Distribution and role of heterotrimeric G proteins in the secretory pathway of polarized epithelial cells

The delivery of newly synthesized proteins to different membrane domains results in a geographic segregation of functions within polarized epithelial and neuronal cells. Similarly, the polarized secretion of newly synthesized proteins from distinct sides of epithelial cells enables differentiation and specialization of separate compartments of the extracellular milieu. Many membrane proteins and soluble proteins synthesized in the rough ER are trafficked via the constitutive secretory pathway, through the Golgi complex, to the cell surface or other destinations (Farquhar, 1985; Kelly, 1985; Palade, 1975; Rothman and Orci, 1992). The trafficking of these proteins is believed to occur as a continuous stream of small quantities of proteins that are shuttled between compartments in small vesicles or other membrane-bound carriers. Although the mechanisms governing the sorting and vesicular transport of proteins through this constitutive pathway have been studied extensively, there is still little understanding of how this dynamic, multi-stage pathway is regulated.

One example of a protein that is secreted in a constitutive and a polarized fashion from many epithelial cells is the heparan sulfate proteoglycan BMHSPG, which is a component of the basement membrane underlying the epithelia. The intracellular processing and secretion of the BMHSPG has been studied in kidney epithelial cell lines (Caplan et al., 1987; de Almeida and Stow, 1991; Stow and Farquhar, 1987; Stow et al., 1989). In polarized monolayers of cultured LLC-PK ₁ cells and MDCK cells, the BMHSPG is sorted and secreted in a polarized fashion from the basal surface of the cells (Caplan et al., 1987; de Almeida and Stow, 1991). The BMHSPG is secreted relatively abundantly and continuously from LLC-PK ₁ cells grown on filters and the sequential post-translational modifications of biosynthetically labelled BMHSPG can be used to follow its intracellular trafficking. Thus, assaying the processing and secretion of BMHSPG in LLC-PK ₁ cells has provided a useful model in which to study the secretory pathway. Importantly it has enabled us to follow an endogenous protein in intact cells for studies of secretion, particularly with respect to the regulation of secretion by G proteins, as described below.

The use of non-hydrolyzable nucleotide analogs such as GTPyS provided the first clue that GTP-binding proteins were involved in various steps of vesicle trafficking (Balch, 1990; Beckers and Balch, 1989; Melancon et al., 1987; Ruohola et al., 1988; Tooze et al., 1990). Addition of GTPγS to permeabilized cells and or cell-free assays was generally found to block constitutive protein transport between membrane-bound compartments, by inhibiting either fusion or formation of vesicle carriers. The effects of GTPγS were initially attributed to the many examples of monomeric G proteins found throughout the secretory pathway (Balch, 1990). However, it has since become known that members of the heterotrimeric G protein family are also present and functional in the secretory pathway.

Two large gene families of guanine nucleotide-binding (G) proteins have been identified. The ras supergene family consists of several families of small molecular mass, monomeric G proteins, which are found in soluble or membrane-attached forms in yeast and higher cells and are involved in a diverse range of cellular events, including protein trafficking and secretion (Downward, 1990). The other main family of G proteins has long been recognized for its involvement in signal transduction pathways and the regulation of receptors at the cell surface (Gilman, 1987; Stryer and Bourne, 1986). Heterotrimeric G proteins exist as αβγ subunit complexes in which the a subunit is responsible for binding and hydrolysis of GTP. The βγ subunit is complexed to α during the inactive portion of the hydrolysis cycle and more recent studies have shown that βγ may have its own independent signalling functions following dissociation from a (Federman et al., 1992; Tang and Gilman, 1991). The Gα subunits have traditionally been classified by their functional sensitivity to bacterial toxins; Gα; and α₀ are ADP-ribosylated by pertussis toxin, whereas Gα_s is specifically ADP-ribosylated by cholera toxin. The heterotrimeric G proteins exist largely as membrane-associated proteins and the individual subunits have myristoylation or other fatty acyl modifications for membrane attachment (Fukada et al., 1990; Jones et al., 1990; Mumby et al., 1990).

Multiple gene products of each of the α, β and γ subunits associate in distinct combinations generating heterogeneity amongst these G proteins (Simon et al., 1991). Discrete deployment of these heterogeneous G proteins to different sites on membranes throughout the cell then may allow for the simultaneous regulation of many different signal transduction pathways. The cellular distribution and intracellular localization of heterotrimeric G protein subunits is thus important to elucidate. The localization of different G proteins on membranes in cells and tissues has recently been studied in our laboratory and by others. Detection and localization of different G proteins have been made possible by the availability of specific peptide antibodies that can distinguish between the different G protein subunits (Spiegel et al., 1990).

The polarized epithelial cell line LLC-PK₁ expresses only two of the known pertussis toxin-sensitive Gα subunits, Gα_i-2 and Gα_i-3. Our studies showed that Gai-2 and Gai-3 are localized on distinct membrane domains and have quite separate functions in these polarized cells (Ercolani et al., 1990; Holtzman et al., 1991; Stow et al., 1991a). Gα_i-2was found to be associated with the plasma membrane, while Gα_i-3was found in a novel position on intracellular Golgi membranes. The immunofluorescence staining of Gai-2 resembles a cobblestone pattern, typical of basolateral plasma membrane staining, in confluent monolayers of LLC-PK_i cells (Ercolani et al., 1990). The Gα_i-2subunit is targeted in a polarized fashion to the basolateral membrane, particularly the lateral membrane, and is not found on the apical membrane of these cells. The localization of Gai-2 on the cell membrane is consistent with its proposed role in regulating adenylyl cyclase in response to hormone receptors (Ercolani et al., 1990). Gα_i-2 is developmentally expressed in LLC-PK _i cells. The expression of Gα_i-2 is under transcriptional control and its appearance on the plasma membrane parallels the gradual development of polarity as the cultures progress from a sparse, non-polar morphology to a confluent, polarized morphology (Holtzman et al., 1991). Due to the lateral membrane localization of Gα_i-2 and its pattern of expression, it is tempting to speculate that this Gα subunit may actually be involved in the acquisition of functional or morphological polarity in these cells.

In contrast to Gα_i-2, its highly homologous relative, Gα_i-3, was found to be localized on the perinuclear Golgi complex in LLC-PK_i cells (Ercolani et al., 1990; Stow et al., 1991a). It is not currently known how different Gα subunits are targeted to distinct membrane domains. Our recent studies, using expression of chimeric proteins, have identified regions of the Gα proteins that appear to be responsible for their targeting (J. Stow, L. Ercolani, D. Ausiello et al., unpublished data). The targeting or membrane association of Gα subunits in LLC-PK_i cells appears to be a saturable process. This was also elegantly demonstrated in fibroblasts where the localization of Gai-3 was dependent on its level of overexpression (Hermouet et al., 1992). While LLC-PK_i cells express Gai-3 only on their Golgi membranes, other cells have Gα_i-3 associated with additional membranes. In the A6 toad kidney cells, Gα_i-3 is found on the perinuclear Golgi complex and is also present near the apical cell surface where it is intimately associated with, and regulates, Na⁺ channel activity (Ausiello et al., 1992; Cantiello et al., 1989). In fibroblasts, Gα_i-3 is present on Golgi membranes and also in some cases at the plasma membrane where it can regulate adenylyl cyclase (Hermouet et al., 1992). The multiple locations and functions for Gα_i-3 in such cells reinforces the concept that a given G protein subunit may be used promiscuously, for different functions in different cells, and for different functions within the same cell.

At a time when heterotrimeric G proteins were thought to function only at the level of the plasma membrane, several studies provided evidence for the association of pertussis toxin-sensitive G proteins on intracellular membranes such as the rough ER (Audigier et al., 1988), the Golgi complex (Ali et al., 1989; Toki et al., 1989) and endosomes (Ali et al., 1989). However, the functions of these intracellular G proteins and their potential involvement in trafficking pathways, such as the secretory pathway, were not known at that time. It is now clear that Gα_i-3 and other G protein subunits are fairly universal components of Golgi membranes. Gα_i-3 has now been specifically detected on the Golgi membranes in different tissues and cells, including different kidney cell lines in our own laboratory (Fig. 1) and in a number of other cell types by other groups (Barr et al., 1991; Hermouet et al., 1992; Ktistakis et al., 1992; Stow et al., 1991a). Gα_i-3 on Golgi membranes is localized on the cytoplasmic face of Golgi cistemae where it is peripherally, but stably, attached to the membrane (Stow et al., 1991a). While weak cytoplasmic staining of Gα_i-3 disappears after cycloheximide treatment of cells, the more intense Golgi staining remains, suggesting that Gα_i-3 is a stable resident of the Golgi membranes and that the cytoplasmic staining may be attributed to a newly synthesized pool of the subunit (Stow et al., 1991a). Gα_i-3 on the Golgi membranes associates with a Py complex, as was shown by its ability to be efficiently ADP-ribosylated by pertussis toxin, which requires heterotrimer formation (Sunyer et al., 1989). In isolated rat liver Golgi membranes, Gα_i-3 is the only pertussis toxin-sensitive protein that can be detected by immunoblotting (Fig. 2) or by ³²P-ADP-ribosylation with pertussis toxin (Stow et al., 1991a). Other G protein subunits are however present on these membranes. Immunoblotting confirmed the presence of the β subunit and of the cholera toxin-sensitive Gα_s subunit on rat liver Golgi membranes (Fig. 2). Golgi membranes from other cells may contain different combinations of Gα subunits. Trani-Golgi membranes from PC 12 cells contain Gα_i-3 and other pertussis toxin-sensitive subunits; Gα₀ and Gα_i-1,2, and the Ga_s subunit have also been found on these membranes (Leyte et al., 1992).

The localization of Gα _i subunits on various intracellular membranes, together with functional studies such as those described below, have now established a role for these G proteins in the secretory pathway.

The distinct intracellular and intercellular distributions of G proteins found in cultured cells are also found in epithelial cells in tissues and this is graphically demonstrated by the distribution of Gα subunits in the kidney cortex. The renal nephron of the kidney is composed of a large number of morphologically and functionally distinct polarized epithelial cells. The apical or lumenal membranes and the basolateral membranes of these cells host a variety of ion channels and hormone receptors that are known to be coupled to G protein-mediated signal transduction pathways. The distribution of different Gα subunits on membrane domains throughout kidney epithelial cells is therefore potentially important for the segregation of functions associated with these membranes.

Using specific peptide antibodies, the distribution of different Gα subunits in cells along the nephron has been studied by several groups (Brunskill et al., 1991; Gupta et al., 1991; Stow et al., 1991b). There is a heterogeneous distribution of Gα; and Gα_s subunits in cells from functionally distinct areas of the tubules (Brunskill et al., 1991; Stow et al., 1991b) and examples of Gα _i, Gα_s and Gα₀ have been found in the renal glomerulus (Gupta et al., 1991). Proximal tubule cells express several different Gα subunits, the largest number appearing to be concentrated in this region of the nephron. Gα_i-2, Gα_i-3 and Ga_s are all heavily concentrated in the sub-apical vesicles of proximal tubule cells (Fig. 3). Gai-3 was also found associated with the basolateral membrane of cells in some segments of the proximal tubules (Stow et al., 1991b). Cells of the cortical collecting duct show a highly polarized distribution of Gα_s, which is on the basolateral membranes of the principal and intercalated cells (Fig. 3). The principal cells also contain Gα_i-2, which is restricted to lateral membranes, in a similar localization to that seen in cultured renal LLC-PK, cells (Ercolani et al.. 1990; Stow et al., 1991a). The distribution of the Gα_i-1 subunit provides the most dramatic example of a highly selective and polarized localization of G proteins. Gα_i-1 is found on the apical membrane of cells in the thick ascending limb of the distal nephron; however in a highly specialized group of cells within this tubule, the macula densa, Gα_i-1 was found at the basal aspect of the cells (Stow et al., 1991b). The inverted polarity of the Gα_i-1 protein is perhaps related to the altered morphological polarity of the macula densa cells (Rouiller and Orci, 1969), which are uniquely positioned for their highly specialized roles in the control of glomerular filtration rate and renin secretion. The functional relevance of the polarized localization of Gα_i-1 in thick ascending limb cells or in macula densa cells is not known.

In general, the functions of these G protein subunits on distinct membrane domains in kidney epithelial cells are not known. However their distributions on specific membrane domains, once again, suggests that geographic segregation of G protein subunits may be important for the control of both membrane protein signalling events and membrane trafficking.

The localization of Gα_i-3 on the Golgi membranes provided us with the first suggestion that this G protein might be involved in Golgi trafficking, and subsequent studies in polarized LLC-PK _i cells produced evidence for a functional role of Gai-3 in secretory trafficking (Stow et al., 1991a). The constitutive polarized secretion of BMHSPG in LLC-PK _i cells was used as a model to determine the effects of overexpression of Gα_i-3. The overexpression of close-to-physiological amounts of Gα_i-3 resulted in correct targeting of this protein to Golgi membranes and functionally it produced a significant retardation of the secretion of BMHSPG. The effect of the overexpressed, and presumably activated, Gα_i-3 was at the level of trafficking through the Golgi complex where precursors of the BMHSPG accumulated. These studies suggested that activated Gai-3 might be imposing a rate-limiting or regulatory step in the transport of proteins, exemplified by BMHSPG, through the Golgi complex. This retardation or block in trafficking could be relieved by pretreating cells with pertussis toxin which increased the rate of intracellular trafficking and secretion of BMHSPG. Pertussis toxin ADP-ribosylates Gα_i-3, functionally uncoupling it from its receptor or activating protein. Since Gα_i-3 is the only known pertussis toxin substrate on Golgi membranes in LLC-PK _i cells, the effects of the toxin can be attributed directly to Gai-3. Furthermore, since pertussis toxin also enhanced secretion in cells expressing only endogenous Gα_i-3, this G protein can be regarded as a normal, intrinsic regulator in the secretory pathway. Interestingly, either blocking or enhancing the rate of intracellular transport disrupted neither the sequence of post-translational processing of the BMHSPG nor the polarity of its secretion, also supporting the concept that the acti-vation/deactivation of Gα_i-3 is perhaps a normal regulatory step in this pathway.

The effects of Gα_i-3 overexpression on constitutive secretion are consistent with the effects of agents that cause a more widespread activation of heterotrimeric G proteins. It has been known for some time that exogenous addition of aluminium fluoride can be used to activate G proteins; in the context of the GDP-binding pocket of Gα, the fluoride ions of this complex mimic the γ phosphate of GTP, prompting activation of the G protein (Higashijima et al., 1991). Recently, aluminium fluoride has been shown to be a more specific activator for heterotrimeric G proteins than for several families of monomeric G proteins (Kahn, 1991) and it has been used specifically to implicate heterotrimeric G proteins in a number of membrane trafficking pathways. For instance, functional assays have shown that heterotrimeric G proteins are involved in regulating several steps of the endocytic pathway (Carter et al., 1993; Columbo et al., 1992). In an earlier study, Melancon and coworkers provided some of the first evidence that heterotrimeric G proteins were required for Golgi trafficking when they showed that both GTPγS and aluminium fluoride blocked in vitro vesicle transport between Golgi compartments (Melancon et al., 1987). Using another cell-free assay, which measures the budding of secretory vesicles and secretory granules off trans-Golgi network (TGN) membranes, Barr and coworkers also provided important evidence for the involvement of heterotrimeric G proteins in secretory trafficking (Barr et al., 1991, 1992). Data showing that vesicle budding was blocked by both GTPγS and aluminium fluoride but enhanced by addition of free βγ subunits or by pertussis toxin (Barr et al., 1991), are consistent with similar effects on secretion produced by overexpression of Gα_i-3 and pertussis toxin in intact cells (Stow et al., 1991). Further studies on vesicle budding from TGN membranes have also shown that, in addition to Gα_i-3, there are other pertussis toxin- and cholera toxin-sensitive G proteins on these membranes (Leyte et al., 1992). Cholera toxin, which ADP-ribosylates and activates Ga_s, also stimulated TGN vesicle budding, suggesting that the Gα_s may act in concert with Ga_s as an opposing and therefore regulatory partner. Taken together, these studies confirm the requirement for a pertussis toxin-sensitive Gα protein in secretory trafficking, although the molecular mechanisms by which it, together with other G proteins, perform this regulation are still unknown.

Due to the presence of multiple G protein subunits on Golgi membranes, it is easy to speculate that there is more than one mode of regulation based on GTP-hydrolysis and perhaps more than a single signal transduction pathway involved. The key proteins to which the heterotrimeric G proteins are functionally connected should be, in traditional terms, a receptor and an effector or their equivalents. Neither of these proteins has yet been identified on Golgi membranes. The receptor on the Golgi could fit into the traditional family of G-protein-coupled, seven membranespanning domain receptors, or possibly the family of single membrane-spanning domain receptors involved in sorting, or it could defy presumption and be a completely different class of membrane protein. Studies to date have also not identified a likely effector on the Golgi. We were not able to find effects on vesicle trafficking with intermediates in pathways of common effectors such as phospholipase or adenylyl cylcase (J. Stow, unpublished data).

There are several lines of evidence to connect heterotrimeric G proteins to other Golgi-associated proteins involved in vesicle trafficking. One set of proteins that appears to be intimately involved in trafficking is the coatomer, including the β-COP protein (Duden et al., 1991; Serafini et al., 1991). β-COP is a component of a cytosolic complex, which binds to Golgi membranes where it forms a specialized coat that may be responsible for structuring Golgi transport vesicles or other vehicles. β-COP and other cytosolic proteins such as γ-adaptin and the monomeric G protein ARF (ADP-ribosylating factor) have two features in common: their binding to Golgi membranes is prevented by the fungal metabolite brefeldin A (BFA) and enhanced by activation of G proteins (Donaldson et al., 1991b; Ktistakis et al., 1992; Robinson and Kries, 1992). The BFA-induced dissociation of β-COP is prevented by aluminium fluoride or mastoparan, which in turn is also altered by pertussis toxin, implying that a heterotrimeric G protein is responsible for stabilizing its membrane association (Donaldson et al., 1991b; Ktistakis et al., 1992). The binding of β-COP and ARF to Golgi membranes was measured directly, in vitro, to show that a heterotrimeric G protein was involved in this binding (Donaldson et al., 1991a). We have identified a cytosolic protein, p200, which binds specifically to Golgi membranes (Narula et al., 1992). The binding of p200 to Golgi membranes is regulated in a similar fashion to the coatomer proteins, and its association with the Golgi is influenced by BFA and by activation of G proteins. The dissociation of p200 in the presence of BFA in NRK cells is prevented by activation of heterotrimeric G proteins by aluminium fluoride (Fig. 4). Our recent studies have shown that p200 binding to Golgi membranes is also affected by pertussis toxin (J. Stow, B. de Almeida, unpublished data). These studies have provided evidence that heterotrimeric G proteins, including specifically pertussis toxin-sensitive G proteins, are involved in regulating the dynamic interaction of soluble vesicle coating proteins with Golgi membranes. The effects of G protein activation/ inhibition on coat protein binding are not yet reconciled into a ready explanation of how G proteins ultimately regulate secretion. Obviously, other proteins in this molecular network need to be identified in order for us to understand how G proteins are able to provide regulation in constitutive pathways of protein secretion and membrane trafficking.

This work was supported by NIH grants DK 42881 and PODK 38452.