Selective expression of an acetylcholine receptor–lacZ transgene in synaptic nuclei of adult muscle fibers

The postsynaptic membrane of the neuromuscular junction forms a distinct and highly specialized domain on the surface of the skeletal muscle fiber. For example, acetylcholine receptors (AChRs) are present at a density of ⩾15 000 μm² in the postsynaptic membrane, but are virtually absent (⩽10 μm²) from extrasynaptic regions of the muscle fiber membrane (Salpeter, 1987; Schuetze and Role, 1987; Bloch and Pumplin, 1988). Other proteins known to be concentrated at postsynaptic sites include N-CAM in the plasma membrane; slaminin, agrin and acetylcholinesterase in the basal lamina; and 43K/RAPSYN, dystrophin-like proteins, and acetylated microtubules in the cytoskeleton (Massoulie and Bon, 1982; Froehner, 1986; Bloch and Pumplin, 1988; Sanes, 1989; Chang et al. 1989; McMahan and Wallace, 1989; Jasmin et al. 1990; Sanes et al. 1990; Ohlendieck et al. 1991). The motor nerve induces these specializations as development proceeds, and contributes to their subsequent maintenance (Schuetze and Role, 1987; Martinou et al. 1991).

In investigating mechanisms that might contribute to the generation and maintenance of the postsynaptic apparatus, we have focussed on the expression of AChR subunit genès. AChRs are composed of α, β, γ, and δ subunits in the embryo, and of α, β, ϵ, and δ subunits at adult synapses (Mishina et al. 1986; Gu and Hall, 1988). We previously showed that mRNAs encoding the AChR α- and δ subunits are more abundant in synapsϵrich than in synapsϵfree regions of adult muscles (Merlie and Sanes, 1985). Subsequent studies have extended our results to the AChR β and ϵ subunits, and used in situ hybridization to demonstrate that AChR subunit mRNAs accumulate at synaptic sites (Fontaine et al. 1988; Fontaine and Changeux, 1989; Goldman and Staple, 1989; Brenner et al. 1990; see also Harris et al. 1989; Bursztajn et al. 1989). The most straightforward interpretation of these results is that AChR subunit genes are selectively transcribed by the few nuclei within each multinucleated muscle fiber that are located at the synaptic site, although it is also possible that AChR mRNA is transported from extrasynaptic to synaptic regions or selectively stabilized at synapses (Merlie and Sanes, 1985).

An important next step in understanding how AChRs – and potentially other synaptic components – are directed to synaptic sites will be to identify nucleotide sequences that are responsible for the observed synaptic accumulation of AChR mRNAs. Regulatory elements from two AChR subunit genes (α- and δ) have recently been isolated and studied in cultured muscle cells (Klarsfeld et al. 1987; Baldwin and Burden, 1988; Wang et al. 1988, 1990; Piette et al. 1990; Prody and Merlie, 1991). However, synapsϵassociated nuclei are not readily identifiable, and may not constitute a distinct subpopulation, in currently available culture systems. Accordingly, we have adopted the alternative approach of studying AChR subunit gene expression in transgenic mice (Merlie and Komhauser, 1989; Merlie and Sanes, 1990). We report here on the generation and characterization of transgenic mice in which regulatory elements of the AChR α-and ϵsubunit genes direct expression of a derivative of E. Coll β-galactosidase (lacZ) that is targeted to the nucleus (we call the fusion gene nlacZ, and the fusion protein nlacZ). We show that the regulatory elements present in both transgenes are sufficient to direct musclϵspecific and developmentally regulated expression of nlacZ, and to direct high levels of nlacZ expression in the intrafusal fibers of muscle spindles. In addition, we show that ϵ-derived sequences direct highly selective expression of nlacZ by synapsϵassociated nuclei in adult skeletal muscle fibers. These results provide new evidence for differences between synaptic and extrasynaptic nuclei, and establish a system in which the molecular basis of the differences can now be studied.

In independent studies reported while this work was in progress, Klarsfeld et al. (1991) described preferential expression of nlacZ by synaptic nuclei in neonatal mice bearing an onlacZ transgene. In our animals, the synaptic preference is considerably less dramatic in α-lacZ than in ϵ-nlacZ mice. Possible explanations for the differences between the α-nlacZ and ϵ-nlacZ transgenes, and between our results and those of Klarsfeld et al. are considered in the Discussion.

The 5’-flanking sequences of the mouse e gene were subcloned from genomic clone λge4 (Buonanno et al. 1989). Initially, a 6.5 kb Sam HI restriction fragment was subcloned into the Bluescript cloning vector, pSK+, (Stratagene, San Diego, CA). This plasmid, pSKe4a, was shown by sequence analysis to contain the first coding exon and part of the first intron. Because no convenient restriction sites were found within this fragment, we prepared a separate subclone of the 5’-end free of protein coding sequences by polymerase chain reaction using synthetic oligonucleotide primers. The noncoding primer extended to the last base before the translation start codon and contained a synthetic BamHI cloning site at its 3’-end, while the coding primer extended from just 5’ of the WmdIII site in exon 1 (see Fig. 1a) and also included a synthetic BamHI cloning site. This fragment was subcloned into the BamHI site of pSK+ and verified by sequence analysis. To prepare pϵ3500, the 3.5 kb WmdIII fragment (Fig. 1a) was subcloned into the HindIII digested subclone of the PCR product. The resultant full-length 3.5 kbp promoter fragment of pϵ3500 was then used to prepare CAT and lacZ fusion genes as follows: An XhoI-BamHI fragment of pϵ3500 was blunted by Klenow treatment and ligated into the blunted HindIII site of pUC9-CAT (Donoghue et al. 1988) to produce pϵ3500-CAT or into the Smal site of pnlacZ (R. Palmiter and J. Peschon, personal communication; Mercer et al. 1991) to yield pϵ3500-lacZ. The predicted structure of the junction region of each fusion gene was confirmed by sequence analysis.

To prepare pα-nlacZ, the 850 bp chicken α-subunit promoter fragment was obtained from a SalI digestion of clone pαAChCAT (Klarsfeld et al. 1987) and ligated into Sah-digested pnlacZ.

To test α-derived sequences for enhancer activity, the α3500 fragment was cloned into the unique BamHI site of an MLC1 promoter CAT vector containing the 400 bp PvuII-HincII promoter fragment fused to CAT in pUC9CAT (Donoghue et al. 1988; Prody and Merlie, 1991).

The start of transcription was determined by RNAase protection and primer extension analyses. For RNAase protection an antisense probe was prepared (Melton et al. 1984) from pSKe4a using the BgBI site to linearize the template (Fig. 1a). RNA was hybridized with ³²P-labeled antisense probe and treated as described in Martinou and Merlie (1991). Primer extension analysis (Ghosh et al. 1980) was performed using a 30 nucleotide synthetic oligonucleotide corresponding to the 3’-end of exon 1 (Fig. 1a). The product was labeled with ³⁵S-dATP.

For transient transfections, supercoiled plasmid DNA was purified twice on CsCl gradients, then introduced into C2 or 3T3 cells using the calcium phosphate method (Graham and Van der Eb, 1973) as described previously (Donoghue et al. 1988; Prody and Merlie, 1991). Cotransfections with an SV40-lacZ plasmid, pCHHO (Lee et al. 1984), were performed to monitor transfection efficiency in each tissue culture dish. β-galactosidase assays were done as described (Miller, 1972). CAT activity in cell extracts was determined (Gorman, 1985) using aliquots of extracts containing equal amounts of β-galactosidase. ¹⁴C-acetylated chloramphenicol spots were cut from thin layer chromatograms and counted in a scintillation counter.

Transgenic mice were prepared by standard methods (Hogan et al. 1986) using linear DNA prepared from kpnI-HindIII and SmaI-SphI digested panlacZ and penlacZ respectively. Embryos for injection were obtained from (C57BL6/J×CBA F₁) × (C57BL/6J×CBAF₁) matings. Transgenic mice were identified by a polymerase chain reaction using tail digests (Hanley and Merlie, 1991). Transgenic founders and subsequent generations were bred by backcrossing to C57BL6/J×CBA F₁ hybrids. Thus, all experiments were performed with mice hemizygous for the transgene.

Mice were anesthetized with ether, and dissected in PBS (150 mM NaCl, 10 UM sodium phosphate, pH 7.3). Whole embryos or individual adult muscles were fixed for 1 h in 2 % paraformaldehyde plus 0.2% glutaraldehyde in PBS, washed thoroughly in PBS, and stained for lacZ as described by Sanes et al. (1986) and Weis et al. (1991). The staining solution contained 2 mM 5-bromo-4-chloro-3-indolyl-/3-D-galactoside (X-gal; United States Biochemicals, Cleveland, OH), 5mM potassium ferrocyanide, 5HIM potassium ferricyanide, and 2mM MgC12 in PBS. Staining was for 16–20 h at 30°C. Tissues were then rinsed several times in PBS and reincubated in PBS at 30–37° for an additional day, which caused an intensification of the staining even though substrate was no longer present. Thereafter, tissues were stored at 4° in 2% paraformaldehyde plus 2 % glutaraldehyde in PBS to minimize accumulation of ‘background’ (nonspecific) staining. Whole embryos or muscles were photographed through a dissection microscope. Single muscle fibers or muscle spindles were teased from fiber bundles, mounted in glycerol, and photographed through a compound microscope with Nomarski optics. In some cases, individual fibers were stained for cholinesterase by the method of Karnovsky and Roots (1964) after they were isolated but before they were mounted.

Cloned DNA spanning the mouse AChR ϵ-subunit gene was available from a previous study (Buonanno et al. 1989). To seek elements that regulate e gene transcription, we first subcloned and partially sequenced a 6.5 kb fragment that extended 5’ from the first exon (Fig. 1a). We then used the method of RNAase protection (Melton et al. 1984) to determine the site at which transcription is initiated. The probe used is indicated in Fig. 1a, and results of a typical experiment are shown in Fig. 1b. These results fixed the site of transcription initiation (marked +1 in Fig. 1c) at 82 bp 5’ of the translation start codon. Attempts to confirm this assignment by the method of primer extension were unsuccessful, perhaps because ϵsubunit mRNA is an extremely rare message -it constitutes <0.001 % of poly(A)+ RNA and is <5 % as abundant as other AChR subunit mRNAs in neonatal, denervated, or cultured muscles (Martinou and Merlie, 1991). However, indirect confirmation of our assignment is that the start site for ϵ -subunit transcription determined by RNAase protection is within a first, short, protein-coding exon, as we and others have found for AChR α, β, γ, δ subunit genes (Klarsfeld et al. 1987; Crowder and Merlie, 1988; Baldwin and Burden, 1988; Wang et al. 1990; Prody and Merlie, 1991; C. Prody and J.P.M., unpublished).

Sequences 5’ of the transcriptional initiation site frequently contain regulatory elements that direct tissuϵspecific and developmentally regulated expression (Mitchell and Tijan, 1989). Sequence analysis of this region of the ϵ -subunit gene revealed that it is similar to that of other AChR subunit genes, and that it contains potential binding sites for both ubiquitous and musclϵspecific transcription factors (Fig. 1c, and see Discussion). Based on these results, we asked whether sequences 5’ to the transcriptional initiation site could function as a promoter or enhancer. A fragment of the e gene containing 3.5kb of 5’ untranscribed sequence (ϵ3500 in Fig. 1a) was fused to a heterologous reporter, chloramphenicol acetyltransferase (CAT); the plasmid was transfected into cultured cells and CAT levels were assayed 96 h later. Initial tests made use of C2 cells (Yaffe and Saxel, 1977), which fuse in culture to form multinucleated myotubes that synthesize a variety of musclϵspecific proteins including AChRs. To seek evidence for cell typϵspecific and differentiationdependent expression, the plasmid was also transfected into C2 myoblasts, which remain mitotically active and express only low levels of musclϵspecific genes; and into 3T3 cells, which are fibroblasts that normally do not express musclϵspecific genes. As shown in Une 1 of Fig. 2, CAT activity was > 100-fold higher in C2 myotubes transfected with this plasmid than in either C2 myoblasts or 3T3 fibroblasts. As a control, we also assayed cells transfected with the parent CAT plasmid, lacking the e3500 fragment. This ‘promoterless’ CAT plasmid was inactive in all three cell types, (Fig. 2, line 2) indicating that bacterial sequences alone were insufficient to direct detectable expression of CAT. Thus, the ϵ-derived sequences tested can promote cell typϵ and differentiation-dependent CAT expression.

To ask whether the ϵ-derived sequences can act as an enhancer of transcription, we constructed plasmids in which e3500 was placed 3’ to the CAT gene. In these constructs, basal promoter function was provided by sequences from the rat myosin light chain 1/3 gene, which direct negligible levels of expression on their own (Fig. 2, fine 3) but are capable of directing high levels of expression in conjunction with enhancers (Donoghue et al. 1988). The ϵ -derived sequences were active and tissuϵspecific in either orientation when placed 3’ to the myosin promoter-CAT gene (Fig. 2, lines 4 and 5). Thus, ϵ -derived sequences satisfy the criteria for an enhancer in that their tissuϵspecific activity is position- and orientation independent.

To study gene expression in vivo, the ϵ3500 fragment described above was fused to a modified lacZ gene. lacZ was substituted for CATbecause its product can be detected histochemically in whole mounts (Sanes et al. 1986). In addition, the lacZ gene was augmented by incorporation of a short sequence that encodes a nuclear localizing signal from the SV40 T antigen (Kalderon et al. 1984; Dingwall and Laskey, 1986; Galileo et al. 1990); we call the fusion protein nlacZ. In previous studies, Ralston and Hall (1989) generated hybrid myotubes from nlacZ-positive and -negative C2 cells, and showed that nlacZ accumulates preferentially in myotube nuclei that contain the nlacZ gene. Thus, we had reason to believe that an nlacZ reporter could serve to mark subsets of nuclei within individual muscle fibers in vivo.

Eight mice bearing the ϵ-nlacZ transgene were generated by standard methods. These mice, or their offspring, were killed at ⩾3 weeks of age, when endogenous e gene expression has reached adult levels (Martinou and Merlie, 1991). Several muscles and a variety of other tissues from each mouse were fixed and stained for nlacZ. We detected expression of nlacZ in four of the eight lines. In all four, nlacZ-positive nuclei were detectable in many skeletal muscles, but not in any of a variety of extramuscular tissues tested, including arteries (smooth muscle) and heart (cardiac muscle). Thus, the ϵ-derived sequences used in the transgene promote skeletal musclϵspecific expression of the reporter.

Most notably, staining within the muscles of ϵ-nlacZ mice was concentrated in narrow bands that corresponded to the ‘end-plate zones’ along which intramuscular nerves run and within which synapses form (Fig. 3a,c). To test the synaptic specificity of nlacZ expression, individual fibers were teased from the muscles and examined with Nomarski optics. Our main observations were the following, (a) All stained nuclei were associated with muscle fibers; nuclei in intramuscular nerves, blood vessels, and connective tissue were unstained. However, nlacZ-positive nuclei occurred in the intrafusal muscle fibers of muscle spindles (see below) as well is in extrafusal (ordinary) muscle fibers, (b) Within each extrafusal muscle fiber, staining was confined to a small group of nuclei in the endplate region (Fig. 3d) -typically 15–30 of the several hundred total nuclei (Schmalbruch and Hellhammer, 1977) per fiber. Nuclei outside of the end-plate zone were not detectably stained, (c) Most muscle fibers contained two distinct types of nlacZ-positive nuclei. Generally, one small group of 3–6 nuclei was most intensely stained in each fiber. These nuclei were tightly clustered, large, and round (Fig. 3e), all features that identified them as the ‘fundamental’, ‘solϵplate’ or ‘synaptic’ nuclei that aggregate directly beneath the postsynaptic membrane of the neuromuscular junction (Couteaux, 1973; Zacks, 1973). This identification was confirmed by restaining teased fibers for the synaptic marker acetylcholinesterase (Karnovsky and Roots, 1964); the nlacZ-positive nuclear clusters invariably lay directly beneath acetylcholinesterasϵstained synaptic sites (Fig. 3f). The other, less intensely stained nuclei occurred singly, and were generally elliptical in shape and smaller than their synaptic counterparts (Fig. 3d, inset); these nlacZ-positive perisynaptic nuclei were histologically indistinguishable from extrasynaptic nuclei throughout the muscle fiber. Whether only synaptic nuclei or also perisynaptic nuclei are specialized is unclear: perisynaptic nuclei might be stained because they express nlacZ or because some nlacZ mRNA or protein diffuses from synaptic to perisynaptic sites. In either case, these results suggest that synapsϵassociated and extrasynaptic nuclei in a single muscle fiber express different genes, and that the ϵsubunit gene contains elements that promote its preferential expression by synaptic nuclei.

In a second set of transgenic mice, we linked nlacZ to sequences from the AChR α-subunit gene that we previously showed to be capable of directing musclespecific expression of CAT in vivo (Merlie and Komhauser, 1989; see also Klarsfeld et al. 1991). Expression of nlacZ was musclϵspecific in α-nlacZ mice, but in this case the reporter accumulated in both extrasynaptic and synaptic nuclei (Figs 3g—j and 4c). In a minority of muscle fibers, synaptic nuclei were significantly more intensely stained than extrasynaptic nuclei (Fig. 3k). In most fibers (>95%), however, synaptic and extrasynaptic nuclei were stained with similar intensity (Fig. 3I). Similar results were obtained in both intensely and faintly stained muscle fibers (compare Figs 3g and h), indicating that the different patterns of staining in α-nlacZ and ϵ-nlacZ mice do not result simply from higher overall levels of expression in the former. The lack of a marked synaptic preference in α-nlacZ mice was unexpected, given the localization of endogenous α-subunit mRNA and the results obtained with ϵ-nlacZ mice; possible explanations are presented below (see Discussion). Whatever the explanation, however, the α-nlacZ mice provided a fortuitous control for the possibility that the dramatic synaptic localization of nlacZ in ϵ-nlacZ mice reflected some peculiarity of the reporter (e.g., differential metabolism of nlacZ in synaptic and extrasynaptic areas) rather than an influence of ϵderived sequences.

Results presented so far indicate that the ϵsubunit gene contains elements important for selective expression in synaptic nuclei. Before accepting this conclusion, however, it is necessary to show that patterns of expression reflect the activity of the ϵderived sequences, rather than influences of the chromosomal site of integration. To this end, we analyzed four independently generated lines of ϵ-nlacZ mice (#s 2, 29, 51 and 75), each of which had presumably inserted the transgene at a different site. The number of muscle fibers and the particular muscles that expressed nlacZ varied from line to line. For example, mice of line 75 reproducibly showed more intense staining in forelimb than in hindlimb muscles, whereas no such difference was observed in line 29 mice. Thus, insertion site or genetic background may subtly affect transgene expression. Importantly, however, nlacZ-positive nuclei were selectively associated with synaptic sites in all four ϵ-nlacZ lines, indicating that sequences within the transgene rather than at the integration site were responsible for this feature of transgene, expression.

We also compared expression of nlacZ in three independently derived α-nlacZ mice -the stable line (<r-nlacZ16) described above, and two other adult founders, which were not bred into lines. In all three cases, synaptic nuclei were more intensely stained than extrasynaptic nuclei in a small minority <5 % of muscle fibers, but the degree of selectivity was low, and synaptic and extrasynaptic nuclei were stained with similar intensity in the majority of fibers. Expression of nlacZ varied from muscle to muscle in αnlacZ mice as was the case in ϵ-nlacZ mice. However, variations among muscles were more readily documented in α-nlacZ than in ϵ-nlacZ mice, because staining extended along entire muscle fibers. In mice of line 16, for example, tibialis anterior (Fig. 3g) and sternomastoid (Fig. 4a) contained large regions in which nearly all fibers were stained, internal intercostals (Fig. 3h) were composed of a mixture of stained and unstained fibers, and soleus (Fig. 4b) contained no stained fibers. Interestingly, we observed similar patterns of staining in the other two α-nlacZ founder mice, suggesting that intermuscular variations in expression may be transgenϵencoded rather than integration sitϵdependent in this case. However, analysis of other α-nlacZ lines will be needed to test this point critically.

An unexpected observation in ϵ-nlacZ mice was that nlacZ was expressed in the intrafusal muscle fibers of muscle spindles as well as in synaptic nuclei of extrafusal or ordinary muscle fibers. Spindles were visible as aggregates of intensely nlacZ-positive nuclei that contained more nuclei than were found at endplates and that lay outside of endplate bands (Fig. 5a). The identity of the spindles was confirmed in teased preparations, in which the individual fibers and the spindle capsule were clearly visible (Fig. 5c). Labeled nuclei occurred in both nuclear bag and chain fibers (Barker and Banks, 1986), although, not all fibers were stained in all spindles. Within spindles, staining was generally confined to the central or equitorial region, which is the area in which both sensory and motor nerve endings are concentrated (Barker and Banks, 1986; see Fig. 7); distal areas, which are poorly innervated, were generally unstained. Thus, nlacZ is preferentially expressed in the most densely innervated portion of intrafusal fibers.

Extrasynaptic expression of nlacZ rendered identification of spindles difficult in αnlacZ mice. In less intensely stained muscles, however, it was possible to recognize nlacZ-positive spindles, and to confirm their identity by isolating them (Fig. 5d). Thus, nlacZ is expressed in intrafusal fibers in n-nlacZ as well as in ϵ nlacZ muscles.

The endogenous AChR α-and ϵsubunit genes are regulated differently during development, αsubunit protein and mRNA levels decline during the first few postnatal weeks, as AChRs are lost from extrasynaptic portions of the muscle fiber surface and become restricted to synaptic sites. In contrast, ϵsubunit levels are low in embryos, then increase selectively at synaptic sites postnatally, as synaptic AChRs convert from the α₂βγδ to the α₂βϵδ form (Schuetze and Role, 1987; Witzemann et al. 1989; Brenner et al. 1990; Gu and Hall, 1989; Martinou and Merlie, 1991).

To ask whether the α-and ϵsubunit-derived sequences in the transgenes were capable of directing developmentally regulated expression, we stained staged embryos and neonatal mice for nlacZ. Results are illustrated in Fig. 6 and summarized in Table 1. Staining in α-nlacZ16 embryos was detectable by embryonic day (E) 11, the earliest age tested. At this time, staining was confined to a series of axial stripes (Fig. 6a) that were identifiable at high power as myotomes and the earliest rudiments of the segmented axial musculature (Fig. 6b). By E14, muscles were nlacZ-positive in the head and limbs as well as in the trunk (Fig. 6c). Levels of staining remained high throughout embryogenesis and in early postnatal life (Fig. 6d), then decreased during the first few postnatal weeks (Table 1). Although this decrease appears to be less dramatic than that observed for endogenous α-subunit mRNA and for an α-CAT transgene (Merlie and Komhauser, 1991), we believe that it represents an appropriate, activity-dependent down-regulation, because we can prevent or reverse the decrease in expression by neonatal denervation (K. Gundersen, J.R.S., and J.P.M., unpublished results).

Differences in staining intensity among muscles were apparent in embryonic and neonatal αnlacZ animals, but we have not yet determined whether the nlacZ-positive and -negative embryonic muscles correspond to those that are positive and negative in the adult. Within individual muscles, however, nlacZ-positive nuclei were present both synaptically and extrasynaptically at all stages of development; in particular, the occasionally selective staining of synaptic nuclei was no more evident in neonates than in adults (Fig. 6d and Table 1).

Expression of nlacZ in ϵnlacZ embryos and neonates differed from that in αnlacZ mice. No expression was detected in extrafusal fibers of embryos in either of the two lines that were studied (Fig. 6e and Table 1). However, muscle spindles were nlacZ positive by E16-18 (not shown), which is nearly as soon as they form (Kozeka and Ontell, 1981). Expression was first detectable in extrafusal fibers during the first postnatal week -at P6 in line 29 and at P2 in fine 75 (Fig. 6f and Table 1). Levels of expression then increased over the first postnatal weeks to reach adult levels by P21. At all times studied, expression was musclϵspecific, and nlacZ-positive nuclei were confined to synaptic regions (Fig. 6f) and to muscle spindles (Fig. 5b).

The principal aim of the studies reported here was to seek evidence for selective expression of ‘synaptic’ genes by synaptic nuclei in muscle fibers. To this end, we generated, analyzed, and compared transgenic mice bearing sequences from two AChR subunit genes, α-and e. We chose these two subunits because their expression is regulated differently. All AChRs in muscle are thought to contain the α-, β-, and δ-subunits. They are present throughout the muscle fiber surface in embryos, are lost extrasynaptically as development proceeds, become restricted to synaptic sites in adults, and reappear extrasynaptically following denervation. In contrast, ϵsubunit-containing AChRs (presumed subunit composition, α₂βϵδ) are almost completely restricted to adult synaptic sites; the ‘fetal’ γ-subunit takes the place of e in developing muscle and in the extrasynaptic AChRs that appear following denervation (presumed subunit composition, α₂βγδ) (Salpeter, 1987; Schuetze and Role, 1987; Block and Pumplin, 1988; Mishina el al. 1986; Gu and Hall, 1988). Thus, even though both α-and ϵsubunit mRNAs are concentrated near synapses in mature muscle (Merlie and Sanes, 1985; Fontaine and Changeux, 1989; Goldman and Staple, 1989; Brenner et al. 1990), it was not obvious that this restriction would be regulated identically.

Because regulatory elements of the ϵsubunit gene had not previously been characterized in any species, we began this study by locating the transcription start site of the mouse ϵgene (Fig. 1) and testing elements 5’ of this site for promoter/enhancer activity. We found a 3.5 kb genomic fragment that can act as both promoter and enhancer to direct cell type-specific and differentiation-dependent expression of a reporter gene in tissue culture cells (Fig. 2). Sequence analysis established that the ϵ gene is similar to other AChR subunit genes (Klarsfeld et al. 1987; Baldwin and Burden, 1988; Piette et al. 1990; Wang et al. 1990; Prody and Merlie, 1991) in several respects: the first exon is proteincoding, the 5’ untranslated portion of the mRNA is relatively short, and the gene lacks a canonical TATA box but bears a weak consensus for several TATα-dependent genes (Smale and Baltimore, 1989). Furthermore, sequences in a 0.2 kb stretch 5’ to the transcriptional initiation site contain potential binding sites for Spl, AP2 (Mitchell and Tijan, 1989), and MyoD-like (Olson, 1990; Weintraub et al. 1991) transactivating proteins. In fact, we now know that this 0.2 kb fragment is sufficient to promote expression in cultured muscle cells (T. Sunyer and J.P.M., unpublished); however, because we were seeking complex regulatory behavior in vivo, we used the entire 3.5 kb fragment to construct the transgene.

In contrast, we used a previously characterized 0.85 kb fragment from the chicken αsubunit gene to generate αnlacZ mice. This fragment was shown by Klarsfeld et al. (1987) to direct cell typϵspecific and differentiation-dependent expression in cultured cells; several groups have subsequently mapped and characterized functionally important elements within this fragment (Wang et al. 1988; Piette et al. 1990; Prody and Merlie, 1991). Most important, we previously showed that these sequences are sufficient to direct musclespecific, developmentally-regulated, and activity-dependent expression of the reporter gene, CAT, in transgenic mice (Merlie and Kornhauser, 1989). However, CAT is a freely soluble enzyme, and is also difficult to localize at a cellular level in situ (but see Donoghue et al. 1991); its distribution within individual muscle fibers and its variation among fibers within a muscle had therefore been inaccessible to study. Accordingly, we constructed an αnlacZ transgene for use in the present work.

Both the α-and the ϵsubunit derived sequences proved to be active in transgenic mice when linked to nlacZ, and the histochemical stain for lacZ provided a sensitive means for studying patterns of expression at cellular and (for multinucleated muscle fibers) subcellular levels. Expression of both a’-nlacZ and ϵnlacZ transgenes was skeletal musclϵspecific; other tissues, including smooth and cardiac muscle were uniformly nlacZ-negative. Furthermore, within skeletal muscles, expression was restricted to muscle fibers. These results extend those previously obtained with the αCAT transgene, and establish that ϵsubunit derived sequences can direct expression that is as cell typϵspecific as that directed by the ϵ-derived sequences (Fig. 3). In addition, both α-and ϵderived sequences direct developmentally regulated expression that is generally appropriate for the respective endogenous genes (Fig. 6): expression of the αnlacZ transgene is detect-able as soon as muscles form, and declines in level postnatally, whereas expression of the ϵ-nlacZ transgene is first detectable perinatally, and then increases in level postnatally.

A striking demonstration of the additional information that can be obtained by use of a histochemically detectable reporter is the finding that both α-nlacZ and ϵ-nlacZ transgenes are expressed in the intrafusal fibers of muscle spindles (Fig. 5). This result may reflect endogenous AChR expression, in that motor axons form conventional cholinergic neuromuscular junctions on intrafusal fibers (Barker and Banks, 1986), and we have detected binding of α-bungarotoxin to intrafusal fibers (Fig. 7); however, the developmental regulation and subunit composition of AChRs in spindles has not, to our knowledge, been studied. The ϵ-nlacZ transgene may prove to be useful in future studies of spindles in both developing and adult muscles, because it provides a convenient marker for locating these small sensory organs and assessing their distribution; we know of no currently available reagents that selectively stain muscle spindles in whole mounts.

In a minority of muscle fibers from αnlacZ mice, synaptic nuclei were significantly more intensely stained than extrasynaptic nuclei. In most fibers, however, the αnlacZ transgene exhibited at best a modest preference for synaptic nuclei. In that endogenous AChR α-subunit mRNA is highly concentrated near synapses (Merlie and Sanes, 1985; Fontaine and Changeux, 1989; Goldman and Staple, 1989), this result might seem unexpected. However, even if ossubunit gene transcription is synapsϵspecific, there are several possible explanations for the pattern of transgene expression we observed. (1) Sequences important for directing high levels of expression at synapses and/or for repressing expression extrasynaptically might be absent from or inactive in the relatively short genomic fragment used to construct the αnlacZ transgene. In this context, it will be interesting to ask whether longer ϵ-subunit genomic sequences direct more markedly synapsespecific expression. (2) We showed previously that significant (albeit low) levels of endogenous αsubunit mRNA are present in extrasynaptic regions of normal adult muscle (Merlie and Sanes, 1985). In contrast, little if any δ- or ϵ-subunit mRNA is detectable extrasynaptically (Merlie and Sanes, 1985; Witzemann et al. 1987; M. Valleca and J.P.M., unpublished). Thus, the α-subunit gene may be less stringently regulated than other subunit genes in this respect. (3) The histochemical method that we used is a powerful one, but it is not quantitative and may not be suited for discriminating relatively small differences in reporter level. In short, we cannot conclude that the αnlacZ transgene exhibits no selective expression at synapses, only that its distribution is far less dramatically synapsespecific than that of the ϵ-nlacZ transgene.

While our studies of the α-nlacZ transgene were underway (Merlie and Sanes, 1990), Klarsfeld et al. (1991) reported the generation of transgenic mice bearing the same AChR α-subunit sequences that we used. They described what appears to be more selective expression in synaptic nuclei of neonates than we observe. However, our methods differ in several respects that make it difficult to assess just how markedly our results differ. (1) Klarsfeld et al. (1991) observe no synaptic selectivity in PO animals, and no expression at all after P4. Although we examined embryos, neonates, and adults, it is possible that we have so far failed to detect a narrow range of ages during which expression of nlacZ is markedly synapsespecific in our α-nlacZ mice. In addition, because we were interested in AChR expression at mature synapses; we screened transgenics as adults; we would therefore have rejected as non-expressers any that displayed precisely the pattern observed by Klarsfeld et al. (2) We generated and examined three separate lines of mice, whereas they obtained only one; it is not unusual to find quantitative differences in expression among lines. (3) Although we both used identical α-subunit regulatory and nlacZ translated sequences, we used different polylinkers and 3’ untranslated sequences; these might subtly affect expression. (4) Klarsfeld et al. describe synapsϵspecific expression in diaphragm, but do not comment on patterns of expression in other muscles. Although we examined numerous muscles in both trunk and limbs, it is possible that we failed to identify some muscles in which nlacZ expression is markedly synapsϵspecific. (5) Klarsfeld et al. (1991) document an ~3-fold difference in peak intensity between synaptic and extrasynaptic nuclei. As noted above, the histochemical method that we used may not reliably discriminate small differences in nlacZ level among nuclei. In summary, both we and Klarsfeld et al. (1991) detect some selective expression of an α-nlacZ transgene by synaptic nuclei, and technical considerations may explain much of the apparent discrepancy between their results and ours.

Our most important observation is that the ϵnlacZ transgene is selectively expressed by synaptic nuclei in adult skeletal muscle fibers. Two observations established that ϵderived sequences are responsible for this selectivity. First, synapsϵspecific expression was observed in four independently derived lines of ϵnlacZ mice, each of which had integrated the transgene at a distinct site, thus rendering highly unlikely the possibility that endogenous sequences near the site of integration are directing expression of the transgene. Second, nlacZ was expressed at similar levels in synaptic and extrasynaptic nuclei of most muscle fibers from three independently-derived αnlacZ mice, thus demonstrating that synaptic localization does not reflect differential metabolism of the reporter protein in synaptic and extrasynaptic areas -for example, selective transport into synaptic nuclei or lability in extrasynaptic nuclei.

The most plausible interpretation of our results is that ϵderived sequences in the ϵnlacZ transgene include elements that promote selective transcription of the transgene - and, by implication, of the endogenous gene -in synaptic nuclei. We cannot formally exclude the possibilities that ϵnlacZ mRNA is preferentially stabilized in or transported to synaptic nuclei, because, as in most transgenes, the regulatory regions that we used extend past the start site of transcription and therefore contribute to the mRNA: the α-nlacZ and ϵ nlacZ mRNAs contain short untranslated sequences (19 and 82 nucleotides, respectively) derived from the AChR genes. However, the two mRNAs are identical over ~3.6kb, including all translated and 3’ untranslated portions of the message. Thus, by far the major difference between these two constructs is in their nontranscribed regulatory elements. Furthermore, insofar as determinants of eukaryotic mRNA stability have been defined, they generally lie in the translated or 3’ untranslated regions (Cleveland, 1988; Bernstein and Ross, 1989; Saini et al. 1990), which are identical in the cr-nlacZ and ϵnlacZ transgenes. Similarly, no precedents exist for selective transport of mRNA mediated by short 5’ sequences. Thus, selective expression of the ϵnlacZ transgene in synaptic nuclei is either mediated transcriptionally by elements in the 3500 bp of ϵderived DNA or posttranscriptionally by the 82 nucleotides of ϵderived sequence in the transgene mRNA; of these alternatives, currently available data on regulation of gene expression strongly favors the former. Experiments to distinguish between these alternatives, and to define more precisely the elements responsible for synaptic localization, are now in progress.

In summary, we have demonstrated that an ϵnlacZ transgene is selectively expressed by synaptic nuclei in muscle fibers. Because the sequences that direct expression of the transgene derive from a synapsespecific endogenous gene, we suggest that transcriptional specializations of synaptic nuclei may play a role in maintaining the highly specialized character of the postsynaptic membrane. More generally, several other examples of regional specialization in muscle fibers (e.g., at tendons and in muscle spindles; see, for example, Katz and Miledi, 1964; Jones and Vrbova, 1974; Kucera et al. 1978; Lomo and Slater, 1980; Salviati et al. 1986) may result from differential gene expression by subsets of nuclei within individual fibers. Finally, the transgenic animals described here should provide a useful starting point for defining nucleotide sequences that promote synaptic localization, for identifying transacting factors that mediate synaptic expression, and for assaying nervϵderived factors (see for example, Usdin and Fischbach, 1986; Fontaine et al. 1986; New and Mudge, 1986; McMahan and Wallace, 1989) suspected to induce synaptic differentiation.

This work was supported by grants from NIH and MDA to John P. Merlie and Joshua R. Sanes. We thank J. Peschon and R. Palmiter for generously providing the pnLacZ vector in advance of publication; and M. Donoghue, K. Gundersen, J. Lichtman, P. Taghert, and M. Velleca for comments.