An alkyne-containing isoprenoid analogue based on a farnesyl diphosphate scaffold is a biologically functional universal probe for proteomic analysis

Abstract

Prenylation consists of the modification of proteins with either farnesyl diphosphate (FPP) or geranylgeranyl diphosphate (GGPP) at a cysteine near the C-terminus of target proteins to generate thioether-linked lipidated proteins. In recent work, metabolic labeling with alkyne-containing isoprenoid analogues including C15AlkOPP has been used to identify prenylated proteins and track their levels in different diseases. Here, a systematic study of the impact of isoprenoid length on proteins labeled with these probes was performed. Chemical synthesis was used to generate two new analogues, C15hAlkOPP and C20AlkOPP, bringing the total number of compounds to eight used in this study. Enzyme kinetics performed in vitro combined with metabolic labeling in cellulo, resulted in the identification of 8 proteins for C10AlkOPP, 70 proteins for C15AlkOPP, 41 proteins for C15hAlkOPP, and 7 proteins for C20AlkOPP. While C10AlkOPP was the most selective for farnesylated proteins and C20AlkOPP was most selective for geranylgeranylated proteins, the number of proteins identified using those probes was relatively small. In contrast, C15AlkOPP labeled the most proteins including representatives from all classes of prenylated proteins. Functional analysis of these analogues demonstrated that C15AlkOPP was particularly well suited for biological studies since it was efficiently incorporated in cellulo, was able to confer correct plasma membrane localization of H-Ras protein and complement the effects of GGPP depletion in macrophages to yield correct cell polarization and filopodia. Collectively, these results indicate that C15AlkOPP is a biologically functional, universal probe for metabolic labeling experiments that has minimal effects on cellular physiology.

Graphical Abstract

INTRODUCTION

Post-translational modifications, including protein prenylation, play crucial roles in protein function and regulation.^1,2 Given the widespread involvement of prenylation in protein localization, membrane trafficking, and cellular signaling, dysregulation of prenylation and prenylated proteins have been implicated in the pathogenesis of numerous diseases.^3–6 Prenylation consists of the modification of proteins with either farnesyl groups from farnesyl diphosphate (FPP,) or geranylgeranyl moieties from geranylgeranyl diphosphate (GGPP), at a cysteine near the C-terminus of target proteins (Figure 1A).⁷ There are 3 canonical subtypes of prenylation including farnesylation, type I geranylgeranylation, and type II geranylgeranylation. Farnesylation and type I geranylgeranylation target proteins with an identifying tetrapeptide “CaaX” motif, where C is the modified cysteine, “a” is an aliphatic amino acid, and “X” is a variable amino acid, for modification with either FPP or GGPP respectively. Type II geranylgeranylation is a more complex process since protein selection occurs in concert with an additional escort protein and involves residues located substantially upstream from the site of prenylation. Consequently, short peptides cannot generally be used as substrates to study that latter enzymatic process. Another key difference between type I geranylgeranylation and type II is that in the latter, Rab proteins are usually modified with two geranylgeranyl moieties. Considering this, the sequence of the modified motif for Rab proteins is more variable and the modified cysteines are in patterns including “CXC”, “CXXC”, “XXCC”, or “CCXX”.⁸ Additionally, two new prenylation processes have been discovered recently. First, type III geranylgeranylation which acts much like type I geranylgeranylation. However, this process has yet to be fully characterized with the number of protein substrates being unclear.^9,10 Second, a specific protein, ALDH9A1, undergoes self-prenylation with geranylgeranial or farnesal resulting in a farnesoylated or geranylgeranoylated protein linked to the cysteine via a thioester bond.¹¹ To date, only one example of this has been reported.

Figure 1: General schematic of prenylation, enzyme function, and comparison of probes.

A) General schematic of the farnesylation and geranylgeranylation type I, two subtypes of prenylation. B) Comparisons of the prenylation probes used in this study compared to the native substrates, purple line shows length of FPP, blue line shows length of GGPP. C) Computational modeling showing the docking of C10AlkOPP (carbon; orange), C15AlkOPP (carbon; purple), C15hAlkOPP (carbon; yellow) and C20AlkOPP (carbon; green) into the three prenylation enzyme binding pockets superimposed on the native substrate FPP (carbon; pink) or GGPP (carbon; blue) respectively. Left FTase binding cavity (1JCR, red mesh), middle GGTase I binding cavity (1TNO, blue mesh), right GGTase II binding cavity (3DST, green mesh). For all ligands, oxygen is red, and phosphorus is orange.

The subtypes of prenylation have been shown to have distinct roles in the development of disease and biological functions. For example, in Alzheimer’s Disease (AD), it has been shown that in an AD mouse model, haploid deficiency of FTase in the forebrain rescued cognitive function and reduced neuroinflammation whereas haploid deficiency of GGTase I only reduced neuroinflammation in the brain tissue.^12,13 This suggested that farnesylation and geranylgeranylation play different roles in the development of AD. Furthermore, it is estimated that there are over 60 Rab proteins in humans, each having distinct roles in cell physiology. Dysregulation of specific Rab proteins has been identified in a variety of different diseases including, cancer (Rab5 in malignant human lung cell adenocarcinoma¹⁴, Rab1 in tongue squamous cell carcinomas,¹⁵ Rab25 in aggressive ovarian and breast cancers¹⁶), choroideremia (which is caused by a mutation in the Rab escort protein, resulting in a buildup in Rab27a¹⁷), and neurological diseases (Rab11 and Rab8 in Huntington’s disease¹⁸, Rab3a and Rab8 in Parkinson’s disease¹⁹, and Rab23 in Carpenter syndrome²⁰). Investigating the role of prenylation in all these disease models, exploring the roles of the different subtypes of prenylation, and identifying and tracking the levels of all the prenylated proteins is a crucial task that has stimulated research for some time.

Fortuitously, the prenyltransferase enzymes manifest some tolerance for structural flexibility in the isoprenoid substrate, allowing efficient incorporation of analogues of FPP and GGPP containing embedded bioorthogonal functionality.^21–23 Such analogues have been successfully used in two different forms including the functional diphosphate form and the precursor alcohol form. The intracellular endogenous kinases responsible for phosphorylation of the alcohol precursor to yield the physiologically relevant diphosphates have not been identified, with one hypothesis being that a two-step process involving farnesyl kinase and farnesyl-phosphate kinase may be responsible.²⁴ That two-step process has recently been reproduced in an in vitro system.²⁵ Despite this gap in knowledge, many alcohol precursors of FPP and GGPP analogues have been used in cell experiments.^26–32 These analogues are the foundation of the chemoproteomic strategy to investigate the role of prenylation in disease frequently denoted as “prenylomics”.^33,34 In that approach, cells, tissues or animals are metabolically labeled with an analogue, typically containing a bioorthogonal alkyne handle, that allows for enrichment of the prenylated proteins for identification with a bottom-up proteomic analysis.^26,35 In work from our lab and others, alkyne analogues have been successfully used to identify all subtypes of prenylation.^35–38 One analogue known as C15AlkOPP (2b, Figure 1B), has been one of the primary tools in prenylomic investigations. That well characterized analogue labels all subtypes of prenylation substrates^28,39,40 which is helpful for investigating the totality of prenylation. However, if investigation of only one subtype of prenylation is desired, a more complex experimental design is required.

It has been previously demonstrated that analogues longer than FPP are more selective for labeling geranylgeranylated proteins.^23,26,39 However, those analogues have either used a terminal azide (instead of an alkyne) or a more synthetically challenging probe structure involving an alkyne linked to a farnesyl scaffold through a carbon-carbon bond. It should be noted that the latter was used in the precursor alcohol form.²⁶ Early metabolic labeling and proteomic work using C15AlkOH (2a) showed less incorporation of the analogue into prenylated proteins compared to more recent work with the fully phosphorylated C15AlkOPP analogue.^28,29,35,39 It is important to note that metabolic labeling with the alcohol precursor C15AlkOH was previously examined in HeLa cells, where lower levels of incorporation were observed compared to other cell lines such as COS7,³⁵ but C15AlkOH has not been previously used to identify prenylated proteins with the current prenylomic methodology.

Described here is an examination and comparison of the use of either the alcohol precursor (C15AlkOH) or the fully phosphorylated (C15AlkOPP) analogue as tools to identify prenylated proteins using the chemoproteomic bottom-up approach. Additionally, two new GGPP analogues and their corresponding precursor alcohols were synthesized and characterized. The selectivities of these new probes were investigated along with a shorter, previously established, analogue, C10AlkOPP (1b) and its alcohol progenitor. That shorter analogue was predicted to be selective for FTase given its smaller two-isoprene unit scaffold. Using these different length analogues, we hypothesized it should be possible to selectively label farnesylated proteins with the shorter analogue and geranylgeranylated proteins with the longer analogues. The labeling profile of each analogue and precursor alcohol was determined via metabolic labeling in COS7 cells, which have been thoroughly investigated in previous work.^35,41,42 The function of these analogues was also examined in several biochemical and cell-based assays to study the impact of the chemical modifications on incorporation efficiency and biological function of proteins modified with these non-natural isoprenoids.

RESULTS AND DISCUSSION

Computational docking experiments predict that longer probes are selective for GGTase.

To explore how the length of the proposed probes might affect enzyme specificity, computational docking studies were performed with C10AlkOPP (1b), C15AlkOPP (2b), C15hAlkOPP (3b), and C20AlkOPP (4b) (Figure 1C and S1). These probes were docked into FTase (1JCR, red), GGTase I (1TNO, blue), and GGTase II (3DST, green), and compared to the native isoprenoid substrates, FPP and GGPP respectively. The top scoring poses for each analogue bound to FTase, GGTase I and GGTase superimposed on the structures of bound FPP or GGPP are also shown in Figures S2 (FTase), S3 (GGTase I) and S4 (GGTase II). For FTase (Figure 1C, red) the binding pocket did allow the diphosphate groups from C10AlkOPP (orange), C15AlkOPP, (purple) and C15hAlkOPP (yellow) to all align well with the corresponding moiety of FPP, suggesting that these analogues could bind in a catalytically productive manner. In contrast, the diphosphate group of C20AlkOPP (green) did not superimpose with that of FPP and instead extended over 6 Å outside of the binding pocket, indicating that it should not be an effective FTase substrate. The conformation of C10AlkOPP (Figure S2A) was almost identical to that of FPP over the first two isoprene units, although there was some divergence in the distal region (C-1’ to C-3’). Both C15AlkOPP (Figure S2B) and C15hAlkOPP (Figure S2C) overlapped with FPP in the first isoprene unit but then deviated after that. Deviation was greater in the case of the C15hAlkOPP with the largest difference observed in the third isoprenoid unit, with carbons C-11, C-12, and C-15 (see Figure S1 for numbering scheme and distance measurements), rotated away from the same atoms in FPP leading to a 2.6 Å translocation between C-15 in the two compounds. In contrast, for C15AlkOPP, that perturbation of C-15 positions was only 1.6 Å. However, the overall isoprenoid lengths of the C15hAlkOPP and C15AlkOPP complexes were still quite similar to that of FPP, with the terminal alkyne carbons extending only 1.5–1.6 Å beyond C-15 in FPP. Overall, these docking results suggest that C10AlkOPP is the best mimic of FPP with C15AlkOPP also providing good complementarity. For the larger analogues, C15hAlkOPP and C20AlkOPP, their larger size limits their ability to effectively duplicate the conformation of the physiological substrate, FPP, in the FTase active site.

The docking of the probes in GGTase I (Figure 1C, blue) was somewhat more complex. While the P-α atoms from all the analogues were essentially superimposable with the P-α atom in GGPP, the P-β atoms were somewhat displaced. For the isoprenoid portion, the conformation of C10AlkOPP was quite similar to that of GGPP (Figure S3A). In contrast, for C15AlkOPP, substantial differences in the conformation of both the first (C-1 through C-4) and second (C-5 through C-8) isoprenoid units but not the third (C-9 through C-12) were observed (Figure S3B). For C15hAlkOPP (Figure S3C) and C20AlkOPP (Figure S3D), differences in the first and third but not the second were observed. The significance of the changes in the first two isoprene units is somewhat unclear since similar changes are known to occur in the GGPP conformation as the enzymatic prenylation reaction proceeds.⁴³ Probably the most useful insight obtained from these calculations come from an analysis of the overall length of the various docked structures compared to the native isoprenoid, GGPP. For C15AlkOPP, the terminal alkyne carbon (C-3’) extends 1.8 Å past the terminal atom (C-20) of GGPP while that value increases modestly to 2.3 Å in the case of C15hAlkOPP. However, that distance (4.4 Å) is dramatically larger in the case of C20AlkOPP. Thus, these docking results suggest that C15AlkOPP and C15hAlkOPP can mimic the bound conformation of GGPP and hence should function as effective substrates for GGTase I with C20AlkOPP performing more poorly.

For GGTase II, the diphosphate and first and second isoprene moieties for all probes aligned well with GGPP although there was substantial divergence at the more distal positions. In the case of C10AlkOPP, differences were minor (Figure S4A). For C15AlkOPP (Figure S4B), the linear alkyne forced it to rotate to a position where C-3’ was 4.1 Å from C-16 in GGPP. In contrast, C15hAlkOPP (Figure S4C) fit better with the comparable C-4’ to C-16 distance reduced to 1.4 Å. For C20AlkOPP (Figure S4D), a major rotation of the alkyne group and fourth isoprene unit occurred, displacing C-20 in the two structures by 2.8 Å, resulting in C-3’ from the analogue being positioned 7.4 Å from C-16 of GGPP. It is likely that this rotation moves the analogue into the protein substrate binding site potentially hindering efficient enzymatic prenylation. Given the large size of the GGTase II active site, it is unclear whether the modest differences observed between C15AlkOPP and C15hAlkOPP binding could impact enzymatic processing of these compounds. However, it is likely that C20AlkOPP binding results in steric clashes that may adversely impact efficient catalysis.

Synthesis of novel geranylgeranylation probes.

The syntheses of both C10AlkOPP and C15AlkOPP have been previously described.^39,40,44 In brief, the linear route starts with commercially available geraniol (12) or farnesol by alcohol protection followed by Riley oxidation of the terminal carbon for the subsequent attachment of the alkyne handle. This is followed by alcohol deprotection, bromination and phosphorylation of the terminal alcohols. A related strategy was employed for the synthesis of C15hAlkOPP (Figure S5) from farnesol. While the initial synthetic steps for preparing 6 were the same as for C15AlkOPP, for incorporation of the homo-propargyl group, the terminal hydroxyl group was converted to a chloride to yield 7 in 72% yield. That more electrophilic allylic chloride facilitated the etherification reaction with homopropargyl alcohol (8) to prepare the ether-linked alkyne (9) in 61% yield. Following deprotection to generate C15hAlkOH (3a), the resulting alcohol was brominated and phosphorylated to give C15hAlkOPP. While the linear synthesis described above is useful for the C₁₅ or C₁₀ scaffold, approaching the synthesis of a C₂₀ analogue via the same route is exceptionally low yielding. This is primarily because of the lack of regio-selectivity associated with the key Riley oxidation reaction which decreases the yield of the liner synthetic route early on. To overcome limitations, inspiration was taken from Yu et al. and others who have reported the preparation of functionalized long-chain oligoprenols.⁴⁵ Applying their strategy here allowed the low yielding step to be avoided through a convergent route using two geraniol-derived (10 and 11) precursors (Figure S6).

Through the regioselective allyl-allyl coupling of one C₁₀ precursor with a terminal sulfone group, 11, to another C₁₀ fragment with an electrophilic terminal halide, 10, a C₂₀ scaffold was produced with differentially protected alcohols at each end of the isoprenoid scaffold, 17 (Figure S7). Removal of the sulfone group from 17 was potentially problematic as it could result in olefin migration in the presence of the palladium catalyst used in the subsequent reaction. This was overcome using a bulky hydride source, Super Hydride, which sterically drives the reaction to the desired E, E, E, E isomer, 18 in 66% yield. Subsequent removal of the silyl protecting group yielded 19, with the desired C₂₀ scaffold. From there, the synthetic protocol was similar to the route used for C15AlkOPP (2b) and C15hAlkOPP (3b) including etherification with propargyl-bromide to introduce the alkyne handle into the molecule, followed by deprotection, bromination of C20AlkOH (4a) and phosphorylation to give C20AlkOPP (4b) in 21% yield.

In vitro investigations of isoprenoid analogues as substrates for prenyltransferase enzymes.

Initial tests with the new C20AlkOPP and C15hAlkOPP probes as potential substrates for the prenylation enzymes were performed with an established in vitro spectrofluorimetric assay. In this assay, an increase in dansyl group (Ds) fluorescence occurs upon prenylation of a dansylated “CaaX” peptide substrate (Figure S8). This assay was previously used to show that C15AlkOPP is a substrate for both yFTase and hGGTase I.^39,40 Accordingly, Ds-GCVIA (22) was used as a rFTase substrate and Ds-GCVLL (23) was used as a rGGTase I substrate. Fluorescence measurements indicated that Ds-GCVIA was efficiently modified by rFTase using C15hAlkOPP, but not C20AlkOPP (Figure S9A). That latter observation was consistent with predictions from molecular docking discussed above. Promisingly in the GGTase I assays (Figure S9B), it was seen that C20AlkOPP and C15hAlkOPP were both enzymatically transferred to the peptide.

To determine the kinetic parameters for these new probes, the compounds were assayed at different concentrations using rGGTase I or rFTase and the data fitted to a simple Michaelis-Menten model (Table 1 and Figures S9C–D). Those values were then compared with data obtained from all the diphosphate probes. Examining the k_cat values, the order of reactivity for rFTase was C15AlkOPP, then C15hAlkOPP, with C20AlkOPP showing less than 1% the reactivity of FPP. Thus, reactivity decreased with increasing chain length. For k_cat/K_M, the order was FPP, C15hAlkOPP, C15AlkOPP followed by C10AlkOPP. Collectively, these results generally show that for rFTase, increasing chain length leads to decreased catalytic activity although a decrease is also observed when the overall probe length is shorter than FPP. For rGGTase, the order of reactivity in terms of both k_cat and k_cat/K_M was C15hAlkOPP, C20AlkOPP and then C15AlkOPP. Thus, for GGTase, increasing chain length leads to increased catalytic activity, although there appears to be an upper limit, perhaps as the probe exceeds the active site size, as was noted in the docking studies.

Table 1:

Kinetic parameters for probes evaluated using rFTase and rGGTase

	rFTase and Ds-GCVLS					rGGTase and Ds-GCVLL
Analogue	kcat (s−1)	KM (μM)	kcat /KM (M−1s−1)	kcat rel. to FPP (%)	kcat /KM rel. to FPP (%)	kcat (s−1)	KM (μM)	kcat /KM (M−1s−1)	kcat rel. to GGPP (%)	kcat/KM rel. to GGPP (%)
FPPa	0.3	1.5	2.0 × 105	100	100	-	-	-	-	-
C10AlkOPPb	-	-	2.4 × 104	-	12	-	-	-	-	-
C15AlkOPPc	0.037 ± 0.002	0.69 ± 0.15	5.4 × 104	12	27	0.02	3.0	6.7 × 103	38	22
C15hAlkOPP	0.028 ± 0.001	0.26 ± 0.06	1.1 × 105	9	55	0.032 ± 0.0006	0.084 ± 0.02	3.8 × 105	62	122
GGPPd	nse	ns	ns	ns	ns	0.052	0.17	3.1 × 105	100	100
C20AlkOPP	~0.0018			<1	-	0.030 ± 0.0009	0.35 ± 0.06	8.5 × 104	58	27

^aData previously reported in [⁴⁶] for FPP with rFTase.

^bData previously reported in [⁴⁴], found with rFTase and Ds-TKCVLS.

^cData previously reported with the C15AlkOPP probe with hGGTase I and Ds-GCVLL in [⁴⁰].

^dData previously reported in [⁴⁷].

^eGGPP has previously been reported as an inhibitor of FTase and not a substrate.^48,49 A dash (−) indicates that no kinetic measurements were performed.

Identifying whether C20AlkOPP or C15hAlkOPP are suitable substrates for GGTase II is a more challenging task. Geranylgeranylation type II requires the involvement of an additional escort protein for prenylation and typically results in the addition of two geranylgeranyl groups to the target protein.⁵⁰ The escort protein, aptly named Rab escort protein (REP), not only recruits Rab proteins to the GGTase II complex for modification but also binds to and orients the C-terminus of the Rab protein for geranylgeranylation of the cysteine residues (Figure S10A).

Small peptide substrates cannot be used to assay this enzyme. To study how C20AlkOPP and C15hAlkOPP behaved as potential substrates, in vitro reactions were carried out using REP1, GGTase II, the isoprenoid diphosphate, and a Rab protein (Rab7 here). To visualize the prenylated products, the presence of the alkyne handles integrated in the substrates were exploited by performing copper catalyzed azide-alkyne cycloaddition (CuAAC) with TAMRA-azide (28) after the enzymatic reaction followed by in-gel fluorescence analysis. Fluorescent labeling was only visible in the presence of all requisite components (Figure S10B, lane 4). To serve as a point of comparison, assays with C15AlkOPP were used since it has been previously shown to label geranylgeranylated Rab proteins in a variety of experiments.^35,39

For these studies, GGTase II-catalyzed prenylation reactions were carried out over a range of isoprenoid concentrations followed by CuAAC reaction with TAMRA-azide, SDS PAGE fractionation and subsequent in gel fluorescence measurements (Figure S10C). Prior to those experiments, the enzymatic reaction time was optimized to ensure that the measurements were conducted during the initial linear phase of the reaction. The resulting fluorescent product data was fit to the Michaelis-Menten equation to obtain the K_M values. Since absolute quantitation of the amount of product is difficult in these types of experiments, the rate data was normalized to the V_max value obtained for C15AlkOPP to obtain relative V_max values for C15hAlkOPP and C20AlkOPP. The plots from those experiments with all three compounds are shown together (Figure 2) and separately for each compound along with the kinetic parameters (Figure S10D–F).

Figure 2.

Kinetic analysis of GGTase II-catalyzed prenylation of Rab 7 using C15AlkOPP, C15hAlkOPP and C20AlkOPP. All plots show relative rate expressed as a percentage of V_max for C15AlkOPP that was set to 100%.

While the V_max values varied less than 2-fold (73% for C15hAlkOPP and 63% for C20AlkOPP) there was more variation in the K_M parameters. The apparent K_M value for C15AlkOPP was 4.3 μM while those obtained for C15hAlkOPP and C20AlkOPP were 0.34 and 0.63 μM, respectively. Thus, these experiments establish that the C15hAlkOPP and C20AlkOPP probes are good substrates for GGTase II with the former being the most efficient. The general conclusions from these kinetic experiments are that the short C10AlkOPP probe is selective for rFTase although it is not a particularly efficient substrate for that enzyme while the long C20AlkOPP probe is the most selective for geranylgeranylated proteins (it has minimal activity with rFTase) but is not the most efficiently incorporated analogue for the GGTase enzymes. That may be due to its excessive length as suggested from the aforementioned docking results. In contrast, the intermediate length C15AlkOPP and C15hAlkOPP analogues appear to be able to serve as effective substrates for all three prenyltransferases.

Metabolic incorporation of probes in COS7 Cells.

After establishing C15hAlkOPP and C20AlkOPP were substrates for GGTase I and II, their utility in metabolic labeling was explored. The shorter, previously reported probe, C10AlkOPP, was also examined along with the well-characterized C15AlkOPP analogue to provide a comprehensive analysis of the effect of probe length on metabolic incorporation. All probes were studied in both their diphosphate and alcohol precursor forms. Thus, a total of 8 compounds were examined. To investigate metabolic labeling, COS7 cells were treated for 24 h with each of the analogues, either with or without a 6 h lovastatin pretreatment. Lovastatin is an HMG-CoA reductase inhibitor of the first committed step in the mevalonate pathway which produces FPP and GGPP. Thus lovastatin pretreatment results in a reduction of the endogenous levels of FPP and GGPP so greater analogue incorporation can be obtained.²⁸ Probe evaluation was carried out by metabolic labeling in cellulo followed by lysis, CuAAC with TAMRA-azide (28), SDS-PAGE fractionation and in-gel fluorescence analysis (Figure 3A).

Figure 3.

Comparison of metabolic incorporation analysis of C10AlkOPP, C15AlkOPP, C15hAlkOPP, C20AlkOPP probes and pro-analogue counterparts. (A) The metabolic labeling workflow for treating cells with analogous with or without lovastatin. The treated cells are lysed then undergo a copper click reaction with TAMRA-Azide (37). The fluorescently tagged proteins are then visualized in-gel to examine success of metabolic incorporation. (B) In-gel fluorescence analysis of metabolic incorporation of 10 μM C10AlkOPP, C15AlkOPP, C15hAlkOPP, or C20AlkOPP with (odd lanes) or without (even lanes) a 6 h Lovastatin (10 μM, 36) into COS7 cells. (C) COS7 treated in the same manner but with 10 μM of C10AlkOH, C15AlkOH, C15hAlkOH, and C20AlkOH instead, with (even lanes) or without (odd lanes) lovastatin pretreatment. In both the top panel for both is a measurement of TAMRA fluorescence, bottom panel is total protein loading visualized with Coomassie stain. Two regions of note are highlighted with brackets: Blue is the 20–25 kDa region which is where many small GTPases and Rab proteins migrate, common GGTase I and II substrates, and Red which is the 40–75 kDa region where a few more distinctive higher molecular weight FTase substrates appear.

As a starting point, experiments were conducted with C15AlkOPP. In general, alkyne-functionalized probes label proteins in two regions including a 20–25 kDa region (labeled in blue) and a 40–75 kDa region (labeled in red) (Figure 3). The 20–25 kDa region contains a mixture of all types of prenylated proteins, mainly small GTPases, including those modified by FTase, GGTase I, and GGTase II while the 40–75 kDa region contains predominantly farnesylated proteins including Lamins (74 kDa) and DnaJA1 (45 kDa). Inspection of the data for C15AlkOPP (Figure 3, lanes 4 and 5) shows that this probe efficiently labels proteins in both the 20–25 kDa and 40–75 kDa regions suggesting that it is able to label all types of prenylated proteins.

A small increase in labeling in the 40–75 kDa range (compare Figure 3, lanes 4 and 5) occurred in the presence of lovastatin suggesting that this compound is a poorer substrate for FTase compared with the GGTases and requires suppression of endogenous isoprenoid production for efficient incorporation into FTase substrates. For the shorter C10AlkOPP analogue, incorporation was much lower (Figure 3, lanes 2 and 3) with minimal labeling in the absence of statin and discernably more in its presence. For this probe, labeling in the 40–75 kDa region was greater than in the 20–25 kDa range, suggesting that this analogue is a better mimic of FPP versus GGPP. That is consistent with the absence of GGTase reactivity observed with this probe in the in vitro enzyme assays. In contrast, the results obtained with the two longer compounds, C20AlkOPP and C15hAlkOPP, were the opposite (Figure 3, lanes 6–9). For both those probes there was minimal labeling in the 40–75 kDa region with more substantial labeling in the lower 20–25 kDa range suggesting that these analogues are not efficient substrates for FTase. However, the labeling obtained with C20AlkOPP was notably less than that observed with C15hAlkOPP suggesting there is an upper limit for what length can be tolerated by GGTases. This trend is again consistent with the results from in vitro enzyme assays where reactivity increased with chain length but decreased when the probe length exceeded the active site size as was observed in the docking experiments. Efforts to improve C20AlkOPP labeling using higher concentrations were not successful (Figure S11). Lovastatin had minimal effects on the labeling with these two analogues although its inclusion did lead to enhancement of one band near 75 kDa in the case of C15hAlkOPP suggesting that this probe does have a limited capacity to be utilized by FTase (Figure 3, lane 9). To explore the generality of the above results, similar metabolic labeling experiments were performed in immortalized astrocytes. The same patterns of labeling for C10AlkOPP, C15AlkOPP, C15hAlkOPP and C20AlkOPP were observed those cells (Figure S12). Taken together, a pattern of reactivity is apparent from these studies where C10AlkOPP labels primarily FTase substrates, C15AlkOPP labels a mixture, and C15hAlkOPP and C20AlkOPP label mainly GGTase substrates. While the C20AlkOPP analogue may be too large to be effectively transferred to many geranylgeranylated protein substrates, some prenylated substrates are still effectively labeled with this analogue. These results mirror the observations made in the in vitro kinetic analyses reported above.

Labeling results using the alcohol precursors were more complex. For the C10AlkOH and C15AlkOH analogues (Figure 3C, lanes 3–6), the labeling was much less efficient compared with the corresponding diphosphates (Figure 3B, Lanes 3–6). However, the longer C15hAlkOH appears to label with comparable efficiency as its diphosphate form (C15hAlkOPP) and better than C15AlkOH. The inclusion of lovastatin also appears to have some effects on the labeling in the 20–25 kDa region for both C15hAlkOH and C20AlkOH. It is important to note here that there are major biochemical differences between experiments employing the diphosphate versus alcohol forms of these probes. Currently, it is unclear how the diphosphates enter cells. Since they are trianions at physiological pH, it is difficult to envision that they simply diffuse across the hydrophobic plasma membrane. Instead, it is likely that a transporter is involved although none has been identified to date. In contrast, for free alcohols, passive diffusion across the membrane and into the cells is a reasonable hypothesis. However, once in the cell, they must be phosphorylated. Since the biosynthesis of FPP and GGPP involves smaller phosphorylated precursors and does not proceed via phosphorylation of farnesol or geranylgeraniol, it is unclear what enzymes are responsible for phosphorylating the probes described here and in other studies for subsequent metabolic incorporation.^26,28 Interestingly, an engineered system comprised of two kinases from bacteria has recently been developed to accomplish this transformation.²⁵

Previous work has shown that cells treated with C15AlkOP (the monophosphate form of the probe) incorporate the alkyne-modified isoprenoid into prenylated proteins as effectively as those treated with C15AlkOPP (the diphosphate form).⁴² Our assumption has been that the monophosphate form of the analogue was rapidly converted to the diphosphate that would be the species utilized by the prenyltransferases. That hypothesis is consistent with previous work performed by Fierke and coworkers who showed that farnesyl monophosphate (FMP) was at least 50-fold less reactive than FPP using in vitro enzyme assays with rFTase.⁵¹ To examine that question in more detail, the efficiency of C15AlkOP as a substrate for rFTase and rGGTase I was examined (Figure S13). Those experiments indicate that the monophosphate form of the probe is substantially slower than the diphosphate form suggesting that the former will not be effectively incorporated in metabolic labeling experiments without conversion to the corresponding diphosphate. In the context of the work reported here, these results indicate that both phosphorylation steps appear to be necessary for activation of the alcohols. Since the enzymes responsible for those transformations are likely to exhibit different efficiencies with different isoprenoid structures, the levels of the final diphosphates formed from various alcohol precursors will almost certainly not be constant across all analogues. This would likely cause different trends in metabolic labeling when comparing the results obtained using the diphosphates versus the alcohols.

Prenylomic identification of proteins labeled by C10AlkOPP, C15AlkOPP, C15hAlkOPP, C20AlkOPP and the corresponding precursor alcohols.

The aforementioned gel-based experiments provided some preliminary understanding concerning the selectivity of the isoprenoid analogues for different subtypes of prenylation. To obtain a more granular perspective, proteomic experiments were undertaken. Using the same metabolic labeling strategy described above, prenylated proteins in COS7 cells were enriched though the CuAAC reaction with biotin-azide (30) and subsequent pull down with neutravidin resin. This process was typically carried out using 3 samples obtained from cells treated with the alkyne probe and 3 samples treated with farnesol or farnesyl diphosphate where no alkyne labeling should occur. After on-bead trypsin digestion, labeling with TMT-6plex reagents for quantitation, multiplexing of the samples, and reversed-phase basic fractionation for decreasing sample complexity, an LC-MS analysis via a data dependent strategy using multinotch MS³ was performed. This strategy has been well explored with C15AlkOPP in multiple cell lines, primary cells, and mice to investigate various biological questions.^33–37 In this work, prenylomic data was gathered for COS7 cells, after a 6 hour lovastatin pretreatment, followed by probe treatment for 24 hours.

For analysis, the results of the two-student t-test of the TMT intensity, plotted as a volcano plot with an FDR = 1% (Figure 4B–E) allowed for identification of prenylated proteins labeled with the alkyne probe. For each probe, the composition of the labeled proteins was then parsed by prenylation subtype and displayed as FTase substrates (red), GGTase I substrates (blue), GGTase II substrates (green), and ALDH9A1, a previously reported self-prenylated protein (yellow).¹¹ When examining the analogue C15AlkOPP, as previously observed, a mixture of proteins from all three types of prenylation was detected (Figure 4B).

Figure 4.

Prenylomic methodology and proteomic identification of prenylated proteins labeled with C10AlkOPP or C15AlkOPP. (A) Schematic of prenylomic workflow. Lovastatin pretreated COS7 cells were treated with one of the alkyne probes and the corresponding control, either FPP or farnesol (FOH) for 24 h. Labeled proteins were enriched using a CuAAC reaction with azide-biotin followed by a neutravidin pulldown. After an on-bead digestion three replicates of the control labeled peptides and three alkyne analogue controls were labeled with TMT6Plex for quantification. (B-E) Volcano plots of prenylomic data for C15AlkOPP, C15AlkOH, C10AlkOPP, and C10AlkOH. Volcano plots were generated from a t-test analyzing the normalized intensity for reporter ions for each of the three replicates for each treatment, with an FDR =1% and s⁰=0.5. (B) Plot of C15AlkOPP compared to FPP. (C) Plot of C15AlkOH compared to FOH. (D) Plot of C10AlkOPP compared to FPP. (E) Plot of C10AlkOH compared to FPP. For all plots; red points are FTase substrates, blue points are GGTase I substrates, green points are GGTase II substrates, yellow points are the self-prenylating protein ALDH9A1, and gray points are proteins that are not known to be prenylated. Square points are proteins found with prenylomics for the first time. (F-G) Venn diagrams of comparisons of ungrouped enriched proteins of (F) C15AlkOPP and C15AlkOH or (G) C10AlkOPP and C10AlkOH. H) Correlation Log₂ fold change values for of proteins found both with C15AlkOPP and C10AlkOPP.

In the work reported here, 60 prenylated proteins, (70 proteins when protein groups were ungrouped) were enriched with C15AlkOPP including 25 farnesylated proteins, 12 type I geranylgeranylated proteins and 22 type II geranylgeranylated (Rab) proteins. Additionally, one enriched protein not previously found using a prenylomic method was identified: centrosomal protein of 85 kDa-like (CEP85L). The CEP85L protein was predicted to be farnesylated using PrePS, a prenylation prediction tool.⁵² This distribution of prenylation types (42% FTase substrates, 20% GGTase I substrates, 36% GGTase II substrates) was similar to previously reported results with this probe (39% FTase substrates, 21% GGTase I substrates, 38% GGTase II substrates),³⁵ further confirming that C15AlkOPP is a particularly versatile probe (Figure 3B).

In comparison to the results described above, proteomic analysis of cells labeled using the alcohol precursor, C15AlkOH, identified only 1 prenylated protein (Figure 4C); this observation is consistent with the in-gel fluorescence experiments reported above where much lower levels of labeling were detected (Figure 3C). That single protein, DNAJA2, was also enriched using C15AlkOPP (Figure 4F). While C15AlkOH was used in an earlier prenylomic analysis that employed a 2D differential gel electrophoresis (DIGE) strategy to identify 19 prenylated proteins not including DNAJA2,²⁹ that approach did not yield the number of prenylated protein substrates obtained here with the more streamlined bottom up prenylomic analysis.

Next, labeling using the shorter C10AlkOPP analogue was examined (Figure 4D). In that case, FTase substrates comprised the majority of the labeled proteins with only 3 geranylgeranylated proteins detected including CDC42, Rac1, and GNG12. Interestingly, even for the farnesylated proteins, the number detected with this shorter probe (5 proteins) was far smaller compared to the number found with C15AlkOPP (25 proteins). Moreover, all the proteins found with C10AlkOPP were also enriched with C15AlkOPP. A comparison of the Log₂ fold change for those shared proteins showed that there is no correlation between the two datasets, and all proteins have a higher fold change with C15AlkOPP (Figure 4H). This combined with the observation that C10AlkOPP labeled 20 fewer farnesylated proteins than C15AlkOPP illustrates that while C10AlkOPP is selective for farnesylation, it is not incorporated efficiently into the majority of farnesylated proteins.

Finally, no prenylated proteins were enriched with the alcohol precursor of this shorter C10AlkOH probe (Figure 4E and 4G). Collectively, those results are consistent with the in-gel fluorescence data described above (Figure 3) and suggest that the C10Alk scaffold is an inefficient substrate even for FTase under these metabolic labeling conditions. Additionally, the minimal labeling manifested by this shorter alcohol analogue indicates that it may be a poor substrate for the endogenous enzymes responsible for activation of the alcohol precursors to the metabolically competent diphosphates.

The properties of the two longer analogues, C15hAlkOPP and C20AlkOPP were next investigated in prenylomic experiments using the strategy described above. Since these analogues are closer to GGPP in their length, the use of both FPP and GGPP was examined as the control isoprenoid. For C15hAlkOPP, prenylomic analysis using GGPP as a control identified 33 enriched prenylated proteins, (41 proteins when ungrouped) including 8 farnesylated proteins, 8 type I geranylgeranylated proteins, and 16 Rab proteins as well as ALDH9A1 (Figure 5A and S14B). Using FPP as the control, slightly fewer proteins (36 proteins when ungrouped) were enriched including 9 farnesylated proteins, 10 type I geranylgeranylated proteins, and 17 Rab proteins as well as ALDH9A1 (Figure S14A–B). Overall, it appears that C15hAlkOPP labeled all types of prenylated proteins although the fraction of those that were farnesylated was smaller than what was obtained with C15AlkOPP.

Figure 5:

Prenylomic results for novel GGPP analogues and comparisons of C15hAlkOPP to C15AlkOPP. (A-D) Volcano plots generated from a t-test analysis of the normalized TMT reporter ion intestines with an FDR=1%, S⁰ =0.5. Volcano plots showing the fold change of proteins labeled with (A) C15hAlkOPP compared to GGPP, (B) C20AlkOPP compared to FPP, (C) C15hAlkOH compared to FOH. (D) C20AlkOH compared to FOH. (E-F) Venn diagrams showing the overlap of ungrouped enriched proteins found when comparing (E) C15hAlkOPP to C15hAlkOH or (F) C20AlkOPP to C20AlkOH. (G) Correlation of the Log₂ fold change of the proteins found both with C15hAlkOPP and C20AlkOPP. Rank spearman correlation was performed on the 6 proteins shared between the two datasets. The spearman correlation coefficient is −0.1, with a p-value 0.4. Solid black line illustrates perfect positive correlation (H) Venn diagram of all (left) ungrouped proteins found enriched in C15AlkOPP or C15hAlkOPP and a Venn diagram of only the farnesylated proteins found with C15AlkOPP or C15hAlkOPP. *ALDH9A1. (I) Correlation of 37 proteins found with both C15AlkOPP and C15hAlkOPP. The spearman correlation coefficient is −0.1, with a p-value = 0.4. The solid black line shows the ideal positive correlation.

In contrast, labeling with C20AlkOPP appears to be much more selective for geranylgeranylated proteins Prenylomic analysis using C20AlkOPP with FPP as a control yielded 1 farnesylated protein, 6 type I geranylgeranylated proteins, and 2 Rab proteins (Figure 5B) while the use of GGPP as a control led to the detection of only 2 type I geranylgeranylated proteins (Figures S14C–D). One of the geranylgeranylated proteins detected in the C20AlkOPP/FPP experiment was GNG13, a protein not previously found in any prenylomic experiment (Figure 5B). The underlying reasons for the different results using FPP versus GGPP as controls is unclear at this time and is worthy of future study. Unfortunately, while the selectivity of C20AlkOPP for labeling geranylgeranylated proteins was notable, that was offset by the relatively small number of proteins that were enriched using this probe (Figure 5B). Finally, as noted for the other probes, the number of proteins identified using the alcohol precursors of these longer probes, C15hAlkOH and C20AlkOH, was small (1 and 7 respectively, Figures 5C–D). However, curiously, the C20AlkOH probe identified the largest number of prenylated proteins of all the alcohol probes (Figure 5F and Table S3). That higher level of labeling may reflect greater efficiency of phosphorylation by endogenous kinases.

There were 6 prenylated proteins identified with both C15hAlkOPP and C20AlkOPP. Correlation analysis of the fold changes for these proteins showed no correlation while the rank-spearman correlation analysis was not statistically significant. The C15hAlkOPP analogue enriched all 6 proteins more than the C20AlkOPP did (Figure 5G). In fact, C15hAlkOPP functioned most like C15AlkOPP with an overlap of 37 prenylated proteins identified in both datasets (Figure 5H). However, when exclusively examining farnesylated proteins, there were 18 proteins found with C15AlkOPP that were not identified with C15hAlkOPP (Figure 5H), confirming that C15hAlkOPP, due to its increased length, is less effective as a substrate for FTase. In contrast, C15AlkOPP and C15hAlkOPP demonstrated similar efficacy with type I geranylgeranylation with 72% of the geranylgeranylated proteins found with both probes (Figure S15).

It is important to note that when comparing the Log₂ fold change for the 37 proteins that were found in both datasets, they were not correlated, although that determination is not statistically significant. However, it does illustrate that nearly all proteins were observed above the perfect positive correlation (identical enrichment) line in favor of C15AlkOPP (Figure 5I). This shows that those proteins were labeled more with C15AlkOPP compared to C15hAlkOPP. However, there were a few proteins including RHOG, RAB9A, and RAB11A;11B, that were found to be enriched more with C15hAlkOPP.

Comparisons of proteins identified and differences in CaaX sequences of enriched proteins.

The above experiments demonstrated that prenyltransferase selectivity can be modulated by varying analogue length. Only two proteins were found with all diphosphate analogues, CDC42 and DNAJA1 (Tables S1 and S2, Figure 6A). Beyond those, there are no other proteins enriched with both the shortest (C10AlkOPP) and longest (C20AlkOPP) probes. A more detailed picture of the length dependence can be obtained by inspection of the distribution of prenylation types where the proportion of farnesylated proteins found decreased with increasing probe length (Figure 6B). Also, of considerable importance, when comparing the enrichment of proteins with different probes relative to that obtained with C15AlkOPP, most proteins were enriched to the greatest extent with C15AlkOPP although there were some proteins that were labeled more so with other analogues (Figure 6C). For example, the fold-enrichment of Rab proteins was typically greater using C15hAlkOPP compared with C15AlkOPP. In both C15hAlkOPP datasets (using FPP or GGPP as controls) Rab9A, Rab11AB, Rab35, and Rab21 were enriched more relative to what was observed with C15AlkOPP. Those results are consistent with the greater catalytic efficiency exhibited by GGTase II for C15hAlkOPP compared with C15AlkOPP as was determined in in vitro kinetic experiments.

Figure 6.

Comparisons of proteins found enriched across all analogues. (A) Venn diagram of ungrouped proteins found enriched with all diphosphate probes. Probe length increases from left to right. (B) Tiered pie chart showing the distribution of prenylation types for the enriched proteins found with each analogue with the smallest probe, C10AlkOPP, in the center, to the largest analogue, C20AlkOPP, on the outside ring. (C) Examination of fold changes for all proteins found across the C10AlkOPP (teal triangle), C15AlkOPP (green diamond), C20AlkOPP (purple square), C20AlkOH (red X), C15hAlkOPP vs GGPP (blue circle), and C15hAlkOPP vs FPP (orange cross). Proteins are ranked from smallest to largest Log₂ fold change found with C15AlkOPP. Proteins not found with C15AlkOPP, were added into the ranking list at the necessary interval point. For example, GNG13 was not found in the C15AlkOPP dataset, so the fold change for GNG13 from the C20AlkOPP analysis was added as an entry in the total ranked list, indicated by a black line.

When comparing the CaaX sequences of farnesylated proteins enriched using the C10AlkOPP, C15AlkOPP, and C15hAlkOPP analogues, the most common residue found in the a1 position was valine (Figure 7 and S16). The C15AlkOPP and C15hAlkOPP analogues labeled proteins with a similar diversity in the amino acids following the modified cysteine, with slightly more variability in the a1, a2, and X positions compared to the farnesylated proteins identified with C10AlkOPP. This variance in C15hAlkOPP compared to C10AlkOPP is likely a reflection of the “alternative prenylation” phenomenon, where farnesylated proteins are geranylgeranylated when FTase is suppressed with specific inhibitors or statins. Given that 50% of the farnesylated proteins identified with C15hAlkOPP have demonstrated alternative prenylation,^10,53–56 the differences in CaaX sequences (Table S1), primarily in the a1 and a2 position which are more likely to be aliphatic residues in proteins enriched with C15hAlkOPP, further underscores the geranylgeranylation selectivity of C15hAlkOPP (Figure 7). Conversely, comparisons of the CaaX sequences for type I geranylgeranylated proteins found with each analogue demonstrated more similarity. For example, analysis of geranylgeranylated proteins found with C15AlkOPP, C15hAlkOPP, and C20AlkOPP all identified leucine be the most frequent residue in the a1, a2, and X positions (Figure S16E–G). Unsurprisingly C15hAlkOPP and C15AlkOPP identified geranylgeranylated proteins with nearly identical CaaX sequences given the significant overlap in identified GGTase I protein substrates.

Figure 7.

Sequence frequency logos for the CaaX motif for FTase protein substrates enriched with different analogues. The sequence frequency logos show the last 4 amino acids, including the modified cysteine for the proteins enriched with each probe. Letter size is related to the frequency of the residue in the submitted sequences. Enriched proteins were ungrouped and substrates of FTase and GGTase I were separated. Farnesylated sequences enriched with C15AlkOPP are shown on the left and with C15hAlkOPP on the right. Sequence frequency logos generated using WebLogo.^57,58

When comparing the sequences of the Rab proteins enriched by C15AlkOPP or C15hAlkOPP in both datasets (FPP or GGPP controls), a larger range from the C-terminus was examined since prenylation of Rab proteins can extend five residues upstream from the C-terminus. Additionally, because most Rab proteins undergo two modifications,⁵⁰ the perturbative impact of using chemically modified isoprenoids is potentially greater. When comparing the sequences of Rab proteins identified with C15hAlkOPP and C15AlkOPP, the former exhibits a preference for Cys at the 6 position whereas the latter has the highest preference for cysteine at the 10 position (Figure S17). Overall, these results confirm the idea that because the isoprenoid and protein substrates directly contact each other in the prenyltransferase active sites, changes in the probe structure may result in compensatory alterations in protein substrate specificity.

Functional studies with isoprenoid probes.

The studies described above highlight the utility of isoprenoid analogues containing alkyne functionality for identifying prenylated proteins. The next step in their application is to employ them in biological experiments aimed at measuring how the levels of prenylated proteins change in response to different drugs, developmental processes or disease states. However, as a prelude to such studies, it is important to determine whether the use of these analogues perturbs cellular physiology since that would make results obtained with them more difficult to interpret. Given that the C15AlkOPP analogue labeled the largest number of prenylated proteins, the functional studies performed here focused on that probe although several of the others discussed above were also investigated. Accordingly, several different types of experiments were conducted.

First, to obtain insight into how efficiently the C15AlkOPP probe was incorporated into farnesylated proteins in cellulo, the incorporation of that probe into H-Ras was evaluated using a using a gel shift assay. It is well established that the prenylated and unprenylated forms of H-Ras can be resolved by SDS PAGE and detected via immunoblotting.⁵⁹ Thus, COS7 cells were treated with lovastatin to inhibit prenylation and then supplemented with different isoprenoids to evaluate their ability to restore prenylation. One example of gel-based data is shown in Figure 8 and densitometric analysis from 3 replicates is shown in Figures S18 and S19. In these cells, in the absence of lovastatin (Figures 8, lanes 1, 7, 13 and 19), approximately 90% of the H-Ras protein was prenylated. Treatment with 10 μM lovastatin gave partial prenylation inhibition (54% prenylation) whereas treatment with 25 μM gave almost complete inhibition (37% prenylation). The conditions used to obtain the latter data are better for studying rescue since they provide a larger range. Using 25 μM statin, addition of 10 μM C15AlkOPP gave an increase from 37% to 66% prenylation (Figure 8, lane 3) suggesting that treatment with C15AlkOPP was able to substantially rescue H-Ras prenylation. In contrast, treatments with FPP were much less efficient at rescuing prenylation. No restoration was observed using 10 μM FPP (lane 14) or at higher concentrations (Figure S18). The behavior of the alcohols C15AlkOH and farnesol was more complex. For farnesol (Figure S18), some restoration was observed at 10 μM (54% prenylation at 10 μM lovastatin) and that increased to 75% at 100 μM. In contrast, for C15AlkOH (Figure S15), some rescue was observed using 10 μM analogue but the extent of rescue decreased at higher probe concentration suggesting that there was some toxicity. Overall, these experiments indicate that C15AlkOPP can substantially rescue statin-induced inhibition of H-Ras farnesylation suggesting that there is efficient incorporation of that probe under the conditions used for metabolic labeling (10 μM statin and 10 μM probe). These data also indicate that similar rescue does not occur to a significant extent, at least in this cell line, using the precursor alcohol of this probe (C15AlkOH) which is consistent with the metabolic labeling and proteomic experiments described above.

Figure 8:

Western blot analysis of H-Ras protein in cells subjected to statin-induced inhibition of prenylation and rescue with different isoprenoid diphosphates and alcohol precursors. COS-7 cells were treated with (A) 10 μM or (B) 25 μM lovastatin for 1.5 h (except for the sample indicated by -) followed by supplementation with different isoprenoids for an additional 18 h while retaining the statin. Cells were then lysed, the lysates fractionated via SD-PAGE, blotted to nitrocellulose membranes and visualized via western blotting followed by imaging.

Similar experiments were explored to study the prenylation of Rap1B, a geranylgeranylated protein. Interestingly, treatment of COS7 cells with 10 μM lovastatin had no effect on Rap1B prenylation while treatment with 25 μM statin resulted in only a small amount of inhibition. When the stain concentration was raised to 50 μM, substantial inhibition of prenylation was observed (Figure S20). Inclusion of 10 μM C15AlkOPP caused a significant amount of rescue while no rescue was observed with FPP. That latter result is expected since Rap1B is a geranylgeranylated protein and at 50 μM lovastatin it is unlikely that there is sufficient IPP present to be used by geranylgeranyl diphosphate synthase to extend FPP to the longer isoprenoid. Collectively, these gel-shift results demonstrate that C15AlkOPP can be efficiently incorporated into proteins that are normally modified with farnesyl or geranylgeranyl groups in cells.

Next, the impact of these probes on cellular localization was examined to serve as a functional readout of prenylation. It is well established that farnesylation of H-Ras causes the protein to localize to the plasma membrane. That can be inhibited by the addition of a statin that suppresses the production of FPP. Here, MDCK cells stably transfected with a GFP-H-Ras construct were used to monitor the localization of H-Ras via confocal microscopy.⁶⁰ In the absence of a statin, GFP-H-Ras localizes to the plasma membrane as evidenced by intense green fluorescence observed at the membrane (Figure S21A, 0 μM). Upon treatment with statin, the images appear radically different with green fluorescence distributed throughout the cell (Figure S21A, 10–100 μM). In the experiments reported here, the ability of different isoprenoid analogues to restore membrane localization was evaluated. While the metabolic labeling experiments described above were performed using 10 μM lovastatin, treatment of the MDCK cells used herein gave an intermediate phenotype with only partial inhibition of plasma membrane localization at that lovastatin concentration (Figure S21A, 10 μM). A lovastatin titration experiment showed that 30 μM lovastatin was necessary to obtain complete inhibition of plasma membrane localization (Figure S21A, 30 μM) and hence, subsequent experiments used that higher statin concentration. In those experiments, removal of the statin by changing the media led to a return to the normal phenotype exhibiting complete membrane localization as evaluated using colocalization with CellMask Orange within 1–3 h (Figure S21B). Hence the experiments reported here used an 18 h treatment with 30 μM lovastatin followed by the addition of isoprenoid (while retaining statin) for 6 h (Figure S22).

Initial experiments examined the ability of the naturally occurring isoprenoids FPP and GGPP to restore plasma membrane localization of GFP-H-Ras. Treatment with FPP did indeed restore membrane localization (Figure 9C) while similar treatment with GGPP did not (Figure 9E). That result was expected given that GGPP is not a substrate for FTase and H-Ras (a farnesylated protein) is not a substrate for GGTase I. Farnesol was also unable to restore membrane localization (Figure 9D). While that alcohol is known to undergo phosphorylation to yield FPP in cells, this observation suggests that under these conditions, FPP is not generated from farnesol at levels sufficient to prenylate most of the GFP-H-Ras within the cells. Next, the ability of the alkyne-containing analogues to restore membrane localization was studied. Treatment with C15AlkOPP resulted in substantial restoration of GFP-H-Ras plasma localization (Figure 9F) while treatment with the corresponding alcohol precursor, C15AlkOH did not (Figure 9G). Some restoration of localization was also observed with C15hAlkOPP (Figure 9H). Quantitative image analysis was performed on cells from each treatment to calculate the integrated GFP-H-Ras volume with a larger volume indicating mislocalization. That analysis revealed no statistically significant differences between treatments with FPP, C15AlkOPP and C15hAlkOPP (Figure S23). Taken together, these results suggest that the plasma membrane localization observed with GFP-H-Ras modified with C15AlkOPP or C15hAlkOPP is the same as that seen with the physiologically relevant farnesylated form. They also confirm the results observed in the metabolic labeling experiments where minimal labeling was obtained with the alcohol precursors of the probes highlighting why they are suboptimal for proteomic investigations.

Figure 9.

Recovery of plasma membrane localization of GFP-H-Ras in MDCK cells treated with lovastatin using various isoprenoids. All cells were treated with lovastatin (30 μM) for 24 h except the untreated sample. Where indicated, cells were supplemented with the isoprenoid indicated after 18 h of statin treatment and incubation continued for an additional 6 h. (A) Untreated cells; (B) Cells treated with statin only for 24 h; (C) Cells treated with statin and FPP (30 μM); (D) Cells treated with statin and FOH (30 μM); (E) Cells treated with statin and GGPP (30 μM); (F) Cells treated with statin and C15AlkOPP (30 μM); (G) Cells treated with statin and C15AlkOH (30 μM); (H) Cells treated with statin and C15hAlkOPP (30 μM). For each column, the GFP channel showing the GFP-H-Ras is at the top, the nuclear stain (Hoechst 34580) in the middle and the merged channels at the bottom.

The above experiments examined the efficiency of alkyne-containing isoprenoid probe incorporation in cellulo and the ability of proteins bearing modified isoprenoids to localize to the plasma membrane. To examine the biological activity of proteins modified with isoprenoid analogues, assays directed at macrophage function were employed. In those experiments RAW 264.7 cells were stimulated using the chemoattractant C5a resulting in a phenotype characterized by cellular polarization and numerous filipodia. Prior treatment with simvastatin inhibits both polarization and filipodia leading to the formation of smooth round cells. Previous work demonstrated that GGTIs cause the same effect as statins suggesting that geranylgeranylated proteins are involved in these complex cellular events.⁶¹ This has been attributed to dysfunction in cytoskeletal actin polymerization as a result of the loss in geranylgeranylation of small GTPases that regulate cytoskeletal remodeling.^62–66 That mechanistic hypothesis has been validated in experiments showing that the rounded cell phenotype resulting from statin inhibition could be rescued via the addition of GGPP. Here, the ability of isoprenoid analogues to rescue statin inhibition of polarization and filipodia formation was evaluated in a similar manner.

As noted above, cells treated with C5a manifest an elongated shape with numerous filipodia (Figure 10A) that were converted to a rounded phenotype upon the addition of simvastatin (Figure 10B). Cotreatment of those cells with 200 μM C15AlkOPP and simvastatin caused them to retain the differentiated phenotype displaying substantial polarization and filipodia (Figure 10C) suggesting that the alkyne-containing analogue was able to mimic the role of GGPP in this biological process. A quantitative analysis of the dose-response of both polarization (Figure 10E) and filopodia (Figure 10F) indicated that the analogue-induced rescue manifested complete recovery at 10 μM C15AlkOPP. That response is similar to what has previously been observed with GGPP (Figure S24A and S24B). Related experiments with C15hAlkOPP showed restoration of polarization at 10 μM (Figure S24C) but not filipodia at concentrations as high as 200 μM (Figure 10D and S24D) while C10AlkOPP rescued both polarization (Figure S24E) and filipodia (Figure S24F) but only at concentrations of 40 μM or greater. For the alcohol precursors, C15AlkOH was able to rescue polarization at 10 μM (Figure S24G) but not filipodia (Figure S24H), even at 200 μM with similar results obtained with C15hAlkOH (Figure S24I and S24J). C10AlkOH required higher concentrations (80 μM) for polarization rescue (Figure S24K) while again showing no filipodia (Figure S24L). Taken together, these observations illustrate that C15AlkOPP, of all the analogues tested here, most closely duplicates the behavior of GGPP in these experiments.

Figure 10:

Macrophage morphology obtained by C5a stimulation is inhibited with simvastatin and restored with C15AlkOPP. Confocal imaging experiments using RAW 264.7 cells stained with phalloidin. (A) Cells stimulated with C5a after treatment with vehicle; (B) Cells stimulated with C5a after treatment with statin; (C) Cells stimulated with C5a after treatment with statin and C15AlkOPP (200 μM). (D) Cells stimulated with C5a after treatment with statin and C15hAlkOPP (200 μM). Quantitative image analysis of confocal images including polarization (cell aspect ratio) and filopodia (number) as a function of C15AlkOPP concentration. Cells were treated with lovastatin (except for the sample indicated by - where vehicle was used) along with different isoprenoids at the concentrations indicated for 24 h followed by stimulation with C5a. (E) Cell polarization in the presence of C15AlkOPP; (F) Number of filopodia observed in the presence of C15AlkOPP. One-way analysis of variance (ANOVA) was used to evaluate the group differences with one independent variable and more than two experimental groups. Statistical confidence in the difference between a given sample and the untreated control sample (no statin, no isoprenoid) is indicated by (*).

CONCLUSIONS

In this report, a series of isoprenoid analogues incorporating biorthogonal alkyne groups was prepared and studied. The experiments described here reveal that these probes exhibit a wide range of behaviors. The shortest compound, C10AlkOPP, while highly specific for farnesylation, failed to label numerous farnesylated proteins modified by the probes intermediate in length. Conversely, while highly selective for geranylgeranylated proteins, the longest compound, C20AlkOPP, did not label many of such proteins that were successfully modified using the intermediate length probes. The C15AlkOPP and C15hAlkOPP analogues were the most versatile compounds, labeling prenylated proteins of all types with the C15AlkOPP probe labeling the largest number of proteins (70) in this study and C15hAlkOPP manifesting some selectivity for labeling geranylgeranylated proteins. Minimal labeling was observed with the alcohol precursors, emphasizing the better performance of the diphosphate forms of the analogues. Functional studies including gel-shift experiments, cellular localization and phenotypic complementation indicated that the C15AlkOPP analogue is particularly well suited for metabolic labeling experiments due to its efficient incorporation, ability to correctly localize prenylated proteins and restore biological function in response to isoprenoid depletion. Collectively, these results indicate that C15AlkOPP is a biologically functional, universal probe for metabolic labeling experiments that has minimal effects on cellular physiology. Consequently, it should be quite useful for experiments aimed at studying the levels of prenylated proteins in a variety of disease models.

MATERIALS AND METHODS

Materials.

Purchased reagents were used without further purification. Immortalized astrocytes were obtained from Dr. G. W. Rebeck at Georgetown University. COS7 cells were purchased from ATCC. For proteomic sample preparation LC-MS grade water was used to prepare all buffers.

Docking analysis of C10AlkOPP, C15AlkOPP, C15hAlkOPP, and the C20AlkOPP analogue:

The structures of (7, 5, 9, 11) in their di-anionic forms were created in ChemDraw, saved as MDL files, imported into Maestro (Schrödinger), and prepared for docking using the OPLS3 force field with a target pH of 7.0 ± 2.0. Similarly, the PDB file 1JCR (rFTase), 1TNO (GGTase I), 3DST (GGTase II) containing the coordinates of the different enzymes and substrates was prepared for docking in Maestro with a target pH of 7.0 ± 2.0 and then the structures were optimized by removing any water molecules with less than 3 H-bonds to non-waters and minimized using the OPLS3 force field. Around the bound FPP or GGPP, a receptor grid was created with a scaling factor of 1.0 and a partial charge cutoff of 0.25. Glide was then used to dock the analogues into the protein using the extra precision mode with flexible ligand sampling and the top 20 poses recorded in the output file and viewed in Pymol. Each analogue was docked independently into each enzyme.

Steady-state kinetic characterization of alkyne analogues using FTase and GGTase I.

The reactivity of analogues C15hAlkOPP and C20AlkOPP were evaluated with rFTase and rGGTase-I using dansylated peptides. The peptide substrates Ds-GCVLS, for FTase, and Ds-GCVLL, for GGTase-I, were employed using a previously reported fluorescence-based prenylation assay in 96-well black low adhesion plates.⁶⁷ In this procedure 20 nM enzyme was incubated with analogues (0.3 – 10 μM) in 1X reaction buffer (50 mM HEPPSO-NaOH pH 7.8, 5 mM TCEP, and 5 mM MgCl₂) for twenty minutes. Reaction were initiated through the addition of 3 μM peptide substrate after preincubation in 1X reaction buffer (RT, 20 min). Fluorescence was measured using a BioTek H1 Synergy plate reader (λ_ex= 340nm and λ_em= 520nm) over time courses from 30 min to 3 h as appropriate. To convert the fluorescence signal to concentration (Fl/ μM product) using a correction factor that was determined by total change in fluorescence. Initial velocities (μM /sec) were determined in the initial linear range of each reaction. Then, the initial velocities were divided by enzyme concentration and plotted against analogue concentration to generate V/E vs S plots. These data were fit to a Michaelis-Menten model to determine k_cat and K_m values using KaleidaGraph for curve fitting (Synergy Software, Reading, PA).

Kinetic analysis of alkyne analogues using GGTase II.

Constructs pGATEV-RabGGTaseAlpha, pET30 RabGGTase Beta, pOPINM_REP1_DanioRerio and pIF-Rab7SSSC were generous gifts from Dr. Kiril Alexandrov (University of New South Wales). Thes plasmids were transformed into E. coli and the expressed proteins purified as previously described.^68,69 In vitro prenylation reactions (100 μL) were performed in 25mM HEPES pH 7.2, 24 mM NaCl and 5mM DTT containing pGGTase II, REP and Rab7 proteins (6 μM each) and initiated by the addition of the isoprenoid diphosphate (0–20 μM). Reactions were allowed to proceed for various times and then terminated by precipitation using a ProteoExtract Protein Precipitation Kit. CuAAC was performed on the protein precipitate by adding the following to each tube: 92.5 μL of PBS containing 1% SDS, 2.5 μL of 1 mM TAMRA, 2 μL of 50 mM TCEP, 1 μL of 10 μM TBTA, and 2 μL of 50 mM CuSO₄. The reaction mixtures were incubated in the dark at rt for 1 h. A 10 μL aliquot from the reaction was diluted with 4X SDS gel loading buffer and heated at 95 °C for 10 min followed by fractionation via a 12% SDS-PAGE gel. A fluorescence scan was performed to detect TAMRA fluorescence and the gel was stained and destained with Coomassie blue to visualize protein loading. Band intensites were quantified using ImageJ.

In-gel fluorescence analysis.

Cell pellets were lysed in lysis buffer (300 μL PBS containing 1% SDS, with the addition of 0.64 μL 50 mM PMSF, 5 μL of protease inhibitor cocktail, and 0.26 μL of benzonuclease per samples). Each sample was sonicated (pulse seven times for 1 sec, at an intensity of 7, with rest for 10 sec between pulses (60 Sonic Dismembrator Fischer Scientific)) on ice. To determine protein concentrations, BCA assays (BioRad) were used. Lysates were diluted to 1.0 μg/μL total protein for click reactions, with 100 ug of protein in each sample. Next the click reactions were carried out by sequentially adding TAMRA-PEG3-N₃ (36, 2.5 μL of 1 mM stock solution in DMSO, 25 μM final concentration, BroadPharm), tris(2carboxyethyl)phosphine (TCEP, 2 μL of 50 mM stock solution in DMSO, 1 mM, final concentration Sigma-Aldrich), tris(1-benzyl-1H-1,2,3-triazol-4-yl-)methylamine (TBTA, 1 μL of 10 mM stock solution in DMSO 0.1 mM final concentration, Sigma-Aldrich), and CuSO₄ (2 μL of 50 mM stock solution in LC-MS grade water, 1 mM final concentration, Sigma-Aldrich). After reacting for 2 h at rt proteins were precipitated using a ProteoExtract protein extraction kit (MilliporeSigma) following the manufacturer’s protocol. Protein pellets were washed, air-dried and frozen. For analysis, samples were thawed, dissolved in Laemmli buffer (2% SDS, 10% glycerol, 62.5 M Tris HCl, pH 6.8, 0.01% bromophenol blue) and heated to 100 °C for 5 min, followed by fractionation using 12% SDS-PAGE gels run at 120 V for 120 mins.

Enrichment of proteins.

Cell lysate (2 – 2.4 mg protein) from each sample was subjected to a CuAAC click biotinylation reaction by sequentially adding 100 μM, biotin-Peg3-N₃ (37, 100 μM, final concentration, BroadPharm), TCEP (1 mM final concentration), TBTA (0.1 mM, final concentration), and CuSO₄ (1 mM, final concentration). All samples were then diluted to 1 mL with PBS containing 1% SDS, followed by a 2 h incubation at rt. Next, proteins were precipitated using a CHCl₃/CH₃OH/H₂O (1:4:3) extraction. To isolate the protein pellets, the samples were centrifuged at 4,000 xg for 7 min, washed with CH₃OH and then air dried. Pellets were stored at −20 °C overnight. Any remaining liquid was removed via aspiration and then pellets were solubilized in 0.55 mL PBS containing 1% SDS. To achieve full solubilization, samples were rotated at rt for 30 mins. Protein concentrations were measured using BCA assays.

Neutravidin agarose beads (200 μL, 50% slurry, Thermo Scientific) were washed with 1 mL PBS/1% SDS, then vortexed and briefly centrifuged. The beads were then allowed to rest for 2 mins. The overlaying liquid was then removed without disturbing the resin. This entire process was repeated once. Then biotinylated protein samples were normalized to 1–1.2 mg/mL in 0.5 mL of PBS/1% SDS then incubated with the pre-washed resin for 120 min. Beads were washed using the same method described above with three washes with PBS/1% SDS (1 mL each time) followed by one wash with PBS (1 mL). Next, three washes with 8 M urea in 50 mM triethylammonium bicarbonate (TEAB) buffer (1 mL each time) with 10 min incubations between washes were performed. Then to prepare the beads for digestion, three washes with 50 mM TEAB buffer (1 mL each time) were performed. Next, the resin was resuspended in 100 μL of 50 mM TEAB buffer, and the retained proteins were digested on-bead with trypsin (1.5 μg, (0.25 μg/μL stock concentration) in provided buffer, Promega Corp.) overnight at 37 °C. The digestion was quenched by adding 20 % HCO₂H (2.5 μL) in H₂O and then incubated are rt for 15 min. The resin was then loaded into Pierce spin columns and the centrifuged to remove the resin and collect the peptides, the resin was washed with 100 μL of 0.5% HCO₂H in H₂O and then 100 μL of 30% CH₃CN in H₂O. The collected peptides and washes were pooled and dried via lyophilization.

Isobaric labeling of peptides and proteomic preparation.

Dried peptide samples were redissolved in 50–60 μL of 100 mM TEAB buffer and their concentration determined via BCA assay. Then, 10 ug of protein was added to low-bind Eppendorf tubes, supplemented with 150 fmol of internal standard (yeast ADH1, Waters) and subjected to tandem mass tag (TMT)-labeling using TMT 6-plex reagents (Thermo Fisher Scientific). TMT 126–128 were used for the control samples (typically FPP) and TMT 129–131 were used for the alkyne analogue-labeled samples. The TMT reactions were rotated at rt for 180 min. The reactions were quenched by the addition of 0.5% hydroxylamine (2.5 μL) followed by a 15 min incubation. Then, the TMT-labeled samples were combined, dried with lyophilization, and dissolved in 200 mM NH₄HCO₂ (pH 10, 300 μL). Multiplexed samples were subsequently fractionated using a homemade stage tip (three layers of SDB-XC extraction disks, (3M, 1.07 mm × 0.50 mm i.d.)) in a 200 μL pipette tip. Peptides were fractionated under high pH reverse phase conditions yielding 7 fractions of 60 μL in volume (5, 10, 15, 20, 22.5, 27.5, 80% CH₃CN in 200 mM NH₄HCO₃, pH 10). The first two fractions were combined, then all fractions were dried by lyophilization. Samples were then dissolved in 30 μL of 0.1% HCO₂H for LC-MS³ analysis.

LC-MS data acquisition.

TMT-labeled peptides were resolved with flow rate of 300 nL/min using a RSLC Ultimate 3000 nano-UHPLC (Dionex) with a reversed-phase column (75 μm i.d., 45 cm) packed in-house. Each fraction from the high pH prefractionation was subjected to varying gradients (7–34 %) of Buffer B (CH₃CN containing 0.1 % HCO₂H) and Buffer A (H₂O containing 0.1% HCO₂H) over 80 min and sprayed directly into the Orbitrap instrument (Thermo Fisher Scientific). MS1 scans were collected at 120,000 resolutions in a 320–2,000 m/z range with 100 ms max injection time (IT) and automatic gain control (AGC) target of 200,000. The subsequent data-dependent MS/MS scans were collected with collision-induced dissociation (CID) at a normalized collision energy (NCE) of 35% with a 1.3 m/z isolation window, with max IT of 100 ms and AGC target of 5000. Acquisition in MS3 was done by selecting the top 10 precursors (SPS) for fragmentation by high-collisional energy dissociation (HCD) in the orbitrap with the following settings: 55% NCE, 2.5 m/z isolation window, 120 ms max IT, and 50,000 AGC target.

Prenylomic Data processing.

Raw MS3 data files were uploaded into MaxQuant (version 1.6.17.0) and searched against a non-redundant human database (UP0000000589) from Uniprot. For all but the following parameters, the default settings were maintained. Trypsin/P was used for digestion with allowance for 3 missed cleavages, minimum peptide length was set to 7, protein FDR was set to 0.5, modifications in search oxidation (M) and Acetyl (protein N-Term), and unique and razor peptides were used for quantification. MaxQuant was run through the MaxQuantCmd.exc at the University of Minnesota Supercomputing Institute. The proteingroup.txt file generated was uploaded into Perseus (version 1.6.14.0). Proteins that were potential contaminants and reverse peptides were removed by site identification. Raw intensities were transformed to log₂ values, and proteins with more than 3 out of 6 values returning “NaN” after transformation were removed. Missing values were imputed from the normal distribution of the remaining values. Reported values (TMT) were normalized by subtracting rows by mean value and columns by median value. Statistical analysis was performed using a two-sample t-test FDR = 1.0% and s0 = 0.1. Data was exported to excel for generation of plots and figures. For cross dataset comparisons protein groups were ungrouped. When two or more proteins were possible identities from the same spectra, they were grouped together. Whether a protein was grouped together was not consistent across datasets so for cross dataset analysis grouped proteins were separated into independent entries with the same data values.

Western blot analysis of cells subjected to statin-induced inhibition.

The COS-7 African Green Monkey kidney cell line was obtained from the American Type Culture Collection. Lovastatin was purchased from Tocris Bioscience and prepared as a 25 mM stock in DMSO. Farnesyl pyrophosphate (FPP) was obtained from Sigma-Aldrich and prepared as a 50 mM stock in 25 mM NH₄HCO₃. Farnesol (FOH) was purchased from Sigma-Aldrich and prepared as a 50 mM stock in DMSO. Rabbit-anti-H-Ras antibody was from Protein Tech, and rabbit anti-Rap1B antibody was from Cell Signaling. COS-7 cells were plated in 20 × 35 mm tissue culture plates at a concentration of 2 × 10⁵ cells/plate in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% heat-inactivated fetal bovine serum and antibiotics. The cells were cultured overnight at 37°C, 5% CO₂ and then the media in different plates was supplemented with different concentrations of lovastatin (as indicated in the figures) or with 0.2% DMSO. After 90 minutes, the media in different plates was supplemented with different concentrations of C15AlkOPP, C15AlkOH, FPP, or FOH (as indicated in the figures). After culturing overnight at 37°C, 5% CO₂, the cells in each plate were collected by scraping into ice-cold TX-100 buffer (1% Triton-X100, 150 mM NaCl, 10 mM Tris, pH 7.4) containing benzonase [50 U/ml] and PMSF [100 mM]. The collected cell suspensions were mixed with an equal volume of SDS buffer (2% SDS, 50 mM Tris, pH 7.4), rocked for 15 minutes at 4C, and centrifuged. The supernatant was collected, and an aliquot of each supernatant was used to determine the total protein concentration using the BCA protein assay. An aliquot of each supernatant was mixed in 5X SDS loading buffer to make a suspension containing 10 mg protein. These suspensions were subjected to SDS-PAGE and immunoblotting using primary antibodies to H-Ras or Rap1B, and the bound primary antibodies in the immunoblots were visualized using horseradish peroxidase-linked secondary antibodies (GE Healthcare). Images of the immunoblots were obtained using an ImageQuant LAS4000 Biomolecular Imager, as previously described⁷⁰.

Analysis of GFP-H-Ras in MDCK cells.

MDCK cells expressing eGFP-H-Ras were plated at 200,000 cells per 35 mm plate then incubated plates for 24 h at 37 °C, 5 % CO₂ before treating cells with 30 μM lovastatin for 18 h. After lovastatin pretreatment, 30 μM of each isoprenoid was added to each plate, then incubated for 6 h. Note the media was not changed after lovastatin pretreatment, thus the isoprenoid analogue was added to media containing lovastatin. Then, Hoechst 34580 was added to each plate (1 μL of a 2 μg/μL solution in DMSO, final concentration of 2 μg/mL) and imaged using a Zeiss Cell Observer spinning disk confocal microscope (Axio Observer). The microscope environmental chamber was set to 37 °C and 5 % CO_2. The images were obtained using a 63 X water immersion objective, with a 1.2 NA, c-apochromat aberration correction, infinity correction with 0.17 mm cover glass, a working distance of 0.28 mm and a Photometrics Quant EM 512SC-high sensitivity 16 bit, 512 × 512 grayscale CCD detector. For image analysis, an illuminated corrected image was generated in Cell Profiler (Broad institute, v 4.2.7) by applying a fit polynomial smoothing method (block size of 60 pixels) and normalizing each histogram for each image. Then, a mask image was created by using an adaptive sauvola thresholding method (smoothing scale 1.0, correction factor 1.0, adaptive window of 100 pixels). The mask image was used to select and measure the GFP area and integrated density of the illumination corrected image (Image J, FIJI v 2.9.0). To measure the number and area of nuclei in each image, the background was subtracted using a rolling ball radius of 50 pixels and applying a median filter with a radius of 5.0 pixels on each nuclei image using Image J. Next, the nuclei were selected in Cell Profiler by thresholding with an adaptive strategy and two class Otsu method, smoothing scale of 1 and correction factor of 0.8. Clumped nuclei were divided into individual objects based on shape and distance between the local maxima of each nuclei. Then, the number of nuclei and area of nuclei were recorded for each image. Objects less than 40 pixels or greater than 500 pixels were discarded to ensure only intact nuclei were measured. Finally, the GFP integrated density was divided by the nuclei area to quantify the amount of GFP-H-Ras PM localization. A smaller value indicates more plasma membrane localization and a larger value indicates more cytosolic localization.

Macrophage morphology analysis.

RAW 264.7 cells (ATCC, TIB-71) were plated at a density of 3×10⁴ cells/mL in 24-well plates (Corning, Cat. No. 09–761-146) on top of coverslips using complete media (DMEM, 5% (v/v) FBS, and 1% penicillin/streptomycin) containing simvastatin (20 μM) and given concentrations of probe for 24 h (37 °C with 95% humidity and 5% CO₂). Following that treatment, cells were stimulated with 100 nM C5a (R&D Biosystems Cat. No. 2150-C5–025) for 10 min, washed with PBS, then fixed with PBS containing 4% paraformaldehyde for 20 min. Subsequently, the cells were permeabilized with 0.1% Triton X-100, stained with 0.5 μL/well (0.5:200) AF-488 phalloidin and DAPI (1 μL/well), then washed. The cover slips were then mounted on to microscope slides for the imaging analysis using an Olympus BX51 microscope, under oil emersion at 60X magnification, for both filopodia and polarization analysis. One experimenter had the role of taking the images and another trained observer was responsible for analyzing the images. Both experimenters were blinded to experimental condition. For each condition, polarization data were generated by measuring eight cells per well, across a minimum of four fields of view. For polarization analysis, the length and width of each cell was measured using ImageJ software. Using these measurements, the length-to-width ratio was determined and reported as the polarization. Two cells per well were averaged for the filopodia data collection. A sample size of four independent wells was used for each treatment group in each experiment. Filopodia, defined as clear filamentous protrusions extending from the membrane of the cell, were manually quantified. One-way analysis of variance (ANOVA) was used to evaluate the group differences with one independent variable and more than two experimental groups.

Data Availability

The proteomic data has been deposited in the PRIDE repository using the proteome Xchange tool with the ascension number PXD055402.