The Molecular Frequency, Conservation and Role of Reactive Cysteines in Plant Lipid Metabolism

Abstract

Cysteines (Cys) are chemically reactive amino acids containing sulfur that play diverse roles in plant biology. Recent proteomics investigations in Arabidopsis thaliana have revealed the presence of thiol post-translational modifications (PTMs) in several Cys residues. These PTMs are presumed to impact protein structure and function, yet mechanistic data regarding the specific Cys susceptible to modification and their biochemical relevance remain limited. To help address these limitations, we have conducted a wide-ranging analysis by integrating published datasets encompassing PTM proteomics (comparing S-sulfenylation, persulfidation, S-nitrosylation and S-acylation), genomics and protein structures, with a specific focus on proteins involved in plant lipid metabolism. The prevalence and distribution of modified Cys residues across all analyzed proteins is diverse and multifaceted. Nevertheless, by combining an evaluation of sequence conservation across 100+ plant genomes with AlphaFold-generated protein structures and physicochemical predictions, we have unveiled structural propensities associated with Cys modifications. Furthermore, we have identified discernible patterns in lipid biochemical pathways enriched with Cys PTMs, notably involving beta-oxidation, jasmonic acid biosynthesis, fatty acid biosynthesis and wax biosynthesis. These collective findings provide valuable insights for future investigations targeting the mechanistic foundations of Cys modifications and the regulation of modified proteins in lipid metabolism and other metabolic pathways.

Introduction: An Overview of Cysteine Biochemistry

Cysteine (Cys) is an essential, sulfur-containing amino acid that plays diverse roles in protein structure and function, metabolism and redox signaling (Poole 2015, Bak et al. 2019, Garrido Ruiz et al. 2022). Its unique functional properties arise from the nucleophilic nature of its sulfhydryl (–SH) side chain, which is incorporated into Cys through sulfur assimilation processes in photosynthetic organisms. The nucleophilicity of the thiol group (especially in the thiolate form S⁻) facilitates the stabilization of protein structures, metal binding, redox activity and catalytic mechanisms across diverse metabolic pathways (Giles et al. 2003). Moreover, its reactivity with antioxidant molecules and thioredoxin (Trx)-like proteins offers multiple avenues for redox-based regulation in signaling and metabolism (Couturier et al. 2013, Held 2020, Martí et al. 2020, Zhou et al. 2023). Proteomics studies have provided compelling evidence that numerous –SH residues are susceptible to thiol-based post-translational modifications (Cys PTMs), including many proteins in lipid metabolism (Li-Beisson et al. 2013), potentially affecting protein structure and function (Corpas et al. 2022). Consequently, understanding the impact of these modifications on Cys residues has become an area of active research, shedding light on the intricate interplay between redox regulation and Cys-driven protein function in both plant lipid metabolism and the broader plant proteome.

The biosynthesis of sulfur compounds is critical for the cell viability of photosynthetic organisms (Kopriva et al. 2007; Takahashi et al. 2011, Jez and Kopriva 2019). Cys is produced through a multistep biochemical pathway incorporating sulfide, derived from the plastid sulfate assimilation pathway, with the amino acid skeleton O-acetylserine. Sulfate is initially acquired through cellular uptake and can have multiple metabolic fates. Sulfate transported into the plastid is activated through adenylation by ATP sulfurylase to form adenosine 5′-phosphosulfate (APS). Two sequential reductions are then catalyzed by APS reductase and ferredoxin-dependent sulfite (SiR) to produce sulfite and sulfide, respectively. A β-replacement reaction of O-acetylserine, formed from the activation of serine with acetyl coenzyme A by serine acetyltransferase, exchanges the acetyl moiety with sulfide by O-acetylserine (thiol) lyase to form Cys. Cys is one of the least abundant amino acids in the plant proteome (∼2% relative abundance, slightly higher than tryptophan in relative abundance) (Lamesch et al. 2012, UniProt 2015). Given the energy expended to synthesize Cys and its critical roles in structure–function (described later), it may be expected that Cys residues are evolutionarily conserved. However, Cys often exhibits the most extreme conservation patterns, i.e. majority highly conserved or poorly conserved (Marino and Gladyshev 2010a), a concept explored in this review within lipid metabolism proteins and the broader plant proteome. In addition to its direct usage in protein biosynthesis, Cys is an important precursor molecule for other essential sulfur-containing, small-molecule thiols, e.g. the proteinogenic amino acid methionine, the antioxidant glutathione, catalytic Fe–S clusters and cofactors such as coenzyme A and thiamine (Romero et al. 2014, Poole 2015).

Many Cys residues have been shown to contribute to protein formation and stability in plants, primarily through formation of disulfide bonds, e.g. Bienert et al. (2012), Skryhan et al. (2015), Xia et al. (2018). Several cell compartments and protein machinery contribute to this oxidative protein-folding system in plant cells (Onda 2013). However, multiple lines of evidence suggest that most plant proteins do not contain intramolecular nor intermolecular disulfide bonds. Foremost, several studies have experimentally examined the prevalence of disulfide bonds (Lee et al. 2004, Nietzel et al. 2020, Doron et al. 2021, Willems et al. 2023). Additionally, sequence analyses and structural data provided within this review show that most Cys are not in proximity to form disulfide bonds. Finally, the distribution of protein biosynthesis and variation of oxidative environments limit the number of disulfide bonds. For example, while about 90% of endoplasmic reticulum (ER)–synthesized proteins have intramolecular disulfide bonds that contribute to their conformational stability (Urade 2019), only about one third of the cellular proteome is synthesized by the ER (Anelli and Sitia 2008). Alternatively, Cys can stabilize α helices using weak hydrogen bonds, i.e. acting as a weak hydrogen bond donor (thiol) or acceptor (thiol or thiolate) with backbone amide groups (Bulaj 2005, Liu et al. 2016, Mazmanian et al. 2016, Garrido Ruiz et al. 2022). Typically, however, the Cys side chain is found in hydrophobic environments where a thiolate anion can be destabilizing. This has led to Cys often being classified as a nonpolar residue although its chemical properties are more consistent with a polar residue, e.g. serine (Poole 2015).

The enhanced reactivity of thiols results in Cys residues participating in several mechanistically distinct redox reactions (Giles et al. 2003), which in concert with other redox-active processes that are essential for plant redox regulation (Couturier et al. 2013) and phytochemical diversity (Horn 2021). Given that each Cys has a unique microenvironment and the microenvironments may themselves vary in dynamic cellular environments, it is often challenging to assign a specific level of Cys reactivity (e.g. reflected by lower pK_a values), especially if the protein’s structure has not been well characterized (Nielsen et al. 2011, Olsson et al. 2011, Roos et al. 2013, Garrido Ruiz et al. 2022). However, an increased number of experimentally determined structures (Berman et al. 2000) as well as improved homology-based structure predictions, e.g. AlphaFold (Varadi et al. 2022), have undoubtedly improved the understanding of Cys structure–function. The valence electrons of the sulfur group of Cys, along with imidazole nitrogen of His, coordinate most of the metal ligands, e.g. zinc (Zn²⁺) and iron–sulfur (Fe–S), in plant proteins catalyzing a diverse set of chemical reactions (Sticht and Rösch 1998, Pace and Weerapana 2014, Willems et al. 2023). In fact, an analysis of disulfide bonds and metal-binding sites in AlphaFold2 structures across the plant kingdom showed that about 5–10% Cys were predicted to be involved in metal binding, with the majority coordinating Zn²⁺ (Willems et al. 2023). The thiolate of Cys not only coordinates metals but also performs nucleophilic attacks on substrates, e.g. in plant lipid metabolism forming an acyl-enzyme intermediate in fatty acid biosynthesis (Moche et al. 2001) and fatty acid elongation (Ghanevati and Jaworski 2001), as well as activating the hydrolysis of the thioester bond in acyl–acyl carrier protein (ACP) thioesterases (Yuan et al. 1996).

A complex cellular network of redox-driven protective mechanisms has evolved to enable plants to continually integrate and dynamically tune energy transduction, metabolism, gene expression and growth through thiol/disulfide exchange reactions (Dietz 2011, Geigenberger et al. 2017, Liebthal et al. 2018, Nikkanen and Rintamäki 2019, Yoshida and Hisabori 2023). For example, in chloroplasts, two thiol-redox systems, ferredoxin (Fdx)-dependent and NADPH-dependent, are crucial for maintaining redox homeostasis. These systems use a family of Trx proteins to mediate thiol–disulfide exchanges via Cys-rich motifs (e.g. CXXC) with a propensity to act as a ‘redox regulatory switch’. While both thiol-redox systems have been implicated in numerous cellular functions, there is increasing evidence for roles in chloroplast lipid metabolism and membrane biology (Hernández and Cejudo 2021), e.g. light-dependent redox activation of acetyl-CoA carboxylase (ACCase) (Sasaki et al. 1997), Trx activation of monogalactosyldiacylglycerol synthase (Yamaryo et al. 2006) and peroxiredoxin q activation of fatty acid desaturase 4 (Horn et al. 2019). Mutant studies in Arabidopsis have demonstrated that a disruption of these thiol–disulfide regulation networks often leads to enhanced stress susceptibility and in some cases lethality (Lamkemeyer et al. 2006, Koussevitzky et al. 2008, Marty et al. 2009, Naranjo et al. 2016).

In addition to thiol/disulfide exchange, many Cys may undergo one of several thiol-based PTMs (Corpas et al. 2022). These reversible PTMs include S-sulfenylation, S-nitrosylation (or nitrosation), persulfidation (or S-sulfhydration), S-acylation, S-glutathionylation, S-cyanylation and S-prenylation (Supplementary Table S1). An unbalanced oxidative environment, e.g. misregulated levels of reactive oxygen species (ROS) and reactive nitrogen species (RNS), can alter the prevalence of Cys PTMs (Dixon et al. 2005, Huang et al. 2019). Furthermore, sulfenic acid may be further oxidized to sulfinic and sulfonic acids, non-reversible damaging redox states (S-sulfination or S-sulfonation). There is also increasing evidence that PTMs often alter protein functionality although in most cases, specific outcomes for each PTM–protein combination and mechanism are not known (Horn 2021, Corpas et al. 2022). Multiple studies using advances in thiol-specific labeling and mass spectrometry–based detection that have revealed many proteins undergo one or more of these PTMs (Dixon et al. 2005, Hemsley et al. 2013, Hu et al. 2015, Aroca et al. 2017, García et al. 2019, Huang et al. 2019, Kumar et al. 2020).

In the ‘Structural Tendencies and Sequence Conservation of Cys PTMs’ section, we discuss an in silico–based analysis of the frequency, microenvironments and evolutionary conservation of Cys in the Arabidopsis thaliana proteome with an emphasis on Cys detected with a PTM by one of three processes: S-sulfenylation (Huang et al. 2019), S-nitrosylation (Hu et al. 2015) and/or S-acylation (Kumar et al. 2020) (Table 1, Supplementary S1 and Supplementary Dataset S1). These sets were chosen among the Cys PTM datasets, as summarized in Corpas et al. (2022), as they were all conducted in A. thaliana, and a substantial but selective number of proteins/peptides and Cys sites were identified (3000+ total proteins and Cys sites identified). With our laboratory studying the role of Cys in plant lipid metabolism, within the ‘Structural Tendencies and Sequence Conservation of Cys PTMs’ section, we will frame our analyses on the 775 proteins annotated as part of the plant lipid metabolism, i.e. ‘AraLip’ database (Li-Beisson et al. 2013), in addition to the broader proteome (‘all’). In the ‘Cysteine PTMs in Lipid Metabolism Proteins and Pathways’ section, in addition to the three datasets discussed earlier, we include one additional S-sulfenylation dataset (Wei et al. 2020) and three persulfidation datasets (Aroca et al. 2017, Jurado-Flores et al. 2021, 2023) to examine which lipid metabolism proteins and pathways are over-represented by Cys PTMs (Table 1, Supplementary Dataset S1). Implications on plant lipid signaling and regulation networks are discussed. Finally, we integrate the analyses from both sections summarizing how this information may inform future studies in Cys biochemistry in plant lipid metabolism and contribute to enhanced plant resilience.

Table 1

Representative lipid metabolism proteins containing 1+ Cys PTM

AGIa	Protein descriptionb	Lipid pathway/functionc	Cys Posd	% Cys IDe	pKaf	Buried % (SASA)g	6-Å pocket, % IDh
S-sulfenylation
AT2G43790	MAP Kinase 6 (MPK6)	Fatty Acid (FA) Elongation and Wax Biosynthesis	201	100	10.8	40 (0)	67
AT2G43710/AT3G02610	SAD (SAD, FAB2) / SAD (DES2)	FA Biosynthesis and Modification	260 267	100 100	11.8 9.2	100 (0) 100 (0)	97 98
AT5G46290	KASI	FA Biosynthesis	135 266 343 401	81 100 100 94	12.4 9.8 12.3 13.0	100 (1) 36 (69) 94 (0) 100 (1)	93 99 79 97
AT4G13050	Acyl-ACP Thioesterase A (FATA)	FA Biosynthesis	300	98	10.8	85 (24)	85
AT3G06860	Multifunctional Protein (MFP2)	β-oxidation	493 251	97 95	11.3 13.1	91 (22) 100 (0)	96 69
AT2G43710	SAD (SAD, FAB2)	FA Biosynthesis and Modification	366	93	10.4	35 (47)	59
AT3G25860	Dihydrolipoamide Acetyltransferase (E2)	FA Biosynthesis	331	93	9.3	5 (10)	87
AT2G35690/AT4G16760	Acyl-CoA Oxidase (ACX1/5)	β-oxidation	376 467	89 65	12.2 10.3	100 (2) 33 (9)	90 72
S-nitrosylation
AT1G04710	Ketoacyl-CoA Thiolase (KAT1, PKT)	β-oxidation	130	100	10.8	100 (14)	99
AT5G46290	KASI	FA Biosynthesis	325 343	99 100	8.9 12.3	11 (83) 94 (0)	89 82
AT3G16950	Dihydrolipoamide Dehydrogenase (E3)	FA Biosynthesis	124	99	DS*	0 (22)	99
AT3G52430	Phytoalexin-deficient 4 (PAD4)	Lipase-like Protein in Salicylic Acid Signaling	450	95	12.5	76 (13)	78
AT1G64400/AT4G23850	LACS (LACS3/LACS4)	FA Activation	587	94	11.4	52 (2)	49
AT3G25860	Dihydrolipoamide Acetyltransferase (E2)	FA Biosynthesis	331	93	9.3	5 (10)	87
AT1G36160	ACCase (ACC1)	FA Biosynthesis	668 871	93 77	10.3 9.75	70 (13) 0 (28)	92 56
AT4G11850 AT4G11830	Phospholipase Dγ (PLDγ)	Membrane Lipid Remodeling	173 833	90 70	10.7 9.9	57 (31) 33 (28)	75 83
AT4G16760	Acyl-CoA Oxidase (ACX1)	β-Oxidation/Jasmonic Acid Biosynthesis	376	89	12.2	100 (5)	90
AT1G06290	Acyl-CoA Oxidase (ACX3)	β-Oxidation	219 489	89 32	9.0 10.5	3 (43) 5 (11)	93 67
AT3G45140	Lipoxygenase (13-LOX, LOX2)	JA Biosynthesis	281	34	11.1	22 (6)	60
AT1G13280	Allene Oxide Cyclase (AOC4)	JA Biosynthesis	110	27	8.9	0 (14)	67
S-acylation
AT2G15050	LTP type 1	Lipid Transport/Wax Biosynthesis	102	100	DS*	0 (13)	54
AT2G37870	LTP type 5	Lipid Transport/Wax Biosynthesis	73 71	100 100	DS* DS*	0 (36) 0 (0)	76 100
AT3G51600	LTP type 1	Lipid Transport/Wax Biosynthesis	54 114 53 100	100 100 100 100	DS* DS* DS* DS*	0 (21) 0 (7) 0 (0) 0 (6)	46 67 80 87
AT5G46290	KASI	FA Biosynthesis	266 343 325 401 131 135 347	100 100 99 94 94 81 28	9.8 12.3 8.9 13.0 13.8 12.3 13.6	100 (69) 94 (0) 11 (83) 100 (1) 100 (1) 100 (1) 100 (0)	99 82 89 97 96 93 92
AT1G01610	sn-2-Glycerol-3-Phosphate Acyltransferase (GPAT4)	Wax Biosynthesis	373	99	14.9	100 (20)	79
AT1G01610/AT4G00400	sn-2-Glycerol-3-Phosphate Acyltransferase (GPAT4, GPAT8)	Wax Biosynthesis	330	89	12.1	98 (13)	83
AT5G57800	Eceriferum 3 (CER3)	Wax Biosynthesis	575	98	11.5	58 (19)	76
AT2G38110	sn-2-Glycerol-3-Phosphate Acyltransferase (GPAT6)	Wax Biosynthesis	375 381 332	97 96 92	15.2 11.3 12	100 (18) 97 (6) 91 (9)	97 95 95
AT1G68530	Ketoacyl-CoA Synthase (KCS6)	Wax Biosynthesis	96 142 103	88 80 72	12.8 11.3 10.6	100 (14) 73 (8) 48 (4)	78 88 52
AT2G47240	LACS (LACS1)	FA Activation	167	75	DS*	0 (12)	51
AT3G45140	Lipoxygenase (13-LOX, LOX2)	JA Biosynthesis	561	72	14.3	100 (2)	83
AT5G42650	Allene Oxide Synthase	JA Biosynthesis	381	32	13.1	100 (6)	65

a

Arabidopsis thaliana gene identification number. See Supplementary Dataset S1 for additional details and full list of lipid proteins with Cys PTMs.

b

Descriptions from AraLip and/or TAIR with common protein abbreviations.

c

General pathway and/or function related to lipid metabolism from AraLip and/or TAIR.

d

Amino acid position in A. thaliana representative model protein sequence.

e

Percentage identity of Cys in genome sequence alignments. Table is sorted from high to lowest relative conservation.

f

Theoretical pK_a of Cys residue calculated using PROPKA. DS* = disulfide bond.

g

Percentage buried of Cys residue (0% = exposed to 100% = buried) and solvent accessible surface area (square angstroms).

h

The mean percentage conservation of residues within 6 angstroms of the –SH group of Cys using AlphaFold protein structures.

Structural Tendencies and Sequence Conservation of Cys PTMs

The functionality of an individual amino acid within a protein is the product of multiple inherent properties (e.g. size, charge and polarity) and its interaction within a microenvironment (e.g. location in a protein, surrounding amino acids, solvent accessibility and cell/solvent properties). Based on recent Cys PTM proteomics datasets (Table 1, Supplementary Table S1, and Supplementary Dataset S2), the majority of Cys residues (>90%) within the plant proteome were not modified (S-sulfenylated, S-nitrosylated and S-acylated) under the conditions tested and the thiols that were modified covered diverse proteins and pathways (discussed in a ‘Cysteine PTMs in Lipid Metabolism Proteins and Pathways’ section). In this section, we summarize multiple in silico analyses comparing PTM versus unmodified (or no PTM) Cys in terms of their propensity to be buried, solvent accessibility, pK_a, amino acid composition in primary sequence and in structures and evolutionary conservation of amino acids nearby each Cys in its primary sequence and associated homology-based AlphaFold structure. Depending on the type of analysis, the data are presented in one of two sets: (i) the full Cys PTM proteome set (all proteins with 1+ Cys PTM) or (ii) the AraLip Cys PTM set (proteins in the AraLip dataset with 1+ Cys PTM) to see whether lipid metabolism proteins exhibit unique PTM physiochemical properties. Additional details on all methods and datasets used are provided in Supplementary Methods and Datasets.

Cys PTM residues occupy unique solvent exposure microenvironments

Cys residues that undergo PTMs are expected to be found in protein microenvironments such that their –SH group has an increased probability of its valence electrons interacting with, at least transiently, another reactive molecule (e.g. H₂O₂ for S-sulfenylation, nitric oxide (NO) for S-nitrosylation and a modifying enzyme for S-acylation). To examine whether modified Cys were more accessible to reactive molecules relative to unmodified Cys, two programs were used to evaluate the solvent accessibility of each Cys (PROPKA and (solvent accessible surface area (SASA)) ; Supplementary Methods). As an example, Fig. 1A shows the AlphaFold structure (UniProt Q43307) of a plastid-localized glycerol 3-phosphosphate acyltransferase (GPAT, AT1G32200) that converts glycerol 3-phosphate to lysophosphatidic acid. In this structure, PROPKA determined that Cys410 was 0% buried (i.e. side chain exposed to the environment), while four other Cys (255, 267, 278 and 368) were between 70 and 100% buried (i.e. Cys side chain surrounded by other residues). Notably, only Cys278 was detected with a PTM (sulfenic acid, Supplementary Dataset S2). This percentage buried calculation was applied to the entire Arabidopsis AlphaFold Proteome (25,000+ proteins, comprising over 206,000 Cys) and grouped according to the type of Cys modification (or no modification) in Fig. 1B, Supplementary Fig. S1 and Supplementary Dataset S2. In the full proteome, the median percentage of buried distribution of S-acylated Cys was 0%, that of Cys with no PTM was 18%, that of S-nitrosylated Cys was 57% and that of S-sulfenylated Cys was 72%. Similar trends were observed in the AraLip set although the values differed: S-acylated Cys 0%, no PTM 48%, S-nitrosylated Cys 47% and S-sulfenylated Cys 70%. The low buried % of S-acylated Cys in both sets suggests that the majority of these Cys side chains were exposed to the surface or open regions/pockets/tunnels within their respective structures. In contrast, the values for S-nitrosylated Cys were more evenly distributed, whereas unmodified Cys had clusters of residues at 0% and 100% buried (Fig. 1B versus Supplementary Fig. S1). Finally, S-sulfenylated residues were consistently the most buried. The overall trends for percentage buried of Cys were consistent with an alternative calculation of SASA (Shrake and Rupley 1973) on the same structural models (Pearson r = −0.59, Supplementary Table S3). Analyzing the full proteome, the median surface area for S-sulfenylated Cys was 3.9 Ų, that for S-nitrosylated Cys was 10.3 Ų, that for S-acylated Cys was 21.9 Ų and that for unmodified Cys was 19.0 Ų. Similar trends were observed in the AraLip subset although the median non-modified Cys was considerably lower at 12.3 Ų. Notably, median values are reported here (and other analyses reported later) to account for the skewed distribution observed within these datasets. Furthermore, the statistical analyses represent an evaluation of whether there is a significant difference between the medians of Cys PTM versus unmodified Cys (i.e. post hoc Dunn’s test after a Kruskal–Wallis H-test) with caveats further discussed in the Supplementary Methods.

Fig. 1

Summary of the structural composition and sequence conservation of Cys PTM. (A) Left: Protein surface view of the plastid GPAT (AlphaFold model, UniProt Q43307). Five Cys are marked in structure with % buried value based on its –SH group solvent exposure. Cys 278 was detected as S-sulfenylated (Table 1), while other Cys residues showed no PTM. (B) Violin plot distribution of the percentage buried of each Cys residue in the entire AlphaFold proteome (all) and for AraLip-specific proteins as well as type of PTM or no PTM (analyzing over 25,000 proteins and over 206,000 Cys in total). Additional details are provided in Supplementary Fig. S1 and Supplementary Dataset S2. (C) Secondary structure view of the plastid 13-lipoxygenase (LOX-2, AlphaFold model, UniProt P38418) with side chain residues colored according to their closest distance from the sulfur group (yellow) of Cys 561 (carbons black): under 4 Å, green (closed to Cys); between 4 and 6 Å, pink and between 6 and 8 Å, light blue (furthest from Cys). (D) Differences in the relative amino acid composition of each PTM type versus no PTM of amino acids under 6 Å to each respective Cys. Amino acids were group according to common classification: (−) Asp, Glu; (+) Lys, Arg, His; Polar: Asn, Gln, Ser, Thr; Cys; nonpolar aromatic: Phe, Tyr, Trp; nonpolar aliphatic: Ala, Gly, Ile, Leu, Met, Pro, Val. Additional details are provided in Supplementary Fig. S6 and Supplementary Dataset S3. (E) Distribution of Cys conservation (% identities) within Arabidopsis reference protein relative to sequence alignments of top homolog from up to 124 genomes (left-to-right, n = 1,442, 117, 2,880, 28,726, 48, 41, 144, 6,035 Cys). (F) Distribution of sequence conservation (% identities) of all amino acids within 6 Å of each Cys (left-to-right, n = 14,008, 10,541, 24,059, 241,207, 440, 395, 1,131, 50,347 amino acids). (G) Distribution of sequence conservation (% identities) of amino acids at a set position within primary sequence surrounding each Cys. Only the middle two quartiles are shown for simplicity (see Supplementary Dataset S4 for more details including number of amino acids analyzed). Asterisks represent P < 0.05 post hoc Dunn’s test after a Kruskal–Wallis H-test relative to unmodified Cys. All plots in this figure use the same PTM legend as shown in (A).

Collectively, using these two metrics, % buried and SASA, it was clear that exposure of the –SH groups within the Cys PTM varied according to the type of modification, especially S-sulfenylated Cys as ‘likely buried’ and S-acylated Cys as ‘likely exposed’. Notably, of the seven readily ionizable amino acids (Arg, Asp, Cys, Glu, His, Lys and Tyr), Cys is considered the most nonpolar using hydrophobicity indices (Poole 2015), i.e. most likely to be found in the interior of a protein. While our focus is Cys residues, an analysis of these other ionizable residues within the Arabidopsis proteome using the % buried metric is consistent with these previous findings (Supplementary Fig. S3). However, Cys with an increased solvent exposure may be more likely to undergo S-acylation that requires catalysis by an interacting enzyme, although exposure is not exclusive to S-acylated Cys. Additionally, an analysis of the relative position of each Cys within its respective protein showed the positional frequency of Cys varied by modification type (Supplementary Fig. S4). While unmodified Cys were uniformly distributed between the N-terminus, middle and C-terminus regions, S-nitrosylated and S-sulfenylated Cys showed ‘Gaussian-like’ distributions with most Cys at the center/middle of the protein sequence and fewer at the termini. S-acylated Cys on the other hand were over-represented at both the N- and C-termini. For S-acylation, this pattern is likely due to acylation being associated with protein–membrane anchoring (discussed later). Of course, the homology-based structures used for these analyses are static and ultimately models (albeit with an impressive level of quality relative to experimentally determined structures in most cases) (Jumper et al. 2021, Varadi et al. 2022, Yin et al. 2022). Exposure to cellular environments results in proteins dynamically moving, i.e. bending, flexing, vibrating and undergoing conformational changes (Henzler-Wildman and Kern 2007), such that a –SH may vary in its exposure to the environment. This may be one explanation for why GPAT only had one of its five Cys detected as PTM (and why the majority of Cys appear resistant to PTMs). It may also explain differences in proteins identified with PTMs under varying conditions (i.e. Supplementary Dataset S1, various S-sulfenylation sets) (Waszczak et al. 2014, Akter et al. 2015, De Smet et al. 2019, Huang et al. 2019, Wei et al. 2020). Regardless, applying the same constraints on each subset show that Cys location and solvent accessibility vary with PTM. Additionally, despite the AraLip subset only being about 3% of the proteome, the trends observed within this subset were consistent with the larger proteome analysis. Ultimately, based on these metrics alone, it remains unclear why most Cys do not undergo PTMs. To further investigate this trend, other factors such as pK_a and amino acid composition were investigated.

The pK_a parameter of an amino acid side chain is a measure of its ability to lose a proton or be ionized, e.g. the intrinsic pK_a for Cys at physiological pH is ∼8–9 depending on the method of determination. In proteins, pK_a values for ionizable amino acids can vary widely within microenvironments with pK_a’s shifting due to contributions of nearby hydrogen bonds, columbic interactions and desolvation effects (Thurlkill et al. 2006, Pace et al. 2009). Theoretical calculations of pK_a for residues considering these microenvironments within proteins are complex and still a work in progress (Sham et al. 1997, Nielsen et al. 2011, Olsson et al. 2011). The PROPKA program was used to address whether a shift in pK_a values was associated with each type of PTM. As a baseline, all ionizable amino acids within the entire proteome were evaluated (Supplementary Fig. S5). The median pK_a value was mostly consistent with the intrinsic values: Arg 12.5 (intrinsic pK_a 12.5), Asp 3.8 (3.8), Cys 10.1 (9.0), Glu 4.6 (4.5), His 6.3 (6.5), Lys 10.5 (10.5) and Tyr 10.8 (10.0). Disulfide bonds were not included for Cys median analysis as they were given a pK_a value of 99.99 by PROPKA. However, the median pK_a for S-sulfenylated Cys, S-nitrosylated Cys, S-acylated Cys and no PTM in the full proteome shifted to 11.4, 11.0, 9.6 and 9.1, respectively (Supplementary Fig. S5). These values generally correlated with the degree of % buried, i.e. the more buried residues of S-sulfenylated Cys (Fig. 1B) were associated with higher pK_a reflective of the dehydration effect. It was surprising not to find a shift in lower pK_a values for Cys PTMs as may be expected with an increased propensity of its thiolate or reactive form. Within the GPAT example in Fig. 1A, all five calculated pK_a values were greater than 10 (Supplementary Dataset S2). However, these results reinforce the challenges of pK_a predictions. As discussed in the section on ‘amino acid composition’, the composition varies widely among Cys PTM. In terms of disulfide bonds and pK_a, rarely were other Cys residues found in proximity to S-sulfenylated and S-nitrosylated Cys and therefore have few predicted disulfide bonds (0.6% and 2.1% of these modified Cys residues in the full proteome, respectively). However, for S-acylation, there were numerous Cys predicted to be disulfide bonds (33.1% of Cys), notably, a larger relative percentage than the no PTM set (14.7% of Cys). Considering individual Cys residues (discussed later) and these pK_a values, at this time, the pK_a calculations cannot be used in isolation to predict modified Cys residues.

Cys PTM residues are physically surrounded by unique amino acid compositions

The distinct microenvironments observed among the different Cys PTMs (via % buried, solvent exposure and pK_a analysis) may create distinct ‘reaction pockets/regions/tunnels’ that would be supported by distinct amino acid compositions. To further examine these microenvironments, both in the context of AraLip proteins and the broader Arabidopsis proteome, the relative composition of amino acids was determined using two approaches: structural models and primary sequences. First, using the AlphaFold models, the surrounding amino acids of every Cys was determined at set distances: under 4 Å, between 4 and 6 Å, 6 and 8 Å and 8 and 10 Å. As an example, Fig. 1C shows the structure and nearby residues of S-acylated Cys561 in 13-lipoxygenase involved in oxylipin signaling (Uniprot P38418, AT3G45140). While only four residues were within 4 Å, expanding the ‘reaction pocket’ to 8 Å revealed 20 residues. To examine the relative amino acid frequency for each type of PTM, residues under each distance criterion, for each Cys within that PTM subset (e.g. S-sulfenylated Cys), were summed and normalized to the no PTM subset (Supplementary Dataset S3). The results under 6 Å are summarized later for amino acid classes (Fig. 1D). While the trends generally held consistent at each distance criteria (Supplementary Fig. S5), the relative differences considerably narrowed further away from the –SH group. For S-sulfenylated Cys, the greatest absolute differences were observed in nonpolar aliphatic amino acids (+10%) and Cys (−6%) with minor differences in polar, basic and nonpolar aromatic groups (−1 to −2%). The general trends in S-nitrosylated Cys matched S-sulfenylated Cys (e.g. Cys −6%); however, nonpolar aliphatic groups were not as prominent (only +4%) and acidic groups were also more common (+4%). Residues surrounding S-acylated Cys were quite different, with nonpolar aliphatic and basic residues less common (−1 to −2%) and polar and Cys more common (+1 to 2%). This representation of amino acid classes is relatively consistent with S-sulfenylated and S-nitrosylated residues more likely to be on the protein interior relative to S-acylated Cys. Analysis of the smaller AraLip dataset showed some differences in relative distribution relative to the full set (Supplementary Dataset S3), notably in the proportion of acidic and basic residues, which may reflect the specific proteins in this subset (discussed later).

An analysis of amino acid frequency in primary sequences (using the full proteome) surrounding each Cys was also performed and compared to the structural approach. In the positional analysis, all amino acids were represented at least once in all positions for all three Cys PTM datasets, supporting the diversity of sequences associated with Cys modification. Supplementary Fig. S7 shows a heat map summary of amino acid enrichment and amino acid class environment for each Cys PTM set relative to no PTM (additional details are provided in Supplementary Dataset S4). For S-sulfenylated Cys relative to unmodified Cys, basic residues were almost uniformly over-represented at all positions (+10 to +60%), while nonpolar aromatic (−40 to −15%), polar (primarily Ser, −38 to −11%) and additional Cys (−95 to −65%) were under-represented. For S-nitrosylation, the amino acid trends were relatively similar although there was an increase in the relative abundance of acidic residues near the modified Cys (+10 to +73% with the majority at positions −4 to −4). For S-acylation, the relative position composition was more similar to unmodified Cys than to other Cys PTM types. An exception was the presence of additional Cys (+14 to +118% depending on the position).

In the original S-sulfenylation proteomics study, an amino acid over-representation analysis was performed from −6 to +6 residues relative to Cys PTM (Huang et al. 2019). Our primary sequence frequency analysis was consistent with their observations of basic residues (His and Lys) enriched around Cys PTM. However, structural models showed that most of these His and Lys were oriented in such a manner that they were not in proximity, at least in terms of over-representation relative to other residues and relative to unmodified Cys. Of course, there is a degree of rotational torsion for most amino acids, and larger protein movements can change the interacting residues. For example, less than 30% of S-sulfenylated Cys even had a His and/or Lys within 6 Å. While the number of His/Lys increased at larger distance cutoffs, they were only slightly enriched relative to Cys with no PTM (Supplementary Dataset S3). Therefore, although it has been proposed that these basic residues would be preferred Cys hydrogen-binding partners effectively increasing their sensitivity to oxidation (Roos et al. 2013), the structural models suggest a more complex microenvironment. For the S-nitrosylation set, our positional and structural frequency analysis came to similar conclusions as the motif-enrichment analysis performed in the original study (Hu et al. 2015), i.e. acidic residues (Asp and Glu) were enriched nearby Cys PTM, while basic residues were slightly under-represented (Supplementary Fig. S6) in contrast to previous proteomics studies on S-nitrosylation finding an enriched acid–base motif. Notably, a structural-based study using experimentally determined structures from the Protein Data Bank showed the complexity of attributing a specific composition and microenvironment to S-nitrosylation (Marino and Gladyshev 2010b). The acid–base motif was less attributed to activating the Cys but rather engaging in protein–protein interactions resulting in trans-nitrosylation. Finally, within the S-acylation study, motif and specific residue enrichment was inconclusive (Kumar et al. 2020). It has been proposed that S-acylation may be more stochastic in nature and a function of accessible Cys (Rodenburg et al. 2017). However, in general, S-acylation showed the opposite relative composition as S-sulfenylation and S-nitrosylation, and the increased presence of other Cys was notable and a reflection of specific proteins enriched in lipid metabolism (discussed later) and the broader proteome. In conclusion, the comparative positional and structural approach shows the complexity in defining specific Cys PTM microenvironments. As the microenvironments varied substantially, this would suggest that the specific Cys residues and the residues surrounding Cys PTM may vary in modes of activation and may also not be evolutionarily conserved.

Cys PTMs are more likely to be evolutionary conserved

Amino acid sequence conservation is often, but not always, an indication that a residue is important to a protein’s structure and/or function (Whisstock and Lesk 2003, Fowler et al. 2010). Relative to other residues, Cys often exhibits the most extreme conservation patterns, i.e. majority highly conserved or poorly conserved (Marino and Gladyshev 2010a). To evaluate whether Cys PTMs in planta have similar extreme conservation patterns and/or differed in conservation relative to unmodified Cys, a sequence conservation analysis was performed using the predicted protein sequences from 124 publicly available genomes across the plant kingdom (Supplementary Table S2, Supplementary Dataset S5) with analysis details provided in Supplementary Methods. Briefly, for each Arabidopsis protein analyzed, the top protein sequence homolog was identified using BLAST. Global sequence alignments were then performed using the multiple sequence comparison by log-expectation algorithm (Edgar 2004) and percent amino acid identities calculated. Due to the computational power required for large global sequence alignments, only a subset of the proteome was used for conservation analysis, i.e. 3,147 proteins which had to satisfy the criteria of containing 1+ modified Cys (S-acylation, S-nitrosylation and S-sulfenylation) somewhere in their primary sequence. To establish a baseline conservation level within this dataset, the conservation of all individual amino acids was performed. Among these 3,147 alignments, Cys was the third most conserved amino acid (n =28,822 Cys with a median value of 90% after Trp 94% and Gly 91%, Supplementary Fig. S8). Furthermore, a distribution plot of Cys conservation values showed that Cys did not have the extreme conservation patterns in plants and was one of the most consistently conserved residues (in both the larger set and AraLip-specific proteins, Supplementary Fig. S8). Given these overall high conservation rates, it was initially unclear that PTM-Cys would show differences in conservation level relative to no PTM. However, within AraLip proteins, S-sulfenylated Cys (mean = 80%, median = 91%, n = 48) and S-acylated Cys (mean = 77%, median = 88%, n = 144) showed slightly higher levels of conservation (and less low conservation outliers) on average relative to unmodified Cys (mean = 72%, median = 87%, n = 1182) (Fig. 1E, P < 0.05). The larger variation in S-nitrosylated Cys conservation meant that it could not be distinguished from other subsets (mean = 72%, median = 89%, n = 43). Similar results were observed analyzing the set of 3,147 proteins where S-sulfenylated and S-acylated residues were relatively more conserved on average (Fig. 1E).

To test whether these trends held consistent within genome subsets, an additional analysis was performed comparing Arabidopsis to one of four subsets: eudicots (79 species), monocots (26 species), non-vascular land plants (5 species) and algae (13 species) (Supplementary Fig. S9, Supplementary Table S2). Interestingly, S-sulfenylated Cys residues within the AraLip protein set retained an increased mean level of conservation (but not the median level of conservation) versus unmodified Cys in eudicots (89.3 versus 80.8%), monocots (84.0 versus 69.3%), non-vascular land plants (80.4 versus 72.8%) and algae (45.1 versus. 42.1%). Results were similar for S-sulfenylated Cys in the full protein set, although the algae mean values were similar. The similarity in median values but higher overall mean values within S-sulfenylated Cys reinforces that these residues tend to be more evolutionarily conserved. The evolutionary conservation in other PTM was less consistent, although S-acylation Cys in AraLip and full protein sets held generally higher conservation levels in eudicots and monocots. The possibility of different rates and roles of PTM across the plant system is unexplored and intriguing from a biochemical evolutionary perspective.

It is intriguing to hypothesize that the higher conservation levels of PTM-Cys versus unmodified Cys support an enhanced functional importance. However, this observation comes with several caveats. Foremost, the sequence analysis only uses the top homolog (via BLAST comparison) from each genome, and the genome clades and families are not uniformly represented [e.g. 79 eudicots (30 Brassicaceae) versus 26 monocots (20 Poaceae) versus. 5 non-vascular land plants versus 13 algal species]. However, this analysis expands upon previous efforts to identify conserved Cys across the plant kingdom (Moore et al. 2020). Second, most of the genomes use predicted protein sequences and do not account for genetic variation within the genome, nor alternative splicing in addition to variations in isoform number and sequence-driven specialization. Third, as described briefly in the Introduction section, Cys may have one of several structural and functional roles. The role of each of the 4,000+ modified Cys and 25,000+ unmodified Cys (considering only the 3,000+ proteins analyzed for conservation) has not been defined. Furthermore, the increased Cys conservation is also associated with increased surrounding residue conservation as well (overall and when comparing modified versus unmodified Cys). Nevertheless, it is possible that many of these Cys with higher levels of conservation are critical for structure–function through enhanced –SH reactivity and these PTMs are associated with this general enhanced reactivity. It has been observed that surface Cys are generally less conserved than buried Cys (Marino and Gladyshev 2010a). Pearson correlation analysis (Supplementary Table S3) of Cys conservation showed very low positive correlation with % buried (r = 0.18) and predicted pK_a (r = 0.29), but a moderate negative correlation with SASA (r = −0.46, lower solvent accessibility associated with higher conservation). Ultimately, at this large of a comparison scale, it is impractical that a single physiochemical parameter can predict the functional importance and conservation of Cys. Lastly, for S-acylation, many of the Cys PTMs appeared clustered around other Cys (Fig. 1D, Supplementary Fig. S7) previously shown to have higher conservation rates in other systems (Marino and Gladyshev 2010a).

Amino acids in proximity to Cys PTM are more likely to be evolutionarily conserved

Often not just a single residue, but rather a region of amino acids is conserved to retain a critical function within a protein. To address whether the increased conservation percentage of Cys PTM versus unmodified Cys held true in surrounding regions, two types of relative conservation analysis were performed: (i) the primary sequence from −8 to +8 residues of each Cys (Fig. 1G) and (ii) residues identified within AlphaFold structures at a set distance from Cys (Fig. 1F). For the primary sequence analysis, each position was independently assessed for conservation, and it was determined that in general, the residues surrounding S-sulfenylation and S-nitrosylation Cys were more conserved than residues surrounding S-acylation or no PTM (Fig. 1G). For example, in the AraLip dataset, the median percent conservation of all positions was 87.6% for S-sulfenylation (n = 736 total residues for −8 to +8 positions around 46 Cys), 87.0% for S-nitrosylation (n = 672 residues), 72.4% for S-acylation (n = 2,272 residues) and 82.1% for no PTM (n = 19,010 residues). Similar results were obtained for the median percent conservation of full protein dataset: 89.4% for S-sulfenylation (n = 22,982 residues), 88.6% for S-nitrosylation (n = 17,856 residues), 77.2% for S-acylation (n = 45,135 residues) and 81.3% for no PTM (n = 380,870 residues). Additionally, the higher the conservation value of an individual Cys residue (regardless of whether it was post-translationally modified or not) generally was associated with a more conserved surrounding primary sequence (Supplementary Table S3). As an example, for S-sulfenylation Cys conserved between 95 and 100% (n = 11 Cys in AraLip, n = 452 Cys in the full set) the median percent conservation was ∼84%, but for Cys conserved between 85 and 95% (n = 19 Cys in AraLip, n = 535 Cys in the full set), the median percent conservation was ∼79%. Furthermore, the extent of increased conservation varied for each type of Cys PTM versus no modification. For example, within the AraLip dataset, for Cys that were highly conserved (>95% identities), the average conservation value of −8 to +8 primary sequence was 83% for S-sulfenylation, 85% for S-nitrosylation, 69% for S-acylation and 77% for no PTM.

While the primary sequence information is important to analyze, it does not give a complete picture into the microenvironment for each Cys. Comparing the residues in the primary sequence −8 to +8 position relative to the amino acids under 6 Å in AlphaFold models, there is a minimal overlap (Supplementary Fig. S10). For example, in the S-sulfenylation Cys set, while positions −1 to +1 overlapped ∼100% of the time with residues within 6 Å, other positions had much lower overlap reinforcing the need for a structural-based analysis: 16–20% overlap (−8 to −5 positions relative to Cys), 45–60% (−4 to −2), 100% (−1 to +1), 28–38% (+2 to +4) and 16–20% (+5 to +8). Similar results at each position were obtained for the two other Cys PTM subsets and unmodified Cys subset (Supplementary Fig. S10). Supplementary Fig. S10 also shows a detailed example, using the lipid transfer protein 2 (LTP2), in which the mean percentage conservation is different depending on which method is used to calculate it (e.g. 63% using the structural method versus 42% using the sequence-only method). Regardless of these variations in absolute conservation percentage, considering residues within 6 Å still showed a higher median conservation for S-sulfenylated and S-nitrosylated Cys (89 and 90% in AraLip proteins, respectively) than S-acylated or no PTM (83 and 84%, respectively, Fig. 1F). Primary sequence consideration is still very helpful in the context of defining the secondary structure for each Cys.

Overall, the sequence conservation analysis provides evidence that these Cys PTM sites are more likely to be evolutionarily conserved than unmodified Cys. If residues surrounding Cys PTM are also more conserved than unmodified Cys, then this may suggest that these Cys PTMs possess critical functions, although it may not be due to their ability to undergo PTM but rather other essential catalytic or structural functions. As discussed later, many of these Cys PTM sites are associated with catalysis. Future mechanistic studies, e.g. via mutagenesis, are necessary to identify the specific function of each Cys PTM. Conversions to Ala and/or Ser are conservative replacements that would address the thiol functionality while likely minimizing other structural effects of these substituted amino acids. Ultimately, while these Cys PTM trends will be interesting to explore further, an examination is needed at the metabolic pathway and protein level to gain additional insights into the role and importance of Cys PTM in lipid metabolism and beyond.

Cysteine PTMs in Lipid Metabolism Proteins and Pathways

Plants use cellular and molecular signaling mechanisms to efficiently adapt to changes in the environment. PTMs are an important molecular mechanism enabling dynamic control at the protein level, i.e. acting as molecular switches to modify the protein structure and function. Thiol-based PTMs of Cys are driven by the unique reactivity properties within Cys (Fig. 1) and linked to fluctuations in the redox state of cells. Each signaling molecule, via a Cys PTM, may result in a distinct change in activity, stability, interactions and subcellular location. To gain a better understanding of the role of these molecular switches, the proteomic data described in ‘Structural Tendencies and Sequence Conservation of Cys PTMs’ were combined with an additional S-sulfenylation dataset (Wei et al. 2020) and three persulfidation datasets (Aroca et al. 2017, Jurado-Flores et al. 2021, 2023) and specifically applied to the AraLip dataset (Li-Beisson et al. 2013) to identify lipid metabolism proteins that are susceptible to these modifications (Hu et al. 2015, Huang et al. 2019, Kumar et al. 2020). Using this approach, 246 acyl lipid metabolism proteins (of 775 total proteins) were shown to be susceptible to persulfidation, S-acylation, S-nitrosylation and/or S-sulfenylation with varying levels of overlap (Fig. 2B, details in Supplementary Dataset S1). Furthermore, within these proteins, 20 Cys sites showed more than one PTM with varying biophysical parameters and conservation rates. These proteins were analyzed for gene ontology (GO) and pathway enrichment (Supplementary Dataset S6). GO addresses enrichment in specific biological processes, molecular functions or localization in specific cellular compartments (Fig. 3) (Tian et al. 2017), while pathway analysis can identify functional enrichment within the analyzed groups (Kanehisa et al. 2023). Representative Cys PTM-enriched pathways and proteins are highlighted here (Table 1, Fig. 4) integrating the structural and conservation analyses with functional outcomes. While each redox PTM study (persulfidation, S-sulfenylation and S-nitrosylation) analyzed the proteome in a control and enhanced oxidative stress environment/genotype, the data were quite variable across studies, making it challenging to draw conclusions about specific stress-induced Cys modifications (Supplementary Dataset S1). Ultimately, although the role of these modifications has not been determined for most Cys, we have begun to explore the implications of these structural changes by identifying the source and stimulus of these signals in the cell.

Fig. 2

Plant cysteine modification pathways and comparisons of lipid metabolism protein PTMs. (A) Persulfidation modifications can be initiated by the production hydrogen sulfide (H₂S) during the photosynthetic sulfate assimilation pathway. S-sulfenylation can be the result of ROS production during light-dependent photosynthetic reactions. S-nitrosylation is a result of RNS produced due to environmental stress. S-acylation is a cellular process that is catalyzed by PATs, and de-acylation is catalyzed by APTs and α/α hydrolase domain-containing protein 17-like APTs (ABAPTs). All these PTMs are reversible and initiated by signals within the cell. (B) In Arabidopsis, 183 lipid metabolism proteins were susceptible to persulfidation, S-acylation, S-nitrosylation and/or S-sulfenylation (Hu et al. 2015, Aroca et al. 2017, Huang et al. 2019, Kumar et al. 2020, Wei et al. 2020, Jurado-Flores et al. 2021, 2023). This Venn diagram shows that most proteins are susceptible to more than one PTM, but there are also several proteins that are only susceptible to one under the conditions tested. Details on specific proteins and Cys PTM residues are provided in Supplementary Dataset S1. Figure created using BioRender.

Fig. 3

GO and pathway analysis were used to determine if there was biological functional enrichment within each PTM dataset. (A) There was a significant enrichment of Cys PTM lipid metabolism proteins involved in several biological processes and molecule functions. Most of these molecular events were related to fatty acid metabolism. In addition, there was a significant enrichment of proteins in these datasets that were localized in the chloroplast, peroxisome and ER. See Supplementary Dataset S6 for additional details on GO analysis and Venn diagram. (B) Pathways analysis showed that most enzymes involved in the JA biosynthesis, cuticular wax and β-oxidation pathway were susceptible to at least one PTM. Most of the enzymes were susceptible to more than one PTM.

Fig. 4

Representative protein structure models containing one or more Cys PTM. (A–F) Six representative AlphaFold predicted models were visualized using Pymol software. Unmodified Cys (light orange) and Cys PTM (dark magenta) are marked in both full and zoomed-in structures. The evolutionary conservation % (in amino acid identities) of each Cys and its position in Arabidopsis are shown in a bar plot with all PTMs detected (n = S-nitrosylation, s = S-sulfenylation and a = S-acylation). Each Cys zoomed-in structure centers on one Cys PTM (or two in the case of LTP2 due to proximity) and residues within 6 Å to illustrate diverse microenvironments: acidic (Asp, Glu; blue), basic (Arg, His, Lys; green), polar (Asn, Gln, Ser, Thr; yellow) and all other nonpolar residues (gray). Proteins: (A) MAPK6, (B) MFP2, (C) PKT (or KAT1), ketoacyl-CoA thiolase, (D) KASI, (E) SAD/FAB2, and (F) LTP2 with two disulfide bonds. See Table 1 and Supplementary Dataset S1 for additional details.

Oxidative Cys PTMs in β-oxidation and jasmonic acid biosynthesis

GO and pathway analysis showed that the β-oxidation and jasmonic acid (JA) pathway contain several genes that are susceptible to PTMs and primarily via chemical oxidation (Fig. 3). β-oxidation is a multistep, metabolic process by which fatty acid molecules are broken down. Long-chain fatty acids are converted into acyl-CoA chains and into progressively smaller fatty acids (Baker et al. 2006). Seed germination, breaking of seed dormancy (Pinfield-Wells et al. 2005, Footitt et al. 2006, Goepfert and Poirier 2007) and embryo development (Rylott et al. 2003, 2006) require β-oxidation. JA biosynthesis also requires β-oxidation (Raza et al. 2021). JA is a phytohormone in plants that is involved in development (Wasternack and Hause 2013, Ghorbel et al. 2021), defense (Vijayan et al. 1998, Macioszek et al. 2023) and response to unfavorable environmental conditions (Raza et al. 2021). Additionally, previous work has suggested that ROS- and JA-mediated signaling pathways may interact to regulate development or defense responses (Hamann et al. 2009, Denness et al. 2011). The potential involvement of Cys PTMs in β-oxidation and JA biosynthesis points to the complexity of essential cellular signaling processes. Representative Cys PTM proteins are discussed later in the context of these two pathways and specific types of Cys PTMs.

One of these cellular signals, hydrogen peroxide (H₂O₂), the most stable form of ROS, is produced during photosynthesis, photorespiration, respiration, β-oxidation and biotic and abiotic stress (Slesak et al. 2007, Chae et al. 2023). Although an accumulation of ROS can result in oxidative stress, this molecule also acts as an essential cellular signal at relatively low concentrations (Sies 2017). One signaling function of ROS is oxidation of specific protein Cys thiols to form an inter- or intramolecular disulfide bonds, sulfenic acid, sulfinic acid and/or sulfonic acid (Fig. 2A) (Chae et al. 2023). S-sulfenylation is reversible and can change protein conformation, activity, stability, interactions and subcellular location (Huang et al. 2019). As a result, this process can act as a redox-based ‘on-and-off’ switch for targeted proteins and regulate downstream stress signaling pathways. Forty-seven lipid metabolism proteins have S-sulfenylated Cys (Fig. 2B) within the two datasets analyzed (Huang et al. 2019, Wei et al. 2020). Over 75% of these proteins have also predicted sites for other PTMs including persulfidation, S-nitrosylation and S-acylation. A significant number of S-sulfenylated proteins are involved in β-oxidation, fatty acid biosynthesis and catabolism, and lipid oxidation and are located in the chloroplast or peroxisome (Fig. 3, Supplementary Dataset S1, Supplementary Dataset S6). Two proteins with highly conserved S-sulfenylated Cys include mitogen-activated protein kinase 6 (MAPK6, Cys201 at 100% amino acid identities, Fig. 4A) and multifunctional protein 2 (MFP2, Cys251 at 95% and Cys493 at 97%, Fig. 4B). Notably, the S-sulfenylated peptide attributed to MAPK6 (detected in both control and H₂O₂-treated cells) here could be mapped to one of several Arabidopsis MAPKs due to a shared conserved motif, and another isoform MAPK4 was discussed extensively in the original S-sulfenylation study (Huang et al. 2019). Regardless of whether one MAPK, both MAPKs or even additional MAPKs undergo S-sulfenylation, each MAPK has mechanistic links to redox signaling and activation by ROS (Liu and He 2017). In terms of lipid metabolism, MAPK6 is required for the biological function of MYB41, a transcription factor controlling cell expansion and cuticle deposition in response to abiotic stress (Lippold et al. 2009, Hoang et al. 2012), while MAPK4 regulates salicylic acid– and JA/ethylene-dependent responses (Brodersen et al. 2006). On the other hand, MFP2, with two Cys modified in both control and H₂O₂-treated cells, is involved in peroxisomal β-oxidation and required for seedling establishment (Rylott et al. 2006). Although it is not known if MFP is regulated by S-sulfenylation, MFP forms a complex with 3-ketoacyl-CoA thiolase KAT2, which has been shown to be redox regulated (Pye et al. 2010). Notably, both KAT2 and KAT1/PKT (Fig. 4C, discussed later) had highly conserved Cys (Cys138 and Cys 130, respectively, at 100%) that were detected as S-nitrosylated. The interplay between different S-sulfenylation and S-nitrosylation modifications on interacting proteins, or in cases on the same protein, has not been thoroughly investigated.

Hydrogen sulfide (H₂S) is a colorless lipophilic gas signaling molecule proposed to be involved in all life processes in plants (Wang et al. 2021). H₂S is produced in cells through metabolism of sulfur-containing molecules and can interact with the protein Cys thiol to form a persulfide (R-SSH) group through a process called persulfidation. Persulfides have enhanced chemical reactivity relative to thiols. Upon formation, they likely impact the protein structure and function with reports of activation, inactivation and modified protein–protein interactions in cells (Wang et al. 2021). Like S-sulfenylation, at high levels of ROS, R-SSH can also be further oxidized [to perthiosulfenic acids (R-SSOH), perthiosulfinic (R-SSO₂H) or perthiosulfonic (R-SSO₃H)] providing an antioxidant protective mechanism. In Arabidopsis, over 3,000 proteins were susceptible to this PTM under normal conditions (Aroca et al. 2017). Additional Cys have been identified as susceptible to persulfidation under nitrogen-deficient and drought conditions (Jurado-Flores et al. 2021, 2023). In the study by Aroca et al., the authors discussed primary metabolism pathways targeted by persulfidation including the tricarboxylic acid cycle, glycolysis and the Calvin cycle. Out of the 775 lipid metabolism proteins in Arabidopsis, 183 have been shown to be persulfidated across all three datasets (Supplementary Dataset S1). Notably, 74 of these proteins were also susceptible to other PTMs (Fig. 2B). Within this group of 183 proteins, there was a significant enrichment of proteins involved in fatty acid biosynthesis and breakdown, JA biosynthesis and defense (Fig. 3A). At least one protein involved in each major step of JA biosynthesis and β-oxidation was susceptible to persulfidation (Fig. 3B). These two processes are an essential part of plant stress responses (Raza et al. 2021). Reversible modifications of the proteins involved in each step of these cellular events provide the cell with efficient and dynamic regulation of these proteins and products. For example, JA has also been mechanistically linked to H₂S in several plant processes including the regulation of stomatal development (Deng et al. 2020) and ascorbate and glutathione metabolism (Shan et al. 2018). Therefore, it is possible that persulfidation of JA biosynthesis enzymes may modulate this pathway as a feedback mechanism. Notably, although proteins were identified as containing a persulfidation site, the original studies did not describe all the specific Cys residues that were modified. Future studies that identify specific modification sites would be highly informative to further differentiate the various types of Cys PTMs. However, as noted earlier, several of the lipid metabolism proteins have both persulfidation modifications and other types of Cys PTMs including 33 with S-nitrosylation. H₂S and NO are coordinated in a number of signaling cascades (Mishra et al. 2021) reinforcing the value of analyzing multiple types of Cys PTMs within diverse molecular processes and environments.

S-nitrosylation is a reversible PTM that includes a covalent linkage of NO to a reactive Cys thiol group (Feng et al. 2019). RNS, the source of NO in plant cells, is known for causing oxidative damage while also acting as a signaling molecule (Kapoor et al. 2019). Although substantial efforts have been made to characterize the enzymes involved in NO biosynthesis and signaling, these pathways are still not well understood (Gas et al. 2009, Lechón et al. 2020). However, studies have pointed to an important role for plastids in the synthesis and signaling of this molecule (Gas et al. 2009). S-nitrosylation has been identified as a RNS-dependent signaling mechanism that can alter protein function including stability, biochemical activity, conformation change, subcellular location and protein–protein interaction (Feng et al. 2019, Mata-Pérez et al. 2023). Only 35 of 775 Arabidopsis lipid metabolism proteins were S-nitrosylated, with the majority (>90%) also susceptible to other modifications (Fig. 2B) (Kapoor et al. 2019). There was an enrichment of proteins involved in oxylipin biosynthesis, defense and lipid oxidation that were S-nitrosylated (Table 1, Supplementary Dataset S6). Although JA is considered an oxylipin species, there are several additional oxylipin species in Arabidopsis involved in defense and stress signaling (Vellosillo et al. 2007, Savchenko et al. 2014, Singh et al. 2022). Although the GO analysis did not statistically identify peroxisomal β-oxidation as enriched, three key proteins were S-nitrosylated: (i) a long-chain acyl-CoA synthetase (LACS, Cys587, 94% conserved) that activates fatty acids to acyl-CoAs for oxidation, (ii) an acyl-CoA oxidase that introduces a trans double-bond α–β to the acyl-CoA-thioester (ACX3, Cys219, 89% conserved) and (iii) ketoacyl-CoA thiolase (KAT1 or PKT Cys130, 100% conserved) that thiolytically cleaves a two-carbon unit from 3-ketoacyl-CoA. KAT1 Cys130 was previously identified as S-nitrosylated in a separate study (Fares et al. 2011) and is an essential catalytic residue in the active site pocket (conservation of 99% within 6 Å). Plant peroxisomes have been shown to contain NO (Corpas et al. 2021). As the β-oxidation pathway is linked to the production of H₂O₂ via acyl-CoA oxidase (through redox cycling of FADH₂), it is possible that increased levels of NO may modulate the activity of this pathway via Cys PTM in efforts to balance levels of ROS and direct specific signaling processes. Ultimately, evaluating the role of Cys PTMs during stress responses with particular attention to β-oxidation and JA biosynthesis will provide unique insights into the role and regulation of lipid metabolism.

Cys PTMs in fatty acid biosynthesis and modification

Multiple proteins within plastid fatty acid biosynthesis and modification exhibited S-sulfenylation, S-nitrosylation and S-acylation. For example, within the pyruvate dehydrogenase complex that converts pyruvate to acetyl-CoA, both the E2 (both dihydrolipoamide acetyltransferase isoforms) and E3 (dihydrolipoamide dehydrogenase) components had a Cys PTM. The S-nitrosylated Cys124 of E3 is conserved (99%) within a conserved region (99%). Its strict conservation is not surprising considering that this Cys is part of a redox-active disulfide bridge (as characterized in other systems) that links the reduction of NAD+ with the oxidation of the E2-lipoate via a flavin cofactor (Hopkins and Williams 1995). In contrast, the only Cys in both isoforms of E2 (Cys331 in AT3G25860 and Cys316 in AT1G34430) were S-sulfenylated in both control and H₂O₂-treated cells. Cys331 is conserved (93%; in a conserved region of 87%) and was also detected as S-nitrosylated (one of the few proteins with multiple modifications at a single Cys, Supplementary Dataset S1). Surprisingly, it has been proposed that this residue is not critical to the protein’s function (e.g. involved in catalysis) based on a comparison with the mitochondrial E2 isoform (i.e. three pieces of experimental data: Cys331 substituted with Trp, mitoE2 has multiple Cys and alkylating agents only inhibited mitoE2 and not plastid E2) (Mooney et al. 1999). Both Cys of E2 were detected as S-sulfenylated in at least one other proteomics study providing confidence in their PTM assignments (Akter et al. 2015, De Smet et al. 2019). Acetyl-CoA is subsequently activated to malonyl-CoA via the ACCase complex via a plastid heteromeric complex and in the cytosol by a homomeric multifunctional protein. In the plastid heteromeric ACCase, multiple residues had Cys PTMs, but in general, they were not highly conserved: (I) the carboxyltransferase β-subunit (β-CT) Cys387 was S-nitrosylated (only 29% conserved), Cys247 and Cys250 (a CXXC motif), and Cys387 were S-acylated (64, 64 and 29% conserved, respectively), (II) the biotin carboxylase subunit Cys269 was both S-nitrosylated and S-acylated (only 7% conserved) and (III) biotin carboxyl carrier protein Cys115 was S-sulfenylated (78% conserved). While it has been shown that ACCase resides with the CT half reaction through a regulatory disulfide bond, these are not the same Cys residues (Kozaki et al. 2001). As these Cys were conserved much lower on average relative to other Cys PTM, they may be associated with regulation specific to Arabidopsis or a subset of species.

The nascent fatty acid chain is then elongated (as an acyl-ACP) to 16 carbons through a reaction cycle including a condensation reaction catalyzed by ketoacyl-ACP synthase I (KASI, Fig. 4D) that had two Cys S-nitrosylated (325 and 343), four Cys S-sulfenylated (135, 266, 343 and 401) and seven Cys S-acylated (131, 135, 266, 325, 343, 347 and 401). The two S-nitrosylated Cys, Cys325 and Cys343 (neither proposed to be a catalytic residue, i.e. Cys224), were highly conserved at ∼100% but had drastically different microenvironments, e.g. predicted pK_a of 8.9 and 12.3, buried % of 11 and 94%, SASA value of 83 and 0 Ų and secondary structure end of beta-strand and part of alpha helix, respectively. A very similar condensing enzyme KASII, specific for 16- to 18-carbon elongation, also had one Cys S-sulfenylated (Cys450) and one Cys S-acylated although it was less well conserved overall (84% for Cys, 81% for nearby residues). 18-carbon acyl-ACPs can be further desaturated by stearoyl-ACP desaturases (SADs). Two SAD isoforms with varying contributions to oleic acid biosynthesis (Kapoor et al. 2019), DES2 and FAB2/SAD, shared a similar sequence motif that was S-sulfenylated in both proteins, Cys260 of DES2 and Cys267 of FAB2/SAD. Furthermore, Cys366 of FAB2/SAD was also detected as S-sulfenylated (Fig. 4E) with different characteristics. While FAB2/SAD Cys260 was 100% conserved and completely buried (100%) within a highly conserved region of DES2 (97% in the 6-Å pocket), Cys366 was 93% conserved but exposed (35%) within a somewhat conserved region (59%) of FAB2. Given that Cys260 is much closer to the active site, it may be involved in catalysis, whereas Cy366 may play a regulatory role such as mediating protein–protein interaction. Notably, a recent protein–protein interaction screen of interactors with peroxiredoxin A (PrxA, a type of 2-Cys Prx redox regulator) revealed SAD as a redox condition-dependent interacting partner (Liebthal et al. 2020). Several reviews have discussed the complex network of proteins involved in redox regulation through thiol-based mechanisms (Dietz 2011, Cejudo et al. 2021, Hernández and Cejudo 2021, Yoshida and Hisabori 2023). Some of these proteins may be directly or indirectly involved in mechanisms on Cys PTMs identified in these proteomics studies. Finally, thioesterases cleave the acyl-ACP chain prior to glycerolipid incorporation. In the Fatty Acyl-ACP Thioesterase A, Cys300 (98% conservation) was S-sulfenylated, and the analogous residue within Jatropha curcas showed that it is critical for enzymatic activity (Liu et al. 2022). The broad level of Cys PTM in each step of the fatty acid biosynthesis and modification pathway likely represents additional forms of redox regulation and would be consistent with increasingly recognizing mechanistic links between photosynthetic performance, redox state and lipid metabolism (Horn et al. 2019, Cejudo et al. 2021, Horn 2021).

S-acylation in cuticular wax biosynthesis

Another PTM that has been shown to alter protein function and subcellular location is S-acylation. S-acylation includes the addition of 16-carbon palmitate or 18-carbon stearate to Cys residues through a thioester bond (Hemsley 2020). Although this process is not dependent on a specific signaling molecule, S-acylation provides a mechanism by which proteins can be moved throughout the cell and targeted to specific organelles or compartments leading to changes in signaling on membrane surfaces (Li et al. 2022). This process is catalyzed by protein S-acyltransferases (PATs) and the removal of acyl groups, or de-acylation enzymes include acyl-protein thioesterases (APTs) and α/α hydrolase domain-containing protein 17-like APTs (ABAPTs) (Rana et al. 2018, Liu et al. 2021, Li et al. 2022). Studies have shown that PATs may have specific substrates and/or distinct roles in cellular processes (Hemsley 2020). Additionally, PATs have been identified in many plant genomes (Hemsley and Grierson 2008) and proteomics studies have shown that between 1,000 and 2,500 proteins are S-acylated in plant systems (Hemsley 2020). When lipid metabolism proteins in Arabidopsis were examined, 79 proteins (144 Cys) were susceptible to S-acylation (Fig. 2, Supplementary Dataset S1) (Kumar et al. 2020). GO enrichment analysis showed a significant and unique enrichment of proteins involved in cutin biosynthesis, response to water deprivation and abscisic acid and lipid transport (Fig. 3, Table 1). Interestingly, these biological processes are linked within Arabidopsis. When Arabidopsis plants are subjected to water deprivation or abscisic acid treatment, there is an increase in the quantity of cuticular wax on leaves (Kosma et al. 2009). S-acylation could provide a dynamic and efficient method of altering cutin biosynthesis and accumulation in response to environmental changes and abiotic stress. Similar to fatty acid biosynthesis, JA biosynthesis and β-oxidation, several proteins in successive biochemical steps to wax biosynthesis are S-acylated. Three ER-localized enzymes for converting standard acyl-CoAs into very-long chain (VLC) acyl-CoAs, VLC primary alcohols and VLC alkanes are S-acylated (Fig. 3B). Some of these compounds are incorporated into glycerolipids prior to extracellular deposition. Multiple ER sn-2 GPATs, required for the synthesis of cutin or suberin (Yang et al. 2012), were S-acylated and may be important for their membrane localization. In fact, an analysis of positional frequency of S-acylation revealed that they were more often found in the termini of proteins (Supplementary Fig. S4). Finally, multiple lipid transfer proteins (LTPs) were detected as S-acylated with a representative LTP shown in Fig. 4F. LTPs transport small numbers of lipids at a time using hydrophobic cavities that stabilize lipid molecules outside membranes (Wong et al. 2019, Missaoui et al. 2022). These proteins contain disulfide bonds to stabilize their structure, and several of these Cys within a subset of LTPs were S-acylated (Table 1). The fact that usually only one of the two Cys within an intramolecular disulfide bond was modified may point to the Cys PTM residue being the redox-sensitive Cys (the peroxidatic Cys). LTPs have been proposed to aid in the transfer and deposition of lipid monomers to the cell wall required for cuticle assembly (Salminen et al. 2018). Reversible S-acylation may be a process that helps these LTPs move between and anchor to different organellar membranes or contact sites. Collectively, the over-representation of S-acylation in wax biosynthesis is intriguing and warrants additional studies. Unlike the other PTMs described, the stimulus or signal leading to S-acylation is not well understood. In addition, the functional impact of this modification for most proteins is largely unknown. If each PAT has specific substrates and/or a distinct role(s), the stimulus and protein alteration caused by this modification could also be unique. Studies focused on the role of this modification in altering lipid metabolism under diverse environmental conditions will be helpful in filling in the large knowledge gaps.

Summary

The implementation of proteomics approaches capable of detecting Cys PTM has resulted in rich datasets to explore, revealing a complex signaling and regulatory environment for lipid metabolism proteins (as well as the broader plant metabolome). The process of integrating physiochemical predictions, model structures and genomics-derived sequence comparisons using published data is now relatively straightforward. However, similar to other integrative PTM studies (Marino and Gladyshev 2010b), it is challenging to conclude specific microenvironments that explain why and how specific Cys residues undergo PTM (and current data suggest that the vast majority of Cys are not modified). In general, it does, however, appear that each type of thiol-based PTM has a propensity for Cys with certain levels of accessibility and surrounding protein microenvironments (Fig. 1). The location of the modified Cys within the protein (Supplementary Fig. S4) may be a key component to its function, which is not only in cases of catalytic function but also in directing subcellular location/trafficking and interactions with other macromolecules. However, proteins are not in isolation within the cell, so the true microenvironments are complex to define. Furthermore, in most cases, the Cys PTM residues are more conserved than their unmodified counterparts. Based on an analysis within lipid metabolism (Figs. 2–4), many of the highest conserved Cys PTMs are known (or likely) catalytic residues (Table 1, Supplementary Dataset S1). Chemical modification of a Cys thiolate within an active site may be expected to be inhibitory in nature. However, exceptions leading to other functional changes, e.g. activation, conformational changes, etc., have been noted. Future studies that characterize the structural and functional changes in specific Cys residues, with and without PTM, will be highly informative at the mechanistic level (and evaluating the Cys PTM proteome more broadly). Collectively, ROS, H₂S and RNS can alter plant growth and development through reversible PTMs of reactive, Cys thiol groups. S-sulfenylation, persulfidation and S-nitrosylation may have synergistic or antagonistic functions from a signaling perspective. Variability in natural Cys PTM patterns and stress-induced patterns within current proteomics datasets make it challenging to mechanistically link stress-induced Cys PTMs and downstream structure–function alternations, at least as applied to lipid metabolism proteins. In addition to these redox-driven PTMs, S-acylation can also modify thiol groups impacting protein localization and function. It would be shortsighted not to consider that there are relatively few proteomics studies that have identified Cys PTMs and these focused on very few plant systems and (a)biotic conditions. It cannot be ruled out that in some cases, these PTMs may be experimental artifacts, and therefore, defining their conditionality and pervasiveness is key. Ultimately, although the presence of several PTMs on lipid metabolism proteins has now been identified, the role and abundance of these changes are largely unknown. Future studies focused on elucidating how proteins are affected by these modifications will be essential to fully understanding the relationship and role of these signaling molecules and protein modifications in plants. Finally, efforts are underway to genetically engineer crops and other plants with enhanced resilience in response to an increasingly destructive climate. Continuing to expand the field and knowledge of plant thiol biochemistry will inevitably inform these engineering efforts and outcomes.

Supplementary Data

Supplementary Data are available at PCP online.

Data Availability

The data underlying this review are available in referenced articles, the AlphaFold structure database and genomes deposited in the public domain as described through the manuscript. Reasonable requests for additional processing data will be provided in a timely manner.

Funding

US Department of Agriculture (grant number 2022-67013-39697 to P.J.H.) and start-up support from the Research and Innovation office at the University of North Texas.

Author Contributions

All authors contributed to the conceptual framework of this review, analyzed the data and wrote the manuscript.

Disclosures

No conflicts of interest declared.