The Mammalian Gene of Acetylcholinesterase-associated Collagen*

The collagen-tailed or asymmetric forms (A) represent a major component of acetylcholinesterase (AChE) in the neuromuscular junction of higher vertebrates. They are hetero-oligomeric molecules, in which tetramers of catalytic subunits of type T (AChET) are attached to the subunits of a triple-stranded collagen “tail.” We report the cloning of a rat AChE-associated collagen subunit, Q. We show that collagen tails are encoded by a single gene, COLQ. The ColQ subunits form homotrimers and readily form collagen-tailed AChE, when coexpressed with rat AChET. We found that the same ColQ subunits are incorporated, in vivo, in asymmetric forms of both AChE and butyrylcholinesterase. A splice variant from the COLQ gene encodes a proline- rich AChE attachment domain without the collagen domain but does not represent the membrane anchor of the brain tetramer. The COLQ gene is expressed in cholinergic tissues, brain, muscle, and heart, and also in noncholinergic tissues such as lung and testis.

Acetylcholinesterase (AChE, EC 3.1.1.7) 1 is highly concentrated at vertebrate neuromuscular junctions. This enzyme is encoded by a single gene, and adult mammalian muscles express a single splice variant, corresponding to the catalytic subunit of type T (AChE T ) (1,2). At the post-translational level, however, quaternary interactions introduce a considerable diversity of molecular forms that are characterized by distinct localizations in cellular structures. These molecules include amphiphilic monomers (G 1 a ) and dimers (G 2 a ), nonamphiphilic tetramers (G 4 na ), as well as hetero-oligomeric structures in which tetramers of catalytic subunits are disulfide-linked with a hydrophobic "tail" (20 kDa) in the membrane-bound G 4 a forms (3,4) or with a collagenous "tail" in the collagen-tailed or asymmetric (A) forms. The latter molecules consist of one, two, or three tetramers (A 4 , A 8 , A 12 ), which are disulfide-linked to the strands of the triple helical collagen tail (see Fig. 1A). G 1 a and G 2 a forms appear to remain mostly intracellular and represent precursors of more complex molecules. The G 4 na form is secreted and hydrophobic-tailed tetramers (G 4 a ) are attached to the plasma membrane. The collagen-tailed molecules are tethered in the basal lamina, and are largely responsible for the high concentration of AChE at the neuromuscular junction.
To understand the biosynthesis of the various AChE forms and its regulation, it is necessary to analyze the association of AChE T catalytic subunits with anchoring subunits, particularly the collagen subunits, which have been named Q, according to the nomenclature of AChE-associated proteins (5). Cloning and expression of the collagen tail subunit of the asymmetric AChE forms from Torpedo electric organ (tQ 1 ) allowed us to show that the structural and catalytic subunits assemble into collagen-tailed molecules when coexpressed in COS cells (6). The primary sequence of the Q subunit comprises an N-terminal region, Q N , a collagen domain, and a C-terminal domain, Q C . We showed that the Q N domain is able to recruit monomers, producing a tailed tetramer (7,8) and defined a proline-rich attachment domain (PRAD) that is sufficient for interaction with AChE T (9).
In the present study, we cloned cDNAs encoding Q subunits from the rat, using cross-hybridization with Torpedo probes. We asked three major questions: Do several genes encode different Q subunits? Are collagenous and noncollagenous AChEassociated proteins, particularly the 20-kDa membrane anchor, generated by the same gene(s)? Is the expression of Q gene(s) exclusively restricted to cholinergic tissues? EXPERIMENTAL PROCEDURES Unless otherwise indicated, reagents were purchased from Prolabo (Paris, France), Sigma, or Appligene (Illkirch, France), and enzymes from New England Biolabs (Ozyme, France).
Preparation of Reinnervating Rat Soleus Muscle and Sternomastoid Muscle-Ten male Wistar albino rats weighing 180 -250 g at the time of surgery were anesthetized by intraperitoneal injection of a mixture of 5 mg/kg xylazine (Rompun, Bayer AG, Leverkusen, Germany) and 90 mg/kg ketamine (Ketalar, Parke-Davis and Co., Berlin, Germany). The sciatic nerve on one side was exposed and crushed in the mid thigh region by a nonserrated hemostat for 30 s. The wound skin was closed, and the animals were left to recover for 21 days to allow for sciatic nerve regeneration and reinnervation of the soleus muscle. The animals were then sacrificed, and the reinnervating soleus muscle was rapidly isolated and frozen in liquid nitrogen.
Preparation and Screening of a cDNA Library-Total RNA was extracted from muscle (10), and a cDNA library was constructed in pCDM8. Briefly, double-stranded cDNA was obtained (11), using a cDNA synthesizing kit (Pharmacia). Then, BstXI adaptors were linked to the blunt-ended cDNA, after purification by Sephacryl 400 chromatography, and the cDNAs were ligated in pCDM8, which had previously been cleaved by BstXI.
1.5 ϫ 10 6 independent transformed MC1061/P 3 bacteria were distributed in pools of 10,000 -12,000 clones and amplified. 60 pools were grown, and for each pool the DNA was extracted, digested by XhoI, run on agarose gels, and blotted to nylon membranes that were hybridized with probes derived from the Torpedo cDNA (nucleotides 1-213 and 916 -1416) (6). The membranes were incubated in 7% SDS, 0.5 M sodium phosphate, pH 7.2, 2 mM EDTA at 50°C with the radiolabeled probes and washed in 1% SDS, 0.2 M sodium phosphate, and 2 mM EDTA at 45°C. One pool was positive with the two probes. Sequential dilutions were used to isolate a positive clone, rQ 1 . A second screening of this library with a probe derived from the rQ 1 clone yielded five other clones with different structures, which will not be described in detail here.
Sequences were determined by the dideoxynucleotide method with the Sequenase kit (U. S. Biochemical Corp.) and analyzed by the Genework program (IntelliGenetics). Comparisons of protein and nucleic acid sequences with data banks were done with the FASTA and BLAST programs.
Analysis of mRNA by Northern Blots-We used the rat multiple tissue Northern blot from CLONTECH, prepared with 2 g of poly(A) ϩ RNA from heart, brain, spleen, lung, liver, skeletal muscle, kidney, and testis. The N-terminal probe (nucleotides 5-412 of rQ 1 cDNA) was obtained by polymerase chain reaction with oligonucleotides 5Ј-CCTT-GCCCTCTGACTTGATA-3Ј and 5Ј-GGGGGCATCAGTAGGCAGCA-3Ј, the C-terminal probe (nucleotides 892-1412) with oligonucleotides 5Ј-GAAAGAGGATTTCCAGGGCC-3Ј and 5Ј-TATCGGCAGGGTGTGG-AGTC-3Ј. Radiolabeled probes of high specific activity were generated by random priming with the Rediprime kit (Amersham Corp.). Hybridization was performed in 5 ml of rapid hybridization buffer (CLON-TECH) for 2 h at 55°C. Washings were done in 1 ϫ SSC at 50°C. The same membrane was first hybridized with a C-terminal probe and exposed, then boiled in 2 mM EDTA and exposed again to verify the completeness of dehybridization. It was finally hybridized with the N-terminal probe. Blots were analyzed with a PhosphorImager (Molecular Dynamics) after 48 h of exposure.
Analysis of Genomic DNA-Rat genomic DNA was extracted by the salting out method (12). For Southern blotting, genomic DNA was digested by restriction enzyme, transferred to nylon membranes, and hybridized as described for screening of the cDNA library.
The mouse 129 cosmid library (Stratagene) was screened with Nterminal (5-412) and C-terminal (892-1412) probes derived from the rQ 1 cDNA. We isolated one clone that hybridized only with the N probe and contained part of the Q gene. All EcoRI fragments from this cosmid were subcloned and sequenced at their extremities. The fragments were assembled by sequencing through EcoRI site, directly on the cosmid. Exons were localized by hybridization with cDNA probes of EcoRI fragment digested with restriction enzymes or deleted by exonuclease III (13). Some junctions were confirmed by sequencing the cosmid with specific oligonucleotides.
RNase Protection Assays-RNA probes were generated by Sp6 or T7 RNA polymerases with [␣-32 P]UTP (800 Ci/mmol; Amersham). Q 1 probe contains the sequence 145-412 of the rQ 1 cDNA, Q R probe 145-321. The probes were gel-purified and eluted in 2 mM EDTA, 0.5% SDS at 37°C. Hybridization of RNA with the radioactive probes, RNase digestion, and inactivation were performed with the Hybspeed TM RNase protection assay kit (Ambion). The protected fragments were analyzed after electrophoresis in a Fuji image analyzer (BAS 1000). The content of A nucleotides was taken into account for calculation of the relative abundance of the fragments.
Preparation of Antiserum against the Q N 35-51 Peptide-A peptide corresponding to residues 35-51 of the rQ 1 deduced primary sequence was synthesized by Joel Vandekerckhove, coupled to keyhole limpet hemagglutinin, and injected into hens (Biocytex, France). IgY antibodies were purified from egg yolks by Biocytex.
Expression in COS Cells-Transfection of COS cells was performed by the DEAE-dextran method, as described previously (7), using 5 g of DNA encoding the rat catalytic subunit AChE T and 5 g of DNA encoding rQ 1 or rQ R . Cells were cultured at 37°C and analyzed 2-4 days after transfection (14).
Expression in Xenopus Oocytes-We constructed plasmids in which the AChE T and rQ 1 coding sequences were inserted between 5Ј-and 3Ј-untranslated sequences of Xenopus globin, in the T S T 7 vector (15), so as to enhance stability and translation efficiency. Synthetic transcripts were prepared with the Ambion mMESSAGE mMACHINE TM in vitro transcription kit. Samples of ϳ50 nl (25 ng of AChE T and 5 ng of Q mRNA) were injected with an air pressure injection system (Injectϩmatic, Geneva, Switzerland) into Xenopus oocytes (16). Analysis of the molecular forms of AChE was performed 1-3 days after injection.
Extraction and Purification of AChE-AChE was extracted from transfected COS cells and Xenopus oocytes in 10 mM Tris-HCl, pH 7, 1 M NaCl, 1% Triton X-100. In the case of rat heart, we performed sequential extraction by a low salt buffer without detergent, a low salt buffer in the presence of 1% Triton X-100 and finally a high salt solution containing 1 M NaCl, as described previously (17). AChE purifications were performed from frozen bovine caudate nucleus and Xenopus oocytes essentially as described previously (3) using an trimethylammonium m-phenylenediamine affinity column (18). All steps were performed at 4°C in presence of 20 mM EDTA and 2 mM benzamidine hydrochloride as protease inhibitors. The activity was assayed by the colorimetric method of Ellman et al. (19).

kDa).
Western Blots-After electrophoresis, proteins from acrylamide gels were electrophoretically transferred to polyvinylidene difluoride membranes (Westran, Schleicher & Schuell, Dassel, Germany) in a tank-blot system (XCell II, Novex, San Diego, CA). Blotting was performed overnight at 20 V in a borate buffer containing 50 mM sodium borate (pH 9.0) and 20% (v/v) methanol. After transfer, polyvinylidene difluoride membranes were rinsed three times with 50 mM Tris/HCl, pH 7.4, 0.15 M NaCl. For blocking, a solution of 20 mM Tris/HCl, pH 7.4, 0.15 M NaCl (buffer A) and 3% bovine serum albumin was added to the blot, which was incubated for 4 h at room temperature. It was then incubated overnight at 4°C with 1/200 diluted antiserum in buffer B (buffer A plus 0.05% Tween 20). The membranes were rinsed four times with buffer B and incubated for 2 h at room temperature with horseradish peroxidase-conjugated rabbit anti-mouse or anti-hen immunoglobulins diluted 1:1000 with buffer B. The strips were washed five times with buffer B and developed with a solution of 0.05% of diaminobenzidine in buffer A containing 1 l/ml of 8.
Sedimentation Analysis in Sucrose Gradients-Sedimentation analysis of AChE forms in sucrose gradients was performed as described previously (7); the 5-20% sucrose gradients contained 0.4 M NaCl and 0.2% Brij-96; they were centrifuged at 35,000 rpm, at 7°C for 16 h, in a SW41 Beckman rotor. AChE was assayed by the colorimetric method of Ellman et al. (19). The sedimentation coefficients were deduced by a linear relationship from the position of internal markers, alkaline phosphatase (6.1 S) and ␤-galactosidase (16 S) from E. coli. For assays of AChE and BChE, AChE activity was specifically inhibited with 10 Ϫ6 M BW284C51, and BChE activity was inhibited with 10 Ϫ4 M tetraisopropyl pyrophosphoramide (21).

Cloning of the Rat Collagen Subunit (rQ): A Modular
Structure-To increase the level of transcripts encoding the collagen tail, we used a rat soleus muscle which was in the process of reinnervation after a crush of the sciatic nerve; in this case, AChE collagen-tailed forms become highly expressed throughout the entire length of muscle fibers (22). We constructed a cDNA library containing 1.5 ϫ 10 6 independent clones from RNA isolated from the reinnervating soleus and screened it with probes corresponding to the Q N and Q C regions of the Torpedo tQ 1 subunit. We thus obtained one cDNA clone, rQ 1 , which cross-hybridized at low stringency with the two probes. The length of rQ 1 was 2,731 nucleotides. It contains 45 nucleotides of the 5Ј-untranslated region, and 1312 nucleotides of the 3Ј-untranslated region. The coding sequence (46 -1419) is entirely homologous to that of Torpedo, tQ 1 (6); rQ 1 and tQ 1 presented 60% identity at the nucleotide level and 52% identity at the amino acid level.
An alignment of the deduced peptide sequences shows that nonconserved regions alternate with remarkably well conserved regions, which define potentially functional domains (Fig. 1B). The N-terminal part of the sequence corresponds to a secretion signal peptide, which is predicted to contain 22 residues in rQ 1 , and 42 residues in tQ 1 (23). The predicted cleavage sites are located at the same position in the aligned rQ 1 and tQ 1 sequences, and this position was recently confirmed in the case of tQ 1 (8). The N-terminal extremity of the mature protein shows only 10 conserved residues out of 28, but it is followed by a very conserved PRAD, containing 15/17 identical residues. The PRAD sequence includes two adjacent cysteines that establish disulfide bonds with AChE T subunits (24) and stretches of five and three prolines. This peptidic domain is sufficient for binding an AChE T tetramer (9). It is separated from the collagen domain by 30 residues, showing essentially no conservation between tQ 1 and rQ 1 , except for the presence of a cysteine residue, which may participate in the stabilization of the triple helix through intersubunit disulfide bonds (tQ 1 contains two cysteines in this region, only one of which is conserved in rQ 1 ).
The collagen domain has approximately the same length in both sequences; it consists of 63 triplets of amino acids GXY in rat, compared with 60 in Torpedo, with an interruption of 10 residues in tQ 1 , 7 in rQ 1 , at the same position in both sequences. Two internal motifs, which have been proposed to represent binding sites for heparan sulfate proteoglycans in tQ 1 (25), are totally conserved, RKGR (128 -131) and KRGK (234 -237).
The C-terminal noncollagenous domain contains three conserved regions of unequal length. The first conserved region consists of 14 residues, beginning immediately at the end of the collagen domain; it contains two conserved cysteines, separated by a single residue, that are probably involved in interchain bonds. The second conserved region contains 31 identical residues, and 9 conservative substitutions, out of 45 residues. It is highly hydrophilic, with a large proportion of negatively and positively charged residues, and shows no similarity with known proteins. Separated from this domain by a short linker, the third conserved region covers the C-terminal 81 residues of rQ 1 (83 in tQ 1 ). It contains 10 conserved cysteines, and presents an imperfect internal repeat, as previously noted in the case of tQ 1 (6).
Another cDNA clone, rQ R was isolated from the soleus muscle library and also obtained by a rapid amplification of cDNA ends 3Ј procedure. In this transcript the PRAD-encoding sequence is followed by a short sequence (355 nucleotides) that is not included in rQ 1 (see Fig. 3). As described in a subsequent section, analysis of the genomic sequence in the mouse showed that this unrelated sequence corresponds to the "readthrough" of the following intron. The rQ R transcript was terminated by a poly(A) sequence, despite the absence of a canonical polyadenylation signal.
rQ Encodes the Collagen Tail of AChE A Forms-The identity of rQ 1 as the collagen tail of AChE asymmetric forms was confirmed in two ways. First, the coexpression of this molecule with rat AChE T led to the production of asymmetric AChE molecules in COS cells (not shown) and in Xenopus oocytes ( Fig. 2A). Second, we developed chicken antibodies directed against peptide 35-51, which precedes the PRAD and shows no conservation with the Torpedo sequence. In sucrose gradients, these antibodies shift the sedimentation of all A forms synthesized in oocytes ( Fig. 2A) and all the A forms extracted from soleus muscle (Fig. 2B). The N-terminal peptide of rQ 1 is therefore present and accessible in the collagen-tailed molecules.
Collagen Q Is Encoded by a Single Gene; Partial Genomic Organization-Rat genomic DNA, cleaved by restriction enzymes, was hybridized in Southern blots at low stringency with probes corresponding to the Q C domain. We observed a single strong signal (not shown), suggesting the absence of closely related genes, since C-terminal sequences are generally well conserved among members of a collagen family (26).
We also screened a human cosmid genomic library with two probes corresponding to the rat Q C domain, at low stringency. We obtained five independent cosmids, all of which overlapped in the same genomic region. 2 In addition, we cloned a 35-kb cosmid containing part of the mouse Q gene, by hybridization with a probe covering rat Q N domain. This cosmid covers less than a third of the coding sequence.  codes the PRAD and flanking sequences, and the next 3Ј exon (E coll 1, 96 nucleotides) extends into the beginning of the collagen domain. Exons encoding the N-terminal part of the collagen domain are small (27-72 nucleotides), and each one encodes a multiple of GXY triplets, starting with the first base of the glycine codon (GGN). The few coding differences between the rat and the mouse sequences are indicated on Fig. 3.
The Asymmetric Forms of AChE and BChE Possess the Same Collagen Tail-Another way to investigate the possible diversity of cholinesterase-associated collagens is to examine whether BChE asymmetric forms cross-react with antibodies raised against the nonconserved peptide that precedes the PRAD, anti-rQ N 35-51. Because collagen-tailed forms of BChE are relatively abundant in Torpedo heart (27), and because rQ 1 appears to be expressed at a high level in rat heart, we thought that it might be possible to characterize asymmetric BChE in this tissue. We performed sequential extractions of rat heart ventricle in low salt and in high salt concentrations, as described previously (17), to obtain an extract enriched in A forms. This high salt extract was analyzed by sedimentation in sucrose gradients, followed by assays of AChE and BChE activities, performed in the presence of the specific inhibitors tetraisopropyl pyrophosphoramide and BW284C51, respectively (Fig. 4). The patterns obtained showed the presence of asymmetric forms of both AChE and BChE, with distinct sedimentation coefficients in the fractions of the same sucrose gradient, 16.5 S for AChE A 12 and 18 S for BChE A 12 . This difference in sedimentation, in complete agreement with previous analyses in rat muscle (28), established that the AChE and BChE molecules were distinct and independent. The sedimentation coefficients of all asymmetric forms, BChE as well as AChE, were increased by about 2 S after incubation with the anti-rQ N 35-51 antibodies, while globular forms were not affected. Although the BChE A 4 and A 8 forms were not entirely resolved from the G 4 peak, these molecules were also clearly shifted by the antibodies.
Analysis of RNA Splicing at the Junction between Exons Encoding the PRAD and the Collagen Domain-The rQ R cDNA clone corresponds to unspliced readthrough transcript, in which the PRAD exon is followed by the subsequent genomic sequence (Fig. 3), instead of the collagen exons. This transcript is polyadenylated and thus represents a mature mRNA, which encodes the PRAD without the collagen domain. By RNase protection assays with a probe covering E PRAD and the readthrough sequence rQ R (Q R probe), we showed that this mRNA represents about 5% of the transcripts in the soleus muscle and about 15% in the heart ventricle (Fig. 5). We also performed RNase protection assays with a probe covering E PRAD and the three following exons, encoding the N-terminal part of the collagen domain (Q 1 probe). As shown in Fig. 5, we observed a major protected fragment corresponding to the structure of Q 1 in both tissues; the fragment corresponding to E PRAD originating from rQ R or other splicing events was nearly undetectable in this experiment but was visible in other experiments. Using reverse transcription-polymerase chain reaction analysis we failed to detect any transcript in which the PRADcontaining exon would be associated with downstream sequences differing from rQ 1 or rQ R . We obtained similar results with brain mRNA (not shown).
Amphiphilic Tetramers from Brain Do Not Contain the PRAD of ColQ-Since the protein encoded by rQ R contains the attachment domain without the collagen domain, it appeared possible that it might represent the hydrophobic subunit that anchors AChE tetramers (G 4 a ) in cellular membranes. To test this hypothesis, we coexpressed rQ R with AChE T ; this produced only nonamphiphilic G 4 na molecules, either in COS cells or in Xenopus oocytes (data not shown). In Western blots after nonreducing electrophoresis, we found that anti-rQ N 35-51 antibodies recognize the heavy dimer of G 4 na produced by coexpression of rQ R and AChE T in oocytes but not the dimers that constitute the G 4 a form extracted from bovine brain (Fig. 6). In addition to the labeling of the heavy dimer, which appears diffuse, we observed a labeled band corresponding to the association of a rQ R with a single AChE subunit. In the brain extract, a labeled band of high molecular weight could correspond to the A form. As positive controls, we verified that the antibodies shifted the sedimentation of bovine A forms in sucrose gradient, demonstrating that they interact with the bovine Q N sequence (not shown) and that the heavy dimer of purified brain G 4 na is recognized by an anti-anchor antibody (29), producing a labeled band at the expected position in Western blot (not shown).
Expression of the COLQ Gene in Rat Tissues-We analyzed the tissue distribution of rQ transcripts by Northern blots, using an N-terminal probe extending into the beginning of the collagen domain (N-terminal probe), and a probe corresponding to the Q C domain (C-terminal probe). As shown in Fig. 7, both N-terminal and C-terminal probes recognized a major transcript of about 2.7 kb, corresponding to the size of rQ 1 , in muscle, heart, and brain, as well as in noncholinergic tissues including lung, spleen, and testis, but not in the liver (Fig. 7,  lane 5). Surprisingly, the rQ 1 transcripts appeared less abundant in skeletal muscle than in lung or spleen, and were particularly abundant in heart. In addition to the 2.7-kb transcript, the two probes hybridized with smaller distinct bands but of much lower intensity; for example, the lung contains a 2.2-kb transcript that hybridizes with the C-terminal but not with the N-terminal probe.

A Single Gene Encodes the Collagen Tail of AChE and BChE:
Collagen of Type Q, ColQ-The asymmetric forms of AChE are hetero-oligomers in which AChE T catalytic subunits are associated with collagenous Q subunits. We report the cloning of a Q subunit from rat muscle, rQ 1 , which is homologous to the previously cloned Torpedo collagen subunit, now renamed tQ 1 (6). The fact that rQ 1 indeed encodes the AChE-associated collagen tail was demonstrated by the production of asymmetric forms with AChE T subunits in COS cells and in Xenopus oocytes and by the fact that antibodies directed against a nonconserved peptide from rQ 1 (anti-rQ N 35-51) interact with all asymmetric forms extracted from tissues.
Southern blotting of rat genomic DNA and screening of cos-mid libraries containing mouse and human genomic fragments showed that Q subunits are encoded by a single gene, COLQ. This conclusion is supported by the fact that the anti-rQ N 35-51 antibodies recognize BChE as well as AChE A forms, indicating that the collagen tails of the two enzymes are generated from the same gene. A partial analysis of the collagen exons of the COLQ gene reveals that they are organized as in the genes of true collagens. We found that collagen exons contain an integral number of codons, corresponding to multiples of GXY triplets, as in the case of several fibrillar collagens (30). An alignment of the Torpedo and rat sequences defines well conserved motifs and nonconserved regions. In the collagen domain, the regular repetition of the GXY triplet is interrupted once by a short noncollagenous motif (10 residues) that could correspond to a hinge between independent parts of the collagen. Note that the N-and C-terminal parts of the collagen domain contain a higher proportion of prolines (essentially triplets of amino acids GXP and GPP), compared with the central part. The two putative heparan sulfate binding sites (25) are localized at the two ends of the first part of the collagen domain preceding the hinge. In addition to the collagen domain itself, the conserved motifs include an N-terminal PRAD that binds an AChE tetramer (9), a short sequence that immediately follows the collagen domain (8 conserved residues out of 14), a highly hydrophilic motif containing a high proportion of charged amino acids (45 residues), and a C-terminal motif of 80 residues, containing 10 cysteines with an internal repeat. The C-terminal non-collagen sequences are usually well conserved within collagen families and may participate in the formation of trimers, prior to the organization of the triple helix. The conserved C-terminal motifs of ColQ have no homology with those of other collagens or of other known proteins; in addition, the length and the structure of the collagen domain have no equivalent in any known type of collagen (26). Therefore the Torpedo and rat collagen subunits of cholinesterase collagentailed forms constitute a distinct type of collagen that we name collagen of type Q or ColQ.
ColQ Is Expressed in Cholinergic and Noncholinergic Tissues-Northern blots showed that the rQ gene is expressed in cholinergic tissues such as brain, muscle, and heart, as expected, but also in noncholinergic tissues, including lung and testis, but not liver. We found that all tissues expressing rQ 1 contained transcripts that hybridized with both 5Ј and 3Ј probes and corresponded to the full length of rQ 1 . However, shorter transcripts that hybridized only with a 5Ј or with a 3Ј probe occurred in a tissue-specific manner, e.g. in the lung (3Ј) The lanes correspond to heart (1), brain (2), spleen (3), lung (4), liver (5), skeletal muscle (6), kidney (7), and testis (8). The same membrane was hybridized successively with C-and N-terminal 32 P-labeled probes, the structure of which is schematically shown at the top. and in the testis (5Ј). The short 3Ј transcript cannot encode the binding domain (PRAD) nor the beginning of the collagen domain, so that the deduced protein would not associate with AChE T catalytic subunits and may not be collagenous in structure. As for the short 5Ј transcript, it could correspond to an alternatively spliced transcript that would encode the PRAD, but not the collagen domain, as in the rQ R cDNA clone (see next section), or to nonspecific hybridization with the PRAD-coding sequence, that contains a repetition of proline codons (CCN).
Extracts from rat heart ventricles were found to contain only a low level of collagen-tailed AChE forms (31). It was therefore surprising to find that heart contains the highest level of COLQ transcripts, which we found to correspond to rQ 1 (not shown). This suggests that ColQ is not exclusively associated with accumulated AChE at synapses, but may also exist as an independent component of extracellular matrices. It should be recalled, however, that collagen-tailed forms of cholinesterases may have been underestimated or undetected, if present in some tissues in a nonextractable state (32,33).
Diversity of AChE Anchor Molecules-The N-terminal noncollagen sequence contains two vicinal cysteines, as well as stretches of consecutive prolines, constituting the PRAD that binds a tetramer of AChE T subunits (7)(8)(9). This very well conserved peptidic motif of 17 amino acids is sufficient to interact with AChE T subunits, organizing tetramers from monomers and dimers.
We found that the exon encoding the binding domain, E PRAD , may be expressed without the collagen exons. This raised the possibility that the 20-kDa hydrophobic anchor of AChE (G 4 a ) (3, 4) might be encoded by a splice variant of the COLQ gene. The same attachment domain, PRAD, would be used for associating AChE T subunits with both collagenous and hydrophobic anchors, allowing the localization of the enzyme in the basal lamina and in membranes, respectively. One cDNA clone (rQ R ) corresponds to a readthrough transcript, in which E PRAD is not spliced to the first collagen exon (E coll 1) as in rQ 1 , but followed by the continuous "intronic" sequence that encodes a short peptide. This transcript is mature since its terminates with a poly(A) stretch. When co-expressed with AChE T , the resulting protein was incorporated in a catalytic tetramer composed of a light dimer and a heavy dimer. The heavy dimer was labeled in Western blots by anti-rQ N 35-51 antibodies, showing that it contains the PRAD domain. In contrast, the heavy dimer of the G 4 a form extracted from brain was not recognized by the anti-rQ N 35-51 antibodies. We conclude that the 20-kDa hydrophobic anchors of brain AChE does not include a peptide encoded by E PRAD and therefore does not derive from the COLQ gene. Although the Q R transcript is not responsible for AChE anchoring in the membrane, it might play a role in the production of soluble tetrameric molecules (G 4 na ), which are secreted at the neuromuscular junction (34) and may somehow become attached to the synaptic basal lamina (35). In skeletal muscle, heart, and brain, the exon E PRAD was found to be associated to collagen exons, in rQ1 or to readthrough sequence, in rQ R . We cannot exclude, however, that different splicing patterns might occur in other tissues, e.g. in testis as indicated above, so that the COLQ gene might generate yet other cholinesterase-associated proteins.
It is worth noting that Xenopus oocytes, like RBL cells (36), produce practically only AChE T monomers and dimers in the absence of Q 1 , while the human HEK 293 cells and the COS cells produce and secrete significant proportions of tetramers (8,14,37). It is possible that the organization of AChE T subunits into tetramers requires the presence of a binding protein that is expressed at variable levels in the different cell types. It is intriguing that such cholinesterase-associated proteins may not be genetically or even structurally related to ColQ, since the hydrophobic anchor of membrane-bound AChE tetramers from brain does not contain the PRAD.