Very Long-chain Acyl-CoA Synthetases

Activation by thioesterification to coenzyme A is a prerequisite for most reactions involving fatty acids. Enzymes catalyzing activation, acyl-CoA synthetases, have been classified by their chain length specificities. The most recently identified family is the very long-chain acyl-CoA synthetases (VLCS). Although several members of this group are capable of activating very long-chain fatty acids (VLCFA), one is a bile acid-CoA synthetase, and others have been characterized as fatty acid transport proteins. It was reported that the Drosophila melanogaster mutantbubblegum (BGM) had elevated VLCFA and that the product of the defective gene had sequence homology to acyl-CoA synthetases. Therefore, we cloned full-length cDNA for a human homolog of BGM, and we investigated the properties of its protein product, hsBG, to determine whether it had VLCS activity. Northern blot analysis showed that hsBG is expressed primarily in brain. Compared with vector-transfected cells, COS-1 cells expressing hsBG had increased acyl-CoA synthetase activity with either long-chain fatty acid (2.4-fold) or VLCFA (2.6-fold) substrates. Despite this increased VLCFA activation, hsBG-expressing cells did not have increased rates of VLCFA degradation. Confocal microscopy showed that hsBG had a cytoplasmic localization in some COS-1 cells expressing the protein, whereas it appeared to associate with plasma membrane in others. Fractionation of these cells revealed that most of the hsBG-dependent acyl-CoA synthetase activity was soluble and not membrane-bound. Immunoaffinity-purified hsBG from transfected COS-1 cells was enzymatically active. hsBG and hsVLCS are only 15% identical, and comparison with sequences of two conserved motifs from all known families of acyl-CoA synthetases revealed that hsBG along with the D. melanogaster and murine homologs comprise a new family of acyl-CoA synthetases. Thus, two protein families are now known that contain enzymes capable of activating VLCFA. Because hsBG is expressed in brain but previously described VLCSs were not highly expressed in this organ, hsBG may play a central role in brain VLCFA metabolism and myelinogenesis.

Free fatty acids must be "activated" to their CoA thioesters before participating in most catabolic and anabolic reactions (1). Processes such as incorporation of fatty acids into phospholipids or triacylglycerols, fatty acid elongation or unsaturation, fatty acylation of proteins, and degradation of fatty acids by ␣-, ␤-, or -oxidation all require activated fatty acid substrates. Activation is catalyzed by acyl-CoA synthetases, also known as acid:CoA ligases (AMP-forming) (EC 6.3.1.x). The acyl-CoA synthetases have been classified by their specificities for fatty acid substrates of different chain lengths. Biochemically, the most well characterized acyl-CoA synthetases are those that activate 16-and 18-carbon long-chain fatty acids (LCFA), 1 compounds that are abundant both endogenously and in the normal human diet. However, synthetases that activate short-(C2-C4), medium-(C6 -C10), and very long-chain fatty acids (VLCFA, ՆC22) have also been described.
Over the last decade, both biochemical and molecular approaches have revealed that more than one enzyme may exist in each chain length specificity class. For example, five rat long-chain acyl-CoA synthetases (LCS) have been purified and/or cDNA-cloned (2)(3)(4)(5)(6). We found that the family of human proteins that contains very long-chain acyl-CoA synthetase (VLCS) includes six members (7). A unifying feature of all acyl-CoA synthetases is the presence of an "AMP-binding domain signature" (PROSITE PS00455). This highly conserved motif consists of 10 -12 amino acids and was first described nearly a decade ago (8). Black et al. (9) proposed that a second highly conserved region present in the LCS family was a "signature motif" for this group of enzymes. We reported that a related motif containing 22 residues is characteristic of the VLCS family (7,10). These observations suggest that other families of acyl-CoA synthetases might be distinguished by the presence of similar domains.
Our understanding of the physiologic role(s) of acyl-CoA synthetases has been facilitated by gene disruption and mutational studies in non-mammalian species (9 -12). The Drosophila melanogaster mutant bubblegum is characterized by neurodegeneration and elevated levels of very long-chain fatty acids (12). These findings are similar to those observed in the human disease X-linked adrenoleukodystrophy, in which there also is reduced peroxisomal VLCS activity (13). The gene defective in the bubblegum mutant encodes a protein whose amino acid sequence had similarities to sequences of acyl-CoA synthetases. However, the Drosophila bubblegum protein (dmBG) lacked both the motif characteristic of the VLCS family and peroxisome targeting signals. Therefore, we cloned cDNA for the human ortholog of bubblegum (hsBGM) and investigated the properties of the human bubblegum protein (hsBG). We report here that hsBG is a VLCS that represents a previously undescribed acyl-CoA synthetase family.

EXPERIMENTAL PROCEDURES
Materials and General Methods-Expressed sequence tag clones were obtained from the American Type Culture Collection (ATCC) and are identified by their ATCC number. COS-1 cells were a gift from Catherine Thompson (Kennedy Krieger Institute). [1-14 C]Palmitic acid was obtained from American Radiolabeled Chemicals. [1-14 C]Lignoceric acid was synthesized from tricosanol and K 14 CN (American Radiolabeled Chemicals) by the method of Muralidharan and Kishimoto (14). DNA sequencing was performed at The Johns Hopkins University Department of Biological Chemistry Biosynthesis and Sequencing Facility using the fluorescent dideoxy terminator method of cycle sequencing on an Applied Biosystems Inc. 377 automated DNA sequencer, following Applied Biosystems protocols. Nucleotide and amino acid sequence alignment, and calculation of sequence identity and similarity, was done using the ClustalW program (15). PCR conditions were as described previously (16). Protein was determined by the method of Lowry et al. (17).
Cloning of hsBGM Full-length cDNA-Homology probing of the Gen-Bank TM nonredundant protein data base identified the 634 amino acid "KIAA0631 protein" (accession number AB014531) as a candidate human homolog of dmBG. Amino acid sequence alignment with that of dmBG demonstrated that the KIAA0631 sequence was incomplete at the amino terminus. We devised a two-step strategy for obtaining full-length hsBG cDNA. First, the KIAA0631 sequence was obtained by PCR amplification; second, 5Ј-RACE was used to obtain the remaining 5Ј-sequence. Two overlapping fragments of hsBG were amplified by PCR from a human brain cDNA library (CLONTECH) based on the GenBank TM KIAA0631 sequence. PCR using oligonucleotide primers P12-1 (5Ј-CTCGAGGCTTCCCTACACTGTGCATC-3Ј, which incorporates an XhoI site) plus P12-2 (5Ј-CGGAATGAAGCCAGACTGAGC-3Ј) for fragment 1 and P12-3 (5Ј-GGTGGAGCCCACATCACACA-3Ј) plus P12-4 (5Ј-TCTAGACCCGGAACGCCTGCCCTCTATT-3Ј, which incorporates an XbaI site) for fragment 2 yielded amplicons of the expected size (891 and 1126 bp, respectively). The region of overlap contained a SmaI restriction site. Fragments 1 and 2 were cloned into a TA cloning vector (pGEM-T Easy, Promega). A ligation mixture containing restriction endonuclease-treated fragment 1 (XhoI/SmaI), fragment 2 (SmaI/ XbaI), and N-myc-pcDNA3 (XhoI/XbaI; this vector is a modified pcDNA3 (Invitrogen) vector that places an in-frame c-Myc epitope at the amino terminus; gift of Stephen Gould, Johns Hopkins University) yielded a construct (P99SJS4-1) encoding the carboxyl-terminal 624 residues of hsBG. To obtain the remaining 5Ј-sequence, Marathon RACE (CLONTECH) was performed according to the manufacturer's instructions using an adult human brain library (CLONTECH) as template. Nested pairs of reverse oligonucleotide primers P12-7 (5Ј-CCAC-TGCCGAGAAGAACCACTCC-3Ј) and P12-6 (5Ј-GAACATCCGATGCA-CAGTGTAGG-3Ј) were used with forward primers AP1 and AP2 (CLONTECH), respectively, in two successive amplification cycles. The resulting 406-bp amplicon was TA-cloned as above. The predicted amino acid sequence of the amplicon overlapped that of KIAA0631 and contained three potential start codons. The 5Ј-RACE product was then used as template for PCR with primers P12-8 (5Ј-CGCGGATCCATGC-CACGCAATTCTGGAGC-3Ј, which incorporates a BamHI site) and P12-7. For assembling a full-length construct of hsBGM cDNA in pcDNA3, the following fragments were ligated: a BamHI/EcoNI digest of the latter PCR product, an EcoNI/XbaI digest of P99SJS4-1, and a BamHI/XbaI digest of pcDNA3. For assembling a full-length construct of hsBGM in N-myc-pcDNA3, a similar ligation was performed substituting the vector fragment of a BamHI/XbaI digest of P99SJS4-1 for the BamHI/XbaI digest of pcDNA3. The latter construct is referred to as N-myc-hsBG in the text.
To place the c-Myc epitope at the carboxyl terminus of hsBG, the full-length hsBGM pcDNA3 construct was used as template for PCR with oligonucleotides P12-5 (5Ј-TGGTGAACCAGGACGCAGAG-3Ј) and P12-12 (5Ј-GCCAAGCTTCATTTTTTGCTCTTGGTA-3Ј, which incorporates a HindIII restriction site). The 653-bp amplicon was digested with AseI and HindIII and ligated with a BamHI/AseI digest of the pcDNA3/hsBG construct and a BamHI/HindIII digest of the pcDNA3.1myc/his-A vector (Invitrogen). This construct is referred to as C-myc-hsBG. All constructs were fully sequenced. The nucleotide se-quence of our hsBGM cDNA clone was submitted to GenBank TM and assigned the accession number AF179481.
To identify potential murine homologs of bubblegum, we screened the GenBank TM expressed sequence tag data base and identified several candidate clones. One I.M.A.G.E. clone (accession number AA111606) was sequenced in its entirety and found to contain a complete open reading frame of 721 amino acids that was 86.5% identical to hsBG.
Northern Blot Analysis-An oligonucleotide probe corresponding to bp 774 -1527 of hsBGM was isolated by digestion of cDNA with EcoNI and SacII. This oligonucleotide and a ␤-actin probe (CLONTECH) were labeled with [ 32 P]dCTP (Amersham Pharmacia Biotech) using the Rediprime DNA labeling system (Amersham Pharmacia Biotech) and hybridized to a human multiple tissue Northern blot (CLONTECH). Hybridization of probes was detected by exposure of either Kodak BioMax film or a BAS-MP 20405 imaging plate and evaluated using a BAS-2500 PhosphorImager and Image Reader Program 1.3E (Fujifilm).
Subcellular Fractionation, Acyl-CoA Synthetase, and Fatty Acid ␤-Oxidation Activity of COS-1 Cells Expressing hsBG-COS-1 cells were transiently transfected with full-length cDNA encoding hsBG, N-myc-hsBG, C-myc-hsBG, or with the pcDNA3 vector alone by electroporation at 250 V as described previously (16). Three days posttransfection, cells were harvested by trypsinization, washed with phosphate-buffered saline, and stored at Ϫ80°C prior to assay. In some experiments, transfected cells were separated into a crude cytoplasmic fraction and a fraction enriched in membranes and subcellular organelles by gently resuspending the washed cells in isotonic sucrose buffer containing digitonin (100 g/ml; Merck). After 10 min on ice, the suspension was centrifuged for 5 min at top speed in a microcentrifuge to separate the cytoplasm (supernatant) from membranes/organelles (pellet). To prepare a true cytosolic fraction, transfected cells were homogenized using a precision ball-bearing homogenizer (18) and a postnuclear supernatant was prepared by centrifugation at 1500 ϫ g for 30 s. A fraction enriched in mitochondria and peroxisomes (ML fraction) was prepared by centrifugation of the postnuclear supernatant for 15 min at 18,000 ϫ g in a Beckman JA-20 high speed centrifuge. The organelle-free supernatant was then centrifuged at 435,000 ϫ g for 15 min in a Beckman Optima TLX ultracentrifuge (TLA 120.2 rotor) to separate microsomes and cytosol. Assay of activation of 1-14 C-labeled fatty acids was performed as described previously (19). Fatty acid ␤-oxidation was measured in freshly isolated COS-1 cells as described previously (20) 3 days post-transfection.
Immunoaffinity Purification of Expressed N-myc-hsBG-The cytosolic fractions from COS-1 cells transiently transfected with N-myc-hsBG or the N-myc-pcDNA3 vector were mixed with anti-Myc antibody covalently bound to agarose beads (Santa Cruz Biologicals) for 2 h at 4°C. Beads were washed extensively with homogenization buffer by centrifugation using Ultrafree MC centrifugal filtration units (Millipore). Washed beads were resuspended in homogenization buffer for assay of acyl-CoA synthetase activity.
Indirect Immunofluorescence-Following transfection with N-myc hsBG, COS-1 cells were plated on glass coverslips. After 3 days, cells were fixed, permeabilized with Triton X-100, and incubated first with mouse anti-Myc (gift of Gerald Johnson, San Diego State University) and then with donkey anti-mouse IgG fluorescein isothiocyanate conjugate (Jackson ImmunoResearch) as described previously (21). For double labeling, cells were transfected with both N-myc hsBG and human adrenoleukodystrophy protein (in pcDNA3). After fluorescein labeling of hsBG as above, coverslips were incubated with anti-adrenoleukodystrophy protein (21) and then with donkey anti-rabbit IgG lissamine rhodamine conjugate (Jackson ImmunoResearch). Fluorescence was visualized using a Leitz Aristoplan fluorescent microscope. Confocal microscopy was performed at The Johns Hopkins University School of Medicine Microscope Facility. Images were acquired with a Noran Oz confocal laser scanning system using Intervision (version 6.5) software on a Silicon Graphics O 2 platform.

RESULTS
Cloning and Characteristics of hsBGM cDNA-Comparison of the amino acid sequence of dmBG to human proteins in the GenBank TM nonredundant data base identified only one protein with significant homology. The "KIAA0631 protein" (an incomplete open reading frame consisting of 634 amino acids) and dmBG (666 amino acids) were 41% identical and 64% similar. Sequence alignment using the ClustalW program revealed that the amino-terminal region of KIAA0631 was incomplete. By using PCR, cDNA encoding the KIAA0631 protein was amplified from a human brain cDNA library, and additional 5Ј-sequence was obtained using a RACE protocol as described under "Experimental Procedures." The complete nucleotide sequence and predicted amino acid sequence of hsBGM cDNA are shown in Fig. 1. Two in-frame stop codons upstream of the predicted initiator methionine were present (bp Ϫ30 and Ϫ75). Although this clone contains only 48 bp of the 3Ј-untranslated region, the nucleotide sequence of KIAA0631 contains extensive 3Ј-untranslated sequence and includes a putative polyadenylation signal (aataaa) at bp 2888 (data not shown). Expressed sequence tags containing portions of the hsBGM cDNA sequence form a UniGene cluster, Hs.277543. This cluster maps to human chromosome 15 between the microsatellite markers D15S114 and D15S115, which corresponds to 15q23-24.
The predicted amino acid sequence of hsBG contains 724 residues. Although this sequence is longer than that of dmBG (647 amino acids), we noted that the hsBG lacked a methionine residue at or near the proposed initiator methionine position of dmBG (12). A molecular mass of 81,290 daltons for hsBG was calculated from the predicted amino acid sequence. The pI of hsBG was calculated to be 5.73. Amino acids 281-290 (YTSGT-TGNPK) comprise the putative AMP-binding domain characteristic of acyl-CoA synthetases and related enzymes (8).
Tissue Expression of hsBG-To determine which tissues normally express hsBG, a human multiple tissue Northern blot was probed with a radiolabeled fragment of hsBGM cDNA as described under "Experimental Procedures." A strong signal corresponding to a 3.0-kb mRNA species was detected in brain ( Fig. 2). Whereas the exact sizes of the 3Ј-and 5Ј-untranslated regions of hsBGM are not known, the 2175-bp open reading frame is compatible with this size transcript. The high level expression in brain is consistent with the observation that mutation of this protein in Drosophila results in neurodegeneration (12). In addition to the 3.0-kb transcript, a larger but less abundant transcript of 4.4 kb was seen in brain. An even Full-length cDNA for hsBG was cloned into the pcDNA3 vector as described under "Experimental Procedures" and was sequenced in its entirety. Two upstream stop codons (indicated by *) were found in the 5Ј-untranslated region at bp Ϫ30 and bp Ϫ75 relative to the putative start codon. An additional stop codon was also found in the 3Ј-untranslated region at bp 2196. Motif 1 (AMP-binding domain; amino acids 281-290) and motif 2 (amino acids 561-585) are shown in bold.
larger transcript of 6.5 kb was observed in heart and skeletal muscle. The significance of these larger transcripts is currently under investigation.
Acyl-CoA Synthetase Activity of COS-1 Cells Expressing hsBG-The presence of an AMP-binding domain and the homology of hsBG and dmBG to other acyl-CoA synthetases suggested that these proteins were also acyl-CoA synthetases. To investigate this, we expressed the full-length cDNA encoding hsBG in COS-1 cells. Three days post-transfection, cell suspensions were assayed for their ability to activate fatty acids to their CoA derivatives. As shown in Table I, cells transiently expressing hsBG activated both palmitic acid (C16:0, a representative LCFA) and lignoceric acid (C24:0, a representative VLCFA) at a greater rate than did vector-transfected cells. Activation of the LCFA was increased 2.4-fold, and activation of the VLCFA 2.6-fold compared with vector-transfected cells. Activities of non-transfected cells and vector-transfected cells were similar (data not shown). No differences in the acyl-CoA synthetase activity of cells expressing hsBG or N-myc-hsBG were observed (data not shown).
Effect of hsBG Overexpression on Fatty Acid ␤-Oxidation-The observation of increased concentrations of VLCFA in the Drosophila mutant bubblegum and the ability of hsBG expressed in COS-1 cells to activate VLCFA suggested that this enzyme may be relevant to cellular VLCFA homeostasis. Excess VLCFA are normally degraded by the peroxisomal ␤-oxidation pathway (22). Therefore, we investigated the possibility that overexpression of hsBG in COS-1 cells might increase the rate of degradation of lignoceric acid, a 24-carbon VLCFA. Despite the increased rate of activation of VLCFA in total cellular extracts of cells overexpressing hsBG (Table I), no increase in the rate of VLCFA ␤-oxidation was found. The potential role(s) of hsBG in other facets of VLCFA metabolism are currently under investigation.
Subcellular Localization of hsBG-Because of the similarities between the phenotype of the fruit fly bubblegum mutant and the human peroxisomal disease X-linked adrenoleukodystrophy, it was hypothesized that hsBG might be a peroxisomal protein. Although no peroxisome targeting signal 1 (PTS1) was present, a potential PTS2 (RIDPSCPQL) was found at amino acid residue 93. When analyzed by SignalP version 1.1 (23) and PSORT (24) prediction programs, no signal sequence or endoplasmic reticulum membrane retention signal was found. No mitochondrial or nuclear targeting signals were detected. The TopPred 2 program (25) identified a single candidate membrane-spanning segment. In contrast, the Predict Protein program (26) indicated that hsBG was a compact globular protein and PSORT predicted a cytoplasmic location.
To investigate the subcellular location of hsBG, we used indirect immunofluorescence and confocal microscopy of cells transiently expressing the protein. Because antibody to hsBG is not yet available, we transfected COS-1 cells with hsBGM cDNA containing the c-Myc epitope in frame at the amino terminus. Fluorescence appeared to be associated with the plasma membrane, but fluorescence of the cytoplasm and intracellular organelles could not be ruled out (data not shown). Therefore, confocal microscopy was used to examine COS-1 cells co-transfected with hsBG and hsALDP (a peroxisomal membrane protein) (Fig. 3). This method did not clearly distinguish between a cytoplasmic and plasma membrane location for hsBG (Fig. 3A). Whereas hsALDP had a punctate immunostaining pattern typical of peroxisomes (Fig. 3B), there was no evidence for localization of hsBG with peroxisomes (Fig. 3C).
Further analysis of cells expressing hsBG was performed by obtaining confocal images of serial sections and using these to reconstruct a three-dimensional image of the cell (Fig. 4). Two populations of COS-1 cells expressing N-myc-hsBG were observed. In some cells, hsBG was predominantly cytoplasmic (Fig. 4, A and B), and in others, the majority of the hsBG immunofluorescence appeared to be close to or associated with the plasma membrane (Fig. 4, C and D). There was no evidence of fluorescent labeling of intracellular organelles. Similar results were obtained when cells were transfected with C-myc-hsBG (data not shown).
The association of some hsBG with the plasma membrane and the possibility that hsBG contains one membrane-spanning region prompted us to conduct additional immunofluorescence studies. COS-1 cells were transfected with either N-myc-hsBG or C-myc-hsBG. When cells were permeabilized with 1% Triton X-100 prior to incubation with anti-Myc antibody (standard procedure), results were identical to those in Fig. 4 (data not shown). When the Triton X-100 step was eliminated (no permeabilization), no immunofluorescent signal was detected in cells expressing either construct (data not shown). If either the carboxyl or amino terminus of hsBG protruded through the plasma membrane, fluorescence would have been detected in the absence of detergent. Thus, we found no evidence that hsBG is an integral plasma membrane protein.
hsBG-dependent Acyl-CoA Synthetase Activity Is Primarily Cytosolic-To investigate the subcellular distribution of hsBG by biochemical methods, digitonin was initially used to fractionate transfected cells into crude cytoplasmic and crude membrane/organelle fractions. Most of the hsBG-dependent  by COS-1 cells expressing hsBG COS-1 cells were transiently transfected with hsBG or the pcDNA3 vector alone as described under "Experimental Procedures." Three days post-transfection, cells were harvested and either assayed immediately for ␤-oxidation activity or frozen at Ϫ80°C for later assay of acyl-CoA synthetase activity. The mean Ϯ S.D. is shown for the number of independent transfection experiments in parentheses. Results with cells expressing hsBG and N-myc-hsBG were indistinguishable (data not shown). acyl-CoA synthetase activity was in the cytoplasmic fraction (Table II). To determine whether this activity was truly soluble, COS-1 cells transiently expressing hsBG were subjected to a more rigorous subfractionation procedure as described under "Experimental Procedures." A peroxisome-and mitochondriaenriched ML fraction, a microsomal fraction, and cytosol were prepared and assayed for acyl-CoA synthetase activity. As expected, nearly all of the acyl-CoA synthetase activity of vectortransfected COS-1 cells was associated with the ML fraction and microsomes (Table II). In contrast, the hsBG-dependent acyl-CoA synthetase activity was mainly in the soluble (cytosolic) fraction (Table II). Small increases in activity of the ML and microsomal fractions in hsBG-expressing cells were also observed. It is not clear whether the plasma membrane-associated hsBG is responsible for these small increases. hsBG Isolated from COS-1 Cells Has Enzyme Activity-Although amino acid sequence homology suggests that hsBG is an acyl-CoA synthetase, it is possible that the increased enzyme activity of COS-1 cells expressing hsBG is a secondary phenomenon. To rule out this possibility, we purified the expressed N-myc-hsBG from the cytosolic fraction using anti-Myc antibody covalently linked to agarose beads. No acyl-CoA synthetase activity was bound to beads incubated with cytosol from cells transfected with the vector (N-myc-pcDNA3) alone (Table III). In contrast, 26 -39% of the cytosolic activity in hBG-transfected cells was found associated with the anti-Myc beads (Table III). These data confirm that the hBG protein itself has acyl-CoA synthetase activity.
hsBG Represents a New Family of Acyl-CoA Synthetases-In our investigations of the VLCS protein family, we noted two highly conserved motifs present in their amino acid sequences. One motif is the previously mentioned AMP-binding domain. The second motif is 240 -250 amino acids downstream and contains 22 residues, 15 of which are completely conserved in all family members from yeast to man (7, 10). This motif overlaps a highly conserved region in the long-chain acyl-CoA synthetase family described by Black et al. (9) as the signature motif for this family. Inspection of the comparable regions in both dmBG and hsBG amino acid sequences revealed that these two proteins lack this motif and thus were not members of either the VLCS family or the LCS family. To determine whether hsBG and dmBG were similar to any other known acyl-CoA synthetases, we screened the GenBank TM nonredundant protein data base. Amino acid sequences of all identifiable vertebrate and yeast acetyl-CoA synthetases, medium-and long-chain acyl-CoA synthetases, and unknown or putative acyl-CoA synthetases were aligned and regions corresponding to VLCS motifs 1 and 2 were identified. In some cases Caenorhabditis elegans, D. melanogaster, Arabidopsis thaliana, and/or bacterial sequences were included. To facilitate assignment of sequences to protein families, we extended motif 1 by 5 amino acids in each direction and motif 2 by 10 amino acids. Alignments of these sequences are shown in Table IV. Visual inspection of motif 2 in most cases confirmed that proteins previously designated as acetyl-CoA synthetases, LCSs and VLCSs, were indeed closely related to each other but represented distinct protein families. Only one sequence for medium-chain acyl-CoA synthetase was found (Bos taurus xenobiotic/medium-chain fatty acid:CoA ligase form XL-III). However, human, rat, and mouse SA protein (a protein associated with essential hypertension and a putative acyl-CoA synthetase (27,28)) clearly belonged to the same family. In addition, proteins identified as "human unknown protein" and "rat kidney-specific protein" belonged to the medium-chain acyl-CoA synthetase family. The protein encoded by the yeast FAT2 gene is also a putative acyl-CoA synthetase, but inspection of both motifs 1 and 2 revealed that it did not clearly belong to any of the families described above. A search for additional members of this family identified yeast, mycobacterial, and plant relatives but no vertebrate sequences.
We next compared the sequences of hsBG, mmBG and dmBG to those of the five protein families described above. As shown in Table IV, the sequences of these three proteins showed considerable homology to each other. However, they did not appear to be members of any of the other families. Thus, the proteins encoded by "bubblegum" genes appeared to represent a distinct, previously unknown family of acyl-CoA synthetases. Furthermore, this family may represent a new family of VLCSs, enzymes capable of activating VLCFAs. DISCUSSION Our previous investigations of enzymes that activate VLCFA to their CoA derivatives suggested that although hsVLCS, hs-VLCS-H2, and mmFATP1 have VLCS activity, additional enzymes with VLCS activity must exist (16, 29 -31). The elevation of VLCFA levels in the bubblegum mutant and the similarity of dmBG to other acyl-CoA synthetases (12) suggested that BG might have VLCS activity. As we report here, hsBG is indeed capable of activating VLCFA and, like hsVLCS, will also activate long-chain fatty acid substrates.
At present, the physiologic role of hsBG in fatty acid metabolism is not known. Because patients with XALD and fruit flies with the bubblegum mutation have similarly elevated VLCFA levels, it was reasonable to investigate the hypothesis that hsBG is a peroxisomal protein that activates VLCFA for degradation by peroxisomal ␤-oxidation. However, neither aspect of this hypothesis was validated in COS-1 cells expressing hsBG. In these cells, hsBG could not be detected in peroxisomes, and a role for this protein in VLCFA ␤-oxidation was not demonstrated. However, if concentrations of other enzymes or proteins required for normal peroxisomal VLCFA ␤-oxidation are rate-limiting, increasing the concentration of hsBG would likely be without effect. Nonetheless, studies to determine whether hsBG is involved in XALD biochemical pathology are currently in progress.
We also considered the possibility that hsBG might play a role in complex lipid synthesis. Although we did not observe increased incorporation of either C16:0 or C24:0 into neutral or polar lipid classes in COS-1 cells expressing hsBG (data not shown), these experiments are subject to the same concerns as the ␤-oxidation experiments, i.e. VLCFA activation may not be rate-limiting. Thus, COS-1 cells overexpressing hsBG may not be an appropriate system for demonstrating a role for this enzyme in complex lipid synthesis. Studies to evaluate the effect on VLCFA metabolism of decreasing the level of hsBG in cells endogenously expressing the protein are currently in progress.
The observation that hsBG expressed in COS-1 cells is primarily soluble was surprising because most endogenous acyl-CoA synthetase activity has been reported to be in either organellar membranes (microsomal, peroxisomal, or mitochondrial) or the plasma membrane (1). Confocal microscopy showed hsBG localizing near the plasma membrane in some transfected COS-1 cells, whereas the protein was clearly cytoplasmic in others. These observations suggest a potential role of hsBG in "trapping" fatty acids inside cells as their CoA derivatives as they enter the cell from the extracellular medium. Furthermore, this raises the question of whether alterations in the physiologic status of cells or the composition of the culture medium could influence the distribution of hsBG within the cell.
Prior to this report, the only enzymes capable of activating VLCFA were members of the VLCS/FATP protein family (16, 29 -31). It is noteworthy that hsBG belongs to a separate protein family. The bubblegum family, established by homology within motifs 1 and 2 of Table IV, includes hsBG (this report), mmBG, 2 dmBG (12), and a homolog of dmBG that derived from the Drosophila genome sequencing project (GenBank TM accession number AAF44850). In addition to these proteins, we have postulated the existence of homologs of hsBG and mmBG based on sequences of expressed sequence tags. 2 These additional proteins are most likely acyl-CoA synthetases and potentially have VLCS activity.
In our analysis of the amino acid sequences of the known 2 A. K. Heinzer, K. D. Smith, and P. A. Watkins, unpublished data. Cytosolic fractions of COS-1 cells expressing N-myc-hsBG or the N-myc-pcDNA3 vector were incubated with agarose beads containing covalently linked anti-Myc antibody as described under "Experimental Procedures." Equal amounts of cytosol and bead suspensions containing bound myc-tagged proteins were assayed for acyl-CoA synthetase activity using C16:0 and C24:0 substrates. Results of two experiments are shown, each using cytosol from an independent transfection. To calculate the percent activity bound to beads, endogenous acyl-CoA synthetase activity (vector-transfected cytosol, A) was subtracted from that of the hsBG-transfected cytosol (B).  IV Amino acid sequence homologies define distinct families of acyl-CoA synthetases Amino acid sequences of known or putative acyl-CoA synthetases were obtained from GenBank™. Sequence alignments were performed using the ClustalW program. Regions corresponding to Motif 1 and Motif 2 of the VLCS/FATP family (7,10) were identified. These sequences were extended in both the 5Ј-and 3Ј-directions by 5 residues (Motif 1) or 10 residues (Motif 2) to improve the accuracy of family assignments. Additional family members were identified by using a representative Motif 2 sequence (from a human protein if available) as query sequence in a BLAST search of the GenBank™ non-redundant protein database. All vertebrate sequences identified are shown below. Representative sequences from yeast and other species are also shown in many cases, but these listings are not exhaustive.

DAMIMFTGGTTGLPKMVPWT
LGSLSAAGDLSIRGRIKELINR-GGEKISPERVEGVLASHPNVME H70982 acyl-CoA synthetases, we also established that Fat2p, a putative acyl-CoA synthetase from Saccharomyces cerevisiae, belonged to yet another protein family (Table IV). Enzyme activity of Fat2p has not been firmly established, and if indeed it has activity it will be of interest to determine its substrate specificity. Although no mammalian members of this family were readily detected in public domain data bases, if such homologs exist they will soon be identified by genome sequencing efforts. It is also likely that additional families of acyl-CoA synthetases may be identified as a result of genome sequencing and that additional candidate VLCSs will emerge. It is significant that hsBG is expressed primarily in brain, since no member of the hsVLCS/FATP family is highly expressed in this tissue. In general, brain has higher concentrations of VLCFA than other tissues (32). VLCFA present in brain include saturated (e.g. C24:0 and C26:0), monoenoic (e.g. C24:1 and C26:1), and polyunsaturated fatty acids (e.g. C24: 46 and C22:63) (32). Concentrations of brain VLCFA increase during development (33). These VLCFA are components of complex lipids such as gangliosides, cerebrosides, sulfatides, sphingomyelin, and other phospholipids. Activation by VLCSs is required for incorporation of VLCFA into these complex lipids. Although their role(s) in brain function are not completely understood, many of these VLCFA-containing lipids are components of myelin membranes and disruption of VLCS activity could be responsible for one or more genetic demyelinating disorders. Thus, investigation of brain-specific VLCSs will significantly enhance our understanding of brain fatty acid homeostasis due to the central importance of VLCFA activation in this process.