Location of S‐nitrosylated cysteines in protein three‐dimensional structures

Although S‐nitrosylation of cysteines is a common protein posttranslational modification, little is known about its three‐dimensional structural features. This paper describes a systematic survey of the data available in the Protein Data Bank. Several interesting observations could be made. (1) As a result of radiation damage, S‐nitrosylated cysteines (Snc) are frequently reduced, at least partially. (2) S‐nitrosylation may be a protection against irreversible thiol oxidation; because the NO group of Snc is relatively accessible to the solvent, it may act as a cork to protect the sulfur atoms of cysteines from oxidation by molecular oxygen to sulfenic, sulfinic, and sulfonic acid; moreover, Snc are frequently found at the start or end of helices and strands and this might shield secondary structural elements from unfolding.


| INTRODUCTION
Protein structures, such as those determined by crystallographic analyses, were long thought to be static objects.Then, over the last three decades, it became clear that proteins are flexible, 1 and that some are completely unfolded, 2,3 and several techniques have been developed to determine the relationship between flexibility and function, ranging from molecular dynamics 4 to normal mode analysis 5 and B-factor analysis. 68][9] Hundreds of PTM have been identified and documented in databases. 10nitrosation of cysteines (the incorrect term S-nitrosylation is widely used in molecular biology and will be used here) is an intriguing example of PTM in which the oxidation state and electronic structure of the sulfur atoms change.[11][12][13] The thiol group R S H is transformed into the nitrosothiol group R S N O, with the sulfur atom's oxidation number increasing from À2 to 0 and the nitrogen atom's decreasing from +2 to +1.
Protein S-nitrosylation is believed to be a prototypic redox-based signaling mechanism in living organisms. 14It is known to occur in all phylogenetic kingdoms, and two thirds of proteins are thought to be reversibly S-nitrosylated. 15S-nitrosylation may occur nonenzymatically through oxidation of NO to more reactive NO + -a process named redox-activation 13 -or through the formation of the more reactive N 2 O 3 -via the reactions 2NO + O 2 !2NO 2 and NO 2 + NO !N 2 O 3. 16 However, it is thought to be regulated in vivo by enzymes. 17,18Three types of enzymes (S-nitrosylases) work together to attach NO to proteins, similar to how ubiquitination works.More in general it was shown that in Escherichia coli S-nitrosylation is dependent on the enzymatic activity of the hybrid cluster protein Hcp, which uses nitrate reductase-produced NO.Anaerobiosis on nitrate induces both Hcp and nitrate reductase and leads to the S-nitrosylation-dependent assembly of an extensive interactome that includes enzymes that generate NO, produce nitrosylated proteins, and propagate SNO-based signaling. 18Consequently, S-nitrosylation is involved in a wide range of biological processes and pathologies, including skeletal muscle regulation, neurological dysfunctions, and cancer. 19,20evious studies, based on very small data sets, suggested that S-nitrosylated cysteines (Snc) tend to be near charged residues (Arg, Lys, His, Glu, and Asp), despite the occurrence of a local hydrophobicity surrounding seems to favor S-nitrosylation. 21Later, based on a few experimental structures and some computational models, a comprehensive structural analysis of cysteine S-nitrosylation revealed that these tendencies are extremely heterogeneous and may be relevant only for subsets of proteins but cannot be generalized. 22Nonetheless, charged and solvent exposed side-chains within 8 Å of the cysteine were found to play a role in S-nitrosylation by regulating proteinprotein interaction and trans-nitrosylation. 22 All of these studies do not provide enough data to create a computational technique for predicting cysteine S-nitrosylation.However, in the age of Artificial Intelligence, several methods for predicting cysteine S-nitrosylation have recently been published.One of them predicts only if a protein is S-nitrosylated 23 and others predict S-nitrosylation sites.[29][30] These computational methods appear to perform well, with accuracies ranging from 0.65 to 0.86 (Table S1 in the Supplementary Material), and an in silico analysis pipeline has been described. 31These prediction strategies, however, do not provide information about the effects of S-nitrosylation on the structure and dynamics of S-nitrosylated proteins, as well as the chemical role of Snc.
Our understanding of the chemical and physical consequences of S-nitrosylation is rather rudimentary.From a very general perspective, they can be classified into two broad categories: on the one hand, S-nitrosylation might be a simple and effective way to protect cysteines from irreversible oxidation to sulfenic, sulfinic, and sulfonic acid 32 ; on the other hand, it may have an impact on protein structure and dynamics, as well as introduce new chemical features that alter protein reactivity.
In the first case, any structural or chemical modification of the protein would be irrelevant, as S-nitrosylation would simply be a transient protection of the thiol.In the second case, on the other hand, sulfur oxidation implies a change in the electronic structure, which may affect protein structure, dynamics, and reactivity.
A systematic comparison of the structure of the S-nitrosylated protein to the structure of the same protein when it is not S-nitrosylated is not possible because the second structure is not generally available among the experimental structures available in the PDB.In principle, the structure of the non-S-nitrosylated protein could be predicted using AlphaFold 33 or it could be found in the AlphaFold Protein Structure Database, 34 which contains more than 200 million computational models predicted using AlphaFold.However, computational models may miss small-but not insignificantdetails of protein structures and data mining from predicted structure databases must still be proven to be an effective and reliable practice.
The few possible comparisons yield multifaceted results.In other cases, S-nitrosylation has very little effects on structure.
Weichsel et al. showed that, in human thioredoxin 1, S-nitrosylation of Cys 62 occurs, despite its burial in the protein core, though a local conformational change-the helix containing this cysteine is unraveled and the cysteine becomes solvent accessible. 36ven that more than 80% of the S-nitrosylated protein structures have been deposited in the Protein Data Bank [37][38][39] after the publication, in 2010, of the last comprehensive structural analysis of cysteine S-nitrosylation, 22 it is important to extend these studies.In particular, we would like to examine where S-nitrosylation occurs and which are the interaction between the Snc side-chain with the rest of the protein.
Considering that the last thorough structural study of cysteine S-nitrosylation was published in 2010 22 and that over 80% of the S-nitrosylated protein structures have been deposited in the Protein Data Bank, 22,[37][38][39] it is important to extend these studies. 22In particular, we would like to examine where S-nitrosylation occurs and which are the interaction between the Snc side-chain with the rest of the protein.

| Data selection
All data were obtained from the Protein Data Bank, [37][38][39] which was searched for crystal structures containing hetero groups called SNCthis is the name of S-nitrosylcysteines.Approximately 90% of these crystal structures were refined at resolution better than 2.5 Å.

| Secondary structure analysis
Secondary structure assignments were performed with Chimera, 40 which has an implementation of DSSP, 41 and three types of secondary structures were considered: helices, strands, and loops.
We examined not only the type of secondary structure adopted by the Snc but also the position of the Snc in the secondary structural elements (SSEs).Two types of SSE were considered: helices and strands.
The following numbering scheme was followed (see Figure 1): 1. the three positions preceding the SSE were numbered as B3, B2, and B1; T A B L E 1 List of the protein crystal structures containing S-nitrosylated cysteines in the Protein Data Bank.The four-letter PDB identification code, the name of the S-nitrosylated cysteine, the resolution (Å), the name of the protein, and the biological source where it is expressed are all provided for each entry.The name of the S-nitrosylated protein (SNC) is preceded, if necessary, by the conformational disorder code, followed by the protein chain identifier, and finally by the residue serial number.F I G U R E 1 Numbering scheme of the sequence positions in the secondary structural element and of the sequence positions preceding and following the secondary structural element.When the Snc was close to both the beginning and the end of a SSE-this may happen, in particular, for β-strands, which are often short-the nearest boundary was preferred; as an example, if it occupies a position that can be considered to be both P2 and P(n-2), the P2 classification was preferred.
If the Snc is equidistant from the beginning and the end of a SSE, it was arbitrarily considered to be closer to the beginning; as an example, if it occupies a position that can be considered to be both P2 and P(n-1), the P2 classification was preferred.

| Miscellaneous
Solvent excluded surface areas were computed with Chimera with a default solvent probe radius of 1.4 Å. 40 Hydrogen and chalcogen bonds were identified visually.For chalcogen bonds, the following quantitative thresholds were also used: distances S O and S S ≤ sum of the van der Waals radii +0.1 Å (van der Waals radii = 1.52 Å for oxygen and 1.80 Å for sulfur); angles X-S O and X-S S ≥ 155 .

| Radiation damage
The C S N O group is rigidly planar according to small molecule crystallographic data and to quantum mechanical computations. 42wever, in most of the protein crystal structures of Table 1, it is not.
Figure 2B shows that often the torsion CB-SG-ND-OE (see Figure 2A for the numbering scheme) is not equal to 0 (cis confirmation) or to ±180 (trans conformation).Although there is a maximum at 0 , no maxima are observed at ±180 , and two maxima at 60-90 and À60-90 clearly appear.4][45][46] This phenomenon has been studied by means of quantum chemistry computations 42 and some scientists have observed degradation of the S-NO group when crystals absorb large X-ray doses in highly brilliant synchrotron beam lines. 47,48Now, the systematic survey of the structural data described in the following paragraphs confirms that reduction from I to II occurs, at least partially.

The electron delocalization of the S C N O moiety vanishes
with the reduction from I to II (Figure 3A).This implies that the hybridization of the sulfur and nitrogen atoms changes from sp 2 in I to sp 3   are quite large, and a perfect correlation between the variables is unlikely.On the other hand, it is possible-even likely-that the reduction from I to II is incomplete, resulting in varying degrees of reduction observed in different crystal structure determinations.This implies that some of the crystal structures contain a mixture of oxidized and reduced S-nitrosylation sites; this type of "conformational disorder," which is difficult to characterize, degrades the quality of the observed stereochemistry.
However, we can conclude that we do not know much about the true structure of this posttranslational modification.The majority of the available structures could be experimental artifacts, at least in part.

| Secondary structure of cysteines that are S-nitrosylated
We examined not only the secondary structure of the S-nistrosylated cysteine, but also its position along helices, strands, or loops, using the numbering scheme shown in Figure 1.
The vast majority of Snc are at the borders of the SSE (Figure 4A).
About 70%-80% of the Snc are in positions B1-P1-P2 or P(n-1)-P(n)-A1 for helices and about 90% for strands.There are no Snc in internal positions for strands, though it must be remembered that they are often short and do not have internal positions.In helices, which are longer, only about 15% of the Snc are in internal positions.However, there are also cysteines that are not S-nitrosylated despite they are close to the borders of a SSE.For example, in the crystal structure of human tyrosine-protein phosphatase non-receptor type 1 (PDB code 3eu0 50 ), three cysteines are not S-nitrosylated despite they are in position P1 of a helix (Cys 32), in position P3 of a helix (Cys 92), or in position A2 after a strand (Cys 121)-in the same structure there is a cysteine that is S-nitrosylated and in position A1 after a strand (Snc 215).
Does this mean that only cysteines near the SSE boundaries can be S-nitrosylated?Based on the limited structural data available at the moment, it is impossible to answer this question.It is also important to remember that S-nitrosylation was achieved in vitro rather than under physiological conditions in many crystal structures.Furthermore, radiation damage during X-ray diffraction data collection may reduce Snc due to NO loss.Consequently, it is possible that some of Snc does not really exist in vivo despite they are observed in vitro, and that some Snc that exist in vivo are denitrosylated in the samples used for determining the crystal structures.

| Solvent accessibility
In theory, a cysteine that is not solvent accessible cannot be S-nistrosylated.In reality, due to protein elasticity, residues buried at the protein interior can be exposed transiently due to local structural fluctuations.The crystal structure of human thioredoxin (PDB code 4oo5), for example, shows a Snc (Snc 62) that is not accessible to the solvent.It is thought that the helix beginning at residue 62 can completely unravel, exposing this residue to the solvent and allowing it to be S-nitrosylated. 36e solvent excluded surface areas (SESA) of CA, CB, SG, ND, and OE atoms were investigated (see Figure 5B).They increase in average as they move away from the backbone: while the average SESA of the CA atom is less than 4 Å 2 , the average SESA of the OE atom is 10 Å 2 .
The latter value is quite large, given that only 25% of the atoms in the protein structures studied here have SESA greater than 10 Å 2 .
This means that the N O molecule, once incorporated in the cysteine, remains quite accessible to the solvent.There are exceptions: for example, the Snc of human thioredoxin (PDB code 2hxk) is completely buried in the protein core. 36However, there is a statistical tendency for N O to remain quite accessible to the solvent, perhaps to protect, as a cork, the underlying Cys from non-reversible oxidation by molecular oxygen.It has been proposed that S-nitrosylation of the active site Cys 215 in human tyrosine-protein phosphatase nonreceptor type 1 (PDB code 3eu0) is necessary to protect it from nonreversible oxidation. 50In the absence of proper protection, oxidation of cysteine to sulfenic, sulfinic, and sulfonic acid may occur. 32triguingly, the fact that N O groups are relatively solventaccessible is consistent with the observation that nitrosylation occurs more frequently near the secondary structural element boundaries.In particular, the boundaries of helices may be more solvent-accessible than the central region, as the hydrogen bonding pattern is disrupted or incomplete at the beginning and end of the helix.

| Hydrogen bonds
Potential hydrogen bonds have been visually identified-an automated procedure would be highly unreliable in this case due to the uncertainty of Snc's oxidation state.Figure 5A shows the number of hydrogen bonds observed in the protein crystal structures studied in this communication.
In agreement with the solvent exposure of the N-O moiety, ND and OE are equally involved in hydrogen bonds, many of which (about 50%) are with water molecules.Surprisingly, approximately 35% of the hydrogen bonds involving ND are with oxygen atoms that are only hydrogen acceptors (backbone oxygen atoms and side chain carboxylate or amide oxygen atoms).This clearly indicates that ND is protonated, at least in part, and that the Snc is reduced from I to II (Figure 3).
On the contrary, OE rarely interacts with oxygen atoms that are only hydrogen acceptors.

| Chalcogen bonds
2][53] The sulfur oxidation due to S-nistrosylation should influence the formation and stability of chalcogen bonds, since the N O moiety is markedly more electronegative than the thiol hydrogen atom.
Chalcogen bonds, like hydrogen bonds, have been visually identi- Additionally, most of the chalcogen bonds connect residues that are close in sequence.Often, they are in a helix and the Snc number X interacts in two ways with the carbonyl of residue number X-4: a canonical hydrogen bond with its backbone nitrogen atom and a chalcogen bond with its side-chain sulfur atom (for example Snc 93 and Ser 89 of chain B in 1buw).In two cases, the calchogen bond involves the carbonyl of the residues that precedes Snc and only in one case the backbone oxygen is far in the protein sequence.
As a consequence of S-nitrosylation, the side-chain sulfur atom cannot behave as a hydrogen donor.It is thus possible that chalcogen bonds are a sort of compensation for the loss of some hydrogen bonds.This cannot be confirmed, however, based on the available data since the exact position of the thiol hydrogen atom is always uncertain. 54hese observations agree with the fact that the S N O moiety is rather exposed to the solvent and interacts, through hydrogen and chalcogen bonds, with hydrogen donors (and hydrogen acceptors if reduction from I to II occurs; Figure 3) and with nucleophilic atoms.

| Chemical composition of the surroundings
Furthermore, because Marino and Gladyshev proposed that protein-protein interaction is related to trans-nitrosylation from one protein to the other, the distribution of charged atoms around Snc was monitored 22 : this hypothesis is based on the observation of Marino and Gladyshev that charged atoms exposed to the solvent are more frequent within 8 Å of Snc than within 8 Å of reduced cysteines.
This should indicate that the electrostatics at the protein surface close to the S-nitrosylation site is important for S-nitrosylation and is related to protein-protein interaction.However, no differences between reduced and S-nitrosylated cysteines were observed in the set of protein structures examined here.We modified the threshold distance to 6, 7, 9, and 10 Å and we modified the criterion for solvent accessibility from 1.0 Å 2 -the criterion adopted by Marino and Gladyshev 22 -to 0.5 Å 2 or 2.0 Å 2 .However, no clear trends emerged other than random fluctuations related to the small size of the available data set.This does not imply that the hypothesis of trans-nitrosylation proposed by Marino and Gladyshev is incorrect but it suggests that S-nitrosylation may occur with several, different reaction mechanisms.

| CONCLUSIONS
S-nitrosylation of cysteines is a common posttranslational protein modification.Few three-dimensional structures have been determined, primarily through X-ray diffraction experiments, in which S-nitrosylated cysteine (Snc) is frequently reduced, at least partially, as a result of X-ray photo-chemical damage.
Snc are frequently found near the beginning or end of helices and strands.This could be related to the function of S-nitrosylation: it could be a defense against thiol irreversible oxidation and, as a result, a defense of protein folding, with secondary structural elements preserved from unfolding.The paucity of the data does not allow a robust estimation of the statistical significance of this observations and further information is needed to rich a deeper understanding of this phenomenon.
Snc are found to be relatively solvent accessible and involved in both hydrogen and chalcogen bonds.The latter are almost always trans to the ND atoms, as one would expect given the NO group's higher electronegativity.
The main source of concern in this analysis is the data quality.
Several S-nitrosylations were obtained in vitro rather than under in vivo experimental conditions.As a result, some of them could be experimental artifacts that do not reflect physiological S-nitrosylation.
Furthermore, radiation damage may not only reduce Snc but also remove the N O group, resulting in some S-nitrosylations going undetected.
More high-resolution experimental structures are expected to shed light on the mechanisms of S-nitrosylation and the role of this posttranslational modification.A better characterization of the redox status of the C S N O moiety, in particular, should be possible at very high resolution using not only X-ray diffraction experiments but also alternative experimental techniques, for example, neutron diffraction.

2 .
the three positions following the SSE were numbered as A1, A2, and A3; 3. the first three positions of the SSE were numbered as P1, P2, and P3; 4. the last three positions of the SSE were numbered as P(n-2), P(n-1), and P(n); 5. in SSEs longer than six residues, all additional positions were considered to be internal, without further classification.
in II.Consequently, both the bond angles CB-SG-ND and SG-ND-OE are expected to widen and the bonds CB-SG, SG-ND, and ND-OE are expected to lengthen in going from I to II.These trends are actually observed in the set of structural data examined here.

FigureF
Figure 3B shows, for example, that there is a good correlation between the two angles CB-SG-ND and SG-ND-OE.The correlation coefficients between these five variables, the two bond angles CB-SG-ND and SG-ND-OE and the three bond distances CB-SG, SG-ND, and ND-OE, are shown in Figure 3C and follow the trends expected for the reduction from I to II.However, the correlation coefficients are not very large in absolute value.There are two possible explanations.On the one hand, the atomic positional standard errors, which are unknown here, are in the range 0.1-0.2;this implies that the bond distance standard errors are in the range 0.15-0.30and the bond angle standard errors are in the range 10-20 .As a result, the bond distances and uncertainties

FigureF
Figure 4B shows an example of S-nitrosylation.The crystal structure of human S-nitroso thioredoxin (PDB code 2hxk 49 ) shows two Snc (Snc 62 and 69), which are at the beginning and at the end of a helix, two oxidized cysteines (Cys 32 and 35), which form a disulfide

F
I G U R E 5 (A) Number of hydrogen bonds between water/protein atoms and the atoms ND and OE of the Snc.(B) Solvent excluded surface areas (Å 2 ) of the atoms CA-OE of the Snc.(C) List of the possible chalcogen bonds involving the SG atoms of Snc.(D) Chemical composition around SG (continuous line), ND (continuous lines with black circles), and OE (broken line).
fied.They are listed in Figure5C.Although they are not very numerous, they indicate some structural trends.First, they involve always a backbone oxygen atom.Second, this oxygen atom is nearly always trans to the ND-SG covalent bond.Third, most of the chalcogen bonds are observed in structures where the C S N O moiety is planar and the reduction from I to II (Figure3) did not occur or, at least, was very limited.

Figure
Figure 5D depicts the chemical composition of spheres of increasing radius and centered on the atoms SG, ND, or OE.At short distances from the SG/ND/OE atom the percentage of backbone atoms is higher than at large distance, especially for SG and ND.Analogous trends are observed for protein and water oxygen

Table 1
lists all these crystal structures and the Snc they contain.There are 35 crystal structures and 80 Snc, with nine of them being conformationally disordered.

Table 1 ,
as well as the amplitude of the bond angles CB-SG-ND and SG-ND-OE.
C S NH OH (II) occurs (or to radicals C S N ˙OH and C S NH O ˙) due to electrons produced by radiation absorption (Figure