Electron Microscopic Analysis of Membrane Assemblies Formed by the Bacterial Chemotaxis Receptor Tsr

The serine receptor (Tsr) from Escherichia coli is representativeof a large family of transmembrane receptor proteins that mediatebacterial chemotaxis by influencing cell motility through signaltransduction pathways. Tsr and other chemotaxis receptors formpatches in the inner membrane that are often localized at thepoles of the bacteria. In an effort to understand the structuralconstraints that dictate the packing of receptors in the planeof the membrane, we have used electron microscopy to examineordered assemblies of Tsr in membrane extracts isolated fromcells engineered to overproduce the receptor. Three types ofassemblies were observed: ring-like "micelles" with a radialarrangement of receptor subunits, two-dimensional crystallinearrays with approximate hexagonal symmetry, and "zippers," whichare receptor bilayers that result from the antiparallel interdigitationof cytoplasmic domains. The registration among Tsr moleculesin the micelle and zipper assemblies was sufficient for identificationof the receptor domains and for determination of their contributionsto the total receptor length. The overall result of this analysisis compatible with an atomic model of the receptor dimer thatwas constructed primarily from the X-ray crystal structuresof the periplasmic and cytoplasmic domains. Significantly, themicelle and zipper structures were also observed in fixed, cryosectionedcells expressing the Tsr receptor at high abundance, suggestingthat the modes of Tsr assembly found in vitro are relevant tothe situation in the cell.

INTRODUCTION

Top

Introduction

The serine receptor (Tsr), one of four methyl-accepting chemotaxisproteins (MCPs) that span the inner membrane of Escherichiacoli, initiates responses and governs adaptation to changesin the serine concentration. MCPs belong to a large class oftransducers (21, 46), which sense a variety of environmentalcues and are the inputs to sensory pathways that bias cell movementtoward favorable environments (12). The chemotaxis pathwaysbelong to the two-component superfamily of signal transductionpathways (17, 42), which are chiefly found in prokaryotes. Atwo-component pathway consists of a sensor, which is frequentlyan integral membrane protein possessing kinase activity, andone or more cytoplasmic phosphate-accepting response regulatorproteins. The transmembrane sensor-kinases of the chemotaxispathways are often noncovalent complexes between MCPs (whichhave no enzyme activity) and two soluble cytoplasmic proteins,namely, an adaptor protein (CheW) and a kinase (CheA) (15, 39).

Elucidation of the structure and distribution of receptors inthe membrane of the cell is integral to understanding the molecularbasis of signaling by the transmembrane sensor (MCP-CheW-CheA)complexes. X-ray structure determination of the soluble domainshas clearly defined the dimeric organization of the 60-kDa receptorsubunits (19, 31, 45), and functional studies have helped toelucidate the role of dimer organization in the mechanism oftransmembrane signaling (references 32 and 12 and referencestherein). Figure 1 summarizes the relationship between X-raystructure data obtained with soluble receptor fragments andan atomic model of the intact receptor and its functional domains.The dimeric organization of the periplasmic ligand-binding domain,which is evident in the X-ray structure, also places the ligand-bindingpocket at the dimer interface (45). The structure of the cytoplasmicdomain, which interacts with the signaling proteins and containsthe sites methylated during adaptation, has proven to be anextended coiled-coil hairpin that also forms dimers and packsin the crystal as a trimer of dimers (19). Using these structures,Kim et al. have constructed a plausible model of the intactreceptor dimer (19, 20). The model is largely consistent withnumerous site-directed sulfhydryl cross-linking studies (reviewedin reference 13) that provide structural information in regionswhere high-resolution data are missing, notably in the transmembraneregion, and in the flexible linker that joins the second transmembranehelix to the cytoplasmic domain (6).

View larger version (38K):
[in this window]
[in a new window]
FIG. 1. An atomic model of the bacterial chemoreceptor dimer (Kim et al. [19]; middle illustration), based on high-resolution X-ray structures (left illustration) of the aspartate receptor (Tar) ligand-binding domain and the Tsr cytoplasmic fragment, which are represented here with structures of the Salmonella Tar ligand-binding domain (45) and the E. coli Tsr cytoplasmic domain fragment (19) (PDB accession numbers 1vlt and 1qu7, respectively.) The two subunits in the dimers are shown in blue and red ribbon representations. The cytoplasmic domain fragment structure also displays two more dimers (blue-green and gray stick representations), which appear in the trimer-of-dimers arrangement in the crystal. The illustration of the receptor dimer (right), in which the two subunits are different shades of gray, depicts the functional domains of the receptor.

Although numerous biochemical studies implicate the involvementof extended interactions among receptor subunits in the membraneduring signaling (i.e., interactions among receptor dimers),direct structural evidence of these interactions is sparse.Biochemical properties of the signaling system that supporta role for such extended interactions include (i) clusteringof soluble cytoplasmic domain receptor fragments (either aloneor in complexes with the kinase) (14, 28, 29), (ii) transmethylationof receptor dimers (22, 26, 44), (iii) the stoichiometry ofreceptor-CheW-CheA complexes and the cooperativity of ligand-mediatedinhibition of kinase activity (5, 24, 25), and (iv) the interdependencyof signaling and response sensitivity among receptors of differentligand specificities (1, 16). These data provide the impetusto obtain direct structural information of intact receptorsin membranes. Electron and light microscopy studies conductedso far localize the receptors and cytoplasmic signaling proteinsat the poles of the E. coli cell (7, 30, 41), and although theserepresent an important step in defining the properties of receptorcomplexes in the cell, the resolution is insufficient for determinationof the disposition of individual subunits. Consequently, littleinformation is available that defines, for example, the orientationof the cytoplasmic domains with respect to the plane of themembrane or the lateral distribution of subunits.

To obtain a better understanding of the arrangement of the receptorsin the cell membrane, we have carried out an electron microscopicstudy on isolated receptor-containing membrane preparationsand on cryosections of an antibody-labeled E. coli Tsr overexpressor.Images recorded from negatively stained specimens of the membraneextracts are sufficiently well defined to generate plausibleinterpretations of receptor subunit orientation and intersubunitarrangements. Although some of these interactions are consistentwith the known packing of the soluble domains, others are novel.Moreover, arrangements of receptor subunits similar to thoseobserved in the membrane extracts also appear to be presentin whole cells, suggesting that such interactions have physiologicalrelevance.

MATERIALS AND METHODS

Top

Materials and Methods

Bacteria strains, plasmids, and biochemical reagents. E. coli strain RP437 is wild type for chemotaxis (35). The plasmidspHSe5.tsr_QEQE and pHSe5.tsr_QQQQ (where QEQE and QQQQ denotethe amino acid composition at the sites of methylation [Tsrresidues 297, 304, 311, and 493]) direct IPTG (isopropyl-ß-D-thiogalactopyranoside)-regulatedexpression of E. coli Tsr in wild-type and fully modified forms,respectively (37). The E. coli strain HCB721 [relevant genotype: ${Delta}$ tsr(7021) trg::Tn10 ${Delta}$ (cheA-cheY)::XhoI(Tn5) (9)] does not expressthe receptors Tar, Tsr, Trg, and Tap or the cytoplasmic signalingproteins CheA and CheW (found in receptor complexes) or CheRand CheB (the enzymes involved in altering the level of receptorcovalent modification). HCB721 transformed with pHSe5.tsr maintainsoverproduced Tsr in the genetically coded state of covalentmodification. The detergents, Tween 80, octyl glucoside (OG),and nonylglucoside (NG), were obtained from Sigma-Aldrich (St.Louis, Mo.) and Anatrace (Maumee, Ohio).

Preparation and electron microscope analysis of inner membrane samples. Inner membranes containing Tsr_QEQE were prepared from HCB721/pHSe5.tsr_QEQEby osmotic shock, isolated on sucrose gradients as describedpreviously (27, 34), and stored at -80°C in pH 8.0 buffer(10 mM Tris HCl, 5% [wt/vol] glycerol, 1 mM phenylmethylsulfonylfluoride, 1 mM EDTA). The inner membranes were treated withdetergent in the manner described by Corless et al. (10) priorto electron microscopic analysis. Although a range of conditions(noted in parentheses as follows) were tested, detergent treatmenttypically consisted of a 4-h incubation (2 to 20 h) of 5 µMTsr (3 to 30 µM) at 20°C with a Tween 80 concentrationof 0.3% (wt/vol) (0.03 to 2.0%) and sometimes also OG or NG(0.001 to 0.1%) in a pH 7.5, 50 mM Tris buffer with 100 mM NaCl,10% (wt/vol) glycerol, 1 mM EDTA, and 1 mM AEBSF [4-(2-aminoethyl)-benzenesulfonyl fluoride]. No degradation of Tsr was observed on sodiumdodecyl sulfate (SDS)-polyacrylamide electrophoresis gels followingthese treatments. To determine the extent to which the detergenttreatment extracted lipid and protein from the membrane, somesamples were subjected to centrifugation (30 min at 180,000x g) and the protein and phospholipid contents in the supernatantand pellet fractions were measured by the Lowry (Bio-Rad DCprotein assay kit; catalog no. 500-0111) and phosphate (8) assays,respectively. The fraction of Tsr in the pellet and the supernatantfollowing detergent treatment was also estimated by densitometricanalysis of Coomassie-stained SDS gels. For electron microscopy,the preparations were deposited on glow-discharged, carbon-coatedgrids. At 30 s after the application of a membrane sample (3µl), grids were prepared for microscopy by two (45 s)washes in detergent-free, low-ionic-strength Tris buffer, stainedin 1% uranyl acetate (10 s), and blotted dry. Images were recordedusing a Tecnai 12 electron microscope operating at 120 kV andequipped either with a tungsten filament or an LaB₆ crystaland a Gatan 2k x 2k charge-coupled device camera.

Cell growth, fixation, and immunogold electron microscopy of frozen sections. Overnight T-broth (1% tryptone, 0.5% NaCl) cultures (2 ml) wereinoculated with colonies from freshly grown LB plates, whichhad been streaked with cells from frozen permanents of HCB721/pHSe5.tsr_QQQQ.T-broth cultures (50 ml) were inoculated with 50 µl ofthe overnight culture and were grown at 35°C until theyentered exponential phase (optical density at 650 nm of $~$ 0.1),at which time tsr expression was induced with 1 mM IPTG. Cellswere grown for 2 h, harvested, and then fixed at room temperaturefor 2.5 h in a mixture of 2% paraformaldehyde and 0.2% glutaraldehydein PHEM buffer {120 mM PIPES [piperazine-N,N'-bis(2-ethanesulfonicacid)], 50 mM HEPES [pH 6.9], 4 mM MgCl₂, 20 mM EGTA}. The fixedcells were collected by centrifugation, resuspended in a minimumof PHEM buffer, and gently mixed with an equal amount of moltengelatin (2%) at 37°C. After solidifying at room temperature,the cell-containing gelatin pellets were cut into 1-mm cubesand infiltrated with a 2.3 M sucrose solution in 0.1 M sodiumphosphate buffer (pH 7.4). Cubes of gelatin were frozen to thesurface of special aluminum pins by being plunged into liquidnitrogen. By using an ${alpha}$ -Tsr antibody directed against the highlyconserved signaling domain of Tsr (1), sections of the frozencubes were cut in a freezing ultramicrotome at -100°C andimmunogold labeled as described previously (36). Sections wereexamined as described above for the membrane preparations.

RESULTS

Top

Results

The detergent-treated inner membranes isolated from E. colicells induced to overexpress the wild-type form of Tsr (Tsr_QEQE)were observed by conventional transmission electron microscopyof negatively stained specimens. Regular assemblies of receptorsubunits were found; Fig. 2 presents a collection of the assembliesseen in a large membrane aggregate. Three distinct modes ofassembly can be identified, which are referred to as crystalline(c), micellar (m), and zippered (z) by virtue of the visualsimilarity each bears to its namesake. Hybrid structures (e.g.,micelle-zipper and micelle-crystallite assemblies) were alsoobserved, which serve to indicate how the different forms joinand/or interconvert. We attribute these structures to the Tsrprotein, since they are not observed in membranes isolated fromcontrol cells from which the major chemoreceptor genes (tar,tsr, trg, and tap) were deleted and since SDS-polyacrylamidegel electrophoresis analysis of the inner membrane protein compositionisolated from the overexpressor demonstrates that Tsr proteinsare present in an overwhelming majority. Treatment with Tween80 solubilized ca. 65% of the membrane phospholipid and lessthan 10% of Tsr, but as a consequence, the large membrane aggregateswere often separated so that the crystalline, micellar, andzippered structures were observable as individual entities.The orientation and gross conformation of the Tsr subunits arereadily apparent in the separated assemblies. As described below,observations of the three types of subunit assembly allow usto conclude that (i) there is significant agreement betweenthe overall structure of the Tsr subunit in the membrane samplesand the model of the chemoreceptor dimer (judged on the basisof X-ray structure data of the periplasmic and cytoplasmic domains)(Fig. 1) (19) and (ii) the packing interactions among Tsr dimersin the membrane assemblies, while compatible with the trimer-of-dimerarrangement, also display a novel and specific interdigitatinginteraction in the zipper-like assembly.

View larger version (168K):
[in this window]
[in a new window]
FIG. 2. Electron micrograph from a negatively stained inner-membrane sample (isolated from an E. coli strain that overexpresses Tsr_QEQE) which was treated with 0.3% Tween 80. Zippers (z), micelles (m), and crystalline patches (c), which are three types of structures that are characteristic of these preparations, can be seen in this micrograph. Scale bar, 100 nm.

Ordered two-dimensional assemblies. Figure 3a, an electron microscope image of a negatively stainedspecimen of detergent-treated Tsr inner membrane, displays twoexamples of ordered receptor arrays connected to micellar structures.The arrays in Fig. 3a can be plausibly interpreted as projectionsof membrane patches viewed from a perspective perpendicularto the plane of the membranes. Although the arrays were limitedin extent ( $~$ 10³ Å), it was possible to conduct a preliminaryanalysis of the subunit arrangement in the membrane by generatinga Fourier transform (Fig. 3b) of the boxed region in Fig. 3a,which was found to be an approximately hexagonal lattice witha unit cell dimension of 75 Å and a corresponding unitcell area of $~$ 5,000 Å². An estimate of the number of Tsrreceptor subunits that is compatible with this unit cell areais based on the following reasoning. The average cross-sectionalarea observed in a well-ordered two-dimensional protein crystalfor an ${alpha}$ -helix perpendicular to the membrane plane is $~$ 180 Å²(18). Since the ligand-binding domain of the Tsr receptor subunitis a bundle of four helices (Fig. 1) that are likely to be nearlyperpendicular to the membrane, each receptor subunit has anapproximate cross-sectional area of 800 Å² in looselypacked crystals like those observed in our samples. Thus, eachunit cell is large enough to accommodate six (or fewer) Tsrmolecules, i.e., three dimers, which is consistent with thetrimer-of-dimer packing suggested by Kim et al. (20).

View larger version (87K):
[in this window]
[in a new window]
FIG. 3. (a) An electron micrograph of a negatively stained, ordered two-dimensional array of Tsr in a membrane sample that was incubated with detergent (Tween 80 = 0.08% [wt/vol], [OG] = 0.1% [wt/vol]) for 12 h. Scale bar, 100 nm. (b) The optical transform generated from the region of ordered Tsr molecules (corresponding to the boxed region in panel a) roughly corresponds to a lattice with constants a = $~$ 75 Å, b = $~$ 75 Å, and ${gamma}$ = $~$ 60°.

Micellar assemblies. In the crystal structure of the Tsr cytoplasmic domain (Fig.1), the trimer-of-dimers arrangement is due to subunit interactionsnear the coiled-coil hairpin. Similarly, in the micellar assembliespresented in Fig. 4, the cytoplasmic domains are analogous tothe spokes of a wheel, which connect at the hub through subunitinteractions that have a strong resemblance to the trimer-of-dimerinteraction. The resemblance is depicted in the enlarged imageof this micellar structure, shown at the right in Fig. 4, inwhich silhouettes of two dimers are placed at the hub. The cytoplasmicdomains radiate outward and are bounded by an annulus that isprobably composed of lipid, detergent, and receptor transmembranesegments. In some cases, the periplasmic domains can be discernedoutside the lipid ring, as expected from this packing arrangement.Characteristic radial dimensions were measured for several micellarstructures, including those shown in Fig. 4, and are summarizedas averages in angstroms ± standard deviations as follows:radius of hairpin contact (r_H), 52 ± 5 (n = 11); radiusat the membrane inner edge (r_I), 245 ± 13 (11); radiusat the membrane outer edge (r_O), 300 ± 16 (11); and outermostradius (r_P), 362 ± 32 (3). Estimates of the characteristicdimensions of the receptor, e.g., the end-to-end length, weredetermined as differences of two radial dimensions, e.g., r_P- r_H, on each micellar assembly individually and then averaged.The result of this analysis is presented in Table 1 for boththe micellar and (as discussed below) the zippered assemblies.Dimensions estimated from the electron micrographs of the twotypes of assemblies are in good agreement. The measured estimatefor the end-to-end length of the receptor (310 Å) is $~$ 20%shorter than the length of the receptor dimer model describedby Kim et al. (19).

View larger version (114K):
[in this window]
[in a new window]
FIG. 4. Nine panels at left: a gallery of isolated Tsr micellar assemblies. Scale bar, 50 nm. Right panel: An enlarged micellar assembly of Tsr, illustrating how characteristic radial dimensions are determined (in this case, r_P). Also superimposed on this assembly are scaled silhouettes of two Tsr cytoplasmic domain dimers, which are patterned after the trimer-of-dimer organization of subunits observed in the crystal structure of the Tsr cytoplasmic domain fragment (residues 286 to 526 [19]) and represent about 70% of the C-terminal portion of the Tsr protein that is located in the cytoplasm. The individual dimers are rendered as blue and red cigar shapes in which the pair of dimers forms a V-shaped assembly. Scale bar, 25 nm.

View this table:
[in this window]
[in a new window]
TABLE 1. Receptor dimensions determined from micelle and zipper assemblies

Zipper assemblies. Two prominent features of the zippered assemblies shown in Fig.5 are the bilateral symmetry and the parallel striations perpendicularto the zipper axis; these features allow us to deduce that thezippered structure represents a bilayer of interdigitating receptors.The transmembrane region of the receptor serves to maintainthe alignment among the individual subunits, which lie alongsideone another in ribbon-like fashion. The alignment among thesubunits facilitates the identification of the receptor domainsand makes clear the correspondence to the domains in the receptordimer model (Fig. 1). The regularity of the overlap (65 ±6 Å [Table 1]) implies the existence of specific contactsbetween the cytoplasmic domains in opposing leaflets and givesrise to the double-layer arrangement. The cross-sectional areasof the periplasmic and cytoplasmic domains in the receptor dimerare determined largely by the number of ${alpha}$ -helices in each domain(eight and four helices, respectively). Interdigitation effectivelydoubles the close-packed area on the cytoplasmic side of themembrane (from four helices to eight), and as a result, thecross-sectional areas on the two sides of the membrane are expectedto be essentially equivalent, facilitating parallel alignmentof the opposing leaflets in the receptor bilayer (as depictedin the illustration in Fig. 5).

View larger version (195K):
[in this window]
[in a new window]
FIG. 5. Top seven panels: a gallery of isolated zippered assemblies. Scale bar, 50 nm. Lower left panel: An electron micrograph of an isolated Tsr bilayer specimen that is labeled in one leaflet of the twin bilayer membrane to illustrate the topology of the receptor subunits. p, periplasmic ligand-binding domain; b, membrane bilayer; c, cytoplasmic domain. Domain assignments in the other leaflet are apparent by symmetry. Two other features are labeled: a zone of interdigitation, or overlap (o), and a feature highlighted by stain that may represent the linker domain (l). Scale bar, 25 nm. Lower right panel: An illustration of a zipper assembly, serving to show only the "bilayer of bilayers" topology and the antiparallel nature of cytoplasmic domain interdigitation.

The interdigitating interaction may prove to be a feature commonto all MCP assemblies of this type, owing to the high degreeof sequence conservation near the hairpin (21). Electron microscopicobservations of two other MCPs are consistent with the presenceof an interdigitating interaction. First, the periodic striationsacross the Tsr zippered assemblies resemble the striations seenin reconstituted lamellar microcrystals of the E. coli ribose/galactosereceptor (Trg) (4). Second, the arrangement of the Salmonellaaspartate receptor (Tar) cytoplasmic domain within soluble supramolecularsignaling complexes (that also contain CheW and CheA [14]) iscompatible with a similar type of interdigitating interaction.

Measurements of the other dimensions in representative zipperedassemblies of Tsr, including an estimate of the probable end-to-endlength of the receptor, are provided in Table 1. As in the caseof micellar assemblies, the measured receptor length is $~$ 20%shorter than that described for the model of Kim et al. (19)(Fig. 1). Given the known limitations for the interpretationof biological macromolecular structure from negatively stainedelectron microscope specimens, it is difficult to assert withcertainty the reason for the difference in the two estimates,but based on several pieces of information, a more compact arrangementof the polypeptide in the linker region (Fig. 1) provides aplausible explanation for the discrepancy between the measuredend-to-end length and the end-to-end length predicted in themodel. First, detailed structural information about the linkeris unavailable, although it is known by a conserved sequencemotif (2) and an amphipathic ${alpha}$ -helical character (6). Second,while the model plausibly depicts the linker as an extensionof the second transmembrane ${alpha}$ -helix (given the available structuralinformation), it is a relatively uncompact conformation thatreduces the cross-section in this portion of the receptor dimerto that of two ${alpha}$ -helices. Third, the zippered assemblies depictedin Fig. 5 often show a region highlighted by the stain at alocation in the cytoplasmic domain that is consistent with theposition of the linker domain in the primary sequence of MCPs(Fig. 1) (2). Finally, a comparison of the measured (Table 1)versus model-generated lengths (19) of the periplasmic, transmembrane,and cytoplasmic domains shows that the agreement is poorestbetween the two lengths estimated for the cytoplasmic domain(190 Å [Table 1] versus 260 Å [19]).

To further assess whether or not this difference is meaningful,we tested for specimen shrinkage under our experimental conditionsby measuring the accurately known repeat dimension in the tailstriations of T4 phage (33). These experiments showed that theshrinkage of the tail striation repeat was no more than $~$ 5% (datanot shown), although we cannot exclude the possibility thatthere is a different extent of shrinkage in the Tsr membraneassembly specimens. Taking all of these observations together,we suggest that the highlighted feature in the zippered assemblies(Fig. 5) reflects a more compact structure in the polypeptidechain than the extended ${alpha}$ -helix depicted in the model. Apartfrom this one (but potentially important) difference, the dimensionsdetermined from the electron micrographs are in reasonable agreementwith the dimensions of the model, lending support both to themodel and to our interpretation of the subunit arrangementsin the electron microscope images.

In situ imaging in cell sections. To determine whether the assemblies observed in the extractedmembranes also occur in the intact cell, we recorded imagesfrom E. coli wild-type and Tsr overexpressor strains which hadbeen fixed with glutaraldehyde, sectioned under cryogenic conditions,and labeled with an ${alpha}$ -Tsr antibody that binds to the highly conservedcytoplasmic signaling domain. Projection images of wild-typeE. coli sections (data not shown) exhibited specific labelingof receptors localized mainly at the cell poles, in agreementwith previous observations (30). In the Tsr overexpressor, receptorlabeling is distributed around the periphery of the cell andlabeled regions also project into the interior of the cell (Fig.6), suggesting that the cell becomes filled with membranousstructures. Moreover, the images from the Tsr overexpressorclearly possess features that bear a striking resemblance tothe zippered and micellar assemblies seen in the isolated membranepreparation (Fig. 6), which demonstrates that the assembliesobserved in vitro have relevance for receptor subunit interactionsin vivo. A detailed investigation (using electron microscopetomography) of the three-dimensional arrangement of zipperedand micellar assemblies in the cell is to be presented elsewhere(J. Lefman et al., submitted for publication).

View larger version (70K):
[in this window]
[in a new window]
FIG. 6. Images recorded from stained 70-nm sections of the E. coli Tsr overexpressor (HCB721/pHSe5.tsr_QQQQ induced with 1 mM IPTG) after immunogold labeling of Tsr (10-nm gold particles). Regions identified with zippered (z) (right panel) and micellar (m) (left panel) assemblies are indicated. Scale bar, 200 nm.

DISCUSSION

Top

Discussion

The work described here demonstrates that membranes containingTsr can arrange themselves into a variety of structures withstrong interactions between neighboring subunits and, in thecase of the zippered assemblies, nonlocal interactions betweenmembranes. In other examples of E. coli membrane protein overexpression(3, 11, 43), extended membrane structures form that are enrichedin the overexpressed protein. Often, these protein-lipid assembliesare long cylinders, an organization that is compatible withproteins in which the major portion of the polypeptide is embeddedwithin the lipid bilayer. In contrast, a prominent feature ofthe Tsr-containing membrane preparations is the extensive interactionamong the extramembranous portions of the receptor subunits,in particular the cytoplasmic domains. Many of these interactionsare compatible with the subunit interactions in the X-ray structuresof the soluble domains (19, 45). Specifically, the hexagonalarrangement of subunits in ordered arrays and the hairpin contactsobserved in micellar assemblies are consistent with the trimer-of-dimerinteraction found in the crystal structure of the Tsr cytoplasmicdomain.

The interdigitating cytoplasmic domain in the zippered assembliesrepresents a novel subunit interaction that may be common toall MCPs, since it occurs within a region of high sequence identity(21) and it is compatible with electron microscope observationson detergent-solubilized, reconstituted Trg microcrystals (4)and soluble signaling complexes containing the Tar cytoplasmicfragment (14). The occurrence of interdigitation in cryosectionsof the Tsr overexpressor strain demonstrates that this typeof interaction can also occur in the cell when Tsr is expressedat a high level. The extent to which the interactions observedwith the elevated Tsr expression levels reflect the behaviorof receptors at normal (wild-type) levels of expression remainsto be determined. This issue has relevance to biochemical investigationsof receptor function, which are often conducted either withcells that have elevated levels of receptor or with membranesisolated from these cells (5, 15, 22-27).

The notion that the receptor protein in the overexpressed stateis functional in vivo has support from the observation thatat moderate levels of Tar overexpression (but normal levelsof the cytoplasmic signaling proteins), the increase in theadaptation time of a cell in response to an aspartate stimulusis approximately commensurate with the degree of overexpression(38). This property is consistent with the interpretation thatin the overexpressed state, the majority of the receptors inthese cells can be methylated. Still, the relationship betweenoverexpression and biochemical function is complex, particularlywith respect to receptor-mediated kinase activation and regulation,in which the subunit stoichiometry in the receptor-kinase complexmay have a significant influence on the signaling propertiesof the complex. With the growing appreciation of the importanceof receptor arrays in signaling (1, 16, 20, 23, 40), these electronmicroscopic observations contribute to and illustrate an approachfor the further development of our understanding of the architectureof signaling complexes in the cell.

ACKNOWLEDGMENTS

We thank John S. Parkinson (University of Utah) for the giftof Tsr antisera, Erik Bos (Netherlands Cancer Institute, Amsterdam)for assistance in preparing the E. coli cryosections, Sung-HoKim (University of California at Berkeley) for the Tsr dimermodel coordinates, and David DeRosier (Brandeis University)for helpful discussions.

This work was supported by grants to R.M.W. from the NIH (RO1NIGMS53210) and to S.S. from the intramural program at the NationalCancer Institute, NIH.

FOOTNOTES

* Corresponding author. Mailing address for Robert M. Weis: Department of Chemistry, LGRT 701, 701 North Pleasant St., Amherst, MA 01003-9336. Phone: (413) 545-0464. Fax: (413) 545-4490. E-mail: rmweis{at}chem.umass.edu. Mailing address for Sriram Subramaniam: Laboratory of Biochemistry, National Cancer Institute, National Institutes of Health, Bethesda, MD 20817. Phone: (301) 594-2062. Fax: (301) 480-3834. E-mail: ss1{at}nih.gov.

REFERENCES

Top