Kinase Suppressor of Ras Couples Ras to the ERK Cascade during T Cell Development

Ras and its effectors are important during T cell development in the thymus. In particular, several lines of evidence indicate the requirement for the Ras/Raf/MEK/ERK pathway in directing positive selection, the transition that a small percentage of immature double-positive (DP²; CD4⁺CD8⁺) cells undergo to become either mature CD4 single-positive (CD4⁺ SP), or mature CD8 SP cells (CD8⁺ SP) (1). The expression of dominant-negative forms of either Ras or MEK in transgenic animals inhibits the generation of mature single-positive thymocytes due to a specific blockade in positive selection (2, 3). Furthermore, ERK1 null mutant animals exhibit similar defects in thymocyte maturation, and the inhibition of ERK activity with chemical inhibitors is sufficient to block positive selection in a fetal thymic organ culture assay (4, 5).

The Ras signaling cascade is essential for many growth and developmental processes. One well-characterized pathway through which Ras mediates its effects is via the sequential activation of the cytoplasmic kinases Raf, MEK, and ERK (6). Upon Ras signaling, membrane-localized Raf is activated. Activated Raf then phosphorylates and activates MEK, which in turn phosphorylates and activates the MAPK, ERK. However, the regulation of the interactions among these effectors is not completely understood. Furthermore, Ras also uses other downstream effector pathways to elicit its multiple effects (7, 8). A central and not well-characterized aspect of Ras signaling concerns how the multiple downstream effector pathways are coordinately activated to induce highly specific effects.

One mechanism that is used to activate specific signaling pathways among closely related molecules involves scaffolding molecules. Scaffolding molecules selectively bind multiple molecular components and specifically enhance signaling by these molecules by facilitating their association with appropriate components and/or by reducing nonproductive interactions (9, 10). Although scaffolding molecules have been well documented in yeast, only recently has it been shown in mammalian cells that molecules such as JNK-interacting protein 1 and MEK partner 1 are used as scaffolds to enhance signaling through the MAPKs JNK and ERK, respectively (11, 12, 13, 14).

Kinase suppressor of Ras (KSR) was initially identified as a positive regulator of Ras signaling in genetic screens performed in Caenorhabditis elegans and Drosophila melanogaster (15, 16, 17), and homologous proteins were discovered in mammals (15). The protein sequence of KSR is most similar to Raf, with the highest homology corresponding to the kinase domains. However, the necessity of the putative kinase domain for KSR function is unclear (18). Epistasis studies suggest that KSR functions either in parallel to, or downstream of Raf (15). KSR is constitutively associated with MEK1 and 2 and can directly interact with ERK1 and 2 (19, 20, 21, 22, 23, 24, 25). KSR is also able to interact with Raf as well as other signaling molecules such as 14-3-3 (21, 22, 25, 26, 27). Association with 14-3-3 may be important for the negative regulation of KSR activity because this interaction prevents KSR membrane translocation (28, 29).

Although KSR was originally found to augment Ras signaling, ensuing experiments demonstrated that KSR can also negatively regulate Ras signaling when expressed at high levels (19, 20, 22, 24, 25, 30, 31). These observations are consistent with a scaffolding function because scaffolds enhance positive signaling by increasing molecular interactions until the point at which the scaffold is in excess, thereby resulting in the formation of nonfunctional complexes (10, 32). Recently, it has been demonstrated in a Drosophila cell line that KSR initiates complex formation between Raf and MEK, thereby allowing for the activation of MEK by Raf (27). These results, in combination with the findings that KSR associates with a number of molecules, indicate that KSR can act as a scaffold to form a molecular bridge between Raf and MEK to facilitate activation of this cascade.

To determine whether KSR expression might play a role in modulating Ras function during T cell development, we overexpressed it in T cell precursors using retroviral infection, and its effect on T cell development was tested in reaggregate fetal thymic organ cultures (rFTOC). Overexpression of KSR in rFTOCs blocks the DP to SP transition during positive selection. The magnitude of this effect is comparable to that of a catalytically inactive MEK (dead MEK (dMEK)) in the same system, and the overexpression of KSR, as well as that of dMEK, blocks ERK activation in a thymoma cell line. Furthermore, structure-function analysis of KSR in this system showed that these effects of KSR overexpression require an interaction between KSR and MEK, suggesting that during positive selection of T cells, KSR links Ras signaling to ERK activation.

A 4.1-kb EcoRI murine DNA fragment from mKSR pBluescript (provided by M. Therrien, Clinical Research Institute, Montreal, Canada) was subcloned into the EcoRI sites of the murine stem cell virus-internal ribosomal entry site-GFP (Mig) vector (provided by L. Van Parijs, Massachusetts Institute of Technology, Boston, MA). The KSR S392A, I397A/V401A, R589K, R589M, C809Y, and cysteine-rich motif (CRM) mutants were generated by PCR using the following primers: S392A, 5′-CTT CGG AGG ACA GAG GCA GTC CCG TCA GAT ATC and 5′-GAT ATC TGA CGG GAC TGC CTC TGT CCT CCG AAG; I397A/V401A, 5′-CCG TCA GAT GCC AAC AAC CCA GCG GAC AGA GCA GCA GAG and 5′-CTC TGC TGC TCT GTC CGC TGG GTT GTT GGC ATC TGA CGG; R589K KSR, 5′-GGC GAG GTG GCC ATT AAG CTG CTG GAG ATG GAC-3′ and 5′-GTC CAT CTC CAG CAG CTT AAT GGC CAC CTC GCC-3′; R589M, 5′-GGC GAG GTG GCC ATT ATG CTG CTG GAG ATG GAC-3′ and 5′-GTC CAT CTC CAG CAG CAT AAT GGC CAC CTC GCC-3′; C809Y, 5′-GTC GGC GAG ATC CTG TCT GCC TAC TGG GCT TTC GAT CTG CAG and 5′-CTG CAG ATC GAA AGC CCA GTA GGC AGA CAG GAT CTC GCC GAC; and CRM, 5′-GTG AAG TCC AAA CAC TCC AGG TTA AAA TGC C and 5′-CCT GGA GTG TTT GGA CTT CAC GCC. DNA sequences encoding the mutated codons are underlined in boldface. These primers were paired with T3/T7 primers, and the resultant 4.1-kb product was digested with EcoRI and subcloned into Mig digested with EcoRI. The entire coding region of each of these products was sequenced using a PerkinElmer (Wellesley, MA) sequencing system.

RNA samples from spleen and thymus from 4- to 6-wk-old C57BL/6 mice were generated using the TRIzol system (Invitrogen Life Technologies, Carlsbad, CA). An equivalent amount of RNA for each sample was subject to reverse transcription, and the resulting cDNA was analyzed for KSR and GADPH expression by standard PCR and gel electrophoresis protocols using the following primers: GADPH, 5′-AAT TCA ACG GCA CAG TCA AGG CCG AGA ATG and 5′-GCG GCA CGT CAG ATC CAC GAC GGA C; KSR, 5′-GCT TAG TGT GAC CCC AAG C and 5′-AGG GCG GAC ACG GAG ATG. The expression level of KSR was determined by standardizing the band intensity from each reaction to that of a GADPH control as measured by densitometry using the Alphaimager densitometry analysis program (Alpha Innotech, San Leandro, CA). As a negative control, a reaction lacking reverse transcriptase is included to demonstrate that the bands obtained are the result of reverse-transcribed RNA. The identities of the reverse-transcribed DNA products were confirmed by sequencing following subcloning into the TOPO-TA cloning vector (Invitrogen Life Technologies).

The 16610 D9 cells (provided by S. Hedrick, University of California, San Diego, CA) were cultured in RPMI 1640/10% FBS and infected with retroviruses by centrifugation at 1400 rpm at room temperature for 1 h. Twenty-four hours postinfection, cells were stimulated with 6 nM PMA for 10 min and harvested according to the protocol in Chow et al. (33). Following fixation, cells were stained with either rabbit or mouse anti-phosphorylated ERK Abs (NEB, Beverly, MA). Following staining with the primary Abs, cells were stained with either Cy5-conjugated anti-rabbit or Cy5-conjugated anti-mouse Abs (Jackson ImmunoResearch Laboratories, West Grove, PA) for 15 min and washed with 1× PBS/4% FBS. Samples were then analyzed by flow cytometry using a FACSCalibur (BD Biosciences, San Jose, CA) and the Flow Jo analysis program (TreeStar Software, Ashland, OR).

Five days before reaggregation, host thymic lobes are microdissected from E15-E16.5 fetuses (from B6D2F₁ pregnant females). Lobes are then washed in excess medium supplemented with 1.35 mM 2′deoxyguanosine (dGuo; Sigma-Aldrich, St. Louis, MO) at 37°C/7% CO₂ for 1 h (34). The lobes are placed onto Millipore (Bedford, MA) 050 PICM filters within a six-well dish (1 ml of 1.35 dGuo/well) and incubated at 37°C/7% CO₂ for 5 days. After dGuo treatment, lobes are washed three times in excess DMEM/10% FBS, once with 1× PBS, followed by 0.05% trypsin/EDTA treatment for 40 min to dissociate stromal cells from the thymic matrix. Postdissociation, cells are incubated in DMEM/10% FBS medium at 37°C/7% CO₂ until donor thymocytes are prepared (described below).

On the day of reaggregation, donor thymocytes are microdissected from E15.5-E16.5 fetuses (from C57BL/6, B6D2F₁, or MHC° pregnant females (as listed in the text and figures)). Thymic lobes are dissociated with type IV collagenase (Sigma-Aldrich; 2 mg/ml) prepared in RPMI 1640/20 mM HEPES at 37°C/7% CO₂ for 0.5 h. Two million thymocytes are cultured with retroviral particles in a 24-well dish and centrifuged (1400 rpm) at room temperature for 1 h. The retroviral supernatant was replaced with DMEM/10% FBS, and cells were cultured for a minimum of 2 h at 37°C/7% CO₂. To reaggregate lobes, 2 × 10⁵ donor thymocytes are combined with 2 × 10⁵ host stromal cells in a total volume of 0.7 μl vol of DMEM/10% FBS and placed on filters for air-liquid interphase culture and incubated for the designated time periods.

Lobes were harvested and treated with 2 mg/ml T4 collagenase in RPMI 1640/10 mM HEPES at 37°C/5% CO₂ for 1 h in a 96-well dish. Cells were centrifuged and resuspended in 1× HBSS/4% FBS plus dilutions of fluorophore-conjugated Abs to TCRβ, CD4, and CD8α (BD Pharmingen, San Diego, CA). Samples were then analyzed by flow cytometry using a FACSCalibur (BD Biosciences) and the Flow Jo analysis program (TreeStar Software).

Activation of Ras and of the ERK MAPK cascade is required during positive selection in the thymus (2, 3, 35), although ERK is probably not the only Ras downstream effector implicated in this process (36). Therefore, the identification of molecules necessary for coordinating and facilitating the activity of this pathway is important to understanding Ras function. Because it has been shown that a number of scaffolding molecules play a role in regulating the activity of Ras in different developmental models, we decided to test whether one of these molecules, KSR, which is normally expressed in all thymic populations (data not shown), is involved in the regulation of Ras MAPK activity during positive selection.

rFTOC of retrovirally transduced precursors provides a kinetically sensitive way to test this hypothesis. The use of fetal thymocytes from embryonic stages E14.5 to E16.5 allows transduction of precursors during the proliferative phase post β-selection (double-negative to DP transition), so that the gene of interest is expressed in DP thymocytes (37). Furthermore, the use of donor thymocytes derived from nonselecting background matings, MHC°, minimizes the background of cells positively selected before the transgenes can be expressed in the rFTOC. The infected thymocytes are immediately reaggregated with stromal cells derived from deoxyguanosine-treated E15.5 fetal thymic lobes, and cultured in rFTOC for up to 12 days to follow the kinetics of thymocyte differentiation from the DP to the CD4 and CD8 SP stages.

To investigate the role of KSR during thymocyte development, these rFTOC cultures were analyzed for TCRβ and GFP expression (Fig. 1,A), and the TCRβ^high populations were subsequently analyzed for CD4 and CD8 expression (Fig. 1,B). Two to five individual lobes were analyzed per time point in each experiment, and the results of individual lobes for two representative experiments are shown in Fig. 1,C, in which we plot for each lobe, the ratio of the total number of TCRβ^high cells to the number of TCRβ^low-mid DP cells for both the GFP⁻ and GFP⁺ populations. In each experiment in this system, the GFP⁻ cells serve as an additional internal control for the GFP⁺-transduced cells within the same lobe. Fig. 1,D shows a comparison of the generation of TCRβ^high cells in GFP⁺ vs GFP⁻ cells for thymocytes infected with KSR or vector alone (Mig) for five different experiments. Infection with the KSR retrovirus results in levels of KSR expression that correlates linearly with the intensity of the GFP staining (data not shown), and, in fetal thymocytes, results in at least 3-fold greater than that detected endogenously in average (data not shown). Altogether, these results show that overexpression of KSR results in a reduction in the number of mature αβ T cells in this rFTOC system. In contrast, we could not detect any significant skewing in the generation of CD4 vs CD8 TCRβ^high cells in this system (Fig. 1,B). The blockade in positive selection seems to occur at early stages in the process, because both TCRβ^high HSA^high and TCRβ^high HSA^low populations are affected (Fig. 1 E).

To test whether this decrease was due to an actual block in positive selection, or just to a delay in the appearance of TCRβ^high cells in the KSR-infected thymocytes, we performed time course experiments. As shown in Fig. 2, the decrease in the generation of TCRβ^high cells was already seen at day 6 of culture and was maintained at later time points (day 12). This effect was also observed in similar experiments as early as day 3 of culture (data not shown).

As this phenotype is reminiscent of the block in positive selection obtained by the overexpression of dominant-negative forms of Ras and MEK in transgenic mice (2, 3, 35), we decided to directly compare the effect of overexpressing KSR and the dominant-negative (dMEK) used in the original studies (MEK_K97A) (35). E15.5 fetal thymocytes from C57BL/6 and MHC° mice were infected in parallel with dMEK, KSR, and Mig, reaggregated, and allowed to develop for 9 days. Two representative rFTOC experiments are shown in Fig. 3. As shown in Fig. 3,A, infection with both dMEK and KSR results in at least a 2-fold reduction in the percentage of TCRβ^high cells. In addition, both dMEK and KSR overexpression result in a reduction in the percentage of SP cells and an increase in the percentage of DP cells (Fig. 3,B). Fig. 3 C depicts the graphical representation of all the analyzed lobes in these experiments and reveals that infection with dMEK or KSR results in the reduction of mature TCRβ^high cells. These results suggest that KSR functions in a manner similar to dMEK, blocking positive selection by inhibiting activation of the MAPK cascade.

To directly test whether KSR could affect activation of the MAPK cascade, the thymus-derived cell line 16610D9 (38) was infected with dMEK, KSR, and Mig. Twenty-four hours postinfection, cells were stimulated for 10 min with 6 nM PMA to activate ERK, and the levels of phosphorylated ERK were analyzed by intracellular staining (in Materials and Methods (33)). Stimulation with PMA under these conditions results in optimal ERK activation in T cells, as assessed by Western blot with phospho-specific Abs and by in vitro kinase assays (data not shown).

As shown in Fig. 3 C, PMA-induced activation of ERK is severely impaired in cells infected with KSR or dMEK when compared with both uninfected cells in the same culture, and with cells infected with the empty Mig vector. There is also a correlation between the expression levels of KSR or dMEK (as assessed indirectly by the GFP expression) and the inhibition in ERK activation. Therefore, these results demonstrate that overexpression of KSR uncouples Ras from the ERK cascade in thymocytes, and this results in a block in positive selection in rFTOCs.

The mechanism by which the overexpression of KSR blocks development and ERK activation most likely involves the formation of nonfunctional complexes with other proteins in the Ras/MAPK signaling cascade. To determine the nature of these complexes, a series of KSR structural mutants were generated (Fig. 4). In particular, our studies have focused on the membrane localization domain, elements important for the association of KSR with other molecules, and the kinase domain.

The CA3 domain is important for the membrane localization of KSR, and in some systems functions has the same effect as wild-type KSR (39, 40). The CRM mutant converts the cysteines at aa positions 359 and 362 (within the CRM/CA3 domain) to serines. This mutant blocks the ability of KSR overexpression to augment Ras signaling in Xenopus oocytes (25, 39). Sites important for the interaction of KSR with the molecules c-TAK, 14-3-3, and MEK were also mutated. KSR may require membrane localization to participate in growth factor receptor signaling (18), and this translocation is regulated by both c-TAK and 14-3-3 (28). Mutations of S392 and I397/V401 to alanines interfere with the c-TAK and 14-3-3 interactions, respectively, resulting in dominant-positive forms of KSR that constitutively translocate to the membrane (28). Another molecular interaction important for KSR activity is its constitutive association with MEK. A KSR loss of function mutation corresponding to the cysteine at aa position 809 was initially identified in C. elegans (16). Mutations at this site abrogate the ability of KSR to interact with MEK, and consequently, its ability to assist in Ras signaling (21).

Based on its homology to Raf kinase, KSR was initially proposed to function as a kinase (41). However, the murine form of KSR lacks the conserved lysine at position 589 that corresponds to the Mg²⁺-ATP-binding motif that is important for phosphotransfer (15). It has been proposed that mutations at this site block the kinase ability of KSR (41, 42, 43), but it has also been suggested that this site is important for mediating the interactions of KSR with other molecules, rather than kinase activity (18). To investigate the importance of this site during T cell development, the arginine at position 589 was mutated to methionine to generate the R589M mutant. A R589K mutant was also generated to determine whether the kinase function is important because this mutant contains the conserved lysine in the Mg²⁺-ATP binding region.

Once generated, all of the KSR mutants (see Fig. 4,A) were subcloned into Mig and tested for their effects on ERK activation using 16610D9 cells, as described above. Fig. 4 B shows that, with the exception of C809Y/Mig and CA3/Mig, overexpression of each of the mutants reproduces the inhibition of PMA-induced ERK activation mediated by KSR in wild-16610D9 cells. Overexpression of all mutants was confirmed by Western blot (data not shown).

Once screened for their effects in the ERK activation assay, S392A/Mig, C809Y/Mig, CA3/Mig, and the CRM/Mig mutants were tested for their effects on T cell development. These mutants were chosen so that representatives of both types of mutants (those that mimic the effects of KSR as well as those that do not) were analyzed. The effects of the mutants were analyzed between days 7 and 11 of culture by examining TCRβ, CD4, and CD8 levels in rFTOC assays, as described above. As seen previously, KSR overexpression results in a decrease in the percentage of TCRβ^high cells (Fig. 4, C and E). Among the KSR mutants tested, CRM/Mig (data not shown) and S392A/Mig had comparable effects to wild-type KSR. In contrast, the overexpression of CA3/Mig (data not shown) and C809Y/Mig did not alter the development of mature T cells in the rFTOC (Fig. 4, C and E). The CD4/CD8 profiles within the uninfected and infected TCRβ^high populations were also examined, demonstrating that as seen with TCRβ expression, the S392A/Mig and CRM/Mig (data not shown) mutants behave similarly to KSR, while only the CA3/Mig (data not shown) and C809Y/Mig mutants do not behave like KSR (Fig. 4 D). This analysis suggests that of the KSR domains analyzed, only the MEK interaction domain is required to block T cell development in rFTOCs, while the membrane localization and the kinase domains are dispensable for this effect. Furthermore, the isolated expression of the membrane localization domain (CA3) does not affect this process. These results also demonstrate that the ability of the mutants to interfere with ERK activation correlates with the ability to block T cell development.

Ras signaling is central in the regulation of cell growth, survival, and differentiation. Ras controls these processes by activating numerous downstream effector pathways, including the MAPK cascade (44, 45). Although this cascade is the best-defined downstream Ras effector pathway, abundant evidence shows that it is insufficient to recapitulate Ras function in numerous processes, including mammalian cell transformation, T cell activation, and development (7, 36, 46). The mechanisms that regulate which Ras effectors are recruited are unclear, but the isolation by genetic screens of a number of proteins that function as scaffolds suggests that they may be integral for the effective activation of specific signaling cascades (47).

Scaffolds coordinate an appropriate response by interacting with other proteins and organizing them into a multimolecular complex. This association is important for fostering interactions with molecular partners as well as sequestering them from inhibitory molecules. To understand in more detail how Ras functions during positive selection of thymocytes, we have addressed the role of KSR in this process. KSR was identified as a modifier of Ras signaling in genetic screens performed in D. melanogaster, and is physiologically expressed in the thymus. Our results show that overexpression of KSR disrupts T cell development, and that this effect correlates with its ability to alter MAPK activation downstream of Ras.

In invertebrates, a number of putative scaffolding proteins capable of affecting Ras signaling have been identified, some of which do not generate obvious phenotypes when mutated. This may be due to their role as fine-tuning components of their respective pathways rather than as primary signaling molecules. In accordance with this observation, KSR mutations do not produce strong phenotypes in flies, and a knockout of KSR does not generate an obvious phenotype in developing murine thymocytes (48). However, upon closer examination, it is clear that KSR does affect signaling in thymocytes, because ERK signaling is compromised in mutant animals (48). The lack of a developmental phenotype in the KSR knockout transgenic animals is likely to be due to the expression of a second KSR gene in the thymus. This hypothesis is supported by the identification of a second KSR gene in C. elegans, both of which have partially redundant functions (49).

Therefore, to further understand the role of scaffolding molecules in thymocytes, approaches other than gene disruption may be informative. For example, in Drosophila, sensitized backgrounds in which the levels of the molecules of interest are below a critical threshold provide a means to study molecules with more subtle effects. As scaffolding molecules coordinate multimolecular complexes, another approach to investigating their function entails overexpression studies. For example, the overexpression of wild-type connector enhancer KSR in the Drosophila eye imaginal disc has a mild dominant-negative effect on Ras signaling, most likely due to the titration of connector enhancer KSR-interacting proteins that are necessary for Ras signal transduction (50).

To study the role of KSR in regulating Ras signaling during positive selection, we chose an ex vivo approach using rFTOC. This system reproduces in vivo fetal T cell development (51) and makes it possible to introduce genes in the developing T lymphocytes using retroviral-mediated gene transfer. This system is useful for studying the function of scaffolding molecules because these molecules can be significantly overexpressed as compared with endogenous protein levels. Furthermore, this method also allows for the manipulation of a large number of multiple constructs, a facet that is prohibitively time consuming by traditional transgenic approaches.

During fetal development, overexpression of KSR has an effect upon T cell development, resulting in a block in positive selection. This phenotype is similar to that observed in adult thymocytes following the overexpression of a dominant-negative form of Ras or of catalytically inactive MEK (dMEK) (2, 3), and to the effect of overexpressing the dMEK construct in rFTOCs. Despite the effect on the generation of single-positive thymocytes, we did not observe any alteration in the CD4/CD8 ratios, although this system is good at detecting such differences (37). Although it has been suggested that the intensity of MAPK signals can contribute to this lineage commitment decision, these results, as well as our previous experiments in transgenic mice, argue against this possibility (reviewed in Refs. 36 , 52 , and 53).

The developmental effects observed by overexpressing KSR correlate well with its biochemical effects: overexpression of KSR blocks ERK activation as effectively as overexpression of dMEK. Interestingly, there is a difference between these two constructs. Although the effect of dMEK seems to be linearly related to the amount of transgene expressed (which roughly corresponds with the amount of GFP expressed by the infected cells), KSR blocks the cascade in an all-or-none fashion. Cells with low expression seem to activate ERK normally, but after a certain threshold level, the cascade is blocked completely. This suggests that the effect of KSR is due to the titering away of a limiting component in the signaling cascade.

To investigate the structural elements of KSR important for this effect, a series of mutations within KSR were generated. Mutations of the membrane localization motif, the kinase domain, the c-TAK, or 14-3-3 interaction domains do not reverse the effects of KSR overexpression in our assays, suggesting that these domains are not critical for the overexpression phenotype. In addition, generating a predicted kinase active form of KSR has no effect in our assays. However, mutations predicted to disrupt the interaction between KSR and MEK render the construct incapable of blocking thymocyte development. Thus, in our system, the effect of KSR appears to be due to the sequestration of MEK. We propose that in thymocytes, the association between KSR and MEK occurs within the cytoplasm before KSR membrane translocation because membrane localization of KSR is not required for its inhibition of Ras signaling.

We thank L. Van Parijs, M. Therrien, M. Han, and S. Hedrick for providing reagents, as well as members of the Alberola-Ila, Baltimore, and Rothenberg laboratories for helpful discussion and technical assistance. We are grateful to T. Brummel, G. Hernandez-Hoyos, D. Leopoldt, J. Pomerantz, and P. Sternberg for critical comments regarding the manuscript.