Evaluating protein prenylation of human and viral CaaX sequences using a humanized yeast system

Post-translational modification of the C-terminal CaaX sequence (herein referring to all C-terminal tetrapeptide sequences regardless of whether they fit the traditional consensus sequence) is important for regulating the localization and function of many proteins, including the Ras GTPases often cited as archetypical CaaX proteins (Campbell and Philips, 2021; Cox et al., 2015; Ravishankar et al., 2023). Ras oncogenic mutants are associated with 90-95% of certain cancers (https://www.cancer.gov/research/key-initiatives/ras), and oncogenic phenotypes are altered when CaaX modification is disrupted (Bergo et al., 2002; Cox et al., 2015; Sousa et al., 2008; Winter-Vann and Casey, 2005). In addition to cancer, CaaX proteins are functionally implicated in numerous other disease states including Alzheimer's disease, progeria, liver disease, and viral infections such as hepatitis D and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) (Jeong et al., 2022; Marakasova et al., 2017; Soveg et al., 2021; Wickenhagen et al., 2021; Yang et al., 2012; Zhao et al., 2020). A recent example demonstrating the importance of protein prenylation for viral infection resulted from studies of OAS1 gene splice variants, which concluded that humans expressing a prenylated variant of the Oas1 protein had higher innate antiviral protection against Sars-CoV-2 than those expressing the non-prenylated protein (Soveg et al., 2021; Wickenhagen et al., 2021). With regard to Alzheimer's disease, the CaaX protein Ydj1 (yeast ortholog of human DNAJA1) has been demonstrated to directly interact with amyloid β42 oligomers and drive amyloid β42 toxicity in model organisms (Ring et al., 2022). Defining the full spectrum of prenylated proteins is thus crucial for understanding the impacts of pharmacological interventions that globally alter prenylation patterns and lead to a combined effect arising from altering the prenylation of many CaaX proteins simultaneously.

The C-terminal CaaX sequence has been traditionally defined as a cysteine (C), two aliphatic amino acids (a₁a₂), and one of several amino acids (X) (reviewed in Jung and Bachmann, 2023; Wang and Casey, 2016). The first step in CaaX modification is covalent attachment of a farnesyl (C15) or geranylgeranyl (C20) isoprene lipid to the cysteine. Farnesyltransferase (FTase) generally targets CaaX sequences, whereas geranylgeranyltransferase type I (GGTase-I) targets the subset of CaaL/I/M sequences (Hartman et al., 2005). Some CaaX sequences are modified by both enzymes, albeit one is typically preferred (Caplin et al., 1998; Fiordalisi et al., 2003; Moores et al., 1991; Trueblood et al., 1993). CaaX sequences with an aliphatic amino acid at the a₂ position commonly follow the canonical pathway that involves two additional modification steps: proteolytic removal of aaX and carboxymethylation of the exposed prenylated cysteine (Fig. 1A). In yeast, the Ras2 GTPase and a-factor mating pheromone are well-characterized examples of canonically modified CaaX proteins. Conversely, CaaX sequences with non-aliphatic residues at a₁ and a₂ can be farnesylated but resist proteolysis, resulting in a biochemically distinct C-terminus compared to that of canonically modified CaaX proteins (Kim et al., 2023). These so-called shunted CaaX sequences are prevalent and occur throughout biology, with the yeast HSP40 Ydj1 protein being the best-characterized example to date (Hildebrandt et al., 2016b).

FTase and GGTase-I are heterodimeric enzymes. They share an α subunit, whereas the β subunits provide isoprenoid specificity and substrate recognition (Lane and Beese, 2006; Maurer-Stroh et al., 2003). Human FTase [Homo sapiens (Hs)FTase] is composed of HsFNTA(α) and HsFNTB(β) subunits; the orthologous subunits of yeast FTase [Saccharomyces cerevisiae (Sc)FTase] are ScRam2(α) and ScRam1(β) (Fig. 1B). Human GGTase-I (HsGGTase-I) is composed of HsFNTA(α) and HsPGGT1B(β) subunits, whereas the orthologous subunits of yeast GGTase-I (ScGGTase-I) are ScRam2(α) and ScCdc43(β). In yeast, RAM1 is not an essential gene (i.e. ram1Δ strains are viable), whereas RAM2 and CDC43 are essential, raising the possibility that GGTase-I is more critical to yeast life processes relative to FTase (He et al., 1991; Ohya et al., 1993). In this study, we describe the manipulation of the yeast genome to allow for expression of human FTase and GGTase-I in lieu of their yeast counterparts.

Using the broadest definition of a CaaX sequence (i.e. cysteine followed by any three amino acids), 1207 human proteins can be identified within UniProtKB/Swiss-Prot. These proteins use 680 of the 8000 possible CaaX sequence combinations. There are also 375 viral proteins with a CaaX sequence that could potentially utilize host enzymes for CaaX modification (Fig. 1C). Resolving the prenylation potential of candidate CaaX sequences has depended on a variety of methods and prediction algorithms, each with its own limitations. Case-by-case methods for verification of protein prenylation include radiolabeling with ³H-mevalonate (i.e. the farnesyl and geranylgeranyl precursor) and analyses based on gel mobility, localization and mass spectrometry (Anderegg et al., 1988; Hancock et al., 1989; Hildebrandt et al., 2016b; Michaelson et al., 2005; Ravishankar et al., 2023). Methods for systematic verification have relied on metabolic labeling with reagents compatible with click chemistry, peptide arrays and genetic screens (Kho et al., 2004; Kim et al., 2023; Rashidian et al., 2013; Storck et al., 2019; Suazo et al., 2018, 2021; Wang et al., 2014). Differentiating FTase and GGTase-I target specificity has mostly relied on in vitro and genetic approaches (Hougland et al., 2010; Kim et al., 2023; Stein et al., 2015). Independent of the methods used, it remains a distinct challenge to confirm the prenylation status of candidate CaaX sequences, especially in organisms in which prenyltransferase specificities have never been investigated or studied in depth.

In this study, we developed yeast strains expressing HsFTase and HsGGTase-I that functionally replace the yeast enzymes in a wide array of tests. Results with a universal reporter for both farnesylation and geranylgeranylation indicate that human prenyltransferases target a wide variety of human and viral CaaX sequences, with human FTase having an ability to modify non-canonical sequences like its yeast counterpart. We also demonstrate the utility of this model system for evaluating the prenylation of heterologously expressed human proteins. This system provides a valuable new resource for protein prenylation research in a genetically amenable cell-based system. This resource is expected to enhance the identification and characterization of previously unreported prenyltransferase targets, including those harboring atypical CaaX sequences.

As a key first step toward developing a yeast system to express HsFTase, complementation studies were performed to assess the functional equivalence of yeast and human FTase subunits. Codon-optimized genes encoding HsFNTA(α) and HsFNTB(β) were introduced into yeast, separately or in combination, and evaluated for their ability to complement for loss of the yeast FTase β subunit Ram1 (i.e. ram1Δ). For this evaluation, we used a highly sensitive and well-characterized biological assay (i.e. yeast mating) that detects production of the farnesylated a-factor mating pheromone (Hildebrandt et al., 2016b; Nijbroek and Michaelis, 1998). There is a direct correlation between a-factor levels and mating efficiency, and the high-sensitivity assay can quantifiably measure a-factor levels as low as 0.001% relative to normal levels. Co-expression of both HsFNTA(α) and HsFNTB(β) as genome-integrated genes resulted in a mating phenotype that was qualitatively and quantifiably indistinguishable from mating exhibited by wildtype yeast, indicative of normal a-factor production (Fig. 2A,B). Similar observations were made with plasmid-encoded genes (Fig. S1A,B, lanes 2 and 5). Importantly, despite observations that human proteins can often substitute for their yeast counterparts, HsFNTB(β) expressed in yeast was unable to support a-factor production (i.e. mating), indicative of no prenyltransferase activity (Fig. S1A,B, lane 3). This result indicates either that ScRam2(α) and HsFNTB(β) subunits do not interact or that a formed hybrid FTase complex [ScRam2(α)-HsFNTB(β)] is non-functional. During development of these strains, differences in mating efficiency were also observed depending on the promoters used to express the subunits (Fig. S1C). This observation ultimately directed us to integrate human FTase subunits into the yeast genome using the constitutive phosphoglycerate kinase promoter (i.e. P_PGK1) rather than orthologous yeast promoters.

As additional phenotypic confirmation of human FTase activity in yeast, we observed that the temperature-sensitive growth phenotype of the parental FTase-deficient strain (i.e. ram1Δ) had been alleviated in the humanized FTase strain (Fig. 2C). To further confirm the ability of yeast-expressed HsFTase to prenylate other CaaX proteins, gel-shift analysis was used to determine the extent of farnesylation for yeast Ydj1, yeast Ras2 and a human oncogenic variant of HRas (Fig. 2D; Fig. S1D). The strains were used directly to prepare cell lysates containing endogenous levels of Ydj1 and Ras2, or they were transformed with a yeast expression plasmid encoding Myc-tagged HRas^Q61L to obtain an HRas-containing lysate. A gel shift was observed for all three proteins when ScFTase or HsFTase was present; the faint upper bands associated with farnesylated Ras2 are occasionally observed and are of unknown origin. Each protein exhibited faster gel mobility than that of its unfarnesylated form. The reason for this counterintuitive but common observation remains unknown. Not all prenylated proteins demonstrate a gel shift upon prenylation, and the direction of the shift can vary for different proteins. Nonetheless, these complementation studies indicate that HsFTase engineered into yeast can support the normal prenylation of both yeast and human CaaX proteins.

Our complementation studies revealed that HsFTase can modify a variety of yeast and human CaaX proteins: a-factor (with the CaaX sequence CVIA), Ras2 (CIIS), Ydj1 (CASQ) and HRas (CVLS). We next used the humanized HsFTase yeast system, a GFP-Ras2 reporter and gel-shift analysis to investigate farnesylation of human Ras CaaX sequences: KRas4A KRas4A (CIIM) and KRas4B (CVIM) (both encoded by KRAS), HRas (CVLS) and NRas (CVVM) (Fig. 3A). In the presence of HsFTase, all the GFP-Ras2-CaaX variants exhibited faster mobility relative to that of a negative control (i.e. GFP-Ras2-SIIS) that cannot be prenylated due to the absence of cysteine within its CaaX sequence. Moreover, the GFP-Ras2-CaaX variants based on HRas (CVLS) and NRas (CVVM) did not shift in the absence of FTase (Δ), whereas those based on KRas4A (CIIM) and KRas4B (CVIM) still shifted. In mammalian systems, some CaaX sequences are prenylated by HsGGTase-I when HsFTase activity is diminished (Mohammed et al., 2016; Whyte et al., 1997). We thus infer that alternate geranylgeranylation occurs for CIIM and CVIM sequences in FTase-deficient yeast that still express endogenous ScGGTase-I.

In yeast and humans, Ras localization to the plasma membrane depends on canonical modification of the CaaX sequence (Boyartchuk et al., 1997; Michaelson et al., 2005; Ravishankar et al., 2023). Bolstering our conclusion that alternate prenylation occurred for GFP-Ras2-CIIM and -CVIM, these proteins exhibited plasma membrane localization in the absence of FTase, which would be predicted if they were geranylgeranylated (Fig. 3B). Importantly, GFP-Ras2-CIIS was only membrane localized in the presence of FTase (either yeast or human), consistent with observations that it is not subject to alternate prenylation, and this localization was dependent on the presence of cysteine within its CaaX sequence as has been previously observed (Dong et al., 2003; Ravishankar et al., 2023). We also used the plasma membrane localization phenotype of the GFP-Ras2 reporter to confirm that HsFTase could modify CaaX sequences associated with the human Ras orthologs (Fig. 3C). Each sequence directed GFP-Ras2 to the plasma membrane, indicating farnesylation by HsFTase. Combined, these observations indicate that mammalian CaaX sequences can be modified by the humanized HsFTase strain and that ScGGTase-I has the opportunistic ability to modify certain CaaX sequences in the absence of FTase activity.

Our complementation studies indicated that HsFTase can modify the Ydj1 CASQ sequence (Fig. 2D). The ability to modify CASQ and a wide-range of non-canonical CaaX sequences, in addition to canonical sequences, is a recently identified property of ScFTase that has not been systematically investigated for FTase from other species (Berger et al., 2018; Hildebrandt et al., 2016b; Kim et al., 2023; Ravishankar et al., 2023).

To investigate the breadth of sequences recognized by HsFTase, we used a modified version of the humanized FTase yeast system to evaluate a set of well-characterized sequences representing different types of expected modification. For these tests, we evaluated farnesylation of a plasmid-encoded Ydj1-CaaX reporter in a HsFTase ydj1Δ strain (Fig. S7). Farnesylation of Ydj1 is required for growth at elevated temperatures (Caplan et al., 1992). Yeast expressing the Ydj1-CaaX variants exhibited growth profiles in the thermotolerance assay that were entirely consistent with previous reports and essentially indistinguishable regardless of whether farnesylation was supported by ScFTase or HsFTase (Fig. 4A) (Hildebrandt et al., 2016b). Ydj1-CASQ and -CAHQ (i.e. shunted sequences) best supported thermotolerance at 40°C; these sequences are derived from yeast Ydj1 and human DNAJA2, respectively. Ydj1-CVIA and -CTLM (i.e. canonical sequences) supported intermediate thermotolerance; these sequences are derived from a-factor and Gγ Ste18, respectively. As expected, Ydj1-SASQ (i.e. the unprenylated protein) and a vector control (i.e. no Ydj1) were unable to support growth at high temperature. Although the shunted status of CAHQ has not been verified through biophysical investigations, sequence homology, prediction algorithms, biochemical assays and the above thermotolerance analyses are all consistent with it being a shunted sequence (Hildebrandt et al., 2016b, 2024; Kim et al., 2023). Moreover, CAHQ is resistant to CaaX proteolysis as inferred from mating assays (Fig. S2). Gel-shift analysis of the Ydj1-CaaX variants revealed that both shunted and canonical CaaX sequences were fully prenylated by both ScFTase and HsFTase (Fig. 4B). Additional support for the ability of HsFTase to modify these sequences is evident when comparing the mobilities of Ydj1-CaaX variants in the absence of FTase (i.e. ram1Δ) (Fig. 4C).

The observations made for Ydj1 were also evident for human DNAJA2 (CAHQ) that was heterologously expressed in the humanized HsFTase yeast system. DNAJA2 has 48% identity to Ydj1 and can complement some ydj1Δ-associated defects (Whitmore et al., 2020). DNAJA2 supported thermotolerance in the context of both ScFTase and HsFTase, whereas DNAJA2-SAHQ did not, indicating that high-temperature growth was prenylation dependent (Fig. 4D). By gel-shift analysis, DNAJA2 was observed to be farnesylated by both ScFTase and HsFTase (Fig. 4E). These results further extend the utility of the humanized HsFTase yeast system for studies of heterologously expressed human CaaX proteins. Moreover, these observations are consistent with the hypothesis that ScFTase and HsFTase have highly similar sequence-recognition profiles that are likely broader than the CaaX consensus sequence.

Studies on the prenylation of CaaX proteins in most systems are hindered by several factors, including difficulty in detecting low abundance proteins, lack of phenotypic readouts, and high labor and time costs for case-by-case studies. Many of these hurdles can be overcome by using the Ydj1 reporter and genetically tractable yeast system described in this study. To extend the utility of our approach, Ydj1-CaaX variants harboring the CaaX sequences of prenylated human and potentially prenylated viral proteins were evaluated by gel-shift assay. Each variant was evaluated in the presence and absence of HsFTase to ensure that gel shifts were due to HsFTase and to assess whether alternate prenylation by ScGGTase-I was possible.

The Ydj1 reporter system was used to evaluate human CaaX sequences derived from chaperones (Nap1L1, DNAJA1 and Pex19), chromosome stability proteins [CENPE, CENPF and Spindly (SPDL1)], nuclear lamins (prelamin A and lamin B), Ras and Ras-like proteins (Rab38, KRas4A, KRas4B and HRas) and Prickle1 (Fig. 5A) (Ashar et al., 2000; Kho et al., 2004; Kim et al., 1990; Leung et al., 2007; Moudgil et al., 2015; Onono et al., 2010; Palsuledesai et al., 2014; Storck et al., 2019; Strutt et al., 2013; Varela et al., 2008). Apart from the Rab38-associated sequence (CAKS), every sequence was completely modified by HsFTase. CAKS is reported to be underprenylated in its natural protein context (i.e. Rab38), which is consistent with our findings (Köhnke et al., 2013). Sequences ending in methionine (CSIM, CAIM, CLIM, CIIM and CVIM) were fully modified in the presence of HsFTase, yet partially modified in the absence of FTase (ram1Δ), suggesting that they are alternately prenylated by ScGGTase-I.

The Ydj1 reporter system was also used to evaluate the prenylation of CaaX sequences associated with human and swine viral proteins (Fig. 5B). The importance of prenylation in viral infection is an emerging line of inquiry for infectious disease therapy, but documentation of viral protein prenylation is limited (Einav and Glenn, 2003; Marakasova et al., 2017). All of the viral CaaX sequences evaluated were modified by HsFTase. These sequences were derived from the influenza virus H1N1 protein PB2 (CRII), Molluscum contagiosum virus subtype 1 protein MC155R (CVLM), human cytomegalovirus protein IRL9 (CRIQ), hepatitis delta virus protein L-HDAg (CRPQ), human adenovirus 1 protein E1A (CLLS), hepatitis C virus non-structural protein 5A (NS5A) (CSMS), human parainfluenza virus 3 protein hemagglutinin-neuraminidase (HN) (CSQS), HIV protein NEF (CVVS) and African swine fever virus (ASFV) proteins A179L (homolog of mammalian Bcl-2) (CNLI), X69R (CKCY) and I9R (CVFM). Of these CaaX sequences, only L-HDAg has been biochemically characterized as being prenylated (Glenn et al., 1992). Several others are inferred to be prenylated based on suppression of viral replication by statins (Pronin et al., 2021).

We also evaluated the prenylation of CaaL/I/M sequences derived from well-characterized human CaaX proteins (Fig. 5C). Sequences adhering to this consensus motif are considered targets of GGTase-I or are interchangeably targeted by FTase or GGTase-I (Krzysiak et al., 2010; Trueblood et al., 1993; Zhang et al., 1994). The CaaX sequences were from small GTPases (CDC42, RAC1, RAP1B, RHOA and RHOB), Gγ subunits (GBG5 and GBG12, encoded by GNG5 and GNG12, respectively), the myelin protein CN37 (encoded by CNP), the F-box protein FBXL2 and the anti-viral protein Oas1 isoform p46 (De Angelis and Braun, 1994; Katayama et al., 1991; Kawata et al., 1990; Kilpatrick and Hildebrandt, 2007; Onono et al., 2010; Soveg et al., 2021; Storck et al., 2019; Wang et al., 2005; Wickenhagen et al., 2021). All Ydj1-CaaL/I/M variants were fully prenylated in the presence of HsFTase, with the exception of CSFL. Comparing band mobilities in the presence and absence of HsFTase revealed that several sequences were prenylated in a predominantly HsFTase-dependent manner (CKVL, CLLL, CLVL, CQLL, CSFL and CVLL). For the remaining sequences, a mobility shift was still apparent in the absence of HsFTase (CTII, CIIL, CTIL and CVIL), suggesting that these sequences are naturally geranylgeranylated or subject to alternate prenylation in the absence of FTase activity.

The substrate recognition profile of yeast FTase has recently been expanded to certain sequences one amino acid longer or shorter than the tetrapeptide CaaX sequence (i.e. CaX and CaaaX). This expanded specificity was first identified using a combination of yeast genetics and in vitro biochemical analyses using synthetic peptides and purified FTase (Ashok et al., 2020; Blanden et al., 2018; Schey et al., 2021; 2024). To extend these observations to HsFTase, previously studied Ydj1-CaX and -CaaaX variants were expressed in the humanized HsFTase yeast system and analyzed by gel-shift assay (Fig. 6). One additional CaX (CHA) sequence was evaluated due to its association with the SARS-CoV-2 protein ORF7b.

The 16 shorter CaX sequences evaluated exhibited varying degrees of farnesylation ranging from >90% (CVI and CTI) to <25% (CLL and CHA), with half exhibiting >50% farnesylation (eight of 16) (Fig. 6A; Table S1). The farnesylation patterns were similar to those previously observed for ScFTase (Ashok et al., 2020). The very weak farnesylation seen for Ydj1-CHA, which was quantified to be ∼8% of the population, was not evident for Ydj1-SHA, consistent with the cysteine-dependent nature of prenylation, but it remains to be determined whether such a low level of prenylation is biologically significant for the biology of SARS-CoV-2.

The 20 longer CaaaX sequences evaluated also exhibited varying degrees of farnesylation ranging from >90% farnesylation (e.g. CMIIM) to no farnesylation (e.g. CASSQ), with nearly half exhibiting >50% farnesylation (nine of 20) (Fig. 6B). Most sequences were similarly modified by ScFTase and HsFTase. Some partially modified sequences exhibiting varied prenylation (Table S1). Of note, the CGGDD sequence (annexin A2, encoded by ANXA2) was unmodified in the humanized HsFTase yeast system, despite being identified as farnesylated through methods involving metabolic labeling with an azido-farnesyl analog (Kho et al., 2004). Among the most dramatic differences in farnesylation by the yeast and human FTases were seen for CQTGP (GLCM1, encoded by GLYCAM1) and CSQGP (SNED1), which were modified by HsFTase but not by ScFTase, and CSLMQ (TCEA3), which was fully modified by ScFTase but <50% was modified by HsFTase (Fig. 6C). The differences in species specificity observed for these and other CaaaX sequences indicate differences in the substrate-binding pockets of these enzymes, which warrants further investigation.

This survey of CaX and CaaaX sequences reveals that HsFTase can indeed modify sequences with non-canonical lengths in a cell-based system. For many of the sequences associated with eukaryotic or viral proteins (see Table S2), it remains to be determined, however, whether they are prenylated in their native context.

Yeast genetic screens using HRas and Ydj1 reporters have comprehensively evaluated all 8000 CaaX sequences for reactivity with ScFTase (Kim et al., 2023; Stein et al., 2015). These genetic strategies are fully compatible with the humanized system developed in this study that could be used to query CaaX sequence space recognized by HsFTase. As proof of principle, we sampled a Ydj1-CaaX plasmid library using the humanized HsFTase system along with thermotolerance and gel-shift assays (Fig. 6D; Table S1). The results for a randomly sampled set of 26 Ydj1-CaaX variants revealed that farnesylation of Ydj1-CaaX by HsFTase correlated with thermotolerant growth properties. Exceptions were Ydj1-CWIM and -CWFC, which were farnesylated as judged by gel shift yet were temperature sensitive. The observations for CWIM and CWFC are consistent with the profile of canonical CaaX sequences (i.e. fully-modified: farnesylated and cleaved, as well as carboxyl methylated at the end of the protein sequence) that temper the ability of Ydj1 to support thermotolerance (Berger et al., 2018; Hildebrandt et al., 2016b).

Studies of FTase specificity have led to the development of predictive models for prenylation (Berger et al., 2022; Kim et al., 2023). The more recent of these models based on ScFTase was used to predict the farnesylation (positive and negative) of the randomly sampled Ydj1-CaaX variants (Fig. 6D; see bottom of the panel). The predictions correlated well with empirical observations for positive and negative modification by HsFTase, with a majority of sequences (20 of 26) being accurately predicted using rigorous thresholds (>50% prenylation as determined by band quantitation was required for positive classification; <10% prenylation was required for negative classification). For classification purposes, CKRC was judged to be unmodified because the gel mobility associated with this sequence also occurred in the absence of FTase (Fig. S3B). Of the outlier sequences, some were predicted to be modified but exhibited prenylation below the 50% threshold for positive classification (i.e. CWNA, CWHS, CSNY and CPGH). By contrast, two outlier sequences that were predicted to be prenylated were unmodified (i.e. CQLP and CFTP), suggesting differences in specificity between yeast and human FTase. Overall, the ScFTase-based prediction model suggested ∼75% accuracy for predicting prenylation of a CaaX sequence by HsFTase (Kim et al., 2023).

An obvious caveat of any predictive model is that the sequence must be accessible in its native context for modification by cytosolic FTase. Although many sequences tested in this study are associated with known prenylated proteins, others that exist on human proteins have not been investigated. For example, querying the UniProtKB/Swiss-Prot database for sequences from the limited-scale genetic screen revealed that CGGN and CKRC exist on human proteins (Table S2), but our data indicate that they cannot be farnesylated (Fig. 6D). We thus predict that they will not be farnesylated in their native contexts. CPAA, which is associated with the human endoplasmic reticulum-localized DNase-1-like 1 protein (DNASE1L1) and the extracellular laminin subunit α4 (LAMA4), was modified by HsFTase. The subcellular localizations of these human proteins likely disqualify them from being substrates of cytosolic FTase. Nevertheless, the observation that the substrate specificities of ScFTase and HsFTase are highly similar but not identical prompts the need for future studies aimed at fully evaluating CaaX sequence space in the context of HsFTase. Such studies are now possible with our humanized FTase yeast system, and such investigations may better refine the target profile of HsFTase, potentially leading to the discovery of previously unreported CaaX proteins.

To develop a yeast system for expressing HsGGTase-I, studies were also performed to investigate the functional equivalence of ScGGTase-I and HsGGTase-I subunits. Because GGTase-I activity is essential for yeast viability, strains lacking either subunit of ScGGTase-I can only be propagated when complemented by a plasmid-encoded copy of the missing gene. Pertinent yeast strains were created and, in a manner similar to FTase studies, they were transformed with plasmid-encoded copies of both orthologous HsGGTase-I subunit genes HsFNTA(α) and HsPGGT1B(β) under different promoter conditions (Fig. 7A). The strains were evaluated using a growth assay [i.e. 5-fluroorotic acid (5FOA) assay] that counter-selects for the URA3-marked plasmid encoding the ScFTase/GGTase-I α subunit gene (i.e. RAM2), resulting in yeast that lack FTase and are reliant on HsGGTase-I for growth. Regardless of whether subunits were expressed from the orthologous yeast promoters or the PGK1 promoter, strains producing HsGGTase-I exhibited robust growth nearly equivalent to yeast expressing endogenous ScGGTase-I (Fig. 7B). As expected, yeast strains lacking RAM2 did not support growth on 5FOA at 37°C because they lack FTase activity.

The 5FOA assay also revealed, as observed for FTase, that neither HsFNTA(α) nor HsPGGT1B(β) alone could complement for the absence of their yeast orthologs ScRam2(α) and ScCdc43(β), respectively, indicating that interspecies subunits do not form functional GGTase-I complexes (Fig. S4A). The results did not change when human subunits were overexpressed using the constitutive PGK1 promoter. This observation contrasts with the situation in which both HsFTase subunits had to be expressed from the PGK1 promoter for functional complementation. The reason for this discrepancy is unknown but might reflect differences in promoter strength (P_RAM1 versus P_CDC43) or β subunit stability.

Evidence suggests that GGTase-I has broader than anticipated substrate specificity at the a₁ and a₂ positions than implied by the CaaL/I/M consensus sequence (Kawata et al., 1990; Kilpatrick and Hildebrandt, 2007; Lebowitz et al., 1997; Onono et al., 2010; Sarkar et al., 2024). We investigated this issue using the humanized HsGGTase-I yeast system, plasmid-encoded Ydj1-CaaX variants, and gel-shift analysis. The yeast strains expressed either endogenous ScGGTase-I or HsGGTase-I for which subunit expression was driven by either orthologous RAM2(α) and CDC43(β) promoters (Hs) or the constitutive PGK1 promoter (_PGK1Hs) (Fig. S7). This analysis revealed that neither GGTase-I could modify Ydj1-CAHQ, a sequence associated with DNAJA2 that is strictly farnesylated (Andres et al., 1997). A mix of canonical (CVLL, CVIM, CIIM and CVLS) and non-canonical (CNLI and CSFL) CaaX sequences exhibited no or limited prenylation by ScGGTase-I and each was better modified by HsGGTase-I (Fig. 7C). This was more evident when HsGGTase-I subunits were expressed using the PGK1 promoter, presumably due to increased protein levels of HsGGTase-I. Collectively, these observations indicate that HsGGTase-I has the potential to modify certain non-canonical CaaL/I/M sequences (e.g. CNLI and CSFL) and that HsGGTase-I expressed in the yeast system has more activity and/or different specificity than that of ScGGTase-I.

To further survey the geranylgeranylation potential of CaaX sequences, we evaluated a test set of 22 sequences similar to the CaaL/I/M consensus using the humanized HsGGTase-I yeast system with lower GGTase-I activity (i.e. the system in which expression is driven using orthologous gene promoters). This was done to reduce the potential of overexpression artifacts. Most sequences exhibited a gel shift, but only a few were fully shifted, indicative of complete (i.e. CIIL, CTIL, CVIL and CVLL) or near-complete (CIIM, CVIM and CTII) geranylgeranylation (Fig. 7D). Within this data set, we observed that the a₁ position influenced prenylation by HsGGTase-I. This is evident when comparing the CxIM set of sequences: CSIM, CAIM and CLIM (∼50% modified) versus CIIM and CVIM (100% modified). A similar observation was made when comparing CxLL sequences: CLLL and CQLL (∼50% modified) versus CVLL (100% modified). Sequences in which the X position did not match that in the CaaL/I/M consensus (CAHQ, CVLS) were unmodified by HsGGTase-I. Lastly, we observed that GFP-Ras2-CIIM (KRas4A) and GFP-Ras2-CVIM (KRas4B) were prenylated by HsGGTase-I, as inferred from previous observations (see Fig. 3 and Fig. S4B).

The ability to humanize yeast for investigations of prenyltransferase activities can be adapted to studies of the CaaX proteases that are gatekeepers for the canonical CaaX modification pathway. Rce1 cleaves certain prenylated CaaX proteins, allowing for carboxymethylation. Its activity is prenyl dependent and target sequences typically have an aliphatic amino acid at a₂ (Kim et al., 2023; Plummer et al., 2006). Ste24 (also known as ZMPSte24) has also been identified as a CaaX protease, despite a lack of widespread CaaX protein substrates and an ability to cleave non-prenylated sequences (Boyartchuk et al., 1997; Hildebrandt et al., 2016a). Ste24 has an unrelated and likely primary function in protein quality control (Ast et al., 2016; Runnebohm et al., 2020).

The yeast mating assay (i.e. a-factor production) has been used to evaluate the specificities of heterologously expressed human Rce1 and ZMPSte24 (Mokry et al., 2009; Plummer et al., 2006). Studies of CaaX protease specificity are critical for resolving the cleavage preferences of medically relevant human proteins. For example, prelamin A (CSIM), which is associated with progeria-like diseases, is subject to two proteolytic events: one within the CaaX sequence and another 15 amino acids from the C-terminus (Barrowman et al., 2012b). Both cleavage events are needed to produce lamin A and have been attributed to ZMPSte24 (Barrowman et al., 2012a; Quigley et al., 2013). With the humanized CaaX protease yeast system, we observed that yeast mating, indicative of a-factor CaaX cleavage, occurred with a-factor-CSIM when HsRce1 was present, and not with ZMPSte24, in accordance with more recent observations (Fig. 8) (Berger et al., 2022; Nie et al., 2020 preprint). We extended these studies to evaluate the lamin B CaaX sequence (CAIM), which led to a mating phenotype with HsRCE1 or ZMPSte24. We also examined peroxisomal chaperone Pex19 sequences. Farnesylation of Pex19 is critical for its interaction with client proteins, and mutations of human Pex19 unrelated to its CaaX sequence are associated with Zellweger syndrome (Emmanouilidis et al., 2017; Matsuzono et al., 1999). Human Pex19 (CLIM) and yeast Pex19 (CKQQ) have strikingly different CaaX sequences. Both were fully prenylated by yeast and human FTases (see Fig. 5A and Fig. S5A). With the humanized CaaX protease system, we observed mating for a-factor-CLIM with HsRCE1 or ZMPSte24, whereas no mating was observed with a-factor-CKQQ. This result suggests that Pex19 proteins have, for reasons unknown, evolved C-termini with distinct biophysical properties. Taken together, our studies fully support the combined utility of the humanized prenyltransferase and CaaX protease yeast systems for profiling the CaaX modification potential of different sequences. Such studies could prove useful for predicting the potential effects of inhibiting the Rce1 CaaX protease, which would impact a narrower range of CaaX protein targets than prenyltransferase inhibitors.

We report the development of yeast systems useful for systematic characterization of prenylation and cleavage of CaaX sequences by human prenyltransferases and CaaX proteases. The humanized HsFTase system, in particular, was phenotypically equivalent to yeast expressing native ScFTase in various assays: (1) a-factor-dependent mating, (2) Ydj1-based thermotolerance, (3) GFP-Ras2 localization and (4) gel shift of distinct reporters. The humanized HsFTase yeast system is an accessible and rapid tool for characterizing the farnesylation potential of candidate CaaX sequences and heterologously expressed human CaaX proteins. Importantly, this system is cell based, which alleviates concerns associated with in vitro farnesylation assays that depend on chemically modified substrates, and unnatural amounts of enzyme and substrate. A limitation of this system, however, is that CaaX sequences determined to be prenylated using any of our reporters (i.e. Ydj1, Ras2 and a-factor) will still need to be validated in their native context. For example, the accessibility of a CaaX sequence and the properties of sequences upstream of CaaX may constrain reactivity (i.e. the subcellular localization of the substrate, buried C-terminus, etc.) (Bond et al., 2017; Hicks et al., 2005; Reid et al., 2004). Thus, any positive results with our system can should be coupled with additional studies. Nonetheless, the ease and convenience of the humanized HsFTase yeast system that we have developed is an important and valuable new resource for initial and supporting investigations on the farnesylation potential of CaaX sequences.

Limited comparative data exists for the specificities of mammalian FTase relative to that of ScFTase. In vitro, ScFTase and rat FTase exhibit similar specificities when evaluated against a CVa₂X peptide library (Wang et al., 2014). Our in vivo results confirm and further extend such observations to reveal that ScFTase and yeast-expressed HsFTase have a striking degree of substrate conservation (see Figs 2, 4 and 5, Fig. S5 and Table S1). This is most evident in the ability of both enzymes to similarly farnesylate canonical and non-canonical sequences (i.e. sequences lacking aliphatic amino acids at a₁ and a₂). These findings indicate that farnesylation-prediction algorithms developed using ScFTase should work reasonably well for predicting HsFTase specificity (Kim et al., 2023). The conservation between human and yeast FTases extends to prenylation of sequences with non-canonical lengths (i.e. CaX and CaaaX), although species-specific differences in substrate recognition of longer sequences were observed. Toward mechanistically understanding such differences, it will be important to compare high-resolution structures of mammalian and yeast FTase, which have limited primary sequence identity (22-34%). Although many mammalian X-ray crystal structures of FTase exist, there is currently no ScFTase structure. Overlay of the HsFTase crystal structure and yeast FTase structural models predicted by AlphaFold, however, revealed excellent conservation of residues that coordinate the a₂ and X positions of CaaX sequences (Fig. S6), consistent with our observations that these enzymes have highly similar substrate specificities.

This study also produced a humanized HsGGTase-I yeast system for characterizing the geranylgeranylation potential of candidate CaaX sequences that, presumably, can be used for analyses of heterologously expressed human CaaX proteins. Human GGTase-I prenylated a wide range of CaaL/I/M sequences but did not prenylate CaaS/Q sequences, in alignment with expectations. Both yeast and human GGTase-I had the ability to alternately prenylate similar sequences in the context of GFP-Ras2 (i.e. CIIM and CVIM) and Ydj1 (i.e. CAIM, CIIM, CKVL, CLIM, CLLL, CLVL, CSIM and CVIM) (Figs 3, 5 and 7; Fig. S4B; Table S1). This typically occurs when FTase activity is reduced through action of farnesyltransferase inhibitors or, in our case, when FTase activity was genetically ablated. This observation demonstrates that recognition of alternately prenylated sequences by GGTase-I is a transferable property (i.e. it occurs in other protein contexts) and is evolutionarily conserved between yeast and human enzymes, and likely among other species as well. A caveat of the humanized HsGGTase-I yeast system, however, is that it displayed more robust activity than that of endogenous yeast GGTase-I in modifying Ydj1-CaaL/I/M variants. This observation could be due to higher HsGGTase-I enzyme levels and/or specificity differences between the yeast and human enzymes. An overlay of the rat GGTase-I crystal structure and yeast GGTase-I AlphaFold models predicts different active site architectures for these enzymes, which could explain our observations (Fig. S6).

The scope of potentially prenylated proteins in protein databases is expansive, with over 6000 annotated proteins in eukaryotes and viruses ending in Cxxx. When expanded to 3-mer and 5-mer length sequences, this number is over 15,000 (Fig. 1C). Not accounted for in this analysis is the potential number of prokaryotic CaaX proteins. Although prokaryotes lack protein prenylation machinery, certain pathogenic bacteria (e.g. Legionella pneumophila and Salmonella typhimurium) encode CaaX proteins that are prenylated by host enzymes as part of their infectious life cycle (Ivanov et al., 2010; Reinicke et al., 2005). Thus, reliable prediction algorithms and reporter systems will be helpful in identifying and verifying the subpopulation of CaaX sequences modified by HsFTase or HsGGTase-I against the backdrop of infectious disease.

Viral genomes alone encode many under-characterized proteins that might be prenylated by host enzymes involved in CaaX protein modification. The viral-derived CaaX sequences tested in this study were all farnesylated. We observed that CRPQ (hepatitis delta virus large delta antigen, L-HDAg) is modified by both human and yeast FTases, but not by GGTase-I. Furthermore, we observed that the CRPQ sequence is not cleaved, consistent with its non-canonical nature (Fig. 5; Fig. S2). Farnesylation of L-HDAg is required for virion assembly, and a farnesyltransferase inhibitor (lonafarnib) is currently in phase 3 clinical trials for use in combination treatments for hepatitis D infections (Khalfi et al., 2023; Koh et al., 2015; Yardeni et al., 2022). We also determined that CRIQ (cytomegalovirus protein of unknown function IRL9) can be farnesylated, cleaved and carboxymethylated. CaaX modification of IRL9 has not been previously reported, but it is a good candidate for additional investigations as statins that block production of mevalonate (a precursor of farnesyl and geranylgeranyl groups) are known to diminish cytomegalovirus replication (Potena et al., 2004). Our yeast systems revealed that other viral CaaX sequences, i.e. CRII (influenza virus H1N1 protein PB2), CVLM (M. contagiosum protein MC155R) and CNLI (ASFV protein A179L), were substrates of both HsFTase and HsGGTase-I. If these sequences are prenylated in their native contexts, dual prenyltransferase inhibitors might have better utility for blocking their prenylation. Given that all the viral-derived CaaX sequences evaluated in our study were prenylated, it is prudent to consider the role that the prenylation of viral proteins may have on viral propagation and infection.

In summary, we developed three humanized yeast systems useful for probing the modification potential of CaaX sequences. These systems provide an accessible and rapid avenue for determining whether a specific CaaX sequence is unmodified, prenylated (i.e. farnesylated or geranylgeranylated) or additionally cleaved. Such investigations are critical first steps for characterizing any CaaX protein for which the modification status is unknown. We expect that individual CaaX sequences determined to be unmodified by our systems will likely be unmodified in their native protein contexts, whereas modified sequences will need to be fully vetted in their natural contexts to confirm predictions derived using these systems. Additionally, the data collected in this study suggest that yeast and human FTase specificities are largely preserved, consistent with their conserved active site architectures, whereas yeast and mammalian GGTase-I specificities are less preserved, consistent with proposed differences in their active site architectures.

The manipulated genomes of humanized prenyltransferase strains are summarized in Fig. S7. Detailed genotypes are listed in Table S3, with detailed strain construction strategies given in the supplementary Materials and Methods. Yeast were typically cultured in synthetic complete (SC) or yeast extract-peptone-dextrose (YPD) liquid and solid yeast media unless otherwise noted. For expression of Myc-HRas^Q61L from the inducible MET25 promoter (p-05547) in methionine auxotroph strains (i.e. yWS2544 and yWS3186), yeast were cultured in SC-leucine containing 20 µg/ml methionine to late log phase, washed and resuspended in SC-leucine containing 2 µg/ml methionine to allow for both growth and induction of the MET25 promoter.

New strains were created for this study by standard genetic manipulations starting with commercially available haploid and heterozygous diploid genomic deletions. KAN^R- and NAT^R-marked gene replacements were confirmed by growth on YPD medium containing G418 (200 µg/ml; Research Products International) or nourseothricin (100 µg/ml; GoldBio), respectively. All gene replacements were further checked by PCR to confirm the presence of the knockout at the correct locus and absence of the wildtype open reading frame (ORF). Plasmids were introduced into strains via a lithium acetate-based transformation procedure (Elble, 1992). Yeast sporulation was carried out in a solution of 2% potassium acetate, 0.25% yeast extract and 0.1% dextrose. Spore enrichment and random spore analysis followed published methods (Rockmill et al., 1991). For the humanized FTase strains (yWS3186 and yWS3220), P_PGK1-FNTB was integrated at the RAM1 locus, replacing the ORF, using a loop-in-loop-out strategy, and P_PGK1-FNTA was integrated at the his3Δ1 locus by homologous recombination using a HIS3-based integrative plasmid.

A comprehensive list of plasmids is given in Table S4 and detailed plasmid construction strategies are in the supplementary Materials and Methods. Plasmids were routinely analyzed by diagnostic restriction digestion and DNA sequencing (Eurofins Genomics, Louisville, KY, USA) to verify the entire ORF and surrounding sequence. Plasmids recovered from the Ydj1-CaaX Trimer20 library are described elsewhere (Kim et al., 2023). New plasmids encoding Ydj1-CaaX and a-factor-CaaX (encoded by MFA1) variants were made by PCR-directed plasmid-based recombinational cloning as previously described (Berger et al., 2018; Oldenburg et al., 1997). Plasmids with GFP-Ras2-CaaX variants were generated using a modified QuikChange site-directed mutagenesis procedure; complementary mutagenic primers were used for whole plasmid PCR of pWS1735 using Q5 polymerase (New England Biolabs) and an 8-min extension time, with products being digested with DpnI (New England Biolobs) prior to E. coli transformation. Plasmids encoding HsFTase and HsGGTase-I subunits were created in multiple steps. First, the ORFs and flanking 5′ and 3′ sequences of yeast prenyltransferase subunits were PCR-amplified from strain BY4741 and subcloned into the multicloning sites of appropriate pRS series vectors (Sikorski and Hieter, 1989). In parallel, synthetic cDNAs for human FNTA, FNTB and PGGT1B that were codon optimized for S. cerevisiae were subcloned into the PstI and XhoI sites of pBluescriptII KS(-); all cDNAs were commercially obtained (GenScript). The human ORFs were engineered to also have 39 bp of flanking sequence on both the 5′ and 3′ends, which equivalently matched the flanking sequences of the orthologous yeast ORFs. Next, DNA fragments encoding the human ORFs and yeast flanking sequences were used with recombination-based methods for direct gene replacement of the plasmid-encoded yeast ORFs. The PGK1 promoter was amplified and used to replace the orthologous yeast promoters by similar recombination-based methods. To change selectable markers, the various FTase and GGTase-I promoter-containing segments were subcloned as XhoI-SacI fragments into different pRS vector backbones by conventional ligation-based cloning.

The assay was performed as previously described (Blanden et al., 2018; Hildebrandt et al., 2016b). In brief, plasmid-transformed strains were cultured in appropriate selective synthetic drop-out (SC−) liquid media to saturation (25°C, 24 h); strains without plasmids were cultured in SC complete media. Cultures were serially diluted into H₂O (10-fold dilutions), and dilution mixtures replica spotted onto YPD plates. Plates were incubated (25°C for 72 h; 37°C for 48 h; 40°C and 41°C for 72 h) then digitally imaged using a Canon flat-bed scanner (300 dpi; grayscale; TIFF format).

Cell-mass equivalents of log-phase yeast [absorbance at 600 nm (A₆₀₀) of 0.95-1.1) cultured in appropriate selective SC− liquid media were harvested by centrifugation, washed with water, and processed by alkaline hydrolysis and trichloroacetic acid precipitation (Kim et al., 2005). Total protein precipitates were resuspended in urea-containing sample buffer (250 mM Tris, 6 M urea, 5% β-mercaptoethanol, 4% SDS, 0.01% bromophenol blue, pH 8) and analyzed by SDS-PAGE (9.5%), followed by immunoblotting. Blots were processed according to standard protocols. Antibodies used for western blots were: rabbit anti-Ydj1 polyclonal (1:10,000; gift from Avrom Caplan, City College of New York); mouse anti-c-Myc 9E10 monoclonal (1:1000, Santa Cruz Biotechnology, sc-40); mouse anti-GFP B-2 monoclonal (1:500, Santa Cruz Biotechnology, sc-9996); mouse anti-DnaJA2(7) monoclonal (1:500, Santa Cruz Biotechnology, sc-136515); mouse anti-ScRas2 monoclonal (1:600, Santa Cruz Biotechnology, sc-365773); mouse anti-HIS his.h8 monoclonal (1:1000, VWR, 101981-852); and goat HRP-anti-rabbit and goat HRP-anti-mouse (1:1000, Kindle Biosciences, R1006 and R1005, respectively) antibodies. Immunoblots were developed with ECL reagent (ProSignal Pico ECL Spray or Kindle Biosciences KwikQuant Western Blot Detection Kit) and images were digitally captured using the KwikQuant Imager system (Kindle Biosciences). Adobe Photoshop was used for cropping and rotating of images; no other image adjustments were applied.

Prenylation of Ydj1-CaaX was quantified using ImageJ and at least two exposures of each digital image having good dynamic range of band intensities. The percentage of prenylation was calculated as the intensity of the lower (prenylated) band divided by the total intensity of the prenylated and unprenylated (upper) bands. Data presented in Table S1 represent the averages of the indicated number of biological replicates. For duplicate samples, values are reported as an average and associated range. For multiplicate samples, the values are reported as an average and standard error of the mean (s.e.m.).

Qualitative and quantitative yeast mating assays were performed as detailed previously (Berger et al., 2018). In brief, SM2331, the MATα strain (IH1793), and yWS164-derived MATa test strains were cultured to saturation at 25°C in YPD or selective SC− liquid media as appropriate. Cultures were normalized by dilution with fresh media (A₆₀₀ 1.0), mixed in a 1:9 (MATa:MATα) ratio in individual wells of a 96-well plate, and cell mixtures were subjected to 10-fold serial dilution using the MATα cell suspension as the diluent. For qualitative analyses, each dilution series was replica spotted onto SC-lysine and minimal synthetic defined (SD) solid media. For quantitative analysis, a portion of an empirically predetermined dilution mixture was spread onto SC-lysine and SD plates in duplicate, and colonies counted after plate incubation (72 h, 30°C). The SC-lysine cell count reports the total number of MATa haploid cells initially in the mixture, whereas the SD cell count reports the number of mating events (i.e. diploids). Mating efficiencies were normalized to an a-factor (MFA1) positive control within each experiment.

Imaging was performed as detailed previously (Ravishankar et al., 2023). In brief, yeast transformed with plasmids encoding GFP-RAS2-CaaX variants were cultured to late log phase (A₆₀₀ 0.8-1) in SC-uracil liquid media and viewed using a Zeiss Axio Observer microscope equipped with fluorescence optics (Plan Apochromat 63×/1.4 N.A objective). Images were captured using AxioVision software and minor image adjustments performed using Adobe Photoshop.

yWS3186 was transformed with the Ydj1-CaaX plasmid library (pWS1775), plated onto SC-uracil solid media, and incubated for 4 days at 25°C. Randomly sampled single colonies of varying size were individually cultured and subjected to a modified temperature sensitivity assay using three spots of 20-fold serial dilutions. Growth at 41°C was scored on a scale of 0 to 5 relative to growth of the control strains yWS3312 (Ydj1-SASQ, score of 0) and yWS3311 (Ydj1-CASQ, score of 5). Plasmids were recovered from candidates scoring 0 or 5 (n=11 and n=15, respectively), sequenced to identify the CaaX sequence and retransformed into yWS3186 for retesting by temperature sensitivity (10-fold dilutions) and gel-shift assays.

BY4741, yWS3106 and yWS3109 were transformed with combinations of HIS3- and LEU2-marked plasmids encoding HsFNTA and HsPGGT1B, and/or empty vector plasmids, such that all strains had the same selectable markers. Strains cultured on SC-histidine-leucine-uracil solid media were used to inoculate SC-histidine-leucine liquid media. After incubation (25°C, 24 h), cultures were normalized with fresh media (A₆₀₀ 2.0), added to wells of a 96-well plate, serially diluted with H₂O (8-fold dilutions), and the dilution mixtures replica spotted onto SC complete solid media containing 1 mg/ml 5-FOA (Research Products International, Mount Prospect, IL, USA).

ScanProsite (https://prosite.expasy.org/scanprosite/) was used to search for CaaX sequences in Eukaryota, H. sapiens, S. cerevisiae and viruses in the UniProtKB/Swiss-Prot database using the search strings ‘CXXX>’, ‘C{C}XXX>’, or ‘{C}CXX>’ (search performed 12 July 2023). For compilation of Table S2, several CaaaX and CaX and sequences had no human- or virus-associated hits in the UniProtKB/Swiss-Prot database. For these sequences, the search was expanded to the UniProtKB/TrEMBL database, inclusive of all eukaryotes.

The active site residues of HsFTase and rat GGTase-I have been reported previously (Reid et al., 2004). The equivalent residues for ScFTase and ScGGTase-I subunits were determined by structure-based sequence alignment using PyMOL and the crystal structures of HsFTase (PDB: 1S63) and rat GGTase-I (PDB: 1N4Q) for reference. The following AlphaFold-derived structures were mapped onto the reference structures: ScRam1, ScRam2 and ScCdc43 (Jumper et al., 2021). EMBOSS Needle (i.e. primary sequence alignment) was used to determine the active site residues of the human GGTase-I β subunit using the rat PGGT1B sequence (Madeira et al., 2022). EMBOSS Needle was also used to determine percentage identity and percentage similarity scores for each subunit relative to the reference HsFNTA, HsFNTB and rat PGGT1B sequences.

We thank Dr Avrom Caplan (City College of New York) for anti-Ydj1 antibody, Dr Jun Wang (Institute Pasteur of Shanghai, Chinese Academy of Sciences) for discussions on ASFV CaaX proteins and Schmidt laboratory members for constructive feedback and technical assistance.