Biosynthesis of the unsaturated 14-carbon fatty acids found on the N termini of photoreceptor-specific proteins.

In the vertebrate retina, a number of proteins involved in signal transduction are known to be N-terminal acylated with the unusual 14 carbon fatty acids 14:1n-9 and 14:2n-6. We have explored possible pathways for producing these fatty acids in the frog retina by incubation in vitro with candidate precursor fatty acids bearing radiolabels, including [3H]14:0, [3H]18:1n-9, [3H]18:2n-6, and [3H]18:3n-3. Rod outer segments were prepared from the radiolabeled retinas for analysis of protein-linked fatty acids, and total lipids were extracted from the remaining retinal pellet. Following saponification of extracted lipids, fatty acid phenacyl esters were prepared and analyzed by high pressure liquid chromatography (HPLC) with detection by continuous scintillation counting. Transducin, whose alpha-subunit (Gt alpha) is known to bear N-terminal acyl chains, was extracted from the rod outer segments and subjected to SDS-polyacrylamide gel electrophoresis and fluorography to detect radiolabeled proteins. Gt alpha was also subjected to methanolysis, and the resulting fatty acyl methyl esters were analyzed by HPLC. The identities of HPLC peaks coinciding with unsaturated species of both phenacyl esters and methyl esters were confirmed by reanalyzing them after catalytic hydrogenation. The results showed that 14:1n-9 can be derived in the retina from 18:1n-9 and 14:2n-6 from 18:2n-6, most likely by two rounds of beta-oxidation, but that neither is produced in detectable amounts from 14:0. Retroconversion of unsaturated 18 carbon fatty acids to the corresponding 14 carbon species showed specificity, in that 18:3n-3 was not converted to 14 carbon fatty acids in detectable amounts. Myristic acid (14:0), 14:1n-9, and 14:2n-6 were all incorporated into Gt alpha. A much less efficient incorporation of 18:1n-9 into Gt alpha was also observed, but no radiolabeling of Gt alpha was observed in retinas incubated with 18:3n-3. Thus, retroconversion by limited beta-oxidation of longer chain unsaturated fatty acids appears to be the most likely metabolic source of the unusual fatty acids found on the N termini of signal transducing proteins in the retina.

In the vertebrate retina, a number of proteins involved in signal transduction are known to be N-terminal acylated with the unusual 14 carbon fatty acids 14: 1n-9 and 14:2n-6. We have explored possible pathways for producing these fatty acids in the frog retina by incubation in vitro with candidate precursor fatty acids bearing radiolabels, including [ 3 H]14:0, [ 3 H]18:1n-9, [ 3 H]18:2n-6, and [ 3 H]18:3n-3. Rod outer segments were prepared from the radiolabeled retinas for analysis of protein-linked fatty acids, and total lipids were extracted from the remaining retinal pellet. Following saponification of extracted lipids, fatty acid phenacyl esters were prepared and analyzed by high pressure liquid chromatography (HPLC) with detection by continuous scintillation counting. Transducin, whose ␣-subunit (G t␣ ) is known to bear N-terminal acyl chains, was extracted from the rod outer segments and subjected to SDS-polyacrylamide gel electrophoresis and fluorography to detect radiolabeled proteins. G t␣ was also subjected to methanolysis, and the resulting fatty acyl methyl esters were analyzed by HPLC. The identities of HPLC peaks coinciding with unsaturated species of both phenacyl esters and methyl esters were confirmed by reanalyzing them after catalytic hydrogenation. The results showed that 14:1n-9 can be derived in the retina from 18:1n-9 and 14:2n-6 from 18:2n-6, most likely by two rounds of ␤-oxidation, but that neither is produced in detectable amounts from 14:0. Retroconversion of unsaturated 18 carbon fatty acids to the corresponding 14 carbon species showed specificity, in that 18:3n-3 was not converted to 14 carbon fatty acids in detectable amounts. Myristic acid (14:0), 14:1n-9, and 14:2n-6 were all incorporated into G t␣ . A much less efficient incorporation of 18:1n-9 into G t␣ was also observed, but no radiolabeling of G t␣ was observed in ret-inas incubated with 18:3n-3. Thus, retroconversion by limited ␤-oxidation of longer chain unsaturated fatty acids appears to be the most likely metabolic source of the unusual fatty acids found on the N termini of signal transducing proteins in the retina.
A number of proteins of both eukaryotic and viral origin are modified by fatty acylation through an amide linkage to Nterminal glycine residues (reviewed by James and Olson (1990), McIlhinney (1990), Towler et al. (1988), Schlesinger (1993), and Gordon et al. (1991)). This modification, which in the great majority of cases studied involves the saturated 14 carbon fatty acid myristate (14:0), has been shown to play an important role in the function of these proteins (Jones et al., 1990;Linder et al., 1991;Yonemoto et al., 1993;Wedegaertner et al., 1995). In vertebrate retinas, the types of N-terminal fatty acids for proteins involved in signal transduction are strikingly different from those found in other tissues. These proteins, which include the ␣-subunit of the G protein transducin (G t␣ ) 1 Kokame et al., 1992;Neubert et al., 1992;Yang and Wensel, 1992), guanylyl cyclase-activating protein (GCAP) , recoverin (Dizhoor et al., 1992;Johnson et al., 1994), and the catalytic subunit of cAMPdependent protein kinase , are heterogeneously acylated with frequent occurrence of 14:1n-9 and 14: 2n-6 in addition to 14:0 and 12:0. This unusual pattern appears to reflect unusual pathways for synthesizing and utilizing fatty acids for N-terminal acylation in the retina rather than specific characteristics of the acylated proteins. The ubiquitous protein kinase is exclusively modified with 14:0 in the brain and heart (Carr et al., 1982;Johnson et al., 1994) but in the retina shows the same heterogeneous pattern of fatty acylation as G t␣ . Indeed, although 12:0 and 14:0 are found in animal tissues, 14:1n-9 and 14:2n-6 are quite rare and have been found in abundance only in marine mammals (14:1n-9) (Markley, 1960) and the Asian plant Evodia rutaecarpa (14:2n-6) (Kurono et al., 1972).
It is conceivable that the two unsaturated fatty acids could arise from the desaturation of 14:0. However, previously described ⌬ 5 -desaturase enzymes prefer long chain polyunsaturated fatty acids as substrates (Numa et al., 1984), and ⌬ 8desaturases have only been found in a few cell types, such as testicular Sertoli (Oulhaj et al., 1992) and brain glioma cells (Cook et al., 1991). Alternatively, these unusual 14-carbon fatty acids might arise by retroconversion from longer chain fatty * This work was supported by NEI National Institutes of Health Grants EY07001, EY00871, EY04149, and EY02520 (to R. E. A.) and EY07981 (to T. G. W.), by grants from the Retina Research Foundation, by Research to Prevent Blindness Inc., by the National Retinitis Pigmentosa Foundation Inc., by the Samuel Roberts Nobel Foundation, Inc., and by the Presbyterian Health Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Preparation of Frog Retinas-All dissection procedures described here were performed under dim red lights. Dark-adapted frogs (16 h) were killed by decapitation followed by immediate pithing of the spinal column. Eyes were enucleated and placed in room temperature Krebs-Ringer buffer (118 mM NaCl, 4.7 mM KCl, 1.17 mM KH 2 PO 4 , 1.17 mM MgSO 4 , 5.6 mM D-glucose, 35 mM NaHCO 3 , and 1.0 mM EDTA, pH 7.4) and then converted to eyecups through removal of the cornea, lens, and vitreous humor. Retinas were detached by further incubation (15 min) of eyecups in room temperature Krebs-Ringer buffer. Retinas were removed and cleaned of adhering retinal pigmented epithelium using jewelers forceps and then stored in ice-cold Krebs-Ringer buffer.
Lipid Extraction and Separation of Lipid Species-Membranes were homogenized in 2 ml of Tris acetate buffer and extracted twice with 2 ml of CHCl 3 /MeOH/12 N HCl (100:100:1, v/v/v) and once with 1.5 ml of CHCl 3 . Following centrifugation, the CHCl 3 layers were pooled, total lipid phosphorus was estimated by the method of Rouser et al. (1970), and total radioactivity was determined by scintillation counting. An equal portion of each extract was separated into phospholipids (phosphatidylcholine, phosphatidylserine, phosphatidylinositol, and phosphatidylethanolamine classes), free fatty acids, and triglycerides by TLC (Kupke and Zeugner, 1978) on a silica gel 60 TLC plate (EM Science), with development up one-third of the plate with CHCl 3 / MeOH/H 2 O (65:30:5, v/v/v) and complete development with hexane/ Et 2 O/HOAc (80:20:1.5, v/v/v). For analysis of total radioactivity in each class, positions were located by I 2 vapor and marked, and after I 2 sublimation, spots were scraped and counted in BCS mixture (Amersham Corp.) 1N HCl (15:1, v/v). For fatty acid analysis, spots corresponding to the lipid classes were located by spraying with 2,7-dichlorofluorescein and illuminating with UV and then scraped and analyzed as described below.
Fatty Acid Phenacyl Ester Preparation and HPLC Analysis-Total lipid extracts or isolated lipid classes were saponified for 45 min at 100°C in 2% KOH/EtOH (w/v), diluted with H 2 O, acidified with 12 N HCl, and extracted three times with hexane. Free fatty acid extracts were used to prepare FAPEs according to the method of Wood and Lee (1983) as modified by Chen and Anderson (1993a). FAPEs were separated on HPLC using a Supelco (Bellefonte, PA) Supelcosil LC-18 column (25 cm ϫ 4.6 mm I.D.) with elution (2 ml/min) by a linear gradient of CH 3 CN/H 2 O starting at 80:20 (v/v), increasing to 92:8 in 45 min, holding at 92:8 for 10 min, and returning to 80:20 in 5 min. Elution of FAPEs was monitored by UV absorbance at 242 nm. Radioactivity profiles were obtained with an on-line continuous scintillation counter (Flo-One: Radioactive Flow Detector A-200, Radiomatic, Tampa, FL) using Flo-scint A mixture (Packard Instrument Co., Inc., Meriden, CT) at a 2.5:1 (v/v) ratio with the column eluant. Preliminary identification of FAPE peaks was based upon retention times obtained for FAPEs prepared from radiolabeled and nonradiolabeled fatty acid standards (Sigma). Fatty acid standards of 14:1n-9, 14:2n-6, and [1-14 C]14:2n-6 were generously provided by Dr. Howard Sprecher (Department of Medical Biochemistry, Ohio State University).
Analysis of FAPEs by Catalytic Hydrogenation-Individual FAPE peaks were collected and extracted three times with hexane, followed by saponification with 2% KOH/EtOH (w/v), as described above. Free fatty acids were solubilized in EtOH/hexane (2:1) and bubbled vigorously with hydrogen (20 min) in the presence of ϳ10 mg of Pt 2 O (Matheson Coleman & Bell Manufacturing Chemists, Norwood, OH) catalyst. Spent Pt 2 O was removed by centrifugation. Hydrogenated fatty acids in the supernatant were phenacylated as described above, and FAPEs were chromatographed on HPLC with the modification of a 30-min elution with 100% acetonitrile immediately following the 98:2 acetonitrile/water, to elute saturated fatty acids larger than 18 carbon atoms.
Isolation of Transducin-Transducin was isolated using a modification of the procedures of Ohguro et al. (1990) and Umbarger et al. (1992). ROS were homogenized at 4°C in isotonic wash buffer (100 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 1 mM EGTA, 2 mM dithiothreitol, and 1 mM phenylmethylsulfonyl fluoride) and centrifuged at 100,000 ϫ g for 30 min. The isotonic buffer wash was repeated once, and the ROS pellet was homogenized at 4°C in hypotonic wash buffer (5 mM Tris-HCl, pH 7.5, 0.5 mM MgCl 2 , 2 mM dithiothreitol, and 1 mM phenylmethylsulfonyl fluoride) and centrifuged at 100,000 ϫ g for 30 min. The ROS pellet was homogenized at 4°C in hypotonic wash buffer containing 200 M GTP␥S and centrifuged at 100,000 ϫ g for 30 min. The hypotonic buffer wash with 200 M GTP␥S was repeated once, and the transducin subunits (G t␣␤␥ ) were recovered in the GTP␥S wash supernatants.
To test for resistance of radiolabel of proteins to hydrolysis by hydroxylamine (NH 2 OH) as evidence of amide linkage, SDS-PAGE was repeated and gels were treated with NH 2 OH using a modification of the methods of Buss et al. (1987) and Olson et al. (1985). The gels were fixed (30 min) in isopropanol/H 2 O/HOAc (25:65:10, v/v/v), washed (45 min) with H 2 O, soaked (16 h) in 1 M NH 2 OH ⅐HCl (pH 6.7), and washed (6 h) with isopropanol/H 2 O/HOAc (10:80:10, v/v/v). Gels were stained, destained, treated with EN 3 HANCE, dried, and autoradiographed. To determine the efficiency of thioester-linked fatty acid removal by NH 2 OH, we performed the procedure on rhodopsin, a protein known to carry two thioester linked 16:0 groups (Papac et al., 1992). Using previously described procedures, frog retinas were incubated in vitro with 150 Ci of [9,10(N)-3 H]16:0 (American Radiolabeled Chemicals Inc., St. Louis, MO), and ROS were prepared, stripped of transducin, and subjected to SDS-PAGE using low reducing conditions (1 h of incubation in sample application buffer containing 2 mM dithiothreitol and 15 M 2-mercaptoethanol). Gel slices (n ϭ 4) containing rhodopsin (ϳ50 g, estimated by BCA* protein assay kit, Pierce) were saponified in 1 M NaOH at 37°C for 2 h and acidified with 12 N HCl, and radioactivity was determined by counting gel slice and hydrolysate together (BCS mixture/hydrolysate, 15:1, v/v). Identical gels were sub- B, Coomassie Blue-stained electrophoretic gel (a) and corresponding fluorogram (b) for transducin (G t␣␤␥ ) isolated from frog retinas labeled with 1 mCi of [ 3 H]14:0. Fluorography was performed for 28 days. G t␣ and G t␤ are the ␣and ␤-subunits of transducin, respectively. The G t␥ (8 kDa) was run off the gel. C, HPLC elution profiles of FAME radioactivity for G t␣ and G t␤ (control) isolated from frog retinas labeled with 1 mCi of [ 3 H]14:0. FAMEs were released from the SDS-PAGE-purified G t␣ and G t␤ by acidic methanolysis. jected to treatment with 1 M NH 2 OH as described above; rhodopsin was excised, saponified, and counted. NH 2 OH treatment removed 84 Ϯ 2% of the radiolabel from this protein.
The Coomassie staining pattern (Fig. 2B, a) and fluorogram (Fig. 2B, b) revealed faint radiolabeling concentrated at a migration position in SDS-PAGE aligning precisely with G t␣ (39 kDa), with some diffuse radiolabeling extending into the region for G t␤ (36 kDa). Another area of diffuse radiolabeling was observed to be focused at a migration position aligning with a faint doublet protein band above 84 kDa. HPLC of methanolysis products (Fig. 2C) showed that the radiolabel on G t␣ was a mixture of 14:1n-9 (108 Ϯ 16 counts/6 s/10 g) and 18:1n-9 (51 Ϯ 6 counts/6 s/10 g), whereas the radiolabel seen in the region for G t␤ was only 18:1n-9 (35 Ϯ 9 counts/6 s/10 g). HPLC of the G t␣ methanolysis products after simultaneous collection and catalytic hydrogenation (Fig. 2C)
The Coomassie staining pattern (Fig. 3B, a) and fluorogram (Fig. 3B, b) revealed moderate radiolabeling at a migration position in SDS-PAGE aligning precisely with G t␣ (39 kDa), with faint radiolabeling also seen aligning with G t␤ (36 kDa). No other areas of radiolabeling were observed in the fluorogram. HPLC of methanolysis products (Fig. 3C) revealed that the radiolabel on G t␣ was only 14:2n-6 (101 Ϯ 8 counts/6 s/10 g), and no detectable radiolabel was associated with G t␤ . Catalytic hydrogenation and HPLC of the methanolysis product for G t␣ confirmed its identity by showing an appropriate shift in retention time to that of 14:0.
Incubation with [9,10,12,13,15,  H]18:3n-3, respectively). HPLC of the 18:4n-3 desaturation product after catalytic hydrogenation (Fig. 4A) confirmed its identity by showing an appropriate shift in retention time to that of 18:0. Likewise, identities of the other desaturation and elongation products were confirmed by catalytic hydrogenation (results not shown). Hydrogenation of the prominent peak at ϳ18 min produced a product with a retention time between that of 18:0 and 20:0 (result not shown), making its identification inconclusive.
The Coomassie staining pattern (Fig. 4B, a) and fluorogram (Fig. 4B, b) showed no radiolabeling corresponding with G t␣ (39 kDa), G t␤ (36 kDa), or any other protein observed on SDS-PAGE. HPLC of methanolysis products (Fig. 4C) revealed that no detectable radiolabel was associated with either G t␣ or G t␤ , consistent with the fluorogram results.
Hydroxylamine Treatment of Radiolabeled Transducin-Shown in Fig. 5 2. Incubation with [9,10-3 H]18:1n-9. A, HPLC elution profile of FAPE radioactivity for total lipids from frog retinas incubated with 1 mCi of [ 3 H]18:1n-9. The y axis ([ 3 H] counts/6-s interval) is shown at a 10ϫ reduction of the original scale; the maximum for the [ 3 H]18:1n-9 peak was at 21,322. Also shown are the HPLC elution profiles of FAPE radioactivity for the 14:1n-9 and 16:1n-9 after being collected separately and subjected to catalytic hydrogenation. B, Coomassie Blue-stained electrophoretic gel (a) and corresponding fluorogram (b) for transducin (G t␣␤␥ ) isolated from frog retinas labeled with 1 mCi of [ 3 H]18:1n-9. Fluorography was performed for 41 days. G t␣ and G t␤ are the ␣and ␤-subunits of transducin, respectively. G t␥ (8 kDa) was run off the gel. C, HPLC elution profiles of FAME radioactivity for G t␣ and G t␤ (control) isolated from frog retinas labeled with 1 mCi of [ 3 H]18:1n-9. FAMEs were released from the SDS-PAGE-purified G t␣ and G t␤ by acidic methanolysis. Also shown is the HPLC elution profile of FAME radioactivity for the 14:1n-9 and 18:1n-9 from the methanolysis of G t␣ , after being collected simultaneously and subjected to catalytic hydrogenation. FIG. 3. Incubation with [9,10,12,13-3 H]18:2n-6. A, HPLC elution profile of FAPE radioactivity for total lipids from frog retinas incubated with 120 Ci of [ 3 H]18:2n-6. The y axis ([ 3 H] counts/6-s interval) of the radioactivity profile is shown at a 10ϫ reduction of the original scale; the maximum for the [ 3 H]18:2n-6 peak was at 32,946. Also shown are the HPLC elution profiles of FAPE radioactivity for the 14:2n-6 and 16:2n-6 after being collected separately and subjected to catalytic hydrogenation. B, Coomassie Blue-stained electrophoretic gel (a) and corresponding fluorogram (b) for transducin (G t␣␤␥ ) isolated from frog retinas labeled with 120 Ci of [ 3 H]18:2n-6. Fluorography was performed for 42 days. G t␣ and G t␤ are the ␣and ␤-subunits of transducin, respectively. G t␥ (8 kDa) was run off the gel. C, HPLC elution profiles of FAME radioactivity for G t␣ and G t␤ (control) isolated from frog retina labeled with 120 Ci of [ 3 H]18:2n-6. FAMEs were released from SDS-PAGE-purified G t␣ and G t␤ by acidic methanolysis. Also shown is the HPLC elution profile of FAME radioactivity for the 14:2n-6 from the methanolysis of G t␣ , after being subjected to catalytic hydrogenation.  4. Incubation with [9,10,12,13,15,16-3 H]18:3n-3. A, HPLC elution profile of FAPE radioactivity for total lipids from frog retinas incubated with 120 Ci of [ 3 H]18:3n-3. The y axis ([ 3 H] counts/6-s interval) of the radioactivity profile is shown at a 4ϫ reduction of the original scale; the maximum for the [ 3 H]18:3n-3 peak was at 7086. Also shown are the HPLC elution profiles of FAPE radioactivity for the 18:4n-3 after being subjected to catalytic hydrogenation. B, Coomassie Blue-stained electrophoretic gel (a) and corresponding fluorogram (b) for transducin (G t␣␤␥ ) isolated from frog retinas labeled with 120 Ci of [ 3 H]18:3n-3. Fluorography was performed for 35 days. G t␣ and G t␤ are the ␣and ␤-subunits of transducin, respectively. G t␥ (8 kDa) was run off the gel. C, HPLC elution profiles of FAME radioactivity for G t␣ and G t␤ (control) isolated from frog retina labeled with 120 Ci of [ 3 H]18:3n-3. FAMEs were released from SDS-PAGE-purified G t␣ and G t␤ by acidic methanolysis. Fig. 6 shows the distribution of isotope in the retinal phospholipid, free fatty acid, and triglyceride pools from the [ 3 H]-14:0, [ 3 H]-18:1n-9, [ 3 H]-18:2n-6, and [ 3 H]-18:3n-3 incubations. The figure also shows the composition of radioactivity contained within the individual lipid classes. In all incubations the majority of the radiolabel was sequestered in the phospholipid pool after 8 h. We did not examine the distribution of radioactivity among the various phospholipid classes. For all incubations, fatty acid analysis of each lipid class revealed all the previously described desaturation, elongation, or retroconversion products, with percentages similar to that found when the total lipid pools were analyzed. There was not a significant enrichment of any metabolic product into a particular lipid class. DISCUSSION N-terminal fatty acylation of a protein is carried out cotranslationally by the enzyme myristoylCoA:protein N-myristoyltransferase (Towler et al., 1988 andSchlesinger, 1993). Human and yeast myristoyltransferases have a high substrate specificity for 14:0-CoA (Kishore et al., 1991(Kishore et al., , 1993Lu et al., 1994). Although 14:2(n-6)-CoA has not been tested, 14:1(n-9)-CoA is utilized by both myristoyltransferases at a ϳ3-fold lower catalytic efficiency (V max /K m ) than 14:0-CoA (Kishore et al., 1993). Despite extensive study, no evidence has been found for myristoyltransferase isozymes with drastically altered substrate specificity, and supporting this mRNA from the single copy human gene appears to be identical in all tissues studied (Duronio et al., 1992). Consequently, to successfully compete for myristoyltransferase, levels of available 14:1n-9 and 14:2n-6 CoAs must be higher than that of 14:0. To attain these levels, the retina would need to possess special pathways for generating 14:1n-9 and 14:2n-6, which are apparently absent or less active in other tissues. However, this does not mean 14:1n-9 and 14:2n-6 would necessarily accumulate in the retina, because their CoAs might be taken up rapidly by the myristoyltransferase during active protein synthesis. This hypothesis is consistent with the lack of detectable (Ͻ0.1%) 14:1n-9 and 14:2n-6 in the total lipid pool of the frog and bovine retina (Chen and Anderson, 1993a; Bartley et al., 1962).
N-terminal acylated retina proteins do not contain 14:3n-3 , even though n-3 polyunsaturated fatty acids are abundant in the vertebrate retina (Fliesler and Anderson, 1983). We found that metabolism of [ 3 H]18:3n-3 did not produce detectable [ 3 H]16:3n-3 or [ 3 H]14:3n-3 retroconversion products and radiolabel was not incorporated into G t␣ . This result shows that the retroconversion process in the frog retina has a selectivity dependent on double bond position, as well as on chain length. The 18:3n-3 did undergo extensive elongation and desaturation, following the steps toward 22: 6n-3, as described for liver (Sprecher, 1972), consistent with previous observations in the frog retina (Wang and Anderson, 1993).
Under the in vitro incubation conditions we used, there was an active uptake and metabolism for all the radiolabeled fatty acids, as indicated by the amount of the original starting radioactivity that was incorporated into the retina glycerolipids ([ 3 H]14:0, 12%; [ 3 H]18:1n-9, 14%; [ 3 H]18:2n-6, 5%; and [ 3 H]18: 3n-3, 7%). In all incubations, the phospholipids contained most of the radiolabel, whereas only a small percentage was incorporated into triglycerides, consistent with previous studies (Wang and Anderson, 1993;Chen and Anderson, 1993b). Incorporation into glycerolipids indicates effective conversion to precursor CoA ester derivatives. Acyl-CoAs are also the substrates for myristoyltransferase, and labeling of G t␣ with [ 3 H]14:0 suggested efficient conversion to the CoA ester.
retina can be modified with these fatty acids, possibly at levels too low to be detected by mass spectrometry.
Amide linkage of the [ 3 H]14:0, [ 3 H]14:1n-9, and [ 3 H]14:2n-6 to G t␣ was supported by finding that the radiolabel was resistant to hydroxylamine, under conditions where 86% removal of rhodopsin's thioester-linked fatty acid was achieved. Taken together with the mass spectrometric results of Johnson et al. (1994) and the well established specificity of myristoyltransferase in yeast and mammals, this result strongly suggests that the most likely site for attachment of these fatty acids is the ␣-amino group of the N-terminal glycine of frog G t␣ .
Because frog G t␣ is reported to be modified exclusively by 14:2n-6, we investigated whether radiolabeled 14:2n-6 would directly incorporate into the protein. We performed in vitro incubations (n ϭ 2) of frog retinas with 20 Ci of [1-14 C]14:2n-6 (data not shown). Analysis of the retina total lipids showed significant chain elongation of the [ 14 C]14:2n-6 to [ 14 C]16:2n-6 and [ 14 C]18:2n-6 (3 and 4% of [ 14 C]14:2n-6, respectively). Metabolism of nonradiolabeled 14:2n-6 to 16:2n-6 and 18:2n-6 has been previously noted in rat liver (Sprecher, 1967). Methanolysis of SDS-PAGE-purified G t␣ (19 Ϯ 2 g) and G t␤ (13 Ϯ 1 g) from the [ 14 C]14:2n-6 labeled retinas failed to release any detectable radiolabeled fatty acids. Because the [ 14 C]14:2n-6 used in our incubations was of very low specific activity (55-60 Ci/mol), it is possible that dilution with endogenously produced 14:2n-6 precluded our detecting radioactivity in protein product. However, it is also possible that 14:2n-6 supplied directly may not be readily available as a substrate for myristoyltransferase to incorporate into G t␣ . It is clearly converted into the necessary chemical form, CoA ester, as evidenced by chain elongation products and its incorporation into glycerolipids. However, subcellular compartmentalization may limit the access of myristoyltransferase and the nascent G t␣ polypeptide to 14:2n-6 when it is supplied directly, while allowing free access to the 14:2n-6 pool produced by retroconversion.
The fatty acid retroconversions seen in our experiments have the limited chain shortening characteristics associated with peroxisomal ␤-oxidation (Schulz, 1991), as shown by the lack of production of 12:1n-9 and 12:2n-6 or other short chain fatty acids. In support of this are electron microscopy studies showing that Mü ller and photoreceptor cells of frog retinas contain significant peroxisome-like organelles (St. Jules et al., 1992). However, more experiments are necessary to determine where the observed ␤-oxidation occurs. Retina peroxisomes may prove to be more vigorous or less stringent in retroconverting these fatty acids compared with those from other tissues. Such differences may account for the absence of 14:1n-9 or 14:2n-6 on myristoylated liver proteins such as cytochrome b 5 reductase (Ozols et al., 1984), even though liver peroxisomes carry out fatty acid retroconversions.
Although we do not know the functional significance for N-terminal fatty acylation of photoreceptor proteins with unsaturated forms of myristate, the retroconversion pathways we have investigated could play a major role in both normal visual function and retinal disease states. Congenital defects such as adrenomyelnoneuropathy (Moser et al., 1987), neonatal adrenoleukodystrophy (Jaffe et al., 1982), infantile Refsum's disease (Poll-the et al., 1987), and Zellweger syndrome (Bowen et al., 1964) are known afflictions where peroxisomal ␤-oxidation is impaired. The general phenotype of these diseases is deterioration of nervous system, often involving the retina. These symptoms may arise in part from impairment of the biosynthesis of docosahexaenoic acid (22:6n-3) (Martinez et al., 1994), an essential component of neuronal cells including retina photoreceptors, which requires the peroxisomal based retroconversion of 24:6n-3 to 22:6n-3 (Voss et al., 1991). Because N-terminal fatty acylation with 14:1n-9 and 14:2n-6 may be required for proper function of phototransduction proteins, impairment of 14:1n-9 or 14:2n-6 production could also have a devastating effect on normal visual function. It is known that many types of retinal degeneration involve lipidated phototransduction proteins, such as rhodopsin (autosomal dominant and autosomal recessive retinitis pigmentosa) (Dryja et al., 1990;Rosenfeld et al., 1992) and the ␤-subunit of cGMP phosphodiesterase (PDE␤) (autosomal recessive retinitis pigmentosa) (McLaughlin et al., 1993). A specific example of defective protein lipidation in retinal degeneration is choroideremia, whose basis is a mutation in geranylgeranyl transferase (Seabra et al., 1993), the enzyme that isoprenylates PDE␤ and G proteins of the Rab family. Therefore, genes encoding proteins involved in the pathways required for heterogeneous fatty acylation of retina protein warrant further consideration as retinal degeneration candidates.
FIG. 6. Distribution of desaturation, enlongation, or retroconversion products in lipid classes derived from retina membrane pellets. Total lipids obtained from [ 3 H]14:0, [ 3 H]18:1n-9, [ 3 H]18:2n-6, and [ 3 H]18:3n-3 radiolabeled retina membranes were resolved into phospholipids (PL), free fatty acids (FFA), and triglycerides (TG) using onedimensional two-step TLC. Lipid classes were directly counted to determine total radioactivity. Lipid classes were then saponified, and the resulting fatty acids were converted to phenacyl esters and chromatographed on HPLC. The data are represented as percentages calculated by taking the ratio of radioactivity for the individual lipid classes or fatty acid species to the total radioactivity in the original total lipid extract or separated lipid class.