Ras signals principally via Erk in G1 but cooperates with PI3K/Akt for Cyclin D induction and S-phase entry

ABSTRACT

Numerous studies exploring oncogenic Ras or manipulating physiological Ras signalling have established an irrefutable role for Ras as driver of cell cycle progression. Despite this wealth of information the precise signalling timeline and effectors engaged by Ras, particularly during G1, remain obscure as approaches for Ras inhibition are slow-acting and ill-suited for charting discrete Ras signalling episodes along the cell cycle. We have developed an approach based on the inducible recruitment of a Ras-GAP that enforces endogenous Ras inhibition within minutes. Applying this strategy to inhibit Ras stepwise in synchronous cell populations revealed that Ras signaling was required well into G1 for Cyclin D induction, pocket protein phosphorylation and S-phase entry, irrespective of whether cells emerged from quiescence or G2/M. Unexpectedly, Erk, and not PI3K/Akt or Ral was activated by Ras at mid-G1, albeit PI3K/Akt signalling was a necessary companion of Ras/Erk for sustaining cyclin-D levels and G1/S transition. Our findings chart mitogenic signaling by endogenous Ras during G1 and identify limited effector engagement restricted to Raf/MEK/Erk as a cogent distinction from oncogenic Ras signalling.

Introduction

35 years of intense research have established an essential role of the GTPase Ras (refers collectively to H-Ras, K-Ras4A, K-Ras4B and N-Ras) in the mitotic program of virtually every nucleated cell. This pivotal role manifests in the high prevalence of Ras mutations in cancer and the strong transforming potential of oncogenic Ras mutants such as RasG12V in preclinical models of cancer. A plethora of studies has categorically shown that oncogenic Ras triggers cell cycle entry and DNA synthesis independently of growth factors [1,2]. Accordingly, inhibition of native Ras by a variety of approaches including microinjection of inhibitory Abs [3], dominant negative RasS17N [4], expression of Ras-GTP scavenging modules [5], small-molecular allosteric inhibitors of Ras function [6] or by preventing upstream steps in Ras activation [7] inhibits DNA synthesis and cell proliferation. More recently, the analysis of engineered MEF lines devoid of all 3 Ras genes (termed Rasless) has underscored the indispensable nature of Ras in cell cycle progression, as cells rendered Rasless were forced into a quiescence-like state and ceased to proliferate [8,9].

The functions of Ras in governing cell-cycle entry and progression are presumably manifold. Ras signalling is key for cells to escape from quiescence and commence the cell cycle in the first place, a function that appears to be conserved from yeast, where Ras signalling is essential for spore germination [10]. In addition, Ras is probably important at other stages of the cell cycle, perhaps best illustrated by the observation that cycling cells stop dividing and accumulate in G0/G1 upon induction of the Rasless state or following RasS17N expression [8,11]. Studies on the cell cycle phase transition governance by Ras are scarce, however, and have centered on a possible function of Ras along G1 for progression into S-phase. In a pioneering series of experiments Hitomi and Stacey reported a requirement for Ras at mid G1 for transition to S-phase by tracing the fate of asynchronously growing cells microinjected with the Ras-neutralizing Ab Y13-259 [3,12]. These and other studies using temperature-sensitive Ras mutants or the controlled expression of dominant-negative RasS17N suggested a critical function of Ras during G1 progression [11,13]. Importantly, elevated Ras-GTP levels are present at mid G1 in fibroblasts and epithelial cell lines surging from quiescence [14–17], a finding that is both a premise and consistent with the notion that Ras relays mitogenic signals in G1.

Numerous studies have attributed the induction of type-D cyclins, a class of key mediators of the G1/S transition and progression through G1 into S to the function of Ras [12,14,18,19]. According to most reports mitogen-induced expression of cyclin D1 requires activation of Erk (collectively for Erk1 and Erk2) downstream of Ras [20–22]. The duration of Erk signaling appears to play a definitive, yet poorly understood role as fibroblasts require sustained Erk activity in order to efficiently induce cyclin D and enter S phase [23,24], while in PC12 and other cells transient Erk activation was sufficient to promote G1/S transition [25–27]. It is however noteworthy that Erk may not always be required for cyclin D production as serum-dependent cyclin D induction is only marginally affected by MEK inhibition in IIC9 and NIH3T3 fibroblasts [14,28]. In line with this view, oncogenic Ras mutants defective in coupling to Raf can drive cell proliferation in the absence of ostensible Erk activation [29,30], suggesting that Erk signalling may be conditionally dispensable for Ras-driven cell cycle progression. Indeed, given the intricate role of Erk in early Ras signaling as cells exit quiescence and enter G1, it can be arduous to catalogue discrete Erk signaling events in G1. In this respect it was emphasized before that Erk activity decays along G1 and exhibits poor correlation with Ras-GTP levels at later stages of G1 [14,15]. These findings among others have fostered the concept that Ras-effectors other than the Raf/MEK/Erk module could mediate, or at least significantly contribute to cyclin D induction and other mitogenic Ras-signals throughout G1.

A second Ras-effector route that has been linked to the mitogen-dependent induction of cyclin D is the PI3K/Akt pathway. PI3K/Akt signaling elevates cyclin D levels by various mechanisms, i.e. via transcriptional regulation [14], post-transcriptionally via elevated translation of cyclin D mRNA [31] and post-translationally by promoting cyclin D stability through GSK3ß [32]. PI3K signaling also affects cyclin D activity through the mediation of Rac [33] or by repressing the cyclin dependent kinase (CDK) inhibitor p27 via phosphorylation of Forkhead transcription factors [34]. Incidentally, PI3K lipid product levels and Akt activity are both elevated at mid-G1 [14,17,35,36], which in principle is consistent with PI3K/Akt acting downstream of Ras to govern cyclin D1 levels and S-phase entry. However, it is difficult to judge if PI3K/Akt signalling at G1 proceeds downstream or in parallel to Ras based on available data because most findings are correlative or involved pharmacological inhibition of PI3Ks and engineered expression of active Ras, PI3K or Akt mutants.

In addition to Erk and PI3K/Akt, a third established Ras effector pathway represented by guanine nucleotide exchange factors for Ral GTPases (RalGEFs, refers collectively to all four mammalian Ras-responsive exchange factors for Ral) also regulates cyclin D expression at the transcriptional level [14], possibly mediated by NF-κB [37]. Thus, while it emerges that at least three Ras effector programs can potentially control cyclin D levels, it is not known if and at which stage along G1 Ras engages each of them to orchestrate cyclin D turnover and S-phase entry. At this point it should be noted that there is also evidence suggesting that cyclin D may not account for all mitogenic signaling downstream of Ras. For example, cyclin D can synergize with Ras in cell transformation under some conditions [38,39]. In fact, the debate about the extent to which cyclin D levels depend on Ras has gained momentum recently owing to the observation that growth-arrested Rasless cells harbored normal levels of cyclin D despite the complete lack of Ras expression [8].

Summarizing the above, Ras plays an undisputed paramount role in commitment to proliferation, initiation of cell cycle and G1/S transition but its relevance to cyclin D induction/stabilization and the precise signaling period(s) and effector(s) engaged by Ras along G1 are obscure. As stated above, this gap in knowledge can be attributed in part to limitations in the experimental approaches to probe and manipulate Ras function. Most work addressing cell-cycle-dependent role(s) of Ras entailed the use of oncogenic RasG12V or the microinjection of the neutralizing Ras-Ab Y13-259, a cumbersome procedure that does not readily permit biochemical analysis of downstream signaling [3,12,40,41]. Moreover, as noted previously [42], data interpretation is hampered by the fact that Y13-259 neutralizes not just K-Ras, H-Ras and N-Ras but cross-reacts with the related R-Ras-class members R-Ras, TC21 and M-Ras [43,44]. Similarly, the specificity of the dominant-negative Ras version RasS17N, probably the most widely used tool for Ras inhibition, may have also been overrated. RasS17N inhibits Ras signaling by occupying GEFs such as Sos, combined with the inability of RasS17N-GTP to activate effectors [45,46]. Since Sos regulates other GTPases and signaling mediators in addition to Ras, RasS17N could exert inadvertent Ras-independent effects. Moreover, RasS17N does not block Ras activation under all circumstances, as illustrated by the case of phorbol ester/PKC-dependent Ras activation in fibroblast and epithelial cells [47–49].

This said, perhaps the most critical shortcoming of prevalent approaches is the inordinate time lag required to exert Ras inhibition, which lies in the range of hours to days and thus often exceeds the duration of the signaling processes under investigation. Ideally, addressing cell cycle-dependent Ras signalling requires a means to instantaneously inhibit Ras at will at any given stage of the cell cycle. We report here the development and implementation of an experimental strategy enabling acute inhibition of Ras combined with biochemical analysis of downstream signalling to delineate Ras signalling in G1. Our findings demonstrate that Ras signaling is required deep into G1 for full-blown cyclin D induction and progression into S phase. Importantly, using this approach we have, for the first time, directly interrogated the engagement of the three major Ras effector pathways during G1.

Results

A heterodimerization strategy for the inducible rapid membrane translocation of a Ras-Gap

We have edited a commercially available, genetically encoded system of inducible protein-protein heterodimerization for its use as an experimental tool for acute Ras inhibition. The strategy is based on the heterodimerization between engineered variants of FK506 binding protein (FKBP) and the binding domain for the FKBP-rapamycine complex on mTOR (FRB) [50]. Heterodimerization is triggered by a cell-permeable, inert rapamycine analog termed iDim and proceeds reportedly within minutes of iDim administration [51]. To enable acute inhibition of Ras, FKBP was cloned in fusion with the C-terminus of K-Ras, composed of the hypervariable region (hvr) and CAAX-box, two motifs responsible for the plasma membrane (PM) anchoring of K-Ras4B [52]. The second dimerization partner, FRB, was appended to the catalytic domain of the Ras-GAP neurofibromin, a Ras-inactivator and gene product of the tumor suppressor NF1 (NF1_cat from here on). We reasoned that inducing heterodimerization with iDim should relocate NF1_cat to the PM, thus raising the concentration of NF1_cat in the vicinity of Ras and consequently promoting Ras inactivation (Figure 1(a)).

Figure 1.

Principle of the heterodimerization strategy for acute Ras inhibition.

(a) Schematic representation of the experimental approach based on the induced recruitment of NF1_cat to the vicinity of Ras at the PM. Major modular components are highlighted. See main text for abbreviations. (b) Modular architecture and numbering of all heterodimerization partner units. #6*denominates a NF1construct containing the inactivating point mutation R1276P. (c) Western Blot detection of heterodimerization constructs expressed in HeLa cells. Immunodetection involved the use of antibodies against FKBP, Cherry or GFP, as indicated. #4* represents a cloning intermediate lacking all those three antigens that was loaded as a negative control for immune detection. Molecular size markers are shown on the left side. (d) Inducing heterodimerization of constructs #1 and #6 with iDim in HeLa cells promotes the rapid recruitment of #6 to the PM. Time scale on top of the panel relates to the time point of iDim addition. Magnification bar: 10 μm.

Previous studies have highlighted the importance of steric considerations in the design of the heterodimerization partners for optimal dimerization and biological functioning of the client proteins [53,54]. To account for possible steric influences we generated a panel of fusion constructs for both dimerization partners giving rise to eight dimer combinations (discounting construct #6*, which was envisaged as a specificity control, see below) (Figure 1(b)). Permutations involved the sequence order of dimerization and functional modules as well as the number of the dimerization units in individual constructs. To enable visualization of all heterodimerization partners, constructs #1 to #4 were tagged with mCherry and constructs #5 and #6 with EGFP (see Figure 1(b)). Transfection into HeLa cells illustrated that all encoded fusion constructs gave rise to polypeptides of the predicted size, albeit reaching varying levels of expression (Figure 1(c)). Confocal imaging of transfected HeLa cells showed that constructs #1 to #4, containing the PM anchoring module, all localized predominantly to the PM (shown representatively for #1 in Figure 1(d)). Accordingly, constructs #5 and #6 exhibited a non-particulate, diffuse cytoplasmic distribution, consistent with the absence of obvious membrane-targeting motifs. Addition of the heterodimerizer iDim induced the rapid relocation of the NF1_cat containing fusion proteins to the PM for the pairings #1#6, #3#6 and #4#6 while no dimerization was observed for all other combinations, even after extended incubation periods with iDim of up to 2 h. Translocation to the PM was most pronounced for the pair #1#6, with translocation being completed 15 min after iDim addition (Figure 1(d)). A similar expeditious and pronounced iDim-dependent redistribution of construct #6 was observed in other cell types (Supplementary Fig S1). We concluded that the inducible heterodimerization approach promoted translocation of the NF1_cat containing polypeptides to the PM within minutes in the context of distinct dimer combinations.

Induced heterodimerization causes Ras inhibition

To understand if heterodimerization-dependent accumulation of NF1_cat at the PM promoted Ras inactivation we generated HeLa cell populations with stable expression of #1 and #6 or #4 and #6 (termed HeLa#1#6 and Hela#4#6, respectively) by sequential antibiotic selection and fluorescence assisted cell sorting. Over 95% of cells in the obtained cell populations featured a uniform fluorescence signal for the constructs #1, #4 and #6. Stimulation of HeLa#1#6 cells with EGF prompted rapid Ras activation with kinetics and intensity indistinguishable from parental HeLa cells (Figure 2(a)), indicating that Ras activation was not markedly affected by the mere presence of the NF1_cat polypeptide or by the selection procedure. Addition of iDim 30 min prior to EGF strongly impaired Ras-GTP accumulation (Figure 2(a)), consistent with the working hypothesis shown in Figure 1(a). Heterodimerization prevented activation of all three Ras proteins, as evidenced by the fact that the Ras-antibody used in Figure 2(a) decorates all three isoforms and as confirmed using isoform specific antibodies (Supplementary Fig S2). Ras inhibition was specific because dimerization did not affect EGF-dependent activation of the NF1-insensitive Ras-relative Rap, and it was selective as it did not alter the activity levels of TC21/R-Ras2, a transforming Ras-related GTPase that is unresponsive to growth factors but sensitive to NF1 [55] (Figure 2(a)). Importantly, heterodimerization did not interfere with upstream steps such as autophosphorylation of the EGFR but abrogated Erk and Akt phosphorylation/activation in response to EGF, showing that blocking Ras activation reflected on its downstream effectors Raf and PI3K. While monitoring the activity of Ral-GTPases, the third established Ras effector pathway, we observed that RalA was not activated by EGF in HeLa cells (Figure 2(a)). Notwithstanding the lack of activation by EGF, the fact that Ral-GTP levels were unaffected by iDim-induced heterodimerization represented further incidental proof of specificity. Interestingly, heterodimerization of the alternative pair #4#6 entailed only marginal Ras and Erk inhibition (Supplementary Fig S3), illustrating that efficient recruitment of NF1_cat to the PM was not sufficient to promote Ras inactivation. We presume that this discrepancy reflected an unproductive steric orientation of NF1_cat at the PM in the context of the #4#6 dimer.

Figure 2.

Heterodimerization of #1 and #6 inhibits Ras activation and Ras-dependent signaling.

(a) Promoting heterodimerization with iDim impairs EGF-induced Ras activation. HeLa or HeLa#1#6 cells were deprived of serum, treated or not for 30 min with 500 nM iDim, as indicated, prior to stimulation with 10 ng/ml EGF for the indicated periods of time. The activation status of GTPases and other signaling proteins was determined as described in the experimental section. TC21-GTP levels were determined from RBD-pulldowns without prior stripping of the Ras-GTP signal in order to highlight the lower mobility of TC21 with respect to Ras. The panel marked RBD shows a Coomassie Blue stain of the GST-RafRBD domain used to collect Ras-GTP and TC21-GTP. (b) Expression control of constructs #1 and #6* in parental and stable HeLa cell lines probed with anti-FRB and anti-Cherry antibodies. (c) HeLa#1#6* cells, expressing the catalytically dead NF1_cat unit, were treated or not with 500 nM iDim for 30 min and challenged with EGF for the indicated lengths of time. Cell extracts were processed for Ras-GTP and Ras effector activity determination as before. (d) HeLa#1#6 cells were treated for 30 min with varying concentrations of iDim prior to EGF administration and assessment of Ras and Erk activation. (e) Same as panel D except cells were treated for varying lengths of time with 500 nM iDim. (f) HeLa#1#6 cells were treated and processed as before except that 10% FCS was used for stimulation. (g) Same experiment as in panel F except cells that cells were challenged with 5 μg/ml Insulin Note that Insulin-driven PI3K/Akt activation is much less sensitive to Ras inhibition than the one in response to EGF.

Extending the test of specificity of our approach, we confirmed that iDim had no effect on Ras, Erk or Akt activation in parental HeLa cells (unpublished observations). For a more stringent control we generated HeLa cells co-expressing construct #1 and a mutant construct #6* (Figures 1(b) and 2(b)) coding for NF1_catR1276P, a GAP variant with a mutation of the catalytic arginine R1267, which lacks enzymatic GAP activity while retaining its high affinity for Ras-GTP [56]. Translocation of NF1_catR1276P to the PM in response to iDim was undistinguishable from that of the wild-type construct #6 (Supplementary Fig S4). Given the absence of enzymatic GAP activity, we postulated that NF1_catR1276P should behave as an inert Ras-GTP binding module such as the Ras binding domain of Raf1 (RBD). It was previously demonstrated and mathematically rationalized that RBD enhances Ras-GTP accumulation by protecting Ras-GTP from Ras-GAPs while at the same time it attenuates downstream Ras-signaling by competing with effectors for Ras-GTP [47,57,58]. As shown in Figure 2(c)), heterodimerization of NF1_catR1276P did not only fail to attenuate Ras activation in response to EGF but instead prolonged Ras-GTP accumulation and lowered Erk activation induced by EGF. This response pattern was reminiscent of the effects caused by RBD and thus exactly as predicted for an inert Ras-GTP binding module. From all this we concluded that Ras inhibition following heterodimerization of #1#6 was specific and contingent on the presence of a catalytically competent NF1_cat module.

To further characterize this system we analyzed the dose-dependent response to iDim. Increasing doses of iDim gradually reduced the activation of Ras and its downstream target Erk by EGF, with full inhibition being achieved at 500 nM iDim (Figure 2(d)). This dose-response curve followed closely the dose-dependent iDim-induced translocation of construct #6 to the PM (Supplementary Fig. S5) and was in good accord with previous studies pursuing the same heterodimerization strategy [50]. The inhibitory effect of 500 nM iDim on the activation of Ras/Erk endured for up to 16 h (Figure 2(e)). We also tested if inhibition could be reversed by the washout of iDim but found that heterodimerization could not be reversed even after extended and repeated rounds of washing with full culture medium (data not shown). Finally, since many of the experiments shown below involved the release of quiescent cells into the cell-cycle by serum, we tested if iDim abrogated Ras signaling induced by fetal calf serum (FCS). As shown in Figure 2(f) FCS induced activation of Ras and Erk, albeit weaker than EGF, and prior treatment with iDim reversed the effect. The same result was also obtained for insulin stimulation (Figure 2(g)), showing that the heterodimerization system prevented Ras activation in response to variable upstream stimuli.

Ras inhibition prevents progression to S-phase

Ras is vital for cell proliferation and more precisely, for progression into S-phase of cells re-entering the cell cycle in response to mitogens. These functions of Ras cannot be studied in HeLa cells because they lack a functional G1/S checkpoint control due to the loss of p53 and pocket protein function. To perform this analysis we resorted to T98G cells, a human cell line that depends on growth factors for proliferation and features an intact G1-checkpoint response/arrest to serum withdrawal or contact inhibition [59,60]. We generated T98G cell populations and clonal lines with stable expression of #1 and #6 as before and validated the functionality of the heterodimerization system. As shown in Figure 3(a) administration of iDim did not affect EGF-induced Ras activation in parental T98G cells but markedly reduced accumulation of Ras-GTP in polyclonal and clonal T98G#1#6 populations, inhibition being more pronounced in the clonal derivative. Heterodimerization also induced a corresponding block of Erk activation in both the mixed population and the clonal derivative as assessed by western blotting (Figure 3(a)) or flow cytometry (Supplementary Fig S6). As observed for Ras-GTP, inhibition at the level of Erk was more robust in the clonal T98G#1#6 cell line. Interestingly, Akt phosphorylation was only modestly affected, indicating that EGF activated PI3K via Ras-independent mechanisms in quiescent T98G cells. All complementary quality control experiments, as performed with HeLa#1#6 cells above, corroborated the specificity of the approach in T98G#1#6 cells.

Figure 3.

Acute Ras inhibition precludes G1/S progression.

(a) Parental T98G cells, a polyclonal pool or a clonal derivative of T98G cells with stable expression of #1 and #6 were treated 30 min with 500 nM iDim as indicated, challenged with EGF and processed for the assessment of Ras, Erk and Akt activity levels. (b) Combined flow cytometry-based analysis of cell cycle distribution and DNA synthesis as a marker of S-phase cells. Cartoon shows a typical representation of the flow cytometry data and the gates for the distinct cell cycle phases, including the characteristic “arc” of S-phase cells. (c) T98G cells were rendered quiescent by serum deprivation for a period of 3 or 6 days as indicated and released back into the cell cycle by serum refeeding. DNA synthesis was monitored 0, 12 or 20 h after release by flow cytometry. Gates delineate S-phase `arcs´ encompassing cells in S-phase. S-phase events ± SD (n = 3) are shown on top of the gates. A cell-cycle cartoon depicting the flow-line of the experiment is shown in this and subsequent panels for the sake of clarity. Async. cells: asynchronously growing cells. (d) Ras blockade impedes progression to S-phase. Polyclonal (Pool#1#6) or 3 different clonal T89G#1#6 cell populations were rendered quiescent by 3d of serum deprivation. Cells were treated or not treated, as indicated, with 500 nM iDim for 30 min before being released into the cell cycle by serum refeeding. Gates delineate cells in S-phase. S-phase events ± SD for pooled #1#6 cells, G4 (both n = 3), E9 and G10 (both n = 1) are indicated on top of the gates. Async.: asynchronously growing cells. (e) Quantification of the 12 h data points from panel D.

To test if the block in Ras activation affected cell cycle progression, we analysed progression to S phase of quiescent T98G#1#6 cells impelled into the cell cycle by serum. For the identification and quantification of S-phase cells we monitored DNA synthesis by pulsing cells with a nucleotide analogue (EdU) and monitoring its uptake via flow cytometry (Figure 3(b)). We first ascertained if serum deprivation rendered T98G cells quiescent. Serum starvation for 3 days or longer resulted in an essentially complete halt of DNA synthesis and loss of Ki-67 as markers of proliferation and cell cycle activity (Supplementary Fig S7A and Supplementary Fig S7B, respectively), confirming that T98G cells readily reached stable G0 [60,61]. Upon serum refeeding around 80% of cells reentered the cell cycle, as judged by Ki-67 reappearance (Supplementary Fig S7B), and 60–70% of all cells progressed to S-phase as determined by DNA synthesis 12 h (early S-phase) or 20 h (cumulative S-phase) after release from quiescence (Figure 3(c)). Addition of iDim 30 min prior to release did not affect progression to S-phase of wild type T98G cells (Supplementary Fig S8) but induced a major block in S-phase entry in T98G#1#6 (Figure 3(d), quantification in Figure 3(e)). In line with the observations made at the level of proximal EGFR signalling, clonal T98G#1#6 lines exhibited a more pronounced block in S-phase entry than the polyclonal cell population. We attributed the more pronounced response of the clones to the heterogeneous and ephemeral expression of construct #6 in the mixed T98G#1#6 population, probably owing to clonal out-selection processes because the expression of construct #6 progressively declined with exceeding cell passages (Supplementary Fig S9).

Ras signaling at mid G1 drives G1-to-S progression in cells emerging from quiescence

As documented in Figure 3(d), Ras activity is crucial for quiescent cells to reenter the cell cycle and to progress to S-phase through the restriction point in G1. It has been postulated that G1/S transition is driven by a second, discrete wave of Ras signaling in mid-G1 [62]. To address this possibility, quiescent T89G#1#6 cells were released into the cell cycle by serum addition in the presence or absence of iDim. To discriminate early from late intervals of Ras signaling iDim was added either before or 1 h after FCS refeeding, i.e. before or after the first wave of Ras signaling. As shown above, inducing heterodimerization with iDim 30 min before release from quiescence strongly suppressed progression to S-phase (Figure 4(a)). This effect was mimicked by MEK inhibition with U0126 and was even stronger in cells treated with the PI3K inhibitor LY294002, indicating that both pathways qualified as potential mediators of the mitogenic Ras signal. Notably, an undistinguishable block in progression to S-phase was observed in cells that received iDim 1 h after serum, which showed that Ras signalling at a time point later than the first Ras signalling interval in G0 was essential for G1 progression and S-phase entry (Figure 4(a), quantification in Figure 4(b)). Western blot analysis of T89G#1#6 cells treated identically confirmed that iDim-induced heterodimerization had repressed Ras and Erk activation at the experimentally preordained time points (Figure 4(c)). In particular, administration of iDim 1 h later than serum spared the first Ras signaling wave, corroborating that Ras signaling later in G1 was necessary for G1/S transition. Of note, we observed that T98G did not feature a genuine, discrete peak of Ras activation during G1 as reported for other cell types [14,15] but rather Ras-GTP levels remained at a constantly low, yet detectable basal level that was sensitive to the heterodimerization system (Figure 4(c)).

Figure 4.

G1/S transition depends on Ras activity at mid G1.

(a) Ras activity is required beyond the first signalling wave in quiescent cells for progression to S-phase. Quiescent T98G#1#6 G4 cells were treated with 500 nM iDim, 50 μM UO126 or 10 μM LY294002 30 min before (−0.5 h) or 1 h after (1 h) being released into the cell cycle. DNA synthesis and cell cycle distribution were assessed 12 h and 20 h following release. (b) Quantification of the data from panel A. (c) Ras activation assay and Western blot analysis of samples from cells treated identically to the experiment shown in A. Short and long exposures of Ras and Erk western blots are shown for better clarity. (d) Quiescent T98G#1#6 G4 cells were treated with 500 nM iDim at the indicated time points relative to release (time point 0). DNA synthesis and cell cycle distribution were assessed at 0 h, 12 h and 20 h following release. (e) Quantification of data (n = 3) of 12 h data from panel D. ***p < 0.001.

In an attempt to narrow down the critical interval of Ras signalling in G1, we administered iDim every hour following release of the cells into the cell cycle (Figure 4(d)). This experiment showed that Ras signalling was essential for as long as 5–6 h after the initial wave of Ras signalling during the G0/G1 transition, as determined by sigmoidal curve fitting of the results (Figure 4(e)). Thus, in addition to the immediate early Ras signaling episode in G0, Ras signalling during early to mid G1 was necessary for cells to progress to S-phase.

Ras signals via Erk, but not PI3K or Ral, at mid G1 to sustain Cyclin D1 levels

In line with a wealth of literature, we found that inhibition of Ras or any of its two downstream effectors Erk and PI3K indistinctively abrogated G1/S phase transition (see Figure 4(a) above). One straightforward interpretation was that Ras engaged both pathways during G1 to propel cells into S-phase. Seeking to positively identify the effectors engaged by Ras in G1 for promoting G1/S progression, we monitored the response of the three major Ras effector pathways, Raf/Mek/Erk, PI3K/Akt and RalGEF/Ral, to iDim-induced Ras inhibition. As shown in Figure 5(a), activation of Akt and Ral in response to FCS was pronounced and prolonged throughout G1, but neither event was markedly affected by inducing Ras inhibition before or after FCS addition. In contrast, heterodimerization induced a marked drop in Erk activity, which was ostensible at any and all preordained time points of iDim addition. Furthermore, inhibiting Ras prior to serum addition caused a strong drop of cyclin D levels (Figure 5(a)), indicating that cyclin D induction and/or stabilization depended on Ras signalling. Accumulation of cyclin D was also prevented in T89G#1#6 cells treated with iDim 1 h or 2 h after release into the cell cycle (see Figure 5(b)), but this effect reversed and cyclin D levels were restored to normal levels if iDim administration was postponed to 6 hours after the FCS stimulation (Figure 5(a)). The same response pattern was observed for pocket protein hyper-phosphorylation, monitored here as the phosphorylation-dependent mobility shift of p130, also known as Rb2, a pocket protein that is particularly abundant at G0/G1 (Figure 5(a)). Taken together, these findings indicated that Ras did not engage the PI3K/Akt or Ral pathways during G1 and were most easily explained by postulating that Ras induced and/or stabilized cyclin D levels exclusively via mediation of Erk at mid G1.

Figure 5.

Ras signals via Erk but not PI3K or RalGEF at midG1.

(a) Quiescent T98G#1#6 cells were treated with 500 nM iDim at the indicated time points relative to release (time point 0). Cells were lysed and processed for Ras or Ral activity assay and western blotting as described. (b) Quiescent T98G#1#6 cells were treated with 500 nM iDim, 50 μM UO126 or 10 μM LY294002 2 h after release from quiescence or left untreated, as indicated. Reactions were quenched by lysing the cells at the indicated time points after release. Cells extracts processed for Ras activity assay and western blotting as before.

Pi3k/Akt provides a necessary, Ras-independent signal for Cyclin D stabilization and progression to S-phase

The findings above indicated that Ras promoted progression through G1and G1/S transition via Erk, independent of PI3K/Akt. To evaluate the significance of PI3K/Akt in cyclin D induction and G1-to-S progression, we used a pharmacological approach. Inhibition of PI3K or Akt 2 h after release lead to a loss of cyclin D (Figure 5(b)), showing that PI3K/Akt activity in G1 was necessary for cyclin D induction/stabilization. This effect was virtually indistinguishable from both Ras/Erk pathway inhibition with iDim as well as MEK1/2 inhibitor UO126 (Figure 5(b)), and was most easily explicable by postulating that parallel, independent signalling by Ras/Erk and PI3K/Akt operates in mid G1. Vice versa these findings also implied that activity of either Erk or PI3K/Akt alone was not sufficient to induce cyclin D and G1/S transition in T98G cells.

Ras signaling at early/mid G1 drives G1/S progression in continuously cycling cells

The data obtained for quiescent cells re-entering the cell cycle raised the question whether or not the same Ras-signalling processes at mid-G1 governed G1/S transition in continuously cycling cells, too. It has been previously reported that cycling cells depend on Ras signals in G2/M, not G1, for progression to S-phase [3,12], indicating that cyclin D and S-phase entry could be regulated differently by Ras in incipiently versus continuously cycling cells. We initially verified the general reliance of cycling cells on Ras activity by inhibiting Ras in asynchronous cells grown in serum-supplemented medium. Inhibiting Ras in cycling T98G#1#6 cells led to the accumulation of cells in G0/G1 at the expense of S-phase cells, a ratio that increased with longer incubation times with iDim for up to 72 h (Figure 6(a)), illustrating that Ras signaling was necessary for cycling cells to continuously proliferate. Of note, longer incubations times with iDim did not augment this effect, probably because non-responsive cells eventually started outgrowing the growth-arrested population.

Figure 6.

Transition to S-phase in cycling cells depends on Ras activity in G1.

(a) Asynchronously growing T98G#1#6 cells were treated with 500 nM iDim for 1–4 days and the cell cycle distribution was determined as before. Quantification is shown on the right (n = 3). C: control; D: dimerizer. (b) T98G#1#6 cells were synchronized at the G1/S border by a thymidine block and released back into the cell cycle in the presence or absence of iDim. The time point of iDim addition relative to the release is indicated on the right side of the panel. Cell cycle distribution and DNA synthesis were monitored 0, 3, 6, 12 and 20 h after release. Quantification is shown on the right (n = 3). (c) Asynchronously growing T98G#1#6 G4 cells were synchronized at G2/M by means of a double thymidine/RO-3306 block and released back into the cell cycle in the presence or absence of iDim. The time point of iDim addition relative to the release is indicated on the right side of the panel. Cell cycle distribution and DNA synthesis were monitored at the indicated time points. Quantification of data (n = 3) is shown on the right side of the panel. *p < 0.05, **p < 0.01, ***p < 0.001. (d) Analysis of Ras effector activity and cyclin D levels in cells treated exactly as in panel C.

In an effort to chart the critical, fate-determining interval of Ras signaling, cycling T98G#1#6 cells were synchronized at the G1/S border by a prolonged thymidine block and released back into the cell cycle in the presence or absence of iDim (Figure 6(b)). This experiment showed that blocking Ras as early as 20 h before the next round of DNA synthesis compromised G1/S phase transition causing an accumulation of cells in G1 at the expense of S-phase events. It is worth noting that while the extent of this G1/S block may appear low at first sight, it is of substantial biological relevance as it manifested as late as 20 h after synchronization in the preceding cell cycle round. Moreover, this experiment also documents that Ras inhibition at G1/S impinged earliest on the subsequent G1/S transition but did not prevent or diminish progression through the ongoing S and G2/M phases. We concluded that Ras activity was required specifically for G1 progression and onset of DNA synthesis but not for progression through S and G2/M in cycling cells.

That said, we wished to achieve synchrony of cycling cells closer to the initiation of G1-phase in order to minimize the time-dependent loss of synchrony inherent to these experiments and hence be able to chart Ras signaling in G1 more accurately. For this purpose we synchronized cells at the G2/M border via a double block of thymidine and the Cdk1 inhibitor RO-3306. RO-3306-arrested cells enter mitosis rapidly upon release from the block and progress promptly to a new round of the cell cycle in the absence of microtubule poisons commonly employed for this kind of experiments [63]. In this experimental setup thymidine/Ro-3306-arrested T98G#1#6 cells progressed to G1 in less than 2 hours after release and started entering S-phase 8 h post-release (Figure 6(c)). Inducing Ras inhibition 30 min in advance of the release strongly prevented progression to the next S-phase by more than 60% (Figure 6(c)). Importantly, starting Ras inhibition 2 h or 4 h after release from G2/M, time points at which 70% and more of cells had progressed to G1, resulted in an undistinguishable block in S-phase entry. Western Blot analysis of extracts from cells treated in the same way confirmed that iDim-induced heterodimerization had prevented Cyclin D accumulation at mid/late G1 (Figure 6(d)). These findings indicated that cycling cells required Ras signaling in G1 for progression to a new round of DNA synthesis (Figure 7).

Figure 7.

Model of cell-cycle-dependent Ras signaling.

Discussion

We have devised an approach for the acute inhibition of endogenous Ras based on inducible recruitment of the catalytic domain of neurofibromin to the vicinity of Ras at the PM. Ras inhibition proceeds within minutes of addition of the heterodimizing agent iDim. This celerity distinguishes this strategy from other available methodologies that require hours or days to exert effects and hence are ill suited to chart discrete Ras signaling episodes. Moreover, since the system is specifically anchored to the PM via the hvr of K-Ras (largely avoiding endomembranes) and given the observed virtually complete prevention of Ras-GTP accumulation, our findings strongly argue for the PM as the predominant and major subcellular platform of Ras activation and signaling, in line with previous reports [9,47,64].

The findings obtained using this approach in quiescent T98G cells reentering the cell cycle evidence that Ras signaling at mid G1 is necessary for efficient progression to S-phase, essentially confirming and extending previous data obtained with Ab microinjection or inducible RasS17N expression [3,4,12]. Based on our findings the period of mitogenic Ras signaling can be narrowed down more precisely than in previous studies to the first 6 hours of G1 in cells surging from quiescence. Furthermore, with the current strategy we were able to biochemically interrogate signal transmission processes downstream of Ras. For example, acute blockade of Ras within the critical 6 hour margin after serum stimulation fully ablated cyclin D induction, evidencing a strict dependency of cyclin D production/stabilization on Ras-signalling at mid-G1. This observation is in line with a major body of literature but would appear to contradict recent findings in Rasless cells, as those cells harboured high cyclin D levels despite the complete absence of Ras expression [8]. We hypothesize that this difference could result from the distinct approaches used to blunt Ras function. Given the acute nature of the Ras-block imposed with our approach, we assume that Ras signaling in G1 is truly necessary for inducing cyclin D in the first place, as suggested by the loss of cyclin D expression not just in response to Ras inhibition with iDim, but also to MEK1/2 inhibition 2 h after release into G1 (Figure 5(b)). This concept must not necessarily be at odds with the findings from Rasless cells, as in those cells cyclin D levels could remain high due to feedbacks or adaptive processes triggered by the long time span of Ras–inhibition and/or the only gradual disappearance of cellular Ras activity as residual Ras protein levels succumb only slowly to protein degradation. In this respect, we note that the critical 6 h interval of Ras signaling in G1 is only slightly in excess of the time point of cyclin D1 appearance. Indeed, inhibiting Ras 6 h after release, a time point at which cyclin D levels have reached their maximum, did not prevent or revert accumulation of cyclin D (Figure 5(a)) or S-phase entry (Figure 4(d)). Similar findings were obtained by others using a MEK-inhibitor washout protocol in MEF cells [65]. These observations indicate that early/mid G1 Ras signaling is necessary for the induction of cyclin D levels but at some point around 6 h into G1 cyclin D levels become independent of Ras, and by inference also of Erk signaling, a situation reminiscent of the scenario reigning in Rasless cells. Conceptually, therefore, the Rasless state could be considered to reflect the situation found in the second half of G1, at least as regards the Ras/cyclin D axis. In more general terms, the different experimental outcomes at the level of cyclin D observed here and in Rasless cells illustrate the importance of employing methods for acute Ras inhibition in order to exclude long-range adaptive processes when it comes to delineating proximal Ras signaling. Conversely, genetic manipulations of Ras can be better suited for disclosing long-term consequences and adaptation programs to changes in Ras signaling.

Our findings argue for the Raf/MEK/Erk cascade as the chief Ras effector pathway propagating the mitogenic Ras signal during G1, because Ras inhibition did not compromise stimulation of PI3K/Akt or RalGEF/Ral. This is not unprecedented as e.g. growth factor-induced Ral activation is known to depend on Ras in some cells such as NIH3T3-A14 cells [66], but not in others, notably MEFs [9]. The same considerations apply to PI3K (see below). Indeed, the current picture on the contribution of the various Ras effector pathways to the mitogenic action of Ras is blurry and has been complicated by variable experimental outcomes depending not just on the studied cell type, but also species (human versus rodent), experimental readout, Ras variant (wild type versus oncogenic) and others. Also, most studies involved the engineered overexpression or knockdown of Ras or Ras-effectors or pharmacological interventions downstream of Ras, approaches that can hardly dissect the Ras signaling topology along the cell cycle. A critical role for the Raf/MEK/Erk pathway in transmitting the mitogenic Ras signal is nonetheless widely accepted for human cells, although this notion has been challenged for rodent cells [14,28] or in the context of oncogenic Ras transformation [30]. Also, some authors have raised doubts as to the contribution of Erk to mitogenic Ras signaling during G1 based on the observation that Erk activity is low at mid-to-late G1 and does not correlate well with Ras-GTP levels [14,15]. On the other hand, the aforementioned studies in Rasless MEFs have documented that Raf/MEK/Erk was the only Ras effector pathway that rescued proliferation of Rasless cells. Concurrent with the latter notion, our data argue strongly in favour of Raf/MEK/Erk as the predominant, if not the only mediator of the mitogenic Ras signal in G1, although they do not help clarifying the species discrepancies as we did not analyze murine cells. In this respect, it is striking that our data highlight a prominent role for Ras-stimulated Erk activity at ≤ 6 h post-release, a time point in G1 when Erk activity is low and does not peak, in fact analogous to Ras activity. However, the low levels of Erk activity at 4 h and beyond in G1 are entirely dependent on Ras as they remain sensitive to heterodimerization-induced Ras inhibition all through G1 passage (see e.g. Fig 4(c) or 5(a)). These findings, i.e. the observation that low, basal levels of Erk activity do obviously mediate mitogenic Ras signaling at G1, throws new light on the discussion about the cell type-dependent differential need for transient versus prolonged Erk activation alluded to in the introduction. Based on our findings, we speculate that the true duration of mitogenic Ras/Erk signaling in quiescent cells released into the cell cycle may have been underestimated in previous studies that based their judgement on the absence or presence of a genuine activity peak. Thus, beyond proving the importance of Ras/Erk signaling at mid-G1 for S-phase entry, our findings emphasize the biological relevance of low, tonic activities for the regulation of biological processes as important as cell cycle progression.

The present data document that although PI3K and Ral do not transmit mitogenic signals downstream of Ras, PI3K/Akt signaling, at least, is essential for cyclin D induction and G1/S transition. This suggests that a mitogenic route towards PI3K/Akt bifurcates upstream of Ras, perhaps at one or more mitogenic receptors engaged by serum mitogens. Alternatively, despite the fact that PI3K/Akt activity obviously rises in respond to FCS, the Ras-independent mitogenic signal elicited by PI3K/Akt could be permissive in nature and originate downstream of constitutively active cues like e.g. integrins [67]. Along these lines, it is attractive to speculate that differences in the cell-intrinsic levels of PI3K activity could dictate whether or not activation of the Raf/MEK/Erk pathway is sufficient to induce cell proliferation (or transformation). This concept, which could at least in part explain why forced activation of the Erk-pathway induces cyclin D levels and S-phase entry in some but not in other settings, is currently under investigation in our laboratory.

Irrespective of the origin and nature of the PI3K/Akt signal, our findings exclude a role for PI3K downstream of Ras during G1/S progression. Intuitively, this notion appears unlikely as PI3K is a firmly established effector of Ras and a large body of literature has documented PI3K/Akt signalling in mitogenesis and oncogenic Ras-driven transformation [68–70]. Still, this concept is in line with other findings such as those obtained from transgenic animal models expressing PI3Ks with point mutations in their RBD that are defective in Ras binding. For instance, fibroblasts carrying a Ras-insensitive mutant of the ubiquitous, proto-oncogenic PI3K classIA catalytic subunit p110α (p110α* from here on) exhibit only a mild proliferative defect in serum [71]. Accordingly, homozygous p110α*animals are viable and born in close to Mendelian ratios with a few animals reaching adulthood. Although rescue effects by other PI3K isoforms cannot be excluded (albeit unlikely since p110β is activated by Rac, not Ras [72], and p110δ and p110γ are highly enriched in leukocytes [73]), these findings would indicate that most cell divisions from embryonic state to adulthood proceed unaffected by the inability of Ras to activate p110α. Similarly, transgenic flies harbouring a mutant Dp110 (the only PI3K gene in Drosophila) that does not interact with Ras-GTP, despite exhibiting growth retardation and small cell sizes, are viable and attain adulthood, again illustrating that the inability of Ras to activate PI3K is no impediment for the bulk of insect cell divisions [74]. Our findings are in line with this notion and provide direct evidence that wild-type Ras does not engage PI3K to drive cells through G1 and into S, rationalizing the findings from transgenic animals. As mentioned above, we speculate that the undoubtedly critical function of PI3K during G1/S progression could be permissive in function or, alternatively, PI3K is possibly engaged by mitogenic cues independently of Ras during G1/S progression. In this respect it is interesting to note that the increased lethality and morbidity of p110α* mice was attributed to a defective lymphatic vasculature, pointing to a cell type-specific role of the Ras/PI3K interaction in lymphatic vascular homeostasis and/or proliferation, possibly downstream of VEGFR type receptors [71,75]. It is indeed intriguing that VEGFR signalling via Ras differs remarkably and fundamentally from other related growth factor receptors as it involves PKC upstream of Ras and it is insensitive to the action of dominant-negative RasS17N under certain conditions [49,76].

Just as Ras-dependent Erk activation alone does not suffice for inducing cyclin D levels and progression through G1, the present data illustrate also that the combined serum-dependent activation of PI3K and Ral, which proceeds unaffected by Ras inhibition, is neither sufficient to drive cell proliferation. In sum, the Raf/MEK/Erk and PI3K/Akt modules contribute two independent and necessary signals for cyclin D induction and G1/S progression, of which Raf/MEK/Erk is the only one acting downstream of Ras. This said, investigating the participation of other ill-described Ras effectors was beyond the purview of this study and cannot be excluded. Indeed, the observation that Ras inhibition exerted a more pronounced block of S-phase entry than MEK inhibition (Figure 4(a)) is consistent with the possibility that Ras engages effectors other than Raf, PI3K or RalGEFs during G1.

An important conclusion of the present study is that Ras-signalling at mid-G1 is required and necessary for S-phase entry irrespective of whether cells emerge from quiescence or from a preceding cell division cycle. The latter would appear to contradict previous findings by Hitomi and Stacey, obtained by tracing asynchronously cycling cells microinjected with neutralizing Ras-Abs, which suggested that cyclin D and S-phase entry depended on Ras activity in the preceding G2 but not G1 [3,12]. We suspect that cell type specific variations or methodological differences could account for the different results. In this respect, it is important to stress the inevitable drawbacks of working with asynchronous or synchronous cell populations. For example, results obtained with synchronized cells, as documented here, can be confounded by side-effects of the employed synchronizers. Also, cells loose synchrony promptly once released from any synchronization block, even if working with homogenous cell pools or clones, effectively limiting the time window for experimentation. It was owing to these considerations that we used two complementary synchronization protocols (G1/S-border and G2/M) to address the role of Ras in cycling cells in order to minimize possible experimental artifacts. Notably, G1/S synchronization showed that Ras activity was not required for transit through the incipient S and G2/M, whereas entry into the subsequent S-phase was markedly diminished with cells accumulating at G0/G1. As such, these data were compatible with the findings from Hitomi and Stacey and they were also consistent with data from Livingston and co-workers who showed that repressing Ras in cells synchronized at G2/M triggered entry into quiescence in T98G cells [60]. In addition, our data reach beyond those findings and highlight that continuously cycling cells depend on a Ras signal in G1 for cyclin D stabilization and transit to S-phase. Altogether, in the simplest sense, cells require an input from Ras/Erk at mid G1 for S-phase entry, regardless of whether cells emerge from G0 or M. In addition, a distinct Ras signal in G2 or in any case prior to M, possibly related to the control of CDK2 activity [77], might determine the propensity to exit the cell cycle and achieve stable G0 versus committing to a new cell cycle round.

In sum, this study highlights that cells depend on a mitogenic Ras signal transmitted by Erk at mid G1 and hence close to the restriction point for G1/S transition. This concept is reminiscent of previous findings from the Kazlauskas lab documenting a similar dependency on mitogens at mid-G1 for efficient G1/S transition [78]. Assuming that Ras-GTP levels at mid-G1 are themselves dependent on continuous mitogen presence, a concept for which there is currently no direct rigorous support in the literature, this could be interpreted in a way that cells read out the “mitogenicity” of their environment at the level of the Ras/Erk signal output to ensure that DNA synthesis proceeds only in a favourable, protractedly pro-mitogenic environment. Our findings further document that Ras, unexpectedly, engages only one of three major effector pathways to drive cells through G1 into S. This raises the question as to the genuine physiological role of Ras/PI3K and Ras/RalGEF signaling. Given the large body of literature documenting the importance of PI3K and Ral signalling in Ras-dependent transformation, our data support the possibility that PI3K and Ral could play supportive, not purely mitogenic roles in Ras-driven cancer e.g. related to apoptosis evasion or metabolic rewiring. Alternatively, it is equally well conceivable that oncogenic Ras usurps PI3K and RalGEFs, re-purposing them and converting them into genuine drivers of cell cycle progression.

Materials and methods

Materials

The ARGENT heterodimerization kit for the generation of genetically encoded heterodimerization constructs was from ARIAD pharmaceuticals. The A/C Heterodimerizer (iDim) was acquired from Clontech. Accutase, Nocodazole, Thymidine, Puromycin and Polyethylenimine (branched) were purchased from Sigma-Aldrich. LY294002, U0126 and Akt inhibitor VIII were from Enzo Life Science. GDP, Glutathione-agarose and Picolyl azide sulfo-Cy5 were purchased from Jena Bioscience. Click-iT Plus EdU Alexa Fluor 647 Flow Cytometry Assay Kit was from Molecular Probes. EGF was from Preprotech and RO3306 was kindly provided by Helmut Pospiech, Jena.

Antibodies

Western blot antibodies

The following primary antibodies were purchased from Cell Signaling Technology: Akt (#9272), phospho-S473Akt (#9271), phospho-T308Akt (#9275S), Cyclin D1 (92G2) (#2978), EGFR (#4267), phospho-Y1068EGFR (#2236), p44/42MAPK (Erk1/2) (137F5) (#4695). Anti-pan-Ras C-4 was from Calbiochem. H-Ras (C-20), K-Ras (F234), N-Ras (F155), TC21 (V-20), phosphoErk (Thr202/Tyr204) (E-4) and HRP-anti sheep were obtained from Santa Cruz Biotechnology. mTOR (human FRB Domain) was from Enzo Life Science. RalA (#610221), Rap1 (clone 3) and RB2 Clone 10 (#610261) were acquired from BD Transduction Laboratories. Abs against Cherry and GFP were kindly provided by Ian A. Prior, Liverpool. Secondary HRP-anti mouse and HRP-anti rabbit Abs were acquired from KPL.

FACS antibodies

Phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204) (E10) (Alexa Fluor 647 Conjugate) (#4375) used for FACS analysis was purchased from Cell Signaling Technology. Alexa Fluor 488 anti-Ki-67, Clone B56 (RUO), was from BD Transduction Laboratories.

Cloning of heterodimerization constructs

All constructs were cloned by standard restriction and PCR cloning techniques using the following sequences: the human K-Ras hypervariable region + CAAX domain (K-Ras-hvr-CAAX; amino acids 164–188), FKBP and FRB from the ARGENT heterodimerization kit, EGFP, mCherry, catalytic domain of human NF1 (NF1_cat, amino acids 1174–1537). All modules were connected by linkers of 4–6 amino acids in length. Heterodimerization constructs #1 to #4 were assembled in the pcDNA3.1+ vector (ThermoFisher) whereas constructs #5 and #6 were constructed in pC4-RHE from the ARGENT kit. All constructs were ultimately subcloned into the lentiviral expression vector pCDH-CMV-MCS-EF1-Puro (System Biosciences) via PCR cloning and verified by sequencing.

Cell culture and cell lines

Cervical cancer epithelial HeLa cell line was purchased from ATCC and cultured at 95% humidity, 37°C and 5% CO₂ atmosphere in DMEM (Dulbecco’s modified Eagle’s medium, Biowest) supplemented with 10% FCS (GIBCO). HEK293T cells were obtained from ATCC and cultured in DMEM/F-12 (Dulbecco’s Modified Eagle Medium/Nutrient Mixture F-12, Biowest) supplemented with 10% FCS. T98G glioblastoma cells were kindly provided by Dr. Helmut Pospiech (FLI, Jena, Germany). T98G cells and all derivate lines were cultured in DMEM (Dulbecco’s modified Eagle’s medium) supplemented with 10% FCS, 1x L-Glutamax, 1 mM Sodium pyruvate and 1x NEAAs (non-essential amino acids) under standard conditions (95% humidity, 5% CO₂, 37°C). T98G cells were split 2–3 times per week, always before they reached confluency. The culture medium of stably transduced HeLa#1#6 and T98G#1#6 cell lines was supplemented with 1.5 μg/ml and 1 μg/ml puromycin, respectively. All used cells were free of mycoplasma contamination.

Lentiviral transduction and stable cell line generation

Stable expression of heterodimerization constructs in HeLa or T98G was achieved through lentiviral transduction using third-generation lentiviral vectors and subsequent selection. Briefly, lentiviral expression plasmids encoding the heterodimerization construct(s) in pCDH-CMV-MCS-EF1-Puro were co-transfected with two packaging vectors (pMDL-g and pRSV-g) and an envelope vector (pVSV-g) into HEK293T via standard lipofection using polyethylenimine (PEI). Infectious lentiviral particles were collected from the supernatant 24–72 hours later, mixed with polybrene and used to infect HeLa or T98G cells in 6-well plates. Cell plates were spun 1 h at 500 g and 48 h later medium was supplemented with 2 µg/ml and 1.5 µg/ml puromycine for selection of HeLa or T98G cells, respectively. After one week of selection, puromycine concentration was reduced to 1.5 µg/ml (HeLa) and 1 µg/ml (T98G) and cells positive for both heterodimerization proteins were further enriched via flow cytometry sorting (FACS sorting).

Synchronization and release techniques

T98G cells were induced to exit the cell cycle and attain G0 by depriving them of serum. To this end, cells were washed 3 x with DMEM and kept in DMEM supplemented with 0.15% FCS for 72 h or longer, as indicated. Quiescent cells were induced to reenter the cell cycle by refeeding with release medium (a 1:1 mix of DMEM/10% FCS and conditioned T98G full medium).

G1/S synchrony was accomplished by 24 hours of treatment with 2 mM thymidine in the presence of serum. Cells were released into S-phase by washing out the drug and refeeding with release medium (a 1:1 mix of growth and conditioned T98G medium) as above.

Cells were synchronized at G2/M phase by a double block with thymidine and RO3306. Briefly, cells were treated with 2 mM thymidine in full medium for 24 hours. After 3 rounds of washing with warm PBS, cells were released for 3 hours in release medium before adding 10 μM RO3306 for further 14 hours to achieve synchrony at G2/M. Cells were released from the G2/M block by removing the inhibitor by 3 rounds of washing with PBS.

Cell stimulations and western blot analysis

HeLa or T98G cells were treated with the heterodimerizer iDim or relevant inhibitors as indicated in the results section and figure legends. All drugs and inhibitors were administered as 1000fold or higher stocks in DMSO. HeLa cells were deprived of serum for16 h prior to growth factor challenge. Cells were seeded on 60 mm dishes (T98G) or 6-well plates (HeLa) and stimulated with 10 ng/ml EGF, 10% FCS or 5 μg/ml Insulin for varying periods of time. Reactions were quenched by removing the medium and addition of 500 μl or 850 μl ice-cold RIPA buffer supplemented with protease and phosphatase inhibitors for 6-well plates or 60 mm dishes, respectively. Samples were cleared by centrifugation, treated with Laemmli buffer and resolved by SDS-PAGE. Equal amounts of protein/sample, as determined using the Pierce BCA Protein Assay Kit, were loaded. Unless stated otherwise, activated/phosphorylated Akt (abbreviated p-Akt in the figure panels) was detected with the phospho-S473Akt Ab.

GTPase activity assays

Cells were seeded in 6-well plates, cultured to around 70–80% confluence and deprived of serum overnight (HeLa) or for three days (T98G) in advance of the assay, unless otherwise stated. Cells were treated and challenged with serum or mitogens as appropriate and lysed immediately (without previous washing of the cell monolayers) in 0.5 ml ice-cold lysis solution [50 mM Tris pH 7.5, 150 mM NaCl, 1 mM EGTA, 5 mM MgCl2, 10% glycerol, 1% NP40 (Nonidet-P40)] supplemented with protease and phosphatase inhibitors, 100 μM GDP and 20 μg/ml soluble recombinant affinity domains for the cognate GTPase. These included the RBD from Raf1 for Ras-GTP and TC21-GTP, the Ras association (RA) domain of RalGDS for Rap-GTP and the RalBP1 (also known as RLIP or RLIP76) domain for Ral-GTP. All affinity domains were expressed and produced in E. coli as soluble GST-fusion proteins by standard procedures, including a final size exclusion chromatography to remove free glutathione. GDP and the relevant affinity domain were included at the time point of lysis to quench post-lytic GTP-loading and GAP-dependent GTP hydrolysis of the cognate GTPases, respectively. Cell material was scraped off and lysates were cleared by centrifugation in the cold. Active GTPase molecules complexed to their cognate affinity domains were collected on Glutathione-agarose (30 min at 4°C on a rotating wheel) and washed once with 750 μl plain lysis buffer. Beads were drained of all liquid before addition of Laemmli sample buffer. Samples were denatured at room temperature for at least one hour and processed for Western Blotting.

Cell cycle analysis & cell proliferation assays

Cell cycle distribution was determined through pulse-labelling with EdU (5-ethynyl-2ʹ-deoxyuridine, a thymidine analogue) combined with DNA staining. Briefly, 2 × 10⁶ T98G cells were supplemented with 10 µM EdU for the last 10 min of treatment. Cells were washed once with 1 ml warm PBS and harvested with 800 μl pre-warmed Accutase. Suspended cells were washed once with 1 ml 1% BSA in PBS (further referred to as 1% BSA solution) and fixed sequentially with freshly prepared 4% PFA (15 minutes of incubation in the dark at room temperature) and ice-cold 100% ethanol. The latter was added dropwise while gently vortexing the sample to avoid cell clumping. Removal of the fixative was achieved by centrifugation for 5 minutes at 300 g and one washing step with 1% BSA solution. Fixed cells were further processed with the Click-iT EdU-AlexaFluor647 Flow Cytometer Assay kit according to the manufacturer´s instructions. Finally, samples were washed once with 1% BSA, resuspended in DAPI/PBS solution and analysed using the LRS Fortessa flow cytometer (BD Bioscience) equipped with blue (488 nm) and red (633 nm) lasers. In some experiments, DAPI was replaced for equivalent dyes like Hoechst 33,342 or Propidium Iodide. FlowJo software was used for quantification and analysis of all measurements. Gating of the total number of cells was done in a FSC-A vs SSC-A dot plot, from which a second dot plot DAPI-A vs DAPI-W was created to eliminate doublets or aggregates. Singlet nuclei were displayed in a third DAPI-A vs APC-A (EdU-Alexa647 or EdU-Picolyl-azide sulfo Cy5) dot plot, where the distinct phases of the cell cycle were further gated.

Fluorescence microscopy

Live-cell imaging was carried out on a Zeiss (Jena, Germany) LSM 510 confocal laser scanning microscope equipped with an inverted microscope (Zeiss Axiovert 100M) and a thermostated stage chamber (IBIDI, München, Germany) as described previously [47]. Briefly, cells were seeded on self-made collagen-coated, glass-bottomed 35 mm cell culture dishes and treated as appropriate. Cells were placed on the microscope stage and challenged with mitogens on-stage. Confocal images (optical slice of ≤ 1 μm) were acquired using a 63x water immersion objective lens. EGFP was excited with the Argon 488 nm line and emitted fluorescence was collected with a 505–550 nm band-pass filter. mCherry was excited with the HeNe 543 nm laser line and fluorescence was recorded with a 560 long-pass filter. All images of a series were subjected to the same processing routine using Zeiss ZEN 2008 Light Edition software.

Statistical analysis

Two-way non-parametric ANOVA was used for statistical analysis; *p < 0.05, **p < 0.01, ***p < 0.001.

Funding Statement

This work was supported by the DAAD [91529654].

Disclosure of Potential Conflicts of Interest

No potential conflict of interest was reported by the authors.

Acknowledgments

We acknowledge helpful discussions and the generous provision of reagents by Helmut Pospiech, Jena, Germany. We thank Dmitriy Chudakov and Toshihiko Oki for the kind gift of plasmid DNAs. We are thankful to Ute Wittig and Yvonne Schlenker for excellent technical assistance.

Author Contributions

L.V., G.P. and I.R. designed the experiments. L.V. performed the majority of experiments. S.B., G.P., and I.R. contributed to the experiments. CB wrote the software for the microscopic image analysis and analyzed the fluorescence microscopy data, I.R. coordinated, conceived and designed the study and wrote the manuscript.

Supplemental material

Supplemental data for this article can be accessed on the publisher’s website.