Protein Domains, Catalytic Activity, and Subcellular Distribution of Neuropathy Target Esterase in Mammalian Cells*

Neuropathy target esterase (NTE), the human homologue of a protein required for brain development in Drosophila, has a predicted amino-terminal transmembrane helix (TM), a putative regulatory (R) domain, and a hydrophobic catalytic (C) domain. Here we describe the expression, in COS cells, of green fluorescent pro-tein-tagged constructs of NTE and mutant proteins lacking the TM or the R- or C-domains. Esterase assays and Western blots of particulate and soluble fractions indicated that neither the TM nor R-domain is essential for NTE catalytic activity but that this activity requires membrane association to which the TM, R-, and C-do-mains all contribute. Experiments involving proteinase treatment revealed that most of the NTE molecule is exposed on the cytoplasmic face of membranes. In cells expressing a moderate level of NTE and all cells expressing (cid:1) C-NTE, fluorescence was distributed in an endoplasmic reticulum (ER)-like pattern. Cells expressing high levels of NTE showed aberrant distribution of ER marker proteins and accumulation of NTE on the cytoplasmic surface of ER-derived tubuloreticular aggre-gates. Deformation of the ER was also seen in cells expressing (cid:1) R-NTE or enzymatically inactive S966A-NTE but not (cid:1) TM-NTE. The data suggest that NTE is anchored in the ER via its TM, that its R- and C-domains also interact with the cytoplasmic face of the ER, and that overexpression allow recovery from trypsinization. In initial experiments, relative expression levels of fluorescent protein were determined using a BD Biosciences FACS Vantage system. Prior to electron microscopy, cells were sorted into pure GFP-expressing populations and were then spun down in a swing-out rotor at 3000 (cid:4) g for 30 min at 4 °C. The pellets were processed for routine electron microscopy and for post-embedding immunogold labeling as described previously (15). Disrupted cell fractions were also used for immunogold labeling as described by Johnston et al. (16).

Neuropathy target esterase (NTE) 1 is the human homologue of a protein required for brain development in Drosophila (1,2). mRNA encoding NTE is expressed in embryonic mouse neurons, suggesting the possibility of a similar function in mammals (3). NTE was identified originally in adult vertebrate neural tissue as a protein reactive with the organophosphates (OP), which cause a syndrome of axonal degeneration (4,5). Elucidation of the part played by NTE in OP-induced neuropathy and neural development is of obvious neurobiological interest. However, NTE is present not only in neurons but also in a variety of non-neural tissues including intestine, placenta, and lymphocytes (5,6). Furthermore, the existence of a putative homologue in Saccharomyces cerevisiae (7) suggests that NTE may be involved in a fundamental process common to cells from yeast to neurons.
NTE is conveniently detected in vitro by its ability to catalyze OP-sensitive hydrolysis of an artificial substrate, phenyl valerate (8). Using this assay, NTE has been shown to be firmly membrane-associated. Differential centrifugation of brain homogenates resulted in an enrichment of NTE in microsomal fractions containing elements of endoplasmic reticulum (ER), Golgi, and plasma membrane; attempts to further resolve NTE by density gradient centrifugation of microsomes were unsuccessful (9). In neural sections, immunoreactive NTE staining fills neuronal cell bodies, excluding the nucleus, and extends into the proximal axon; in addition, the rate of accumulation of NTE at a peripheral nerve ligature indicates that it is conveyed along axons by vesicular fast transport (10). These observations are consistent with an ER/Golgi location for NTE.
In keeping with the membrane-bound character of NTE, hydropathy analysis of its primary sequence predicts a transmembrane helix (TM) near the amino terminus (residues 10 -32). Further examination of the sequence of the 1327 residues of NTE indicates two functional domains: 1) an amino-terminal putative regulatory domain of ϳ700 residues that includes areas with similarity to cyclic AMP-binding proteins; and 2) a carboxyl-terminal catalytic/esterase domain containing the active site serine residue (Ser-966), which reacts with OPs (7). Various carboxyl-terminal constructs of NTE have been expressed in Escherichia coli to define the minimum polypeptide with OP-sensitive phenyl valerate esterase activity. A hydrophobic recombinant protein called NEST (NTE amino acids 727-1216) had this property and, although lacking the aminoterminal TM of NTE, associated firmly with phospholipid membranes and required this association for its esterase activity (11,12). In the present study, we expressed various green fluorescent protein (GFP)-tagged constructs of NTE in COS cells to relate the protein's molecular features to its enzymatic activity and intracellular distribution in eukaryotic cells.

EXPERIMENTAL PROCEDURES
Cells, Antibodies, and Other Products-COS-7 cells were obtained from the European Collection of Animal Cell Culture (ECACC number 87021302). Antibodies to GFP and to calnexin were purchased from Zymed Laboratories Inc. and StressGen, respectively. TRITC and peroxidase conjugates of anti-rabbit IgG were from Sigma, and colloidal gold (10 nm)-conjugated anti-mouse IgG was from British Biocell International Ltd. [ 35 S]methionine (43.5 TBq/mmol) was from PerkinElmer Life Sciences.
DNA Cloning and Mutagenesis-For generation of a full-length NTE carrying the GFP tag at the C terminus, two primers were designed to bring the human NTE cDNA in-frame into the pEGFP-N1 vector (Clontech) between SalI and BamHI sites (forward primer, 5Ј-AGATCGGTC-GACCAGCTGGAATC-3Ј; reverse primer, 5Ј-TGTCGAGGATCCCAG-GCATCTGT-3Ј). A 4-kb PCR product was generated from the human * This work was supported by the Medical Research Council. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  NTE clone D16 (7) using Pfu DNA polymerase (Stratagene) and cloned into pEGFP-N1 to produce pNTE-GFP (Fig. 1).
To generate a construct deleting the first 42 amino acid residues, which include the TM at the N terminus, we designed a forward primer (5Ј-TGCCAAGAATTCCAGCCATGGATGGCCCCC-3Ј) to bring in a Kozak consensus, an EcoRI site, and a translation start codon. This primer was paired with a reversed primer (5Ј-TGTCGAGGATCCCAG-GCATCTGTGGCTGAG-3Ј) with a BamHI site to remove the original stop codon to amplify a 4-kb PCR fragment from NTE clone D16. The PCR fragment was cloned into the EcoRI/BamHI sites of vector pEGFP-N1.
To construct an expression vector for the catalytic domain of human NTE lacking the first 680 residues of the full-length protein, a forward primer (5Ј-ATAGCCAAGCTTCCCGAGGCCGCCATGGGT-3Ј) was designed to make use of an internal HindIII site of human NTE cDNA and bring in a Kozak consensus for optimal transcription. This forward primer, with a reverse primer (5Ј-TGTCGAGGATCCCAGGCATCTGT-GGCTGAG-3Ј), was used in a PCR reaction to amplify a 2-kb DNA fragment, which was cloned into the HindIII and BamHI sites of the vector pEGFP-N1 to make pNEST-GFP (Fig. 1).
To generate a construct of NTE-GFP (designated ⌬R-NTE; Fig. 1) which retained the first 157 amino acids (including the predicted TM) but lacked residues 158 -673 (i.e. the putative regulatory domain), we designed two primers (forward, 5Ј-AAGCTTGCTAGCGAATTC-CCCCACGC-3Ј; reverse, 5Ј-GCGGCAAAGCTTCAGGAAGAGTGGCT-TCT-3Ј) to amplify a short fragment of DNA from human NTE clone D16 containing the 5Ј untranslated region, the translation start codon, and the TM. This PCR product was cut with NheI and HindIII and ligated into the long arm of pNTE-GFP pre-cut with NheI and HindIII and recovered from an agarose gel.
To generate a construct lacking the catalytic domain (⌬C-NTE-GFP; Fig. 1) the appropriate region of human D16 NTE cDNA was amplified by PCR with the same forward primer used to make ⌬R-NTE (above) and a new reverse primer (5Ј-TTGATGGGATCCAAGGTGCCCTCGG-GAA-3Ј). The PCR product was cut with NheI and BamHI and cloned into the vector backbone of the construct pNTE-GFP to replace the full-length NTE-encoding sequence.
Mutation of the active site serine (Ser-966) to alanine by site-directed mutagenesis was achieved by using QuikChange TM site-directed mutagenesis kit (Stratagene) and the primers 5Ј-GTGGGCGGCACGGC-CATTGGCTCTT-3Ј and 5Ј-AAGAGCCAATGGCCGTGCCGCCCAC-3Ј. The mutation was verified by DNA sequencing and enzymatic activity assay.
Cell Culture and Fluorescent Microscopy-COS cells were cultured in Dulbecco's modified Eagle's medium with Glutamax-I (Invitrogen), 10% fetal bovine serum, and 2% of both penicillin and streptomycin at 37°C, 5% CO 2 . Transfection was carried out using PolyFect Transfection Reagent (Qiagen) according to the manufacturer's protocol. For fluorescent microscopic study, cells were plated at 4 ϫ 10 4 /well in an 8-well Lab-Tek II chamber slide (Nalge Nunc International) and transfected with various constructs after 24 h of culture. 24 -48 h after transfection, cells were fixed with 2% paraformaldehyde in phosphate-buffered saline for 10 min at room temperature and permeabilized by methanol/ acetone (50:50) at Ϫ20°C for 10 min. After three washes with phosphate-buffered saline, the cells were blocked with 3% bovine serum albumin in phosphate-buffered saline for 2 h at room temperature and probed with anti-calnexin (1:400) followed by TRITC-labeled anti-rabbit IgG (1:400) with three washes between and after antibody reaction. Slides were mounted with Vectorshield ® mounting medium (Vector Laboratories). Fluorescent images were acquired by confocal scanning using an argon-krypton laser and a Leica TCS-4D confocal imaging system. a Relative fluorescence levels were calculated as percentage of those of full-length NTE-GFP based on mean levels determined by FACS analysis.
b Particulate fractions were isolated from COS cells 48 h after transfection with the indicated constructs, and NTE (phenyl valerate) esterase activities were determined as described under "Experimental Procedures." Data are the mean and standard deviation of three separate experiments. Subcellular Fractionation, Esterase Assay, and Western Blotting-COS cells were plated at 0.8 ϫ 10 6 in 10-cm dishes and cultured for 24 h before transfection. Forty-eight hours after transfection, cells were harvested by trypsinization. The cell pellet was resuspended in TE buffer (50 mM Tris-HCl, 1 mM EDTA, pH 8.0), homogenized with 10 passages through a 25-gauge hypodermic needle, and centrifuged at 100 ϫ g at 4°C for 2 min. The supernatant fraction was further centrifuged at 100,000 ϫ g at 4°C for 45 min in an Optima TM TLX ultracentrifuge using a TLA120 rotor (Beckman). After removing the soluble cytosolic fraction, the particulate fraction was washed once by resuspension and centrifugation and finally resuspended in the original volume of TE buffer. Protein concentration was measured using a Bio-Rad protein assay kit (Bio-Rad Laboratories). NTE (phenyl valerate) esterase assay was carried out as described previously (11). Soluble and particulate fractions were run on SDS-PAGE (4 -20% gradient gels), blotted, and probed with anti-GFP antiserum (1:1000) followed by peroxidase-labeled anti-rabbit IgG (1:1000) with final detection by enhanced chemiluminescence using Pierce reagents essentially as described previously (11). Relative levels of immunoreactive GFP in soluble and particulate fractions were determined by densitometry of the Western blots.
Proteinase K Digestion of Sealed Membrane Vesicles Isolated from COS Cells Expressing NTE-GFP-Cells transfected with NTE-GFP for 48 h were harvested and homogenized in hypotonic TE buffer, and the post-nuclear supernatant was subjected to 100,000 ϫ g ultracentrifugation as above. The membrane pellet was resuspended in the same volume of TE buffer containing 0.25 M sucrose and re-homogenized and used as sealed membrane vesicles. For proteinase K digestion, an aliquot of the membrane vesicles was incubated on ice for 30 min with proteinase K (4 g/ml) with or without 1% Triton X-100. The digestion was terminated with 4 mM phenylmethanesulfonyl fluoride (PMSF) and a 30-min incubation on ice to allow complete inhibition of proteinase K before SDS-PAGE and Western blotting with anti-GFP. As a control, the same blots were probed with an antibody (StressGen, catalog number SPA-865; 1:2500) directed to the N terminus of calnexin, an ER protein; this procedure has been reported to detect a 70-kDa intralumenal fragment of calnexin (13).
In Vitro Transcription/Translation and Protection from Proteinase K Digestion by Pancreatic Microsomes-cDNA encoding NTE was subjected to coupled transcription/translation in vitro at 30°C for 75 min using [ 35 S]methionine and a STP3 kit from Novagen, according to the manufacturer's instructions, in the absence or presence of canine pancreatic microsomes (Promega; 1 l). Subsequently, proteinase K (0.2 g) was added, and mixtures were incubated for 30 min at 4°C. Reactions were stopped with 2 mM PMSF. After incubation for a further 20 min, samples were run on SDS-PAGE and blotted onto a nitrocellulose membrane, which was then subject to autoradiography. As a control, parallel reactions were run using cDNA encoding MADM, a mammalian protein cloned in this laboratory (14), which (unlike NTE) has an amino-terminal signal peptide and a carboxyl-terminal TM.
Relative Fluorescence Analysis, Cell Sorting, and Electron Microscopy-Transfected COS cells were harvested by trypsinization and resuspended in culture medium containing 10% fetal bovine serum and incubated at 37°C for 30 min to allow recovery from trypsinization. In initial experiments, relative expression levels of fluorescent protein were determined using a BD Biosciences FACS Vantage system. Prior to electron microscopy, cells were sorted into pure GFP-expressing populations and were then spun down in a swing-out rotor at 3000 ϫ g for 30 min at 4°C. The pellets were processed for routine electron microscopy and for post-embedding immunogold labeling as described previously (15). Disrupted cell fractions were also used for immunogold labeling as described by Johnston et al. (16).

Amino-terminal TM Facilitates Membrane Association, but Neither TM nor the R-domain Is Essential for a Catalytically Active Conformation of NTE in COS Cells-Constructs of NTE
tagged at the carboxyl terminus with GFP, including the fulllength protein, deletion mutants with either the amino-terminal TM segment (⌬TM), the putative regulatory (⌬R) domain or the catalytic (⌬C) domain removed, and a truncated polypep-

FIG. 3. Microsomal membranes do not protect NTE from proteinase K digestion.
A, sealed membrane vesicles were isolated from COS cells transfected with NTE-GFP and treated with proteinase K (PK) as detailed under "Experimental Procedures." As a control, the ER protein calnexin was probed with an antibody to its intralumenal N terminus. The C terminus of calnexin is cytoplasmic and therefore cleaved by proteinase K, reducing its molecular mass from 90 to 70 kDa (see Ref. 13). Calnexin is completely degraded by proteinase K in the present of Triton X-100 (TX). By contrast, the membrane vesicles did not protect NTE-GFP from proteinase K digestion. B, cDNAs encoding either NTE or MADM (a lumenal protein used here as control) were transcribed and translated in the presence of pancreatic microsomes (MS) and then exposed to proteinase K (PK) and analyzed by SDS-PAGE and autoradiography as described under "Experimental Procedures." The presence of microsomal membrane partially protects lumenal MADM, but not NTE, from degradation by proteinase K. tide of similar length to bacterially expressed NEST (11) were made as shown in Fig. 1. FACS analysis revealed that, in the transfected cells, the relative mean levels of fluorescent protein expression were quite similar; mean values for cells transfected with NTE-GFP were about 2-fold higher than those with NEST-GFP, whereas cells transfected with the ⌬TMand ⌬Rconstructs showed intermediate values (Table I).
Phenyl valerate esterase assays on soluble and particulate fractions from the cells transfected with the various constructs indicated that all the activity was confined to the latter fraction (data not shown); this may reflect the fact that association with phospholipids is required for NTE esterase activity (11,12). Esterase activities in particulate fractions from NTE-GFP-and ⌬R-NTE-GFP-transfected cells were roughly proportional to the relative mean fluorescence intensity of the cells (Table I). The fact that ⌬R-NTE has essentially identical catalytic activity as the full-length protein indicates that the R-domain is neither required for, nor does it substantially inhibit, NTE phenyl valerate esterase activity. Sequence homology within the R-domain with proteins that bind cyclic AMP (7) suggests that this nucleotide might modulate NTE catalytic activity. However, we have found no effect of cyclic AMP on phenyl valerate esterase activity in particulate fractions from NTEtransfected COS cells (data not shown).
Mean fluorescence levels in NEST-GFP-transfected COS cell populations were about half of those in NTE-GFP transfected

FIG. 4. Distribution of various GFPtagged NTE protein constructs and calnexin in transfected COS cells.
Cells were transfected with GFP vector or various NTE-GFP constructs as indicated on each panel image for 48 h and then fixed, permeabilized, immunostained for the ER-marker calnexin (red), and visualized by confocal microscopy as described under "Experimental Procedures." Note the marked differences in calnexin distribution in cells expressing ⌬R-NTE-GFP or high levels of NTE-GFP compared with the relatively normal calnexin staining in cells expressing other deletion constructs of NTE.
cells, but esterase activity in particulate fractions from NEST-GFP-expressing cells was only ϳ5% of those from cells expressing the full-length protein (Table I). However, Western blotting of soluble and particulate fractions indicated that, whereas ϳ90% of immunoreactive GFP in NTE-GFP-transfected cells was confined to the latter fraction, only ϳ30% of GFP in NEST-GFP-transfected cells was particulate (Fig. 2). Thus normalized for the amount of NEST present, the esterase activity in particulate fractions of NEST-transfected COS cells is actually about one-third of that in NTE-transfected particulates. Similarly, ϳ60% of immunoreactive GFP was present in the particulate fractions of cells transfected with the ⌬TM construct (Fig.  2) and, thus, when data in Table I are normalized on this basis, the esterase activity in particulate fractions from ⌬TM-NTE-GFP-expressing cells is about half of that in NTE-GFP-transfected cells.
When expressed in E. coli, NEST is able to fold to a catalytically active conformation in association with the bacterial membrane (11). The present results suggest that recombinant constructs of NTE lacking the N-terminal TM can also adopt a catalytically active conformation once they become associated with membranes in eukaryotic cells; thus, the major function of the TM is to facilitate more efficient association with membrane. Interestingly, hydropathy analysis predicts at least one TM near the amino terminus of all the eukaryotic NTE homologues but none in YCHK, a 34-kDa E. coli protein with homology to a ϳ200 residue region surrounding the active site of NTE (7).

Most of the NTE Molecule Is Exposed on the Cytoplasmic
Surface of Intracellular Membranes-Because NTE lacks a signal sequence, it is likely to associate with intracellular membranes with the great majority of the molecule exposed to the cytoplasm. To confirm this possibility, we showed that proteinase K treatment (either in the absence or presence of Triton X-100) of sealed membrane vesicles from NTE-GFP-transfected COS cells reduced the size of the GFP-immunoreactive polypeptide from ϳ180 kDa (NTE-GFP) to ϳ30 kDa (similar to GFP itself), indicating that the membranes did not protect NTE from proteolysis (Fig. 3A). By contrast, a 70-kDa fragment of calnexin (i.e. its intralumenal domain; see Ref. 13) was clearly protected until the membrane vesicles were disrupted by treatment with Triton X-100 (Fig. 3A). In a second approach to the question of the topology of NTE with intracellular membranes, we subjected NTE cDNA to in vitro transcription and translation in the absence or presence of pancreatic microsomes. The resulting polypeptide was not protected by microsomes from degradation by added proteinase K, suggesting a predominantly cytoplasmic disposition (Fig. 3B). By contrast, in a parallel control experiment, the polypeptide formed by transcription/translation of MADM (a mammalian membrane protein with an amino-terminal signal sequence TM; Ref. 13), which is imported into the lumen of the microsomes, was protected from proteinase K digestion (Fig. 3B).
NTE Has an ER-like Distribution, and Its Overexpression Disrupts the ER via a Non-enzymatic Property of the Cytoplasmic C-domain-In COS cells expressing either NEST-GFP or ⌬TM-NTE-GFP, fluorescence was distributed in a pattern distinct from that in cells expressing GFP itself and was partially cytoplasmic and partially coincided with that of the ER marker calnexin (Fig. 4). This pattern is consistent with the soluble/ particulate distribution of these recombinant proteins detected by Western blotting and supports the view that both the C-and R-domains of NTE contribute to its association with the cytoplasmic face of the ER membrane.
An ER-like localization was observed for fluorescence in essentially all cells expressing ⌬C-NTE-GFP but in only a minority of cells expressing full-length NTE-GFP (Fig. 4). In a majority of COS cells expressing NTE-GFP, intense fluorescence was observed in the juxtanuclear area, and those expressing ⌬R-NTE-GFP commonly showed even more intense and bizarre patterns of fluorescence (Fig. 4). Calnexin appeared to colocalize with juxtanuclear areas of intense expression of NTE-GFP or ⌬R-NTE-GFP in a pattern that was clearly abnormal (Fig.  4). Although some cells expressing ⌬TM-NTE-GFP also showed relatively intense juxtanuclear fluorescence, calnexin distribution in these cells was relatively normal (Fig. 4). Similar morphological observations, (and those relating to esterase activity) were also made with transfected human HeLa and N2a mouse neuroblastoma cells using protein disulfide isomerase rather than calnexin as an ER marker (data not shown). Furthermore, the intense juxtanuclear fluorescence was not readily dispersed by treating the cells with brefeldin A (data not shown), an indication that NTE-GFP was not associated with the Golgi apparatus (16).
The ultrastructure of COS cells with intense juxtanuclear expression of NTE-GFP (Fig. 5b) was compared with that of GFP-vector control cells (Fig. 5a) to investigate reasons for the abnormal distribution of calnexin. Cells expressing high levels of NTE-GFP exhibited fine tubular structures (20 -40 nm in diameter) contiguous with the ER, which was often distended to result in vesicles containing fine flocculent material (Fig.  5b). In many of these cells the tubular structures were aggregated in the cytoplasm to form irregular clusters up to 7 m in diameter. Immunogold labeling of resin sections demonstrated the localization of NTE-GFP to these membrane clusters (Fig.  5c). Greater resolution was obtained by immunogold labeling of disrupted cell fractions (16), which clearly revealed the presence of NTE-GFP on the cytoplasmic face of these membrane clusters, whereas adjacent Golgi, mitochondrial, lysosomal, and nuclear membranes were not labeled (Fig. 5d). Electron microscopy of cells expressing ⌬R-NTE-GFP revealed even more markedly abnormal membrane structures than those observed with the full-length protein. These tubuloreticular membrane structures were up to 3 m in diameter and contiguous with the ER (Fig. 5, e and f).
Recently, we demonstrated that purified recombinant NEST catalyzes hydrolysis of membrane lipids in vitro (17). We wondered whether the abnormal membrane structures observed in COS cells overexpressing NTE and ⌬R-NTE, but not in those with ⌬C-NTE, might reflect hydrolysis of membrane lipid. However, the same aberrant membrane clusters were observed in cells expressing the enzymatically inactive S966A mutant forms of both NTE-GFP and ⌬R-NTE-GFP (data not shown). Thus, disruption of the ER and formation of the tubuloreticular structures appear to reflect a non-enzymatic property of the catalytic domain of overexpressed NTE.
The abnormal ER morphology induced by overexpression of NTE is reminiscent of that described in COS cells overexpressing the inositol 1,4,5-triphosphate receptor (18), malfolded cytochrome P450 (19), and microsomal aldehyde dehydrogenase (20). All of these proteins are anchored in the ER membrane by at least one TM and have large cytoplasmic domains. It has been suggested that the ER membranes aggregate by the headto-head association of the cytoplasmic domains of these proteins (20). Our present experiments showing that microsomal membranes do not protect NTE from proteinase K digestion (Fig. 3) indicate that the protein is probably anchored in the ER membrane via its amino-terminal TM with residues 33-1327 exposed to the cytoplasm. This topology is also consistent with the pattern of immunogold labeling of disrupted cell fractions (Fig. 5d). When NTE is overexpressed, intermolecular association of the cytoplasmic hydrophobic C-domains could give rise to ER aggregation by the general mechanism noted above. The fact that overexpression of ⌬TM-NTE-GFP does not cause gross redistribution of calnexin indicates that this ER aggregation requires NTE to be anchored via its TM. The leading role of the C-domain in the aggregation is emphasized by the relatively normal morphology of cells expressing ⌬C-NTE. To some degree, the R-domain may hinder intermolecular association of C-domains, and this may reflect the exacerbated ER abnormality in cells expressing ⌬R-NTE.