primer-261 BLAST (34), where the melting temperature (Tm) and product size were adjusted to 59-61°C, 262 and 90-150 bp, respectively (Table S1). The slides were imaged using 20x magnification with extended plane, using 3DHISTECH 291 Pannoramic 250 FLASH II digital slide scanner. Images were inspected in Qupath v. 0.3.0 (37) 292 and 1-2 regions containing visceral adipose tissue from each image were saved in tiff format. 293 Tiff-images were analysed in ImageJ to measure the size of particles with a size of 1000-294 100000 and circularity of 0.2-1.00 as size and shape filters (used macro in Appendix S2). The 295 selected cells were visually inspected to exclude selections that were merged of multiple cells, 296 were stretched (during sectioning) or had a very irregular shape because of excess dye. The 297 remaining selected cells were measured (area in µm 2 ) (see Fig. 2a for a representative image). 298 The number of cells identified with this method ranged from 40 to 306 between individuals due 299 to the variable size and quality of the sections. However, the number of measured cells was not 300 correlated with mean (Spearman-rho = -0.24, p = 0.426) or median (Spearman-rho = -0.26, p = 301 0.388) adipocyte size, indicating that there was no size bias due to the number of cells 302 measured (Fig. S2). 303 304

Data analysis 305
Respiration was successfully measured from adipose tissues of 14 individuals (nine females, 306 five males) after data from two males, both with the early maturation vgll3 genotype, were 307 omitted due to abnormal noise in the data. Leak CI was not determined for three individuals due 308 to abnormal fluxes. Cryosections for adipocyte size measurements were obtained from 14 309 individuals (seven of each sex, after two females were omitted from the analysis due to low 310 section quality). The males whose respiration data were excluded were included in adipocyte 311 size measurements. For functional coherence, and to avoid biases in the female data due to 312 different maturation status, we excluded data from the one female that displayed more 313 advanced maturation than all others, because salmon consume the energy stored in adipose 314 tissue during maturation (22). Given that all the remaining females in this study were immature, 315 and all males were maturing, sex effects are confounded with maturation status throughout the 316 study. The size, genotypes and respiration data for each individual are provided in Table S2. 317 318 Oxygen fluxes were normalised to tissue mass as well as to both tissue mass and mtDNA 319 amount. To compare the respiration data of individuals with different vgll3 genotypes, we 320 focussed on data from females, as we only obtained data from two early maturation (vgll3*EE) -321 genotype males. 322

323
The data were analysed in the R software environment (38), and visualised using ggplot2 (39). 324 We tested the statistical significance of feed treatment and sex effects and of genotype effects 325 in females using non-parametric Wilcoxon rank sum tests and Spearman's rank correlations to 326 avoid violations of linear model assumptions due to low sample sizes. In line with the 327 exploratory nature of this study, multiple-test correction, which is also restrictively conservative 328 at small samples sizes, was not employed. were detected on fish morphology (Table 2) or respiration traits (see Fig. 2b and Fig. 3-4, where 335 feed treatments are shown with different symbols, and Table S3) except adipocytes were 336 significantly larger in salmon reared under a high-fat diet than a low-fat diet, as expected (Table  337 2). Data from the two feeding treatments were combined to improve statistical power of the 338 subsequent analyses. 339 340 Table 2. Summary of final data sets and fish phenotypes (mean ± SD) from control (C) and low fat (LF) 341 feed treatments. All differences between treatments were non-significant, apart from adipocyte size 342 (results shown in footnotes). Resp = mitochondrial respiration, hist = cryohistology.   Atlantic salmon females and males (nfem = 6, nmale = 7) (Wilcoxon rank sum test,

Sex differences 354
All males had enlarged gonads indicating maturation had been initiated, but females were 355 immature (see also Materials and methods). Males had a higher relative mtDNA amount than 356 females (Table 2; Fig. 3a), but mitochondrial respiration did not differ between the sexes after 357 the data were normalised either with tissue mass, or tissue mass and relative mtDNA amount 358 (Table 3; Fig. 4a-b). This result was corroborated also by a lack of difference between the sexes 359 in mitochondrial respiratory coefficients (Fig. S3). Interestingly, in both sexes CI-mediated 360 respiration (NADH pathway cofactor) was substantially higher than CII-mediated respiration 361 (succinate pathway cofactor), suggesting a generally higher preference for NADH-driven 362 electron transfer in salmon visceral adipose tissue. There were no sex differences between any 363 other morphological phenotypes measured nor for adipocyte size (Table 2; Fig. 2b).

Vgll3 effects on mitochondrial traits 386
To explore the relationship between vgll3 genotypes and mitochondrial respiration, we focussed 387 only on female salmon (Table 4) due to the limited availability of data from males with vgll3*EE 388 genotype. We found that CI-and CI&CII -mediated respiration were significantly higher in 389 individuals with the late maturation genotype compared to those with the early maturation 390 genotype of vgll3 (p = 0.036; Table 3; Fig. 4c). In line with the higher respiration, CI&CII -linked 391 respiratory leak (Leak Omy) was significantly elevated in individuals with the late maturation 392 genotype (p = 0.036; Table 3; Fig. 4c). Consequently, there was no difference between the 393 genotypes in ATP synthesis -linked respiration rate, i.e., Leak Omy subtracted from CI&CII 394 Wilcoxon rank sum test W = 4, p = 0.371). We then asked whether the genotype effects were 396 mediated by the mtDNA amount -as a proxy of mitochondrial density-and found no significant 397 differences (p = 0.551; Fig. 3b). Finally, when respiration was normalised to mtDNA amount 398 within the same individuals, the differences were insignificant in CI&CII -mediated respiration 399 but remained significant in CI-mediated respiration (p = 0.036; Table 3; Fig. 4d). There were no 400 significant effects of genotype on the other respiration traits (Table 3; Fig. 4c, d). Vgll3 genotype 401 effects were also mostly absent in respiration coefficients. However, the coupling capacity, P/E, 402 of individuals carrying the late maturation genotype was marginally higher than of those carrying 403 the early maturation genotype (p=0.071, Fig. 5

Correlations among fish phenotypes and mitochondrial respiration 419
We calculated Spearman's rank correlations between mitochondrial respiration values and fish 420 phenotypic data, including body mass, condition factor, relative mtDNA amount and adipocyte 421 size. There was a marginally significant negative correlation between mtDNA amount and CIV -422 mediated respiration (rho = -0.49, p = 0.089, Fig. S4a), but none of the other correlations were 423 significant at alpha = 0.1 (Table S4) substrates for CI and CII -mediated electron transfer, as expected. We found that visceral 443 adipose tissue mitochondria relied more on NADH-driven CI-mediated than succinate-driven 444 CII-mediated respiration (Fig. 4 & 5). Since NADH is generated via many different cellular 445 processes including glycolysis, citric acid cycle and fatty acid oxidation, the activity of these 446 cellular processes can regulate adipose tissue OXPHOS. The respiration rates were much 447 lower than what had been earlier reported in salmonids from aerobically more active tissues, 448 such as muscle or intestine (40, 41), but this was an expected outcome since adipose tissue is 449 mostly composed of lipids (42). Despite the lower rate of mitochondrial respiration in adipose 450 tissue, the L/P ratio indicated that approximately 30% of mitochondrial respiration was related to 451 proton leak, which is similar to the levels reported previously in fish intestine (41), but higher 452 than that reported in gill (43), or muscle and liver (44), though it should be noted that the 453 protocols and used in measuring Leak respiration often differ between studies. As expected, 454 CI&CII-linked respiratory leak through entire electron transport chain (Leak Omy) was higher 455 than CI-linked respiratory leak (Leak CI) in our study. This confirms that it is feasible to measure 456 both coupled and uncoupled respiration in visceral fat from Atlantic salmon. 457

458
Although we showed that about 40 mg of visceral adipose tissue, measured at 12°C, provides a 459 reasonable output as stated above, the demanding nature of the procedure due to logistic and 460 physiological complexities should be noted. For example, adipose tissue quantity varies 461 between individuals, and comparatively large tissue samples are required (previous studies 462 have used, e.g., 8 mg from gill tissue (43)). We also lost data points from the lowest respiration 463 activity, i.e., three out of 16 of our measurements were unreliable for Leak CI, which could have 464 been avoided by using a higher amount of initial material. Even more tissue would be required 465 to repeat the measurements for each biological replicate, which was not feasible here due to 466 taking samples for several analyses from the same fish, though previous studies in fish have 467 found consistent mitochondrial respiration between technical replicates (41, 43). Finally, it 468 should be noted that since the sampled tissue can be stored only up to a few hours before the 469 respiration measurements (unless a longer storage time is validated for this tissue (45)), 470 coordination between the wet-lab and the field (or rearing facilities) is demanding when research 471 is conducted on adult-sized salmonids. 472 473 Despite low power of the analyses due to low sample size, we detected significantly higher 474 CI&CII-mediated respiration and CI&CII-linked respiratory leak, and marginally higher coupling 475 capacity in immature female salmon with the vgll3 late maturation genotype than in those with 476 the early maturation genotype. These results could provide mechanistic explanations for how 477 variation in maturation timing is linked to vgll3 genotype, for which we postulate below two 478 distinct but potentially complementary hypotheses. 479 480 The first hypothesis concerns resource allocation to and from the lipid deposits in adipose 481 tissue. Maturation in Atlantic salmon is a physiological trait with a genetic threshold mediated by 482 condition factor possibly via lipid accumulation (46). Concordantly, a previous study (20) found 483 that vgll3-associated early maturation in males was mediated by a higher condition factor. In line 484 with these previous findings, and with our results, we hypothesise that the higher mitochondrial 485 respiration in the adipose tissue of immature salmon with a late maturation genotype leads to 486 reduced lipid storage, contributing to delayed maturation. Mitochondrial respiration typically 487 increases during fasting -a state that is characterized by active catabolic metabolism. It is 488 therefore tempting to speculate that the mitochondrial phenotype of the vgll3 late maturation 489 genotype could indicate active catabolic metabolism in adipose tissue. The high mitochondrial 490 respiratory capacity could enhance the oxidation of energy substrates and subsequently reduce 491 the size of adipose tissue depots in salmon carrying the vgll3 late maturation genotype 492 compared to the early maturation genotype. In line with this hypothesis, increased mitochondrial 493 fatty acid oxidation in adipose tissue has been observed to lead to a lean phenotype in mice 494 (47), and reduced mitochondrial respiration is related to adipocyte hypertrophy in a cell line (48). 495 Further, a negative correlation between adiposity (body mass index, BMI) and respiration of 496 isolated mitochondria from adipose tissue was observed in humans (49). In our study, neither 497 genotype nor mitochondrial respiration was associated with body condition of fish (analogous to 498 BMI). A lack of vgll3 genotype effect on body condition in salmon was also found in the same 499 cohort as our study fish earlier (29). However, these results do not contradict our hypothesis 500 since body condition effects may be manifested at a different time or life-stage (20), or 501 alternatively, was not observed due to low statistical power. For instance, salmon with the late 502 maturation genotype could be burning their adipose tissue at a higher rate during the winter 503 (differences in lipid utilisation have also been shown in relation to migration in juvenile salmon 504 (50)), which would result in faster depletion of energy reserves compared to the early maturation 505 genotype. Subsequently, the depletion of lipid reserve could delay maturation because the 506 amount of adipose tissue in the spring is an important determinant of salmon maturation 507 probability the following autumn (22). 508

509
The second hypothesis we propose is based on the finding that salmon with the vgll3 late 510 maturation genotype tended to have more actively working mitochondria and lower reserve 511 capacity (i.e., a higher coupling capacity, P/E ratio). Electron transfer in fish with late maturation 512 genotype was almost maximally coupled to ATP synthesis while the coupling was only up to 513 ~70% in fish with the early maturation genotype. In the white adipose tissue of obese humans, 514 coupling capacities of 77-83% (with an increasing trend during weight loss) have been observed 515 (51). Thus, the adipose tissue coupling capacity in salmon with the early maturation genotype 516 was lower compared to that in obese humans, and the higher coupling capacity in salmon with 517 late maturation genotypes is in line with active catabolic metabolism. Because coupling capacity 518 reflects the ability of mitochondria to adjust ATP synthesis to changes in energy demand and 519 supply, we also hypothesise that it could affect the resource acquisition of salmon, especially if 520 the effect is consistent across tissues. Specifically, the lower coupling capacity in early vs. late 521 maturation genotype could allow the fish with early maturation genotypes to increase ATP 522 synthesis rate more during high energy demand and acute changes in environmental 523 conditions. In contrast, salmon with the late maturation genotype with possibly higher catabolic 524 metabolism may respond more poorly to acute increases in energy demand or stress (12), thus 525 having less energy available to invest into survival, growth, and maturation. The mitochondria of 526 salmon with late maturation genotype were already near-maximally coupled despite the 527 relatively high food availability in this study (except for a 2d-period without feeding prior to 528 sample collection). Such inherent high coupling capacity of salmon with the late maturation 529 genotype may indicate that they could be more vulnerable to low food conditions such as during 530 winters in freshwater (52) or in the sea (53). In line with this, a lower aerobic scope at the whole 531 animal level -indicating lower capacity for aerobic metabolism beyond self-maintenance-in 532 juvenile salmon with the late maturation genotype was also found by Prokkola et al. (19). 533 Further, the lower coupling capacity of the late maturation genotype matches the distribution of 534 salmon with contrasting age-at-maturity in the wild, where individuals maturing later spawn 535 typically in larger rivers that are likely more environmentally stable. Further studies are required 536 to test these hypotheses linking adipose tissue mitochondrial function and mitochondrial 537 coupling to whole animal performance in the wild and in conditions of low food availability and 538 fasting. 539

540
The molecular pathways that could link vgll3 to mitochondrial respiration are not well known. In 541 humans and mice, vgll3 is a cofactor binding to TEA domain -containing (TEAD) transcription 542 factors that regulates tissue differentiation pathways, such as adipogenesis and myogenesis, as 543 well as pathways that regulate development and remodelling of tissue composition and organ 544 and cell size, such as hippo signalling pathway (25, 54, 55). In Atlantic salmon, the widespread 545 expression of vgll3 is correlated with factors in the Hippo pathway (i.e. YAP, and TEAD) (56) 546 suggesting a functional analogy in salmon with mammals. Intriguingly, the same pathways also 547 control mitochondrial biogenesis and function (57-59), further supporting that vgll3 genetic 548 variation might affect mitochondrial functional variation. 549

550
We also detected sex and/or maturation effect on mtDNA amount, where (mature) males had a 551 higher mtDNA amount (relative to gDNA) in adipose tissue than (immature) femalesalthough 552 there were no sex differences in mitochondrial respiration. Differences in adipose tissue 553 processes may emerge between mature and immature individuals, because salmon consume a 554 large part of their adipose tissue to support the high energy demand of maturation (22, 60). A 555 previous study also suggests that sex-specific maturation schedules could be mediated by non-556 visceral lipid storage, e.g., in muscle (21). Hence, future studies that partition sex and 557 maturation status effects across tissue types would be valuable to assess the role mitochondrial 558 variation in relation to these two phenotypes. 559 560 Conclusions 561 Adipose tissue is central to maturation as well as energy homeostasis but very little is known 562 about how these two processes could be genetically interlinked. Our proof of principle study 563 showed the feasibility of studying adipose tissue cellular respiration in salmon and yielded 564 insightful results from which we generated informed hypotheses for future research. To further 565 integrate adipose tissue metabolism into a life-history evolution framework, measurements of 566 mitochondrial respiration in salmon with different vgll3 genotypes could be combined with 567 analyses of lipogenesis and lipolysis (i.e., lipid synthesis and release, respectively) for example 568 using gene expression, and lipid quantification with histochemistry. Moreover, mitochondrial 569 respiration measurements could be combined with measurements of reactive oxygen species 570 and the effectiveness of ATP synthesis (ATP/O) (7, 61). Ultimately, to generalise the role of 571 mitochondrial respiration and of coupling capacity in life-history evolution, studies would need to 572 address life-stage specific genetic effects, and measurements should be extended to non-573 adipose tissues. Given the common physiological roles and functions of visceral white adipose 574 tissue in salmon and humans (28), a better understanding of these functions in salmon may also 575 facilitate its use as a new model species for obesity research. Conversely, our study provides an 576 example of how a more advanced understanding of metabolic disorders and obesity can be 577 harnessed to address questions relevant for ecology and evolution. 578 579 Acknowledgements 580 We thank Petra Liljeström, Mikko Immonen, Paul Bangura, and numerous interns for help in fish 581