Tubulin binds specifically to the signal-transducing proteins, Gs alpha and Gi alpha 1.

Participation of cytoskeletal elements in regulation of hormonal response and responsiveness has been suggested by several laboratories. Addition of dimeric tubulin to rat cerebral cortex synaptic membranes causes stable inhibition of adenylyl cyclase, and the molecular basis for this effect appears to require a direct interaction between tubulin and G proteins. To test whether such tubulin-G protein interaction occurred, several purified G proteins were bound to nitrocellulose, and 125I-tubulin overlay studies were performed. 125I-Tubulin bound to the alpha subunits of Gs and Gil with high specificity and an apparent Kd of approximately 130 nM. Other G protein alpha subunits (alpha i2, alpha i3, alpha 0, and transducin) displayed a much lower affinity for tubulin, despite the much closer relationship of those proteins to alpha il than to alpha s. Association of beta gamma subunits with alpha il or alpha s did not alter the binding of tubulin to these G protein heterotrimers, and the binding of a hydrolysis-resistant GTP analog to the alpha subunits was similarly without effect. These results suggest that tubulin forms complexes with specific G proteins and these complexes might provide a locus for the interaction of cytoskeletal components and signal transduction cascades. These results also provide evidence of a functional distinction among the closely related alpha i subtypes.


Participation
of cytoskeletal elements in regulation of hormonal response and responsiveness has been suggested by several laboratories. Addition of dimeric tubulin to rat cerebral cortex synaptic membranes causes stable inhibition of adenylyl cyclase, and the molecular basis for this effect appears to require a direct interaction between tubulin and G proteins.
To test whether such tubulin-G protein interaction occurred, several purified G proteins were bound to nitrocellulose, and 1261-tubulin overlay studies were performed.
'251-Tubulin bound to the (Y subunits of G. and Gil with high specificity and an apparent Kd of approximately 130 nM. Other G protein a! subunits (ai2, ai3, (rot and transducin) displayed a much lower affinity for tubulin, despite the much closer relationship of those proteins to ail than to (Y.. Association of & subunits with ail or (Y. did not alter the binding of tubulin to these G protein heterotrimers, and the binding of a hydrolysis-resistant GTP analog to the cr subunits was similarly without effect. These results suggest that tubulin forms complexes with specific G proteins and these complexes might provide a locus for the interaction of cytoskeletal components and signal transduction cascades. These results also provide evidence of a functional distinction among the closely related ei subtypes. In a variety of transmembrane signaling systems, cell surface receptors are linked to effecters by GTP-binding (G) proteins which function as signal transducers (1). Several of these proteins have been described, and their classification was (and is) based, primarily, on the effector molecules which each species of G protein appeared to engage. Originally, these effecters included adenylyl cyclase (stimulated by G, and inhibited by Gi) and rod outer segment phosphodiesterase * This work was suunorted bv United States Public Health Service Grant MH 39595 and National Science Foundation Grant BNS 87-19758. The costs of publication of this article were defraved in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. (activated by transducin (G,'). Although the family of signaltransducing G proteins had been considered to be a small one (including the above mentioned proteins plus G,, a protein of unknown function which is abundant in mammalian brain), current evidence suggests that several subspecies of these proteins exist. Recent advances in molecular cloning of cDNAs for (Y (GTP-binding) subunits of G proteins have revealed four forms of (Ye (produced by alternative splicing of a single gene (2, 3) and three forms of (Y~ (4, 5). Concurrent with the discovery of increasing numbers of G protein subtypes has been the assignation of new and varied physiological functions for these molecules. For Gi proteins, these roles include the gating of K' channels as well as the inhibition of adenylyl cyclase. Despite the clear demonstration of molecular heterogeneity among the ai protein isoforms, no distinction with respect to physiological roles has been demonstrated. Tubulin is also a GTP-binding protein, and significant functional similarity and amino acid sequence homology appear to exist between tubulin and the signal-transducing G proteins (6)(7)(8). Functional interaction between tubulin and synaptic membrane G proteins has also been demonstrated. In those studies, incubation of Gpp(NH)p-liganded dimeric tubulin with rat cerebral cortex synaptic membranes resulted in inhibition of adenylyl cyclase which persisted subsequent to washing of those membranes.
This effect appeared to involve the transfer of guanine nucleotide from tubulin to Gi, and this process was thought to represent a unique feature of the regulation of neuronal adenylyl cyclase (9). These data suggested that tubulin molecules interacted physically with G proteins. Further suggestions of such interactions included observations that tubulin bound to a G protein affinity column and, as G,, to a tubulin affinity column (10, 11). The current study was undertaken in an effort to elucidate the molecular details of tubulin-G protein interaction. In this study, the formation of tubulin-G protein macromolecular complexes is demonstrated.
It appears that the ability to form these complexes reveals a clear distinction among the subtypes of (Y~.  Tub&n-G Protein Interactions Tubulin was prepared by two cycles of polymerization from rat brain (15), and microtubule-associated proteins were removed by phosphocellulose chromatography (16). Tubulin was iodinated with IODO-GEN to a sDecific activitv of 1800 Ci/mmol. and functional integrity of '251-tudulin was verif;ed by copolymerizstion with unlabeled freshly depolymerized (second cycle) tubulin (containing microtubule-associated proteins). Under those conditions, 91% of the freshly depolymerized (control) tubulin repolymerized and 88.5% of the iodotubulin repolymerized. Thus, relative to the controls, the iodotubulin used in these experiments was 97% polymerization-competent. Purity of the 1251-tubulin was verified by silver staining and autoradiography. When 90 yg were run on 8.5% SDS-polyacrylamide gel electrophoresis, no impurities were detected by staining or autoradiography.
Photoaffinity Labeling-Membranes were incubated with [3ZP]p3-1,4-azidoanilido-p'-5'-GTP [AAGTP] or unlabeled AAGTP (0.12 NM of either compound) for 3 min at 23 "C, washed, resuspended, UVirradiated (for 5 min), and subjected to SDS-polyacrylamide gel electrophoresis as described (17). Samples (100 ~1) of partially purified G pro&ins containing approximately 5 pg of G protein in a buffer consisting of 50 mM Tris, pH 8.0, 1 mM MgC12, 1 mM EDTA, 1 mM dithiothreitol, 150 mM NaCl, and 0.6% Lubrol PX were incubated under yellow light in a 1.8-ml microcentrifuge tube with 0.1 GM ["'PI AAGTP  at 23 "C for 20 min. The mixture was then UVirradiated on ice for 5 min with a g-watt Mineralight at a distance of 3 cm. The reaction was quenched by adding dithiothreitol to a final concentration of 5 mM. Samples were loaded on a l-ml P6 desalting column (from Bio-Rad) which was equilibrated with Tris buffer. The unbound nucleotide was removed by ultrafiltration, and the protein soiution (-90% of proteins were recovered) was used in the immunoprecipitation experiments.

Immunoblotting and Tub&n
Overlays-After electrophoresis, proteins were transferred to nitrocellulose (PH 75, 0.1~pm pore size; Scheicher and Schuell) by washing in blotting buffer (25 mM Tris, 192 mM glycine, pH 7.6) and transferred at 30 mA for 12 h at 4 "C. Transfer efficacy was verified by silver staining of the dried gels. Nitrocellulose was incubated at room temperature for 2 h in 4% bovine serum albumin (BSA) blocking buffer containing 10 mM sodium phosphate, pH 7.0.150 mM NaCI, 1 mM EDTA, 1 mM sodium azide, and 012% T&on X-100. Nitrocellulose strips were then incubated with iodinated '251-tubulin or '251-ovalbumin (1.3 up/l0 ml unless indicated) at room temperature for 2 h. After repeatehwashing of the nitrocellulose with blocking buffer, nitrocellulose strips were dried and autoradiography was performed using Kodak XAR film. Dot Blotting-Purified G proteins (obtained from the indicated sources) were applied to nitrocellulose in the amounts indicated in the figure legends. Quantification of proteins bound after washing was performed by applying 100 ng of iodonated tubulin, ovalbumin, or 01, to nitrocell;lose. These proteins (each at a concentration of 0.5 ualul) were in a 100 mM PIPES buffer containing 2 mM EGTA. 1 II. mM MgC12, pH 7.4, and one of the following detergents: Lubrol PX, Triton X-100, SDS, CHAPS, sodium cholate (each 0.5%), or no detergent. Nitrocellulose was subjected to the washing procedure described above. Protein remaining on the nitrocellulose was quantified in an LKB Rackgamma counter. In each case, regardless of protein or detergent used, about 40% (39.8% average of two triplicate experiments for each protein) of the applied protein remained on the nitrocellulose after the washing steps.
Immunoprecipitotion-Labeled G proteins (0.5 pg) were incubated with 10 rg of purified tubulin at 23 "C for 20 min followed by the addition of indicated amounts of anti-tubulin antiserum (ICN 65 095) and incubation at 23 "C for another 20 min. After that, 15 ~1 of Staphylococcus aweus cell wall solution (Pansorbin, Calbiochem) were added, and the antibody was collected by centrifugation. The pellet was washed twice, dissolved in 2% SDS sample buffer containing 0.15 M dithiothreitol, and boiled for 5 min. The supernatant was recentrifuged and subjected to SDS-polyacrylamide gel electrophoresis in 10% gels and autoradiography. Heterotrimeric (unlabeled) G -_ _ proteins were incubated and immunoprecipitated under the same conditions. Anti-G6 antibodv (suonlied bv Dr. Yee-Kin Ho. University of Illinois, Chicago) ani iZ51$otein"A (Amersham C&p.) was used to detect heterotrimeric G protein complexes.

RESULTS
Tubulin Binding Proteins in Synaptic Membranes-Initial attempts to demonstrate tubulin-G protein interaction involved SDS-polyacrylamide gel electrophoresis of rat cerebral cortex membrane proteins and transfer of those proteins to nitrocellulose. Subsequent incubation of that nitrocellulose with 1251-tubulin showed 1251-tubulin binding to eight proteins, four of which had electrophoretic mobilities corresponding to proteins labeled covalently by the photoaffinity GTP analog, AAGTP, and ADP-ribosylated by cholera or pertussis toxin (17). However, a number of subspecies of (Y~ exist; these proteins and o(, are not separated well in SDS gel electrophoresis of synaptic membrane preparations. Furthermore, apparent co-migration in one dimension does not provide sufficient evidence to implicate G proteins as the targets of tubulin binding. Thus, it was necessary to examine the ability of tubulin to bind to purified G proteins.

Tub&in
Binding to Specific G Proteins-Tubulin bound preferentially to (Y,(recombinant,) and ail but only slightly to ai2, ais, LY,, and (Ye (Fig. 1). Recombinant ail and ail purified from porcine brain bound '251-tubulin to a similar extent. Furthermore, both activated (GTPTS-liganded) and inactive (GDP-liganded) forms of ail, oi2, and (Y~ were tested. This did not alter the ability of '251-tubulin to bind (or not bind) to these proteins. The binding of tubulin to G, was specific, as either excess cold tubulin or denaturation of G, by boiling abolished the binding of tubulin. Although tubulin did not bind to isolated /3r subunits on nitrocellulose, Gil (the a/3r heterotrimer) bound tubulin with affinity similar to that of ail, and heterotrimeric G, or G, were no more effective than their (Y subunits in the binding of tubulin (Fig. 2). Furthermore, whereas the (Ye used was a recombinant protein (expressed in E. coli), G, was purified from human erythrocytes (Fig. 2).

Affinity
and Stoichiometry of Tubulin-G Protein Interaction-The affinity of tubulin for ail was determined by immobilizing (Y,~ upon nitrocellulose and performing overlay experiments with varying concentrations of '251-tubulin. The binding was saturable and 86% specific at 0.5 nmol of tubulin. Calculated & values for tubulin binding to ai ranged from 110 to 140 nM, and immobilized ail bound 0.3-0.6 mol of tubulin/mol of ail (Fig. 3) 1. ?-Tubulin binding to G proteins. Purified and recombinant G protein cy subunits were dotted on a nitrocellulose sheet in the dilution series indicated. Proteins were dissolved in the buffer and concentrations indicated under "Experimental Procedures" and 0.5% Lubrol PX. The nitrocellulose sheet was then incubated and washed with blocking buffer followed by incubation with '251-tubulin (80 nM) as described in the text. The autoradiograph of the dried nitrocellulose is shown. Heterotrimeric G,l bound tubulin to the same molar extent as (Y,~. Similar results were obtained when nitrocellulose was incubated with different concentrations of '251-tubulin ranging from 30 to 800 nM. The "dot-blot" illustrated above is representative of one of six experiments combining G, subunits from different sources. '251-ovalbumin did not bind to the dotted G proteins, and boiled G protein failed to bind tubulin. Excess (400 X) unlabeled tubulin completely eliminated lz51 binding to G proteins. binding was assessed by overlay with ""I-tubulin and autoradiography as described in Fig. 1. H, purified G protein subunits were dotted on a nitrocellulosr sheet in the dilution series indicated (the protein amount of heterotrimeric G, roughly doubled that of (I! and $,) followed by overlay with ""I-tubulin and autoradiography as described in Fig. 1 whether these proteins could form complexes in solution. Therefore, immunoprecipitation was performed with an antitubulin ant,ibody which did not interfere with the tubulinmediated inhibition of adenylyl cyclase or the binding of tubulin to G proteins on nitrocellulose.
(yil formed complexes with tubulin and was immunoprecipitated only in the presence of tubulin (Fig. 4). Substitution of nonspecific antibody or elimination of tubulin from the incubation blocked (Y; immunoprecipitation. Substitution of chicken uterus myosin and rabbit anti-myosin (both from Dr. P. de Lanerolle, University of Illinois, Chicago) for tubulin and anti-tubulin resulted in precipitation of myosin but not (Y,, indicating that the formation of a precipitable immune complex which has no spe- (; proteins (G, and G,,) were partiall> purified from rat brain and labeled with ['"P]AA(;TP as described in the text. Labeled CI proteins (0.5 p,u) were incubated with 10 pp of purified tubulin in a total volume of 100 ~1 at 23 "C for 20 min followed by addition of indicated amounts of anti-tubulin antiserum (ICN 65 09,5,) and immunoprecipitation.
electrophoresis. and autoradiography as described in the text. cific interaction with N, will not precipitate (Y,. AAGTP-labeled (Y,,, although present in these preparations in amounts roughly equal to (yi, was not immunoprecipitated even in the presence of tubulin, implying that no complexes were formed between tubulin and N,,, a result consistent with the dot blotting experiments (see Fig. 2). Since (vi, immunoprecipitated with tubulin was prelabeled with ['"PIAAGTP before incubation with tubulin, it was assumed that this was dissociated from & subunits. A similar experiment was performed in the presence of 100 PM GTPrS, a condition which should dissociate all N, from By. In this experiment, tubulin and (Y, were immunoprecipitated with anti-tubulin, but @r subunits were not. To detect whether the t$y heterotrimers of Gi formed complexes with tubulin, unlabeled Gi was incubated with tubulin and subjected to immunoprecipitation with anti-tubulin as described above. Immunoprecipitated /3 was detected by anti-b antibodies. Since tubulin did not bind to /+y subunits of G proteins and purified pr subunits were not precipitated by anti-tubulin antibody (in the presence or absence of tubulin) those proteins detected by the specific anti-b antibody were a part of G,-tubulin complexes. DISCUSSION These data suggest that tubulin binds with high affinity to N, and mil but with much lower affinity to W, W, (Y,,, and N,.

1242
Tub&n-G Protein Interactions Given the very high (87-94%) amino acid sequence homology among the ai subtypes (1819) and the high (MO%) homology among the pertussis toxin substrates (i.e. (Y;, (Y,, and 01~ (20)) yet relatively low (-45%) homology between (Ye and the others, this finding is surprising. In fact, the binding of tubulin represents the first demonstration of a functional similarity between aB and one ai subtype which is not shared among the three ai subtypes. Some functional as well as structural differences among the Gi subtypes have been demonstrated. Gil and Giz have been copurified from porcine brain, and only the GTPyS-activated ail could inhibit membrane-bound adenylyl cyclase (21). Furthermore, an affinity-purified polyclonal antibody against ail did not recognize ai2, implying significant difference in the dominant immune determinants of ail and ai (22). Secondary structure analysis of the most diverse region of ai subtypes, amino acid residues 80-130, revealed different predicted secondary structures among these three ai subtypes (4). These structural differences may contribute to the binding specificity of tubulin to ai subtypes. It is noteworthy that tubulin binding represents the first demonstration of a functional distinction between ail and ai3.
The distribution of ai subtypes among mammalian tissues presents an intriguing correlation with the results presented in this report. ail is abundant in brain but less abundant in peripheral tissues (5, 23). The inhibitory effect of tubulin on adenylyl cyclase, which appears to involve transfer of a hydrolysis-resistant GTP analog from tubulin to ai, appears specific for neuronal tissues (9, 11). As the regulation of neuronal adenylyl cyclase differs in several important aspects (e.g. calmodulin dependence and loss of receptor activation subsequent to cell disruption) from that of other tissues, it is possible that membrane-associated tubulin dimers participate in the intracellular regulation of neuronal adenylyl cyclase. Such a process would participate in the modulation of neurotransmitter response (or responsiveness) and could provide a locus for second messenger interaction (7,9).
Tubulin itself is a GTP-binding protein with intrinsic GTPase activity and some structural homology with other GTP-binding proteins (8). Both tubulin and G proteins undergo GTP/GDP-dependent reversible complex formation, represented by microtubule formation in the case of tubulin and G& heterotrimer formation for the signal-transducing G proteins. Tubulin is also a substrate for both pertussis toxin-and cholera toxin-catalyzed ADP-ribosylation (24,25).
Unlike the signal-transducing G proteins, the primary biological role of tubulin is the formation of microtubules. During such an event, several "contact regions" on the tubulin dimer engage in a reversible association with other tubulin dimers. Given the various regions of homology extant between tubulin and G proteins, it is possible that (Ye and ail contain domains which are sufficiently "tubulin-like" that a reversible complex between these G proteins and tubulin is formed (7,9). Since the binding of /3r subunits to ail did not interfere with tubulin binding, it is likely that tubulin and & bind to (Y on separate domains.
While it is possible that a common sequence among G, subunits is capable of being recognized by tubulin, certain conformational features need be preserved for this recognition to occur. This argument is supported by the observation that boiling of ail abolished the ability of (nil to bind tubulin.
Recently, our laboratory suggested that there is a direct interaction and guanine nucleotide transfer between ai and (Y. and that this process is involved in the regulation of neuronal adenylyl cyclase (17). Functional complex formation between G, proteins has also been proposed as a mechanism to explain some of the kinetics involved in receptor-effector coupling (26). The interaction and nucleotide exchange between G proteins and between G protein and tubulin may represent a form of intracellular regulation of neuronal adenylyl cyclase. The significance of specific binding of tubulin to ail and (Ye may be due to the likelihood that it is these proteins which stimulate and inhibit adenylyl cyclase, and it is that enzyme which tubulin appears to modulate. Furthermore, this tubulin-G protein interaction represents a possible role for tubulin dimers which is independent of microtubule formation. The precise relationship between tubulin and G proteins as well as the contribution of tubulin and/or other cytoskeletal components to the regulation of signal transduction is not yet known. Nonetheless, the possibility that an ordered presentation of members of the adenylyl cyclase system and tubulin orchestrate cellular responses to external and internal signals remains intriguing.