Regulation and Function of the RasGRP Family of Ras Activators in Blood Cells

Abstract

Ras guanyl nucleotide releasing proteins (RasGRPs) are guanyl nucleotide exchange factors that activate Ras and related GTPases such as Rap. Like Sos proteins, RasGRPs have a catalytic region composed of a Ras exchange motif (REM) and a CDC25 domain. RasGRPs also possess a pair of atypical EF hands that may bind calcium in vivo and a C1 domain resembling the diacylglycerol (DAG)–binding domain of protein kinase C. DAG directly activates RasGRPs by a membrane recruitment mechanism as well as indirectly by PKC-mediated phosphorylation. RasGRPs are prominently expressed in blood cells. RasGRP1 acts downstream of TCR, while RasGRP1 and RasGRP3 both act downstream of BCR. Together, they regulate Ras in adaptive immune cells. RasGRP2, through Rap, plays a role in controlling platelet adhesion, while RasGRP4 controls Ras activation in mast cells. RasGRP malfunction likely contributes to autoimmunity and may contribute to blood malignancies. RasGRPs might prove to be viable drug targets. The intracellular site of RasGRP action and the relationship between RasGRPs and other Ras regulatory mechanisms are subjects of lively debate.

Early Models of Ras Regulation in Lymphocytes and the Discovery of RasGRP1

The small GTPase Ras is converted from its inactive GDP-bound form to its active GTP-bound form immediately after T cells are activated.¹ In the first Ras activation studies, T cells were labeled in vitro by incubating with ³²Pi to tag intracellular pools of guanyl nucleotides and then stimulated with antibodies directed against components of the T cell receptor (TCR) to simulate the interaction between a T cell and an antigen-presenting cell (APC). Following immunoprecipitation of Ras from cell lysates with an anti-Ras antibody, the status of Ras at the time of lysis could be determined by separation and quantification of GDP and GTP recovered from the immune precipitates. In an alternative approach, cells were rendered permeable, and Ras was loaded in situ with [α-³²P]-GTP and then analyzed as above. These studies showed that the ratio of Ras-GTP/Ras-GDP increased very rapidly after T cell stimulation. Furthermore, treatment with PMA, which mimics the diacylglycerol (DAG) produced by phospholipase C downstream of TCR, also potently activated Ras.¹ When these experiments were published in 1990, the classic and novel forms of protein kinase C (PKC) were the sole known targets of DAG and PMA.

The Ras guanyl nucleotide equilibrium can be regulated by GTP hydrolysis and by guanyl nucleotide exchange. Because Ras in permeable T cells rapidly associated with exogenous nucleotide regardless of TCR stimulation, decreased Ras GTPase activity was implicated in Ras activation following TCR stimulation.¹ Based on the observation that total cell lysate activity of Ras GTPase activating protein (Ras GAP) was decreased, PKC was fingered in transducing a positive signal to Ras by negatively regulating Ras’ negative regulator. This hypothesis was strengthened by a subsequent study, which showed that peptide inhibitors of PKC introduced into permeable cells also dampened Ras activation.² The precise mechanism whereby PKC might inhibit Ras GAPs was not apparent, and indeed, the relative roles of different forms of PKC and Ras GAPs in T cells are still uncertain.

Subsequent experiments with growth factor responses in nonlymphocytes provided an alternate view that stimulation would induce the recruitment of the Ras guanyl nucleotide exchange factor (Ras GEF) Son of Sevenless (Sos) to a signaling complex at the plasma membrane and thereby facilitate a productive interaction between Sos and Ras.³ Adaptor molecules such as Grb-2 and SHC, for example, would bridge Sos proteins to membrane protein phosphotyrosine docking sites on the inner surface of the plasma membrane. Close proximity of Sos to plasma membrane–bound Ras would encourage Sos to evict the guanyl nucleotide from Ras-GDP, allowing the relatively prevalent free cellular GTP to bind Ras.³

Research with T cells sought confirmation of this general mechanism, and attention soon focused on the transmembrane adaptor protein LAT.⁴ LAT is a substrate for upstream protein tyrosine kinases and serves as a nucleus for the assembly of a signaling complex including adaptors Grb-2, SLP76, GADS, as well as the enzyme phospholipase C–gamma1 (PLC-γ1). According to current theory, TCR engagement with peptide antigen bound to major histocompatibility molecules on APCs triggers the activation of the Src family protein tyrosine kinase Lck in the T cell.⁵ Lck phosphorylates immunoreceptor tyrosine–based activation motifs (ITAMs) on the intracellular signaling domains of TCR-associated proteins such as CD3. Subsequently, the Syk family kinase ZAP70 is recruited to the TCR complex through its SH2 domain and activated by Src family kinases. In turn, this leads to ZAP70-mediated LAT phosphorylation on 4 tyrosine residues: Y136, Y175, Y195, and Y235 (numbered according to the mouse polypeptide). Mutation studies have shown that Y136 phosphorylation controls PLC-γ1 recruitment. Phosphorylation of the distal sites is required for Grb-2 binding, which recruits Sos to the plasma membrane.⁴ These studies indicated how Sos proteins could participate in Ras activation in lymphocytes. Complicating this genetic dissection, however, mutation analysis indicated that distal tyrosine phosphorylation sites also contribute to PLC-γ1 binding, possibly indicating cooperative binding of multiple proteins to LAT. Furthermore, Ras activation models based on Sos could not explain the observation that DAG analogs could activate Ras in T cells.

LAT recruitment of PLC-γ1 and subsequent activation by Tec family protein kinases results in cleavage of phosphatidylinositol 4,5-bis phosphate (PIP2). PIP2 cleavage leads to accumulation of inositol-3-phosphate (IP3), which mediates a rise in cytosolic calcium. Calcium activates calcineurin, leading to mobilization of NFAT transcription factors to the nucleus. PIP2 cleavage also leads to accumulation of DAG in the plasma membrane, which has traditionally been associated with activation of DAG-responsive forms of PKC. The discovery of a DAG-responsive Ras activator, RasGRP1, provided a straightforward explanation for how DAG could be linked to RAS signaling in T cells and challenged the view that Sos is responsible for Ras activation downstream of TCR.^6,7 For simplicity, and at the expense of chronological accuracy, current knowledge about the regulation and function of each of the 4 RasGRP family members is summarized separately below.

Properties of RasGRP1

Like Sos proteins, RasGRP1 possesses a REM (Ras exchange motif) and CDC25-related domain that constitute the catalytic region (Fig. 1). Additionally, RasGRP1 possesses a pair of EF hands. EF hands are calcium-binding elements with a helix-loop-helix structure. RasGRP1 has a single C1 domain similar to the DAG-binding region of PKC. RasGRP1 has a carboxyl-terminal tail of unknown significance. The structure of RasGRP1 suggested the hypothesis that the protein links PLC activity to Ras.⁶ Cleavage of PIP2 by PLC would lead to an increase in the concentration of DAG in the plasma membrane. Binding of RasGRP1 to the membrane DAG would concentrate RasGRP1 near Ras, encouraging conversion of Ras-GDP to Ras-GTP. The other consequence of PIP2 hydrolysis, rising cytosolic calcium concentration, could hypothetically change the occupancy level of the EF hands, perhaps favoring activation of RasGRP1 by an allosteric mechanism or by encouraging membrane association.

Figure 1.

Domain structure of RasGRP1. Positions of the conserved domains including REM (Ras exchange motif), CDC25-related domain, potential calcium-binding EF hands, and diacylglycerol-binding C1 domain are indicated. RasGRP2 to RasGRP4 have similar domain structures.

Several types of experiments provided support for this general hypothesis. RasGRP1-encoding genes were first identified as transforming cDNA sequences selected from brain and T cell cDNA libraries.^6,7 The transformed phenotype of cells expressing a RasGRP1 cDNA sequence resembled that seen with weakly activated Ras transforming alleles, and the degree of transformation was enhanced in the presence of overexpressed wild-type H-Ras. The RasGRP1 catalytic domain was shown to promote H-Ras guanyl nucleotide exchange in vitro.⁶ At least one EF hand bound calcium in vitro, and the C1 domain robustly bound [³H]-phorbol ester dibutyrate, an analog of DAG.⁶ Fibroblasts expressing RasGRP1 exhibited elevated basal levels of Ras-GTP. Upon stimulation with PMA, these cells rapidly increased their levels of Ras-GTP. They showed activation of the downstream Raf-Mek-Erk kinase cascade, and they assumed a highly transformed morphology. PMA induced translocation of RasGRP1 to the membrane fraction in subcellular fractionation experiments, and a GFP-tagged RasGRP1 demonstrated PMA-induced translocation to cellular membranes.^6,7 Under these conditions, RasGRP1 signaling was markedly sustained with no signs of attenuation.⁶ Subsequent studies showed that the C1 domain binds DAG and DAG analogs with low nanomolar affinity and shows a preference for DAG complexed with acidic phospholipids.⁸

Using a polyclonal antibody directed against the N-terminus, expression of the predicted 90-kDa RasGRP1 could be detected in various T cell lines and mouse thymocytes.⁹ The Jurkat T cell line was used to explore the possible involvement of RasGRP1 downstream of TCR. Inhibitors of PLC-γ1 decreased the response to TCR stimulation at the level of Ras activation and Erk phosphorylation. For these studies, an anti–K-Ras antibody was used to monitor Ras activation. Overexpression of RasGRP1 in Jurkat T cells resulted in enhanced sensitivity to TCR stimulation of Ras-Erk signaling over a wide range of stimulatory antibody concentrations. Overexpression of RasGRP1 also increased the production of IL-2 in response to treatment with a combination of a calcium ionophore and DAG analogs such as PMA or bryostatin-1. As with PMA stimulation in fibroblasts, TCR stimulation increased the ratio of RasGRP1 associated with the particulate versus the soluble fraction in subcellular fractionation experiments.⁹

The above studies provided suggestive evidence that RasGRP1 might transduce signals from TCR and PLC-γ1 to Ras. Further support for this idea was obtained from the analysis of mice lacking RasGRP1.¹⁰ Ras signaling through the canonical Raf-Mek-Erk pathway had been previously implicated in T cell development.¹¹ Young Rasgrp1–/– mice on the mixed B6/129J genetic background demonstrated a substantial deficiency of mature thymocytes. Splenic T cells were also reduced in number in young mice.

Thymocytes develop from very immature double-negative (CD4⁻, CD8⁻) to intermediate-maturity double-positive (CD4⁺, CD8⁺) cells through a process called β-selection. In most thymocytes, signals emanating from the pre-TCR, which include a TCRβ chain encoded by a somatically rearranged TCRβ gene and an invariant pre-TCRα chain, drive this transition. Pre-TCR signaling likely requires Ras activity, and artificial Ras signaling provided by an activated mutant Ras can also promote this early transition in Rag1–/– mice that fail to express a pre-TCR.¹² Immature thymocytes express RasGRP1, and the expression level increases as thymocytes mature.^10,13 However, Rasgrp1–/– mice display roughly normal numbers of double-negative and double-positive thymocytes, showing that RasGRP1 has no obligatory role in β-selection.¹⁰

Double-positive thymocytes that experience productive TCRα gene rearrangement express a mature TCR that is potentially capable of interacting with self-antigens presented by various thymic APCs. Cells that express TCR-recognizing self-antigens with intermediate affinity receive survival signals and differentiate in a process called “positive selection”. Following lineage commitment, they down-regulate either CD8 or CD4 to become mature single-positive thymocytes and exit the thymus to become either CD4⁺ helper T cells or CD8⁺ cytotoxic T cells, respectively. Thymocytes that fail to express a functional TCR die in a process called “death by neglect”. Those that express strongly self-reactive TCRs undergo apoptosis in a process called “negative selection”. The substantial deficiency of single-positive thymocytes in RasGRP1 mice is consistent with the idea that RasGRP1 plays an important role downstream of newly minted TCR to promote positive selection. Previous work implicates Ras, Raf, Mek, and Erk in this process.¹¹

Further work with the Rasgrp1–/– mice combined with TCR transgenic strains confirmed this proposal.¹⁴ The production of single-positive thymocytes in female mice expressing the weakly selecting class I–restricted HY TCR transgene depended strongly on the presence of RasGRP1, implying that low-grade signals critically depend on RasGRP1 for transmission. In contrast, the more strongly selecting class I–restricted 2C TCR transgene produced single-positive cells in the absence of RasGRP1. Interestingly, although decreased positive selection was paralleled by decreased phospho-Erk in double-positive thymocytes, the single-positive thymocytes that did develop in the absence of RasGRP1 expressed abundant phospho-Erk. This result implies the existence of a RasGRP1-independent signaling pathway that is more efficiently utilized by the more strongly selecting 2C TCR. In male mice, the HY TCR transgenic cells undergo negative selection in response to self-antigen. This process proceeds efficiently in the absence of RasGRP1.¹⁴ Similar studies showed that development of CD4⁺ single-positive thymocytes expressing the class II–restricted AND TCR depends almost entirely on RasGRP1, which was interpreted to reflect the effect of sustained Erk activation in favoring a commitment to the CD4⁺ lineage.¹⁵

Peripheral RasGRP1-deficient T cells exhibit many abnormalities. They exhibit inefficient homeostatic expansion.¹⁴ By surface marker analysis, they show signs of activation and exhaustion.¹⁵ Rasgrp1–/– mice fail to mount a robust T cell immune response to either viral or bacterial pathogens.¹⁵ Rasgrp1–/– mice develop fewer CD4⁺ CD25⁺ Treg cells, but these mutant cells appear to be more actively suppressive than their wild-type counterparts.¹⁶ Naïve peripheral CD8⁺ T cells expressing the 2C TCR have a much higher threshold for responding to antigen when they lack RasGRP1. They also produce less IL-2 and proliferate less than their wild-type counterparts.¹⁷ Whether these phenotypes are direct effects of RasGRP1 loss in mature T cells or the indirect consequences of a RasGRP1-deprived upbringing cannot be easily discerned.

The properties of transgenic mice that overexpress RasGRP1 in thymocytes support the idea that the protein influences thymocyte development.¹³ Thymocytes expressing RasGRP1 from the Lck proximal promoter developed double-positive thymocyte cells even in a Rag1–/– background that precludes TCRβ gene rearrangement and expression of the pre-TCR. RasGRP1 overexpressing thymocytes developed preferentially into the CD8⁺ lineage, and they were hypersensitive to TCR-induced proliferation in vivo.

A curious property of older Rasgrp1–/– knockout mice is that they develop an autoimmune lymphoproliferative disorder associated with pronounced splenomegaly, although this phenotype has not been observed in all colonies.¹⁵ The enlarged spleens contain abnormally high numbers of CD4⁺ cells. Sera from such mice had antinuclear antibodies and increased levels of some IgG isotypes. The lymphoproliferative/autoimmune-like phenotype of Rasgrp1–/– mice is completely suppressed by combining the mutation with the nude mutation, which ablates the thymus and thereby prevents development of mature T cells.¹⁸ Thus, loss of RasGRP1 likely causes homeostatic expansion of aberrant T cells, which have an indirect effect on the B cell compartment. A RasGRP1-deficient strain, Rasgrp1^lag/lag, which arose as an incidental mutation in an unrelated gene targeting study, has similar phenotypes.¹⁹ Although there are some controversial aspects of this latter study, similar conclusions have been drawn by different groups using different methods and different mouse strains. For example, transfer of the Rasgrp1^lag T cells into recipients was sufficient to evoke the autoimmune lymphoproliferative disease,¹⁹ confirming the pathogenic role of RasGRP1-deficient T cells.

Despite these complexities at the whole animal level, the data summarized so far are consistent with the original hypothesis of RasGRP1 regulation and function. The hypothesis that RasGRP1 represents a key target of DAG downstream of TCR signaling is now widely accepted and has achieved the status of curriculum fodder for university undergraduates. However, the roles of the EF hands and the C-terminal region of RasGRP1 have not been deciphered, and inevitably, further studies have necessitated a much more complex view of RasGRP regulation and function.

There have been attempts to implicate the EF hands as calcium sensors using calcium ionophores and chelators, but the effects have been negative or subtle. Such crude probes cannot distinguish direct from indirect effects of calcium on RasGRP function. Analysis of truncated RasGRP1 proteins expressed in chicken DT40 B cells has provided evidence that the EF hands can regulate plasma membrane association.²⁰ The EF hands may function analogously to the C2 domains of PKC, which are structurally distinct calcium- and phospholipid-binding domains thought to work in conjunction with the C1 domain to further membrane attachment. However, genetic analysis has provided conflicting data about whether the EF hands are positive or negative regulators of RasGRP1 function. Without knowing the affinity of the EF hands for calcium, it is not possible to know whether the calcium occupancy changes as a function of cellular stimulation. EF hands in vivo may be constitutively bound to calcium, and some prefer magnesium.²¹ The atypically short spacer between the 2 EF hands of RasGRP1 suggests that it may not function as a typical calcium sensor.

Besides direct recruitment by DAG through its C1 domain, DAG can indirectly influence RasGRP1 by means of PKC-mediated phosphorylation.²² RasGRP1 is phosphorylated on T184 in TCR-stimulated T cells, as demonstrated with an anti–phospho-T184 antibody, and PKC inhibitors prevent Ras activation in T cells.^23,24 A substitution mutation in a RasGRP1 cDNA that encodes a nonphosphorylated protein, T184A, has impaired activation even when the cells are stimulated with PMA to induce direct activation through the C1 domain.²³ The molecular mechanism whereby T184 phosphorylation contributes to RasGRP1 activity is unknown. The phosphorylation site is located near the amino-terminal limit of the CDC25 domain and might serve to enhance catalysis perhaps by altering the tertiary association of CDC25 with REM. Cytotoxic T cell clones that are stimulated with immobilized anti-TCR antibodies show prolonged Erk activation, and they release cytotoxic granules. The late phase of Ras-Erk activation appears to depend more on PKC than does the initial phase.²⁵

The biological significance of both direct and indirect DAG regulation of RasGRP1 is mysterious. Dual regulation of RasGRP1 by DAG might endow RasGRP1 signaling with nonlinear properties, with low-level DAG stimulation leading to roughly proportional membrane recruitment and GEF activity and with higher levels of DAG boosting RasGRP1 output disproportionately as a reflection of co-concentrating PKC and RasGRP1 on the same membrane surfaces.²⁶ However, an alternative mechanism for possible RasGRP1 signaling nonlinearity has been proposed (see below).

The identity of the responsible PKC form(s) that positively regulates RasGRP1 is not known with certainty, and it is plausible that PKC activates some minion kinase to do the job. In Jurkat T cells, PKCθ seems able to modulate RasGRP1 activation,²² and the most recent studies, at least, have documented a role for this kinase in Erk signaling in thymocytes.^27,28 Whether PKCθ–/– mice have a defect in RasGRP1 phosphorylation remains to be determined.

RasGRPs are large multidomain proteins, and they could possess protein-protein interaction domains that contribute to their regulation. SKAP55 is an adaptor protein with PH and C-terminal SH3 domains that appears to be a particularly versatile regulator of RasGRP1 in T cells, in that one group claims SKAP55 is a negative regulator²⁹ and another claims just the opposite.³⁰ Both studies implicated the C-terminal region of SKAP55. In the former study, SKAP55 null mutant cells were shown to have enhanced Ras-Erk signaling and altered intracellular distribution of RasGRP1 compared to wild-type.²⁹

Additional insights into RasGRP1 regulation have come from the study of enzymes involved in DAG synthesis and metabolism. Conditional PLC-γ1 knockout mice have been used to demonstrate that this enzyme is required for thymocyte positive selection as well as Erk activation, although it also plays a role in negative selection.³¹ Likewise, chicken DT40 B cells that lack PLC-γ2 are unable to activate Ras in response to BCR stimulation.³² DAG can also be synthesized from membrane phospholipid in a 2-step process: phospholipase D (PLD) can generate phosphatidic acid that is then dephosphorylated by phosphatidate phosphatases (lipins).³³ This pathway has been implicated in RasGRP1 activation and may be the source of DAG in late stages after cell stimulation.³⁴ However, not all studies support the idea that DAG generated in this manner has signaling properties.³⁵ Certainly, DAG is a complex family of molecules with diverse structures, metabolism, and biophysical properties,³⁶ and not all forms of DAG may be equivalent regulators of C1 domain–signaling proteins. Interestingly, studies with Jurkat T cells overexpressing RasGRP1 have led to the suggestion that exogenous fatty acids can influence the side chain composition of DAG and thereby influence Ras-Erk signaling.³⁷ Dietary fatty acids are thought to influence immune function. Conceivably, RasGRP1 activity may be influenced by diet.

The signaling molecule Vav acts as an adaptor as well as a guanyl nucleotide exchange factor for Rho family GTPases and somehow functions to regulate PLC-γ1 in T cells and PLC-γ2 in B cells. Not surprisingly, then, mouse thymocytes and chicken DT40 B cells deficient in Vav are defective in RasGRP1 activation.^38,39 Overexpression studies in Jurkat T cells also point to a connection between Vav, PLC-γ1, and RasGRP1 signaling.⁴⁰ There is more to the story, however, at least in DT40 B cells. Vav can influence the actin cytoskeleton through its ability to activate the Rho family member Rac1. This process contributes to RasGRP1 plasma membrane localization in DT40 cells following BCR stimulation.⁴¹

DAG kinases (DGK) convert DAG to phosphatidic acid. By depleting membrane DAG, these enzymes can contribute to negative regulation of RasGRP1 in T cells. Overexpression of DGKα in Jurkat T cells attenuates RasGRP1 signaling and membrane association, while ectopic expression of a catalytically inactive mutant form of the enzyme has the opposite effect presumably by interfering with the endogenous enzyme.^42,43 DGKζ has also been implicated in attenuating RasGRP1 in T cells in vitro.⁴⁴ Analysis of mice lacking DAGKα and DAGKζ has confirmed that T cells use both enzymes to attenuate the levels of signaling DAG in anergic T cells.⁴⁵ Furthermore, primary mouse T cells rendered anergic in vitro by TCR stimulation in the absence of coreceptor stimulation show a striking increase in DGKα and DGKζ levels and defective Ras-Erk signaling.⁴⁶ Muted RasGRP1 activation in the presence of excess DAG kinase likely explains the defective Ras signaling and relative antigen nonresponsiveness of anergic cells. Interestingly, anergy can be effectively induced in some T cells by calcium ionophore treatment, and this is also associated with increased activity of NFAT transcription factors and increased expression of DGKα.⁴⁷ The latter findings highlight how calcium signals could act indirectly to influence RasGRP1 activity without necessarily involving its EF hands.

RasGRP1 regulation by the B cell receptor (BCR) in B cells is likely similar to that by TCR in T cells, but the coexpression of RasGRP1 and RasGRP3 in B cells complicates the functional analysis.¹⁸ Additionally, RasGRP1 appears to have functions in some blood lineages that do not express either TCR or BCR. RasGRP1 has been detected by immunoblotting in primary mouse mast cells.⁴⁸ RasGRP1-deficient mast cells display defects in FceR1-mediated mast cell degranulation and cytokine production. Rasgrp1–/– mice are resistant to anaphylaxis.⁴⁸ Interestingly, RasGRP1 signaling in mast cells does not involve the Raf-Mek-Erk effector arm. Rather, RasGRP1 was proposed to regulate Ras-mediated stimulation of the PI3 kinase pathway and granule exocytosis. Ras-GTP is known to interact with the catalytic domain of PI3 kinase in other systems.⁴⁹ In contrast to this idea, however, a concerted attempt to tie RasGRP activation to the PI3 kinase signaling system in lymphocyte cell lines provides no supporting evidence for the existence of this signaling axis.⁵⁰

NK cells express ITAM-containing activating receptors that are linked to PLC-γ1, and they express RasGRP1. Using a siRNA knockdown approach, human NK cells were shown to depend on RasGRP1 for Ras-Erk signaling, cytokine production, and cytotoxic killing.⁵¹ Mice lacking the Gfi1 transcription factor have impaired myelopoiesis and lymphocyte development. RasGRP1 transcript levels and G-CSF–dependent signaling to Erk are decreased in Gfi1–/– bone marrow–derived myeloid cells.⁵² Transduction of RasGRP1 cDNA into such cells rescues the signaling defect. The results were interpreted to mean that Gfi1 normally up-regulates Rasgrp1 gene transcription, which when adequately expressed supports G-CSF signaling from the G-CSF receptor to Erk. How the G-CSF receptor communicates with RasGRP1 is unknown. The literature does not support the proposal that DAG is a messenger in this cytokine receptor system.

Properties of RasGRP2

RasGRP2 was originally identified in a genome sequencing project and named HCDC25L (human CDC25-like).⁵³ The sequence was rediscovered the following year in differential display experiments aimed at identifying novel brain transcripts and was called CalDAG-GEFI.⁵⁴ This name refers to its presumed mode of regulation by DAG and calcium. The protein was independently rediscovered and renamed RasGRP2.⁵⁵ As in the case of RasGRP1, the evidence for calcium regulation through the EF hands is weak. More troubling is the repeated assertion that DAG regulates RasGRP2 as it does RasGRP1. The C1 domain of RasGRP2 does not bind DAG.⁵⁶ RasGRP2 fails to translocate to membranes in vivo in response to DAG analogs.⁵⁷ RasGRP2 may be expressed in 2 forms, a standard form and a variant longer form.⁵⁵ The existence of the longer form was imagined to arise from alternative splicing on the basis of an expressed sequence tag. When ectopically expressed in COS cells, the long form possesses a unique N-terminus that undergoes lipidation. However, the splice sites proposed to generate the longer form are not conserved in other species, and endogenous proteins corresponding to the long form have not been identified.

Despite initial reports that RasGRP2 can activate Ras,⁵⁵ there is a general consensus that it acts directly on Ras-related GTPases of the Rap category.⁵⁴ Rap had an early career as a Ras antagonist, but more recent proposals center around the idea that Rap-GTP can bind molecules like RapL and Riam and modulate the affinity and avidity of integrins. In T cells, for example, RasGRP2 and Rap may be able to regulate the interaction of LFA-1 with ICAM molecules on APC and thereby control the stability of the lymphocyte-APC partnership. Rap has also been proposed to activate B-Raf in some cells, much the same as Ras acts on Raf-1. Detailed RasGRP2 expression studies in different blood cell lineages have not been published. Based on the chromosomal position of the RasGRP2-encoding sequences and on the idea that RasGRP2 and Rap might regulate surface properties of lymphocytes, RasGRP2 was proposed to be the gene mutated in some cases of the familial disease leukocyte adhesion deficiency III (LADIII).⁵⁸ Unfortunately, the described mutations turned out to be natural polymorphisms. Careful analysis by 2 groups identified functionally defective LADIII mutations in the closely linked Kin3 gene.^59,60

Considering the confusion over many aspects of RasGRP2 research, it comes as a relief to learn that the encoded protein has an unambiguous and important function. Studies by Bergmeier and colleagues have shown that platelets from Rasgrp2–/– mice have defects in Rap activation and integrin-mediated adhesion and thrombus formation in response to thrombogenic agents.⁶¹ RasGRP2 has been proposed to serve as a hub for integrating calcium and DAG signals and, by means of Rap activation, governing platelet adhesion through adaptors and integrins such as αIIbβ3. Additionally, Rap can regulate release of platelet granule contents, including potent mediators such as thromboxane A(2), which can further activate platelets through surface receptors.⁶² Lest there be any doubt about the generality of these studies with mice, veterinary research has turned up 3 different dog breeds with bleeding disorders associated with RasGRP2 mutations, including truncating mutations in the RasGRP2 coding region.⁶³ We even have cows. A Simmental calf with a bleeding disorder had a RasGRP2 gene mutation.⁶⁴ Remarkably, a role for RasGRP2 and Rap in platelet adhesion by means of integrin regulation had earlier been inferred from RNA expression profiling and RasGRP2 cDNA transfer studies using megakaryocytes generated from cultured embryonic stem cells.⁶⁵ Because the RasGRP2 C1 domain does not bind DAG directly, the idea of PKC-mediated RasGRP2 activation warrants attention (Fig. 2). However, because RasGRP2 deficiency and PKC blockade synergistically block αIIbβ3 activation and platelet aggregation, PKC likely has other roles regulating integrins and adhesion in platelets.⁶⁶

Figure 2.

Regulation of RasGRP family members by DAG. Direct regulation through the C1 domain (solid rectangles) and indirect regulation by means of PKC-mediated phosphorylation in the N-terminal region of the CDC25 domain are indicated. PKC-mediated regulation of RasGRP2 is hypothetical.

Neutrophils isolated from Rasgrp2–/– mice also exhibit defects in Rap activation.⁶⁶ They are defective at β1 and β2 integrin activation, and they adhere poorly to fibronectin, fibrinogen, and mesenteric venules. They are also defective at migration into the peritoneal cavity after induced inflammation. Additionally, RasGRP2-deficient neutrophils exhibit deficiencies in in vitro chemotaxis assays, and these likely reflect underlying problems with the organization of the actin cytoskeleton.⁶⁷

Like Ras, Rap is lipidated and membrane bound. Because the C1 domain of RasGRP2 fails to bind DAG appreciably, and because the lipidated, long variant of RasGRP2 is entirely conjectural at this point, other membrane translocation mechanisms seem more plausible for RasGRP2. Vav-induced actin polymerization may contribute to RasGRP2 membrane localization,⁶⁸ as proposed for RasGRP1. Indeed, actin polymerized in vitro can bind to the amino-terminal region of RasGRP2 directly. Although this study mostly employed COS cells, which do not express endogenous RasGRP2, functional tests confirmed that Vav and RasGRP2 coexpressed in Jurkat T cells can cooperate to control surface integrin activation.⁶⁸

Finally, mRNA knockdown and overexpression studies in Xenopus laevis embryos revealed a clear role for the RasGRP2 ortholog in angiogenesis.^69,70 That phenotypic data supporting this proposed function have not been reported from mouse studies might reflect functional redundancy between RasGRP2 and RasGRP3.

Properties of RasGRP3

RasGRP3 was first identified in a large-scale cDNA sequencing project. Using an anti-RasGRP3 antibody, expression of the endogenous protein was detected in numerous B cell lines.²⁶ RasGRP3 is also expressed in some endothelial cells, as learned from an elegant gene trap strategy.⁷¹ Expression of RasGRP3 in B cells suggested that the protein serves downstream of BCR and that its function and regulation were analogous to those of RasGRP1 in T cells. To an extent, this seems to be the case. The RasGRP3 C1 domain binds DAG and DAG analogs with high affinity.⁷² RasGRP3 translocates to cellular membranes in response to DAG analog treatment.⁷² RasGRP3 is positively regulated by a DAG-responsive protein kinase(s) and, in fact, was the first RasGRP member to divulge this family secret.²⁶ RasGRP3 phosphorylation in chicken DT40 cells, in human B cell line Ramos, and in mouse primary B cells involves T133, which is homologous to T184 in RasGRP1.^23,73 Experiments exploiting ectopic expression in adherent cells have shown that novel PKC forms, such as PKCθ or PKCδ, might qualify as the responsible protein kinases.^26,74 Inhibitor studies also suggest that a novel form of PKC might be involved in RasGRP3 phosphorylation, at least when the protein is directly activated by a DAG analog. However, most B cells do not express PKCθ. In Ramos B cells, at least, RasGRP3 undergoes multisite stoichiometric phosphorylation. Therefore, RasGRP3 regulation by PKC-mediated phosphorylation is incompletely understood.

In other respects, RasGRP3 displays properties not shown by RasGRP1. Remarkably for a Ras activator, RasGRP3 has the ability to act additionally on Rap, at least when overexpressed in HEK293 cells.⁷⁵ Using a 2-hybrid screen, RasGRP3 was shown to interact through its C-terminal tail with dynein light chain (LC8).⁷⁶ This result might point to an alternate cell localization mechanism in addition to membrane recruitment by DAG.⁷⁶ Neither of these properties has been demonstrated with RasGRP3 endogenously expressed in blood cells.

Supporting the idea that RasGRP3 functions downstream of BCR, Rasgrp3–/– mice show modestly lower serum levels of some immunoglobulin types.¹⁸ Isolated Rasgrp3–/– B cells have reduced basal and BCR-regulated Ras-Erk signaling in vitro. They also show modestly impaired proliferation in response to stimulation with anti-IgM antibodies and costimulation with anti-CD40 antibodies. Rasgrp3–/– mice show no obvious defect in B cell differentiation.¹⁸

A complication in interpreting RasGRP3 function arises from the fact that RasGRP3 and RasGRP1 are coexpressed in most B cells.¹⁸ Indeed, analysis of Rasgrp1–/–; Rasgrp3–/– mice shows that double-mutant B cells have a greater defect in Ras-Erk signaling and proliferation than either single mutant. RasGRP1 and RasGRP3 appear to be at least partially redundant at the level of B cell function. However, the relationship between these 2 RasGRP family members is much more complex at the whole animal level.

As mentioned above, older RasGRP1 mice develop a lymphoproliferative autoimmune condition. This phenotype is associated with hypergammaglobulinemia, the presence of antinuclear antibodies and splenomegaly. Importantly, the condition depends on the activity of RasGRP1-deficient T cells, although some of the effects are manifest by B cells. As mentioned, Rasgrp1–/–; nu/nu–/– double-mutant mice, which lack a thymus and as a consequence have no mature T cells, do not develop splenomegaly or express antinuclear antibodies.⁷⁷ A remarkable feature of the Rasgrp1–/–; Rasgrp3–/– double-mutant mice is that the lymphoproliferative autoimmune phenotype characteristic of the older Rasgrp1–/– mice is absent or at least delayed.¹⁸ Rasgrp3–/– acts as a genetic suppressor of the late-onset Rasgrp1–/– mutant phenotype. This result was interpreted to mean that the double-mutant B cells were nonresponsive to the pathogenic influence of RasGRP1-deficient T cells.

To a certain extent, this intriguing T cell–B cell dialogue has been recapitulated in vitro. Rasgrp1–/– T cells possess a greater capacity to induce proliferation of cocultured B cells than do wild-type T cells. Reciprocally, B cells lacking either RasGRP1 or RasGRP3 are less responsive to Rasgrp1–/– T cells, while double-mutant B cells are virtually nonresponsive. At least one component of this T cell influence on B cell proliferation appears to be IL-4, as a neutralizing anti-IL4 antibody could negate the effect of T cells on B cell proliferation in vitro.⁷⁷

RasGRP3 may play a role in macrophage phagocytosis.⁷⁸ In macrophage cell lines, RasGRP3 is recruited to the cytoplasmic surface of the phagocytic cup by a mechanism involving PLC and DAG. Because phagocytosis is not influenced by Mek inhibitors, RasGRP3 may contribute to phagocytosis by acting on Rap rather than Ras. A role for RasGRPs in macrophage function has not been demonstrated using Rasgrp3–/– mice, however.

Disruption of RasGRP3 coding sequences by a retrovirus gene trap method revealed that RasGRP3 is expressed in endothelial cells.⁷¹ RasGRP3 loss does not impair normal vascular development, nor does it impair the ability of a model mammary tumor to recruit a supporting vasculature. However, blood vessel development by wild-type embryoid bodies in vitro was impaired by exposure to DAG analogs, whereas Rasgrp3–/– cells were indifferent to the presence of DAG analogs. RasGRP3 might be part of some redundant signaling network that controls angiogenesis.

Properties of RasGRP4

RasGRP4 was identified as a novel transcript expressed in mast cells and is claimed to be mast cell specific.⁷⁹ RasGRP4 cDNA sequences were also isolated from peripheral white blood cells from patients with acute myeloid leukemia.^80,81 Whether these later findings reflect a contamination of the malignant cells with mast cells, ectopic expression of a leukemogenic sequence in malignant cells, or a broader distribution of RasGRP4 expression in the normal myeloid lineage is unclear. In any case, RasGRP4 is likely the most restricted family member in terms of its expression.

RasGRP4 expressed ectopically in rodent fibroblasts shows all the hallmarks of a DAG-regulated Ras activator. PMA treatment of RasGRP4-expressing fibroblast cells results in morphological transformation and robust Ras and Erk activation.^79,80 Interestingly, the region corresponding to the PKC phosphorylation sites in RasGRP1 to RasGRP3 is not conserved in RasGRP4. Rather, this region in RasGRP4 is proline rich. Thus, RasGRP4 regulation appears to involve direct membrane recruitment by DAG but not PKC-mediated phosphorylation.²³ The proline-rich sequence may serve to activate RasGRP4 constitutively, contingent on independent binding of DAG to the C1 domain. The source of regulatory DAG for RasGRP4 in mast cells has not been identified, but PLC-γ1 activation downstream of c-kit/CD117 seems a good bet.

RasGRP4 transcripts are subject to alternative splicing, and such events may reflect a normal regulatory mechanism governing RasGRP4 protein expression levels as well as a mechanism leading to diseases such as mastocytosis, mast cell leukemia, and asthma.^79,82 Introduction of RasGRP4 cDNA into the RasGRP4-deficient mast cell line HMC-1 enhanced the ability of the cells to express prostaglandin D2 synthetase.⁸³ Similar studies showed that PMA treatment of RasGRP4-expressing HMC-1 cells resulted in massive transcriptional changes affecting hundreds of genes.⁸⁴ Of these, the interleukin-13 receptor (IL-13Rα2) was prominent, and the results were confirmed with protein expression studies. Activation of normal mast cells might similarly lead to a RasGRP4-dependent up-regulation of IL-13Rα2, making the mast cells less responsive to IL-13 secreted by Th2 cells; this receptor is thought to act as a decoy that functions by sequestering IL-13 and dampening mast cell activity. The phenotype of Rasgrp4–/– mice will hopefully shed light on the function of the protein in mast cells and other lineages.

In summary, although vertebrate RasGRP1 to RasGRP4 each exhibits a similar domain structure, they have distinct modes of regulation and GTPase specificity. They have distinct but partially overlapping expression patterns with prominent expression in blood cells. A single RasGRP gene exists in the genome of Caenorhabditis elegans. With a GFP reporter construct driven by putative RasGRP 5′ regulatory sequences, expression of the RasGRP-encoding gene was shown to start at the 3-fold stage of embryogenesis, after neurogenesis is complete, in all neurons.⁸⁵ Perhaps the ancestral metazoan function of RasGRP initially evolved to perform some postdevelopment neural function. After repeated sequence duplication, the protein family members have diversified in terms of regulation and function and have found employment in limited vertebrate cell types, especially those of the blood system. Notably, the site corresponding to threonine phosphorylation in RasGRP1 to RasGRP3 is occupied by a functionally conservative serine in the C. elegans protein, so dual regulation by DAG may be an ancient adaptation.

Whither RasGRP?

One of the most vexing questions about RasGRPs concerns the intracellular position of the proteins while active. Besides imaging GFP-tagged RasGRP1, or detecting endogenous RasGRP1 using immunological probes, some studies have exploited the different properties of K-, N-, and H-Ras to infer where RasGRPs work. H-Ras and N-Ras are subject to C-terminal processing including prenylation in the endoplasmic reticulum. They are additionally palmitoylated on the Golgi membranes, before mature protein reaches the inner surface of the plasma membrane. K-Ras is expressed as 2 proteins. Alternative splicing events select one or the other of the 3′ coding exons, 4A or 4B. K-Ras4A protein is a minor species in most tissues and resembles H- and N-Ras in its processing. K-Ras4B, the major form, is prenylated but not palmitoylated. It passes from endoplasmic reticulum membranes directly to the plasma membrane. A basic C-terminal peptide region cooperates with the prenylation signal to secure plasma membrane association. The differential lipidation suggests that Ras proteins might display different coupling efficiency to downstream effectors, such as Raf and PI3K, which need membranes for activation and to access substrate, respectively.

Supporting the idea of nonequivalence among Ras proteins, K-Ras knockout mice have an embryonic lethal phenotype, while N-Ras null mutant mice, H-Ras null mutant mice, and even N-Ras, H-Ras double null mutant mice, have very subtle phenotypes.⁸⁶ However, mice lacking K-Ras coding sequences, but expressing H-Ras from the endogenous K-Ras locus, are viable.⁸⁷ It appears that the more ubiquitous expression of K-Ras is largely the basis of its preeminent function. Returning to the question of where RasGRPs function, it should be clear that, especially in light of their membrane association with DAG, these Ras GEFs might interact with different Ras proteins in different cellular compartments.

The first experiments aimed at determining the cellular localization of RasGRP1 used a GFP-tagged protein expressed in fibroblasts and the DAG analog PMA to elicit protein activation.⁷ The protein accumulated rapidly on cellular membranes, including the plasma membrane with nuclear membrane association observed after several minutes. Likewise, RasGRP3 was localized to internal membranes as a perinuclear accumulate in PMA-stimulated HEK293 cells.⁷²

Studies of TCR-stimulated Jurkat T cells provided support for the view that RasGRP1 can function on internal membranes.⁸⁸ By cotransfecting Jurkat T cells with a Ras-GTP probe composed of the Ras-binding domain of Raf fused to GFP and YFP-tagged RasGRP, RasGRP1 and Ras-GTP accumulation on the Golgi in a PLC-γ1 and calcium-dependent manner were documented.⁸⁸ The translocation of RasGRP1 was rapid and did not appear to involve retrograde endosome trafficking. This proposed mode of RasGRP1 regulation appears to function surprisingly well in many cells that do not normally express detectable RasGRP1, such as fibroblasts, COS cells, and PC12 cells.

Additional studies with Jurkat T cells showed that low-intensity TCR signaling was exclusively linked to N-Ras activation by RasGRP1 on the Golgi.⁸⁹ In contrast, stimulation with anti-CD3 antibodies plus costimulation using ICAM to engage LFA-1 resulted in activation of K-Ras as well as N-Ras, leading to accumulation of Ras-GTP on both the Golgi and the plasma membrane.³⁴ The costimulatory effects of ICAM were observed in a Jurkat T cell clone lacking PLC-γ1 and were dependent on the activity of PLD2, although a compensatory role for PLC-γ2 in DAG production in these cells cannot be ruled out.⁹⁰

Some of these studies required overexpression of Ras for detection of Ras-GTP, and some studies used overexpression of RasGRP1, introducing uncertainty about the generality of the results. Not all RasGRP1 localization studies are subject to this criticism. In SKAP55-deficient T cells, for example, increased Ras-Erk signaling was associated with increased association of endogenous RasGRP1 with endomembranes.²⁹

Most in vitro models of immune receptor signaling offer only a rough approximation of the situation in vivo. An elegant study using graded stimulation of TCR transgenic thymocytes with a related series of peptide antigens complexed with MHC tetramers supported the idea that RasGRP1 signaling might occur on internal membranes.⁹¹ In thymocytes stimulated with positively selecting peptides, RasGRP1 was recruited to internal membranes, and phospho-Erk was distributed throughout the cell. Negatively selecting peptides, despite their marginal increase in TCR affinity/avidity, caused thymocytes to accumulate RasGRP1, Grb-2, Sos, and phospho-Erk nearer the plasma membrane. Several previous studies had shown that positively selecting peptides engender modest but sustained Erk activation while negatively selecting ones evoke transient but robust Erk phosphorylation. Thus, at the boundary of antigen-TCR encounters defining positive versus negative selection, RasGRP1 participates in a digital biochemical switch with qualitative positional features as well as quantitative biochemical ones, potentially enabling the cell to make a life or death decision. Importantly, however, the localization of Ras-GTP in this system has not been reported.

Not all studies agree with the view that RasGRPs activate Ras on internal membranes. The initial characterization of Ras activation in Jurkat T cells showed that K-Ras was activated at both high- and low-level TCR stimulation, implying at least some activation outside of the Golgi under both conditions.⁹ Likewise, PLC-γ2–deficient DT40 B cells show defects in BCR-stimulated activation of K-, N-, and H-Ras.³² GFP-tagged RasGRP1 ectopically expressed in chicken DT40 cells was localized to the plasma membrane upon BCR stimulation.^20,41 Jurkat T cells stimulated with anti-CD3 antibodies showed accumulation of RasGRP1 at the plasma membrane, although there was a tendency for the molecules to accumulate in a cap-like structure.⁴⁰ When mouse TCR transgenic T cells were conjugated to APC-bearing cognate antigen, RasGRP1 was localized to the immunological synapse in a DAG-dependent fashion.⁴⁶ Similarly, in Jurkat T cells conjugated to microspheres coated with anti-CD3 and anti-CD28 antibodies, RasGRP1 accumulated at the synapse.⁹² When T cells contact mast cells, increased levels of RasGRP1 and Ras-GTP are visualized on the T cell plasma membrane.⁹³ Localization of RasGRP4 in mast cells was determined to be either on the inner surface of the plasma membrane or in the cytoplasm using immunoelectron microscopy.⁸²

The recent work from Rubio’s group bears careful consideration. Using an ultrasensitive probe composed of tandem repeats of GFP and RBD, these researchers were able to detect endogenous Ras-GTP solely on the plasma membrane of TCR-stimulated Jurkat T cells.⁹⁴ In primary T cells activated by contact with superantigen-loaded B cells, Ras-GTP accumulated only on the plasma membrane within the immunological synapse. Furthermore, a palmitoylation-defective mutant version of N-Ras, which resides on internal membranes, was incapable of TCR-induced activation. Collectively, these studies cast serious doubt on the idea that RasGRP1 acts on the Golgi, at least in mature T cells, as do the findings that active PLC-γ1,^95,96 PKCθ,^92,97 and DAG⁹⁸ all colocalize to the synapse while DAGKα is activated by translocation to the plasma membrane.^43,99 Finally, it should be kept in mind that mice deficient in N-Ras, or both N-Ras and H-Ras, have subtle phenotypes that do not include obvious defects in thymocyte differentiation, implying that activation of these Ras isoforms by RasGRP1 on the Golgi cannot be a critical process in positive selection.⁸⁶

Besides the controversy concerning RasGRP1 localization, attempts to confirm the expectation that the C1 domain controls membrane association of the protein have turned up several surprises. A functional C1 domain contributes to DAG- and DAG analog–dependent membrane attachment and signaling, but experiments with truncated proteins have shown that this is not the whole story. In DT40 cells, the distal C-terminal tail also has a plasma membrane–targeting domain, PT, which is negatively regulated by another region in the C1 proximal part of the C-terminal tail, called SuPT.¹⁰⁰ Mutations expected to abolish conserved calcium-binding residues in the first EF hand decreased membrane attachment of full-length RasGRP1. From the analysis of truncated proteins, a mechanism was proposed whereby EF1 normally counteracts the SuPT domain, thereby activating PT.²⁰ This property of the EF1 hand does not depend on calcium concentration. Rather, alternative splicing that skips exon 11 was proposed to generate a protein missing a substantial portion of EF1, thereby negatively regulating overall RasGRP1 activity. Indeed, the RasGRP1 Δex11 protein showed defective plasma membrane targeting in DT40 cells that was largely corrected by additional deletion of SuPT.²⁰ The analysis of truncated proteins in DT40 cells additionally implicated both the REM and CDC25 domains in BCR-stimulated plasma membrane targeting, leaving only the extreme N-terminus of RasGRP1 without a putative role in membrane localization. While there is some concern that protein truncation studies may be much cruder than point mutant analysis, these studies provide hints that control of RasGRP1 targeting to membranes may be much more complex than initially envisaged. Importantly, association of the catalytic regions with substrate GTPase might influence RasGRP localization studies. While the normal physical interaction of a Ras GEF and its client GTPase is expected to be brief, in cells overexpressing both partners, artificial Ras-RasGRP1 complexes might accumulate on membranes festooned with overexpressed Ras. Alternatively, Ras-GTP might bind to an allosteric site on RasGRP1,²⁰ further complicating localization studies.

Relationships between RasGRPs and Other Ras Regulatory Mechanisms

Regulated Ras GEF activity is thought to promote Ras activation in most situations. According to this view, Ras GAPs could act constitutively as brakes on Ras signaling, returning Ras to its GDP-bound inactive state. However, although the original PKC–Ras GAP hypothesis has generally fallen into disfavor, it is noteworthy that PMA stimulation of a human neuroblastoma cell line leads to Ras activation and PKC-mediated phosphorylation within the GAP-related domain of neurofibromin,¹⁰¹ a Ras GAP that is implicated in lymphocyte regulation.^102,103 Further work is required to identify the key Ras GAPs and assess their modes of regulation in blood cells.

Specific inhibitors of Sos proteins are not available. The presence of 2 very similar proteins, Sos1 and Sos2, in most vertebrate cells has impeded genetic analysis. Consequently, experimental evidence for Sos involvement in Ras activation in lymphocytes and other blood cells has mostly revolved around the analysis of signaling complexes detected by coimmunoprecipitation. Using the venerable logic that proteins that stay together must play together, a number of studies have concluded that Sos is recruited by Grb-2 to tyrosine-phosphorylated LAT, leading to Ras activation. Assuming this recruited Sos exists in physiological levels in intact cells, questions arise as to why Sos does not provide a function redundant to that of RasGRP1 and why Rasgrp1–/– mice have any T cell lineage phenotype.

Sos is subject to positive feedback regulation whereby Ras-GTP binds to the catalytic domain in an allosteric fashion and increases the Ras GEF’s catalytic activity.¹⁰⁴ Roose and colleagues have argued that RasGRP1 and Sos function in a hierarchical fashion in T cells, with RasGRP1 furnishing the initial Ras-GTP that primes Sos. Early evidence for this idea stemmed from the observation that a constitutively membrane-targeted form of Sos1 acts as a constitutively activated Ras GEF in wild-type Jurkat T cells but fails to do so in a Jurkat clone that is deficient of RasGRP1.²⁴ Similarly, signaling from tethered Sos appears to depend on DAG levels and PKC activity, hallmarks of RasGRP regulation. A previous claim that Sos1 and Sos2 were not important for Ras signaling in DT40¹⁰⁵ was modified to allow for Sos involvement at moderate BCR stimulation levels.²⁴ Computer simulation studies have demonstrated that such a hierarchical RasGRP-[Ras-GTP]-Sos-[Ras-GTP] signaling system has intrinsically nonlinear properties.¹⁰⁶ In thymocytes, according to this ingenious model, graded analog signaling from antigen-TCR can be converted to a digital response when RasGRP1-activated Ras reaches a key level that sufficiently engages Sos and triggers the positive feedback loop. This mechanism could allow thymocytes to discriminate between positively and negatively selecting TCR ligands and to translate these antigen-TCR encounters into the transient and intense versus prolonged and moderate Erk signaling that distinguish negative and positive selection, respectively.¹⁰⁷ How this model lends itself to integration with the spatial control model described above remains to be seen. RasGRP1 could work alone, internally, during positive selection. Alternatively, RasGRP1 could prime Sos on the plasma membrane, and signaling would be subject to negative feedback regulation by unknown means to yield intense but transient signals characteristic of negative selection. However, in DT40 cells at least, it is the suboptimal signaling that depends on both RasGRP and Sos.¹⁰⁶

Obviously, the proposal is untenable if Sos cannot be more directly implicated in TCR signaling. Grb-2 is much more abundant than Sos¹⁰⁸ and may have Sos-independent functions in lymphocytes. Grb-2 partners with dynamin to control antigen-BCR internalization,¹⁰⁹ and it binds the thymocyte positive selection protein Themis.¹¹⁰ Therefore, Grb-2 mutant phenotypes cannot be readily used to implicate Sos. Indeed, analysis of a conditional Grb-2 knockout mutation has revealed an unexpected role for this adaptor at the top of the TCR signaling cascade.¹⁰⁸

It may also be worth noting that while molecular immunologists were trying to shoehorn Sos into the TCR signaling network, our view of Sos regulation in nonblood cells evolved significantly. Substantial Sos output still remains when the Grb-2–binding domain is deleted.¹¹¹ Besides the positive feedback loop afforded by allosteric binding to Ras-GTP, membrane lipids such as PIP3¹¹² and PA¹¹³ have been implicated in Sos activation. Thus, the existence of LAT–Grb-2–Sos complexes in lysates might be viewed as something less than synonymous with Sos activity. Furthermore, results obtained with Jurkat T cells, which are PTEN deficient and have high basal levels of PIP3, might lead to overestimates of the role of Sos in TCR signaling. The analysis of T cells lacking both Sos1 and Sos2 by conditional genetic ablation or siRNA knockdown will be required to clarify the role of Sos proteins in TCR signaling. Interestingly, RasGRP1-deficient mature T cells are capable of activating Ras in response to IL-2.¹⁷ Therefore, we cannot rule out the proposals that Sos functions independently in cytokine signaling pathways or that Sos and RasGRP1 act redundantly, for example, in pre-TCR signaling.

RasGRF1 and RasGRF2, comprising the third type of Ras GEF, are regulated by calcium calmodulin and protein phosphorylation. Of the two, RasGRF2 is expressed in T cells and appears to play a minor role in Ras activation,¹¹⁴ further complicating attempts to define the relative importance of Ras GAPs, RasGRPs, and Sos proteins in Ras regulation.

RasGRPs in Autoimmune Disease

Systemic lupus erythematosus (SLE) is a complex autoimmune disease characterized by high levels of serum antinuclear antibodies. A number of studies have led to the hypothesis that defective Erk signaling in CD4⁺ T cells could typify SLE. Reduced Erk signaling is thought to lead to reduced expression of DNA methyltransferase (DNMT1).¹¹⁵ Decreased replication-dependent methylation of CpG islands is thought to result in increased expression of genes that contribute to T cell autoreactivity, such as that encoding LFA-1.

The presence of antinuclear antibodies in aged Rasgrp1–/– mice suggests that aberrant RasGRP1 signaling might be at the root of SLE. Indeed, CD4⁺ T cells from SLE patients have abnormally low levels of RasGRP1.¹¹⁶ The underlying molecular defect is thought to be aberrant splicing arising from unknown causes rather than from coding mutations. More recently, reduced RasGRP1 was confirmed in CD4⁺ T cells from SLE patients as well as SLE-like phenotypes in MLR/lpr mice.¹¹⁷ However, a completely different mechanism underlying RasGRP1 deficiency was invoked. Abnormally high expression of 2 micro-RNAs, miR-21 and miR-148a, was shown to negatively regulate DNMT1 expression. RasGRP1 mRNA possesses a 3′ target sequence, rendering it susceptible to miR-21–induced degradation, while miR-148a acts directly on the DNMT1 mRNA. These studies provide important clues about the origins of SLE and suggest treatment strategies.

RasGRP2 was identified as a gene differentially up-regulated in active SLE CD4⁺ T cells,¹¹⁸ an observation that may be important in light of the evidence that RasGRP2 and Rap can influence LFA-1 activity. Familial studies have implicated genetic variants at or near the RasGRP3 locus in SLE susceptibility, possibly reflecting parallels to the double-mutant mice observations described above.¹¹⁹ Perplexingly, gene mapping studies have also implicated a gene variant at or near the RasGRP1 locus in another autoimmune disease, type I diabetes.¹²⁰ Perhaps genetic background or environmental factors can determine whether a common RasGRP1 variant contributes to SLE or type I diabetes. Alternatively, RasGRP1 might normally provide a quantitative range of signaling levels conducive to normal T cell development and function. Quantitative variants yielding signaling activity either above or below this range could contribute to one or the other of these autoimmune diseases.

RasGRPs in Cancer

Several animal studies support the idea that RasGRPs could contribute to cancer cell phenotypes. Retrovirus insertions adjacent to RasGRP1 and RasGRP2 were identified in large-scale screens for mouse leukemogenic genes.^121-124 Transgenic expression of RasGRP1 in the mouse thymus using the Lck proximal promoter leads to thymic lymphoma.¹²⁵ DGKα–/–; DGKζ–/– double-deficient animals also have increased propensity for developing thymic lymphoma.¹²⁶

Two groups isolated RasGRP4 cDNAs from malignant cells of acute myeloid leukemia (AML) patients.^80,81 In a mouse model of AML based on excess Ras signaling, resistance to Mek inhibitors was associated with retrovirus insertions that up-regulated either RasGRP1 or RasGRP4 expression.¹²⁷ Global transcript analysis identified RasGRP1 as one component of a 2-gene classifier scheme that best predicts the response of AML patients to treatment with a farnesyl transferase inhibitor.¹²⁸ How RasGRP1 RNA levels could help predict the sensitivity of a cancer cell to a failed anti-Ras drug is anyone’s guess. The role of abnormal RasGRP signaling in the development of human blood cancers is unclear, although very recent results indicate that RasGRP3 influences prostate cancer cell phenotypes.¹²⁹

RasGRPs as Drug Targets

Calphostin C is thought to block C1 domain function and has been used to inhibit both PKC and RasGRPs.⁸ However, the compound is not potent, and there is some doubt about the specificity of this compound for C1 domain proteins. Inhibition of RasGRPs can be achieved indirectly by blocking their phosphorylation with inhibitors that block PKC catalytic activity.²⁶ Such an approach may be fraught with problems because PKCs may have diverse targets and multiple, redundant PKCs may regulate individual RasGRPs. Nonetheless, some of the effects of promising PKCθ inhibitors in T cells, for example, may arise from diminished RasGRP1 activity.

In contrast to the situation with inhibitors, RasGRP-activating compounds appear bountiful. These include both naturally occurring DAG analogs, such as bryostatin-1, and unrelated DAG analogs synthesized by chemists. In the latter category, considerable success has been achieved in designing DAG analogs with selectivity for RasGRP versus PKC C1 domains.¹³⁰ Chemical synthesis has also afforded a relatively inexpensive alternative to natural product isolation, a good example being the synthesis of structurally simplified analogs of bryostatin-1.¹³¹

As reviewed elsewhere,^132,133 bryostatin-1 was derived from a marine invertebrate and was identified on the basis of preventing growth of the p388 murine lymphoma cell line. In preclinical studies, bryostatin-1 demonstrated a number of properties that highlighted its potential as an anticancer compound. These included intrinsic effects on malignant cells, either alone or in combination with conventional chemotherapeutic drugs. A number of studies demonstrated that bryostatin-1 also had the potential to act on cancer cells indirectly, for example, by activating normal lymphocytes to attack cancer cells. These early studies were largely driven by empirical evidence with scant understanding of how bryostatin-1 impacted cellular behavior. DAG analog action was thought to proceed from either increased PKC activity or from decreased PKC activity secondary to induced PKC proteolysis and down-regulation. Even the means whereby altered PKC activity was related to cellular properties was virtually opaque. Considering this lack of understanding, it is somewhat surprising that substantial efforts were made to evaluate the potential of bryostatin-1 in cancer patients. In retrospect, it seems likely that many of the experimental effects of bryostatin-1, at least in lymphocytes, were mediated by the combined activities of PKC and RasGRPs as well as other C1 domain proteins.

Bryostatin-1 is capable of activating RasGRP1 signaling in Jurkat T cells, and it binds directly to the C1 domain.^8,9 Bryostatin-1 and the related DAG analog “pico” are capable of activating RasGRP1 in several signaling assays, including those employing primary human peripheral T cells.¹³⁴ The B cell non–Hodgkin’s lymphoma cell line Toledo undergoes virtually 100% cell death after exposure to low doses of either DAG analog.¹³⁵ This process is blocked by PKC and Mek inhibitors and therefore almost certainly involves DAG analog activation of RasGRP1/3. Cell death is by apoptosis, and evidence was provided that Erk-dependent proapoptotic phosphorylation of Bim was involved.

Bim is a BH3-only member of the Bcl-2 family. It is an important player in chemotherapeutic mechanisms as well as a key regulator of normal B cell apoptosis.¹³⁶ Besides being of interest from the point of view of B-NHL therapy, RasGRP-dependent apoptosis in Toledo B-NHL cells might offer insights into the mechanism of B cell negative selection. Toledo B-NHL cells appear to represent germinal center–stage B cells, which normally undergo BCR affinity maturation and clonal deletion. Interestingly, apoptosis in Toledo B-NHL cells demonstrated a strong dependence on sustained RasGRP-Erk signaling. The high mutual affinity of Bcl-2 family members likely results in a fairly stable antiapoptotic deployment in healthy lymphocytes, and reassortment to a proapoptotic pattern may occur slowly. Possibly, this situation permits the B lymphocyte to interpret signals for high-avidity, high-affinity antigen interactions and channel the cellular response to negative selection.¹³⁵

DAG analogs and RasGRP signaling appear to have distinct properties in other blood cells that might be relevant to controlling cancer cell phenotypes. RasGRP1 expression in an immature B cell line, WEHI-231, promotes apoptosis.¹³⁷ Strangely, this involves inhibition of NFκB signaling rather than activation of Erk signaling. Equally intriguing, cell killing requires the EF hands of RasGRP1. EL4 mouse lymphoma cells are growth arrested by PMA treatment. This must involve RasGRP1 because selected PMA-resistant clones lacked RasGRP1 expression and had reduced PMA-induced signaling to Erk.¹³⁸ Furthermore, down-regulation of RasGRP1 with siRNA confers a PMA-resistant phenotype on otherwise sensitive EL4 clones. Bryostatin-1 induces Mek-dependent differentiation in the acute lymphoblastic B cell leukemia cell line Reh,¹³⁹ presumably by activating RasGRP1/3. Besides affecting cancer cell phenotypes directly, DAG analog–induced RasGRP signaling might also favorably modify cancer cell responses to other therapeutic agents. Further interrogation of the RasGRP family will surely provide additional information about their normal functions and modes of regulation as well as their contributions to disease and potential as drug targets.