Gene Inactivation in Lec35.1 (Mannosylation-defective) Chinese Hamster Ovary Cells

In Lec36. 1 CHO mutants, mannose-P-dolichol is syn- thesized but does not participate in the production of glycosylphosphatidylinositol (GPI) anchor precursors or dolichol-linked oligosaccharides. We tested Lec36.1 cells for stable expression of a cDNA encoding GPI-anchored human folate-binding protein (FBP) with the eukaryotic expression vector pJB20. All normal transfectants, but no Lec35.1 transfectants, expressed FBP activity. However, rather than an inability to produce GPI anchors, lack of FBP expression in Lec36.1 was caused by gradual inactivation of the FBP cDNA. FBP cDNA became fully inactive after 2-3 months of cul- ture, and FBP activity was not restored upon correc-tion of the Lee35 mutation. Southern blot analysis revealed that inactivation was associated with gross rearrangement of FBP cDNA. The cellular FBP gene remained intact. Because the Lec36.1 cell line has the ability to in-activate transfected human FBP cDNA, caution should be exercised when expressing transfected cDNAs in Lec36.1 and similar GPI anchor/glycosylation mutants. Interestingly, these results suggest similarities between the Lec36.1 defect and the human paroxysmal hemoglobinuria, which may involve gradual inactivation of a gene necessary for man- nosylation obtained after digestion were collected and analyzed by liquid scintillation counting. Southern Blotting-Isolation of genomic DNA, restriction enzyme digestion, agarose gel electrophoresis, transfer to Pall A membranes, and blotting were performed as described earlier (8).

The most extensively characterized Lec35 line has been termed Lec35. 1 (18). In this report, we examined the ability of these cells to express GPI-anchored proteins by transfection with a cDNA encoding the human folate-binding protein (FBP) (4), which has a well characterized GPI anchor (5), using the eukaryotic expression vector pJB2O. In contrast to normal cells, stable Lec35.1 transfectants failed to express FBP. However, this was not due to an inability to produce GPI anchors; in fact, the Lec35 mutation was found to be sufficiently leaky to produce small amounts of these structures (18). Rather, lack of FBP activity was caused by an unexpected inactivation of the FBP cDNA in every Lec35.1 transfectant that was examined. Genetic analysis of the Lec35.1 transfectants revealed that inactivation was time-dependent and associated with disruption of the transfected cDNA. The implications of these results for transfection experiments with other glycosylation mutants, as well as the similarity between the Lec35.1 mutation and the mutation causing the human disease paroxysmal nocturnal hemoglobinuria, are discussed.

EXPERIMENTAL PROCEDURES
Reagents-Sera, nutrients, antibiotics, and G418 sulfate (Geneticin) used to prepare culture medium were from GIBCO. Hygromycin was from Calbiochem. Bacillus cereus PI-specific phospholipase C was from Boehringer-Mannheim. Folic acid-free M199 medium and [3H]folic acid (40 Ci/mmol) were gifts from B. Kamen [3ZP]dATP (3000 Ci/mmol) used for DNA probes was from Du Pont. Polybrene was obtained from Sigma and Aldrich. Pall A membranes were from ICN.
Plasmids-pJB2O (6.2 kb) is a member of the cytomegalovirus family (6) of eukaryotic expression vectors and was derived from the initial vector of this series, pCMV1. Starting with pTZ18R (Pharmacia LKB Biotechnology Inc.), pCMV1 was created in the laboratory of D. Russell (UT-Southwestern) by adding (in relative order as they exist in the plasmid) a fragment containing nucleotides -760 to +3 of the promoter/enhancer of the human cytomegalovirus major immediate early gene, a synthetic polylinker, a fragment containing the transcription terminator and polyadenylation signal (nucleotides 1533-2157) of the human growth hormone gene, and a fragment from pcD-X containing the SV40 origin of replication and early promoterenhancer (6). pCMVl was further modified (C. Brewer and M. Roth, UT-Southwestern) by splicing a 1.3-kb fragment containing a neomycin resistance gene downstream of the SV40 early promoter. pJB20 was created by P. Beck (UT-Southwestern) by altering a BgnI site in the polylinker to form a unique EcoRI site and by inserting a 0.35-kb HindIII-BamHI fragment into the polylinker (downstream of the EcoRI site) containing the SV40 small T intron of pMSG (Pharmacia LKB Biotechnology Inc.).
pF53 was constructed in S. Lacey's laboratory (UT-Southwestern) by ligating a 0.99-kb EcoRI fragment encoding the human folic acidbinding protein cDNA (4) into the EcoRI site of pJB20 in the sense orientation.
Cell Lines-Normal CHO cells (CHO-K1) were from ATCC (CCL 61). Lec35.1 cells were previously termed PIR and SwR-100 and are described in the accompanying paper (18) and in prior publications 6729 (1)(2)(3). The h c 3 5 designation was given by Dr. P. Stanley.* Lec35 cells are resistant to concanavalin A/swainsonine (1) and have a recessive mutation (2, 3) that allows the synthesis of MPD but not its participation in mannosyltransferase reactions (2). These lines were free of mycoplasma as judged by the Mycotect test (GIBCO/ BRL) .
Cell Culture and Cell Fusion-Cells were normally maintained in Ham's F-12 medium buffered at pH 7.2 with 15 mM Na/Hepes and containing 100 units/ml penicillin, 100 pg/ml streptomycin, 2% fetal bovine serum, and 8% calf serum as described (1). This medium was used for transfection experiments as well. For experiments involving selection with 10 pg/ml concanavalin A, 10% fetal bovine serum was used in place of the fetal bovine/calf mixture. Cell fusion was performed exactly as described (2,3) by selecting for resistance to G418 and hygromycin.
Transfection-The Polybrene method (7) was used to create stable G418-resistant transfectants. Normal CHO cells and Lec35.1 cells were transfected with either pJB2O or pF53. Transfectants were selected with 1 mg/ml G418. After 14 days, well resolved colonies were isolated with glass cloning cylinders and designated by the parental line, plasmid, and a letter to indicate the order picked, e.g. CHO-F53B. Several were further subcloned by limiting dilution and labeled with the number of the subclone, e.g. CHO-F53B2. As indicated in the text, pools of GIN-resistant transfectants were sometimes taken. G418 was used during the first 3 weeks of selection after transfection and then omitted from the medium. During the course of these studies, pJB2O and pF53 transfectants were periodically rescreened with G418 and concanavalin A/swainsonine (1) to ensure that they retained the expected phenotypes.
Assay and PI-specific Phospholipase C Digestion of Cell Surface Human Folic Acid-binding Protein-Assays were typically performed with 24-well plates as described (4), but they were also performed with larger culture vessels by increasing the volumes of the various wash and incubation solutions. Approximately 4 X 10' cells were plated in 24-well plates and allowed to grow 3-4 days until confluent. Plates were chilled on ice and all steps were performed with ice-cold reagents. Monolayers were washed twice with 1.5 ml of PBS and then treated for 30 s with 1.5 ml of acid saline (0.15 M sodium chloride adjusted to pH 3 with acetic acid) to displace bound folic acid that originated from the F-12 medium. After washing twice with 1.5 ml of PBS, monolayers were incubated with 1.2 ml of folic acid-free M199 medium containing 2.8 g M ['HJfoIic acid for 1 h with gentle rocking. The medium was removed and the monolayers washed again with two 1.5-ml portions of PBS. At this point, some samples were used in PI-specific phospholipase C assays (see below). Bound [3H]folic acid was recovered by incubating the monolayers with 1.5 ml of acid saline for 30 s and rinsing the wells with 1 ml of PBS. The acid saline wash and rinse were pooled, mixed with 10 ml of liquid scintillation fluid and counted in a liquid scintillation counter. Nonspecific backgrounds were determined by omitting the first acid-saline wash and were subtracted out. The cellular FBP gene does not express detectable amounts of FBP.
To determine the sensitivity of FBP to PI-specific phospholipase C, ['Hlfolic acid-labeled monolayers were suspended by scraping and subjected to digestion with 30 units/ml PI-specific phospholipase C at 37 'C in the presence of leupeptin and phenylmethylsulfonyl fluoride essentially as described earlier (5). Both the cell pellets and supernatants obtained after digestion were collected and analyzed by liquid scintillation counting.
Southern Blotting-Isolation of genomic DNA, restriction enzyme digestion, agarose gel electrophoresis, transfer to Pall A membranes, and blotting were performed as described earlier (8).

FBP cDNA Is
Stably Expressed in Normal CHO Transfectants but Not in Lec35.1 Transfectants-In order to test for GPI anchor biosynthesis in the Lec35.1 mutant, we chose to express the human FBP (4) by ligating FBP cDNA into the expression vector pJB20. The noteworthy features of this vector include a cytomegalovirus promotor/enhancer to drive expression of the cloned cDNA (6), a SV40 small T-antigen intron t o enhance RNA expression, a marker for G418 resistance, and a polylinker including an EcoRI site. The resulting * P. Stanley, personal communication.
construct, pF53, was transfected into normal or Lec35.1 CHO cells, and individual transfected colonies were isolated by selection for G418 resistance to ensure that each contained the transfected cDNA. Well resolved transfectants were picked and in some cases subcloned by limiting dilution. Throughout these studies, transfectants were periodically rechecked both for G418 resistance to ensure that they contained the plasmid and for concanavalin A/swainsonine resistance (1) to ensure that they had the expected phenotype. Each transfectant was then assayed for binding of [3H]folic acid as described under "Experimental Procedures." Fig. 1 shows the results obtained with normal and Lec35.1 cells transfected with pF53. Two observations can be made. First, the expression levels among normal transfectants varied greatly, over a range of 30-fold. In addition to the three normal transfectants shown, 10 others were assayed, and all were found to have FBP activities within this range. This variation of expression was not surprising; in our hands, stable expression of N-acetylglucosamine-1-phosphate transferase subcloned into pJB20 and transfected in CHO cells varied 40fold (9). Second, the expression of FBP activity in all Lec35.1 transfectants was much lower than in normal cells and was generally insignificant (less than 2% of the positive control value), as compared with the background binding in normal and Lec35.1 vector-transfected controls. This includes the eight Lec35-F53 transfectants shown in Fig. 1 and six others (data not shown).
To determine whether or not the absence of FBP activity in Lec35.1 transfectants was due to the phenotypic effects of the Lec35 mutation, presumably by preventing GPI lipid biosynthesis, several FBP-transfected Lec35.1 lines were fused to normal hygromycin-resistant CHO cells (followed by selection with G418 and hygromycin) to correct the Lec35 defect. It has been shown that fusion of Lec35 cells with normal CHO cells efficiently corrects the Lec35 mutation by complementation (2, 3). As shown in Table I  to normal cells G418-resistant transfectants (pJB2O or pF53) were hybridized to hygromycin-resistant CHO cells as described previously (2). Hybrids were plated in duplicate wells of a 24-well plate and grown to confluence. Binding of [3H]folic acid was determined as described under "Experimental Procedures." The averages of duplicate determinations are given. The 100% value was 24,000 cpm, and the limit of sensitivity of these assays was estimated to be 0.5% of this value.

Transfectant hybridized Relative
FBP activity a One transfectant, Lec35-F53J, was unusual in that it initially had a low level of activity (not shown) and give rise to a hybrid that also had measureable activity (shown above). After additional time in culture, Lec35-F53J lost its activity (see Fig. 1). Therefore, Lec35-F53F appears to have been a Lec35 transfectant with an unusually slow rate of FBP inactivation.
cells. This result suggested that the absence of FBP activity in Lec35.1 transfectants was not due to a post-translational defect affecting the synthesis of GPI anchor structures. In fact, as will be shown in the next section, under the appropriate conditions, transfected Lec35.1 cells were able to express PI-specific phospholipase C-sensitive FBP.

Lec35.1 FBP Transfectants Express FBPActiuity Soon after
Transfection-The results of the previous section could be explained by one of two possibilities. It was possible that during the transfection process and chromosomal integration, the FBP cDNA was destroyed in each of the Lec35.1 transfectants. Alternatively, the FBP cDNA in Lec35.1 transfectants could have been initially active after transfection, followed by a loss of activity later on.
To determine whether the FBP cDNA in Lec35.1 cells was functional at early times after transfection, pooled colonies of stable Lec35.1 transfectants were assayed 3 weeks after transfection. In the experiments described in Fig. 1, individual transfected colonies were analyzed, and in general, they were not assayed until at least 2 months after transfection. As indicated in Table 11, the Lec35.1 transfectant pool expressed significant amounts of FBP activity 3 weeks after transfection, although the activity diminished with time. The FBP activity was susceptible to PI-specific phospholipase C digestion, indicating the presence of a GPI anchor structure. Thus, as indicated in the accompanying paper (18), the Lec35.1 allele was sufficiently leaky to allow production of some GPIanchored FBP. The basis for the fraction of FBP that was releasable without PI-specific phospholipase C (about 18% from normal transfectants, 33-42% from the Lec35.1 transfectant pool) is not clear, but may represent FBP that failed to acquire a GPI anchor and was therefore easily displaced from the cell surface. Table I1 and other similar experiments (data not shown), we estimated that the half-time for inactivation was somewhere between 1 and 2 weeks. Unfortunately, it was not feasible to obtain a more accurate rate because of several sources of error. For example, it must be assumed that the relative fractions of individual transfectants that make up the pool fluctuate with time. We also found that the FBP With one exception (Lec35-F530), the eight Lec35-F53 transfectants lacked an intact copy of FBP cDNA and, in some cases (including Lec35-F530), had additional crosshybridizing fragments. Thus, it appears that in each Lec35.1 transfectant, the FBP cDNA has undergone rearrangement. in

Lec35.1 Cells
It is possible that the apparently intact copy in Lec35-F530 is actually an abnormal fragment that comigrates with the normal fragment. Interestingly, two pairs of Lec35.1 transfectants (Lec35-F53Al, C1; Lec35-F53J, M) had abnormal fragments of the same size, suggesting that "hot spots" for rearrangement may exist. Since each of these lines remained resistant to G418, they all retained some portion of the vector.
Compared with the transcripts in CHO-F53B2 and CHO-F53F, which were of normal size, RNA blot analysis of five Lec35.1 transfectants revealed either the absence of detectable FBP mRNA (Lec35-F53B, J, K) or an abnormally truncated mRNA (Lec35-F53A1, C1) (data not shown), giving further support to the conclusion that a structural alteration of the FBP cDNA had occurred.

DISCUSSION
This study was initiated to assess GPI anchor production in Lec35.1 cells. As shown in the accompanying paper (18), these cells are defective for MPD-dependent mannosyltransferase reactions of GPI lipid biosynthesis and dolichol-linked oligosaccharide biosynthesis. The initial experiments with cloned cDNA for FBP, a GPI-linked protein, appeared to support this conclusion because stable Lec35.1 transfectants did not express FBP activity. However, subsequent experiments showed that this was due to inactivation of the FBP cDNA. In fact, it was shown that prior to cDNA inactivation, the Lec35 mutation was leaky enough to allow GPI-anchored FBP to be produced in Lec35.1 cells.
It should be stressed that the choices of the pJB2O expression vector and FBP cDNA were based only on technical considerations, without regard to their potential for gene rearrangement. In cases involving normal CHO cells, with either human FBP (Fig. 1) or hamster GlcNAc-1-P transferase cDNA (9), pJB2O proved to be an effective expression vector with no indication of rearrangement. Additional studies will be required to determine which features of the vector and/or cDNA are required for the observed rearrangements in Lec35.1 cells. This would clearly include a survey of several different cDNAs and vectors. It will also be necessary to map the rearrangements and "reconstruct" the event, as has been done for gene deletions involving Alu-repetitive elements (8).
The obvious implication of these results is that caution must be employed when expression of proteins from transfected cDNA is used to assess post-transcriptional or posttranslational events in mutant cell lines. This also includes expression cloning strategies because it may not be possible to rescue an intact plasmid. Since it is not clear to what extent the mannosylation defect is responsible for gene inactivation, investigators studying glycosylation mutants should be particularly cautious.
The gene inactivation phenotype that has been described here is quite distinct from mutator phenotypes that have been described previously in CHO cells (10,ll). In these cases, the phenotypes were due to alterations in nucleotide synthesis (10) or DNA synthesis (ll), leading to mutations in the DNA. In these mutator strains, the mutation frequencies for marker genes generally increased lO-lOO-fold, as compared with normal cells (10,11), to reach mutation frequencies in the range of lo" (10). However, this does not compare with the mutation frequency for FBP cDNA in Lec35.1 cells because an absence of FBP activity was observed in every Lec35.1 transfectant.
The actual basis for the observed gene inactivation is unclear, and at least two possibilities should be explored. In one model, a MPD-dependent reaction may be required for maintenance of certain genes. Nuclear glycoproteins have been described (12), and thus it is possible that a nuclear protein required for repair synthesis of DNA must be mannosylated for function. The MPD could be synthesized in the endoplasmic reticulum and be transported to the nuclear envelope by contiguous membranes. Such a reaction would be prevented in a Lec35 mutant. Another model would require inactivation of a "primary" gene that is required for maintenance of a group of "secondary" genes, including the Lec35 gene and genes with sequences similar to those found in pF53.
It is worth noting that Lec35.1 was isolated by a gradual acquisition of phenotype (Fig. 3, Ref. 1) on a time scale that was similar to the inactivation of the FBP cDNA. Thus, the original selection that gave rise to the Lec35.1 cells may have indirectly selected for the loss of a primary maintenance gene. The effect of the gradual selection with diploid CHO cells would have been to allow the sequential mutation of both copies of the primary gene, followed by inactivation of the secondary genes. It should be possible to distinguish between these two models by repeating this study with Lec35.2 cells (18), which were not isolated by gradual selection, and Lecl5 cells, which fail to make MPD. Although attempts have been made, we have never been able to isolate a revertant of Lec35.1, so we were not able to determine if the Lec35.1 phenotype strictly correlated with gene inactivation.
One unexpected consequence of this work is to bring to light the similarity between the phenotype of the Lec35.1 mutant and the defect in the human disease paroxysmal nocturnal hemoglobinuria (PNH) (13). This similarity exists at both the biochemical and genetic levels. Biochemically, it has been shown that hematopoietic cells from individuals with PNH are deficient for GPI-anchored proteins. This includes erythrocytes, platelets, lymphocytes, and granulocytes, and clinical manifestations appear to result from the extreme sensitivity of the PNH red cells to complementmediated hemolysis, due to a deficiency of GPI-anchored decay accelerating factor (13). As a result, PNH patients suffer from anemia and frequent venous thrombolytic events and are generally expected to survive less than 10 years. Biochemical studies indicate that synthesis of GlcNAc-PI and GlcN-PI is normal in PNH polymorphonuclear cells (14,15), but that intact PNH granulocytes fail to make the mature GPI anchor precursor lipid, even though they appear to synthesize MPD (15,16). Thus, one or more of the mannosylation reactions required for the GPI anchor precursor are suspect. This biochemical phenotype is consistent with the one in Lec35 cells, in which the first mannosylation reaction is defective in viuo. Clearly, additional biochemical studies on PNH cells will be necessary to test this hypothesis. Since the Lec35 mutation may affect all three MPD-dependent mannosylation reactions in GPI lipid biosynthesis, it might be expected that cells from patients with differing extents of the disease would accumulate different GPI intermediates that vary by the number of mannose residues.
The genetic defect in Lec35.1 cells may also help to explain the unusual genetics of PNH. PNH is most likely caused by a somatic mutation in a clone of a totipotent stem cell, giving rise to abnormalities in all hematopoietic cells that arise from this clone (13). PNH patients have mixtures of normal (PNHI), moderately affected (PNHII), and severely affected (PNHIII) cells, which presumably arise from different stem cell clones. The proportions of these populations change with time. Interestingly, the PNHII population has been noted as particularly unstable, and appears to give rise to PNHIII cells (for example, see Fig. 31.6 in Ref. 13). This change in severity of the mutation is reminiscent of that observed in the Lec35.1 cells. Thus, PNHI, PNHII, and PNHIII clonal cells could differ by the extent of inactivation of the required gene (possibly the Lec35 homologue).