Partial phenotypic suppression of a peroxisome-deficient animal cell mutant treated with aminoglycoside G418.

Certain enzymes normally associated with peroxisomes, such as the dihydroxyacetone phosphate (DHAP) acyltransferase involved in plasmalogen biosynthesis, are present at low levels in peroxisome-deficient mutants of Chinese hamster ovary (CHO) cells. We now show that the aminoglycoside G418 increases the residual DHAP acyltransferase in mutant ZR-82 by 60-fold. This is accompanied by a dose- and time-dependent restoration of the plasmalogen content. G418 treatment of ZR-82 also increases residual peroxisomal beta-oxidation activity by 3.8-fold. G418 does not affect wild-type CHO cells (CHO-K1) or a different peroxisome-deficient mutant, ZR-78.1. The effects of G418 on ZR-82 are transient, since plasmalogens and DHAP-acyltransferase decline to basal levels 5 days after G418 withdrawal. Other aminoglycosides and lysosomotropic agents do not alter plasmalogen levels in ZR-82. The subcellular distribution of catalase (an enzyme of the peroxisomal matrix which is present in normal amounts in peroxisome-deficient mutants but is mislocalized in the cytosol) is unaffected by G418 treatment of ZR-82, demonstrating that G418 does not restore peroxisomes. Localization of catalase by immunofluorescence microscopy confirms a total absence of intact peroxisomes in ZR-82, either before or after exposure to G418. This study is the first to demonstrate that some peroxisome-deficient mutants can be induced to accumulate functional DHAP acyltransferase and other peroxisomal enzymes, usually missing in the absence of peroxisomes. G418 may have some therapeutic value in selected patients with inborn errors of peroxisome assembly, such as Zellweger syndrome.

Certain enzymes normally associated with peroxisomes, such as the dihydroxyacetone phosphate (DHAP) acyltransferase involved in plasmalogen biosynthesis, are present at low levels in peroxisomedeficient mutants of Chinese hamster ovary (CHO) cells. We now show that the aminoglycoside G418 increases the residual DHAP acyltransferase in mutant ZR-82 by 60-fold. This is accompanied by a dose-and time-dependent restoration of the plasmalogen content. G418 treatment of ZR-82 also increases residual peroxisomal /3-oxidation activity by 3.8-fold. G418 does not affect wild-type CHO cells (CHO-K1) or a different peroxisome-deficient mutant, ZR-78.1. The effects of G418 on ZR-82 are transient, since plasmalogens and DHAP-acyltransferase decline to basal levels 5 days after G418 withdrawal. Other aminoglycosides and lysosomotropic agents do not alter plasmalogen levels in ZR-82. The subcellular distribution of catalase (an enzyme of the peroxisomal matrix which is present in normal amounts in peroxisome-deficient mutants but is mislocalized in the cytosol) is unaffected by G418 treatment of ZR-82, demonstrating that G418 does not restore peroxisomes. Localization of catalase by immunofluorescence microscopy confirms a total absence of intact peroxisomes in ZR-82, either before or after exposure to G418. This study is the first to demonstrate that some peroxisome-deficient mutants can be induced to accumulate functional DHAP acyltransferase and other peroxisomal enzymes, usually missing in the absence of peroxisomes. G4 18 may have some theraputic value in selected patients with inborn errors of peroxisome assembly, such as Zellweger syndrome.
A key function of peroxisomes is the detoxification of hydrogen peroxide by catalase within the peroxisomal matrix (de Duve and Baudhuin, 1966). In addition, peroxisomes play a unique role in mammalian lipid metabolism. Peroxisomes contain dihydroxyacetone phosphate (DHAP)' acyltransferase and alkyl-DHAP synthase, the first two enzymes of plasmalogen (ether lipid) biosynthesis (Hajra and Bishop, 1982), and they also possess a unique P-oxidation pathway for degrading very long chain fatty acids that are not metabolized by mitochondria (Lazarow, 1987).
All peroxisome-associated enzymes and proteins studied thus far are synthesized on free polyribosomes and imported, after translation, into existing peroxisomes (Borst, 1986). In peroxisome-deficient cells from patients with Zellweger syndrome (Suzuki and Hashimoto, 1986;Suzuki et al., 1987;Santos et al., 1988a, 198813;Wiemer et al., 1989;Schram et al. 1986), peroxisomal proteins are synthesized normally, but most are rapidly degraded. Presumably, the accelerated turnover of peroxisomal enzymes accounts for their reduced activity in these mutant cells.
Tartakoff and co-workers (Gupta et al., 1988) (Stevens and Raetz, 1990) has demonstrated that the class F mutants are deficient in peroxisomal DHAP acyltransferase and ether lipids, although they retain intact peroxisomes. These data imply that ether lipid synthesis or some other peroxisomal function might be required for surface expression of Thy-1. In view of these findings, we examined the effects of G418 on peroxisome-deficient CHO cell mutants. We now show that G418 treatment increases DHAP acyltransferase activity, plasmalogen synthesis, and peroxisomal @-oxidation activity in the peroxisome-deficient CHO cell mutants ZR-82 and S3 but not in mutant ZR-78.1. The G418-treated mutants do not regain intact peroxisomes. The effects of G418 are transient and cannot be duplicated by other aminoglycosides or lysosomotropic agents. These data imply that G418 may have theraputic value in some patients with peroxisome-deficiency diseases.
Cells and Culture Conditions-All cells were grown in Ham's F-12 medium containing L-glutamine (Whittaker Bioproducts, Walkersville, MD) supplemented with 10% fetal bovine serum (HyClone Laboratories, Logan, UT), penicillin G (100 units/ml), and streptomycin (100 pg/ml) in an atmosphere of 5% CO, in air. Cells were passaged using trypsin. Wild-type CHO cells (CHO-Kl) were obtained from the American Type Culture Collection (CCL-61). Strains ZR-82 and ZR-78.1 are peroxisome-deficient mutants derived from mutagen-treated stocks of CHO-K1 (Zoeller and Raetz, 1986;Allen et al., 1989). S3 is a spontaneous peroxisome-deficient mutant derived from CHO-K1 (Allen et al., 1990). Strains of CHO-K1, ZR-82, and ZR-78.1 cotransfected with pKOneo and the bone-liver-kidney form of alkaline phosphatase (designated KlnAP2.2, 82nAP8.3, and 78.1nAP1.5, respectively) were the generous gift of Dr. Victoria L. Stevens (Merck, Sharp and Dohme Research Laboratories, Rahway, NJ). Transfected cells were grown in F-12 containing 0.5 mg/ml G418 to ensure retention of the neomycin resistance gene. All amounts of G418 indicate the active drug concentration. Incorporation of [2-14C]Ethanolamine into Phospholipids-Cells were plated in 60-mm tissue culture dishes a t 3-5 X 10' cells/dish in 3 ml of medium containing [2-14C]ethanolamine (0.2 pCi/ml). Half of the dishes also received G418 to 50 pg/ml from a stock solution (10 mg/ml in PBS (pH 7.4)). After 72 h at 37 "C, the medium was removed, each cell monolayer was rinsed with 2 ml of PBS, and the cells were scraped into 0.8 ml of PBS. Lipids were extracted using a neutral two-phase system (Bligh and Dyer, 1959;Morand et al., 1988).
Ethanolamine-containing phospholipids were separated using a two-step, one-dimensional thin-layer chromatography (TLC) system . This treatment separates phosphatidylethanolamine, plasmenylethanolamine, and lysophosphatidylethanolamine in a single dimension. The labeled lipids were visualized by autoradiography, and individual lipid species were scraped off the plates into scintillation vials containing 1 ml of methanol. Samples were counted in Biosafe I1 liquid scintillation mixture.
For dose dependence experiments, ZR-82 was plated in 60-mm tissue culture dishes (5 X lo' cells/dish) in 3 ml of medium containing 0.2 pCi/ml [2-"C]ethanolamine and 10, 25, or 50 pg/ml G418. After 1, 2, 3, or 5 days at 37 "C, the cells were rinsed with PBS and harvested. The lipids were extracted using neutral Bligh-Dyer con-ditions as described above. In control incubations, cells were labeled for 3 days at 37 "C with [2-'4C]ethanolamine (0.2 pCi/ml) in the absence of G418. Plasmenylethanolamine content was determined using the two-step one-dimensional TLC, followed by autoradiography as described above, and quantified by scraping and counting.
To examine the reversibility of the G418 effect, cells were plated at 2 X lo5 cells/lOO-mm tissue culture dish in 10 ml of medium containing 50 pg/ml G418 and 0.2 pCi/ml [2-''C]ethanolamine. After 3 days at 37 "C, the cells were trypsinized, collected by centrifugation, and resuspended in PBS. Portions of each culture were extracted using neutral Bligh-Dyer conditions as described above. The remaining cells were plated into five new dishes in F-12 containing 0.2 pCi/ ml [2-''C]ethanolamine without G418. After replating, cells were grown at 37 "C for an additional 1, 2, 3, 5, or 6 days at 37 "C with medium changes after 3 days. After removal of G418, cells were harvested at the indicated times, lipids were extracted, and plasmenylethanolamine content was determined as above. Control cultures were labeled for 3 days at 37 "C with [2-''C]ethanolamine in the absence of G418. Enzyme Assays-To quantitate peroxisomal (pH 5.5) and microsomal (pH 7.4) DHAP acyltransferase activities, 2 X lo5 cells were plated into each of at least four 100-mm tissue culture dishes in 10 ml of medium. Half of the dishes also received 50 pg/ml G418. After 3 days at 37 "C, the medium was removed, and the cells were rinsed with 4 ml of PBS. Cells were detached from each dish at 25 "C by a 10-min incubation in 5 ml of PBS containing 0.5 mM EDTA, harvested by low speed centrifugation, and washed with 10 ml of buffer containing 20 mM Hepes (pH 7.4), 250 mM sucrose, and 1 mM EDTA. Cells were resuspended in the same buffer (-0.5 rnl/lOO-mm dish) and frozen at -20 "C to break the cells. All cells were thawed and kept on ice prior to assay (-30 min). DHAP acyltransferase activity was assayed as described by Schlossman and Bell (1976) with the following modifications. Reaction mixtures contained 50 mM KCI, 100 mM MES, and 100 mM TES (pH 5.5 or 7.4, respectively), 5 mM MgCh, 8 mM NaF, 2 mM KCN, 2 mg/ml bovine serum albumin, 0.1 mM palmitoyl-CoA, 0.8 mM ["PIDHAP (3-4 pCi/pmol), and 50-100 pg of cell protein in a total volume of 0.3 ml. Some reactions also contained 0.63 mg/ml N-ethylmaleimide. After 15-20 min at 30 "C, reactions were stopped by the addition of 3 ml of chloroform/methanol (12, v/v). 0.5 ml of water, 400 pg of rat liver phospholipids, and 50 p1 of concentrated HCl were also added. After a 10-min incubation at 25 "C, 1 ml of chloroform and 1 ml of water were added to form a two-phase system. The phases were separated by centrifugation, and the lower phase was washed twice with 3 ml of preequilibrated acidic upper phase (Bligh and Dyer, 1959). The final lower phases were collected, evaporated to dryness in glass scintillation vials, and counted in 10 ml of Biosafe I1 liquid scintillation mixture.
Catalase was assayed in cellular fractions (see below) as described previously (Peters et al., 1972;Zoeller and Raetz, 1986), except that all incubations contained 250 mM sucrose to maintain isotonicitiy. For latency determinations, assays were performed in the absence of Triton X-100.
Protein was determined by the method of Bradford (1976) or Smith et al. (1985. Subcellular Fractionation-Six 100-mm dishes of each cell type were grown up, with or without 50 pg/ml G418, for 3 days at 37 "C as described above and then incubated for 1 additional day at 37 "C in medium without G418. The cells were rinsed with 4 ml of PBS, harvested by incubation with 5 ml of 0.5 mM EDTA in PBS for 10 min at 25 "C, and collected by centrifugation at 400 X g , , for 10 min. All subsequent steps were performed at 4 "C. Cells were washed once with 10 ml of buffer containing 20 mM Hepes (pH 7.2), 250 mM sucrose, and 1 mM EDTA and then resuspended in 5 ml of the same buffer. Nitrogen cavitation was used to break the cells (200 p.s.i., 10 min). Homogenates were centrifuged for 10 min a t 400 X g,,, and the postnuclear supernatants were held on ice. The nuclear pellets were resuspended in 1 ml of buffer and centrifuged a t 400 X go" for 10 min. The two postnuclear supernatants were pooled and then recentrifuged at 400 X ga. for 10 min to remove residual nuclei and intact cells. The final postnuclear supernatant fractions were centrifuged in a 50-Ti rotor (Beckman) a t 100,000 X gav for 90 min in a Beckman ultracentrifuge. The supernatants (soluble fractions) were collected, and the pellets (particulate fractions) were resuspended in 0.5 ml of buffer. Catalase was assayed in all fractions prior to freezing.
Macromolecular Synthesis-Cells were plated a t 3 X 10' cells/60mm dish in 3 mi of medium, in the presence or absence of 50 pg/ml (nontransfected cells) or 500 pg/ml (transfected cells) G418. After 3 days a t 37 "C, cells were labeled with [methyl-:'H]thymidine or L-[:''S] methionine as follows. For [methyl-'H]thymidine labeling, cells were incubated in 2.5 ml of fresh medium containing 1 pCi/ml [methyl-"Hlthymidine. After 2 h a t 37 "C, the medium was removed, each dish was rinsed with 2 ml of PBS, and ['HJthymidine incorporation was evaluated as described previously by Mosley et al. (1981). T o assess [:"S]methionine incorporation, another set of dishes received 2.5 ml of fresh medium containing 1 pCi/ml [%]methionine. After 2 h a t 37 "C, the medium was removed, and the cells were rinsed with 2 mi of PBS/dish. The cells were scraped into 1 ml of PBS, and 1 ml of ice-cold 20% (w/v) trichloroacetic acid was added. After a 1 h incubation on ice, the solutions were filtered through Whatman GF/B filters, and each filter was washed with 3 ml of ice-cold 10% trichloroacetic acid. Filters were dried prior to liquid scintillation counting. For some experiments, cells were incubated for 3 days a t 37 "C with 50 pg/ml G418 and then reincubated for 24 h in the absence of G418 before labeling with ['Hlthymidine or [:'"S]methionine. Parallel dishes were used for protein determinations.

RESULTS
Colony Autoradiography of Peroxisomal DHAP Acyltransferase-DHAP acyltransferase is a peroxisomal enzyme that catalyzes the first step in the plasmalogen biosynthetic pathway (Hajra and Bishop, 1982;Lazarow, 1987;Zoeller et al., 1988). CHO cell mutants lacking morphologically identifiable peroxisomes (Zoeller et al., 1989) are deficient in DHAP acyltransferase activity and contain reduced levels of plasmalogens relative to wild-type CHO cells (Zoeller and Raetz, 1986;Allen et al., 1990). CHO cells overlaid with polyester cloth discs proliferate upward and form colonies in the polyester cloth (Raetz et al., 1982). The immobilized colonies can then be assayed for DHAP acyltransferase using colony autoradiography (Zoeller and Raetz, 1986). By this assay, peroxisome-deficient CHO cells lack detectable DHAP acyltransferase activity (Zoeller and Raetz, 1986;Zoeller et al., 1989;Allen et al., 1989). When the aminoglycoside G418 was included in the culture medium during the last 3 days of the polyester overlay, the colonies of the peroxisome-deficient mutant ZR-82 regained measurable DHAP acyltransferase activity (Fig. 1, column A ) , although the level of activity was reduced somewhat relative to CHO-K1. All of the colonies on the disc responded to G418 treatment. Similar data were obtained for a spontaneous peroxisome-deficient mutant, S3 (data not shown), isolated by a photosensitization-based selection method (Morand et al., 1990). However, not all peroxisome-deficient mutants responded to G418, as illustrated for strain ZR-78.1 (Fig. 1, column A ) .
Plasmenylethanohmine Content of G418-treated Cells-The recovery of DHAP acyltransferase activity in ZR-82 suggested that plasmalogen levels might also increase in response to G418. Plasmenylethanolamine is the major ether lipid in CHO cells and normally comprises -50% of the ethanolaminelinked phospholipid (Zoeller and Raetz, 1986). In peroxisomedeficient cells, plasmenylethanolamine levels are reduced, and phosphatidylethanolamine content increases, but the total amount of ethanolamine-containing phospholipid remains Autoradiographic detection of peroxisomal DHAP acyltransferase in colonies treated with G4 18. Cells were grown into HD-17 polyester cloth discs for 8-10 days at 37 "C, with medium changes every 4 days, as decribed previously (Zoeller et al., 1989). Some dishes received 50 pg/ml G418 during the final 3 days of polyester overlay. Polyester cloths were harvested, rinsed with PBS (pH 7.4), air dried, and frozen at -20 "C to lyse the cells. The immobilized colonies were assayed for peroxisomal DHAP acyltransferase at pH 5.5 in the presence of N-ethylmaleimide as described previously (Zoeller and Raetz, 1986;Allen et al., 1989). Polyester cloths were exposed to x-ray film for -24 h a t -80 "C. Following autoradiography, the polyester cloths were stained with Coomassie Blue to visualize the colonies. A and C, colony autoradiography of peroxisomal DHAP acyltransferase; B and D, the corresponding polyester cloths stained with Coomassie Blue. A and B, +G418; C and D, controls.
FIG. 2. Effect of G418 on the plasmenylethanolamine content of CHO cells. Cells were grown for 3 days at 37 "C in medium containing 0.2 pCi/ml [2-''C]ethanolamine, in the presence or absence of 50 pg/ml G418. The lipids were extracted, and ethanolaminecontaining phospholipid species were separated using the two-step, one-dimensional TLC described under "Experimental Procedures." Lipid species were visualized using autoradiography and quantified by liquid scintillation counting. Solid bars, controls; hatched bars, +G418. Error bars indicate the range of duplicate cultures.
constant (Zoeller and Raetz, 1986). To assess the effect of G418 on plasmalogen levels, cells were labeled with [2-"C] ethanolamine in the presence or absence of G418 for 3 days at 37 "C, and plasmalogen levels were quantified using TLC as described under "Experimental Procedures." Incubation of CHO cells with 50 pg/ml G418 increased plasmenylethanolamine levels in ZR-82 4-fold to -20% of the ethanolamine phospholipid pool (Fig. 2), about half the amount found in CHO-K1. Similarly, plasmenylethanolamine levels increased -2-fold in S3, but G418 treatment did not increase plasmenylethanolamine levels in ZR-78.1, consistent with the observation that G418 did not affect DHAP acyltransferase activity in this strain (Fig. 1). The plasmenylethanolamine content of CHO-K1 was also not affected by G418.
Time Course and Concentration Dependence of the Effect of (3418 on Plasmalogen Biosynthesis-To determine optimal conditions for G418 treatment, ZR-82 was incubated at 37 "C Peroxisomal Enzyme and Plasmalogen Accumulation Induced by G418 with [2-I4C]ethanolamine and increasing concentrations of G418. Plasmenylethanolamine levels were measured after 1-5 days. Fig. 3 shows that G418 caused a dose-and timedependent increase in plasmenylethanolamine in ZR-82, with plasmalogen content approaching wild-type levels after 5 days in 50 pg/ml G418. Similar data were obtained for mutant S3 (not shown).
Experiments by Tartakoff and co-workers (Gupta et al., 1988) revealed that the ability of G418 to restore surface Thy-1 in the class F lymphoma mutants is transient. Therefore, we investigated the stability of plasmenylethanolamine levels in CHO cells exposed to G418. Wild-type and peroxisomedeficient cells were treated with 50 pg/ml G418 for 3 days at 37 "C and then reincubated for up to 6 days in medium lacking G418. Cells were labeled with [2-14C]ethanolamine throughout the course of the experiment, and plasmenylethanolamine content was determined using TLC. Care was taken to ensure that all cells remained in logarithmic growth throughout the course of the experiment. As shown in Fig. 4, plasmalogen levels increased in ZR-82 and S3 in the presence of G418 and continued to rise for 1-2 days after G418 removal. Plasmalogen levels then declined, reaching basal levels -5 days after G418 was removed. Again, mutant ZR-78.1 was unaffected. Plasmenylethanolamine levels also increased slightly in CHO-K 1 after G418 removal (Fig. 4).
Specificity for G418"In another set of experiments, other aminoglycosides were tested for their ability to induce plasmalogen synthesis in ZR-82. Cells were grown for 3 days in the presence of [2-14C]ethanolamine and 50 pg/ml G418, or the maximum nontoxic dose of paromomycin (500 pg/ml), hygromycin B (5 pg/ml), or gentamycin (200 pg/ml). As shown in Fig. 5, none of the other aminoglycosides altered the plasmenylethanolamine content of ZR-82. This is interesting in view of the fact that gentamycin is nearly identical to G418 in structure (Gupta et al., 1988). In addition, all of these aminoglycosides inhibit protein synthesis (Vazquez, 1978) and lysosome function (Mingeot-Leclercq et al., 1988); and paromomycin and G418 promote misreading of mRNAs (Eustice and Wilhelm, 1984b;Buchanan et al., 1987). Lysosomotropic agents, such as chloroquine or ammonium chloride (at their maximum nontoxic doses), and low concentrations of cycloheximide also had no effect on the plasmalogen content of ZR-82 (Fig. 5).
Plasmenylethanolamine Synthesis in ZR-82 Transfectants  lane 7, 50 p~ chloroquine; lane 8, 1 mM ammonium chloride. At the end of the incubation period the lipids were extracted using neutral Bligh-Dyer conditions (Esko and Raetz, 1980), ethanolamine-containing lipid species were separated by TLC, and the amount of plasmenylethanolamine was quantified as described under "Experimental Procedures." Error bars indicate the range of duplicate cultures.
Resistant to G418"G418 is "slowly toxic" to eukaryotic cells, i.e. it accumulates in cells over time, eventually causing cell death. Although there was no apparent cytotoxicity caused by G418 in any of the experiments described above, we wanted to assess more clearly whether cytotoxicity was responsible for the elevated plasmalogen content of ZR-82. CHO cells transfected with a neomycin resistance gene (Santerre et al., 1984;Danielson et al., 1989) acquire the ability to grow indefinitely in concentrations of G418 that would kill nontransfected cells. The neomycin resistance gene encodes a phosphotransferase that catalyzes the transfer of the y-phosphate from ATP to G418 (Danielson et al., 1989). The phosphorylated G418 product is nontoxic (Danielson et al., 1989). This suggests that the concentration of unmodified G418 in transfected cells should be lower than in nontransfected cells. for 3 days a t 37 "C in medium containing 0.2 pCi/ml [Z-"C)ethanolamine in the presence or absence of 500 pg/ml G418. The lipids were extracted using neutral Bligh-Dyer conditions (Esko and Raetz, 1980), and the amount of plasmenylethanolamine associated with the cells was quantified as described above. Solid bars, controls; hatched bars, +500 pg/ml G418. Error bars indicate the range of duplicate cultures.
CHO cells cotransfected with pKOneo and the bone-liverkidney form of alkaline phosphatase (designated KlnAP2.2, 82nAP8.3, and 78.1nAP1.5, respectively) can be propagated indefinitely in 0.5 mg/ml G418 (data not shown). The transfectants were labeled for 3 days with [2-14C]ethanolamine in the presence or absence of 0.5 mg/ml G418, and their plasmalogen content was measured as described above. Plasmalogen levels in KlnAP2.2 and 78.1nAP1.5 were unchanged after growth in G418 (Fig. 6). On the other hand, plasmenylethanolamine levels in 82nAP8.3 increased to -15-16% of the ethanolamine phospholipid pool in the presence of 0.5 mg/ml G418 (Fig. 6). The magnitude of the plasmalogen increase in 82nAP8.3 was proportional to the concentration of G418 present in the medium (data not shown), as was shown for ZR-82 (Fig. 3). However, higher concentrations of G418 were required to elevate plasmalogens in the transfectants. This is consistent with the idea that the effective intracellular concentration of (3418 is lower in the transfected cells. The fact that the transfectants can be grown indefinitely in medium containing 0.5 mg/ml G418 renders unlikely the possibility that cytotoxicity was responsible for the plasmalogen increase in the mutant strains. It is also unlikely that the alkaline phosphatase gene present in the transfectants contributed to the plasmalogen increase because similar data were obtained for ZR-82 transfected with only the vector.2 Macromolecular Synthesis-Previous studies have shown that G418 partially inhibits protein synthesis and slows the growth of eukaryotic cells (Vazquez, 1978). T o evaluate the sensitivity of CHO cells to these effects of G418, CHO-K1, ZR-82, ZR-78.1, and their corresponding transfectants were grown for 3 days in 50 or 500 pg/ml G418, respectively. Protein and DNA synthesis were measured after a subsequent 2-h incubation, as described under "Experimental Procedures." G418 treatment inhibited protein synthesis as much as 50% (Table I) and DNA synthesis up to 40%. There was no apparent correlation between increased plasmalogen levels (Figs. 2 and 6) and the degree of inhibition of protein and and DNA synthesis (Table I), as revealed by comparing data for ZR-82 and 82nAP8.3, or ZR-78.1 and 78.1nAP1.5. These results suggest that the effects of G418 on macromolecular synthesis are not related to its effects on ether lipid synthesis. In addition, the effects of G418 on protein and DNA synthesis ' T. Kobayashi and C. Raetz, unpublished observation.

TABLE I
Synthesis of macromolrcuh in crlls treated with G41R Cells were grown a t 37 'C in the presence or ahsence of 50 pg/ml (nontransfected cells) or 500 pg/ml (transfected cells) G418. After 3 days the G418-containing medium was removed, and the cells were labeled for 2 h a t 37 "C in fresh medium containing 1 pCi/ml L-[ "'SI methionine or 1 pCi/ml [meth.yl]-"HJthymidine as descrihed under "Experimental Procedures." In each case, the incorporation of label into trichloroacetic acid-precipitable material was determined as described under "Experimental Procedures." In some cases cells were reincubated in medium without G418 (+ G418 + wash) for 24 h prior to label incorporation. Parallel cultures were used for protein determinations. Each value is the mean of three independent cultures, and the data varv bv less than 20%. ND. not determined. were readily reversible (Table I). One day after C418 removal, macromolecular synthesis in ZR-82 and CHO-K1 was indistinguishable from control cells. Phospholipid Composition-Because G418 increased plasmalogen levels in ZR-82, we thought that G418 might also alter the levels of other phospholipids in CHO cells. CHO-Kl and ZR-82 were labeled for 3 days with ["ZP]orthophosphate, in the presence or absence of 50 pg/ml G418, and phospholipid species were separated by two-dimensional TLC. As shown in Fig. 7 and Table 11, G418 treatment signficantly increased the plasmenylethanolamine content of ZR-82, without affecting other phospholipid classes. Quantitation of the radiolabeled phospholipids (Table 11) revealed that G418 increased plas-

Phospholipid composition of CHO-Kl and ZR-82 cells treated with G418
CHO-K1 and ZR-82 cells were labeled for 3 days at 37 "C in medium containing [32P]orthophosphate in the presence or absence of G418, and labeled phospholipids were separated by two-dimensional TLC as indicated in the legend to Fig. 7. Individual lipid species were scraped off the thin-layer plates into scintillation vials containing 1 ml of methanol and counted in 10 ml of Biosafe I1 liquid scintillation mixture. Each value is the mean f S.D. of three independent cultures. SM, sphingomyelin; PC, phosphatidylcholine; PS, phosphatidylserine; PI, phosphatidylinositol; PE, phosphatidylethanolamine; PlasE, plasmenylethanolamine. menylethanolamine levels 5-fold in ZR-82 to -5% of the total phospholipid, consistent with the data obtained using [2-"C] ethanolamine labeling . The phospholipid composition of CHO-K1 was not altered by G418 ( Fig. 7 and Table   11). However, G418 induced a generalized phospholipidosis in CHO cells, resulting in a 30-50% increase in the phospholipid/protein ratio in CHO-K1 and ZR-82 (data not shown). This is consistent with previous studies (Aubert-Tulkens et al., 1979) which demonstrated that G418 and other aminoglycosides cause a generalized pdospholipidosis in rat fibroblasts without affecting phospholipid composition.
Quantitation of Peroxisomal Enzyme Activities in Cells Treated with G418"The ability of G418 to restore plasmalogens (Fig. 7) and DHAP acyltransferase activity (Fig. 1) in ZR-82 suggested that G418 might affect the activity of additional peroxisomal enzymes. To address this question, we first quantified the amount of DHAP acyltransferase activity in cells grown in the presence or absence of (3418. CHO cells contain two DHAP acyltransferase activities (Ballas et al., 1984;Webber et al., 1987). The peroxisomal form has an acidic pH optimum and is relatively resistant to N-ethylmaleimide. The microsomal form has a neutral pH optimum and is inhibited by N-ethylmaleimide. As shown in Table 111, G418 treatment increased peroxisomal DHAP acyltransferase activity in ZR-82 more than 60-fold, whereas ZR-78.1 and CHO-K1 were unaffected. G418 did not perturb microsomal DHAP acyltransferase activity in any of the strains (Table   111).
In another set of experiments, the effect of G418 on peroxisomal @-oxidation was evaluated. The peroxisomal @-oxidation pathway can degrade palmitate, as well as very long chain fatty acids ( X 2 2 ) which are not substrates for the mitochondrial pathway (Lazarow, 1987;Singh et al., 1984). Another distinguishing featuw of the peroxisomal pathway is its resistance to inhibition by potassium cyanide (Mannaerts et al., 1979). We found that G418-treated ZR-82 and S3 regained the ability to metabolize lignoceric acid in vitro (Table IV) via the peroxisomal @-oxidation pathway. Taken together, these data suggest that G418 was able to restore the activity of a t least five peroxisomal enzymes in mutant ZR-82 (DHAP acyltransferase and the four enzymes of the peroxisomal @-oxidation pathway) (Lazarow, 1987). Peroxisomal @-oxidation activity was not increased in ZR-78.1 after G418 treatment (Table IV).
Peroxisome Content of G418-treated Cells-The simultane-

Effect of G418 on peroxisomal p-oxidation of lignoceric acid
Cells were grown up at 37 "C in the presence or absence of 50 pg/ ml G418 and broken by sonication as described under "Experimental Procedures." Reaction mixtures contained 0.3 M mannitol, 50 mM MOPS (pH 7.4), 10 mM ATP, 1 mM NAD, 0.1 mM FAD, 0.2 mM CoA, 5 mM MgCl,, 5 mM KCN, 10 p~ [1-"Cllignoceric acid (46 mCi/ mmol), and 50-200 pg of cell protein in a total volume of 0.2 ml. After 1 h at 37 "C, reactions were stopped, and the amount of ["C] acetate produced was quantified essentially as described . All assays contained KCN to inhibit mitochondrial 0oxidation. Data are the mean f S.D. of three determinations. ous recovery of multiple enzyme activities normally associated with peroxisomes in G418-treated ZR-82 suggested that G418 might transiently restore intact peroxisomes, i.e. cause a temporary suppression of the mutant phenotype. In peroxisome-deficient cells catalase, in contrast to the peroxisomal enzymes discussed thus far, remains active but is located in the cytosol rather than the peroxisomal matrix (Zoeller and Raetz, 1986;Zoeller et al., 1989;Allen et al., 1989). Therefore, we investigated the ability of G418 to restore intact peroxisomes in ZR-82, using the localization of catalase as an indicator of peroxisome content. CHO-K1 and ZR-82 were grown up at 37 "C, lysed using nitrogen cavitation, and separated into soluble and particulate fractions as described under "Experimental Procedures." As shown in Table V, the specific activity of catalase was similar in CHO-K1 and ZR-82. HOWever, the activity was primarily associated with the particulate fraction of wild-type cells but the cytosolic fraction of ZR-82. Although a small amount of the catalase activity was associated with the particulate fraction in ZR-82, this activity did not show any latency, strongly suggesting that some of the enzyme was adsorbed onto cellular membranes rather than being contained in the peroxisomal matrix. Cells were also cultured in G418 using conditions that restored nearly normal amounts of plasmalogens in ZR-82 (Fig. 4, day 41, i.e. 3 days

TABLE V
Catalase distribution in cells treated with G418 CHO-K1 and ZR-82 were grown up at 37 "C in the presence or absence of 50 pg/ml G418 as described under "Experimental Procedures." Cells were broken by nitrogen cavitation (200 p.s.i., 10 min), and the homogenates were separated into membrane and soluble fractions as described under "Experimental Procedures." Catalase was assayed in all fractions in a mixture of 10 mM imidazole-HCl (pH 7.2), 0.5 mg/ml BSA, 8.82 mM H202, 20 pg of cell protein, and 0.1% Triton X-100, in a total volume of 0.3 ml. For latency determinations the Triton X-100 was omitted. All reactions also contained 250 mM sucrose to maintain isotonicity. After 30 s a t 25 "C, each reaction was terminated by the addition of 3 ml of a saturated titanyl sulfate solution (Leighton et al., 1968). One unit of catalase activity is defined as the amount of enzyme required to degrade 90% of the H202 in 1 min (Peters et al., 1972). Each value is the mean + S.D. of three independent preparations.
Since a small number of intact peroxisomes might not be detected in subcellular fractionation experiments, we used immunofluorescence microscopy (Santos et al., 198813;Wiemer et al., 1989) to assess the subcellular location of catalase more closely in cells treated with G418. By this assay (Fig. 8), CHO-K1 contained numerous punctate fluorescent structures, i.e. peroxisomes, whereas ZR-82 lacked these structures altogether. Similarly, no peroxisomes were detected in ZR-82 after 3 days of growth in the presence of 50 pg/ml G418 (Fig.   8) or in 82nAP8.3 grown continuously in medium supplemented with 0.5 mg/ml G418 (data not shown). The peroxisome content of CHO-K1 was not affected by G418 (Fig. 8). These results are consistent with the subcellular fractionation data and demonstrate that G418 treatment does not restore intact peroxisomes in mutant ZR-82.

DISCUSSION
Although the aminoglycoside G418 is used extensively in transfection experiments to select for eukaryotic cells that have acquired neomycin resistance genes (Santerre et al., 1984;Danielson et dl., 1989), the effects of G418 on cell physiology are complex and not fully characterized. We now report that G418 treatment restores plasmalogen biosynthesis and multiple peroxisomal enzyme activities in some peroxisome-deficient CHO cell mutants. The effects of G418 are time-and concentration-dependent. Plasmalogen levels in mutant ZR-82 approximate those of wild-type cells after a 5day incubation in medium containing 50 pg/ml G418 (Fig. 3). Similarly, in extracts of G418-treated ZR-82 cells, DHAP acyltransferase is increased up to 60-fold (Table III), and peroxisomal P oxidation is increased -4 fold (Table IV). Every colony derived from a population of ZR-82 cells appears to respond to (3418 (Fig. l), and the effects of G418 are transient (Fig. 4). Stable revertants are not induced.
The independently isolated CHO mutants used in this study differ considerably in their ability to increase plasmalogens and peroxisomal enzyme activities in response to G418. After 3 days in G418, plasmenylethanolamine levels increase 4-5fold in ZR-82 and -2-fold in S3 but are unchanged in ZR-78.1 and CHO-K1 (Fig. 2). Similar results are obtained with DHAP acyltransferase (Table 111) and peroxisomal P-oxidation activities (Table IV). The molecular basis for this difference is not obvious given that all the mutants appear to belong to a single complementation group (Zoeller et al., 1989;Allen et al., 1990). We are currently examining the sequence of the gene encoding the 35-kDa membrane protein of peroxisome assembly, recently cloned from rat liver (Tsukamoto et al., 1991), in ZR-82 and ZR-78, since we have found recently that the rat cDNA encoding this protein corrects all aspects of the peroxisome deficiency phenotype of ZR-82 and ZR-78." Although G418 restores the activities of multiple peroxisomal enzymes in ZR-82, the cells do not appear to regain intact peroxisomes. Under conditions in which the plasmenylethanolamine content of ZR-82 reaches -75% of the wild-type value (Fig. 4, day 4 ) , catalase latency and distribution are not significantly altered (Table V). Similarly, when catalase is visualized using immunofluorescence microscopy, CHO-K1 contains numerous fluorescent peroxisomes, whereas ZR-82 lacks these structures, either before or after G418 treatment (Fig. 8). These data show that G418 does not cause a complete suppression of the mutation in ZR-82, but we cannot eliminate the possibility that G418 restores peroxisomes devoid of catalase.
An important implication of these data is that the peroxisome-deficient CHO cell mutant ZR-82 retains the ability to synthesize peroxisomal proteins even though it lacks intact peroxisomes (Zoeller et al., 1989). All peroxisomal proteins studied thus far are synthesized on free polyribosomes and imported after translation into preexisting peroxisomes (Borst, 1986). In the absence of peroxisomes peroxisomal proteins are synthesized normally but are rapidly degraded Santos et al., 1988a), presumably because they cannot be targeted to peroxisomes. It is likely that this by G418 rapid turnover of peroxisomal proteins accounts for the low activity of peroxisomal enzymes in these cells.
Two lines of evidence support the conclusion that the effects of G418 on peroxisomal enzymes in ZR-82 are not related to gross cytotoxicity. (i) G418 (0.5 mg/ml) increases plasmenylethanolamine levels -3-fold in ZR-82 stably transfected with pKOneo (Fig. 6). These cells can be grown indefinitely in this concentration of G418. (ii) Although G418 partially inhibits protein and DNA synthesis in all of the CHO strains, including CHO-K1 (Table I), these effects are readily reversible within 24 h of G418 removal, and there is no apparent correlation between the degree of inhibition of protein synthesis and the plasmalogen content of the CHO mutants ( Table I, data for ZR-82 and 82nAP8.3 or ZR-78.1 and 78.lnAP1.5). In addition, although all aminoglycosides inhibit protein synthesis (Eustice and Wilhelm, 1984a) and lysosome function (Aubert-Tulkens et al., 1979) in eukaryotes, the other aminoglycosides and lysosomotropic agents tested fail to mimic the effects of G418 on plasmalogen synthesis in ZR-82 (Fig. 5).
G418 (but not other aminoglycosides or lysosomotropic agents) has been shown previously to restore transient, partial surface expression of Thy-1 antigen in class F, B, and A Thy-1-deficient T-cell lymphoma mutants (Gupta et al., 1988).
These mutants lack surface expression of Thy-1 because they cannot properly assemble the glycosyl phosphatidylinositol anchor of Thy-1 (Hyman, 1989;Stevens and Raetz, 1991). Thy-1 protein is synthesized normally, but its turnover is accelerated (Gupta et al., 1988;Hyman, 1989). The common features of the Thy-1-deficient and the peroxisome-deficient mutants suggest that G418 somehow slows the turnover of proteins that cannot be targeted to their proper locations. If this were the only explanation for the G418 effect, however, some response should have been observed with mutant Although the mechanism of action of G418 is unknown, one interesting possibility is that a heat shock response is induced in G418-treated cells. Recent work by Buchanan et al. (1987) suggests that heat shock proteins are produced in human fibroblasts treated with G418 but not in fibroblasts treated with other aminoglycosides. Therefore, it is tempting t o hypothesize that heat shock proteins are synthesized in G418-treated ZR-82 cells and that the heat shock proteins increase the activity of peroxisomal enzymes by promoting their proper folding and assembly in the absence of peroxisomes. Correctly folded peroxisomal proteins might have a longer half-life, even if they are not contained in intact peroxisomes. To explain the difference between mutants ZR-82 and ZR-78.1, one would have to postulate that the latter does not accumulate heat shock proteins in response to G418, perhaps because of a secondary mutation.
Given that peroxisomes are not present in the CHO mutants treated with G418, it will be necessary to establish the location of the accumulated peroxisomal proteins in these cells. Santos et al. (1988b) discovered that in cells lacking peroxisomes, such as fibroblasts from patients with Zellweger syndrome, peroxisomal integral membrane proteins are not cytosolic but are found in small, apparently empty, vesicles termed "peroxisomal ghosts." Although ghosts are present in other peroxisome-deficient cells (Wiemer et al., 1989), including our CHO mutants (Zoeller et al., 1989), peroxisomal ghosts have not been isolated, and their exact composition is unknown.
Our findings indicate that one must be very careful when using G418/neoR as a selectable marker in transfection experiments because phenotypic suppression may arise from ZR-78.1.
G418 treatment which is unrelated to the cotransfected DNA.
Regardless of the mechanism of action of G418, this study is the first demonstration that peroxisome-deficient cells can, at least transiently, regain substantial activity of peroxisomal enzymes and the ability to synthesize plasmalogens. If cultured skin fibroblasts from patients with Zellweger syndrome, or any of the other human peroxisome-deficiency diseases, also regain plasmalogens when grown in the presence of G418, it may be possible to use G418 or G418 analogs to alleviate some of the symptoms associated with these diseases.