Lithium influences whole‐organism metabolic rate in Drosophila subobscura

Lithium is widely used to treat bipolar disorder. However, the efficacy and vulnerability as to its side effects are known to differ. Although the specific biochemical mechanism of action is still elusive, lithium may influence mitochondrial function, and consequently, metabolism. Lithium exposure in this study was conducted on a unique set of mito‐nuclear introgression lines of Drosophila subobscura to disentangle the independent effects of mitochondrial DNA (mtDNA) against a common nuclear DNA background. The study addressed three issues: (a) whether lithium has a dose‐dependent effect on whole‐organism metabolic rate, (b) whether mtDNA haplotypes show divergent metabolic efficiency measured by metabolic rate to lithium exposure and (c) whether lithium influences the whole‐organism metabolic rate across sexes. The results confirm that lithium influenced the whole‐organism metabolic rate, showing a subtle balance between efficacy and adverse effects within a narrow dose range. In addition, lithium exposure was found to influence metabolism differently based on mtDNA haplotypes and sex. This preliminary research may have a range of biological implications for the role of mitochondrial variability in psychiatric disease and treatment by contributing to the understanding and predicting of the lithium treatment response and risk for toxic side effects.


| INTRODUC TI ON
Lithium has been used for half a century as a first-line treatment for bipolar disorder (BD) (Geddes & Miklowitz, 2013), especially for acute mania and mood-stabilizing maintenance treatment (Malhi, Gessler, & Outhred, 2012). Lithium appears to reduce the risk of suicide in patients with BD (Malhi et al., 2012;Cipriani, Hawton, Stockton, & Geddes, 2013;Song, Sjölander, & Joas, 2017), and may have other benefits, such as reducing the risk of developing neurocognitive disorders (Jakobsson et al., 2017). However, lithium works best in BD patients with typical symptoms (e.g., extreme shifts in mood) and its efficiency may differ between partial and excellent responders (Alda, 2015). Despite the wide use of lithium and welldocumented efficacy in psychiatry, the biological mechanism of its action remains poorly understood.
There is a linear relationship between lithium levels in blood plasma and lithium ingestion, but regulatory systems to buffer the concentration are not present (de Roos, de Vries, & Katan, 2001).
Applying the D. subobscura MNILs, our study aimed to investigate: (a) whether lithium has a dose-dependent effect on the whole-organism metabolic rate, (b) whether mtDNA haplotypes show divergent whole organism metabolic rate to lithium exposure and (c) whether lithium influences the whole-organism metabolic rate differently across sexes.

Significance
This study provides novel evidence that metabolic parameters (O 2 consumption and CO 2 production) were influenced by lithium in a dose-dependent fashion that may have direct relevance for the narrow therapeutic index with a subtle balance in patients between adverse effects and efficacy. Our experimental design used a unique set of pure mitochondrial lines of Drosophila subobscura that allowed disentangling the independent effects of two main haplotype groups of mitochondrial DNA against a common nuclear DNA background. We confirmed in vivo that the whole organism metabolic rate is influenced by lithium exposure and dependent on mitochondrial genetic background and sex. We suggest that future efforts in this system focus on identifying the mechanistic cause for narrow dose lithium effects in pure mitochondrial lines. This preliminary research has a wide range of biological implications for mitochondrial variability in psychiatry as it may contribute to understanding divergent lithium efficiency as well as predict the risk of toxic side effects.

| Experimental population
The experimental species is D. subobscura and the flies were genotyped for their mtDNA haplotype using the methods described in

| Lithium sulfate dose concentration, fitness assays
The  Eggs were collected from oviposition cages after 5 days. The old food was removed and the fresh food medium introduced (standard, no lithium added). Immediately thereafter, the flies were allowed to oviposit eggs for 24 hr. Eggs were gently removed from the petri dishes with a lancet under a dissection microscope and transferred to 10 × 3 cm ∅ vials provided with 10 ml medium across the six lithium concentrations (C, L2-L6). Petri dishes with a fresh TA B L E 1 Mean (± SD) for viability across lithium concentration (C, L2-L6) for mitochondrial haplotypes (I and II) across MNILs

| Experimental design (lithium exposure)
As mentioned earlier, the crossing design allowed us to explore the effect of lithium on the whole organism metabolism of exposed MNILs ( Figure 1b). The MNILs were created to have two distinct mitochondrial haplotypes (HI and HII) expressed on the same nuclear background (D). HI and HII haplotypes were represented with three lines each.
In total, six MNILs were reared to adulthood (from egg to adult) for one generation at three concentrations of lithium sulfate (C, L1, and L2) according to the dose concentration curve. Flies were exposed to lithium during their developmental cycle (egg-larvaepupae-adult) and then transferred to a glass chamber (without lithium added) to perform measurements of metabolic rate. Flies were 3-5 days old when subjected to respirometry. Each biological replica (MNIL) had three technical replicas. The sexes were scored in the metabolic rate assay separately to include the sex-specific response to lithium exposure. In total, 108 samples were scored (6 MNILs × 3 lithium exposures × 2 sexes × 3 replicas; Figure 1b).
Whole-organism metabolic rates were measured using a Sable Systems (Las Vegas, NV, USA) flow-through respirometry system (Lighton, 2008; see supporting information for technical setup and calibration). This system pumps air at a precisely regulated flow rate through a sealed chamber containing animals with a known weight.
Downstream gas analyzers were then used to measure the amount of CO 2 produced and O 2 consumed by the flies; these measures, then, provided the estimates of metabolic parameters. Briefly, the respirometry system was established in stop-flow mode (Lighton, 2008) with each chamber sealed for 60 min and then flushed for 2.5 min.

| Statistical analysis
We explored the data set using descriptive statistics and a one-sample t-test with 95% confidence intervals. To allow proper estimation of the effects of lithium exposures across MNILs we fitted a linear mixedeffects model in which lines were nested within mtDNA types (haplotypes I and II). The model was created using restricted maximum likelihood (REML) to estimate variance. The model included the following fixed effects: the MNIL nested within mtDNA haplotypes (I and II), lithium exposure, and sex. Random effects were replica and cycle.
Because body mass and activity are unwanted sources of variance in metabolic data, we created an experimental and statistical setup that allowed assess and control of activity and body weight (metabolic rate increases with activity, Lighton, 2008) when estimating resting metabolic rate (RMR). All statistical analyses were performed in R (version 3.5.3., 2019).

| LD 50 for lithium sulfate for D. subobscura
Prior respirometry fitness assay data for each lithium dose concentration were gathered by extracting average egg to adult developmental time and egg-to-adult viability (proportion of eclosion). Descriptive statistics for viability is presented in Table 1 and egg to adult developmental time in Table 2. Lithium has been shown to reduce survival with an increased concentration of lithium in substrate, as well as extending the developmental period. An LD 50 concentration for lithium sulfate for D. subobscura corresponds to an L3 concentration in our experiment (0.0006113 mg/ml).

| Lithium has a dose-dependent effect on the whole-organism metabolic rate
In total, we secured respirometric data from three cycles from each of 108 replicate samples. Overall, the amount of CO 2 produced was highly correlated with O 2 consumed (n = 324, r = 0.734, p < 0.001).
Activity was a strong predictor of O 2 (r = 0.155, p = 0.005) but not of CO 2 (r = 0.019, p = 0.739). Moreover, body weight was significantly influenced by O 2 consumption (r = 0.162, p = 0.003) and production of CO 2 (r = 0.382, p = 0.000). We corrected for activity and body weight in the inferential model.
The descriptive statistics for the lithium effects on O 2 consumption and CO 2 production are summarized in Table 3. A one-sample t-test (95% significance interval) revealed a significant difference in MNIL reared on different lithium concentrations for O 2 consumption (df = 323, t = 35.014, p < 0.001) and CO 2 production (df = 323, t = 37.959, p < 0.001).
The resulting inferential model for both O 2 and CO 2 is given in Table 4. As the table shows, lithium had no significant effect on O 2 consumption (df = 2, F = 3.9708, p = 0.1373) but had a significant effect on CO 2 production (df = 2, F = 18.7982, p ≤ 0.001).
The data indicate that the relationship between lithium and metabolic performance is nonlinear. We showed that the L1 concentration reduced the whole organism metabolic rate compared with the control condition (no lithium treatment). In contrast, when compared with L1, the L2 concentration had a significantly positive effect on the whole-organism metabolic rate (Figures 2 and 3).

| MtDNA haplotypes exhibit divergent metabolic efficiency to lithium treatment
While the mtDNA haplotype was not found to have a significant main effect on any of the metabolic parameters (see Table 4), significant lithium exposure effects were noted for CO 2 production for all MNILs nested within the mtDNA haplotypes for both O 2 (df = 8, F = 18.1733, p = 0.0199) and CO 2 (df = 8, F = 36.7605, p < 0.001).

| Lithium influences the whole-organism metabolic rate differently across sexes
Sex did not have a significant main effect on any of the metabolic parameters (see Table 4). Figures 2 and 3

| D ISCUSS I ON
This study provides novel evidence demonstrating that metabolic parameters (O 2 consumption and CO 2 production) were dose-dependently influenced by lithium, with a narrow optimal dosage window with a subtle balance between adverse effects and efficacy. Our experimental design used a unique set of pure mitochondrial lines of D. subobscura (MNIL) that allowed unraveling of the independent effects of two main haplotypes of mtDNA against a common nuclear DNA background. We confirmed in vivo that lithium affects metabolism efficiency (O2 consumption and CO2 production) and that lithium effects are dependent on mitochondrial genetic background, MNIL variability, and sex.
We found that the lower dose (L1) decreased O 2 consumption and CO 2 production. In contrast, doubling the dose (L2) significantly improved metabolic performance in the O 2 and CO 2 response variables. These results are consistent with the suggestion that lithium TA B L E 3 Mean (± SE) of whole organism metabolic rate in females and males across exposures (C, L1, and L2) for MNILs and mtDNA haplotypes. Given are mean amount of VO 2 produced per milligram per hour (upper) and VCO 2 production (lower) may have diverse effects on metabolism and an extremely narrow range for therapeutic concentration (Machado-Viera et al., 2014;Timmer & Sands, 1999). Lithium is known to have a very narrow dose interval and toxic side effects in several organ systems can prove lethal. For instance, both acute and chronic lithium exposures are linked to cardiac arrhythmia, where the proposed mechanism is via disrupted voltage-gated channels (Mehta & Vannozzi, 2017). One animal study, however, indicates that lithium may instead induce oxidative stress in heart tissue (Mezni, Aoua, Khazri, Limam, & Aouani, 2017).
Lithium's influence on the whole-organism metabolic rate was divergent for the MNILs and across two main mtDNA haplotypes (HI and HII) in D. subobscura. An explanation for these results is that lithium influences OXPHOS enzyme activity, which differs across mtDNA haplotypes and may be affected by mtDNA mutations, proving a causal link between mitochondrial genetic effects and metabolic efficiency (Ballard et al., 2007;Kurbalija Novicic et al., 2015).
Our experimental design allowed us to measure a proxy for metabolism using distinct entities-MNIL distributed across the two main and most abundant haplotypes (Kurbalija Novicic et al., 2020). The sequences for all six MNILs across HI and HII haplotype groups have recently been published and showed two consistent differences in the ND5 gene. The first difference was a synonymous substitution within ND5, which has served as the diagnostic basis for previous genotyping using restriction enzymes (Castro et al., 2010;Jelic et al., 2012). The other difference is a line-specific mutation in the ND5 gene that may influence complex I function of the mitochondrial electron transport system, potentially impairing those tissues that require significant energy input (Malfatti et al., 2007;Steffen, Gemperli, Cvetesic, & Steuber, 2010). Although the second polymorphism in ND5 was synonymous, it could have significant effects on codon usage and availability of different tRNA. Additionally, the MNILs showed several SNPs, some of which were nonsynonymous (Kurbalija Novicic et al., 2020). A critical difference was also found in a SNP in 12S rRNA, located in a region homologous to the base of stem 15 in the mammalian 12S rRNA. The region is polymorphic in humans (Jacobs, 2003) and proven to be adaptive polymorphisms (Ruiz-Pesini & Wallace, 2006). The authors predicted the secondary structure of the rRNA, which differs for the specific HI/HII SNP in the 12S rRNA of D. subobscura. This could alter ribosomal protein synthesis and therefore also alter all rate-dependent life-history traits dependent on the abundance of metabolic enzymes.  where females are exposed to higher evolutionary pressure than males, with the later selecting an optimal balance between energy metabolism and reproduction challenges (Maggi & Della Torre, 2018). It has also been shown that females have a lower RMR (than males influencing body mass index (Buchholz, Rafii, & Pencharz, 2001).
Sex differences in the lithium response have not been extensively studied and the few reports that exist are contradictory and inconclusive. Viguera, Tondo, and Baldessarini (2000) reported no significant difference between the sexes in clinical response to lithium treatment of BD and related affective disorders and, more recently, Öhlund et al. (2018) found no difference in the proportion of women and men who had continued lithium than those who had discontinued lithium treatment. Another study, however, showed that the lithium response could be predicted using a sex-specific genetic score for differential expression (Eugene, Masiak, & Eugene, 2018).
For men, it includes such genes as RPS4Y2, a mitochondria-specific ribosomal protein. For women, it encompasses XIST (X-inactive specific transcript), which, in mammals, regulates mitochondrial maintenance across generations and in aging (Tower, 2015). The present results revealed that females, in the majority of pure mitochondrial lines, had a higher metabolic rate than males. Our results are congruent with very recent results (Nagarajan-Radha et al., 2020) that uncovered a male-specific negative correlation across haplotypes, between metabolic rate and longevity. Evolutionary theory could explain these results: the "mother's curse" hypothesis proposes that maternal inheritance of mitochondria will facilitate the accumulation F I G U R E 3 Metabolic rate in females (F) and males (M) and across all mito-nuclear introgression lines (1B, 3B, 5B, 21B, 25B, and 29B), mitochondrial haplotypes (MtHI and MtHII) and exposures to different lithium concentrations (C, L1, and L2). Given are amount of CO 2 produced per milligram per hour, after adjusting for the effects of activity (predicted mass-specific resting metabolic rate) ) of (mtDNA) mutations that are harmful to males but benign or beneficial to females (Frank & Hurst, 1996).
In conclusion, our study adds divergent effects after lithium exposure to a recent body of research documenting functional differences between mtDNA haplotypes under different exposure conditions to lithium. In this study, we documented in vivo that the effects of lithium differ between mtDNA metabolic phenotypes, which represents a fundamental cornerstone for the hypothesis that the mtDNA haplotype may have clinical relevance for lithium treatment. We also show that within-population variation in mtDNA haplotypes during lithium exposure in D. subobscura is associated with sizeable differences between the sexes in the whole-organism metabolic rate. Finally, lithium, within a very narrow dose range, dramatically switched the metabolic rate from lower to higher than baseline. We suggest that future efforts in this system focus on identifying the mechanistic cause for narrow dose lithium effects in pure mitochondrial lines.
This research has a broad range of biological implications for mitochondrial variability in psychiatry as it may contribute to understanding the differing efficacy of lithium as well as predicting the risk of toxic side effects.

DECL AR ATION OF TR ANS PAREN C Y
The authors, reviewers, and editors affirm that in accordance with the policies set by the Journal of Neuroscience Research this manuscript presents an accurate and transparent account of the study being reported and that all critical details describing the methods and results are present.

ACK N OWLED G M ENTS
We