The influence of developmental spinal stenosis on the risk of re-operation on an adjacent segment after decompression-only surgery for lumbar spinal stenosis.

AIMS: The aim of this study was to determine the influence of developmental spinal stenosis (DSS) on the risk of re-operation at an adjacent level. PATIENTS AND METHODS: This was a retrospective study of 235 consecutive patients who had undergone decompression-only surgery for lumbar spinal stenosis and had a minimum five-year follow-up. There were 106 female patients (45.1%) and 129 male patients (54.9%), with a mean age at surgery of 66.8 years (sd 11.3). We excluded those with adult deformity and spondylolisthesis. Presenting symptoms, levels operated on initially and at re-operation were studied. MRI measurements included the anteroposterior diameter of the bony spinal canal, the degree of disc degeneration, and the thickness of the ligamentum flavum. DSS was defined by comparative measurements of the bony spinal canal. Risk factors for re-operation at the adjacent level were determined and included in a multivariate stepwise logistic regression for prediction modelling. Odds ratios (ORs) with 95% confidence intervals were calculated. RESULTS: Of the 235 patients, 21.7% required re-operation at an adjacent segment. Re-operation at an adjacent segment was associated with DSS (p = 0.026), the number of levels decompressed (p = 0.008), and age at surgery (p = 0.013). Multivariate regression model (p < 0.001) controlled for other confounders showed that DSS was a significant predictor of re-operation at an adjacent segment, with an adjusted OR of 3.93. CONCLUSION: Patients with DSS who have undergone lumbar spinal decompression are 3.9 times more likely to undergo future surgery at an adjacent level. This is a poor prognostic indicator that can be identified prior to index decompression surgery.

A photic visual cycle of rhodopsin regeneration is dependent on Rgr.

During visual excitation, rhodopsin undergoes photoactivation and bleaches to opsin and all-trans-retinal. To regenerate rhodopsin and maintain normal visual sensitivity, the all-trans isomer must be metabolized and reisomerized to produce the chromophore 11-cis-retinal in biochemical steps that constitute the visual cycle and involve the retinal pigment epithelium (RPE; refs. 3-8). A key step in the visual cycle is isomerization of an all-trans retinoid to 11-cis-retinol in the RPE (refs. 9-11). It could be that the retinochrome-like opsins, peropsin, or the retinal G protein-coupled receptor (RGR) opsin12-16 are isomerases in the RPE. In contrast to visual pigments, RGR is bound predominantly to endogenous all-trans-retinal, and irradiation of RGR in vitro results in stereospecific conversion of the bound all-trans isomer to 11-cis-retinal. Here we show that RGR is involved in the formation of 11-cis-retinal in mice and functions in a light-dependent pathway of the rod visual cycle. Mutations in the human gene encoding RGR are associated with retinitis pigmentosa.

Interaction of 11-cis-retinol dehydrogenase with the chromophore of retinal g protein-coupled receptor opsin.

Vertebrate opsins in both photoreceptors and the retinal pigment epithelium (RPE) have fundamental roles in the visual process. The visual pigments in photoreceptors are bound to 11-cis-retinal and are responsible for the initiation of visual excitation. Retinochrome-like opsins in the RPE are bound to all-trans-retinal and play an important role in chromophore metabolism. The retinal G protein-coupled receptor (RGR) of the RPE and Müller cells is an abundant opsin that generates 11-cis-retinal by stereospecific photoisomerization of its bound all-trans-retinal chromophore. We have analyzed a 32-kDa protein (p32) that co-purifies with bovine RGR from RPE microsomes. The co-purified p32 was identified by mass spectrometric analysis as 11-cis-retinol dehydrogenase (cRDH), and enzymatic assays have confirmed the isolation of an active cRDH. The co-purified cRDH showed marked substrate preference to 11-cis-retinal and preferred NADH rather than NADPH as the cofactor in reduction reactions. cRDH did not react with endogenous all-trans-retinal bound to RGR but reacted specifically with 11-cis-retinal that was generated by photoisomerization after irradiation of RGR. The reduction of 11-cis-retinal to 11-cis-retinol by cRDH enhanced the net photoisomerization of all-trans-retinal bound to RGR. These results indicate that cRDH is involved in the processing of 11-cis-retinal after irradiation of RGR opsin and suggest that cRDH has a novel role in the visual cycle.

Expression of a recombinant human RGR opsin in Lentivirus-transduced cultured cells.

PURPOSE: Our goals were to produce a functional recombinant RPE retinal G protein-coupled receptor (RGR) opsin for biochemical studies and to test the efficiency of a lentiviral vector for transgene expression of human RGR. METHODS: A human RGR cDNA was cloned into a replication-defective lentiviral vector, and recombinant hRGR-Lentivirus was prepared for transduction of the ARPE-19, a human retinal pigment epithelium (RPE) cell line, and COS-7 cells. Recombinant RGR expression was detected by Western blot analysis, and functionality of the protein was tested by a [3H]all-trans-retinal binding assay. RESULTS: RGR protein was detected in each cell type after transduction with recombinant virus and was not observed in untreated cells. RGR expression in ARPE-19 cells increased steadily for up to 10 days after transduction and was stable for at least 6 months. The transduced ARPE-19 cells produced approximately 100-fold higher amounts of RGR protein than the transduced COS-7 cells. When cell membranes from the ARPE-19 cells were incubated with [3H]all-trans-retinal, the chromophore bound specifically to the expressed protein. Uptake of [3H]all-trans-retinol into the ARPE-19 cells was followed by specific binding of radiolabeled retinoid to RGR. CONCLUSIONS: Using a Lentivirus-derived gene delivery system, we were able to express high amounts of human RGR protein in the ARPE-19 human RPE cell line. The transduced ARPE-19 cells remain able to process all-trans-retinol, and the expressed protein is capable of binding to the all-trans-retinal chromophore. The Lentivirus-based expression of functional RGR can be used to study RGR in cultured cells and to test in vivo transduction of quiescent RPE cells.

Enamel biomineralization defects result from alterations to amelogenin self-assembly.

Enamel formation is a powerful model for the study of biomineralization. A key feature common to all biomineralizing systems is their dependency upon the biosynthesis of an extracellular organic matrix that is competent to direct the formation of the subsequent mineral phase. The major organic component of forming mouse enamel is the 180-amino-acid amelogenin protein (M180), whose ability to undergo self-assembly is believed to contribute to biomineralization of vertebrate enamel. Two recently defined domains (A and B) within amelogenin appear essential for this self-assembly. The significance of these two domains has been demonstrated previously by the yeast two-hybrid system, atomic force microscopy, and dynamic light scattering. Transgenic animals were used to test the hypothesis that the self-assembly domains identified with in vitro model systems also operate in vivo. Transgenic animals bearing either a domain-A-deleted or domain-B-deleted amelogenin transgene expressed the altered amelogenin exclusively in ameloblasts. This altered amelogenin participates in the formation an organic enamel extracellular matrix and, in turn, this matrix is defective in its ability to direct enamel mineralization. At the nanoscale level, the forming matrix adjacent to the secretory face of the ameloblast shows alteration in the size of the amelogenin nanospheres for either transgenic animal line. At the mesoscale level of enamel structural hierarchy, 6-week-old enamel exhibits defects in enamel rod organization due to perturbed organization of the precursor organic matrix. These studies reflect the critical dependency of amelogenin self-assembly in forming a competent enamel organic matrix and that alterations to the matrix are reflected as defects in the structural organization of enamel.

The endogenous chromophore of retinal G protein-coupled receptor opsin from the pigment epithelium.

The recent identification of nonvisual opsins has revealed an expanding family of vertebrate opsin genes. The retinal pigment epithelium (RPE) and Müller cells contain a blue and UV light-absorbing opsin, the RPE retinal G protein-coupled receptor (RGR, or RGR opsin). The spectral properties of RGR purified from bovine RPE suggest that RGR is conjugated in vivo to a retinal chromophore through a covalent Schiff base bond. In this study, the isomeric structure of the endogenous chromophore of RGR was identified by the hydroxylamine derivatization method. The retinaloximes derived from RGR in the dark consisted predominantly of the all-trans isomer. Irradiation of RGR with 470-nm monochromatic or near-UV light resulted in stereospecific isomerization of the bound all-trans-retinal to an 11-cis configuration. The stereospecificity of photoisomerization of the all-trans-retinal chromophore of RGR was lost by denaturation of the protein in SDS. Under the in vitro conditions, the photosensitivity of RGR is at least 34% that of bovine rhodopsin. These results provide evidence that RGR is bound in vivo primarily to all-trans-retinal and is capable of operating as a stereospecific photoisomerase that generates 11-cis-retinal in the pigment epithelium.

Structure and developmental expression of the mouse RGR opsin gene.

PURPOSE: The aim of this study is to isolate and characterize cDNA clones and the genes that encode mouse RPE retinal G protein-coupled receptor (RGR) and to analyze expression of the RGR gene in the developing mouse retina. The conserved amino acid sequences of RGR from various mammals can be compared to the amino acid sequence motif of G protein-coupled receptors. METHODS: Mouse RGR cDNA and gene clones were isolated from a retina cDNA library and 129SV genomic DNA library, respectively. The expression of RGR in the developing C57BL/6J mouse retina was analyzed by immunohistochemical staining with a polyclonal antipeptide antibody. RESULTS: The deduced amino acid sequence of mouse RGR is 78% and 81% identical to that of bovine and human RGR, respectively. The mouse RGR gene is split into seven exons and extends about 11 kb. Two predominant mRNA transcripts, 1.9 and 1.7 kb in length, and a third, relatively faint, 5.5-kb transcript were detected in mouse eye by hybridization to a RGR cDNA probe. Frozen sections of C57BL/6J mouse retina at various stages of development were incubated with a mouse RGR antipeptide antibody. RGR immunoreactivity was first seen at postnatal day 2 (P2) in centrally located RPE cells. From day P6 to P12, there was an increase in the number and intensity of immunoreactive RPE cells in the central and mid-peripheral regions of the retina, while the most peripheral RPE cells were still negative. By day P16, the length of the RPE monolayer was immunoreactive, and staining of the central RPE cells was markedly more intense than at younger ages. CONCLUSIONS: Mouse and human RGR are highly conserved. A gradient of RGR expression in RPE extends from the central to the peripheral retina during development. In reference to the appearance of melanin-positive differentiated RPE cells, the induction of RGR expression is a relatively late event in the maturation of the retina.

Localization of the human RGR opsin gene to chromosome 10q23.

The human RGR gene encodes an opsin protein (retinal G protein-coupled receptor), which is expressed in Müller cells and the retinal pigment epithelium and is thought to play a role in the visual process. To investigate a possible linkage of the RGR gene to retinal dystrophies, the locus of the gene was mapped on human metaphase chromosomes. Genomic and cDNA fragments of the human RGR gene were used as probes for fluorescence in situ hybridization. Analysis of the fluorescence signals on high-resolution banded chromosomes showed that the RGR gene is localized to human chromosome 10q23. This result now provides for the rapid analysis of this gene with respect to inherited diseases of the retina.

Blue and ultraviolet light-absorbing opsin from the retinal pigment epithelium.

The retinal pigment epithelium (RPE) contains an abundant opsin that is distinct from rhodopsin and cone visual pigments and is able to bind the retinaldehyde chromophore. The putative retinal G protein-coupled receptor (RGR) was isolated in digitonin solution from bovine RPE microsomes and copurified consistently with a minor 34-kDa protein. The absorption spectrum of RGR revealed endogenous pH-sensitive absorbance in the blue and near-ultraviolet regions of light. Membrane-bound RGR was incubated with exogenously added all-trans-retinal and formed two long-lived pH-dependent photopigments with absorption maxima of 469 +/- 2.4 and 370 +/- 7.3 nm. The effects of hydrogen ion concentration suggest that the blue and near-UV photopigments are tautomeric forms of RGR, in which an all-trans-retinal Schiff base is protonated or unprotonated, respectively. The RPE pigment was also demonstrable by its reactivity to hydroxylamine in the dark. The retinaldehyde-RGR conjugate at neutral pH favors the near-UV pigment and is a novel light-absorbing opsin in the vertebrate eye.

Molecular cloning and characterization of the G protein gamma subunit of cone photoreceptors.

The phototransduction process in cones has been proposed to involve a G protein that couples the signal from light-activated visual pigment to the effector cyclic GMP phosphodiesterase. Previously, we have identified and purified a G beta gamma complex composed of a G beta 3 isoform and an immunochemically distinct G gamma subunit (G gamma 8) from bovine retinal cones (Fung, B. K.-K., Lieberman, B. S., and Lee, R. H. (1992) J. Biol. Chem. 267, 24782-24788; Lee, R. H., Lieberman, B.S., Yamane, H. K., Bok, D., and Fung, B. K.-K. (1992a) J. Biol. Chem. 267, 24776-24781). Based on the partial amino acid sequence of this cone G gamma 8, we screened a bovine retinal cDNA library and isolated a cDNA clone encoding G gamma 8. The cDNA insert of this clone includes an open reading frame of 207 bases encoding a 69-amino acid protein. The predicted protein sequence of G gamma 8 shares a high degree of sequence identity (68%) with the G gamma (G gamma 1) subunit of rod transducin. Similar to rod G gamma 1, it terminates in a CIIS motif that is the site for post-translational modification by farnesylation. Messenger RNA for G gamma 8 is present at a high level in the retina and at a very low level in the lung, but is undetectable in other tissues. Immunostaining of bovine retinal sections with an antipeptide antibody against the N-terminal region of G gamma 8 further shows a differential localization of G gamma 8 to cones with a pattern indistinguishable from that of G beta 3. This finding suggests that G beta 3 gamma 8 is a component of cone transducin involved in cone phototransduction and color vision.

Alternative splicing in human retinal mRNA transcripts of an opsin-related protein.

An opsin-related gene encodes a putative RPE-retinal G-protein-coupled receptor (RGR) that is most homologous to the visual pigments and invertebrate retinochrome. A splice variant of human RGR mRNA can be demonstrated by the sequence of isolated cDNA clones and by the amplification and analysis of human retinal mRNA. The shortened transcript contains a deletion of 114 nucleotides that correspond exactly to the sequence of exon 6 in the human rgr gene. The predicted RGR variant lacks the putative sixth transmembrane domain and has a calculated molecular weight of 27,726. Variable amounts of a 28-kDa protein were found in the retinas of some individuals by immunoblot assay. Since a similar shortened RGR transcript was not detected in bovine retina or RPE, the RGR variant is not essential for vertebrate vision. Analysis of the structure of the rgr gene and of the sequences of cDNA clones indicates that the truncated mRNA may be produced through alternative splicing of pre-mRNA from which a cassette exon is removed and the predicted RGR variant is radically altered in primary structure.

A human opsin-related gene that encodes a retinaldehyde-binding protein.

The ligand-binding property of a cytoplasmic membrane-bound protein from bovine retinal pigment epithelium (RPE) has been demonstrated. The putative RPE-retinal G protein coupled receptor (RGR) covalently binds both all-trans- and 11-cis-retinal after reduction by sodium borohydride. The 32-kDa receptor binds all-trans-retinal preferentially, rather than the 11-cis isomer. The amino acid sequence of the opsin-related protein in humans is 86% identical to that of bovine RGR, and a lysine residue, analogous to the retinaldehyde attachment site of rhodopsin, is conserved in the seventh transmembrane domain of RGR in both species. The human gene that encodes the novel retinaldehyde receptor spans 14.8 kb and is split into seven exons. The structure of the gene is distinct from that of the visual pigment genes. These findings support the notion that the rgr gene represents the earliest independent branch of the vertebrate opsin gene family. A second form of human RGR in retina is predicted by alternative splicing of its precursor mRNA. This RGR variant results from the alternative use of an internal acceptor splice site in the second intron of the human gene, and it contains an insertion of four amino acids in the connecting loop between the second and thrid transmembrane domains. Since RGR binds all-trans-retinal preferentially, one of its functions may be to catalyze isomerization of the chromophore by a retinochrome-like mechanism.

Cytoplasmic retinal localization of an evolutionary homolog of the visual pigments.

A rhodopsin-related protein is preferentially expressed at high levels in retinal pigment epithelium (RPE) and in Müller cells. The putative RPE-retinal G protein-coupled receptor (RGR) was localized in light-adapted bovine retina by means of electron microscopic immunocytochemistry. In the RPE, the protein was localized to a widespread intracellular compartment. Except for the region adjacent to the basal surface, the RPE cytoplasm was labeled throughout the cell including the apical surface. In Müller cells also RGR was found in the intracellular compartment, especially in the cytoplasm in the region of the Müller cell endfeet and proximal cell processes. Subcellular fractionation studies of bovine RPE and neural retina indicated that RGR is a membrane-bound protein. The intracellular localization of RGR is a unique variation in the subcellular distribution of seven-transmembrane-domain receptors and suggests an unconventional role for RGR in the signal transduction process.

An opsin homologue in the retina and pigment epithelium.

PURPOSE: The aim of this project was to investigate the retinal pigment epithelium (RPE) at the molecular level by identification of novel RPE-specific cDNAs that may encode proteins of signal transduction pathways or other proteins that are expressed preferentially in the RPE. METHODS: A bovine RPE cDNA library was constructed in bacteriophage lambda g10 using RPE-enriched poly(A)+ RNA. The library was screened by differential hybridization to bovine RPE and kidney cDNA probes. RESULTS: A member of the hepatahelical receptor family was identified in bovine RPE by molecular cloning. Its deduced amino acid sequence predicts a protein that has 291 amino acid residues and resembles most closely the family of visual pigments. A lysine residue, analogous to the retinaldehyde attachment site in rhodopsin, is conserved in the seventh hydrophobic segment of the novel sequence. Messenger RNA encoding the putative G protein-coupled receptor was detected by in situ hybridization in the RPE, inner nuclear layer, and specific cells of the ganglion cell layer. Immunohistochemical staining of bovine retina showed that the receptor protein is localized in Müller cells, as well as in the RPE. CONCLUSIONS: A novel heptahelical receptor defines a distant evolutionary branch of the visual pigment tree. The selective localization of this putative receptor, its abundance in RPE and retina, and its homology to the visual pigments suggest that the function of this receptor is important in a visual process involving the RPE and Müller cells.

Structure of the bovine transducin gamma subunit gene and analysis of promoter function in transgenic mice.

Transducin, the major photoreceptor guanine nucleotide-binding protein (G protein), is composed of three polypeptides: alpha, beta and gamma subunits. The transducin gamma subunit (T gamma) is expressed preferentially in photoreceptors. To study the control mechanisms for photoreceptor-specific expression of the T gamma gene, clones of the bovine T gamma gene were isolated from a bacteriophage genomic library, and the structure of the gene, including a portion of its 5'-flanking region, was characterized. The gene consists of three exons and two introns. The first intron is 91 base pairs (bp) long and is located in the region corresponding to the 5'-untranslated sequence of the T gamma mRNA. The second intron is 5.3 kilobases (kb) long and splits the protein-coding region centrally. A bovine Alu-type repetitive sequence and putative Ret-1 and AP-1 binding site sequences are located in the 5'-flanking region. To investigate promoter function, 1.4 kb of DNA from the 5'-flanking region was joined to the prokaryotic chloramphenicol acetyltransferase (CAT) gene, and the chimeric bovine T gamma gene was used to generate a line of transgenic mice. CAT activity was readily detected in the retinas of the transgenic mice, but was absent in brain, heart, kidney, liver, lung, spleen and other tissues. These results suggest that the 1.4 kb 5'-flanking region of the bovine T gamma gene contains conserved sequence elements that direct tissue-specific expression. Human T gamma cDNA clones were characterized, and a short homologous region of the human T gamma gene promotor was obtained by polymerase chain reaction (PCR) amplification for comparison with the bovine promoter.

Guanine nucleotide-binding regulatory proteins in retinal pigment epithelial cells.

The expression of GTP-binding regulatory proteins (G proteins) in retinal pigment epithelial (RPE) cells was analyzed by RNA blot hybridization and cDNA amplification. Both adult and fetal human RPE cells contain mRNA for multiple G protein alpha subunits (G alpha) including Gs alpha, Gi-1 alpha, Gi-2 alpha, Gi-3 alpha, and Gz alpha (or Gx alpha), where Gs and Gi are proteins that stimulate or inhibit adenylyl cyclase, respectively, and Gz is a protein that may mediate pertussis toxin-insensitive events. Other G alpha-related mRNA transcripts were detected in fetal RPE cells by low-stringency hybridization to Gi-2 alpha and Gs alpha protein-coding cDNA probes. The diversity of G proteins in RPE cells was further studied by cDNA amplification with reverse transcriptase and the polymerase chain reaction. This approach revealed that, besides the above mentioned members of the G alpha gene family, at least two other G alpha subunits are expressed in RPE cells. Human retinal cDNA clones that encode one of the additional G alpha subunits were isolated and characterized. The results indicate that this G alpha subunit belongs to a separate subfamily of G proteins that may be insensitive to inhibition by pertussis toxin.

Novel localization of a G protein, Gz-alpha, in neurons of brain and retina.

Recently, a cDNA coding for a novel G protein alpha-subunit, Gz-alpha, was isolated from a human retinal cDNA library and shown by Northern blot analysis to be expressed at high levels in neural tissues. We have prepared affinity-purified antibodies specifically directed against synthetic Gz-alpha peptides and employed immunohistochemical methods to map the localization of Gz-alpha in human, bovine, and murine retina and brain. By light microscopy, Gz-alpha was localized to the cytoplasm of neurons, with predominant reactivity in ganglion cells of the retina, Purkinje cells of the cerebellum, and most neurons of the hippocampus and cerebral cortex. Reactivity was confined to perikaryon, dendrites, and a very short segment of proximal axons, except for the retinal ganglion cells, in which the axons in the nerve fiber layer showed intense Gz-alpha immunoreactivity proximal to the lamina cribrosa. Pre-embedding immunoelectron microscopy demonstrated the presence of focal Gz-alpha immunoreactivity on the nuclear membranes, endoplasmic reticulum, and plasma membranes of Purkinje cell perikarya and in association with microtubules in their proximal dendrites. Subcellular fractionation studies confirmed the association of Gz-alpha with plasma and intracellular membranes. The localization of Gz-alpha and its unique amino acid sequence suggest that it may have a specialized function in neural tissues.

Gz, a guanine nucleotide-binding protein with unique biochemical properties.

Cloning of a complementary DNA (cDNA) for Gz alpha, a newly appreciated member of the family of guanine nucleotide-binding regulatory proteins (G proteins), has allowed preparation of specific antisera to identify the protein in tissues and to assay it during purification from bovine brain. Additionally, expression of the cDNA in Escherichia coli has resulted in the production and purification of the recombinant protein. Purification of Gz from bovine brain is tedious, and only small quantities of protein have been obtained. The protein copurifies with the beta gamma subunit complex common to other G proteins; another 26-kDa GTP-binding protein is also present in these preparations. The purified protein could not serve as a substrate for NAD-dependent ADP-ribosylation catalyzed by either pertussis toxin or cholera toxin. Purification of recombinant Gz alpha (rGz alpha) from E. coli is simple, and quantities of homogeneous protein sufficient for biochemical analysis are obtained. Purified rGz alpha has several properties that distinguish it from other G protein alpha subunit polypeptides. These include a very slow rate of guanine nucleotide exchange (k = 0.02 min-1), which is reduced greater than 20-fold in the presence of mM concentrations of Mg2+. In addition, the rate of the intrinsic GTPase activity of Gz alpha is extremely slow. The hydrolysis rate (kcat) for rGz alpha at 30 degrees C is 0.05 min-1, or 200-fold slower than that determined for other G protein alpha subunits. rGz alpha can interact with bovine brain beta gamma but does not serve as a substrate for ADP-ribosylation catalyzed by either pertussis toxin or cholera toxin. These studies suggest that Gz may play a role in signal transduction pathways that are mechanistically distinct from those controlled by the other members of the G protein family.

Identification of a GTP-binding protein alpha subunit that lacks an apparent ADP-ribosylation site for pertussis toxin.

Recent molecular cloning of cDNA for the alpha subunit of bovine transducin (a guanine nucleotide-binding regulatory protein, or G protein) has revealed the presence of two retinal-specific transducins, called Tr and Tc, which are expressed in rod or cone photoreceptor cells. In a further study of G-protein diversity and signal transduction in the retina, we have identified a G-protein alpha subunit, which we refer to as Gz alpha, by isolating a human retinal cDNA clone that cross-hybridizes at reduced stringency with bovine Tr alpha-subunit cDNA. The deduced amino acid sequence of Gz alpha is 41-67% identical with those of other known G-protein alpha subunits. However, the 355-residue Gz alpha lacks a consensus site for ADP-ribosylation by pertussis toxin, and its amino acid sequence varies within a number of regions that are strongly conserved among all of the other G-protein alpha subunits. We suggest that Gz alpha, which appears to be highly expressed in neural tissues, represents a member of a subfamily of G proteins that mediate signal transduction in pertussis toxin-insensitive systems.