Optimization of experimental parameters and analysis of the prenylation behaviour of the protein Rap2A
The proposed new procedure starts with a PCR-amplification of the GST-Rap2A open reading frame (Genbank accession of Rap2A BC070031) followed by in vitro transcription and translation using rabbit reticulocyte lysate in the presence of a 3H-labeled isoprenoid donor. The GST-tagged target protein is purified utilizing glutathione sepharose 4B beads and concentrated by precipitation with acetone. The sample is subjected to SDS-page gel electrophoresis and transferred to a nitrocellulose membrane by electroblotting. Detection of incorporated radioactive label is performed by scanning with the TLC analyzer (scanning time: 20 minutes per lane). Afterwards, the amount of target protein is evaluated by Western blotting with an anti-GST-antibody on the same membrane.
Experiments with wildtype GST-Rap2A fusion protein and [3H]mevalonic acid were performed using various reaction times and amounts of radioactive label. The optimal conditions we found were 20–40 μCi [3H]mevalonic acid and at least four hours reaction time, which is in agreement with previous studies .
Experiments with 20 μCi [3H]mevalonic acid, 10 μCi [3H]farnesylpyrophosphate (FPP) and 10 μCi [3H]geranylgeranylpyrophosphate (GGPP) allowed affirmation of prenylation of Rap2A and identification of the preferred isoprenoid attached to Rap2A as a farnesyl-group. However, geranylgeranylation did also occur under the given conditions, but with much lower efficiency (Figure 1). The respective peak area for FPP incorporation is about 15 times the one for GGPP integration. The mutated version (C180A) was used as a negative control reaction to rule out unspecific attachment and to confirm the location of the modification.
All results obtained with our new method were consistent with previously reported data on Rap2A , demonstrating the functionality of the assay. It should be noted that the time consumption of the scanning procedure (1–2 hour per gel) is markedly reduced compared with autoradiography (weeks-months). To allow direct comparison of methods, a film has been exposed with the same Western membrane used for detection with the TLC-Scanner after application of En3Hance Spray from PerkinElmer for one and for three weeks at -80°C. An exposure of three weeks was necessary to get a distinct signal from all bands, which had given a strong signal with the scanner. However, it wasn't sufficient to detect the low amount of GGPP incorporated (Figure 2). If the expression of the target protein is lower than that of Rap2A, the autoradiography can require months of exposure time.
To determine the enzyme prenylating Rap2A, we performed the same assay with and without inhibitors of prenyltransferases. The signal yielded by incorporation of [3H]FPP was reduced to background level upon addition of 50 μM of the FTase inhibitor FTI-277. In addition, the already weak signal of [3H]GGPP-incorporation was diminished to background level by FTI, while application of a GGTase inhibitor (GGTI-298) left a small peak (Figure 3). These data suggest that Rap2A is recognized only by FTase, but the enzyme can also transfer a geranylgeranyl-group, albeit with drastically reduced efficiency (1–2 orders of magnitude) as suggested before on the basis of peptide substrate exposure to FTase .
Analysis of the in vitro prenylation of RasD2, v-Ki-Ras2 and RhoA63L with the TLC scanning method
These three candidates have been selected to show the ability of our new technique to detect alternative prenylation modes. RasD2 (synonym: Rhes, BC013419) is suggested to be a farnesylation target but only due to indirect evidence . Whereas K-Ras homologues such as v-Ki-Ras2 (the Q61H oncogene mutant of K-Ras4B, BC013572) are thought to be modified both by FTase and GGTase1 , the RhoA protein (NM_001664.2) is primarily a GGTase1 target .
The same in vitro assay was performed on all three targets. Because of lower translation efficiency, the reaction mix had to be upscaled by a factor of 5 for RasD2 and RhoA63L and a factor of 7 for K-Ras4B compared with the recipe used for Rap2A. For RasD2 and RhoA63L, we used 50 μCi of [3H]mevalonic acid and 25 μCi of [3H]FPP/[3H]GGPP. In the case of v-Ki-Ras2, we applied 60 and 30 μCi respectively. The results for RasD2 were similar to Rap2A with significant incorporation of a product of mevalonic acid as well as FPP, while GGPP yielded only a ca. 40 times weaker signal (as measured via area below peaks, Figure 4). Thus, we have shown with direct arguments that RasD2 is indeed a target for farnesylation .
On the contrary, while also showing preference for FPP, incorporation of GGPP into v-Ki-Ras2 in the absence of FPP is only 2–3 times lower (Figure 5). These results provide strong evidence for the hypothesis of alternative prenylation while inhibiting FTase. RhoA yielded strong signals for the reactions with mevalonic acid and GGPP (Figure 6). The efficiency of FPP attachment is lower than that with GGPP by a factor of 2. Since the amount of protein detected in the Western blot under condition of FPP addition (lane 3) is considerably larger than in the case of exposition to GGPP (lane 4), we suggest that GGPP is indeed the preferred substrate. This is in accordance with the literature that RhoA is geranylgeranylated  and K-Ras can be modified by both isoprenoids .
Electrophoretic mobility shifts of in vivo prenylated proteins
The simplest in vivo test for prenylation is performed with comparative electrophoretic shift analysis of non-prenylated and prenylated protein forms. The differential shift is typically not caused by the prenyl anchor attachment itself but rather by the in vivo post-prenylation processing steps such as subsequent palmitoylation, proteolytic cleavage of the C-terminal tripeptide of the CaaX box or C-terminal methylation. These mobility shifts are generally small and not easily detectable for all proteins due to their differential post-prenylation processing and possible variable degradation of the prenylated and non-prenylated forms.
Clear electrophoretic mobility shifts have been observed for Rap2A providing an indirect argument for its farnesylation (Figure 7). In the case of the wild-type protein, we see two bands corresponding to the non-farnesylated (slow) and farnesylated (fast) forms (lane 1). As a result of application of lovastatin (lane 2), the fast band representing the farnesylated Rap2A disappears (and the slow band grows in intensity). This effect can be reversed by application of growing amounts of exogenous FPP.
In vivo subcellular localization of N-terminally GFP-tagged constructs
To confirm the biological relevance of the results from our in vitro assays, we analyzed the subcellular localization in HeLa cells of the same proteins as N-terminal GFP-fusion constructs. We tested the wildtype forms, the variants with a mutation at the prenylation site and the wildtype forms together with FTI and GGTI (Figure 8). Fluorescence microscopic views of Rap2A and RasD2 expression showed definite membrane localization for the wildtype protein without and with GGTI. The mutant proteins and the wildtype proteins treated with FTI mislocalized and accumulated in the nucleus. A GFP-fusion protein of RhoA63L, which has been shown to be a primary geranylgeranylation target and which has been previously used for localization studies , was used to demonstrate the functionality of the GGTI treatment. Membrane localization is observed for wildtype protein without and with FTI, nuclear mislocalization is found for the mutant and wildtype protein with GGTI. These observations agree with the results from the in vitro prenylation assay.
Further, we investigated the subcellular localization of the GFP-v-Ki-Ras2 fusion protein in HeLa cells. As shown in Figure 8 (part 9), fluorescence microscopy clearly revealed that the fusion protein was co-localized with cellular membranes. A GFP fusion construct harboring a Cys to Ala mutation within the CaaX box predominantly accumulated in the nucleus (Figure 8, part 10). When using specific inhibitors of farnesylation (FTI-277) or of geranylgeranylation (GGTI-298), we surprisingly found that v-Ki-Ras2 was present mainly in the nucleus with FTI-277 (Figure 8, part 11), whereas GGTI-298 did not show any affect on the localization of the fusion protein (Figure 8, part 12).
In the literature, K-Ras4A and K-Ras4B have been reported to be predominantly farnesylated in vivo. In the presence of potent FTIs, both proteins were alternatively prenylated by geranylgeranyltransferase-1 in the human colon carcinoma cell line DLD-1 and COS cells . Respectively, K-Ras4A and K-Ras4B were found to be associated with the membrane fraction independent of the kind of prenylation in COS cells. For complete inhibition of K-Ras4B prenylation, a combination for FTI-277 and GGTI-298 was required as examined in five different human carcinoma cell lines from pancreatic, pulmonary, and bladder origins . The differing results can be due to differences in cell lines, Ras substrates or GFP-v-Ki-Ras2 overexpression. In the latter case, the ratio of prenylpyrophosphate to substrate protein is skewed. Indeed, at high expression levels, GFP-v-Ki-Ras2 was always found predominantly in the nucleus, independent of the presence of FTIs, GGTIs or the Cys-to-Ala mutation within the C-terminal CaaX box. In support of our interpretation, Rilling et al.  reported that protein prenylation in Chinese hamster ovary cells can vary as a function of the extracellular mevalonate concentration. Fortunately, only for v-Ki-Ras2, we found the localization studies to be technically tricky, fragile and the results difficult to reproduce. Whereas cells were sensitive for overexpression of wildtype GFP-v-Ki-Ras2 resulting in low transfection efficiency and, consequently, the number of transfected cells was low, no similar difficulties could be observed for GFP-vi-K-Ras2 mutant C185A or any of the other GFP-fusion constructs of RasD2, Rap2A or RhoA.
Analysis of the in vivo prenylation of Rap2A with the TLC scanning method
It would be desirable to test whether the TLC scanning method is applicable also for the investigation of protein targets exposed to metabolic labelling with radioactive precursors in vivo and purified with immunoprecipitation from cell culture, SDS-page gel electrophoresis and Western transfer. Since we expected the translation efficiency to be critical for the success of the experiment, we selected Rap2A as test target (Figure 9). Indeed, it was possible to clearly show incorporation of radioactive FPP into Rap2A overexpressed in HeLa cells and recovered by immunoprecipitation with anti-HA-antibodies (lane 1) and the absence of the anchor in the C180A mutant treated identically (lane 2). It is especially notable that the amount of purified protein can be evaluated with an anti-HA-antibody on the same Western blot that was used previously for TLC-scanning similarly to the in vitro protocol with the anti-GST-antibody.
Attempts to detect the prenylation status of Rap2A with HPLC-based purification methods
We have also attempted the proof of prenylation with chromatographic methods. In one of these variants, we utilized coupled in vitro transcription/translation and prenylation. But labelling with 3H-marked isoprenoids was replaced by usage of [35S]methionine during translation, while the added isoprenoid was not radioactively marked. Purification via GST-beads was performed in analogy to the method described above but, after precipitation with acetone, the protein was resuspended in a denaturation buffer containing 50 mM Tris-HCl pH 8.0 as well as 4 mM dithiothreitol (DTT) and 8 M urea. After denaturation, the solution was diluted by addition of 10 volumes 50 mM NH4HCO3. Following digestion with trypsin, the peptides were separated using a C18 column in reverse phase HPLC. Radioactivity was detected by scintillation counting of the collected fractions after the UV signal was recorded. Since the C-terminal peptide of Rap2A contains a methionine residue, a radioactive signal should be found at a retention time characteristic for farnesylated peptides, while it should be absent for the C180A mutant, because the non-prenylated peptide would elute much earlier. Although the expected behavior was observed in singular experiments, we have not been able to select experimental conditions for the reproducible detection of prenylated peptides (see discussion).