Abstract
Social affinity and collective behavior are nearly ubiquitous in the animal kingdom, but many lineages feature evolutionarily asocial species. These solitary species may have evolved to conserve energy in food-sparse environments. However, the mechanism by which metabolic changes regulate social affinity is not well investigated. In this study, we used the Mexican tetra (Astyanax mexicanus), which features riverine sighted surface (surface fish) and cave-dwelling populations (cavefish), to address the impact of metabolic shifts on asociality and other cave-associated behaviors in cavefish, including repetitive turning, sleeplessness, swimming longer distances, and enhanced foraging behavior. After 1 month of ketosis-inducing ketogenic diet feeding, asocial cavefish exhibited significantly higher social affinity, whereas social affinity regressed in cavefish fed the standard diet. The ketogenic diet also reduced repetitive turning and swimming in cavefish. No detectable behavioral shifts were found regarding sleeplessness and foraging behavior, suggesting that other evolved behaviors are not regulated by ketosis. We further examined the effects of the ketogenic diet via supplementation with extragenic ketone bodies, revealing that ketone bodies are pivotal molecules associated with social affinity. Our study indicated that fish that evolved to be asocial remain capable of exhibiting social affinity under ketosis.
Introduction
Wild animals experience frequent fasting because of daily, seasonal, and yearly changes in food availability. Physiologically, fasting can increase the secretion of appetite-related hormones (e.g., ghrelin, peptide Y, orexin) and induce a metabolic shift into nutritional ketosis (McCue, 2010). Concerning behavioral outputs, fasting also induces shifts including boldness in foraging involving risk-taking (Padilla et al., 2016) and a shift from avoiding to approaching prey (Filosa et al., 2016). Interestingly, fasting also induces non-foraging–related behaviors including aggression toward cohorts (Fokidis et al., 2013; Solianik et al., 2016) and engagement in social dominance (Nakajo et al., 2020). These non-foraging behaviors could be evoked by the metabolic changes that occur in a state of nutritional ketosis instead of the increased production of appetite-related hormones. However, it is not fully understood whether ketosis itself in the absence of hunger drives these non-foraging behaviors. Such knowledge will open a path to understanding the effects of different dietary intakes on changing environments, such as switching from ketosis-inducing very low-carbohydrate diets to glycolysis-inducing carbohydrate-rich diets or vice versa.
Recently, the ketosis-inducing ketogenic diet (KD), which contains a high amount of fat, sufficient protein, and very low amount of carbohydrates, gained popularity among humans because of its neuroprotective and anti-inflammatory effects without effects on appetite-related hormone levels (Deemer et al., 2020; Ludwig, 2020; Sumithran et al., 2013). The KD is an effective treatment for refractory seizures, and there is some evidence that it may be beneficial for other nervous system-based disorders such as Alzheimer’s disease, Perkinson’s disease, and autism (Lee et al., 2018; McDonald and Cervenka, 2018; Phillips et al., 2018; Ruskin and Masino, 2012). Because modern humans evolved to acquire resistance to starvation (Bellisari, 2008), our body physiology and behavioral tendencies possibly evolved to accommodate drastic metabolic changes. However, the major molecular mechanisms for these positive are largely unknown (Ludwig, 2020; Qin et al., 2021). We were therefore motivated to explore the effects of metabolic shifts, particularly from glycolysis to ketosis, on behavioral outputs such as social affinity using a single species consisting of typical and starvation-resistant populations.
A suitable model system for this purpose is the Mexican cavefish (Astyanax mexicanus). A. mexicanus has emerged as a model system of diverse aspects of evolution and development, including those with relevance to human medicine, e.g., cataract formation, diabetes, albinism-related syndrome, and insomnia (Aspiras et al., 2015; Bilandžija et al., 2018, 2013; Duboué et al., 2012, 2011; Jaggard et al., 2017; Keene et al., 2016; Ma et al., 2014; McGaugh et al., 2014; Riddle et al., 2018; Rohner et al., 2013; Strickler et al., 2007). A. mexicanus consists of surface riverine epigean (surface fish) and cave-dwelling hypogean (cavefish) populations. Cavefish diverged from their surface-dwelling relatives 20,000–200,000 years ago (Fumey et al., 2018; Herman et al., 2018), and they have subsequently evolved many distinct morphological and behavioral phenotypes in the food-sparse cave environment, including eye regression/loss, pigment reduction, increased mechanosensory lateral line activity, adherence to vibration stimuli, sleeplessness, hyperactivity, repetitive circling, and reduced social affinity (Iwashita and Yoshizawa, 2021; Keene et al., 2016; Yoshizawa, 2015; Yoshizawa et al., 2018). Compared to cavefish, surface fish exhibit typical teleost phenotypes, including typical eyed and pigmented morphologies, no strong adherence to vibration stimuli, nocturnal sleep patterns, and social affinity. Many cavefish traits are believed to have evolved to adapt to food-sparse dark environments. Indeed, wild cavefish were estimated to be exposed to approximately 6 months of food-sparse conditions annually (Espinasa et al., 2021), and they are likely to have the ability to withstand starvation via increased fat storage (Aspiras et al., 2015), slower weight loss during starvation (Huppop, 1986), reduced energy-costing circadian activities, and lack of eyes (Moran et al., 2015, 2014).
Concerning social-like behavior, cavefish exhibit no detectable schooling behavior (Kowalko et al., 2013; Patch et al., 2020; Pierre et al., 2020) or hierarchal dominance (Elipot et al., 2013). By contrast, surface fish school/shoal with cohorts and model fish (Kowalko et al., 2013) exhibit group hierarchical dominance (Elipot et al., 2013). Because social behaviors in many fish (e.g., zebrafish) are promoted by visual stimuli, blind cavefish might not express social-like activities because of the absence of visual acuity. However, a recent detailed study illustrated that surface fish exhibit a high level of social-like nearby interactions (one-by-one affinity) in the dark, and these of which were promoted by mechanosensory lateral line inputs (Iwashita and Yoshizawa, 2021). Interestingly, blind cavefish displayed much lower levels, albeit significant, of nearby interactions than surface fish (Iwashita and Yoshizawa, 2021). Further, cavefish exhibited plasticity in the level of nearby interactions in which they increased plasticity in a familiar environment in comparison with an unfamiliar environment (Iwashita and Yoshizawa, 2021), which is similar to the findings in patients with autism (Helt et al., 2020; Runco et al., 1986).
Thus far, similarities between cavefish and patients with autism have been investigated in terms of gene regulation- and innate behavior-profiles. First, the cavefish gene expression profile is closer to that of patients with autism than to that of other known model systems (cavefish and surface fish transcriptomes exhibited the same directional gene expression changes observed in the brains of patients with autism [>58.5% of 3152 cavefish orthologs]). Conversely, other proxy systems (BTBR mice [classic autism model] and shank3 knockout mice) exhibit much less overlap (<11%) (Lee et al., 2019; Provenzano et al., 2016; Yoshizawa et al., 2018). Second, cavefish’s evolved behaviors—asociality, repetitive behavior, sleeplessness, higher swimming activity, adherence to a particular vibration stimulus, and higher anxiety-related plasma cortisol levels—are similar to those in patients with autism (Yoshizawa et al., 2018). Third, cavefish and human ancestors are starvation-resistant, and they could share some metabolic pathways (Aspiras et al., 2015; Bellisari, 2008; Huppop, 1986; Riddle et al., 2018). These similarities and a fact that KD feeding increases sociality in patients with autism (Evangeliou et al., 2003; Lee et al., 2018; Li et al., 2021; Napoli et al., 2014) prompted us to study the effects of ketosis on social affinity in asocial cavefish.
In this study, we assessed the effects of the KD on an evolutionarily asocial cave population of A. mexicanus. The time-course experiment revealed that 1 month of KD feeding promoted and sustained the juvenile level of nearby interactions, whereas control diet (CD)-fed cavefish exhibited diminished nearby interactions. KD feeding also reduced repetitive turning and swimming activity, which are the hallmarks of the autism-like condition. However, the effects of the KD were limited. For example, sleeplessness and high adherence to a particular vibrating stimulus were not detectably changed under the 1-month KD treatment. To reveal the molecular basis of the effects of the KD, we provided supplementation with a major ketone body, beta-hydroxybutyrate (BHB). This experiment indicated that BHB supplementation promoted nearby interactions and reduced repetitive turning, covering the major effect of the KD.
Overall, ketosis appears to be capable of significantly shifting the asociality of evolved cavefish toward the surface fish phenotype, providing new insights into the contribution of the diet to the evolution of behavior.
Results
From our observation in their wild habitat (the Mexican cave Pachón, Supplemental Movie 1), cavefish swam slower and remained near each other more frequently than the lab population. Because the cave environment has a limited diet compared to that of the surface, we predicted that cavefish experience frequent ketosis induced by fasting.
To avoid appetite-related behavior, we developed a ketogenic diet (KD) based on a human milk formula (KetoCal3:1® with Zeigler zebrafish standard irradiated diet at a 5:1 weight ratio [nutritionally complete, ketogenic medical food; Nutricia North America, Inc. Gaithersburg, MD, USA]; Table 1; Materials and Methods). To test whether our KD could induce a shift in the balance of ketone body and glucose levels, we measured blood ketone and glucose levels after chronic dietary treatment. Three-month-old fish (juvenile–young adult stage) were used in this study because many adult-type behaviors of cavefish emerge in this stage, including higher adherence to a vibration stimulus (vibration attraction behavior [VAB]) (Yoshizawa et al., 2010), less social affinity, and longer swimming distances compared to the findings in surface fish. After 5 weeks of KD feeding, both ketone and glucose concentrations were decreased compared to the findings in CD-fed fish (KetoCal3:1 and Zeigler zebrafish diet at a 1:5 weight ratio; Table 1; Fig. 1A–C). For both diets, surface fish exhibited a significantly higher ketone body level than cavefish (Fig. 1B), whereas cavefish exhibited a higher glucose level than surface fish (Fig. 1C). The glucose ketone index , a medical index proposed for brain cancer diagnosis (Meidenbauer et al., 2015), was lower in surface fish than in cavefish, and the value was reduced under KD feeding in both surface fish and cavefish compared to that in their CD-fed counterparts (Fig. 1D). This result indicated that KD feeding more strongly reduced glucose levels than ketone body levels, resulting in a lower GKI in KD-fed fish than in CD-fed fish (Fig. 1D) and suggesting that KD feeding could shift the metabolic state from glycolysis toward ketosis.
Regarding this dietary treatment, we first examined its ontogenic (developmental) effects on collective social-like behavior (Iwashita and Yoshizawa, 2021). Many adult behaviors emerge in the transition from juvenile to young adult (adolescent) in 3–4-month-old A. mexicanus fish, including foraging behavior, VAB (Yoshizawa, 2015; Yoshizawa et al., 2010), adult-type regulation of sleep (independent from catecholamine) (Duboué et al., 2012; Jaggard et al., 2020, 2017; Yoshizawa et al., 2015), and the collective behavior in young adults (under higher Reynold’s number; Iwashita and Yoshizawa, 2021). We therefore investigated the shift of collective behavior in 3–4-month-old fish using criteria based on the vicinity of two fish (≤5 cm) and duration of nearby interactions (≥4 s) during tracking in a four-fish group (Iwashita and Yoshizawa, 2021) (Fig. 2B). At 3 months old, (‘Pre-treatment’ in Fig. 2), surface fish exhibited social-like nearby interactions for 17.0 ± 4.4 s (Fig. 2C) and 3.1 ± 0.4 bout number of nearby interactions (Fig. 2D) during the 5-min assay. Conversely, cavefish exhibited an approximately 50% shorter interaction duration (8.3 ± 1.5 s; Fig. 2C) and a smaller bout number of interactions (1.8 ± 0.3; Fig. 2D). To track the effect of the KD treatment, fish were fed the KD for 5 weeks, followed by CD feeding during weeks 6–9 to assess the persistence of the effects of the KD (Fig. 2A, C, and D).
The nearby interactions of surface fish did not differ between CD and KD feeding (Fig. 2C and D). By contrast, the nearby interactions of cavefish were significantly decreased by CD feeding compared to the effects of KD feeding in weeks 4 and 5 (Fig. 2C and D), and interactions remained depressed through week 9 by CD feeding. This effect of the KD diet on nearby interactions did not persist. After KD deprivation and CD feeding, the nearby interactions of KD-fed cavefish were indistinguishable from those of CD-fed fish (6–9 weeks, Fig. 2C, and D), suggesting that KD has a promotive/supportive effect on collective behavior in genetically asocial cavefish.
To investigate in detail whether KD-promoted nearby interactions have social-like properties, we explored swimming speed, which is an indicator of the movement of fish in the vicinity of each other. Fish are more likely to have an opportunity to express affinity toward each other at a slower swimming speed (Iwashita and Yoshizawa, 2021). Indeed, surface fish moved slower during nearby interactions than during the non-nearby interaction period (Iwashita and Yoshizawa, 2021) (Fig. 3A). Consistently, KD-fed cavefish swam slower during the nearby interaction period than during the non-nearby interaction period (‘5 weeks;’ Fig. 3B). Overall swimming speed was also slower in the KD group than in the CD group (‘5 weeks;’ Fig. 3B). These findings indicated that KD-fed cavefish exhibited more social-like nearby interactions with a similar speed profile as surface fish. In surface fish, there was no major difference in swimming speed profiles between CD and KD feeding (‘5 weeks;’ Fig. 3A). To address whether the slower swimming speed was sufficient to increase nearby interactions, we tracked the total swimming distance within 5 min from pre-treatment to week 9 of feeding (Supplementary Fig. 1). KD-fed cavefish exhibited a significantly shorter swimming distance (slower swimming speeds) from the first week of feeding (Supplementary Fig. 1), which was much earlier than when the higher level of nearby interactions emerged (weeks 4–5). This result suggests that although a slower swimming speed is associated with nearby interactions (Fig. 3) and KD feeding reduced swimming speeds in cavefish, a slower speed itself is not sufficient to induce nearby interactions.
Repetitive turning is frequently observed in an antagonistic relationship with nearby interactions in cavefish (Iwashita and Yoshizawa, 2021). That is, individuals with few nearby interactions frequently exhibit a high level of turning bias or ‘repetitive turning.’ Accordingly, CD-fed cavefish with few nearby interactions (4–5 weeks) exhibited significantly higher turning bias than KD-fed cavefish (Fig. 4A, B). KD-fed cavefish displayed a comparable level of balanced turning as surface fish (close to a score of ‘1’ in Fig. 4B). In summary, these results suggest that KD feeding could reduce repetitive turning while maintaining longer nearby interactions.
Tracking behavioral changes each week (Fig. 2) may result in confounding factors such as fish remembering the recording environment. To clarify whether our results captured the genuine effects of KD feeding, we repeated 4–5-week dietary treatment in a new set of fish (Supplementary Fig. 2). Similarly, surface fish did not exhibit a detectable change in the duration and number of nearby interactions between CD and KD feeding (Supplementary Fig. 2A, B). By contrast, CD-fed cavefish displayed fewer nearby interactions, whereas the level of nearby interactions was retained in KD-fed cavefish, resulting in a higher level of nearby interactions in KD-fed cavefish (Supplementary Fig. 2A, B). In this repeated experiment, the results for swimming distance and repetitive turning were also similar to those in the previous experiment; specifically, CD-fed cavefish swam longer distances than KD-fed cavefish (Supplementary Fig. 2C), whereas surface fish did not exhibit a detectable shift in swimming distance between the CD and KD groups. Additionally, CD-fed cavefish displayed a high level of turning bias, whereas KD-fed cavefish exhibited balanced turning (Supplementary Fig. 2D).
We then explored other changes induced by KD feeding, including changes in sleep, 24-h swimming distance, and adherence to a vibrating stimulus, which are distinct between surface fish and cavefish. Cavefish largely exhibit reduced sleep duration and swim almost all day, perhaps to find nutrients in the food-sparse environment (Duboué et al., 2012, 2011; Jaggard et al., 2017; Yoshizawa et al., 2015). After 5 weeks of dietary treatment on the 3 – 4 months old fish, both surface fish and cavefish exhibited a shorter sleep duration than observed before treatment regardless of the diet (Fig. 5A, particularly at night), suggesting growth between 3 – 4 and 4 – 5 months old exerted a negative effect on the sleep duration. However, there was no detectable difference between CD and KD feeding.
Animals’ sleep is usually fragmented, involving repeated sleep/awake cycles during the night (diurnal animals) or day (nocturnal animals) (Campbell and Tobler, 1984). Then, the structure and regulation of sleep are typically analyzed according to the number of events (bout). Our detailed sleep analysis illustrated that KD-fed cavefish displayed fewer sleep bouts during the night than CD-fed cavefish (Fig. 5B). However, the number of sleep bouts did not differ between CD and KD feeding (5 weeks; Supplementary Fig. 3). Overall, the sleep phenotype was not dramatically changed by KD feeding, and cavefish exhibited a shortened sleep duration.
The sleep duration is negatively correlated with the 24-h swimming distance (Yoshizawa et al., 2015). Cavefish displayed overall higher activities, which was consistent with previous findings (Duboué et al., 2011; Yoshizawa et al., 2015) and consistent with the findings of longer swimming distance in the nearby interaction assay (Fig. 5C). CD-fed cavefish swam longer distances than KD-fed cavefish. Surface fish, in contrast, did not exhibit a detectable difference in swimming distance between KD and CD feeding. Overall, the KD treatment induced little changes in sleep behaviors in both surface and cavefish.
In general, the KD is assumed to induce ketosis without increasing appetite. We then checked the shift of foraging behavior under KD feeding. Cavefish evolutionarily exhibit increased foraging behavior, termed VAB, in which fish adhere to a particular vibration stimulus (35–40 Hz) in the dark (Yoshizawa et al., 2010). VAB is advantageous for prey capture in the dark. Cavefish and surface fish did not exhibit a detectable difference in VAB between CD and KD feeding, whereas VAB was significantly increased during 1 month of growth (pre-treatment vs. 5 weeks; Supplementary Fig. 4A).
The swimming distance in 3-min VAB assays was decreased in KD-fed cavefish, consistent with the nearby interaction and sleep studies (5 weeks; Supplementary Fig. 4B). In summary, the VAB analysis indicated that KD feeding did not increase a foraging behavior.
Although the KD diet induced significant changes in some behavioral outputs, it suppressed growth during treatment. The average weights of KD-fed surface fish and cavefish were 55.5 % and 69.9 % of those in their CD-fed counterparts, respectively (5 weeks; Fig. 6B). The standard length of KD-fed surface fish was also significantly reduced (5 weeks; Fig. 6A).
Are these behavioral and growth changes induced by ketosis? The KD contains high amounts of fat and other ingredients. This question motivated us to test the molecular basis of the effects of the KD by adding major ketosis metabolites to the standard diet.
In humans, the KD induces ketosis, in which the liver releases the major ketone body, beta-hydroxybutyrate (BHB), via beta-oxidation of fat (Evans et al., 2017). Instead of supplying a massive amount of fat using the KD, BHB (sodium salt form of racemic BHB: 50%L-form and 50%D-form; only the D-form is considered to be biologically active) might be responsible for the majority of effects observed after KD feeding. With this idea, sodium salt BHB was provided to both surface fish and cavefish for 4 weeks. The result indicated that the BHB supplemental diet promoted nearby interactions in cavefish (Fig. 7A and B) and reduced the duration of nearby interactions in surface fish (Fig. 7A). Turning bias tended to be reduced by BHB supplementation in cavefish, although significance was not achieved (Fig. 7C). Swimming distance was not reduced in surface fish or cavefish (Fig. 7D). The body growth of BHB-treated surface fish and cavefish was comparable to that in control fish (standard length and weight; Supplementary Fig 5A and B, respectively). The night-time sleep and VAB did not exhibit detectable differences between control and BHB treatment (Supplementary Fig. 6A and B, respectively) In summary, BHB treatment covered many effects of KD treatment did, including changes in social interactions and repetitive turning. BHB had fewer negative effects on growth and swimming activities, suggesting that ketone bodies are responsible for the ‘positive’ effects of KD feeding.
Discussion
In this study, we examined the behavioral shifts induced by KD feeding and BHB supplementation. Ketosis is expected to occur frequently in wild animals because of a failure to find food (fasting) or an absence of carbohydrate inputs/synthesis (available nutrients). Certain levels of socialness can be beneficial to animal species for mating and finding food. Under KD feeding, cavefish maintained their juvenile level of nearby interactions until the treatment ended (5 weeks). Nearby interactions were then reduced to an indistinguishable level from the control levels within 1 month after stopping KD feeding. Surface fish exhibited a higher number of nearby interactions than cavefish, and no detectable difference was observed in nearby interaction levels between CD- and KD-fed surface fish. KD feeding also effectively reduced repetitive turning in cavefish, whereas CD-fed treated cavefish exhibited a high level of repetitive turning. Under KD feeding, both surface fish and cavefish significantly reduced swimming distances during development, during which the swimming distance typically becomes longer. There were no detectable changes in sleep and foraging behavior (VAB) under 1 month of KD feeding. These patterns in behaviors and growth were not changed in two replicated experiments (social affinity and repetitive turning), supporting the consistency of the observed effects under KD feeding. Finally, the major KD metabolite, BHB, could cover the KD effect, indicating the ketone body has the pivotal role on this treatment.
Effects of the KD on blood ketone levels and body growth
Under 4–5 weeks of KD feeding, blood ketone and glucose levels were reduced compared to the effects of the CD in both surface fish and cavefish, contradicting our expectation that ketone levels would be higher in the KD group. However, GKI (Meidenbauer et al., 2015) was significantly lower under KD feeding than under CD feeding. These significant changes in GKI in both surface fish and cavefish suggest that the metabolic condition was shifted toward ketosis by KD feeding. In general, cavefish had a higher GDI than surface fish under both diets, suggesting that the cavefish physiology was constitutively biased toward glycolysis. For example, blood glucose levels in cavefish under KD feeding were similar to those in surface fish under CD feeding, whereas cavefish had 3-fold lower ketone levels than surface fish under CD feeding, resulting in a higher GKI even under KD feeding.
KD feeding for 4–5 weeks also resulted in slowed body growth. This growth retardation has been observed in patients with epilepsy chronically fed a KD (Coppola et al., 2010; Napoli et al., 2014), and these results were consistent with our observations in KD-fed fish. The detailed molecular/physiological mechanisms are largely unknown, but this study identified BHB as a potential candidate mediating KD-associated phenotypes (see below).
Effects of ketones in the TCA cycle and epigenetics in the brain
In mammals, KD feeding causes a ‘starvation’-like state, causing the liver to release ketone bodies into the bloodstream. BHB is the major ketone body produced by the liver through beta-oxidation. The gut epithelia also absorb and circulate ketone bodies from the diet and/or gut microbiota. Both liver- and gut-derived ketone bodies can cross the blood–brain barrier and exert two functions: (i) inhibit histone deacetylase, which influences epigenetic regulation and induces gene expression in neurons; and (ii) act as a general energy source that is converted into acetyl-CoA to fuel the aerobic TCA cycle in neurons. Both pathways have the potential to alter brain function. The facts that cavefish easily tolerate high blood glucose levels, at which surface fish was paralyzed (Riddle et al., 2018), and Wnt signaling is upregulated in cavefish, potentially resulting in high glycolytic activity as humans do (Vallée and Vallée, 2018; Yoshizawa et al., 2018), support the aforementioned hypothesis that cavefish exhibit high blood glucose levels and generate energy via glycolysis. Also, ketone bodies can also promote behavioral shifts by changing the epigenetic state by inhibiting histone deacetylase (Krautkramer et al., 2017; Szyf, 2015). Histone deacetylase inhibition increases gene expression in general. This possibility is supported by the fact that cavefish have more downregulated genes (2913 genes, log2 < −1.0) than upregulated genes (1643 genes, log2 > 1.0) in the transcriptome at 72 h post-fertilization (Gross et al., 2013; Yoshizawa et al., 2018). In addition, more methylated loci are found in the eye genes of cavefish than in surface fish, which could also be true for other tissues (Gore et al., 2018), and most of these methylated gene loci were downregulated. The brains of patients with autism are also expected to be hypermethylated, resulting in a transcription-less condition (Zhu et al., 2014). Therefore, these two pathways, namely metabolism and epigenetics, were highlighted as possible targets of ketone bodies during behavioral shifts under ketosis. Future research should address these possibilities to clarify the metabolism-based evolution of behavior (cf. (Qin et al., 2021)
Ontogeny of nearby interactions and the KD
In this study, 3-4 months-old cavefish exhibited a weak but detectable level of nearby interactions (social affinity), and this social affinity decayed under CD feeding. Interestingly, KD-fed cavefish and surface fish fed either diet maintained a similar level of nearby interactions during 5 weeks of dietary feeding. The reduction of nearby interactions in CD-treated cavefish can be explained by (1) quicker exhaustion under CD feeding (aerobic ketosis produces more adenosine triphosphate than anaerobic glycolysis), (2) greater anxiety in the recording environment (Iwashita and Yoshizawa, 2021), and (3) less social motivation. The first explanation is unlikely because CD-fed cavefish swam longer distances than KD-fed cavefish. The higher level of anxiety could explain the findings because cavefish exhibited increased repetitive turning, which is related to higher anxiety in mammals (Langen et al., 2011). In addition, cavefish displayed fewer nearby interactions in an anxiety-associated unfamiliar environment in prior research (Iwashita and Yoshizawa, 2021). In the future, the anxiety level should be monitored using plasma cortisol levels in the future (Gallo and Jeffery, 2012). Less motivation regarding social affinity is also a possible cause, and this variable can be monitored by assessing activities in social decision-making networks including the preoptic area, nucleus accumbens, and striatum (O’Connell and Hofmann, 2012, 2011). Explanations (2) and (3) are not mutually exclusive, and co-occurrence is possible. These possibilities will be assessed in our future study.
Possible target system for ketosis
Under KD feeding and BHB treatment, we observed increased social affinity and reduced repetitive turning and swimming distances. However, we detected no changes in sleep and VAB. From the knowledge of neurotransmitters and their associated behaviors, these behavioral phenotypes indicated the possible involvement of the dopaminergic and serotonergic systems but less involvement of the cholinergic, orexin/hypocretinergic, histaminergic, or adrenergic system. The mechanism by which ketosis or ketone bodies more strongly affect the dopaminergic and serotonergic systems than the other systems is unclear. However, the findings that the dopaminergic and serotonergic systems are sensitive to ketosis in terms of evolution are extremely interesting in its mechanism and for future therapy applications.
Ketosis in the cave environment
Cave-dwelling animals usually experience less temperature fluctuation and fewer dietary inputs (Culver and Pipan, 2009), but these features can vary. The diets of cave-dwelling animals in the dry season (approximately 6 months/year) could be organic matter in the pool bottoms, bat guano (larger adults), or small crustaceans (smaller fish), whereas food is sparse in the rainy season (approximately 6 months/year) (Espinasa et al., 2021, 2017). These available diets contain extremely low amounts of carbohydrates, and they can be high in protein and fat (e.g., crustaceans). Although some amino acids, lactate, and glycerol can be used for glucose synthesis in fish (Polakof et al., 2012), cavefish are expected to be exposed to carbohydrate-deprived diets or frequent fasting and therefore frequent ketosis. In prior research, wild cavefish swam slower and exhibited similar social affinity as observed in KD-fed cavefish in this study (Movie 1). Although these observations and dietary inputs suggested that wild cavefish may be under frequent ketosis, recent multiple reports indicated that cavefish may be under anaerobic glycolysis to adapt to anaerobic cave water conditions because of the approximately 20% lower oxygen level in cave pools (Boggs et al., 2022; van der Weele and Jeffery, 2022). In addition, cavefish tend to store lipids instead of using them through the enhanced PPARg pathway (Xiong et al., 2022). These expectations of low ketosis appear to contradict expectations in the wild—starved ketosis conditions. However, it appears to fit well with the findings in cavefish, namely high blood glucose and low ketone levels even under KD feeding in this study, as well as higher GKI in cavefish than in surface fish in both the CD and KD groups (Fig. 1D). Cavefish appear to have evolved to maintain a high GKI (high blood glucose and low ketone levels); therefore, the physiology of cavefish may allow them to survive in the low-oxygen condition by using anaerobic glycolysis. KD-fed cavefish behave similarly as wild cavefish because the balance between ketosis and glycolysis could reach a similar level as that in the wild after KD feeding. By contrast, if cavefish are fed a typical carbohydrate-rich lab fish diet, it may overactivate glycolysis and result in a higher GKI, which may lead to reduced social affinity and increased repetitive circling and swimming distance. The future use of a pharmacological glycolysis inhibitor (e.g., 2-deoxy-D-glucose; Yao et al., 2011) can reveal the relationship between GKI and cavefish behaviors.
Summary statement
Solitary animals surprisingly share a set of dysregulated genes and behavioral outputs. In this study, we demonstrated that a diet that induces ketosis shifts these behaviors toward the surface fish phenotype regardless of the presence of many dysregulated genes. In addition to the gene therapy approach, ketone body-based treatment may open a path for sustainable and less toxic therapy for multigenic psychiatric disorders, including autism, although the target pathways remain unclear. Concerning the genetics of behavior, differentially expressed metabolic genes have been largely overlooked because it was difficult to interpret. Because mitochondria-based disorders are highlighted in neuroscience (Chauhan et al., 2012; Rajasekaran et al., 2015), the balance between glycolysis and ketosis could be the starting point for identifying a therapeutic target. The known evolved behaviors should be also revisited by investigating whether the metabolic shift promotes the variations of behaviors.
Materials and Methods
Fish maintenance and rearing in the lab
A. mexicanus surface fish used in this study were the laboratory-raised descendants of original collections created in Balmorhea Springs State Park in Texas in 1999 by Dr. William R. Jeffery. Cavefish were laboratory-raised descendants originally collected from Cueva de El Pachón (Pachón cavefish) in Tamaulipas, Mexico in 2013 by Dr. Richard Borowsky.
Fish (surface fish and Pachón cave populations) were housed in the University of Hawai’i at Mānoa Astyanax facility with temperatures set at 21 ± 0.5°C for rearing, 24 ± 0.5°C for behavior experiments, and 25 ± 0.5°C for breeding (Elipot et al., 2014; Yoshizawa et al., 2015). Lights were maintained on a 12-h:12-h light:dark cycle (Elipot et al., 2014; Yoshizawa et al., 2015). For rearing and behavior experiments, the light intensity was maintained at 30–100 Lux. Fish husbandry was performed as previously described (Elipot et al., 2014; Keene et al., 2016; Yoshizawa et al., 2015). Fish were raised to adulthood and maintained in standard 42-L tanks in a custom-made water-flow tank system. Adult fish were fed a mixed diet to satiation twice daily starting 3 h after the lights were turned on (Zeitgeber time 3 [ZT3] and ZT9; Zeigler Adult zebrafish irradiated diet, Zeigler Bros, Inc, Gardners, PA; TetraColor Tropical Fish Food Granules, Tetra, Blacksburg, VA, USA; Jumbo Mysis Shrimp, Hikari Sales USA, Inc., Hayward, CA, USA). All fish in the behavioral experiments were between 2.5 and 5 cm in standard length and between 3 and 12 months old. Fish ages were stated in each experiment. All fish care and experimental protocols were approved under IACUC (17-2560) at the University of Hawai’i at Mānoa.
Ketogenic diet
To prepare the KD, we used a mixture of a human KD (KetoCal3:1) and zebrafish standard diet (adult zebrafish irradiated diet) in a 5:1 ratio. The gross caloric amounts were 6.99 kcal/g for KetoCal3:1 and 3.89 kcal/g for the zebrafish diet. Regarding the CD, we used the same KetoCal3:1 and zebrafish irradiated diet but mixed at a 1:5 ratio. The KetoCal3:1 powder and ground zebrafish irradiated diet were mixed in the aforementioned ratios and solidified with 1% agar at a final concentration of 20% w/v (2 g of mixture in 10 mL of 1% agar). After solidification, both KD and CD agar was cut into 3-mm3 cubes, and each four-fish group was given 1–2 pieces every morning (ZT 0:00–3:30) and afternoon (ZT 8:00–12:00). The fish were fed ad libitum in each feeding and the remaining amount was removed 1 h after feeding using a pipette.
Behavior assays
The protocol for social-like nearby interactions was described previously (Iwashita and Yoshizawa, 2021). Briefly, four fish raised in a home tray (15.6 × 15.6 × 5.7 cm3 Ziploc containers, S. C. Johnson & Sons, Inc, Racine, WI, USA) were released in a recording arena (49.5 × 24.2 × 6.5 cm3) with a water depth of 3 cm on the stage of a custom-made infrared (IR) back-light system within a custom-built black box (75 × 50 × 155 cm, assembled with polyvinyl chloride pipe frame and covered by shading film). The IR back-light system was composed of bounce lighting of IR LED strips (SMD3528 850 nm strip: LightingWill, Guang Dong, China). The video was recorded at 20 frame/s using VirtualDub2 software (build 44282; http://virtualdub2.com/) with the x264vfw codec for 6 min, and the last 5 min were used for the analysis. After the recording, the fish were returned to the home tray. The X-Y coordinates of each fish were calculated using idTracker software (Pérez-Escudero et al., 2014) after the video image was processed for background subtraction using ImageJ (Iwashita and Yoshizawa, 2021). This X-Y coordinate was also used for the turning bias analysis. The duration and number of nearby interactions and swimming speed during and after nearby interaction events were calculated using custom-made MATLAB script (MathWorks Inc., Natick, MA, USA) (Iwashita and Yoshizawa, 2021).
The turning bias rate was calculated as , where Ns and Nl represent a smaller (Ns) or larger (Nl) number of left or right turns. This turning bias rate indicates the extent to which fish turning is biased to the left or right, and ranging from ‘1’ (L-R balanced) to exponential (L or R biased). The numbers of left or right turns were calculated as changes in the angles of swimming directions in every five frame window (0.25 s) as described previously (Iwashita and Yoshizawa, 2021). An automatic calculation of the total number of the left or right turns is implemented in the aforementioned homemade MATLAB script.
Analyses of sleep and swimming distance were described previously (Yoshizawa et al., 2018, 2015). Briefly, fish were recorded in a custom-designed 10.0-L acrylic recording chamber (457.2 × 177.8 × 177.8 mm3 and 6.4 mm thick) with opaque partitions that permit five individually housed fish per tank (each individual chamber was 88.9 × 177.8 × 177.8 mm3). The recording chamber was illuminated with a custom-designed IR LED source the light-controlled room on a 12-h: 12-h cycle. The room light was turned on at 7:00 am and turned off at 7:00 pm each day. Behavior was recorded for 24 h after overnight (18–20 h) acclimation beginning 1–2 h after turning the light on (ZT1–2). Videos were recorded at 15 frames/s using a USB webcam with an IR high-pass filter. Videos were captured by VirtualDub2 software with the x264vfw codec and subsequently processed using Ethovision XT (Version 16, Noldus Information Technology, Wageningen, Netherlands). Water temperature was monitored throughout the recordings, and no detectable differences were observed during the light and dark periods (24.0 ± 0.5°C). The visible light during behavior recordings was approximately 30–100 Lux.
The tracking parameters for detection were as follows: the detection was set to ‘subject brighter than background’ and brightness contrast was set from 20 to 255; the current frame weight was set to 15; the video sample rate was set to 15 frames/s; and pixel smoothing was turned off. We monitored sleep, activity, and arousal thresholds via protocols previously established for A. mexicanus (Yoshizawa et al., 2015). The X-Y coordinates of each fish were subsequently processed using custom-written Perl (v5.23.0, www.perl.org) and Python scripts (3.8).
We assayed VAB as described previously (Yoshizawa et al., 2015, 2012, 2010). Briefly, fish were permitted to acclimate for 4–5 days prior to the assay in a cylindrical assay chamber (325-mL glass dish, 10 cm in diameter and 5 cm in height, VWR, Radnor, PA, USA) filled with conditioned water (pH 6.8–7.0; conductivity 600–800 μS). During the assays, vibration stimuli were created using a glass rod that vibrated at 40 Hz. The number of approaches to the vibrating rod was video recorded during a 3-min period under infrared illumination. The number of fish approaches in a 1.3-mm radius from the vibrating glass rod were analyzed using the X-Y coordinate of each fish head detected by the trained DeepLabCut model (Mathis et al., 2018; Fernandez et al., submitted)
Blood ketone and glucose measurements
The fish used to measure ketone and glucose levels were the siblings of the fish used for the behavioral assay. Eight fish were used for each treatment group (CD vs. KD, surface fish vs. cavefish; four groups in total). All blood samples were collected 2 h after feeding. Fish were then deeply anesthetized in ice-cold water, and blood was collected from the tail artery. Blood ketone and glucose levels were measured using an Abbott Precision Xtra blood glucose & ketone monitoring system (Abbott Laboratories, Abbott Park, IL, USA)
Measurement of tissue structure
Fish were anesthetized with 0.2 mg/mL ethyl 3-aminobenzoate methanesulfonate (MS-222: MilliporeSigma, Burlington, MA, USA) in ice-cold conditioned water (pH: 7.0; conductivity: 700 μS), and weight was measured after taking pictures with a standard camera (Pentax K-1 DSLR with 35-70 mm zoom lens, Ricoh, Tokyo, Japan). The standard body length and body depth were measured using ImageJ.
Statistical analysis
Regarding the power analysis, we designed our experiments based on three-way repeated-measures ANOVA with a moderate effect size (f = 0.25), alpha-error probability of 0.05, and power of 0.80, and the number of groups was eight (surface fish vs. cavefish × non-treated vs. treated × pre-treatment vs. post-treatment). G*Power software (Erdfelder et al., 2009; Faul et al., 2007) estimated that the sample size needed for this experiment was nine per group. We thus aimed to use at least 12 fish in each group for all experiments in this study.
For statistical comparisons of our data, we performed tests including Student’s t-test and two- or three-way generalized linear model analyses to compare surface and cavefish, treatment and non-treatment, and pre-treatment and post-treatment. We calculated Akaike’s information criterion (AIC) for each linear and generalized linear model and chose the model with the lowest AIC. Holm’s post hoc correction was used to understand which contrasts were significant (Holm, 1979).
Regarding replicates of experiments, we used different individuals for the replicates, namely two biological replicates, using different individuals in each trial (e.g., Fig. 2 and Supplementary Fig. 2). There was no repeated usage of individual fish excluding the time-course experiment (Fig. 2). For the experiments measuring sleep and VAB in addition to nearby interactions and turning bias (Figs. 5–6, Supplementary Figs. 2–4), we used two biological replicates and confirmed that the averages of experimental data did not largely differ from each other. We then merged the data acquired in two biological replicates and presented the data as a single set of results.
The aforementioned calculations were performed using R version 4.0.4 software (packages of car, lme4, and lmerTest) (Bates et al., 2015; Fox and Weisberg, 2019; Kuznetsova et al., 2017), and all statistical scores are available in Supplementary Table 1, the figure legends, or the text.
Figure Legends
Acknowledgement
We thank C Balaan for constructive comments regarding insights on social-like interactions in cavefish. We are grateful to V Crystal, J Choi, L Lu, J Nguyen, VFL Fernandes, K Lactaoen, M Worsham, H Hernandez, N Doeden, J Kato, M Ito, R Balmilero-Unciano, E Doy, A Martinez, and D Mones for fish care assistance. We gratefully acknowledge supports from the National Institute of Health (P20GM125508) to MY, Hawaii Community Foundation (18CON-90818) to MY.
Footnotes
Conflict of Interest: The authors declare no competing financial interests.