Vitamin B2 enables peroxisome proliferator-activated receptor α regulation of fasting glucose availability

Flavin adenine dinucleotide (FAD) interacts with flavoproteins to mediate oxidation-reduction reactions required for cellular energy demands. Not surprisingly, mutations that alter FAD binding to flavoproteins cause rare inborn errors of metabolism (IEMs) that disrupt liver function and render fasting intolerance, hepatic steatosis, and lipodystrophy. In our study, depleting FAD pools in mice with a vitamin B2 deficient diet (B2D) caused phenotypes associated with organic acidemias and other IEMs, including reduced body weight, hypoglycemia, and fatty liver disease. Integrated discovery approaches revealed B2D tempered fasting activation of target genes for the nuclear receptor PPARα, including those required for gluconeogenesis. Treatment with the PPARα agonist fenofibrate activated the integrated stress response and refilled amino acid substrates to rescue fasting glucose availability and overcome B2D phenotypes. Overall, these findings reveal PPARα governs metabolic responses to FAD availability and nominate its pharmacologic activation as strategies for organic acidemias.


Introduction
Glucose production rate 98 In vivo glucose production was performed, beginning with the insertion of a microcatheter into the 99 jugular vein under anesthesia, followed by 4-5 days rest for complete recovery. Overnight-fasted 100 (16h) conscious mice received a priming bolus of HPLC-purified [3-3 H] glucose (10μCi) and then  Genes in the 95th percentile of a given node consensome were designated high confidence 157 transcriptional targets (HCTs) for that node and used as the input for the HCT intersection analysis 158 using the Bioconductor GeneOverlap analysis package implemented in R. For both consensome 159 and HCT intersection analysis, p values were adjusted for multiple testing using the method of 160 Benjamini and Hochberg to control the FDR as implemented with the p. adjust function in R, to 161 generate q values. Evidence for a transcriptional regulatory relationship between a node and a gene 162 set was inferred from a larger intersection between the gene set and HCTs for a given node or node 163 family than would be expected by chance after FDR correction (q < 0.05). The HCT intersection 164 analysis code has been deposited in the SPP GitHub account at https://github.com/signaling-  The metabolite extraction from the samples was monitored using pooled mouse serum or liver 184 samples and spiked internal standards. The matrix-free internal standards and serum and liver 185 samples were analyzed twice daily. The median coefficient of variation (CV) value for the internal 186 standard compounds was 5%. To address overall process variability, metabolomic studies were 187 augmented to include a set of nine experimental sample technical replicates, which were spaced 188 evenly among the injections for each day.   For lipid separation, 5mL of the lipid extract was injected into a 1.8 mm particle 50 × 2.1mm 234 Acquity HSS UPLC T3 column (Waters). The column heater temperature was set at 55°C. For 235 chromatographic elution, a linear gradient was used over a 20 min total run time, with 60% Solvent 236 A (acetonitrile/water (40:60, v/v) with 10mM ammonium acetate) and 40% Solvent B 237 (acetonitrile/water/isopropanol (10:5:85 v/v) with 10mM ammonium acetate) gradient in the first 238 10 min. The gradient was ramped linearly to 100% Solvent B for 7 min. Then the system was were analyzed separately using relative abundance of peak spectra for the downstream analyses.

254
The identified lipids were quantified by normalizing against their respective internal standard.

280
To determine the relative abundance of FAD and FMN in mouse liver tissue, extracts were

296
To determine the relative abundance of ubiquinone (oxidized CoQ10), ubiquinol (reduced 297 CoQ10), ubiquinone-9 (CoQ9), and ubiquinol-9 (reduced CoQ9) in mouse liver samples, extracts 298 were prepared and analyzed by Thermo Scientific TSQ triple quadrupole mass spectrometer 299 coupled with a Dionex UltiMate 3000 HPLC system. Approximately 20 to 30mg of tissue were 300 pulverized in liquid nitrogen then homogenized with a Precellys Tissue Homogenizer. Coenzymes 301 were extracted with 500µL ice-cold 100% isopropanol. Tissue extracts were vortexed, centrifuged 302 at 17,000xg for 5 min at 4°C, and supernatants were transferred to clean autosampler vials. The  Riboflavin deficiency alters body composition and energy expenditure. 326 In mammals, diet furnishes vitamin B2 to synthesize all the FAD for electron transfer in the 327 mitochondria and redox reactions required for cellular homeostasis (Powers, 2003). Amongst key 328 metabolic organs, ad libitum FAD levels were highest in the liver, heart, and kidney (Figure 1a). 329 To study how FAD depletion influences energy balance, we exposed male mice to vitamin B2-330 deficient (B2D) or control diets for four weeks and performed metabolic phenotyping (Figure 1b). 331 We found 99% vitamin B2 depletion (B2D) was sufficient to reduce liver FAD levels by 70%  Figure S1). These results identify B2 requirements that make FAD available for energy 340 expenditure requirements and body weight in male mice.  Table S3). To determine the effects of B2D on gluconeogenesis in vivo, we as robustly attenuated (Figure 2e) in B2D-exposed mice compared to controls. Hepatic steatosis 375 (Figure 2e) also accompanied B2D effects, presumably due to increased TG (Supplemental 376   Table S4). Accordingly, we hypothesized that altered expression of genes in the liver of B2D  (Shimano et al., 1997). 390 Collectively, our unbiased approach converged metabolic phenotypes and regulatory networks of 391 B2D with those driving NAFLD and macrosteatosis observed in organic acidemias.  B2D (Figure 3a). Likewise, we identified a set of  To explore the physiological intersections between PPARa and B2D, we phenotyped male Ppara 406 whole-body knockout (pKO) mice exposed to control or B2D for one month (Figure 3c). As   (Figure 4b). At the end of the 426 experiment, fasted mice received fenofibrate two hours before blood glucose measurements.

427
Fenofibrate increased blood glucose in both groups of mice far above pre-gavage levels ( Figure   428 4c). When liver histology was examined, we noticed hepatic steatosis was reversed by fenofibrate 429 in B2D mice (Figure 4d). Fenofibrate also decreased hepatic TG and cholesterol in both groups 430 relative to control treatments (Supplemental Table S4).

432
Given the ability of PPARa to regulate flavoprotein gene expression, we pursued additional RNA-433 seq studies to understand the mechanisms that allowed fenofibrate to rescue hypoglycemia in B2D.

485
PPARa activity favors conversion of proteins to provide amino acids as substrates for anabolic 487 processes (Kersten et al., 2001). Our phenotyping analysis of fenofibrate suggested alternative 488 carbon sources might supply the substrates to support glucose production during B2D (Figure 4c), 489 including pyruvate, which recovered blood glucose in B2D to pre-gavage control levels ( Figure   490 6c). Amino acids, such as alanine and serine, are also significant contributors to de novo synthesis  metabolism (Figure 6d). 503 We also observed a gene signature for the integrated stress response (ISR) further supported by 504 increased expression of the master transcription factor regulator Atf4, as well as its key target genes  B2D (Figure 6f).  we demonstrated some of these effects may be mediated through disturbance of nuclear receptor 534 activity and altered glucose availability. One important caveat of these experiments is that the 535 kidney also contributes to glucose production during fasting (Joseph et al., 2000). While our in-

554
Our lipidomics also revealed that B2D caused accumulation of deoxysphingolipids, which also 555 become more abundant in NAFLD from incomplete fat oxidation and accrual of toxic 556 intermediates (Gai et al., 2020). These findings are particularly relevant for discovering new 557 biomarkers for fatty liver disease.