The 3-hydroxyacyl-CoA dehydratases HACD1 and HACD2 exhibit functional redundancy and are active in a wide range of fatty acid elongation pathways

Differences among fatty acids (FAs) in chain length and number of double bonds create lipid diversity. FA elongation proceeds via a four-step reaction cycle, in which the 3-hydroxyacyl-CoA dehydratases (HACDs) HACD1–4 catalyze the third step. However, the contribution of each HACD to 3-hydroxyacyl-CoA dehydratase activity in certain tissues or in different FA elongation pathways remains unclear. HACD1 is specifically expressed in muscles and is a myopathy-causative gene. Here, we generated Hacd1 KO mice and observed that these mice had reduced body and skeletal muscle weights. In skeletal muscle, HACD1 mRNA expression was by far the highest among the HACDs. However, we observed only an ∼40% reduction in HACD activity and no changes in membrane lipid composition in Hacd1-KO skeletal muscle, suggesting that some HACD activities are redundant. Moreover, when expressed in yeast, both HACD1 and HACD2 participated in saturated and monounsaturated FA elongation pathways. Disruption of HACD2 in the haploid human cell line HAP1 significantly reduced FA elongation activities toward both saturated and unsaturated FAs, and HACD1 HACD2 double disruption resulted in a further reduction. Overexpressed HACD3 exhibited weak activity in saturated and monounsaturated FA elongation pathways, and no activity was detected for HACD4. We therefore conclude that HACD1 and HACD2 exhibit redundant activities in a wide range of FA elongation pathways, including those for saturated to polyunsaturated FAs, with HACD2 being the major 3-hydroxyacyl-CoA dehydratase. Our findings are important for furthering the understanding of the molecular mechanisms in FA elongation and diversity.

The enzymes responsible for the first and third steps of the FA elongation cycle have multiple isozymes in mammals (ELOVL1-7 and HACD1-4). The substrate specificities of the FA elongases ELOVL1-7 have already been determined; each exhibits characteristic substrate specificity toward acyl-CoAs with different chain lengths and numbers of double bonds (6,7,11,22). In contrast, the substrate specificities of HACD1-4 are unknown, mainly due to limitations in the commercial availability of 3-OH acyl-CoA species. We previously demonstrated that purified HACD1-4 all exhibit activity in vitro toward the only commercially available 3-OH palmitoyl-CoA substrate (26). In that assay, HACD2 showed the greatest activity and that of HACD4 was the lowest (ϳ16-fold lower than that of HACD2 in terms of V max ). HACD2 mRNA is ubiquitously expressed in tissues (27), and HACD3 mRNA is expressed in many tissues, such as brain, kidney, liver, and placenta (26). In contrast, expression of HACD1 and HACD4 mRNA is restricted to muscle tissue (skeletal muscle and heart) and leukocytes, respectively (26,28). Mutations in HACD1 cause myopathy in humans and dogs (29,30). Hacd1 KO mice also exhibit a myopathic phenotype, with reduced body weight, muscle mass, muscle force, and muscle fiber diameter (31). During myogenesis, myoblasts are fused into multinucleated myotubes, followed by maturation into myofibers. The fusion process is retarded in Hacd1 KO myoblasts (31).
In this study, we aimed to identify as-yet-undetermined substrate specificities of HACDs. For this purpose, we performed lipidomics analyses on newly generated Hacd1 KO mice. Furthermore, the 3-OH acyl-CoA dehydratase activity of each HACD was investigated by FA elongation assay, in which commercially available acyl-CoAs/FAs were used as substrates, instead of direct measurement of 3-OH acyl-CoA dehydratase activity using 3-OH acyl-CoA substrates. Our results indicate that HACD1 and HACD2 exhibit broad substrate specificities. They are active toward saturated, monounsaturated, and polyunsaturated 3-OH acyl-CoAs of long-to very long-chain FAs, with HACD2 exhibiting greater activity than HACD1. In con-trast, HACD3 showed only weak activity in saturated and monounsaturated FA elongation pathways, and no HACD4 activity was detected.

Moderate reduction in 3-OH acyl-CoA dehydratase activity in Hacd1 KO mice
To reveal the substrate specificity of HACD1 and the pathogenesis of the myopathy caused by HACD1 mutations, Hacd1 KO mice were created. HACD/Phs1 family members contain the active-site residues Tyr and Glu within the fifth transmembrane segment (32,33). The Hacd1 KO construct was designed to replace exon 6, which encodes the active-site residues, with a neomycin-resistant gene ( Fig. 2A). Gene disruption was confirmed by genomic PCR (Fig. 2B). The Hacd1 KO mice exhibited smaller body size (Fig. 2C) and reduced body and skeletal muscle (gastrocnemius) weight at 1 and 6 months of age (Fig. 2, D and E), compared with the control mice, as has been reported previously (31).
In WT skeletal muscle, expression levels of HACD1 mRNA were much higher than those of other HACDs (Fig. 3A). How-  The positions of the primers (p1, p2, and p3) used for genomic PCR are denoted with arrows. DTA, diphtheria toxin A gene for negative selection; neo, neomycin-resistant gene for positive selection. B, genomic DNAs prepared from the tails of Hacd1 ϩ/ϩ , Hacd1 ϩ/Ϫ , and Hacd1 Ϫ/Ϫ mice were subjected to PCR using primers p1, p2, and p3. The amplified fragments were separated by agarose gel electrophoresis, followed by staining with ethidium bromide. C, Hacd1 ϩ/ϩ and Hacd1 Ϫ/Ϫ mice at 3 months old. D and E, body weight (D) and gastrocnemius weight (E) of female Hacd1 ϩ/ϩ and Hacd1 Ϫ/Ϫ mice at 1 and 6 months old. Values represent the means Ϯ S.D. from 5 to 8 mice. Statistically significant differences are indicated (**, p Ͻ 0.01; *, p Ͻ 0.05; Student's t test). mo, month.
Next, we examined the effect of Hacd1 gene disruption on the product levels. Because the acyl-CoA products elongated via the FA elongation cycle are mainly used for membrane lipid synthesis, we measured the levels of three major lipid classes in skeletal muscle by liquid chromatography (LC)tandem mass spectrometry (MS/MS). These were the sphingolipid sphingomyelin (Fig. 3D) and the glycerophospholipids phosphatidylcholine (Fig. 3E) and phosphatidylinositol (Fig.  3F). There were no differences in the levels of these lipids among WT (Hacd1 ϩ/ϩ ), Hacd1 heterozygous KO (Hacd1 ϩ/Ϫ ), and Hacd1 homozygous KO (Hacd1 Ϫ/Ϫ ) mice, irrespective of chain lengths (Fig. 3, D-F).
To further examine the effect of Hacd1 gene disruption on the FA elongation cycle, myoblasts were prepared from WT and Hacd1 KO mice. After differentiation into myotube cells, they were incubated with deuterium (d)-labeled FAs for 24 h, and their metabolism was traced by LC-MS/MS analysis. Exogenously added FAs are metabolized to other FAs via elongation and/or desaturation within cells (Fig. 4A). When cells were labeled with d 31 -palmitic acid (C16:0-COOH) (palmitic acid with 31 deuteriums (d 31 )), d 31 -labeled saturated and monounsaturated C16 -C26 FAs were detected in WT cells (Fig. 4B). Almost no differences were observed in the compositions of d 31 -labeled FAs between WT and Hacd1 KO cells. Furthermore, tracer analyses using d 9 -oleic acid (C18:1-COOH) indicated that d 9 -oleic acid metabolism was indistinguishable between WT and Hacd1 KO cells (Fig. 4C). Mammals cannot produce n-6 and n-3 polyunsaturated FAs endogenously, due to a lack of FA ⌬12 desaturase and FA ⌬15 desaturase, and must therefore obtain these FAs from foods. Food-derived polyunsaturated FAs are subjected to repetitive elongation and desaturation in cells and are converted to other polyunsaturated FAs A, total RNAs prepared from the gastrocnemius of female WT mice at 1 month old were subjected to SYBR Green-based real-time quantitative RT-PCR using specific primers for Hacd1, Hacd2, Hacd3, Hacd4, and Gapdh. Values represent the means Ϯ S.D. relative to Gapdh expression levels from three independent reactions. Statistically significant differences are represented (**, p Ͻ 0.01; Tukey's test). B, total membrane fractions (20 g of protein) prepared from the gastrocnemius of female Hacd1 ϩ/ϩ , Hacd1 ϩ/Ϫ , and Hacd1 Ϫ/Ϫ mice at 1 month old were incubated with 0.01 Ci of [ 14 C]3-OH palmitoyl-CoA for 10 min at 37°C. Lipids were saponified, acidified, extracted, separated by normal-phase TLC, and detected using a BAS-2500 image analyzer. Values represent the means Ϯ S.D. of the ratio of the product lipid radioactivity to total lipid radioactivity from three independent reactions. Statistically significant differences are represented (**, p Ͻ 0.01; Tukey's test). C, total RNAs prepared from the gastrocnemius of female Hacd1 ϩ/ϩ , Hacd1 ϩ/Ϫ , and Hacd1 Ϫ/Ϫ mice at 1 month old were subjected to SYBR Green-based real-time quantitative RT-PCR using specific primers for Hacd2, Hacd3, Hacd4, and Gapdh. Values represent the means Ϯ S.D. relative to Gapdh expression levels from seven (Hacd1 ϩ/ϩ and Hacd1 Ϫ/Ϫ ) or five (Hacd1 ϩ/Ϫ ) independent experiments. No statistically significant differences were detected between samples using the Tukey-Kramer method. D-F, lipids were extracted from the gastrocnemius of Hacd1 ϩ/ϩ , Hacd1 ϩ/Ϫ , and Hacd1 Ϫ/Ϫ mice at P0, and sphingomyelin (D), phosphatidylcholine (E), and phosphatidylinositol (F) species were analyzed by LC-MS/MS. Values are the means Ϯ S.D. of the quantity of each lipid species containing FAs with the indicated chain length and desaturation number from three independent experiments. No statistically significant differences were detected between the samples by Tukey's test.

Redundant substrate specificities of HACD1 and HACD2 in saturated and monounsaturated FA elongation pathways
Relatively weak effects on the 3-OH acyl-CoA dehydratase activity as a result of Hacd1 gene KO (Fig. 3B) suggest redundancy among the HACDs. We previously reported that ectopic expression of HACD1 or HACD2 in a PHS1-shutoff yeast strain complemented growth defects, whereas neither HACD3 nor HACD4 exhibited such activity (26). In WT yeast Saccharomyces cerevisiae, the VLCFAs are almost exclusively C26:0 (34), whereas the production of at least some C24:0 VLCFAs is necessary for normal growth (11). Consequently, the results of the growth complementation analysis suggest that both HACD1 and HACD2 possess the ability to produce C24:0 or longer VLCFAs, and therefore that there is some redundancy in their function.
We first examined the activity of HACD1 and HACD2 in the saturated FA elongation pathway, using a yeast system. To minimize the endogenous 3-OH acyl-CoA dehydratase activity, we

Substrate preferences and redundancy of HACDs
created phs1⌬ htd2⌬ cells expressing human ceramide synthase CERS5. In these cells both yeast HACD homolog PHS1 and the mitochondrial 3-OH acyl carrier protein dehydratase HTD2 were deleted. Htd2 is a component of FA synthase type II (35). Because VLCFA production is essential for yeast growth, any genes involved in FA elongation, including PHS1, cannot be deleted under normal conditions (4). In yeast, most VLCFAs are used for sphingolipid synthesis, and the lethality of the VLCFA-deficient cells can be attributed to the loss of sphingolipid production due to the substrate specificity of yeast ceramide synthases. Therefore, genes of the FA elongation machinery can be deleted if yeast cells are engineered to produce ceramides/sphingolipids, such as by ectopic expression of the human ceramide synthase CERS5, which uses LCFAs for ceramide synthesis (36). Phs1, HACD1, or HACD2 was expressed as triple FLAG (3xFLAG)-tagged protein in the phs1⌬ htd2⌬/CERS5 cells. Expression levels of HACD2 were the highest, and those of Phs1 were the lowest (Fig. 5A). Although the only commercially available 3-OH acyl-CoA species is 3-OH palmitoyl-CoA, several acyl-CoA species can be obtained. Therefore, we performed an in vitro FA elongation assay, in which [ 14 C]malonyl-CoA and acyl-CoA were used as substrates for the FA elongation cycle, instead of the 3-OH acyl-CoA dehydratase assay using a 3-OH acyl-CoA substrate. Membrane fractions were prepared from the yeast cells and were subjected to in vitro FA elongation assay using stearoyl-CoA (C18:0-CoA) as an acyl-CoA substrate. After the reactions, the products were converted to FA methyl esters (FAMEs), separated by reversephase TLC, and detected using autoradiography. In the control Figure 5. HACD1 and HACD2 exhibit broad substrate specificities toward saturated and monounsaturated 3-OH acyl-CoAs. A-F, total membrane fractions were prepared from the yeast KMY81 (phs1⌬ htd2⌬/pAB119 (3xFLAG-CERS5)) cells bearing the pAK739 (vector), pRF6 (PHS1-3xFLAG), pYS10 (HMF-HACD1), or pTN28 (HMF-HACD2) plasmid and subjected to immunoblotting with anti-FLAG and anti-Pma1 (membrane protein loading control) antibodies (A) and FA elongation assays using [ 14 C]malonyl-CoA (B) or [ 13 C]malonyl-CoA (C-F). B, total membrane fractions (20 g of protein) were incubated with 75 nCi of [ 14 C]malonyl-CoA and 20 M stearoyl-CoA for 30 min at 37°C. After the reaction, lipids were saponified, converted to FAMEs, separated by reverse-phase TLC, and detected using a BAS-2500 image analyzer. The acyl-CoA and 3-OH acyl-CoA intermediates of the FA elongation cycle, starting from stearoyl-CoA (C18:0), are illustrated in the right panel. C-F, total membrane fractions (20 g of protein) were incubated with 100 M [ 13 C]malonyl-CoA and 10 M stearoyl-CoA (C and D) or oleoyl-CoA (E and F) for 30 min at 37°C. After the reaction, lipids were saponified, derivatized to AMPP amides, and analyzed by LC-MS/MS. Values are the means Ϯ S.D. of the quantities of acyl-CoA derivatives (C and E) or 3-OH acyl-CoA derivatives (D and F) with the indicated chain length and desaturation number from three independent experiments. Statistically significant differences are calculated by Tukey's test (**, p Ͻ 0.01; *, p Ͻ 0.05). Significant differences are represented by asterisks above the lines connecting the bar graphs. Otherwise, the asterisks represent significant differences from the vector control. vec, vector.

Substrate preferences and redundancy of HACDs
phs1⌬ htd2⌬/CERS5 cells bearing the vector, C18:0-CoA was converted to 3-OH 20:0-CoA via 3-keto 20:0-CoA by endogenous FA elongases (Fen1 and Sur4) and 3-ketoacyl-CoA reductase Ifa38 (Fig. 5B). No further conversion was observed, due to the lack of 3-OH acyl-CoA dehydratase. When the PHS1 gene was added back, the cells recovered the ability to produce C20:0-to C26:0-CoAs. Note that 3-ketoacyl-CoA and trans-2enoyl-CoA intermediates were not observed in this assay, because the second and fourth reactions in the FA elongation cycle are rapid. Expression of HACD1 caused production of C20:0-to C26:0-CoAs as well, although their levels were lower than with Phs1 expression. Instead, more 3-OH acyl-CoA intermediates were accumulated. Expression of HACD2 resulted in production of C20:0-to C26:0-CoAs with similar efficiency to Phs1. The levels of 3-OH acyl-CoAs were higher than seen with Phs1, but lower than with HACD1. These results were assessed in more detail by LC-MS/MS using the stable isotope 13 C-labeled malonyl-CoA. Again, C20:0-to C26:0-CoAs were produced under expression of Phs1, HACD1, or HACD2, and 3-OH acyl-CoA levels were the highest in HACD1-expressing membranes (Fig. 5, C and D).
We also examined the activities of HACD1 and HACD2 in the monounsaturated FA elongation pathway using [ 13 C]malonyl-CoA and oleoyl-CoA (C18:1-CoA) as substrates. Expression of Phs1, HACD1, or HACD2 resulted in production of C20:1-to C26:1-CoAs (Fig. 5E). In most cases, the levels of the acyl-CoAs produced by HACD1-and HACD2-expressing membranes were slightly lower than those produced by Phs1expressing membranes. The levels of 3-OH acyl-CoAs were higher in HACD1-and HACD2-expressing membranes than those in Phs1-expressing membranes, but there were no significant differences between those produced by HACD1-and HACD2-expressing membranes (Fig. 5F). Together with the results obtained from the FA elongation assay using C18:0-CoA, these results indicate that both HACD1 and HACD2 exhibit activities toward C20-to C26 saturated and monounsaturated 3-OH acyl-CoAs. However, the HACD1-mediated FA elongation cycle produces more saturated 3-OH acyl-CoA intermediates (but not monounsaturated 3-OH acyl-CoAs) than the HACD2-mediated cycle.

Broad and redundant substrate specificities of HACD1 and HACD2
Functions of HACD1 and HACD2 in the polyunsaturated FA elongation cycle could not be determined by the above yeast system, because yeast FA elongation machinery cannot elongate polyunsaturated acyl-CoAs. We therefore used a mammalian system, in which the HACD gene(s) were disrupted in the haploid human cell line HAP1 by CRISPR/Cas9 genome editing. Quantitative real-time RT-PCR analysis revealed that expression levels of HACD3 mRNA were the highest in HAP1 cells, followed by HACD1 and HACD2 (Fig. 6A). No expression of HACD4 was detected in HAP1 cells.
HACD1 KO, HACD2 KO, HACD3 KO, and HACD1 HACD2 double KO (DKO) HAP1 cells were created. These cells were incubated with d 31 -C16:0-COOH for 6 h. Lipids were then extracted and hydrolyzed to FAs by alkaline treatment, and d 31labeled FAs were quantified by LC-MS/MS analysis. Almost no differences were observed in the compositions of d 31 -labeled FAs between the control, HACD1 KO, and HACD3 KO cells (Fig. 6B). In contrast, ՆC18 saturated and monounsaturated FAs were reduced in HACD2 KO cells, concomitant with increases in C16 FAs. The HACD1 HACD2 DKO caused a greater decrease in C18:0 to C22:0 FAs and increase in C16 FAs than the HACD2 single KO. Trace amounts of d 31 -labeled ՆC18 saturated and monounsaturated FAs were still detected, even in the HACD1 HACD2 DKO cells, suggesting that HACD3 or other unknown 3-OH acyl-CoA dehydratases produced them under the HACD1 and HACD2-null conditions.
When cells were labeled with d 9 -C18:1-COOH, similar results were obtained as with d 31 -C16:0-COOH labeling. There was a decrease in ՆC20 monounsaturated FAs in HACD2 KO cells and HACD1 HACD2 DKO cells, whereas C18 monounsaturated FA was increased in these cells (Fig. 6C). d 11 -C18: 2(n-6)-COOH (Fig. 6D) and d 5 -C18:3(n-3)-COOH (Fig. 6E) labeling experiments indicated that C20, C24, and C26 polyunsaturated FAs were decreased in HACD2 KO cells relative to control cells, and further decreased in HACD1 HACD2 DKO cells in most cases, whereas C18 polyunsaturated FAs were increased. The exception was C20:5(n-3)-COOH, which was slightly increased in HACD2 KO cells relative to control cells (Fig. 6E). The levels of C20 polyunsaturated FAs are determined by the balance between the increase due to C18-to-C20 conversion and the decrease due to C20-to-C22 conversion. HACD2 disruption appeared to have a stronger effect on C20-to-C22 polyunsaturated FA conversion than on C18-to-C20 polyunsaturated FA conversion, which might explain the slight increase in C20:5(n-3)-COOH levels in HACD2 KO cells. Single gene disruption of HACD1 or HACD3 had almost no effect on polyunsaturated FA elongation (Fig. 6, D and E). Thus, HACD2 is the major 3-OH acyl-CoA dehydratase, not only for saturated and monounsaturated FA elongation, but also for polyunsaturated FA elongation, and HACD1 has a redundant function, albeit weaker than HACD2.

Discussion
In the FA elongation cycle in mammals, there are multiple isozymes able to catalyze the first step, condensation (ELOVL1-7), and the third step, dehydration (HACD1-4). Although knowledge relating to the differences in the substrate specificities and physiological functions of the ELOVLs has been accumulated (6,7,11,22), such information was limited for HACD1-4. Although it was known that HACD1-4 exhibits activities toward 3-OH palmitoyl-CoA in vitro (26), their activities toward other 3-OH acyl-CoA species, as well as their actual activities within cells, had not been determined. ELOVL1-7 exhibit characteristic substrate specificities: ELOVL1, saturated and monounsaturated C20-to C24-CoAs; ELOVL2, polyunsaturated C20-to C22-CoAs; ELOVL3, C16to C22-CoAs; ELOVL4, ՆC24-CoAs; ELOVL5, polyunsaturated C18-and C20-CoAs; ELOVL6, saturated and monounsaturated C12-to C16-CoAs; and ELOVL7, C16-to C20-CoAs (6,7,11,22,37). This study revealed that, in contrast to the ELOVLs, the substrate preferences of HACD1 and HACD2 are quite broad; both were active in all steps of the FA elongation pathways of saturated, monounsaturated, and n-6 and n-3 polyunsaturated FAs (Figs. 5-7). In mouse skeletal muscle, the expression levels of HACD1 mRNA were much higher (ϳ75fold) than those of HACD2 mRNA (Fig. 3A). However, 3-OH acyl-CoA dehydratase activity toward 3-OH palmitoyl-CoA in Hacd1 KO skeletal muscle was reduced only moderately (to ϳ60% of the control), and the compositions of the membrane lipids and metabolism of exogenously added FAs were unchanged (Figs. 3 and 4). It is likely that HACD2 was responsible for the remaining activity in Hacd1 KO muscle. In HAP1 cells, HACD2 was the predominant 3-OH acyl-CoA dehydratase over HACD1 (Fig. 6, B-E), although the expression levels of HACD2 mRNA were 71% of those of HACD1 mRNA (Fig.  6A). Together with the ubiquitous tissue distribution of HACD2 (27), these results suggest that HACD2 is the major 3-OH acyl-CoA dehydratase in the whole body.
The need for HACD1 is unclear, because HACD2 has the same function as and higher activity levels than HACD1. The only difference that we observed between HACD1 and HACD2 in this study was in the levels of saturated 3-OH acyl-CoA intermediates during the FA elongation process; 3-OH acyl-CoAs were accumulated at higher levels in HACD1-expressing membranes than in HACD2-expressing membranes (Fig. 5, B and  D). In the FA elongation assay, acyl-CoAs are the main detectable products, although 3-OH acyl-CoAs are also detected at

Substrate preferences and redundancy of HACDs
low levels (38). This indicates that the first (condensation) step catalyzed by ELOVLs is rate-limiting, and the third (dehydration) step catalyzed by HACDs is secondarily rate-limiting in the FA elongation cycle. Neither 3-ketoacyl-CoAs nor trans-2enoyl-CoAs are detected under normal conditions (38), indicating that the second step, catalyzed by KAR, and the fourth step, catalyzed by TECR, are rapid. These fast reactions seem to be achieved by the interplays between the ELOVLs and KAR as well as those between the HACDs and TECR, which enable direct transfer of the FA elongation cycle intermediates from the ELOVLs/HACDs to KAR/TECR without release from the FA elongation machinery (38,39). However, the detection of 3-OH acyl-CoA intermediates in the FA elongation cycle (Fig.  5) (38) implies that some 3-OH acyl-CoAs are released from the FA elongation machinery and used for other purposes. We hypothesize that 3-OH acyl-CoAs or their derivatives have certain regulatory functions in skeletal muscle, such as cell growth, differentiation, and fusion, and that this may explain the high levels of HACD1 in skeletal muscle.
Mutations in HACD1 cause myopathy in humans and dogs (29,30). Hacd1 disruption also causes a myopathic phenotype in mice (31), although the molecular mechanism remains

Substrate preferences and redundancy of HACDs
largely unclear. We also observed decreased body size, body weight, and skeletal muscle weight in Hacd1 KO mice (Fig. 2, C-E). Because HACD1 is involved in FA elongation, it had been hypothesized that membrane lipid compositions, especially those of longer lipids, might be affected by Hacd1 disruption in skeletal muscle. However, we could not detect any differences in the compositions of membrane lipids (Fig. 3, D-F). This suggests that certain changes in lipid composition may occur during the developmental stages, such as the mesoderm to muscle or myoblast to myofiber differentiation stages.
Although the expression levels of HACD3 mRNA were the highest among the HACDs in HAP1 cells, gene disruption of HACD3 had no effect on FA elongation (Fig. 6). The activity of HACD3 toward saturated and monounsaturated 3-OH acyl-CoAs was only detected when it was overproduced in HACD1 HACD2 DKO cells (Fig. 7). In this study, no HACD4 activity was detected. To date, HACD4 activity toward 3-OH C16:0-CoA has only been detected in vitro when HACD4 was solubilized with the nonionic detergent Triton X-100, although the V max value was the lowest among the HACDs (26). In the cellbased assay performed in this study, we examined the activity of HACD4 toward 3-OH acyl-CoAs with ՆC18, but not toward that with C16, in the saturated FA elongation pathway. Considering the low and zero activity of HACD3 and HACD4, respectively, in the cell-based assay, we hypothesize that their natural substrates are not 3-OH acyl-CoAs with ՆC18 but rather are specialized forms of 3-OH acyl-CoAs, such as those containing a short or branched chain. Because HACD4 is expressed specifically in leukocytes (26), it is possible that it functions in the metabolism of pathogen-derived 3-OH short FAs. For example, the lipid A portion of lipopolysaccharides of the Gram-negative bacteria E. coli contains 3-OH C14:0-COOH (40). Further studies are needed, however, to elucidate the physiological functions and exact substrates of HACD3 and HACD4.

Generation of Hacd1 ؊/؊ mice
The Hacd1 KO targeting vector contains the upstream region (ϳ7,000 bp) of exon 6, the Pgk-neo (neomycin-resistant gene under the control of the Pgk promoter) cassette flanked by two loxP sequences, the downstream region (ϳ2,000 bp) of exon 6, and the Tk-DTA (diphtheria toxin A under the control of thymidine kinase promoter) cassette ( Fig. 2A). The targeting vector was constructed using the recombineering method (41) with the bacterial artificial chromosome clone bMQ-431D1 (BACPAC Resources Center; Oakland, CA) prepared from chromosomal DNAs of Mus musculus AB2.2 (129S7/SvEvBrd-Hprtb-m2) in ES cells (42). After transfection of the linearized targeting vector into E14 ES cells, G418-resistant clones were selected. Genomic DNAs were prepared from each clone, and homologous recombination was confirmed by genomic PCR using primers p1 and p2 (Table 1 and Fig. 2A). Recombination was also confirmed by Southern blotting. Positive ES clones were injected into C57BL/6J blastocysts, and the resulting chimera mice were crossed with C57BL/6J mice to obtain Hacd1 ϩ/Ϫ mice. After crossing Hacd1 ϩ/Ϫ mice with C57BL/6J mice repeatedly (Ն10 generations of back-crossing), Hacd1 Ϫ/Ϫ mice were generated by intercrossing the Hacd1 ϩ/Ϫ mice. Genotyping was performed by PCR using genomic DNAs and primers p3 and p2 for detection of the Hacd1 WT allele and p1 and p2 for detection of the Hacd1 KO allele (Table 1 and Fig.  2A). All mice were kept at 23 Ϯ 1°C in a 12-h light/dark cycle with a standard chow diet (PicoLab Rodent Diet 20; LabDiet, St. Louis, MO) and water available ad libitum. The animal experiments performed in this study were approved by the institutional animal care and use committees of Hokkaido University and RIKEN Brain Science Institute.

In vitro 3-OH acyl-CoA dehydratase assay
Total cell lysates were prepared from mouse gastrocnemius by homogenizing the tissues, using a homogenizer, in buffer A (50 mM HEPES/NaOH (pH 7.5), 150 mM NaCl, 10% glycerol, 1ϫ protease inhibitor mixture (Complete, EDTA-free; Roche Diagnostics, Basel, Switzerland), 1 mM PMSF, and 1 mM DTT), followed by sonication and removal of cell debris by centrifugation (400 ϫ g, 3 min, 4°C). They were then subjected to centrifugation at 100,000 ϫ g for 30 min at 4°C, and the resulting pellet (total membrane fraction) was suspended in buffer A. The in vitro 3-OH acyl-CoA assay was performed as described previously (32) in reaction buffer I (total volume of 50 l; buffer A containing 1 mM CaCl 2 , 2 mM MgCl 2 , and 0.1% digitonin). The total membrane fractions (20 g of protein) were incubated with 0.01 Ci of [ 14 C]3-OH palmitoyl-CoA (55 mCi/ mmol; American Radiolabeled Chemicals, St. Louis, MO) for 10 min at 37°C. After terminating the reactions by adding 25 l of 75% KOH (w/v) and 50 l of ethanol, the lipids were saponified for 1 h at 70°C. Samples were then acidified by adding 100 l of 5 M HCl and 50 l of ethanol. Lipids were extracted with 700 l of hexane, dried, suspended in 20 l of chloroform, separated by TLC on Silica Gel 60 high performance TLC plates (Merck, Darmstadt, Germany) with hexane/diethyl ether/acetic acid (30:70:1, v/v) as the solvent system, and detected using a BAS-2500 image analyzer (GE Healthcare, Little Chalfont, UK).

LC-MS/MS analysis
Mouse gastrocnemius muscles (16 mg) were chopped and suspended in 144 l of CHCl 3 /MeOH/formic acid (100:200:1, v/v). Lipids were extracted by mixing samples vigorously in tubes containing zirconia beads for 1 min at 4°C using Micro Smash MS-100 (TOMY Seiko, Tokyo, Japan). After centrifugation (1,000 ϫ g, 10 min, 4°C), the supernatant was mixed with 48 l of chloroform and 86.4 l of water, successively, with vigorous mixing. Phases were separated by centrifugation (9,100 ϫ g, 1 min, room temperature), and the organic phase was recovered and dried. Lipids were dissolved in 100 l of methanol for LC-MS/MS analysis.
FA elongation assay using [ 13 C]malonyl-CoA was performed as follows. Total membrane fractions (20 g of protein) were incubated with 100 M [ 13 C]malonyl-CoA (Sigma) and 10 M acyl-CoA (stearoyl-CoA or oleoyl-CoA; Avanti Polar Lipids, Alabaster, AL) complexed with 0.2 mg/ml FA-free bovine serum albumin in 50 l of reaction buffer III (50 mM HEPES/ NaOH (pH 7.5), 150 mM NaCl, 10% glycerol, 1ϫ protease inhibitor mixture, 1 mM PMSF, 1 mM DTT, 2 mM MgCl 2 , and 1 mM CaCl 2 ) containing 200 g/ml cerulenin and 1 mM NADPH for 30 min at 30°C. The reactions were terminated by adding 25 l of 75% KOH (w/v) and 50 l of ethanol, then saponified for 1 h at 70°C, and acidified by adding 100 l of 5 M formic acid with 50 l of ethanol. Lipids were extracted with 750 l of hexane and dried. FAs were derivatized to AMPP amides using the AMP ϩ mass spectrometry kit (Cayman Chemical, Ann Arbor, MI), according to the manufacturer's protocol, and resolved by LC-MS/MS as described above.

Cell culture
HAP1 is a near-haploid human cell line derived from myelogenous leukemia (50) and was purchased from the American Type Culture Collection (Manassas, VA). HAP1 cells were grown in Iscove's modified Dulbecco's medium (12440-053; Thermo Fisher Scientific) containing 10% FBS, 100 units/ml penicillin, and 100 g/ml streptomycin. Transfections were performed using Lipofectamine Plus Reagent (Thermo Fisher Scientific), according to the manufacturer's instruction. The transfected cells were subjected to selection in 2 g/ml puromycin for 2 days.
Mouse primary cultured myoblasts were prepared essentially as described previously (51). The gastrocnemius, tibialis anterior muscles, and thigh muscles were collected from the hindlimbs of the 1-month-old mice, shredded in PBS, and incubated with 5 ml of cell separation solution I (DMEM (D6429; Sigma) containing 10% FBS and 1.4 units/ml collagenase D (Roche Diagnostics)) for 1 h at 37°C. The muscle cells were then disaggregated by passing them through a syringe with an

Lipid labeling assay
Cells were incubated with 1 M deuterium-labeled FA (d 31palmitic acid (Cayman Chemical), d 9 -oleic acid (Avanti Polar Lipids), d 11 -linoleic acid (Cayman Chemical), or d 5 -␣-linolenic acid (Cayman Chemical), where d 31 means that the FA contained 31 deuteriums) for 24 h at 37°C. Cells were then washed with PBS, detached from the dish by incubating them in 0.05% trypsin/EDTA solution, and resuspended in 100 l of water. Lipids were extracted by mixing with successive additions of 375 l of chloroform/methanol/formic acid (100:200:1, v/v), 125 l of chloroform, and 125 l of water. Phases were then separated by centrifugation, and the organic phase was recovered and treated with 71 l of 0.5 M NaOH to hydrolyze the ester bonds (release FAs from lipids) for 1 h at 37°C. After neutralization with 35.5 l of 1 M formic acid, lipids were extracted via successive additions of 175 l of chloroform and 250 l of water with mixing. Phases were separated by centrifugation, and the organic phase was recovered, dried, derivatized with an AMP ϩ mass spectrometry kit, and resolved by LC-MS/MS as described above.