Nudt19 is a renal CoA diphosphohydrolase with biochemical and regulatory properties that are distinct from the hepatic Nudt7 isoform

CoA is the major acyl carrier in mammals and a key cofactor in energy metabolism. Dynamic regulation of CoA in different tissues and organs supports metabolic flexibility. Two mammalian Nudix hydrolases, Nudt19 and Nudt7, degrade CoA in vitro. Nudt19 and Nudt7 possess conserved Nudix and CoA signature sequences and specifically hydrolyze the diphosphate bond of free CoA and acyl-CoAs to form 3′,5′-ADP and 4′-(acyl)phosphopantetheine. Limited information is available on these enzymes, but the relatively high abundance of Nudt19 and Nudt7 mRNA in the kidney and liver, respectively, suggests that they play specific roles in the regulation of CoA levels in these organs. Here, we analyzed Nudt19−/− mice and found that deletion of Nudt19 elevates kidney CoA levels in mice fed ad libitum, indicating that Nudt19 contributes to the regulation of CoA in vivo. Unlike what was observed for the regulation of Nudt7 in the liver, Nudt19 transcript and protein levels in the kidney did not differ between fed and fasted states. Instead, we identified chenodeoxycholic acid as a specific Nudt19 inhibitor that competed with CoA for Nudt19 binding but did not bind to Nudt7. Exchange of the Nudix and CoA signature motifs between the two isoforms dramatically decreased their kcat. Furthermore, substitutions of conserved residues within these motifs identified amino acids playing different roles in CoA binding and hydrolysis in Nudt19 and Nudt7. Our results reveal that the kidney and liver each possesses a distinct peroxisomal CoA diphosphohydrolase.

CoA is an obligate cofactor utilized as an acyl carrier in hundreds of metabolic reactions. Numerous CoA thioesters are also involved in the post-translational modification of histones and thousands of other proteins (1)(2)(3)(4)(5). Three major subcellular CoA pools are found in the cytosol, mitochondria, and peroxisomes to support specific metabolic pathways (6,7). These include, among others, fatty acid synthesis in the cytosol, the TCA cycle and oxidation of long-chain and medium-chain fatty acids in the mitochondria, and bile acid conjugation and oxidation of very long-chain and branched-chain fatty acids in the peroxisomes. A dedicated pool of acetyl-CoA is also found in the endoplasmic reticulum, where it is involved in protein quality control and autophagy (8), whereas nuclear acyl-CoAs, which freely equilibrate with the cytosolic pool across the nuclear pores, contribute to the regulation of gene expression (5).
Tight control over the concentration of CoA in different organs is essential to maintain normal metabolism and organ function. Indeed, elevated CoA levels in the liver of diabetic mice promote excessive gluconeogenesis and hyperglycemia (9), whereas a supraphysiological concentration of CoA in skeletal muscle leads to decreased ATP levels and exercise performance (10). Adverse consequences are also associated with lower-than-normal CoA levels. Decreased synthesis of this cofactor in the nervous system is associated with a rare neurological disorder in humans (11,12), whereas decreased synthesis of CoA in the liver leads to fasting hypoglycemia and hepatic triglyceride accumulation (13). Dynamic regulation of CoA within the normal physiological range is important to support the metabolic reprogramming that underlies the capacity to respond to changes in the metabolic state. For example, the increase in CoA that characterizes the fed-to-fasted transition in the liver is driven by the activation of the biosynthetic pathway and is required to sustain the high rates of hepatic fatty acid oxidation and gluconeogenesis under fasting conditions (13,14). On the other hand, the net decrease in the concentration of CoA observed upon refeeding after a fast requires both inhibition of de novo CoA synthesis and active degradation of the cofactor accumulated in the fasted state.
The existence of a CoA-degradation pathway in the liver and kidneys is supported by the fact that mice treated with an inhibitor of the CoA biosynthetic pathway exhibit a dramatic reduc- cro ARTICLE tion in hepatic and renal CoA, indicating rapid CoA turnover in these tissues (15). Liver and kidneys contain high transcript levels of two Nudix hydrolases, Nudt7 and Nudt19, respectively, which exhibit CoA diphosphohydrolase activity in vitro (16,17). Nudix hydrolases hydrolyze the diphosphate bond of a variety of substrates consisting of a nucleoside diphosphate linked to another moiety, x (18,19). All 22 N U (where U is a hydrophobic residue, X is any residue, and superscript N denotes the position of a residue of the signature motif within a specific protein sequence) (20,21). Additionally, Nudix hydrolases that specifically hydrolyze CoA also contain a putative FPGG (PROSITE accession number PS01293, formerly UPF0035), upstream of the Nudix box motif (22). The crystal structure of human NUDT7 has recently been released (PDB 3 code 5T3P). Similar to the structure of the bacterial homolog from Deinococcus radiodurans (24), the crystal structure of the human enzyme does not contain CoA bound. Thus, the contribution of the CoA motif to CoA binding and the organization of the CoA-binding site in these enzymes are currently unknown.
Independent studies and approaches have demonstrated the localization of Nudt7 in the peroxisomes (17,25). Nudt7 mRNA and protein levels in the liver are regulated by the nutritional state and are inversely correlated with the tissue concentration of CoA (9,26). Additionally, overexpression of a catalytically active, cytosolic form of Nudt7 has also recently been shown to cause a decrease in CoA levels in vivo (25).
Significantly less is known about Nudt19. Nudt19 exhibits modest sequence similarity to Nudt7 (ϳ30%), is larger than a typical Nudix hydrolase, including Nudt7, and contains a unique 45-49-amino acid insertion of unknown function within its Nudix box (16). This enzyme, also known as RP2p, was originally reported to be a kidney protein whose transcript levels were robustly up-regulated by androgens (27). Later, recombinant Nudt19 was shown to specifically hydrolyze CoA species (16), although modest activity against capped RNA has also been detected (28). Based on proteomics studies, Nudt19 localization has been reported in both peroxisomes and mitochondria (16,29); however, it is currently unclear whether the enzyme has dual localization or exclusively resides in one of the two subcellular compartments, and which one. Given the presence of dedicated CoA pools and complementary metabolic processes in peroxisomes and mitochondria, this distinction is important to gain insight into the physiological role of Nudt19, even in relation to the other mammalian isoform, Nudt7. Indeed, it is presently unclear whether Nudt7 and Nudt19, which seem to be differentially expressed in liver and kidneys, differ in key aspects of their function.
We have investigated the effect of Nudt19 deletion on CoA levels, the localization of this enzyme in intact cells, and its regulation by selected metabolites. Additionally, we used a combination of computational modeling, substrate docking, and site-directed mutagenesis to identify residues involved in CoA binding and hydrolysis in Nudt19 and analyzed the role of equivalent residues in Nudt7. We show that Nudt19 is a peroxisomal enzyme that contributes to the regulation of kidney CoA levels in vivo and provide evidence supporting the conclusion that the CoA-binding sites of Nudt19 and Nudt7 are different.

Nudt19 regulates kidney CoA levels in vivo
Previous work has shown that Nudt19 hydrolyzes free CoA and acyl-CoAs to 3Ј,5Ј-ADP and 4Ј-(acyl)phosphopantetheine in vitro (16). We confirmed these findings by measuring the activity of recombinant Nudt19 against several acyl-CoAs and comparing it to the better characterized Nudt7 (17,25) (Fig. 1,  A and B). We determined that the optimal concentration of MgCl 2 for Nudt7 and Nudt19 was 4 and 10 mM, respectively, when free CoA was used as the substrate (data not shown). However, in experiments where we tested multiple acyl-CoAs, it was necessary to keep the concentration of MgCl 2 at 4 mM for both Nudt7 and Nudt19 to prevent the precipitation of lauroyland stearoyl-CoA. Under these conditions, the specific activity per pmol of Nudt19 was lower than that of Nudt7, but the enzyme readily hydrolyzed short and medium chain acyl-CoAs, malonyl-and succinyl-CoA, in addition to free CoA (Fig. 1A). Exceptions were acetyl-CoA and the synthetic monobromobimane (mBB)-CoA, which were excellent substrates for Nudt7, but not for Nudt19. Nudt19 also had no detectable hydrolytic activity against ATP, ADP, NAD ϩ NADH, NADP ϩ , and NADPH (data not shown), confirming the specificity of this enzyme for CoA (16).
The mRNA levels of Nudt19 are highly abundant in mouse kidneys (16,30). To examine the distribution of this protein in mouse tissues, we generated a polyclonal antibody against the full-length mouse Nudt19. Consistent with the mRNA expression pattern, Nudt19 protein showed the highest abundance in the kidneys, with lower but still detectable levels in skeletal muscle and brain (Fig. 1C). This expression pattern was strikingly different from Nudt7, which was expressed at the highest levels in the liver and, to a significantly lower extent, in white and brown adipose tissue (Fig. 1C). The high expression level of Nudt19, combined with the minimally detectable Nudt7 protein in the kidneys (Fig. 1C), suggested that Nudt19 was the major of the two isoforms in these organs.
To determine whether Nudt19 regulated CoA levels in vivo, we obtained Nudt19 ϩ/Ϫ mice from the KOMP repository and bred them to generate Nudt19 Ϫ/Ϫ mice and wildtype littermate controls. The Nudt19 Ϫ/Ϫ mice were viable, fertile, and outwardly normal. Successful deletion of the gene was verified by PCR analysis of tail biopsies ( Fig. 2A), whereas Western blotting analysis confirmed the lack of Nudt19 protein expression (Fig.  2B). Additionally, enzymatic assays confirmed the lack of any residual formation of 3Ј,5Ј-ADP in kidney homogenates obtained from Nudt19 Ϫ/Ϫ mice (Fig. 2C). The concentration of CoA was measured in the kidneys of Nudt19 Ϫ/Ϫ and control mice in the fed state and following an overnight fast. In the fed state, the kidneys of the Nudt19 Ϫ/Ϫ mice exhibited a significant 20% increase in the concentration of CoA compared with the wildtype mice (Fig. 2D). Fasting induced an increase in kidney CoA levels in both wildtype and Nudt19 Ϫ/Ϫ mice. Under these conditions, the concentration of CoA in the Nudt19 Ϫ/Ϫ kidneys tended to be higher than control mice, but the difference did not reach statistical significance. No compensatory increase in Nudt7 transcript levels was detected in the kidneys of the Nudt19 Ϫ/Ϫ mice in either the fed or fasted states (Fig. 2E). Combined, these data showed that Nudt19 was the major CoA diphosphohydrolase in the kidneys and that this enzyme contributed to the regulation of kidney CoA levels in vivo.
Mouse Nudt7 and other CoA diphosphohydrolases from Saccharomyces cerevisiae and Caenorhabditis elegans localize to the peroxisomes (17,22,31), which suggests that these enzymes might specifically regulate the peroxisomal pool of CoA (32, 33). Nudt19 contains a C-terminal peroxisome targeting signal type 1 (PTS1), Ala-His-Leu, and was originally identified as a peroxisomal enzyme by the proteomic analysis of kidney peroxisomes (16). Recently, proteomic studies conducted on isolated mitochondria have also led to the annotation of Nudt19 as a mitochondrial enzyme in the Mitocarta2.0 inventory (29). Given the common contamination of isolated peroxisomes and mitochondria with variable amounts of other organelles, we used immunofluorescence and confocal microscopy to determine the intracellular localization of Nudt19 in whole HEK 293 cells (Fig. 3). HEK 293 cells were transiently transfected with constructs expressing Nudt19 with N-or C-terminal GFP tags. Nuclei were stained with DAPI, mitochondria were stained with MitoTracker Orange CMTMRos, and peroxisomes were visualized with an antibody against the endogenous peroxisomal protein PMP70. When expressed with an N-terminal GFP tag that left the PTS1 exposed, Nudt19 colocalized with PMP70, as shown by the yellow pixels resulting from the overlap of the green and red contributions (Fig. 3, A-C). Masking the PTS1 of Nudt19 with a C-terminal GFP tag prevented the peroxisomal localization (Fig. 3, G-I) but did not direct the protein to the mitochondria (Fig. 3, J-L), leading instead to a diffuse cytoplasmic localization. Overall, these results supported the conclusion that Nudt19 was a peroxisomal enzyme. As such, the 20% increase measured in the whole kidney homogenates from fed Nudt19 Ϫ/Ϫ mice (Fig. 2D) could be an underestimation of the local accumulation of CoA in the kidney peroxisomes of these animals.

Nudt19, but not Nudt7, is competitively inhibited by specific bile acids
Deletion of Nudt19 led to a significant increase in kidney CoA levels in fed but not fasted mice (Fig. 2D). This suggested that, in a wildtype mouse, the activity of Nudt19 might be higher in the fed state compared with the fasted state. We analyzed Nudt19 mRNA and protein expression levels in the kidneys of mice fed ad libitum and mice fasted for up to 48 h, but we did not detect any difference (Fig. 4, A-C). We then focused our attention on metabolite regulation.
We screened a panel of 30 compounds, including a large number of substrates and products of peroxisomal metabolism, for their effect on the activity of recombinant Nudt19 ( Table 1). None of the compounds tested significantly stimulated Nudt19 activity. Instead, we found that a select group of bile acids inhibited the enzyme by Ͼ50%. More specifically, chenodeoxycholic acid (CDCA) and its conjugated derivatives, taurochenodeoxycholic acid and glycochenodeoxycholic acid, decreased Nudt19 activity by 60 -75% with respect to vehicle control (Table 1). More hydrophobic bile acids, such as lithocholic acid and ursocholanic acid, were also potent inhibitors of Nudt19, and ␣-muricholic acid, which derives from CDCA by the addition of M of each CoA substrate, and the 3Ј,5Ј-ADP produced was quantified as described under "Experimental procedures." The data are plotted as the mean (bars) of duplicate measurements from two or three independent experiments (circles) Ϯ S.D. C, immunoblot analysis to detect Nudt19 and Nudt7 proteins in mouse tissues. Tissue lysate proteins (150 g) were fractionated by SDS-PAGE, transferred to a PVDF membrane, and blotted with the Nudt19 antibody. The same membrane was then stripped and blotted with antibodies against GAPDH (loading control) and Nudt7. Note that although the same amount of protein was loaded, different tissues contain different amounts of GAPDH. Panc., pancreas; WAT, white adipose tissue; BAT, brown adipose tissue; NS, nonspecific band.

Differential regulation of Nudt19 and Nudt7
a 6␤-hydroxyl group, still retained significant potency against the enzyme. In contrast, cholic acid and its glycine and taurine conjugates exhibited decreased potency against Nudt19 compared with CDCA. Furthermore, other steroid compounds, such as progesterone, pregnenolone, and corticosterone, had no significant effect on the activity of the enzyme, confirming the specificity of the inhibitory effect of CDCA and derived bile acids on Nudt19 (Table 1).
Interestingly, none of the compounds tested, including CDCA, significantly affected the activity of Nudt7 (Table 1). In addition, isothermal titration calorimetry (ITC) studies showed no detectable binding of CDCA to the liver enzyme, whereas the K d of the Nudt19-CDCA complex was determined to be 8.3 M (⌬G ϭ Ϫ6.9 kcal/mol) (Fig. 5, A and B). Further kinetic analysis of the mechanism of Nudt19 inhibition revealed that CDCA was a competitive inhibitor with respect to free CoA (Fig. 5C). Combined, these results identified CDCA as a specific competitive inhibitor of Nudt19 in vitro.

The CoA boxes and the Nudix signature sequences of Nudt19 and Nudt7 are not interchangeable
Bile acids are primarily synthesized and conjugated in the liver. Although the potential relevance of Nudt19 inhibition by a select group of these compounds in vivo is currently not known, the lack of CDCA binding to Nudt7 and the competitive nature of the CDCA inhibition of Nudt19 strongly suggested that the CoA-binding sites of the two enzymes might be different. Nudix hydrolases that degrade CoA contain a conserved motif, FPGG, here referred to as the CoA box, which is proposed to play a role in CoA binding (24) (Fig. 6A) (21,34,35). In several validated CoA diphosphohydrolases, the Nudix box exhibits a variable number of X residues between E 7 N and R 15 N . The most extreme case is represented by mouse and human Nudt19, which contain a 45-49-amino acid insertion (Fig. 6A). Given its position within the Nudix box, this insertion would occur between strand ␤4 and helix ␣2, i.e. in the first loop of the Nudix ␤-strand-loop/helix/loop structural motif (Fig. 6A).
We swapped the CoA box in Nudt7 and Nudt19, generating a chimeric Nudt7 with the Nudt19 CoA box, Nudt7-19-CB, and

Differential regulation of Nudt19 and Nudt7
a chimeric Nudt19 with the Nudt7 CoA box, Nudt19-7-CB. We also substituted the Nudix box of Nudt7 with that of Nudt19 producing a Nudt7-19-NB chimera. All three chimeras were expressed and purified in amounts similar or higher compared with the wildtype Nudt7 and Nudt19. Conversely, exchanging the Nudix box in Nudt19 with the shorter Nudt7 Nudix box produced a protein that could be expressed but not purified as a soluble enzyme. Interestingly, a similar result was obtained when we expressed a Nudt19(⌬58 -103) mutant lacking the unique insertion sequence in the Nudix box (data not shown), suggesting that this amino acid stretch could be important for the folding and/or stability of Nudt19. The three soluble chimeras were analyzed by analytical gel filtration to assess their folding. Based on their calculated molecular weight, ϳ70% of Nudt19-7-CB, Nudt7-19-CB, and Nudt7-19-NB eluted as glob-ular monomers, similar to wildtype Nudt7 and Nudt19 (Fig.  6B). Smaller peaks, indicating the formation of multimers, were also detected in both the chimeras and the wildtype enzymes (Fig. 6B). Despite evidence of protein folding, the specific activities of Nudt19-7-CB, Nudt7-19-CB, and Nudt7-19-NB were decreased 2-3 orders of magnitude compared with the respective wildtype enzymes (Fig. 6C). In particular, the substitution of the CoA box of Nudt19 with that of Nudt7 caused Ͼ5,000fold decrease in the k cat and a 3-fold increase in the K m for free CoA compared with Nudt19 (Table 2). Similarly, swapping the Nudt7 CoA box with the CoA box of Nudt19 in Nudt7-19-CB led to a 500-fold decrease in k cat and a 5-fold increase in K m . Finally, exchanging the Nudix box in Nudt7-19-NB caused a 200-fold decrease in k cat and a 3-fold increase in K m compared with Nudt7 (Table 2).

Differential regulation of Nudt19 and Nudt7
Wildtype Nudt19 and Nudt7 readily degrade free CoA and lauroyl-CoA but exhibit different activities against acetyl-CoA and mBB-CoA, which are good substrates for Nudt7 but not Nudt19 (Fig. 1, A and B). We next determined the effect of swapping the Nudix and CoA boxes on the substrate specificity of the three soluble chimeras Nudt19-7-CB, Nudt7-19-CB, and Nudt7-19-NB (Fig. 6D). For each enzyme, the activity data with the different substrates were plotted relative to free CoA to allow for comparison across enzymes with dramatically different specific activities. All three proteins, including the Nudt19 chimera Nudt19-7-CB, degraded acetyl-CoA and mBB-CoA, with a significantly higher relative preference for the latter substrate compared with wildtype Nudt7 and Nudt19. Combined with the dramatic reduction in k cat and robust decrease in K m (Table 2), the altered substrate specificity of the chimeras suggested that swapping the Nudix and CoA boxes between Nudt19 and Nudt7 might have caused a substantial deformation of the respective active sites.

Mutagenesis-guided modeling of Nudt19
The recently released crystal structure of human NUDT7 is the only currently available structure of a mammalian CoA diphosphohydrolase. This structure does not contain any divalent cation, essential to correctly orient the diphosphate bond of the substrate for hydrolysis, or CoA bound (PDB code 5T3P). To identify Nudt19 residues involved in CoA and possibly CDCA binding, we decided to build a model of the putative structure of Nudt19 in complex with Mg 2ϩ and free CoA using a computational approach guided by mutational analysis.
Within the Nudix box, the carbonyl group of G 1 N and two of the three glutamates in the E 16 N , E 19 N , and E 20 N series, participate in the coordination of the essential divalent cation (24, 34 -39). E 16 N , E 19 N , and E 20 N correspond to Glu-112, Glu-115, and Glu-116 in Nudt19, respectively (Figs. 6A and 7C). Mutation of these residues to alanine abolished the enzymatic activity of Nudt19 E112A and Nudt19 E116A ( Table 2). The activity of Nudt19 E115A was detectable; however, the enzyme exhibited a ϳ2000-fold decrease in k cat and a 3-fold increase in K m ( Table 2). Interestingly, the effect of this mutation on the K m suggested that this residue of the Nudix box contributed to CoA

Table 1 Effect of selected metabolites on the activity of Nudt19 and Nudt7
Nudt19 and Nudt7 were assayed in the presence of 250 M of each compound, unless otherwise indicated, and free CoA as substrate. A minus sign indicates a decrease in activity with respect to the vehicle control. A plus sign indicates an increase in activity with respect to the vehicle control. NMN, nicotinamide mononucleotide.

Nudt19
Nudt7 Ϫ15 ϩ14 a Metabolites that decreased or increased the activity Ն50%.

Differential regulation of Nudt19 and Nudt7
binding. We also mutated the equivalent Nudt7 residues E 16 92 , E 19 95 , and E 20

96
. Similar to the Nudt19 mutants, Nudt7 E92A and Nudt7 E96A did not show any detectable activity. Mutation of Nudt7 Glu-95 to alanine decreased the k cat Ͼ10-fold, but unlike the equivalent mutant in Nudt19, it had no significant effect on the K m for CoA. Overall, these results were consistent with mutagenesis studies conducted on other Nudix hydrolases (36,40,41) and confirmed the importance of E 16 N , E 19 N , and E 20 N for the activity of both Nudt19 and Nudt7. The crystal structure of the CoA diphosphohydrolase from D. radiodurans, CoAse, contains Mg 2ϩ bound (24). Although unable to fit completely, docking of free CoA onto the "closed" structure of the bacterial CoAse suggested the existence of a potential contact between the side chain of Arg-55, the first arginine in the CoA box (Fig. 6A), and the diphosphate bond of CoA (24). Additionally, mutation of the equivalent residue to a glutamine in the Arabidopsis thaliana homolog AtCoAse led to a dramatic decrease in activity (42). This arginine residue is highly conserved across CoA diphosphohydrolases, and corresponds to Arg-34 in Nudt19 and Arg-60 in Nudt7 (Fig. 6A). Mutation of these residues to a methionine, an amino acid with a comparable side chain but no charge, dramatically decreased the k cat of the Nudt19 R34M and Nudt7 R60M mutants by 2000-and 200fold, respectively, and caused a 2-3-fold increase in the K m . Combined, these data supported the importance of this residue for catalysis.
We used this information as a guide to evaluate computational models of Nudt19 generated by using the YASARA modeling software suite (43). Specifically, the final model that we selected exhibited 1) the correct positioning of E 16 112 , E 19 115 , and E 20 116 with respect to Mg 2ϩ , 2) the appropriate orientation of the diphosphate moiety of free CoA with respect to Mg 2ϩ to allow for catalysis, and 3) interaction of Arg-34 with the diphosphate group of the substrate CoA (Fig. 7, B and C). In this model, the CoA and Nudix boxes of Nudt19 folded into the expected structural motifs with very good superposition with the CoA and Nudix boxes of the D. radiodurans CoAse and human NUDT7 (Fig. 7A). Within the ␤-strand-loop/ helix/loop structural motif of the Nudix box, the 45-amino acid insertion in Nudt19 was modeled as an expansion in the loop connecting the strand ␤4 to the helix ␣2 and was mostly unstructured.

Identification of additional residues involved in CoA binding and hydrolysis
Based on the model, we selected and mutated a set of Nudt19 residues predicted to participate in the binding of CoA (Table  2). In particular, we mutated Arg-39, Phe-40, Leu-41, which are part of the CoA box, and Arg-189 to alanine (Figs. 6A and 7C). In the Nudt19 model, the side chains of Phe-40 and Leu-41 extended toward the free thiol end of the CoA molecule, whereas the side chain of Arg-39 extended toward the 3Ј-phosphate of the CoA adenine ring (Fig. 7C). Mutation of Phe-40 and Leu-41 to alanine caused a robust 3-4-fold increase in the K m , confirming that these residues played a role in CoA binding. In addition, the Nudt19 L41A mutant also showed a ϳ3-fold decrease in k cat . Mutation of Arg-39 to an alanine did not have a significant effect (Ͻ2-fold) on the K m ; however, removal of the positive charge caused a 3-fold decrease in k cat , indicating that this residue was nevertheless important for CoA hydrolysis. Arg-189 was located in a loop region outside the CoA box and stretched toward the adenosine moiety of CoA (Fig. 7C). The Nudt19 R189A mutant exhibited a 4-fold increase in K m and a 2-fold decrease in k cat , confirming that Arg-189 was important for both CoA binding and hydrolysis. We further investigated the role of this residue by mutating Arg-189 to a methionine. Unlike alanine, the side chain of methionine was able to functionally replace Arg-189, thus suggesting the existence of potential non-polar side chain interaction(s) between Arg-189 and CoA (Table 2). Because Nudt19

Differential regulation of Nudt19 and Nudt7
R34M, F40A, L41A, and R189A mutants exhibited a reduced affinity for CoA (Table 2), we tested the sensitivity of these mutants to inhibition by the competitive inhibitor CDCA (Fig.   5C). At a concentration of 250 M of free CoA, the IC 50 of CDCA was determined to be 125 M against wildtype Nudt19. Among the above mutants, Nudt19 R34M retained ϳ80% of C, specific activity of the recombinant Nudt19, Nudt7, and chimeric proteins. Specific activities Ϯ standard error were calculated from the linear regression of five-point protein curves obtained for each enzyme from at least two independent experiments in duplicate. D, substrate specificity of Nudt19, Nudt7, and chimeric proteins. Activity with each CoA species is expressed as percentage relative to the activity with free CoA. The data are reported as the mean (bars) of duplicate measurements from two or three independent experiments (circles) Ϯ S.D.

Differential regulation of Nudt19 and Nudt7
activity in the presence of CDCA (Fig. 7D), suggesting that Arg-34 might be involved in both CoA and CDCA binding.
Although one of the variable X residues in the Nudix box, Nudt19 Asp-53 is highly conserved among the CoA diphosphohydrolases (Fig. 6A). In the Nudt19 model, this residue showed no interaction with CoA, but we nevertheless mutated it to alanine to determine the effect of the mutation on K m and k cat . Consistent with the model, the Nudt19 D53A mutant exhibited kinetic parameters similar to the wildtype enzyme, and we obtained similar results with the equivalent Nudt7 D79A mutant ( Table 2). The sequence alignment of the CoA diphosphohydrolases also showed that it was common to find a positively charged residue in the same position as Nudt19 Arg-39, a residue involved in catalysis but not CoA binding (Table 2). We mutated the equivalent residue in Nudt7, Lys-65, to alanine. The Nudt7 K65A mutant exhibited a ϳ3-fold increase in both K m and k cat , showing a differential effect on both kinetic parameters with respect to Nudt19. Finally, because Phe-40 in Nudt19 was involved in substrate binding, we mutated the equivalent Nudt7 Arg-66 to both alanine and methionine. Nudt7 R66A showed a Ͼ10-fold increase in K m , whereas the Nudt7 R66M mutant exhibited kinetic parameters similar to the wildtype enzyme (Table 2). These results supported the conclusion that the side chain of Arg-66 participated in CoA binding through hydrophobic interactions and suggested that Phe-40 in Nudt19 and Arg-66 in Nudt7 might be functionally similar.

Discussion
To gain insight into the physiological role of Nudt19, a Nudix enzyme with CoA diphosphohydrolase activity, one the major goals of this study was to determine the ability of the enzyme to regulate CoA levels in vivo. The present work demonstrates that Nudt19 degrades CoA and contributes to the regulation of this cofactor in the kidneys. Indeed, Nudt19 Ϫ/Ϫ mice exhibited a significant 20% increase in total CoA levels in whole kidney homogenates obtained from fed mice (Fig. 2D). This result is consistent with the very narrow tissue distribution of the protein, which showed the highest expression levels in the kidneys, with brain and skeletal muscle containing barely detectable levels of Nudt19 (Fig. 1C).
The second goal of this study was to address conflicting reports on the peroxisomal and mitochondrial localization of Table 2 Kinetic

parameters of Nudt19 and Nudt7 mutants and chimeras
The experiments were conducted in duplicate as described under "Experimental procedures." Kinetic constants were determined by nonlinear regression analysis of the data and expressed Ϯ S.E.

Differential regulation of Nudt19 and Nudt7
Nudt19, because peroxisomes and mitochondria contain separate pools of CoA and are the sites of distinct CoA-dependent metabolic processes (7,44). Using immunofluorescence and constructs expressing Nudt19 with either a C-terminal or an N-terminal GFP tag, we showed that the enzyme resides in the peroxisomes and not in the mitochondria (Fig. 3), thus suggesting a specific role for Nudt19 in peroxisomal metabolism. In the peroxisomes, free CoA and acyl-CoAs are imported through dedicated transporters (45,46) and participate in several processes including the ␣and ␤-oxidation of fatty acids that are not substrates for the mitochondrial oxidative machinery, the synthesis of ether phospholipids and, in the liver, the final steps in bile acid synthesis and conjugation. Additionally, because imported CoA species and other molecules Ͼ400 Da cannot freely exit the peroxisomes, the Nudix-mediated hydrolysis of CoA species to 3Ј,5Ј-ADP and 4Ј-(acyl)phosphopantetheine likely prevents CoA buildup in this organelle (32, 33). Indeed, it has been proposed that the function of peroxisomal CoA-degrading enzymes, including Nudt7 and now Nudt19, is to regulate the peroxisomal CoA pool and, in turn, peroxisomal lipid metabolism (32). Consistent with this proposed role, Nudt19 activity was highest with free CoA, which is required in multiple peroxisomal pathways, and with succinyl-CoA and C8-C12 acyl-CoAs, which are the oxidation products of dicarboxylic and very long-chain fatty acids, respectively (Fig. 1A). Nudt19 was previously reported to possess mRNA decapping activity in vitro (28). Although detectable, the hydrolytic activity of Nudt19 against capped RNA was minimal when compared with that of established mRNA decapping enzymes such as Dcp2 and Nudt16 (28). Additionally, mRNA decapping occurs in cytosolic foci (47). Thus, the peroxisomal localization of Nudt19, combined with its effect on CoA levels in vivo, argues against a role for this enzyme in mRNA decapping. The peroxisomal localization of Nudt19 and fact that the enzyme could affect just one of the subcellular CoA pools could also explain the relatively small increase in total CoA levels observed in whole kidney homogenates obtained from Nudt19 Ϫ/Ϫ mice in the fed state. In the fasted state, the concentration of kidney CoA was higher than in the fed state and similar between wildtype control and Nudt19 Ϫ/Ϫ mice. These results suggest that the increase in kidney CoA levels that occurs upon fasting may be promoted more by activation of the CoA biosynthetic pathway than by a decrease in CoA degradation. The flux through the CoA biosynthetic pathway is controlled by the pantothenate kinase (PanK) isoforms, and PanK1, which is the isoform that drives the increase in CoA levels observed in the liver of fasted mice, is also the major PanK isoform in the kidneys (13,48). Upon refeeding, a net decrease in CoA levels requires not only inhibition of CoA synthesis, but also active degradation of the cofactor accumulated in the fasted state. This might explain why the effect of deleting Nudt19 on kidney CoA levels was mostly detected in fed mice. Interestingly, although the Nudt19 Ϫ/Ϫ mice exhibited higher kidney CoA levels in the fed state compared with control mice, this concentration of the cofactor was still significantly lower than in the fasted state, suggesting the existence of residual CoA-degrading activity in the absence of Nudt19. Nudt7 and Nudt19 are the only currently confirmed CoA-degrading enzymes, and in the kidneys, Nudt7 is a minor isoform (Fig. 1C, and compare Fig. 2E to Fig. 4A). Under our experimental conditions, the production of 3Ј,5Ј-ADP in kidney extracts from Nudt19 Ϫ/Ϫ mice was undetectable (Fig. 2C), and we did not observe any compensatory up-regulation of Nudt7 (Fig. 2E). Despite this, an accurate estimate of the contribution of Nudt7, if any, to the regulation of kidney CoA levels, will require the generation of mice with a deletion in both Nudt7 and Nudt19. Alternatively, CoA might be hydrolyzed to products other than 3Ј,5Ј-ADP and 4Ј-(acyl)phosphopantetheine by a novel class of CoA-degrading enzymes that may or may not reside in the peroxisomes.
The narrow and non-overlapping tissue distribution of Nudt19 and Nudt7 in kidneys and liver, respectively, is intriguing. Another major goal of this study was to investigate the distinctive biochemical and regulatory properties of these enzymes to gain insight into the physiological roles that Nudt19 and Nudt7 may play in their respective organs. Nudt19 (42 kDa) is significantly larger than Nudt7 (27 kDa) and contains unique sequence regions, the most notable of which is the 45-amino acid insertion in the Nudix box, which are not present in Nudt7 (Fig. 6A). Nudt7 expression in the liver decreases in the fasted state and increases in the fed state, showing an inverse correlated with the hepatic concentration of CoA (9,26). Conversely, Nudt19 mRNA and protein levels in the kidneys did not change between the fed and fasted states, despite the fact that this enzyme contributes to the regulation of kidney CoA levels.
While searching for potential small molecule regulators, we identified CDCA as a competitive inhibitor of Nudt19 that did not bind to Nudt7, strongly suggesting that these two enzymes possess distinct CoA-binding sites (Fig. 5). Interestingly, binding of CDCA to Nudt19 depended, at least in part, on Arg-34 (Arg-60 in Nudt7), a highly conserved residue involved in CoA binding and hydrolysis in both Nudt19 and Nudt7 (Figs. 6A and 7D and Table 2). Additional evidence in support of a different organization of the CoA-binding sites between the two isoforms is provided by the different activities that Nudt19 and Nudt7 displayed against substrates such as acetyl-CoA and mBB-CoA and by the dramatic decrease in k cat observed upon exchange of the CoA and Nudix boxes between the two enzymes. Furthermore, the invariant E 19 N in the Nudix box was found to participate in CoA binding in Nudt19 but not Nudt7, whereas the conserved (Arg/Lys) residue in the CoA box (corresponding to Lys-65 in Nudt7 and Arg-39 in Nudt19) was found to be involved in CoA binding in Nudt7 but not Nudt19.
No structural information on CoA diphosphohydrolases with CoA bound is currently available, and this represents a limitation to the accurate prediction of the mode of CoA binding through homology models. Despite this, mutational analysis based on the Nudt19 model that we generated allowed us to identify residues that are important for CoA binding and hydrolysis (Table 2). These residues were found both within and outside the CoA box. The Nudt19 model also shows that the 45-amino acid insertion within the Nudix box may be mostly disordered. Within a protein, disordered regions can exert different functions, including providing display sites for posttranslational modification and/or mediating protein-protein interactions (49). There are only a few known examples of per-

Differential regulation of Nudt19 and Nudt7
oxisomal proteins regulated through reversible post-translational modifications (50,51), and the role of this potentially disordered sequence, if any, in the regulation Nudt19 activity remains to be elucidated. We also docked Mg 2ϩ and free CoA to the NUDT7 crystal structure to determine whether we could find a pose that would agree with the mutational analysis on the closely related mouse Nudt7 and provide us with a model of CoA binding to NUDT7. As for Nudt19, we selected the most likely models using the position of E 16 92 , E 19 95 , E 20 96 , Arg-60, CoA, and Mg 2ϩ . In these models, Arg-66 (equivalent to Arg-66 in mouse Nudt7) would form a salt bridge with the diphosphate moiety of CoA (data not shown). Although the 10-fold increase in K m of the Nudt7 R66A mutant supports the importance of Arg-66 for substrate binding, the lack of effect of the R66M mutation on the kinetic parameters of the enzyme indicates that the requirement for Arg-66 is likely spatial and geometric and not about the charge (Table 2). Thus, in this case, the docking of CoA to the structure of NUDT7 does not seem to provide a close representation of how CoA binds to this enzyme. Crystallization of NUDT7 and possibly Nudt19 with a substrate bound will be required to yield further details of CoA binding.
Finally, bile acids are synthesized de novo exclusively in the liver, with the last steps in the synthesis and conjugation to glycine and taurine occurring in the peroxisomes (52). Although it may be unlikely for Nudt19 to encounter high concentrations of CDCA in kidney peroxisomes, Nudt7 is presumably exposed to high concentrations of CDCA, cholic acid, and conjugated derivatives in liver peroxisomes. Thus, it is tempting to speculate that the ability of Nudt7 to remain active even at high micromolar concentrations of bile acids is one possible reason for the selection of this enzyme as the major CoA diphosphohydrolase in the liver. As for Nudt19, further studies, including an in depth characterization of the Nudt19 Ϫ/Ϫ mice, will be required to determine what selective advantage(s) this isoform may confer to the kidneys.

Materials
Reagents were purchased from the following suppliers: oligonucleotides, Lipofectamine 2000, cell culture reagents, Alexa Fluor 555-and horseradish peroxidase-conjugated goat antirabbit IgG from Thermo Fisher Scientific; restriction enzymes, DNA polymerases, and the Q5 site-directed mutagenesis kit from New England Biolabs; HEK 293 cells from the American Type Culture Collection; and the GAPDH antibody from Cell Signaling. The PMP70 antibody was purchased from Sigma-Aldrich. The antibody against the full-length mouse Nudt19 was raised in rabbit and purified by antigen affinity chromatography. The Nudt7 antibody was generated as previously described (9). mBB-CoA was synthesized from free CoA and monobromobimane (Echelon Biosciences) as described by Shimada and Mitamura (53). The conversion of free CoA to mBB-CoA was Ͼ99.0%, as estimated by HPLC. All other reagents were of analytical grade or better and were purchased from Sigma-Aldrich or Fisher Scientific, unless otherwise stated.

Animal studies
Nudt19 tm1(KOMP)Vlcg embryos, obtained by replacing the Nudt19 gene with a ZEN-UB1 cassette harboring the lacZ reporter and neo resistance gene, were purchased from the Knock-out Mouse Project (KOMP) repository and implanted into pseudopregnant foster mothers by the Transgenic Animal Core Facility at West Virginia University. The progeny was screened for the deletion of the Nudt19 gene by multiplex PCR analysis using the Accustart II PCR genotyping kit (QuantaBio) and the following KOMP primers: Reg-NeoF (GCAGCCTCT-GTTCCACATACACTTCA), Reg-Nudt19-R (TTCACAACC-TGCCAAAATGGCTACC), Reg-Nudt19-wtF2 (GGAAACG-GGAACAGAGAAACAGG), and Reg-Nudt19-wtR (AAGGA-CTCAACTCTCACCCTAAGCG). The Nudt19 knockout allele yielded a product of 466 bp. Although smaller than the size of 584 bp predicted by the KOMP repository, this 466-bp amplicon contained the expected 3Ј end of the ZEN-UB1 cassette (124 bases) and a sequence corresponding to the genomic region downstream of the Nudt19 gene (342 bp), as confirmed by sequencing. The wildtype allele yielded a product of the expected size of 185 bp. Homozygous Nudt19 Ϫ/Ϫ mice and wildtype C57BL6/N littermate controls were obtained through heterozygous ϫ heterozygous breeding. The mice were fed a standard chow diet (Tekland 2018S) and maintained at a room temperature of 72.0 Ϯ 0.3°F, room humidity of 40% Ϯ 2%, and a 12-h light/12-h dark cycle, with the dark cycle starting at 18:00 h. For experiments that required fasting, the mice were transferred to clean cages containing floor grids, and food was removed for the indicated time. All studies were conducted on 10 -14-week-old male mice and were approved by the Institutional Animal Care and Use Committees of West Virginia University.

CoA analysis, immunoblotting, and RT-PCR
The concentration of total CoA (free CoA plus CoA thioesters) in kidney homogenates was determined after conversion of the cofactor to the monobromobimane derivative, as previously described (25). For Western blotting analysis, flash frozen tissues were homogenized in ice-cold radioimmunoprecipitation assay buffer supplemented with protease inhibitors (Biotool) and centrifuged at 10,000 ϫ g for 10 min at 4°C. Proteins were fractionated on 4 -12% bis-Tris polyacrylamide gels and transferred onto PVDF membranes. The GAPDH antibody was used at a 1:3000 dilution; the Nudt7 and Nudt19 antibodies were used at a 1:10,000 dilution. Bound primary antibodies were detected by chemiluminescence with horseradish peroxidase-conjugated goat anti-rabbit IgG at a 1:45,000 dilution. RNA was isolated from flash frozen tissue as previously described (54). Following the removal of genomic DNA with Turbo DNase-free kit (Thermo Fisher Scientific), Nudt19 mRNA levels were quantified in triplicate by RT-PCR using the Quantitect SYBR Green RT-PCR kit (Qiagen), and the primers were previously reported (54). The relative abundance of the Nudt19 mRNA was calculated using the C T method, and the amount (2 Ϫ⌬CT ) was reported relative to the mRNA of ␤-glucuronidase (GusB) (25).

Plasmid construction and mutagenesis
Wildtype mouse Nudt7 and Nudt19 were expressed as Nterminal hexahistine-tagged proteins from plasmids pKM204 (25) and pKM249, respectively. Plasmid pKM249 was a kind gift from Dr. Suzanne Jackowski (St. Jude Children's Research Hospital). The above plasmids were mutagenized with the Q5 sitedirected mutagenesis kit to generate single-point Nudt7 and Nudt19 mutants and a Nudt19 deletion mutant lacking amino acids 58 -103. Insertion of the correct mutations was confirmed by sequencing analysis by the Genomics Core Facility at West Virginia University. Sequences encoding Nudt7 and Nudt19 chimeras were custom synthesized (Thermo Fisher Scientific) to contain NdeI and NotI restriction sites at the 5Ј and 3Ј ends, respectively, codon-optimized for expression in Escherichia coli and subcloned into pET-28a(ϩ) (Millipore). For the expression and localization of mouse Nudt19 in HEK 293 cells, custom pcDNA3.1(ϩ)-derived constructs (Gen-Script) were assembled to express Nudt19 proteins, codon-optimized for expression in human cells, with N-or C-terminal GFP tags.

Protein purification, analytical gel filtration chromatography, and ITC
All Nudt7 and Nudt19 proteins with N-terminal hexahistidine tags were expressed in BL21 CodonPlus (DE3)-RIPL (Agilent Technologies) cells, following induction with 1 mM isopropyl ␤-D-1-thiogalactopyranoside and growth for 18 -20 h at 18 -25°C. For enzymatic assay, the enzymes were purified as previously described (25). Fractions containing ϳ90% pure enzymes were combined and dialyzed overnight at 4°C in buffer containing 20 mM Tris-HCl, pH 8.0, 300 mM NaCl, 1 mM EDTA, and 1 mM dithiothreitol. Glycerol was added to a final concentration of 50%, and proteins were stored at Ϫ20°C. Protein samples analyzed by analytical gel filtration chromatography were loaded (50 l) onto a Superdex 75 10/300 GL column and eluted at 0.3 ml/min with 50 mM Tris-HCl, 150 mM NaCl, pH 8.0. Protein elution was monitored by measuring the absorbance at 280 nm. For ITC studies, the purification procedure included three chromatographic steps to increase the purity of the proteins to Ͼ98%, and the hexahistidine tag was removed. Briefly, cell pellets were lysed in 50 mM Na 2 HPO 4 /NaH 2 PO 4 , pH 8.0, 300 mM NaCl, 10 mM imidazole and centrifuged at 13,000 rpm for 45 min at 4°C. The clarified supernatant was allowed to bind to the nickel-nitrilotriacetic acid resin for 2 h with stirring, before the suspension was transferred into a column. The resin-bound proteins were washed with 50 mM Na 2 HPO 4 /NaH 2 PO 4 , pH 8.0, 300 mM NaCl, 20 mM imidazole, followed by elution with wash buffer containing 50 -100 mM of imidazole. Protein-containing fractions were combined, dialyzed in buffer A (50 mM Na 2 HPO 4 /NaH 2 PO 4 , pH 8.0, 200 mM NaCl), and incubated with 0.04 unit/ml thrombin (Sigma-Aldrich) overnight at room temperature to remove the hexahistine tag. The salt concentration of the protein samples was then lowered with an additional dialysis step in buffer B (50 mM Na 2 HPO 4 /NaH 2 PO 4 , 50 mM NaCl), before applying the proteins onto a Resource15Q anion exchange column equilibrated in buffer B. Proteins were eluted with a 0 -100% gradient of buffer C (50 mM Na 2 HPO 4 / NaH 2 PO 4 , pH 8.0, 500 mM NaCl), and the fractions containing the proteins of interested were pooled, dialyzed in buffer A, and concentrated to 5 ml. The concentrated protein samples were subsequently loaded onto a HiLoad 26/60 Superdex 75-pg size exclusion column equilibrated in buffer A, eluted in the same buffer, concentrated, and stored at Ϫ80°C.
Binding affinity of wildtype Nudt7 and Nudt19 for CDCA was measured by ITC at 25°C using a VP-ITC instrument (Microcal Inc., Northhampton, MA) and 1.4 ml of purified Nudt7 and Nudt19 dialyzed in ITC buffer (30 mM PIPES, pH 7.5, 50 mM NaCl, 2 mM MgCl 2 ) and diluted to 124 and 55.5 M, respectively. Protein solutions were degassed for at least 20 min. CDCA stock solutions were prepared in the same ITC buffer. After an initial 2-l injection of CDCA solution with a 240-s equilibration time, the compound was titrated in 30 injections of 10 l each, added at 300-s intervals. The duration of each injection was 12 s, and the filter period was 2 s. The stirring speed and Feedback Mode/gain option were set up to 300 rpm and "High" respectively. The heat evolved at 25°C after each ligand injection was obtained by integration of the calorimetric signal after subtracting the average heat of dilution. Binding parameters were determined from titration binding curves using the MicroCal Data Analysis software (Malvern, Westborough, MA).

Enzymatic assays
The CoA diphosphohydrolase activity of the recombinant enzymes was assayed, in duplicate, as previously described (25), with minor modifications. The standard assay contained 10 ng of recombinant Nudt7 or 30 ng of Nudt19, 100 mM Tris-HCl, pH 8.0, 0.1 mg/ml ␥-globulin, an optimized concentration of MgCl 2 , which was 4 mM for Nudt7 or 10 mM for Nudt19, and 250 M of free CoA or acyl-CoA substrates, in a total volume of 40 l. The reactions were incubated at 37°C for 10 min and stopped by the addition of perchloric acid. Following neutralization with potassium carbonate, the reaction mixtures were diluted 5-fold with water, and the UV-visible 3Ј,5Ј-ADP product was separated and quantified by HPLC coupled with UV detection on a 4.6 ϫ 150 mm, 3-m C-18 column (Acclaim 120) kept at 35°C. Elution was performed at 0.7 ml/min as follows: 0 -2.5 min, 100% buffer A (50 mM KH 2 PO 4 , pH 4.6); 2.5-9.0 min linear gradient to 50% B (100% acetonitrile); 9.0 -10.0 min isocratic at 50% B; 10.0 -15.0 min linear gradient from 50% B to 75% B; 15.0 -19.0 min isocratic at 75% B; 19.0 -32.0 min return to 100% A, and column equilibration. A standard curve of 3Ј,5Ј-ADP (Sigma-Aldrich) subjected to the same reaction conditions was used to quantify the amount of product formed. For each mutant and chimeric enzyme, protein curves were generated to select concentrations within the activity linear range. The length of the assay was also extended to 2 h for mutants exhibiting marginal activity. The kinetic parameters K m , V max , and k cat were determined using free CoA as the substrate, with concentrations ranging from 31.25 M to 10 mM, and obtained from curves fitted to a Michaelis-Menten equation using GraphPad Prism 6 (GraphPad Software).

Differential regulation of Nudt19 and Nudt7
To monitor the production of 3Ј,5Ј-ADP in kidney extracts, flash-frozen tissue (ϳ50 mg) was homogenized in 750 l of 20 mM Tris-HCl, pH 8.0. The homogenate was incubated on ice for 10 min and centrifuged at 20,000 ϫ g for 10 min. The supernatant was loaded on to a PD10 column equilibrated in 20 mM Tris-HCl, pH 8.0, to remove the small molecules. The eluted proteins (50 -200 g) were incubated with 250 M free CoA for 5 min at 37°C. The reactions were then stopped and analyzed as described above.

Immunofluorescence
HEK 293 cells were cultured at 37°C and 5% CO 2 in phenol red-free Opti-MEM I medium supplemented with 4% FBS, 100 units/ml penicillin, 100 g/ml streptomycin, and 0.25 g/ml amphotericin B. To localize Nudt19, the cells were seeded on 12-mm coverslips at low density and grown in the same medium to ϳ80% confluence. Transfections were performed with 1.25 g of plasmid DNA using Lipofectamine 2000, as per the manufacturer's instructions. Where indicated, the cells were incubated with MitoTracker Orange CMTMRos at a concentration of 25 nM for 30 min at 37°C. The cells were fixed with 4% paraformaldehyde and blocked in PBS containing 5% FBS and 0.3% Triton X-100 for 1 h at room temperature. The PMP70 antibody was used at a 1:1000 dilution in 1% BSA, 0.3% Triton X-100 for 1 h at room temperature. Following 10 washes with PBS, the cells were incubated for 1 h at room temperature with Alexa Flour 555 donkey anti-mouse IgG, used at a dilution of 1:1000 in 1% BSA, 0.3% Triton X-100. Slides were mounted using Prolong Gold Antifade with DAPI (Thermo Fisher). Imaging experiments were performed in the West Virginia University Imaging Facilities using a Nikon A1R/N SIM-E confocal microscope. Image analysis was conducted using ImageJ.

Molecular modeling
Homology models with bound Mg 2ϩ were built using the software suite YASARA (43) (www.yasara.org). 4 4 After an induced fit docking step, the top five poses were evaluated for the correct relative orientation of the diphosphate moiety of free CoA, the bound Mg 2ϩ , and Nudt19 Arg-34 or NUDT7 Arg-60. Sequence alignments were obtained using Clustal W (55) and combined with secondary structure elements using ESPript 3 (56). The Nudt19 model and the crystal structures of human NUDT7 and the bacterial CoAse were aligned using Phenix Ensembler (23).

Statistical analysis
Unless otherwise stated, all data are reported as the means with Ϯ S.D. Statistical significance was calculated by unpaired two-tailed Student's t test using GraphPad Prism 6 (GraphPad Software).