Acyl-CoA Synthetase in Rat Liver Peroxisomes COMPUTER-ASSISTED ANALYSIS OF CELL FRACTIONATION EXPERIMENTS*

The subcellular distribution of the acyl coenzyme A synthetases of rat liver was reinvestigated in order to determine whether part of this activity occurs in per-oxisomes. Rat liver was fractionated by differential centrifugation and by equilibrium density centrifugation. Acyl-CoA synthetase was assayed using a new, simple extraction procedure on three substrates: pal-mitate, laurate, and octanoate. Comparison of the resulting synthetase distributions with the distributions of marker enzymes for peroxisomes, mitochondria, and endoplasmic reticulum demonstrated the presence of some synthetase activity in each of the three organelles. These trimodal synthetase distributions were evaluated quantitatively by means of a computer program that calculated optimal linear combinations of marker enzymes using a least squares criterion. Peroxisomes were found to contain 7% of the liver’s palmitoyl-CoA synthetase activity and 6% of its lau-royl-CoA synthetase activity, but no demonstrable octanoyl-CoA synthetase activity. The remainder of these activities are divided between the mitochondria and endoplasmic reticulum, in a manner that agrees with previous studies. The chain length specificity of the synthetase(s) of each organelle appears to be unique. The absolute activity of the peroxisomal palmitoyl-CoA synthetase is sufficient to maintain maximal peroxisomal &oxidation. Clofibrate treatment of the rats caused a 2.6-to 3.1-fold increase in the liver’s total acyl-CoA synthetase activities. The subcellular distribution was not greatly affected

o whom correspondence should be addressed.
paper. oxidation of fatty acids: the well known mitochondrial system and the recently described peroxisomal system (7, 8). The peroxisomal system contains acyl-CoA oxidase, crotonase, /?hydroxybutyryl-CoA dehydrogenase, and thiolase activities.
It is active on long-chain acyl-CoA's and catalyzes the formation of acetyl-coA. Moreover, its activity is increased by 1 order of magnitude in rats treated with a variety of hypolipidemic drugs (7,9).
In the present study, we have reinvestigated the subcellular distributions of the fatty acid:CoA ligases in rat liver in order to determine whether any of these enzymes are located in peroxisomes. We find that, indeed, about 6 to 12% of the palmitate:CoA ligase, but little or no 0ctanoate:CoA ligase is found in this organelle. These results extend the role of rat liver peroxisomes in lipid metabolism. An abstract of these results has been published (10).

EXPERIMENTAL PROCEDURES
Pa1mitate:CoA Ligase Assay-The reaction conditions are based on those described by Marcel and Suzue (11) and Bar-Tana et al. (12). The final reaction mixture of 0.2 ml contained 150 mM Tris-HC1 buffer (pH 7.4), 6.2 n m MgC12, 0.5 mg/ml of Triton X-100, 2 m~ EDTA, 2.5 r m ATP, 600 PM CoA, 1.0 nm dithiothreitol, 100 p~ ['4C]palmitic acid (1.5 to 2.5 Ci/mol, uniformly labeled), and 0 to 20 pg of enzyme protein. The palmitic acid was added in heptane (1.5 cJ/0.2 ml of reaction mixture). Enzyme samples were diluted in 1 mM dithiothreitol, 10 n m Tris-HC1 buffer (pH 7.4), and added to the icecold reaction mixture. After incubation at 37°C for 6 min, the samples were returned to the ice bath.
Unreacted palmitate was separated from palmitoyl-CoA by the following one-step procedure: 3.25 ml of room temperature M/C/H* (1.41:1.25:1.00, v/v) (13) was mixed with each sample after which 1.05 ml of 0.1 M sodium acetate (pH 4.0). was added. 3 The samples were vigorously shaken for 20 min ( K r a f t model 5-500 shaker) and then centrifuged at room temperature for 10 min at 1500 rpm. The palmitoyl-CoA formed in the reaction remained in the upper, aqueous phase (2.45 ml volume) and 99.9% of the unactivated palmitic acid was extracted into the lower, organic phase. Aqueous phase (0.50 m l ) was withdrawn, mixed with 10 ml of Triton X-lOO/toluene scintillation mixture (14), and its radioactivity was determined.
Control experiments where the enzyme was added after termination of the reaction with M/C/H were systematically included in each series. The radioactive palmitoyl-CoA measured in the aqueous phase was corrected for the small amount of unextracted substrate determined in these controls. Other controls where the reaction was carried out without CoA or without ATP gave similar results.
Almost all the synthetase assays reported in this paper were carried out on freshly isolated fractions. Although the synthetase activities in many purified fractions are stable to freezing, those in L and X fractions (see below) are very labile. 0ctanoate:CoA Ligase and Laurate:CoA Ligase Assays-The same conditions were used as for the palmitate:CoA ligase assay except that the fatty acid concentration was 50 PM, the specific activity was 4 Ci/mol, and the label was only in the first carbon. The same type of controls were systematically carried out.

Peroxisomal Acyl-CoA Synthetase
Analysis of Ligase Reaction Products-The standard reaction mixture was modified by the use of [14C]palmitic acid with a specific activity of 613 Ci/mol and a 10-min incubation. After extraction, 0.5 ml of the aqueous phase was lyophilized, dissolved in 50 pl of water, and spotted on Whatmann No. 3MM paper. ['4C]Palm1tate (in heptane) and [L4C]palmitoyl-CoA (in water) were spotted as standards. The chromatogram was then developed in butanolacetic acid:water (523, v/v) for 4 h (15), dried, and then autoradiography was performed with Dupont Cronex x-ray film.
Enzyme units are micromoles/min except for catalase and cytochrome oxidase which obey first order reaction kinetics (see Ref. 16). Protein (Lowry) is expressed as milligrams based on a bovine serum albumin standard.
Differential Centrifugation-Fiat liver was fractionated by differential centrifugation exactly according to de Duve et al. (19)  Equilibrium Density Centrifugation-Rat liver was fractionated by differential centrifugation (16) to prepare a h fraction, containing peroxisomes and smaller mitochondria (similar to an L fraction). The h fraction was then further fractionated by centrifugation in a steep linear sucrose gradient, exactly according to Leighton et al. (16). Peroxisomes and mitochondria separated by this method are about 93 to 95% pure (16). The preinjection of rats with Triton WR-1339 (16) was omitted from two of the experiments on normal rats and from the fractionation done on clofibrate-treated rats.
Animals-Female Sprague-Dawley rats were obtained from Charles River Breeding Laboratories (Wilmington, MA) and male Fisher F-344 rats were obtained from Microbiological Associates (Walkersville, MD). Rats were fasted overnight before they were killed by means of a guillotine.
Clofibrate Treatment-Male F-344 rats were fed Purina chow containing 0.5% (w/w) clofibrate for 2 to 3 weeks. As in previous experiments (7,9), the clofibrate was added to ground chow in ether. This time, after evaporating the ether, the chow was moistened with water, shaped into pellets and dried.

Materials-[U-14C]
Palmitic acid, [l-14C]octanoic acid, [l-'4C]butyric acid, and [1-14C]palmitoyl-CoA were purchased from New England Nuclear (Boston, MA); [l-'4C]lauric acid was obtained from Amersham/Searle (Arlington Heights, IL). Nonradioactive fatty acids were products of Applied Science Labs (State College, PA). The radioactive fatty acids were mixed with unlabeled fatty acids in heptane to obtain the specific activities given above. Substantial errors were found to occur if labeled and unlabeled fatty acid were added separately to the reaction mixture.
Computer Calculations-Our experimental data consisted of distributions of acyl-CoA synthetase activities among various subcellular fractions, together with distributions of marker enzymes. The amount of the synthetases in each organelle was determined from these data by applying the principle of calculating the linear combinations of marker enzyme distributions that would fit the measured synthetase distributions (19).
We calculated the unique set of coefficients (one coefficient for each marker enzyme distribution) for which the sum of (coefficient 1 X distribution 1 + coefficient 2 X distribution 2 + . . .) best fits the experimentally measured test distribution. The criterion of goodness of fit was the sum of the squares of the differences between calculated and measured values over all the fractions. The best coefficients were calculated explicitly by differentiating the equation for the goodness of fit with respect to each of the coefficients, setting these partial derivatives equal to zero, and solving the resulting simultaneous equations for the coefficients.
The computation was carried out on a digital computer based on the method of Bevington (20). The essence of the method may be illustrated with an example of 15 fractions, three marker enzymes, and one test enzyme. One constructs a 15 X 4 array (A (15,4)) where the first three columns contain the marker enzyme activities in the 15 fractions, and the fourth column contains the test enzyme activities. Each element of a 3 X 4 array (B (3,4)) is then computed: The B array is split in two between the third and fourth column, the I= 1 left-hand 3 X 3 matrix is inverted, and this inverse is multiplied by the right-hand 3 X 1 matrix. The product is a 3 X 1 array containing the desired Coefficients: Further matrix arithmetic yields calculated test enzyme activities that may be plotted and compared to the measured test enzyme activities.
The data used may be in the form of activities, concentrations, specific activities, etc. If normalized values (16,19) are used, the resulting coefficients are the fractional contributions of each organelle to the test activity. We used normalized total activities for the differential centrifugation data, and normalized concentrations for the sucrose gradient data.
The use of this method is described below under "Computer Calculations" and is illustrated in Figs. 9 and 11. We will gladly provide our computer program upon request.

RESULTS'
Assay Procedure-As shown in Table I, a single extraction with M/C/H gives a cleaner separation of palmitoyl-CoA from palmitic acid than do five extractions with Dole's Medium (12). The M/C/H extraction procedure is also effective with lauric and octanoic acids, but it is unsuitable for butyric acid (Table I).
In preliminary experiments, highly purified peroxisome preparations were found to contain some palmitate:CoA ligase activity. Before investigating the subcellular distribution of the enzyme activity in detail, we first determined the optimal assay conditions using purified peroxisomal, mitochondrial, and microsomal fractions. As shown in Fig of Mg2+ and Triton X-100 were critical, with 6 mM MgClZ and 0.5 m g / d of Triton X-100 giving optimal results. With the final assay conditions adopted (given under "Experimental Procedures"), the rate of reaction is proportional to the amount of added protein up to a limit of 20 pg ( Fig. 2A and other experiments not illustrated). The reaction is linear for at least 10 min (Fig. 2B).
Distribution of Acyl-CoA Synthetases after Differential Centrifugation of Normal Rat Liver-Three fractionations were carried out. The means and standard deviations of the enzyme distributions are given in Table I1 and the mean relative specific activities are plotted versus cumulative protein according to de Duve et al. (19) in Fig. 3. The liver content of protein and of the marker enzymes (Table 11) is in the normal range (16,17,19). The separations of nuclei, mitochondria, microsomes, and cell sap are reproducible and similar to what has been described previously ( 19). The recoveries in the fractionations were also satisfactory ( Table 11).
The distribution of palmitate:CoA ligase is most similar to the distribution of esterase, the marker enzyme for the microsomes. However, there is more ligase than esterase in the M fraction. These results imply that the palmitate:CoA ligase is located principally in the microsomes, and to a lesser extent in the mitochondria, in agreement with all previous studies In contrast, the distribution of 0ctanoate:CoA ligase is sim- Bevington (20) discusses various schemes to weight the data during the calculation; however, we found by trial and error that the best fits were obtained without any weighting.
Portions of this paper (including Figs  ilar to that of the mitochondrial marker enzyme cytochrome oxidase. However, there is more ligase than cytochrome oxidase in the P and N fractions. These results imply that this short chain fatty acid ligase is largely in the mitochondria, and to a lesser extent in the microsomes, which is also in agreement with results of previous studies (6).
The distribution of lauroyl-CoA synthetase is intermediate between those of the palmitoyl-CoA and octanoyl-CoA synthetases. This enzyme appears to be partitioned approximately equally between mitochondria and microsomes.
In no case is there an enrichment of ligase activity in the L  Table 111, together with the recoveries. The 0ctanoate:CoA and 1aurate:CoA ligases were assayed on only four of the five gradients. fraction, as there is for enzymes that are located exclusively in the lysosomes or in the peroxisomes (e.g. catalase). However, these results do not exclude the possibility that a small fraction of one or more of the ligases could be associated with the peroxisomes. The differential centrifugation method does not sufficiently separate peroxisomes and mitochondria to detect this.

Distribution of Acyl-CoA Synthetases after Equilibrium Density Centrifugation of a Light Mitochondrial Fraction-
X fractions consisting primarily of peroxisomes, lysosomes, and small mitochondria (16) were subjected to isopycnic centrifugation in a sucrose gradient in order to separate the organelles on the basis of their different densities (16). The mean composition of five X preparations is summarized in Table 111. Fig. 4 illustrates the averages of the five isopycnic gradients. The mitochondria are located in the center of each gradient and are responsible for the major peak of protein.
The peroxisomes are located at greater densities (to the right in Fig. 4) and are responsible for the shoulder in the protein distribution. Microsomes constituted only a small portion of the protein in the X fractions loaded on the gradients; those present have a broad distribution extending throughout the gradient.
The distribution of palmitate:CoA ligase activity has two peaks: a large one coincident with the mitochondria and a small one coincident with the peroxisomes. In contrast, octanoate:CoA ligase has a distribution similar to that of cytochrome oxidase; it has no second peak at high density. The 1aurate:CoA ligase distribution is intermediate with a shoulder in the peroxisomal region of the gradient. These results demonstrate that some long chain, but no short chain, fatty acid: CoA ligase is located in peroxisomes in normal rat liver. These results will be evaluated quantitatively under "Computer Calculations'' (below).
Characterization of the Peroxisomal Pa1mitate:CoA Ligase-The peroxisomal enzyme appears to be specific for ATP. Substituting 2.5 mM GTP for ATP in the reaction mixture gave only slight activity.
The product of the peroxisomal reaction was identified as palmitoyl-CoA by paper chromatography (Fig. 5). The RF values of the peroxisomal and mitochondrial reaction products were similar to each other and to that of the palmitoyl-CoA standard. The slight differences observed may be due to the presence of buffer and salts in the reaction product samples, or possibly to desaturation of the palmitoyl-CoA formed.

Effects of Clofibrate on Ligases-
The activities of the ligases were increased 2.6-to 3.1-fold in the livers of rats treated with clofibrate for 2 to 3 weeks (Table IV). In addition, palmitoyl-CoA oxidation was increased 11-fold, which is consistent with previous results (7,9). The distributions of the ligase activities among fractions prepared by the differential centrifugation of a clofibratetreated rat liver are shown in Table V and Fig. 6. The percentage of protein in the mitochondrial (M) and light mitochondrial (L) fractions is increased relative to controls. A corresponding decrease in protein is seen in the microsomal fraction (P). At least part of this change probably is due to a broadening in the size distribution of the microsomes: considerably more esterase activity is found in the M and L fractions than in the controls.
The distribution of catalase activity is also markedly altered by clofibrate treatment, with some 58% appearing to be soluble.
Comparison of the ligase distributions with the marker enzyme distributions (Fig. 6)  Equilibrium density centrifugation of a X fraction from clofibrate-treated rat liver ( Fig. 7 and Table VI) also gives results similar to what we observed with normal liver (Fig. 4). Palmitoyl-CoA synthetase, but not octanoyl-CoA synthetase, shows a shoulder of activity in the region of the gradient occupied by the peroxisomes.
Chain Length Specificity of the Ligases in the Three Organelles-The specific activities of the acyl-CoA synthetases toward each of the substrates were calculated for the most highly purified fractions that we obtained for each organelle.
As shown in Fig. 8A, the chain length specificity of each organelle appears to be unique. The endoplasmic reticulum is most active on palmitate and the activity drops with decreasing chain length. The mitochondria are most active on laurate and are only 43% as active toward the other two substrates. The peroxisomes, like the mitochondria, are most active on laurate. However, the peroxisomes have nearly as much activity toward palmitate as towards laurate, but only 8% as much towards octanoate.
The chain length specificities of the three organelles show similar patterns in the livers of the clofibrate-treated rats, but the specific activities are all higher (Fig. 8B). Since the amount of peroxisomal protein increases 2.5 to 5 times during clofibrate treatment (7), the total peroxisomal synthetase per g of liver apparently increases even more than the others.   < 0.02). B, clofibrate. elles, we made use of de Duve's postulate of biochemical homogeneity (21). According to this postulate (which has received considerable experimental verification), if all the palmitoyl-CoA synthetase were to be located in the mitochondria, for example, then the distribution of this enzyme throughout our various fractions would be identical to the distribution of any marker enzyme for mitochondria (e.g. cytochrome oxidase). If, however, the palmitoyl-CoA synthetase were located in several organelles, then its distribution' should be a linear combination of the distributions of the marker enzymes for each of the organelles in which it is located (19). It appears from our results that the ligases are present in mitochondria, endoplasmic reticulum, and peroxisomes. Therefore, we investigated whether or not the ligase distributions could be quantitatively accounted for by linear combinations of the marker enzymes for these three organelles and, if so, what was the exact contribution of each organelle. We wrote a computer program to do this, described under "Experimental Procedures," that determines optimal linear combinations by a least squares criterion. Our calculations will be described in detail for palmitoyl-CoA synthetase to make the method clear.

Computations of Subcellular Distributions of Acyl-CoA Synthetases in Normal Liver
We first analyzed the palmitate:CoA ligase distribution of Fig. 3 (differential centrifugation). The computer determined that the optimal linear combination was (0.2789 X cytochrome oxidase distribution + 0.7615 X esterase distribution) (Fig. 9).
This combination is shown in Fig. 1OA with circles superimposed on the measured palmitate:CoA ligase distribution of Fig. 3. The agreement is very good, except in the light mitochondrial (L) fraction, which contains the least activity. The two coefficients are listed in the first column of Table VII, together with their sum, which is 1.0404. The sum of the coefficients should of course be 1.oooO; deviations may be due to inaccuracies in the measured distributions. We consider a value of 1.0404 evidence of an acceptably low noise level. The computer's least squares coefficients were divided by their sum to get the normalized coefficients of 0.27 and 0.73 shown in Table VII. We interpret these values to mean that 73% of  This sum is plotted as open circles in Fig. 1OA.   FIG. 10 (right). Computed optimal linear combinations of cytochrome oxidase and esterase marker enzymes to fit acyl-CoA synthetase distributions of Fig. 3. The solid lines indicate the experimental measurements and the circles represent the computed optimal linear combination. The coefficients of this combination are given in Table VII. the palmitoyl-Cot\ synthetase is located in the endoplasmic reticulum and 27% of this synthetase is located principally in mitochondria, but some of this 27% may be located in peroxisomes which largely co-sediment with the mitochondria during differential centrifugation.
We next considered the palmitoyl-CoA synthetase distribution of Fig. 4, in which peroxisomes and mitochondria were separated on the basis of their densities. The computer calculated the following optimal linear combination: (0.291 X catalase distribution + 0.576 X cytochrome oxidase distribution + 0.146 X esterase distribution)6 (Fig. 11). The sum of the coefficients was 1.013 (Table VIII), and the observed fit was good (Fig. 12A).
Of the palmitoyl-CoA synthetase activity assigned to per-  Lines 6 and 7). This observed partition between peroxisomes and mitochondria must be adjusted to account for the fact that we did not load equal amounts of the two organelles onto the gradients. On the contrary, we loaded an average of 28% of the peroxisomes and 19% of the mitochondria (Table 111) Table VIII.
We are now in a position to calculate the distribution of the palmitoyl-CoA synthetase among all three organelles. Of the 27% of the activity found by differential centrifugation to be in mitochondria and/or peroxisomes, 25% of it is in peroxisomes according to isopycnic centrifugation. Thus: These results are summarized in Table IX. We performed exactly the same computations on the lauroyl-CoA and octanoyl-CoA synthetase distributions of Figs. 3 and 4. The calculated coefficients are given in Tables VI1 and VI11 and the goodness of fit can be examined in Figs. 10 and 12. We find that 6% of the lauroyl-CoA synthetase is peroxisomal with the balance divided evenly between mitochondria and endoplasmic reticulum. Octanoyl-CoA synthetase is mainly mitochondrial; peroxisomes could contribute 1% of this activity.
The calculations just described were based on the mean results of three differential centrifugation experiments, and on the mean results of five density gradient experiments. In order to estimate the reliability of these values, we performed the same computations of optimal linear combinations on all of the individual experiments. The results are given in the lower parts of Tables VI1 and VIII. Substantial variation in the calculated coefficients is observed. However, as shown in It was necessary to include esterase in this computation because even though microsomes constituted little of the protein on the gradient, they accounted for 14.6% of the palmitoyl-CoA synthetase.
' Let X be the true fraction of enzyme activity in Organelle A and  + B ( l -X ) ) . If one measures A, B,  and Y (as we do), one can solve for X X = BY/ (BY + A ( l -Y ) ) .  Fig. 4. The marker enzyme distributions of Fig. 4 are multiplied by the coefficients of Table VIII Table   VIII. Table IX, the effect on the final conclusions is moderate. In the case of palmitoyt-CoA synthetase, the peroxisomal contribution ranged from 6 to 12% in the individual experiments. The mean of these individual values was 876, in good agreement with the value of 7% obtained above.
Analysis of the individual experiments indicated that the injection of Triton WR-1339 into the rats used for some of the density gradient experiments did not affect the results. Sex and strain of rat did not affect the differential centrifugation results. However, the one isopycnic centrifugation of male F-344 rats showed somewhat larger peroxisomal contributions to the ligase activities than did the female Sprague-Dawley rats. We feel cautious about attaching too much significance to this result since there was only one such experiment, and have therefore averaged it in with the rest.
One important fact that is apparent from Table IX is that the peroxisomal activities calculated on the means of the fractionation data are near the lower ends of the ranges of values obtained on the individual experiments. These mean values are 7 and 6% of the liver's palmitoyl-CoA and lauroyl-CoA synthetase activities, respectively, and we consider them the most reliable estimates of the peroxisomal contributions.
The true values might be somewhat higher, but they are unlikely to be lower.
In the case of octanoyl-CoA, the fractionation data of Fig.  4 gave no hint of any peroxisomal activity. We conclude that none is detectable. The quantitative calculations demonstrate that we cannot exclude a possible peroxisomal contribution of 1%. This is the limit of the present methods.
Computations on Clofibrate-treated Rats-The same calculations were performed on the fractionation data of the clofibrate-treated rats. The coefficients and conclusions are summarized in Table X, the goodness of fits (not illustrated) were similar to those seen with the controls. In general, the contributions of the three organelles to the synthetase activities are close to or within the range of values seen with the normal rats. The peroxisomes have 12% of the palmitoyl-CoA synthetase activity, or about twice the control value, suggesting that the peroxisomal synthetase is elevated more than the others. This is consistent with the observed increases in both the specific activity of the peroxisomal palmitoyl-CoA synthetase (Fig. 8) and in total peroxisomal protein (7). On the other hand, the clofibrate experiment was done on only one male F-344 rat and the results are close to the values observed for control male F-344 rats. Therefore, no f i i conclusion can be drawn. In general, clofibrate seems to increase the activities of the synthetases in the various organelles approximately in parallel.

DISCUSSION
Number of Different Synthetases-Since the synthetase(s) in each organelle has a unique chain length specificity (Fig.  8), rat liver must contain a minimum of three of these enzymes. Each organelle could also contain more than one synthetase, each active on different chain length substrates. Aas (6) reported the presence of five synthetases in liver. Recently, Tanaka et al. (22) have reported isolating identical long chain acyl-CoA synthetases from mitochondria and microsomes. Philipp and Parsons (23), on the other hand, have described a very different long chain mitochondrial synthetase. The kinetic properties that we determined for the palmitoyl-CoA synthetase activities in our three purified organelles (Fig. 1) were all quite similar. This does not, of course, mean that the enzymes are the same.
It is noteworthy that our observed chain length specificities are consistent with other functional properties of the organelles. The microsomes are involved in esterifying long chain fatty acids into a variety of lipids, and they activate the long chain fatty acids.
The chain length specificity of the peroxisomal synthetase(s) resembles the chain length specificity of the peroxisomal /I-oxidation system which is active on long chain fatty acids and inactive on short ones (8).
The mitochondria are active towards all the chain lengths tested, including octanoate, the shortest one studied. This is consistent with the fact that mitochondria oxidize short fatty acids as well as long ones.
Absolute Activity of Peroxisomal Fatty Acid Actiuation-The peroxisomes contain 7% of the three units of palmitoyl-CoA synthetase present in 1 g of liver; this amounts to 0.21 pmol of palmitoyl-CoA/min/g of liver. The peroxisomal fatty acid oxidizing system oxidizes acyl-CoA's at an average rate of approximately 1 pmo1 of acetyl-coA produced/min/g of liver in the chain length range of 8 to 16' (7,8). With palmitoyl-CoA as substrate, it can carry out five cycles of P-oxidation (8). Since only one activation is required per five acetyl-coA's formed, an activation of 0.2 pmol of palmitate/min/g of liver would be required to keep pace with a maximally functioning peroxisomal /I-oxidation system. Thus, the available peroxisomal palmitoyl-CoA synthetase is sufficient.
Nothing is known at present about the source of ATP for peroxisomal fatty acid activation. Nor is it known whether some palmitate is activated within the peroxisome while other palmitate is activated elsewhere and enters as palmitoyl-CoA. The observed latency of the peroxisomal synthetase (little activity in the absence of detergent, Fig. 1D) suggests that the peroxisomal membrane is impermeable to one (or more) of the substrates of the reaction.
Comparison with Other Znuestigations-Our results on the properties of the microsomal palmitoyl-CoA synthetase are generally consistent with most previous observations insofar as K,,, values are concerned. Our observed specific activity is at the low end of the range observed by other workers (1-6).
Our findings on the subcellular distributions of the various synthetase activities are consistent with previous reports, especially those of Farstad et al. (1) and Lippel et al. (5), except that we find that some of the activity previously attributed to mitochondria is in fact in the peroxisomes.
While this manuscript was in preparation, Shindo and Hashimoto (24) reported the presence of some acyl-CoA synthetase activity in rat liver peroxisomes. Differences exist between their results and ours, probably due to the impurity of their fractions. They report that their peroxisomal fraction contains mitochondria but claim it is free of microsomes. However, this claim is unverifiable because of the lack of marker enzyme distributions and recoveries in their paper.
Daae and Aas (25) have previously investigated the effect of clofibrate treatment of rats on the total liver activity of octanoyl-CoA and palmitoyl-CoA synthetase. They found that both activities were increased by the drug treatment, and our results are in agreement with theirs.
There are precedents for our finding acyl-CoA synthetase in rat liver peroxisomes. Acyl-CoA synthetase activity has been reported in the peroxisomes of unicellular organisms (26,27) as well as in glyoxysomes (28), closely related plant organelles.
Computations-The use of a computer program to determine whether linear combinations of marker enzymes can account for experimental distributions permitted more precise calculations of the subcellular distributions of the ligases than would otherwise have been possible. This method makes use of all the data of the fractionation procedure instead of merely the few purest fractions (usually assumed to be pure). The method requires that recoveries in the fractionations be good and assumes the truth of de Duve's postulate of biochemical homogeneity (21). In the present study, the calculations gave good fits to the experimental data and yielded consistent and reproducible quantitative conclusions. The success of the computations confirms the correctness of the underlying assumptions in this case. This method should be generally useful for analyzing cell fractionation data. A similar method was recently used independently by Schneider et al. The one consistent deviation in the fits that we observed occurred in the light mitochondrial (L) fractions in the differ-than the linear combination's value. This suggested the possibility that the L fraction, which is also rich in lysosomes, might be hydrolyzing some of the product (or otherwise converting it to an M/C/H-soluble compound). We tested this hypothesis by assaying conversion of [14C]palmitoyl-CoA to M/C/H-extractable material by an L fraction under our usual reaction conditions. Some conversion was observed, but not enough to account for all of the discrepancy.
This example illustrates one other virtue of the linear combinations analysis: discrepancies may point the way to further experiments.

X of Recovered Radioactivity I n aqueoYI
Phase organic phase