Polyunsaturated Fatty Acid Desaturation Is a Mechanism for Glycolytic NAD⁺ Recycling

SUMMARY

The reactions catalyzed by the delta-5 and −6 desaturases (D5D/D6D), key enzymes responsible for highly unsaturated fatty acid (HUFA) synthesis, regenerate NAD⁺ from NADH. Here, we show that D5D/D6D provide a mechanism for glycolytic NAD⁺ recycling that permits ongoing glycolysis and cell viability when the cytosolic NAD⁺/NADH ratio is reduced, analogous to lactate fermentation. Although lesser in magnitude than lactate production, this desaturase mediated NAD⁺ recycling is acutely adaptive when aerobic respiration is impaired in vivo. Notably, inhibition of either HUFA synthesis or lactate fermentation increases the other, underscoring their interdependence. Consistent with this, a type 2 diabetes risk haplotype in SLC16A11 that reduces pyruvate transport (thus limiting lactate production) increases D5D/D6D activity in vitro and in humans, demonstrating a chronic effect of desaturase mediated NAD⁺ recycling. These findings highlight key biologic roles for D5D/D6D activity independent of their HUFA end-products and expand the current paradigm of glycolytic NAD⁺ regeneration.

In Brief

Kim et al. find that highly unsaturated fatty acid (HUFA) synthesis is a mechanism for glycolytic NAD⁺ recycling, analogous to lactate fermentation. This finding highlights a key biologic role for lipid desaturation independent of the HUFA end-products and provides insight on genetic studies linking HUFA desaturation with human disease.

Graphical Abstract

INTRODUCTION

The delta-5 desaturase (D5D) and delta-6 desaturase (D6D), encoded by FADS1 and FADS2 respectively, are required for the synthesis of highly unsaturated fatty acids (HUFAs) which are then esterified into cellular lipids (Nakamura and Nara, 2004). These enzymes introduce double bonds at fixed positions from the carboxyl end during the successive elongation and desaturation of the essential fatty acid linoleic acid (18:2n-6) to arachidonic acid (20:4n-6), and α-linolenic acid (18:3n-3) to eicosapentaenoic (20:5n-3) and docosahexaenoic (22:6n-3) acid. The FADS1 and FADS2 genes are oriented head-to-head on chromosome 11, and common variants in this region associated with alterations in HUFA content in circulating lipids (Gieger et al., 2008; Rhee et al., 2013) have also been associated with fasting glucose and type 2 diabetes (Dupuis et al., 2010; Fujita et al., 2012), anthropometric traits (Fumagalli et al., 2015), and cancer risk (Wei et al., 2014; Zhang et al., 2014). In addition, we have found that lipid HUFA content can change rapidly, increasing in plasma triacylglycerols (TAGs) within two hours of various glycolytic stimuli, including oral glucose ingestion, sulfonylurea administration, and exercise (Rhee et al., 2011). Why HUFA synthesis is so dynamic, and why genetic variation in this response has such a broad impact on human metabolic and proliferative phenotypes is unknown.

In glycolysis, glucose metabolism is coupled to the reduction of cytosolic nicotinamide adenine dinucleotide (NAD⁺) to NADH. Under aerobic conditions, the transfer of electrons into mitochondria and ultimately to the mitochondrial electron transport chain (ETC) can regenerate NAD⁺, whereas the cytosolic reduction of pyruvate to lactate can regenerate NAD⁺ when mitochondrial respiration is impaired. In either case, the flow of electrons from cytosolic NADH to an available electron acceptor restores cytosolic NAD⁺ and permits ongoing glycolysis. Notably, the reactions catalyzed by D5D and D6D also recycle NADH to NAD⁺. These enzymes contain an N-terminal cytochrome b₅ domain that is required for electron transfer from NADH cytochrome b₅ reductase and the substrate fatty acid to molecular O₂, yielding a product fatty acid with an additional double bond, H₂O, and NAD⁺ (Cho et al., 1999a; Cho et al., 1999b; Napier et al., 2003). In addition, D5D and D6D are endoplasmic reticulum (ER) membrane spanning enzymes, with catalytic domains that face the cytoplasmic pool of NAD⁺ and NADH (Fujiwara et al., 1984; Park et al., 2015; Park et al., 2012; Watanabe et al., 2016).

Here, we test the hypothesis that HUFA production provides a mechanism for glycolytic NAD⁺ recycling, analogous to lactate fermentation. These studies utilize liquid chromatography-mass spectrometry (LC-MS) based profiling methods that readily differentiate cellular lipids on the basis of double bond content (Jain et al., 2014; Rhee et al., 2011). In addition, they capitalize on the recent development of key tools for studying cytosolic NAD⁺ recycling, including a genetically encoded fluorescent sensor of the cytosolic NAD⁺/NADH ratio (Zhao et al., 2015), a recombinant NADH oxidase isolated from Lactobacillus brevis (Titov et al., 2016), and the use of alpha-ketobutyrate (AKB) to drive lactate dehydrogenase (LDH) activity (Sullivan et al., 2015). Together, these studies highlight D5D and D6D as an alternative to LDH for the flow of reducing equivalents generated during glycolysis, in vitro and in vivo. Finally, we demonstrate how this mechanism converges with the recent description of SLC16A11 as a pyruvate transporter, shedding light on the association between type 2 diabetes risk variants in SLC16A11 and altered lipid metabolism (Rusu et al., 2017; Williams et al., 2014).

RESULTS

Inhibition of Aerobic Respiration Increases Cellular HUFA Content

To test if inhibition of aerobic respiration and the subsequent increase in glycolysis impacts cellular lipid HUFA content, we treated mouse renal epithelial cells (IMCD3) with the mitochondrial ETC complex I inhibitor rotenone and profiled lipids using LC-MS. Along with the expected increase in media glucose consumption and lactate secretion (Figure 1A), twenty-four hour rotenone treatment causes a preferential increase in media HUFAs, e.g. arachidonic acid, eicosapentaenoic acid, and docosahexaenoic acid (Student’s t-test, P < 0.001 for all) (Figure 1B). Within cells, the increase in HUFA content is captured as a pattern of greater increases in highly unsaturated lipids (Figures 1C–1F). This pattern is evident among cholesterol esters, phosphatidylcholines, diacylglycerols and most dramatically TAGs (Figure 1F), where HUFA-containing TAGs such as TAG 54:9, TAG 56:10, and TAG 58:11 increase >50-fold following rotenone treatment (Bonferroni adjusted P < 0.05 for all; for each TAG, the first number denotes the total number of carbons and the second number denotes the total number of double bonds in the three acyl chains). Although different combinations of three fatty acyl chains can sum to these carbon and double bond totals, by definition these TAGs contain HUFAs in order to reach these high total double bond counts; we have also confirmed this experimentally using MS/MS fragmentation (Rhee et al., 2011). In contrast to the large increases in HUFA-containing TAGs, which represent a small fraction of total cellular TAGs, total triglycerides only increase 1.8-fold (Student’s t-test, P = 0.001) with rotenone treatment (Figure 1G and S1A). Similar findings were observed with different cell types and other methods of inhibiting aerobic respiration such as oligomycin treatment (mitochondrial ATP synthase inhibitor) and hypoxia (Figure S1B–S1D).

Figure 1. Inhibition of aerobic respiration increases lipid HUFA content in vitro.

(A-G) Effect of 24h rotenone (Rot, 200 nM) treatment of IMCD3 cells on (A) media glucose consumption and lactate secretion; (B) media free fatty acid levels, including linoleic acid (LA), alpha-linolenic acid (α-LA), gamma-linolenic acid (γ-LA), arachidonic acid (AA), eicosapentaenoic acid (EPA), and docosahexaenoic acid (DHA); intracellular (C) cholesterol esters, (D) phosphatidylcholines, (E) diacylglycerols, and (F) triacylglycerols (TAGs) measured by LC-MS, differentiated along the x-axis on the basis of total acyl chain carbon and double bond content; and (G) total TAGs measured by colorimetric assay. Values are means ± SEM; *P < 0.05, **P < 0.01; ***P < 0.001; n = 3.

See also Figure S1.

Inhibition of Aerobic Respiration Increases D5D and D6D Activity

Whereas the decrease in β-oxidation that results from impaired aerobic respiration would be expected to cause some TAG accumulation, the magnitude and pattern of HUFA-containing TAG increases following rotenone treatment suggest a specific increase in HUFA synthesis as well. Indeed, this rotenone induced increase in HUFAs occurs rapidly, within 2 hours (Figure 2A), and is not associated with an increase in Fadsl or Fads2 mRNA or D5D protein levels (Figure 2B). In addition, rotenone does not change the distribution of D5D within the cell, with immunofluorescence demonstrating the enzyme’s diffuse expression throughout the ER (Figures 2C and S2A–S2D). We confirmed that the increase in HUFA-containing lipids following rotenone is due to an increase in D5D and D6D activity by measuring the conversion of isotope-labeled substrate (linoleic acid-¹³C₁₈) to isotope-labeled products (gamma-linolenic acid-¹³C₁₈ and arachidonic acid-¹³C₁₈) (14.5-fold and 4.4-fold increase at 24h, respectively, Student’s t-test, P <0.001 for both) (Figures 2D and 2E). In turn, enzyme inhibition with either CP-24879 (dual D5D and D6D inhibitor) or SC-26196 (D6D inhibitor) prevents the rotenone induced increase in HUFA-containing TAGs (Figures 2F, S2E, S2F, and Table S1) (Obukowicz et al., 1998a; Obukowicz et al., 1998b). Thus, our data demonstrate that inhibition of aerobic respiration increases cellular lipid HUFA content by increasing D5D and D6D activity.

Figure 2. Inhibition of aerobic respiration increases D5D and D6D activity.

(A) Time course of rotenone effect on cellular TAGs.

(B) Top, western blot of D5D in IMCD3 cells following 2, 6, 12, and 24h of rotenone (200 nM) or vehicle. Representative gel from one of two independent experiments. Bottom, qPCR of Fads1 and Fads2 in IMCD3 cells following 6h rotenone (200 nM) or vehicle.

(C) Confocal micrograph of D5D (green), the ER marker reticulon 4 (RTN4, red), or merged image (yellow) in IMCD3 cells following 6h of 200 nM rotenone or vehicle.

(D-E) Isotope-labeled (D) free fatty acid products and (E) substrate of D5D and D6D activity in cultured media of IMCD3 cells following 6h and 24h of rotenone (200 nM) or vehicle.

(F) Intracellular TAGs in IMCD3 cells following 24h of rotenone (200 nM) with or without the desaturase inhibitors SC-26196 (10 μM) or CP-24879 (30 μM). Values denote means ± SEM; ns, P > 0.05.; **P < 0.01; ***P < 0.001; n = 3.

See also Figure S2 and Table S1.

D5D and D6D Activity Increase in Response to Decreased Cytosolic NAD⁺/NADH Ratio

We next investigated the mechanism by which impaired aerobic respiration regulates D5D and D6D activity. The reactions catalyzed by D5D and D6D regenerate NAD⁺ from NADH (Figure 3A). Inhibition of aerobic cellular respiration increases cytosolic NADH levels, decreasing the cytosolic NAD⁺/NADH ratio; by contrast, rotenone does not alter the NADP⁺/NAPDH ratio (Figure 3B) (Zhao et al., 2011). To determine whether this decrease in cytosolic NAD⁺/NADH is responsible for the increase in D5D and D6D activity, we tested the effect of increasing the NAD⁺/NADH ratio using LbNOX, a recombinant NADH oxidase isolated from Lactobacillus brevis (Titov et al., 2016). This enzyme has a very low K_m for O₂, is compartment-specific (i.e. cytosolic), and is highly selective for NADH over NADPH. Using SoNar (Zhao et al., 2015), a genetically encoded fluorescent sensor for cytosolic NAD⁺/NADH, we confirmed that LbNOX expression causes a sustained increase in the cytosolic NAD⁺/NADH ratio (Figure 3C). Further, whereas LbNOX expression has no impact on FADS1 or FADS2 mRNA or D5D protein levels (Figure 3D), its impact on cytosolic NAD⁺/NADH is sufficient to reduce TAG HUFA content (Figure 3E). In cells treated with rotenone, LbNOX expression attenuates both the decrease in cytosolic NAD⁺/NADH (Figure 3C) and increase in HUFA-containing lipids (Figure 3F).

Figure 3. Increasing the cytosolic NAD⁺/NADH ratio with an NADH oxidase reduces D5D and D6D activity.

(A) Schema for D5D and D6D mediated desaturation and NAD⁺ recycling.

(B) Effect of rotenone on total cellular NAD⁺/NADH and NADP⁺/NAPDH ratios. NAD⁺/NADH ratio (left) and NADP⁺/NADPH ratio (right) measured using an enzymatic assay of cell lysates from IMCD3 cells treated with rotenone (200 nM) or vehicle for 6 hours.

(C) Effect of rotenone (100 nM) treatment of HeLa cells, with or without LbNOX expression, on cytosolic NAD⁺/NADH measured using the ratio of fluorescence intensities excited at 485 nm and 430 nm in SoNar expressing cells.

(D) qPCR (left) of FADS1 and FADS2 in HeLa cells with or without LbNOX expression (+/− Dox); western blot (right) of D5D in HeLa cells with or without LbNOX expression (+/− Dox), 6h following rotenone (200 nM) or vehicle.

(E and F) Effect of LbNOX expression on HeLa cell TAGs measured by LC-MS (E) relative to vehicle, or (F) with or without rotenone (100 nM at 6h). Values are means ± SEM; ns, P > 0.05; ***P < 0.001; n = 3 for all experiments except n = 10 for (C).

An alternative method to increase the cytosolic NAD⁺/NADH ratio is supplementation with alpha-ketobutyrate (AKB) or pyruvate (King and Attardi, 1989; Sullivan et al., 2015). Both molecules are substrates for lactate dehydrogenase (LDH), accepting electrons from NADH to regenerate NAD⁺ and produce either alpha-hydroxybutyrate (AHB, from AKB) or lactate (from pyruvate) (Figure 4A). An important feature of AKB is that it acts only as an electron acceptor and does not directly contribute carbon molecules to metabolism, e.g. fatty acid synthesis (Sullivan et al., 2015). In cells treated with AKB, AHB levels increase 110-fold (Student’s t-test, P< 0.001), and this conversion is further augmented to >340-fold (Student’s t-test, P< 0.001) in the presence of rotenone (Figure 4B). In turn, increasing LDH activity with either AKB or pyruvate attenuates both the decrease in cytosolic NAD⁺/NADH ratio (Figures 4C, S3A, and S3B) and the increase in HUFA-containing TAGs following rotenone treatment (Figures 4D and S3C–E); under experimental conditions where the cytosolic NAD⁺/NADH is normalized for the entire duration of the experiment, the effect of rotenone on HUFA TAG synthesis is rescued completely (Figures S3F and S3G). By contrast, AKB supplementation has minimal impact on tricarboxylic acid cycle intermediate levels, which decrease ~90% following treatment with rotenone (Figure 4E). These findings are consistent with AKB modulating HUFA synthesis through its impact on cytosolic NAD⁺ recycling rather than a direct effect on mitochondrial metabolism. In isotope studies, AKB reduces the conversion of linoleic acid-¹³C₁₈ to gamma-linolenic acid-¹³C₁₈ (0.66-fold at 24h, Student’s t-test, P < 0.001) and arachidonic acid-¹³C₁₈ (0.40-fold at 24h, Student’s t-test, P = 0.002) following rotenone, without reducing the overall consumption of linoleic acid-¹³C₁₈ (Figures 4F and 4G).

Figure 4. Modulating the cytosolic NAD⁺/NADH ratio via LDH activity modulates D5D and D6D activity.

Schema for LDH mediated NAD⁺ recycling; alpha-ketobutyrate (AKB), alpha-hydroxybutyrate (AHB), NHI-2 (LDH inhibitor).

(B-G) Effect of rotenone (200 nM) treatment of IMCD3 cells, with or without AKB (ImM), on (B) intracellular AHB at 24h; (C) cytosolic NAD⁺/NADH measured with SoNar; (D) intracellular TAGs at 24h; (E) intracellular citrate, isocitrate, and succinate at 24h; and (F and G) isotope-labeled free fatty acids in media at 6h and 24h.

(H-I) Effect of NHI-2 (30 μM) and rotenone (100 nM) treatment of IMCD3 cells on (H) cytosolic NAD⁺/NADH measured with SoNar, where box plot shows upper and lower quartiles and whiskers show maximum and minimum values; and (I) intracellular TAGs at 24h. Values denote means ± SEM; ns, P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001; n = 3 for all experiments except n = 10 for (C, H).

See also Figure S3.

In contrast to LDH activation with AKB or pyruvate, treatment with the LDH inhibitor NHI-2 decreases the cytosolic NAD⁺/NADH ratio (Figure 4H) and significantly increases HUFA-containing TAGs (Figure 4I) (Calvaresi et al., 2013). The combination of NHI-2 and rotenone results in even greater increases in HUFA-containing TAGs than either alone, e.g. with Bonferroni adjusted P < 0.05 for TAG 54:9, TAG 56:10, and TAG 58:11 (Figure 4I). Collectively, these studies of ETC inhibition, NADH oxidase expression, and LDH activation and inhibition demonstrate the dynamic relationship between the cytosolic NAD⁺/NADH ratio and cellular HUFA synthesis.

Modulation of D5D and D6D Alters the Cytosolic NAD⁺/NADH Ratio, Lactate Production, and Cell Proliferation

Because our data implicate D5D and D6D in glycolytic NAD⁺ recycling, we next tested the effect of modulating D5D and D6D on cell metabolism. Transient overexpression of D5D and D6D increases the cytosolic NAD⁺/NADH ratio (Student’s t-test, P< 0.001 for both at 48 hours after transfection), whereas transient knock-down of D5D and D6D reduces the cytosolic NAD⁺/NADH ratio (Student’s t-test, P< 0.001 for both at 48 hours after transfection) (Figures 5A and S4A–S4D). Pharmacologic inhibition with either SC-26196 or CP-24879 also reduces the cytosolic NAD⁺/NADH ratio (Student’s t-test, P< 0.001 at 24 hours for both compounds), with a more significant reduction with the dual D5D/D6D inhibitor CP-24879 (Figure 5B). Blocking NADH oxidation with D5D/D6D inhibition would be expected to divert reducing equivalents elsewhere. Following treatment with CP-24789, we observe an increase in cellular ROS (Figure S4E) as well as a dose-dependent increase in media lactate and decrease in media pyruvate (Figure 5C). The latter finding demonstrates an increase in LDH mediated pyruvate consumption that partially compensates for the loss of desaturase mediated NAD⁺ regeneration. Consistent with this, augmenting LDH mediated NAD⁺ regeneration with AKB prevents the decrease in media pyruvate/lactate following CP-24879 treatment (Student’s t-test, P< 0.001) (Figure 5D).

Figure 5. D5D and D6D inhibition reduces cytosolic NAD⁺/NADH, increases LDH activity, and impairs cell proliferation.

(A-B) Cytosolic NAD⁺/NADH measured with SoNar in (A) HeLa cells transfected with FADS1 or FADS2 cDNA or shRNA targeting FADS1 or FADS2; and (B) IMCD3 cells treated with rotenone (200 nM) or the desaturase inhibitors SC-26196 (10 μM) and CP-24879 (30 μM).

(C-D) Media lactate, pyruvate, and/or pyruvate/lactate ratio from IMCD3 cells treated with (C) rotenone (200 nM) or increasing concentrations of CP-24879 (CP); and (D) CP-24789 (30 μM) with or without AKB (1 mM).

(E-H) Cell proliferation at 3 days for (E) IMCD3 cells treated with increasing concentrations of CP-24789; (F) IMCD3 cells treated with vehicle, CP-24789 (30 μM), CP-24789 (30 μM) and pyruvate (1 mM), CP-24789 (30 μM) and AKB (1 mM), rotenone (200 nM), and rotenone (200 nM) and AKB (1 mM); (G) HeLa cells transfected with shRNA targeting FADS1 and FADS2 or control; and (H) IMCD3 cells treated with CP-24789 (30 μM) with AKB (1 mM) and/or a mixture containing 10 μM each of arachidonic acid (AA), eicosapentaenoic acid (EPA), and docosahexaenoic acid (DHA).

(I) Schema for D5D and D6D activity and NAD⁺ recycling in relation to glycolysis and lactate fermentation. Values are means ± SEM; ns, P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001; n = 20–30 for (A), n = 10 for (B), n = 3 for (C, D), and n = 4 for (E-H).

See also Figure S4.

In addition to decreasing the cytosolic NAD⁺/NADH ratio, CP-24879 causes a dose-dependent decrease in cell proliferation (Student’s t-test, P = 0.002 at 3μM, P< 0.001 at 10 μM, P< 0.001 at 30μM) (Figure 5E). The deleterious impact of D5D/D6D inhibition on cell growth is mitigated by relatively richer media conditions (Figure S4F and Table S2), an effect we hypothesize is due in part to greater pyruvate content (there is no difference in HUFA end- products between the different media). Indeed, co-treatment with either AKB or pyruvate significantly attenuates the antiproliferative effect of desaturase inhibition (Figure 5F). Similarly, AKB co-treatment rescues the decrease in cell proliferation observed with transient knockdown of D5D and D6D (Figure 5G). Importantly, AKB has a significantly greater salutary effect on cell proliferation than a mixture of arachidonic acid, eicosapentaenoic acid and docosahexaenoic acid in CP-24879 treated cells, highlighting the relatively greater importance of NAD⁺ regeneration than HUFA end-products for mediating the effects of D5D and D6D activity on cell growth (Figure 5H). Taken together, our findings show that the interaction between cytosolic NAD⁺/NADH and D5D and D6D activity is bidirectional, whereby decreasing cytosolic NAD⁺/NADH increases HUFA synthesis and decreasing HUFA synthesis decreases the cytosolic NAD⁺/NADH ratio, increases lactate production, and reduces cell growth (Figure 5I).

D5D and D6D Activity is Adaptive During Acute Inhibition of Aerobic Respiration In Vivo

Given the impact of desaturase mediated NAD⁺ regeneration in vitro, we next examined its role in vivo, using rotenone to inhibit aerobic respiration in C57BL/6J mice and then profiling tissue homogenates 10 hours later (Figure S5A). We found that rotenone increases TAGs in kidney, liver, and skeletal muscle (ANOVA, P < 0.0001 for each), but not heart, with greater increases in HUFA-containing TAGs (Figure 6A) than total triglycerides (Figure 6B). Notably, no changes were observed in tissue lactic acid levels (Figure 6C). The tissue specificity of these findings may in part be attributable to relative levels of desaturase expression as D5D is most strongly expressed in kidney and liver (Figure 6D). Rotenone does not impact D5D protein levels (Figure 6E), whereas Fadsl and Fads2 mRNA levels decrease following rotenone treatment (Figure 6F), consistent with the known feedback inhibition of HUFA end-products on gene expression.

Figure 6. Acute increase in HUFA synthesis is adaptive when aerobic respiration is impaired in vivo.

(A-C) Effect of 10 h rotenone (2.8 mg/kg, intraperitoneal) treatment in 11-week old C57BL/6J mice on (A) heart, kidney, liver, and skeletal muscle (S. Mus) tissue TAGs measured by LC-MS, (B)total tissue TAGs measured by colorimetric assay, (C) tissue lactate measured by LC-MS.

(D and E) Western blot of D5D for (D) different tissue samples from one vehicle (V, Veh) and one rotenone (R, Rot)-treated mouse and (E) kidney homogenates from all vehicle and rotenone-treated mice. Representative gels from one of two independent experiments.

(F) Kidney tissue Fadsl and Fads2 mRNA from all vehicle and rotenone-treated mice. Values are means ± SEM; ns, P> 0.05; *P <0.05; **P < 0.01; n = 3 mice per group.

(G-J) Effect of co-treatment of CP-24879 (5 mg/kg, intraperitoneal) and rotenone (1.7 mg/kg, intraperitoneal) in 11-week old C57BL/6J mice on (G) NAD⁺/NADH ratio in kidney tissue; (H) renal tubular damage assessed by H&E staining (asterisk: protein casts, arrow head: apoptotic epithelial cells, arrow: tubular necrosis); (I) plasma creatinine and blood urea nitrogen (BUN); and (J) plasma aminotransferases (ALT and AST). Values are means ± SEM; *P < 0.05; **P < 0.01; ***P < 0.001; n = 3 mice per group.

See also Figure S5 and S6.

A similar increase in tissue lipid HUFA content is observed in kidney and liver homogenates from mice exposed to 8% hypoxia versus normoxia (Figures S5B–S5F) and we have also observed this pattern of HUFA rich TAG accumulation in mouse kidney tissue following ischemia reperfusion injury (Tran et al., 2016). Although cellular lipid accumulation in these contexts has traditionally been viewed as deleterious, our studies suggest that acute increases in HUFA synthesis may represent an adaptive mechanism to support cellular metabolism and viability. Indeed, co-treatment of animals with CP-24879 and rotenone results in a lower tissue NAD⁺/NADH ratio and significantly greater kidney injury than either CP-24879 or rotenone alone (Figures S5A and 6G–6I); a trend for similar findings was observed with liver injury as well (Figure 6J). These findings show how desaturase mediated NAD⁺ recycling supports tissue viability when aerobic respiration is acutely impaired. Finally, to extend our findings to a non-injury in vivo model, we show that metformin significantly increases HUFA-containing TAGs in kidney and liver at a dose known to reduce tissue NAD⁺ levels (Figure S6A) (Gui et al., 2016); these results were further corroborated in vitro, where metformin reduces the cytosolic NAD⁺/NADH ratio and increases cellular HUFA TAGs (Figures S6B and S6C).

D5D and D6D Activity are Chronically Increased in Individuals with the Type 2 Diabetes SLC16A11 Risk Haplotype

Inhibition of aerobic respiration results in large increases in glucose consumption and a marked reduction in the cytosolic NAD⁺/NADH ratio. To assess the interaction between cytosolic NAD⁺/NADH and D5D and D6D activity in a more physiologic context, we examined cells exposed to high (25 mM) versus normal glucose (5.5 mM) media for 24 hours. Exposure to high glucose results in a 3% decrease in the cytosolic NAD⁺/NADH ratio (Student’s t-test, P = 0.017) and increase in HUFA containing TAGs (ANOVA, P < 0.0001), although the changes are of lesser magnitude compared to rotenone treatment (Figures S7A and S7B). Further, LbNOX expression attenuates both the decrease in cytosolic NAD⁺/NADH (Figure S7A) and increase in HUFA-containing lipids (Figure S7C) in cells exposed to high glucose.

Recently, a type 2 diabetes-associated haplotype in SLC16A11 was identified that explains ~20% of the increased type 2 diabetes prevalence in Mexico (Williams et al., 2014). These studies showed that SLC16A11 functions as a pyruvate transporter, and that by lowering both SLC16A11 expression and localization to the cell membrane, type 2 diabetes risk variants reduce SLC16A11 transport activity (Rusu et al., 2017). Genetic perturbation of SLC16A11 induces rapid changes in cellular lipid levels, but the mechanism connecting plasma membrane solute transport to lipid metabolism is unclear. Here, we tested whether reduced pyruvate transport attributable to the type 2 diabetes risk variants reduces cytosolic NAD⁺/NADH ratio, as occurs with NHI-2 mediated LDH inhibition. We expressed SLC16A11 in HEK293 cells co-transfected with SoNar and compared the non-risk SLC16A11 reference protein to the type 2 diabetes-associated protein that contains four missense variants (V113I, D127G, G340S, and P443T) and one silent variant (L187L) (Figures 7A and 7B). We find that the type 2 diabetes variant-associated reduction in pyruvate transport is associated with a 4% fall in the cytosolic NAD⁺/NADH ratio (Student’s t-test, P < 0.001) (Figure 7C), as well as an increase in TAG HUFA synthesis that can be reversed with pyruvate supplementation (Figure 7D). To extend these findings to humans, we used LC-MS to profile plasma from 677 individuals in the Mexico City Diabetes Study who had been genotyped at the SLC16A11 locus; at the time of plasma sampling, no study subjects had developed diabetes mellitus. We find a clear pattern of increased TAG HUFA content among the 364 carriers of the SLC16A11 risk haplotype compared to 313 controls (Figure 7E), although the magnitude of differences is attenuated relative to the in vitro comparison. By contrast, there was no difference in body mass index, fasting glucose, total cholesterol, high density cholesterol, or total triglycerides (Figure 7F). In conjunction with our cellular studies of LDH activation, LDH inhibition, and SLC16A11 expression, these human data underscore the biochemical interplay between lactate and HUFA production during glycolysis and provide a mechanism for how a reduction in pyruvate transport across the cell membrane can directly and chronically modulate lipid metabolism (Figure 7G).

Figure 7. The SLC16A11 diabetes risk haplotype reduces cytosolic NAD⁺/NADH ratio and increases HUFA synthesis in humans.

(A) Western blot of SLC16A11^T2D-V5 and SLC16A11^REF-V5 in HEK293 cells, representative gel from one of two independent experiments.

(B) qPCR of SLC16A11 in HEK293 cells expressing empty vector, SLC16A11^T2D-V5, or SLC16A11^ref-V5.

(C) Cytosolic NAD+/NADH measured with SoNar in HEK293 cells expressing either SLC16A11^T2D-V5 or SLC16A11^REF-V5.

(D) Intracellular TAGs in HEK293 cells expressing SLC16A11^T2D-V5 or SLC16A11^REF-V5 (red), or SLC16A11^T2D-V5 with or without 1mM pyruvate supplementation (blue). Values are means ± SEM; ***P < 0.001; n = 3 for (B, D), and n = 40 for (C).

(E) Plasma TAGs among carriers of the SLC16A11 risk haplotype (n=364) relative to controls (n=313). Values are means; *P<0.05; **P < 0.01.

(F) Mexico City Diabetes Study clinical characteristics at time of lipid profiling

(G) Schema for interaction between SLC16A11, cytosolic NAD⁺/NADH, and HUFA synthesis.

See also Figure S7.

DISCUSSION

In addition to modulating membrane fluidity, HUFAs can be released from membrane lipids and then converted to eicosanoids and other bioactive molecules that play diverse roles in health and disease (Funk, 2001; Wassall and Stillwell, 2009). As such, the significance of D5D and D6D activity have largely been viewed in terms of the biologic actions of their enzymatic end- products. Dietary intake and transcriptional control of FADS1 and FADS2 expression are established determinants of cellular HUFA content (Nakamura and Nara, 2004). Here, we show that changes in cytosolic NAD⁺ and NADH redox states also influence D5D and D6D activity, establishing a bidirectional link between glycolysis and polyunsaturated fatty acid desaturation. These findings alter the existing paradigm of NAD⁺ regeneration in glycolysis, and highlight a key biologic role for D5D and D6D action independent of their end-products.

Interest in FADS1 and FADS2 has been motivated by recent genome-wide association studies implicating desaturase activity in glycemic, growth, and proliferative phenotypes. More specifically, common variants that are associated with reduced desaturase expression are associated with decreased HUFA content in circulating lipids (Gieger et al., 2008; Rhee et al., 2013). These same alleles result in lower fasting glucose (Dupuis et al., 2010), lower risk of type 2 diabetes (Fujita et al., 2012), decreased height and weight among Inuits (Fumagalli et al.,2015), and decreased risk of colon cancer and laryngeal cancer in East Asians (Wei et al., 2014; Zhang et al., 2014). Some of these associations may be attributable to reduced synthesis of HUFA end-products; for example, reduced eicosanoid availability for cell signaling could have anti-proliferative effects (Wang and Dubois, 2010). Alternatively, our results raise the possibility that reduced D5D and D6D activity reduces cell proliferation by reducing the cytosolic NAD⁺/NADH ratio, as we show that AKB, but not a cocktail of HUFAs, reverses growth inhibition by dual D5D and D6D inhibition. Whether desaturase mediated NAD⁺ recycling plays a role in Warburg (aerobic) glucose metabolism in cancer cells requires further investigation (Koppenol et al., 2011)—interestingly, recent studies have shown that some cancer cells are net consumers of glucose and lactate, raising the possibility that alternative mechanisms to support glycolytic NAD⁺ recycling may be required (Faubert et al., 2017). Ultimately, both enzymatic activity and end-products are likely to be pertinent to how HUFA synthesis impacts cellular metabolism, and we acknowledge that our studies do not disentangle their relative contribution to the FADS-related phenotypes established by human genetics.

Interest in FADS1 and FADS2 has been further enhanced by lipidomic studies of human physiology. Using an earlier iteration of the platform utilized herein, we profiled plasma from individuals before and after oral glucose tolerance testing, oral sulfonylurea administration, and exercise treadmill testing (Rhee et al., 2011), using each individual as his or her own biologic control. In each case, we observed an increase in the HUFA content of circulating TAGs within two hours, before any changes in FADS1 or FADS2 expression might occur. These observations raised the question of why diverse glycolytic stimuli would acutely increase D5D and D6D activity, with particular interest in the liver where lipoproteins destined for release into plasma are assembled (Quehenberger and Dennis, 2011). Our findings here provide an explanation for the dynamic changes observed in humans, connecting glycolysis to HUFA synthesis through cytosolic NAD⁺ and NADH. We observe increases in D5D and D6D activity, with modest increases when glycolysis is stimulated with high glucose, and marked increases when glycolysis is stimulated with inhibition of mitochondrial respiration, including in the liver and kidney in vivo. D5D and D6D activity also increases with LDH inhibition, demonstrating that it is the cytosolic NAD⁺/NADH ratio, rather than glucose utilization per se, that triggers increased desaturase activity. The combined inhibition of mitochondrial respiration and LDH yields the most dramatic increases in cellular HUFA levels, outlining the capacity of this system for accommodating the flow of reducing equivalents diverted from their canonical pathways.

Among the tissues we examined, the expression of D5D and the subsequent increase in HUFA levels in mice treated with rotenone or exposed to hypoxia was greatest in liver and kidney, with no accompanying change in tissue lactate. Inhibition of desaturase activity further lowered the tissue NAD⁺/NADH ratio and enhanced the renal toxicity of rotenone (with a trend for increased hepatic toxicity). Thus, the accumulation of intracellular HUFA may be acutely adaptive by facilitating ongoing glycolysis when aerobic respiration is impaired. The relatively higher expression of D5D and D6D in kidney and liver is notable, as these two organs are responsible for the majority of whole-body lactate clearance under normal conditions, and thus have a high threshold for net lactate production (Kraut and Madias, 2014) (Bellomo, 2002).

The impact of desaturase mediated NAD⁺ regeneration is likely to be context dependent. As noted, prior GWAS have outlined the potential salutary consequences of decreased desaturase expression and HUFA synthesis, including lower fasting glucose and type 2 diabetes risk; conversely, a long term increase in desaturase activity contributes to increased metabolic disease. Here, we find that the SLC16A11 risk haplotype—associated with a significant increase in type 2 diabetes among Mexicans, and shown to reduce the protein’s pyruvate transport capacity at the plasma membrane by ~50%—results in a lower cytosolic NAD⁺/NADH ratio, with a subsequent increase in HUFA synthesis (Halestrap, 2013). This phenotype can be reversed with pyruvate supplementation, although it is unclear if the added pyruvate enters cells through SLC16A11 or through a different monocarboxylate transporter. Remarkably, the increase in TAG HUFA content observed in vitro is recapitulated in plasma obtained from Mexican individuals who have the SLC16A11 risk haplotype compared to controls. These results provide mechanistic insight into how coding variants in SLC16A11 alter lipid metabolism, i.e. that reduced LDH activity attributable to reduced pyruvate availability lowers the cytosolic NAD⁺/NADH ratio, triggering a sustained increase in HUFA synthesis. Why increased D5D and D6D action promotes type 2 diabetes requires more investigation, but may relate to an increase in intracellular TAG accumulation, for example in the liver or the beta cell. Whether the increase in HUFA content itself, for example through an effect on eicosanoid production or cell signaling, confers some harm in this context also warrants consideration.

Our results raise several questions. Stearoyl-CoA desaturase (SCD) performs a similar chemical reaction as D5D and D6D, but acts on saturated fatty acids, yielding monounsaturated fatty acids such as palmitoleic and oleic acid. However, our lipidomics data do not demonstrate an increase in SCD activity in response to inhibition of mitochondrial respiration or LDH. More work is required to explore the differential response of SCD versus D5D and D6D—however, it is notable that SCD lacks an N-terminal cytochrome b5-like domain present in D5D and D6D, and that NADPH, rather than NADH, is believed to be the major electron donor for the SCD-mediated enzymatic reaction (Guillou et al., 2004). D5D and D6D are responsible for the synthesis of both ω−6 and ω−3 HUFA. Increased enzymatic activity triggered by a reduction in the cytosolic NAD⁺/NADH ratio would not be expected to have a differential impact on their relative production, and indeed, we found increases of both, best appreciated with the increase in individual free fatty acids or cholesterol esters that contain a single acyl chain. However, more formal study of the relative production of ω−6 and ω−3 HUFA is warranted, as these molecules have disparate and sometimes antagonistic effects, for example in inflammation and hemostasis (Zarate et al., 2017).

Also of interest is the stoichiometry of glycolytic NAD⁺ regeneration via mitochondrial respiration, lactate fermentation, and HUFA synthesis. Clearly, this will depend on the experimental settings, including glucose, pyruvate, and polyunsaturated fatty acid availability, as well as glycolytic capacity and desaturase expression in the chosen cell line. For a given set of conditions, measurement of glucose consumption and end-products alone is not sufficient. As demonstrated by our lipidomics data, HUFA end-products are incorporated diffusely across cellular lipids, including some that are not readily measured, as well as secreted into media. Our results show that NADH produced during glycolysis can drive HUFA synthesis, but the entire pool of cytosolic NADH can contribute reducing equivalents, including cytosolic NADH generated from non-glycolytic dehydrogenase reactions. Finally, HUFA synthesis requires the alternating action of both desaturases and elongases (Jakobsson et al., 2006). Fatty acid elongation also requires NADH as a substrate, regenerating NAD⁺; thus it may be more appropriate to view the entirety of HUFA synthesis, including desaturation and elongation, as a mechanism of cytosolic NAD+ recycling. However, because average daily intake of the HUFA precursors linoleic and alpha-linolenic acids is ~0.07 mol/day for adults (U.S. Department of Agriculture, Agricultural Research Service, 2018), this mechanism can only account for 0.03~0.08 mol/day of NAD⁺ regeneration, assuming 16–36% of ingested lipid is converted to HUFA end-products (Burdge et al., 2002; Burdge and Wootton, 2002). By contrast, whole body lactate turnover is likely >20–80 fold greater (Bergman et al., 1999; Katz et al., 1993). Nevertheless, our data support a physiologically relevant role for desaturase mediated NAD⁺ recycling, particularly in select tissues such as kidney and liver, and in the context of chronic changes in cytosolic redox status as may occur with genetic variants in FADS1–3 or SLC16A11.

In sum, our findings advance knowledge of metabolism along several axes. First, they show that polyunsaturated fatty acid desaturation increases in response to a reduction in the cytosolic NAD⁺/NADH ratio. Second, they characterize this response as a mechanism for glycolytic NAD⁺ recycling that operates in parallel with lactate fermentation and is acutely adaptive in vivo when aerobic respiration is impaired. In this role, the importance of polyunsaturated fatty acid desaturation expands beyond the downstream biologic roles of HUFA end-products. Third, our results provide mechanistic insight on how variants in the monocarboxylate transporter SLC16A11 may impact cellular HUFA production, and thus contribute to the pathogenesis of type 2 diabetes. And finally, our results provide insight on human genetic studies linking HUFA desaturation with glycemic, growth, and cancer phenotypes as well as human physiologic studies demonstrating rapid changes in plasma HUFA levels.

Limitations of Study

Although we use established knowledge about PUFA consumption and lactate turnover to outline the greater magnitude of lactate fermentation, we do not systematically compare and quantitate the two processes in vivo. More work is required to fully characterize in which tissues and under what conditions HUFA synthesis makes a substantial contribution to glycolytic NAD⁺ recycling and to understand the functional consequences of increasing cellular HUFA content via this mechanism.

STAR Methods

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Eugene P. Rhee (eprhee@partners.org). Certain materials are shared with academic and non-profit research organizations for research only under an MTA.

EXPERIMENTAL MODELS AND SUBJECT DETAILS

Cell Lines and Cell Culture

Mouse inner medullary collecting duct (IMCD3) and human (female) embryonic kidney (HEK) 293 cells were obtained from ATCC. IMCD3 cells were maintained in DMEM F-12 medium (30–2006, ATCC) and HEK293 cells were maintained in MEM medium (30–2003, ATCC), both containing 10% fetal bovine serum (FBS, HyClone) and 100 units/ml penicillin/streptomycin (Life Technologies). Doxycycline-inducible human (female) cervical carcinoma HeLa-LBNOX cells (provided by D. Titov and V. Mootha) were cultured in 10% FBS and 100 units/ml penicillin/streptomycin-containing DMEM medium (D5921, Sigma-Aldrich) supplemented with 1 mM pyruvate, 4 mM L-glutamine, 200 μg/mL geneticin and 1 μg/mL puromycin (all from Thermo Fisher Scientific). A week before experiments, geneticin and puromycin were removed from the culture media. All cells were grown at 37°C in a humidified atmosphere with 5% CO₂. Cells with fewer than 20 passages were used.

Animal Studies

All studies were approved by the Institutional Animal Care and Use Committee of the Massachusetts General Hospital and conducted under their guidelines. Ten-week-old C57BL/6J male mice were purchased from the Jackson Laboratory. Mice were housed in a controlled environment with a 12/12-h light-dark cycle (7am to 7pm) and fed a normal chow diet, Prolab^® Isopro^® RMH 3000 (LabDiet). Mice were used for experiments one week after arrival. For rotenone treatment, mice underwent i.p. injection with 2.8 mg/kg rotenone (1.4 mg/ml stock) in DMSO or DMSO vehicle (n=3, each group). In another set of experiments, mice were co-treated with a single dose of rotenone (1.7 mg/kg, i.p.) and multiple doses of CP-24879 (5 mg/kg, i.p.) for 2 days (n=3, each group). Following euthanasia 10 hours after rotenone treatment, tissues were collected and immediately snap frozen in liquid nitrogen. Chow was removed 4 hours prior to sacrifice. For metformin treatment, mice were treated with metformin (500 mg/kg, i.p.) in saline or saline vehicle (n=3, each group) and chow was removed. Following euthanasia 4 hours after metformin treatment, tissues were collected and immediately snap frozen in liquid nitrogen. For hypoxia, mice were divided into two groups (n=4, each group) and were exposed to hypoxia (8% O₂) or normoxia (21% O₂) at ambient sea-level pressure as previously described (Jain et al., 2016). In brief, mice were placed in 60 liter plexiglass chambers. Each chamber was ventilated from an inlet using a continuous mixture of nitrogen and air. The mixture was adjusted to obtain the desired oxygen concentration using an oxygen analyzer (MiniOx I) placed at the outlet of the chamber. The oxygen analyzer was calibrated using a reference tank with a concentration of 8.55% O₂ (Airgas). The flow of the mixture was 5 l/min and was adjusted to maintain a CO₂ concentration below 1000ppm inside the chamber. The CO₂ concentration in the chamber as well as the temperature and the humidity were monitored continuously using a dedicated infrared CO2 analyzer, thermometer, and humidity meter (Extech CO200 Monitor, Extech Instruments). To ensure fasting status at the time of sample collection, the chamber was briefly opened at 4 hours prior to sacrifice and chow was removed. Following euthanasia, tissues were collected and immediately snap frozen in liquid nitrogen.

Human Studies

Study participants in the Mexico City Diabetes Study, as well as genotyping using the Illumina OMNI2.5 array and exome-sequencing, have been previously described (Estrada et al., 2014; Williams et al., 2014). Informed consent was obtained from all subjects, and the study was approved by the Comité de Etica e Investigatión committee at The Centro de Estudios en Diabetes. In sum, plasma from a total of 677 individuals with genotyping, including 364 carriers of the SLC16A11 risk haplotype (mean age 52.6 ± 7.6; 40.4 % male) and 313 controls (mean age 52.6 ± 7.6; 37.4 % male), underwent lipid profiling as described below. No individuals at the time of lipid profiling had type 2 diabetes. An analysis of the influence of sex on the association between SLC16A11 status and TAG pattern was not performed because sex is not known to influence the association between SLC16A11 status and diabetes risk

METHOD DETAILS

In Vitro Experimental Conditions

Inhibition of mitochondrial respiration

IMCD3 and HEK293 cells were seeded at 200,000 cells per well in 6-well plates and incubated overnight. The following day, HEK293 cells were washed once with 2 mL of pyruvate-free MEM (M5650, Sigma-Aldrich) and then the media was changed to 2 mL of pyruvate-free MEM containing 10% dialyzed FBS (Thermo Fisher Scientific), penicillin/streptomycin and 2 mM L-glutamine premixed with either 200 mM rotenone or 0.1% DMSO. Similarly, IMCD3 cells were washed once with 2 mL of pyruvate-free DMEM (D5671, Sigma-Aldrich) and then the media was changed to 2 mL of pyruvate-free DMEM containing 10% dialyzed FBS, penicillin/streptomycin and 2 mM L-glutamine premixed with 200 nM rotenone, 5 nM oligomycin, 2 mM metformin, or 0.1% DMSO. For D5D/D6D inhibition, media was changed to2 mL of pyruvate-free DMEM containing 10% dialyzed FBS, penicillin/streptomycin and 2 mM L-glutamine premixed with 200 nM rotenone ± 10 μM SC-26196 or 200 nM rotenone ± 30 μM CP-24879. For modulation of cytosolic NAD⁺/NADH, media was changed to 2 mL of pyruvate- free DMEM containing 10% dialyzed FBS and penicillin/streptomycin premixed with 200 nM rotenone ± 1 mM alpha-ketobutyrate (AKB) or 200 nM rotenone ± 1 mM pyruvate. For LDH inhibition, media was changed to 2 mL of pyruvate-free DMEM containing 10% dialyzed FBS, penicillin/streptomycin and 2 mM L-glutamine premixed with 30 mM NHI-2 ± 100 nM rotenone. For all experiments, cells and media were collected for analysis at 24 hours unless otherwise specified.

Hypoxia

IMCD3 cells were seeded at 200,000 cells per well in 6-well plates and incubated overnight. The following day, cells were washed once with 2 mL of pyruvate-free DMEM and media was changed to 2 mL of pyruvate-free DMEM containing 10% dialyzed FBS, penicillin/streptomycin and 2 mM L-glutamine. For the induction of hypoxia, one plate of cells was placed in a hypoxia chamber (STEMCELL Technologies) containing a 1% O₂ gas mixture and incubated at 37°C. At 1 hour, the chamber was re-gassed to eliminate O₂ contained in the media or trapped in the plastic ware; a petri dish with 20 mL of water was placed inside the chamber to prevent excessive evaporation of culture medium. Control cells were placed outside of the hypoxia chamber, but within the same incubator. Cells were collected for analysis at 24 hours.

LbNOX expression

HeLa-LbNOX cells were seeded at 100,000 cells per well in 6-well plates and incubated overnight. The following day, cells were washed once with 2 mL of pyruvate-free DMEM and media was changed to 2 mL of pyruvate-free DMEM containing 10% dialyzed FBS, penicillin/streptomycin and 4 mM L-glutamine with or without 300 ng/mL doxycycline. Twenty-four hours after doxycycline addition, media was exchanged to 2 mL of pyruvate-free DMEM containing 10% dialyzed FBS, penicillin/streptomycin and 4 mM L-glutamine premixed with 100 nM rotenone ± 300 ng/mL doxycycline or 0.1% DMSO ± 300 ng/mL doxycycline. Cells were collected for analysis at 6 hours.

Dose-effect of CP-24879

IMCD3 cells were seeded at 200,000 cells per well in 6-well plates and incubated overnight. The following day, cells were washed once with 2 mL of pyruvate-free DMEM and media was changed to 2 mL of pyruvate-free DMEM containing 10% dialyzed FBS, penicillin/streptomycin and 2 mM L-glutamine premixed with 3 μM CP-24879, 10 μM CP-24879, 30 μM CP-24879, 200 nM rotenone, or 0.1% DMSO. Media was collected for analysis at 24 hours.

Transient transfection

HeLa and HEK293 cells were seeded at 100,000 and 300,000 cells per well in 6-well plates, respectively. The following day, media was replaced with fresh antibiotic-free MEM and cells were transfected with FuGENE HD transfection reagent (Promega). More specifically, cells were incubated with 1 μg of plasmid DNA encoding FADS1, FADS2, shFADS1, shFADS2, SLC16A11^REF, or SLC16A11^T2D and 3 μL of FuGENE HD transfection reagent for 24 hours (human FADS cDNA and shRNA expression vectors were purchased from OriGene Technologies). After transfection, cells were washed once with 2 mL of pyruvate-free MEM and then the media was changed to 2 mL of pyruvate-free MEM containing 10% dialyzed FBS and 2 mM L-glutamine. Cells were collected for analysis at 24 hours (2 days after the start of transfection).

Conversion of linoleic acid-¹³C₁₈ to gamma-linolenic acid-¹³C₁₈ and arachidonic acid-¹³C₁₈

IMCD3 cells were seeded at 200,000 cells per well in 6-well plates and incubated overnight. The following day, cells were washed twice with 2 mL of pyruvate-free DMEM and media was changed to 2 mL of pyruvate-free DMEM containing 10 mM linoleic acid-¹³C₁₈, 1 mg/mL fatty acid-free bovine serum albumin (Sigma-Aldrich), penicillin/streptomycin and 2 mM L-glutamine premixed with 0.1% DMSO ± 1 mM AKB, 200 nM rotenone ± 1 mM AKB, or 30 μM CP-24879. Media was collected for analysis at 6 and 24 hours. For D5D/D6D activity measurements after transient knockdown of FADS1 and FADS2 in HeLa cells, 0.5 μg of each pRS-shFADSl and pRS-shFADS2 vectors were transfected using Fugene HD for 24 hours. After transfection, cells were washed twice with 2 mL of pyruvate-free DMEM and media was changed to 2 mL of pyruvate-free DMEM containing 10 μM linoleic acid-¹³C₁₈, 1 mg/mL fatty acid-free bovine serum albumin (Sigma-Aldrich) and 2 mM L-glutamine. Media was collected for analysis at 24 hours.

Total protein measurement

For each experimental condition, a duplicate experiment was performed for total protein quantitation. For these 6-well plates, cells were washed with 2 mL of cold PBS once, then scraped with 0.3 mL RIPA buffer containing Halt protease inhibitor cocktail (all from Thermo Fisher Scientific), and protein levels were measured using the DC Protein Assay Kit (Bio-Rad). Results of cell culture analyses, specifically LC-MS based measurement of lipids, fatty acids, lactate, pyruvate and other polar metabolites; total triglycerides; and media glucose were all normalized to protein levels.

LC-MS Analyses

Sample preparation and extraction

As described above, cells were grown in 6 well plates for LC-MS analyses. After vacuum aspiration of media, cells were washed with cold PBS (without Mg²⁺ and Ca²⁺) and extracted with either 800 μL isopropanol cooled to 4°C (for lipid profiling) or 800 μL of 80% methanol cooled to −80°C (for polar metabolite profiling). Cell debris and extraction solution were transferred to a 1.5 mL centrifuge tube and extracted for lh (at 4°C for lipid profiling, on dry ice for polar metabolite profiling), vortexed, centrifuged (10,000 rcf, 4°C, 10 min), and then supernatants were transferred to a fresh 1.5 mL centrifuge tube and stored at −80°C until analysis. For polar metabolite profiling, pellets were further extracted using 100 μL of 80% methanol and secondary extracts were pooled with primary extracts. For fatty acid-¹³C₁₈ measurements of cultured media, 30 μL of media were extracted with 120 μL of 80% methanol cooled to 4°C; for pyruvate and lactate measurements, 30 μL of media were extracted with 120μL of 80% methanol containing 2 μg/mL l-lactate-d3. Extracted media were then vortexed, centrifuged (10,000 rcf, 4°C, 10 min), and supernatants were transferred to a fresh 1.5 mL centrifuge tube and stored at - 80°C until analysis. For mouse samples, snap frozen tissue samples were cut to ~60 mg and then homogenized in 4 volumes of ice-cold water using a Tissue Lyser II (Qiagen). For lipid profiling of mouse samples, 10 μL of tissue homogenate were extracted in 190μL of isopropanol containing 0.2 ng/μL l,2-dilauroyl-sn-glycero-3-phosphocholine, vortexed, and centrifuged (10,000 rcf, RT, 10 min), and supernatants were stored at −80°C until analysis. For lactate measurements in mouse samples, 30 μL of tissue homogenate were extracted in 170μL of 80% methanol containing 2 μg/mL l-lactate-d3, vortexed, and centrifuged (9,000 × g, 4°C, 10 min), and supernatants were stored at −80°C until analysis. For lipid profiling of human plasma samples, 10 μL of plasma were extracted in 190 μL of isopropanol containing 0.2 ng/μL 1,2- dilauroyl-sn-glycero-3-phosphocholine, vortexed, and centrifuged (10,000 rcf, RT, 10 min), and supernatants were stored at −80°C until analysis.

Lipid profiling

10μL of extracted cell lysate or 2 μL of extracted mouse tissue or human plasma were separated by reversed phase chromatography using an Acquity UPLC BEH C8 column (Waters); mobile phase A: 10 mM ammonium acetate in 95% water/5% methanol/0.1% formic acid; mobile phase B: 0.1% formic acid in methanol. The column was eluted isocratically with 20% mobile phase B for 1 minute followed by a linear gradient to 80% mobile phase B over 2 minutes, a linear gradient to 100% mobile phase B over 7 minutes, then 3 minutes at 100% mobile phase B. MS data were acquired with a Q Exactive Plus orbitrap mass spectrometer (Thermo Fisher Scientific) in the positive ion mode using electrospray ionization and full scan MS over m/z 200–1100 (Williams et al., 2014). Raw data were integrated and visually inspected using TraceFinder 3.3 software (Thermo Fisher Scientific). For each lipid analyte, the first number denotes the total number of carbons in the lipid acyl chain(s), and the second number (after the colon) denotes the total number of double bonds in the lipid acyl chain(s).

Fatty acid-¹³C₁₈ measurement

2μL of extracted cultured media were separated using reverse phase chromatography using a 150 × 2.1 mm Acquity UPLC BEH Cl8 column (Waters); mobile phase A: 0.01% formic acid in water; mobile phase B: 0.01% acetic acid in acetonitrile. The column was eluted isocratically with 20% mobile phase B for 3 minutes followed by a linear gradient to 100% mobile phase B over 12 minutes, then 3 minutes at 100% mobile phase B. MS data were acquired with a Q Exactive hybrid quadrupole orbitrap mass spectrometer (Thermo Fisher Scientific) in the negative ion mode using electrospray ionization and full scan MS over m/z 70–850. Raw data were integrated and visually inspected using TraceFinder 3.3 software (Thermo Fisher Scientific), with monitoring for the loss of hydrogen at the following masses: 297.2933 (linoleic acid-¹³C₁₈), 295.2777 (gamma-linolenic acid-¹³C₁₈), and 321.2933 (arachidonic acid-¹³C₁₈).

Polar metabolite profiling

Negatively charged polar analytes including alpha-hydroxybutyrate and citric acid cycle intermediates were measured in 10 μL of extracted cell lysates separated using a 150 × 2.0 mm Luna NH2 column (Phenomenex); mobile phase A: 20mM ammonium acetate, 20mM ammonium hydroxide in water; mobile phase B: l0mM ammonium hydroxide in 25% methanol/75% acetonitrile. The column was eluted isocratically with a linear gradient from 90% to 0% mobile phase B over 10 minutes, then 2 minutes at 0% mobile phase B. MS data were acquired with multiple reaction monitoring in the negative ion mode on a 5500 QTRAP MS (SCIEX). Raw data were integrated and visually inspected using MultiQuant software (version 2.1, SCIEX).

Lactate and pyruvate measurement

10 μL of extracted cultured media were separated using a 150 × 2.0 mm Luna NH2 column (Phenomenex); mobile phase A: 20mM ammonium acetate, 20mM ammonium hydroxide in water; mobile phase B: l0mM ammonium hydroxide in 10% water/22.5% methanol/67.5% acetonitrile. The column was eluted isocratically with a linear gradient from 100% to 0% mobile phase B over 10 minutes, then 2 minutes at 0% mobile phase B. MS data were acquired with multiple reaction monitoring in the negative ion mode on a TSQ Quantiva triple quadrupole MS (Thermo Fisher Scientific). Raw data were integrated and visually inspected using TraceFinder 2.1 software (Thermo Fisher Scientific). MRM transitions were as follows: pyruvate (87/43), lactate (89.1/43.1), lactate-d3 (92.1/45.1). Metabolite identifications were confirmed using synthetic mixtures of reference compounds.

Other Measurements

Total Triglyceride Measurement

In the comparison of 24h treatment of 200 nM rotenone versus 0.1% DMSO in IMCD3 cells, the same experiment as described for 6-well plates was repeated, except scaled for 1,000,000 cells at seeding in 10 cm culture dishes. Total triglycerides were measured using a TAG quantification kit (K622–100, Biovision). For mouse samples, 10 μL of tissue homogenate or plasma were extracted in 90 μL of 5.5% Nonidet P-40 solution (Thermo Fisher Scientific) and total TAG levels were measured as per the manufacturer’s instructions.

Measurement of Cytosolic NAD⁺/NADH Ratio Using SoNar

IMCD3 and HeLa-LbNOX cells were seeded at 600,000 cells in 10 cm culture dishes and incubated overnight. The following day, media was replaced with fresh antibiotic-free DMEM and cells were transfected with Fugene HD transfection reagent. More specifically, IMCD3 cells were incubated with 5 μg of pCDNA3.1-SoNar (provided by Y. Yang and J. Loscalzo) and 18 μL of Fugene HD transfection reagent for 24 hours, then transfected cells were resuspended with trypsin and replated at 20,000 cells per well on a 96-well clear bottom black microplate (Coming) in FluoroBrite DMEM (Life Technologies) containing 5% dialyzed FBS, 1 mM pyruvate and 2 mM L-glutamine. Twenty-four hours after transfected cells had been seeded on the 96-wells plate, cells were washed once with 100 μL of FluoroBrite DMEM and media was changed to 100 μL of FluoroBrite DMEM containing 2% dialyzed FBS and 2 mM L-glutamine. One hour later, the following compounds were added (final concentration): DMSO (0.1%), rotenone (1 nM to 200 nM), AKB (1 mM), rotenone (200 nM) + AKB (1 mM), pyruvate (1 mM), rotenone (200 nM) + pyruvate (1 mM), NHI-2 (30 μM), or metformin (2 mM). Dual-excitation ratios of the SoNar sensor were measured immediately afterwards and at the specified time points using an Envision multi-label reader (PerkinElmer) with 430 nm or 485 nm excitation and 535 nm emission for both excitation wavelengths. The procedure for LbNOX- expressing HeLa cells was the same, except that at the time of replating on the 96-well plate, media was also supplemented with or without 300 ng/mL doxycycline. For transient overexpression or knockdown of FADS1 and FADS2, 3 μg of each vector was co-transfected with 3 μg of pCDNA3.1-SoNar in HeLa cells, and subsequent steps were as described above. For SoNar measurements after transient overexpression of SLC16A11 variants in HEK293 cells, 3 μg of pLX304-SLC16A11^REF or pLX304-SLC16A11^T2D was co-transfected with 3 μg of pCDNA3.1-SoNar using Fugene HD as described above for 24 hours, then transfected cells were replated at 60,000 cells per well on a 96-well clear bottom black microplate in DMEM (D5921) containing 5% dialyzed FBS, 2 mM L-glutamine and MEM non-essential amino acids. Twenty- four hours after transfected cells had been seeded on the 96-wells plate, SoNar signals were measured.

Intracellular NAD⁺/NADH and NADP⁺/NADPH measurements

NAD (NAD⁺ and NADH) and NADP (NADP⁺ and NADPH) were measured using Biovision colorimetric kits (K337–100 and K347–100, respectively). To remove NADH or NADPH consuming enzymes, cell or mouse kidney tissue lysates were filtered through 10 kDa molecular weight cut off filters (Amicon Ultra, EMD Millipore) and the filtrate was subsequently used for further analysis.

Media Glucose Concentration

Media glucose levels were measured using a glucose colorimetric assay kit according to the manufacturer’s instructions (K606–100, Biovision).

Quantitative PCR

Total RNA was extracted using a RNeasy mini kit (Qiagen) and cDNA was synthesized using an AMV first-strand cDNA synthesis kit (Thermo Fisher Scientific). PCR reactions were performed using a CFX Connect Real-Time PCR machine (Bio-Rad) and SYBR green master mix (Thermo Fisher Scientific) according to the manufacturer’s instructions. The primer sequences are listed in Table S3.

Western Blots

A standard protocol was used for immunoblotting of cell lysates. In brief, samples were prepared using 4× Laemmli sample buffer (Bio-Rad), separated on a SDS polyacrylamide gel, transferred onto a 0.45 μm PVDF membrane (Thermo Fisher Scientific), and incubated with the primary and HRP-conjugated secondary antibodies. Immunoreactive proteins were visualized using a SuperSignal West Dura detection kit (Thermo Fisher Scientific). Following detection of D5D, FLAG-LbNOX, FLAG-D5D, FLAG-D6D, or V5-SLC16A11 the membrane was stripped using Restore^™ western blot stripping buffer (Thermo Fisher Scientific) for 15 min, then reprobed with a α-tubulin or β-actin antibody for the detection of loading control. We were unable to identify a specific antibody for endogenous D6D.

Microscopy

IMCD3 and HeLa cells were grown on Lab-TEK II chamber slides (Nunc). A standard immunocytochemical method was used for immunostaining of D5D or subcellular marker proteins. In brief, cells were fixed in pre-cooled (−20°C) 100% methanol for 5 min at 4°C, then blocked with 0.2% BSA and 10% normal goat serum (NGS) in PBST (PBS + 0.05% Tween 20) for 1 hour at room temperature. For endogenous D5D visualization, cells were incubated with Alexa Fluor-488-labeled anti-D5D and Alexa Fluor-568-labeled anti-reticulon 4 or anti-calnexin or anti-β-actin in 5% NGS in PBST overnight at 4°C. The smooth ER marker reticulon 4 (Shibata et al., 2010; Shibata et al., 2006) antibody was labeled with Alexa Fluor-568 using the APEX antibody labeling kit (Thermo Fisher Scientific). For transient FADS1 and FADS2 overexpression, cells were stained with anti-FLAG and anti-β-actin. After incubating the primary antibodies, cells were further incubated with Alexa-Fluor-568-conjugated goat anti-mouse IgG H&L (Thermo Fisher Scientific) and Alexa Fluor-488-conjugated goat anti-rabbit IgG in 5% NGS in PBST for 1 hour at room temperature. Stained cells were mounted with ProLong diamond antifade mountant with DAPI (Thermo Fisher Scientific) and examined using a laser-scanning confocal microscope (LSM 510, Zeiss) or conventional fluorescence microscope (Eclipse Ni, Nikon).

Cell Proliferation

IMCD3 and HeLa cells were quantitated using the CyQUANT cell proliferation assay kit (Molecular Probes). Fluorescence measurements were made with an Envision multi-label reader with excitation at 485 nm and emission detection at 535 nm.

Mouse Histology and Blood Chemistry

Hematoxylin & eosin (H&E) staining was used to assess renal tubular damage. Plasma creatinine, blood urea nitrogen (BUN), alanine aminotransferase (ALT), and aspartate aminotransferase (AST) were analyzed using a DRI-CHEM^® 7000 Analyzer (Heska).

Cellular ROS Production

The levels of cellular ROS were determined in IMCD3 cells using chloromethyl-2’,7’- dichlorofluorescein diacetate (CM-H₂DCFDA, Thermo Fisher Scientific), a cell permeable, non- fluorescent precursor of DCF. IMCD3 cells were incubated with 25 mM CM-H2DCFDA for 45 minutes. After loading the probe, cells were washed once with FluoroBrite DMEM and then the media was changed to FluoroBrite DMEM containing 2% dialyzed FBS and 2 mM L-glutamine premixed with 30 μM CP-24879, 200 μM tert-butyl hydroperoxide, or 0.1% DMSO. Fluorescence measurements were made with an Envision multi-label reader with excitation at 485 nm and emission detection at 535 nm.

QUANTIFICATION AND STATISTICAL ANALYSIS

Quantification

For comparative analysis of D5D expression in mouse kidney tissues, western blots images were quantified with ImageJ (NIH).

Statistical Analyses

Total triglycerides measured by colorimetric assay, as well as linoleic acid-¹³C₁₈, gamma-linolenic acid-¹³C₁₈, and arachidonic acid-¹³C₁₈, glucose, other free fatty acids, lactate, pyruvate, other polar metabolite levels, qPCR results, NAD⁺/NADH ratio, cell proliferation, and ROS production were analyzed using Student’s t-tests. For comparison of individual TAGs measured by LC-MS, a Student’s t-test was performed. When noted, Bonferroni-adjustment was made for the total number of TAGs assayed, P-value/50. Tissue lysate lipid content measured by LC-MS was compared using two-way ANOVA. Statistical analyses were performed using Graph Pad Prism 6 and statistical parameters can be found in the figure legends.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-D5D (FADS1) [EPR6898]	Abcam	Cat#ab126706; RRID: AB_11130088
Anti-D5D [EPR6898], Alexa Fluor 488 conjugate	Abcam	Cat#ab214251
Anti-D6D (FADS2), N-terminal	Abcam	Cat#ab170665
Anti-D6D	Abcam	Cat#ab72189; RRID: AB_2041175
Anti-Nogo A+B (reticulon 4)	Abcam	Cat#ab47085; RRID: AB_881718
Anti-Calnexin [AF18]	Thermo Fisher Scientific	Cat#MA3–027; RRID: AB_2069043
Anti-DYKDDDDK (FLAG)	Cell Signaling Technology	Cat#2368; RRID: AB_2217020
Anti-V5 [D3H8Q]	Cell Signaling Technology	Cat#13202; RRID: AB_2687461
Anti-β-actin [8H10D10]	Cell Signaling Technology	Cat#3700; RRID: AB_2242334
Anti-α-tubulin [DM1A]	Cell Signaling Technology	Cat#3873; RRID: AB_1904178
Goat anti-mouse IgG, Alexa Fluor 568 conjugate	Thermo Fisher Scientific	Cat#A11004; RRID: AB_141371
Goat anti-rabbit IgG, Alexa Fluor 488 conjugate	Thermo Fisher Scientific	Cat#A11034; RRID: AB_2576217
Bacterial and Virus Strains
Biological Samples
Human plasma samples	This paper	N/A
Chemicals, Peptides, and Recombinant Proteins
Rotenone	Sigma-Aldrich	Cat#R8875; CAS: 83-79-4
Oligomycin A	Sigma-Aldrich	Cat#75351; CAS: 579-13-5
Metformin	EMD Millipore	Cat#317240; CAS: 1115-70-4
SC-26196	Sigma-Aldrich	Cat#PZ0176; CAS: 218136-59-5
CP-24879	Sigma-Aldrich	Cat#C9115; CAS: 10141-51-2
NHI-2	Sigma-Aldrich	Cat#SML1463; CAS: 1269802-97-2
α-Ketobutyrate	Sigma-Aldrich	Cat#K0875; CAS: 2013-26-5
Sodium pyruvate	Thermo Fisher Scientific	Cat#11360070; CAS: 113-24-6
Doxycycline	Sigma-Aldrich	Cat#D3447; CAS: 10592-13-9
Arachidonic acid	EMD Millipore	Cat#181198; CAS: 506-32-1
Docosahexaenoic acid	Sigma-Aldrich	Cat#D2534; CAS: 6217-54-5
Eicosapentaenoic acid	EMD Millipore	Cat#324875; CAS: 10417-94-4
Fatty acid-free bovine serum albumin	Sigma-Aldrich	Cat#A8806; CAS: 9048-46-8
Sodium L-lactate-d3 solution	Sigma-Aldrich	Cat#616702; CAS: 867-56-1 (unlabeled)
Linoleic acid-13C18	Cambridge Isotope Laboratories	Cat#CLM-6855; CAS: 60–33-3 (unlabeled)
Prostaglandin E2-d4	Cayman Chemical	Cat#314010; CAS: 34210-10-1
1,2-dilauroyl-sn-glycero-3-phosphocholine	Avanti Polar Lipids	Cat#850335; CAS: 18194-25-7
Acetic acid, for LC/MS	Fisher Scientifics	Cat#A113-50; CAS: 64-19-7
Formic acid, for LC/MS	Honeywell Fluka	Cat#56302; CAS: 64-18-6
Ammonium acetate, for LC/MS	Sigma-Aldrich	Cat#73594; CAS: 631-61-8
Ammonium hydroxide solution, for HPLC	Honeywell Fluka	Cat#17387; CAS: 1336-21-6
Water, LC/MS grade	Fisher Scientifics	Cat# W6-4; CAS: 7732-18-5
Methanol, LC/MS grade	Fisher Scientifics	Cat#A456-4; CAS: 67-56-1
Isopropanol, LC/MS grade	Fisher Scientifics	Cat#A461-1; CAS: 67-63-0
Acetonitrile, LC/MS grade	Fisher Scientifics	Cat#A955-4; CAS: 75-05-8
tert-Butyl hydroperoxide	Sigma-Aldrich	Cat#458139; CAS: 75-91-2
CM-H2DCFDA	Thermo Fisher Scientific	Cat#C6827; CAS: 4091–99-0 (H2DCFDA)
Critical Commercial Assays
DC protein assay kit	Bio-Rad	Cat#5000116
Triglyceride quantification kit	BioVision	Cat#K622
NAD/NADH quantitation kit	BioVision	Cat#K337
NADP/NADPH quantitation kit	BioVision	Cat#K347
Glucose assay kit	BioVision	Cat#K606
RNesay mini kit	Qiagen	Cat#74104
AMV first-strand cDNA synthesis kit	Thermo Fisher Scientific	Cat#12328
SYBR green master mix	Thermo Fisher Scientific	Cat#A25741
SuperSignal West Dura detection kit	Thermo Fisher Scientific	Cat#34076
APEX Alexa Fluor 568 antibody labeling kit	Thermo Fisher Scientific	Cat#A10494
CyQUANT cell proliferation assay kit	Thermo Fisher Scientific	Cat#C7026
Deposited Data
Experimental Models: Cell Lines
mIMCD3	ATCC	CRL-2123
HEK293	ATCC	CRL-1573
Doxycycline-inducible HeLa-LbNOX cells	Titov et al., 2016	N/A
Experimental Models: Organisms/Strains
C57BL/6J mice	The Jackson Laboratory	JAX: 000664
Oligonucleotides
Primers for qPCR, see Table S3	This paper	N/A
Recombinant DNA
pCMV6-Entry-empty	OriGene Technologies	Cat#PS100001
pCMV6-Entry-human FADS1	OriGene Technologies	Cat#RC229252
pCMV6-Entry-human FADS2	OriGene Technologies	Cat#RC223780
pRS-scrambled non-effective shRNA control	OriGene Technologies	Cat#TR30012
pRS-human shFADS1	OriGene Technologies	Cat#TR313102; Tube ID: D
pRS-human shFADS2	OriGene Technologies	Cat#TR313101; Tube ID: B
pLX304-empty	Rusu et al., 2017	N/A
pLX304-SLC16A11REF-V5	Rusu et al., 2017	N/A
pLX304-SLC16A11T2D-V5	Rusu et al., 2017	N/A
Software and Algorithms
TraceFinder 3.3 (or 3.2)	Thermo Fisher Scientific	https://www.thermofisher.com/order/catalog/product/OPTON-30493
MultiQuant 2.1	SCIEX	https://sciex.com/products/software/multiquant-software
Prism 6	GraphPad Software	https://www.graphpad.com/scientific-software/prism
ImageJ	NIH	https://imagej.nih.gov/ij
Other
Dialyzed FBS	Thermo Fisher Scientific	Cat#26400044
Nonidet P-40 solution	Thermo Fisher Scientific	Cat#85124
FuGENE HD transfection reagent	Promega	Cat#E2311
MEM non-essential amino acids solution	Thermo Fisher Scientific	Cat#11140050
Prolong diamond antifade mountant with DAPI	Thermo Fisher Scientific	Cat#P36966
Amicon Ultra 10 kDa cut off filters	EMD Millipore	Cat#UFC501096
100 × 2.1 mm Acquity UPLC BEH C8 column	Waters	Cat#186002878
150 × 2.1 mm Acquity UPLC BEH C18 column	Waters	Cat#186002353
150 × 2.0 mm Luna NH2 column	Phenomenex	Cat#00F-4378-B0

Highlights.

• Blocking ETC or lactate production reduces cytosolic NAD⁺/NADH and increases HUFAs

• HUFA synthesis by D5D and D6D is a mechanism for glycolytic NAD⁺ recycling

• D5D and D6D mediated NAD⁺ regeneration can be acutely adaptive in vivo

• SLC16A11 and FADS1–3 may influence metabolic risk via impact on cytosolic NAD⁺/NADH