Profiling cysteine reactivity and oxidation in the endoplasmic reticulum

Abstract

The endoplasmic reticulum (ER) is the initial site of biogenesis of secretory pathway proteins, including proteins localized to the ER, Golgi, lysosomes, intracellular vesicles, plasma membrane, and extracellular compartments. Proteins within the secretory pathway contain a high abundance of disulfide bonds to protect against the oxidative extracellular environment. These disulfide bonds are typically formed within the ER by a variety of oxidoreductases, including members of the protein disulfide isomerase (PDI) family. Here, we establish chemoproteomic platforms to identify oxidized and reduced cysteine residues within the ER. Subcellular fractionation methods were utilized to enrich for the ER and significantly enhance the coverage of ER-localized cysteine residues. Reactive-cysteine profiling ranked ~900 secretory pathway cysteines by reactivity with an iodoacetamide-alkyne probe, revealing functional cysteines annotated to participate in disulfide bonds, or S-palmitoylation sites within proteins. Through application of a variation of the OxICAT protocol for quantifying cysteine oxidation, the percentages of oxidation for each of ~700 ER-localized cysteines were calculated. Lastly, perturbation of ER function, through chemical induction of ER stress, was used to investigate the effect of initiation of the unfolded protein response (UPR) on ER-localized cysteine oxidation. Together, these studies establish a platform for identifying reactive and functional cysteine residues on proteins within the secretory pathway, as well as for interrogating the effects of diverse cellular stresses on ER-localized cysteine oxidation.

The endoplasmic reticulum (ER) is the largest intracellular organelle and contains cisternae that are contiguous with the nuclear membrane and extend toward the plasma membrane. The ER is separated into the smooth and rough ER, each of which performs distinct cellular functions. The smooth ER is absent of ribosomal proteins and is responsible for lipid biogenesis, Ca²⁺ storage and signaling, and xenobiotic metabolism. The rough ER is characterized by a high density of ribosomes and predominantly mediates protein synthesis, folding, and export^{1, 2}. Proteins destined for the secretory pathway are co-translationally translocated into the ER lumen, where protein folding, glycosylation, and disulfide-bond formation occurs to generate the mature protein³. Properly folded proteins are transported to the Golgi via coated protein complex II (COPII)-mediated anterograde transport for further processing. From the Golgi, proteins are either retrograde transported back to the ER in COPI vesicles⁴, or assembled into secretory granules for transport to the plasma membrane or extracellular environment³.

Protein maturation within the ER involves several folding chaperones, including heat shock proteins (HSPs), and lectin-based chaperones. HSPs, such as HSPA5 (BiP), stabilize intermediate structures to promote folding and minimize aggregation⁵. The lectin-based chaperones, calreticulin (CALR) and calnexin (CALX), recognize N-linked glycan intermediates as a quality-control mechanism to promote proper folding⁶. Dysregulation of protein folding in the ER leads to the accumulation of unfolded proteins and upregulation of the unfolded protein response (UPR). The UPR attenuates translation and upregulates expression of folding chaperones and endoplasmic reticulum associated degradation (ERAD) components⁷. ERAD degrades irreversibly misfolded proteins through ubiquitin-mediated protein degradation upon translocation to the cytosol⁸. Dysregulated protein folding in the ER is associated with several diseases, including diabetes, cancer, and neurodegeneration^9–13.

A critical component of ER-localized protein folding, is the proper formation of disulfide bonds to stabilize protein structure. The highly oxidizing extracellular environment can spontaneously introduce inter- and intramolecular disulfide bonds¹⁴, necessitating the controlled formation of disulfide bonds prior to extracellular export. As one of the most oxidizing organelles in the cell, the ER is the favored location for the installation of disulfide bonds on secretory proteins^{15, 16}. Disulfide-bond formation in the ER is mediated by the protein disulfide isomerase (PDI) family through thiol-disulfide exchange reactions¹⁷. Dysregulated disulfide-bond formation can lead to the accumulation of misfolded proteins, and contribute to various diseases, including transmissible spongiform encephalopathies (TSEs) and amyotrophic lateral sclerosis (ALS), where improper disulfide-bond formation leads to protein misfolding and aggregation^11–13.

Methods to identify functional cysteines and quantify cysteine oxidation in the ER can illuminate the (dys)regulation of disulfide-bond formation, and protein-folding processes within the secretory pathway. Cysteine-reactivity profiling using isotopic tandem orthogonal proteolysis – activity-based protein profiling (isoTOP-ABPP) has enabled the interrogation of cysteine reactivity, modifications and inhibition^18–20. Briefly, isoTOP-ABPP analysis utilizes a cysteine-reactive iodoacetamide-alkyne (IA) probe to covalently modify reactive cysteines within a proteome. Incorporation of distinct isotopic labels to IA-tagged peptides from two proteomes allows for assessing cysteine reactivity changes induced by oxidation, or covalent inhibition. Typical isoTOP-ABPP studies are performed in whole-cell lysates¹⁸, and afford poor coverage of the ER proteome; 42 ER-localized cysteine residues (from 35 proteins) are identified in a prototypical study¹⁸. Recently, an ER-localized chloroacetamide probe was generated to selectively target ER cysteines. In total, 195 proteins were identified, of which 180 were annotated as ER and secretory pathway proteins. However, despite enriching proteins using a cysteine-targeted probe, this study did not identify or evaluate reactivity changes to the individual cysteine residues that were probe labeled²¹. Therefore, with currently available methods, the number of ER-localized cysteine residues that has be identified remain minimal, necessitating platforms that combine ER enrichment with cysteine profiling. Potential strategies to enrich ER proteins, include the use of engineered ascorbate peroxidases (APEX)^{22, 23}, proximity-dependent biotin identification (BioID/TurboID)²⁴, and subcellular fractionation by differential centrifugation²⁵. Previously, mitochondrial isolation using differential centrifugation provided enhanced coverage of reactive mitochondrial cysteines²⁶, without the need to engineer the cellular system, as required for APEX and BioID/TurboID. Differential centrifugation methods to isolate the ER have previously identified ~469 ER-associated proteins from rat pancreas²⁷. Combining these established ER fractionation methods with isoTOP-ABPP has the potential to significantly enhance coverage of ER-localized cysteines for reactivity profiling.

Although cysteine-reactivity profiling can indirectly inform on cysteine oxidation state through an observed loss in cysteine reactivity across two samples, several proteomic methods exist to more directly monitor cysteine oxidation within a single biological sample^28–30. One such method is OxICAT (isotopically coded affinity tags), which adapts thiol-trapping techniques with isotopically encoded alkylating reagents to interrogate cysteine oxidation state³¹. Briefly, all reduced cysteine residues are capped with an isotopically light alkylating reagent, followed by treatment with a reducing agent to expose previously oxidized cysteines for tagging with an isotopically heavy reagent. Light:heavy (L:H) ratios for identified cysteine-containing peptides directly reflect the percentage of oxidation for each cysteine residue. OxICAT methods have been performed in E. coli³¹, yeast³², and mammalian cells³³, as well as in enriched mitochondria³⁴. More recently, OxICAT was implemented to study the effects of ethanol consumption on redox homeostasis in the ER of murine pancreas tissue³⁵. In this study, the oxidation states of ~ 500 secretory pathway localized cysteines were identified and evaluated for changes due to ethanol consumption.

Here, we report the use of ER isolation by differential centrifugation coupled to isoTOP-ABPP and OxICAT analyses to monitor cysteine reactivity and oxidation, respectively. For cysteine-reactivity studies, we apply the isoTOP-ABPP platform to rank 923 ER-localized cysteine residues by reactivity. Our studies revealed a subset of hyperreactive ER-localized cysteines that are functionally implicated in protein stabilization through disulfide bonds, and membrane localization through S-palmitoylation. To directly interrogate the oxidation state of cysteines in the ER, OxICAT analysis afforded the quantification of percent oxidation of 736 cysteines on secretory proteins. Together, these studies establish methodological guidelines for in-depth proteomic interrogation of cysteine function and oxidation within the ER. We demonstrate the utility of this platform by evaluating the effects of UPR upregulation on the cysteine oxidation state of ER-localized proteins.

Results and Discussion

Isolation of the endoplasmic reticulum using differential centrifugation.

The low abundance of ER proteins, relative to highly abundant cytosolic and nuclear proteins, results in poor coverage of ER-localized cysteines in typical isoTOP-ABPP analyses of whole-cell lsates^{18, 36} (Supplemental Figure S1A³⁷ and B). To enrich for ER proteins prior to isoTOP-ABPP analysis, a differential centrifugation method developed for isolation of rat pancreas ER²⁷ was adapted to mammalian cells (Figure 1A). Briefly, SKOV-3 cells were gently homogenized to disrupt the cell membrane while preserving the integrity of organelles. Multiple differential centrifugation steps were used to pellet dense organelles, such as nuclei and mitochondria, after which a high speed centrifugation step (150,000g × 30 min.) afforded a microsomal fraction. This microsomal fraction was stripped of ribosomal proteins, and a final centrifugation step (48,000g × 30 min.) generated ER proteomes for isoTOP-ABPP and OxICAT-based proteomic studies.

Figure 1.

Proteomic platform for the enrichment and identification of ER proteins and peptides. A) Schematic representation of the differential centrifugation steps used to isolate the ER proteome, and proteomic workflow to analyze ER-derived trypsin-digested peptides. B) Western-blot analysis of isolated organellar fractions against ER markers, PDIA4 and CALR, cytosolic marker, GAPDH, ribosomal marker, RPS6, nuclear marker, H3FA, and mitochondrial marker, ATPIF1. Subcellular fractions analyzed include: whole-cell lysate (WC), nucleus (Nuc.), mitochondria (Mito.), cytosol (Cyto.), ribosomes (Ribo.), and ER. C) Percentage of proteins Uniprot annotated as ER proteins identified from MS analyses of trypsin digests from whole-cell and ER fractions. D) Percentage of proteins Uniprot annotated as ER and secretory pathway proteins (ER, cell surface, extracellular, intracellular vesicle, and golgi) from MS analyses of trypsin digests from whole-cell and isolated ER fractions .

ER enrichment was confirmed by western blot of each fraction, including whole-cell, nucleus, mitochondria, cytosol, ribosomes, and ER (Figure 1B). Two ER markers, protein disulfide isomerase A4 (PDIA4) and CALR, a cytosolic marker, glyceraldehyde 3-phosphate dehydrogenase (GAPDH), a ribosomal marker, 40s ribosomal protein S6 (RPS6), a nuclear marker, histone H3 (H3FA), and a mitochondrial marker, ATPase inhibitor (ATPIF1), were used to confirm successful organelle fractionation. Relative to the whole-cell fraction, the isolated ER fraction demonstrates a strong enrichment for ER resident proteins, PDIA4 and CALR, and a significant decrease of proteins found in other subcellular locations, confirming successful enrichment of ER proteins through differential centrifugation.

Confirmation of ER enrichment using mass spectrometry.

In addition to western-blot analysis, mass-spectrometry (MS) analyses was performed to determine the major protein constituents of the ER-enriched samples. To achieve this, whole-cell lysates and ER fractions were analyzed by LC/LC-MS/MS after trypsin digestion. Filtering of MS data for proteins with ≥2 identified peptides, generated 1,582 and 1,844 proteins for whole-cell and ER fractions, respectively (Supplemental Table S1). Gene ontology (GO) enrichment analysis³⁸ (Supplemental Figure S2A and Supplemental Table S2 and S3) demonstrated that the ER fraction was enriched for ER-associated processes, such as protein localization to the ER, vesicle cargo loading, N-linked glycosylation, and chaperone-mediated folding. None of these aforementioned biological processes were significantly enriched in the whole-cell sample.

Subcellular localization of identified proteins was determined by mining the Uniprot database for cellular locations. In total, MS analysis of the ER fraction identified 1,926 ER peptides from 212 proteins, in contrast to the 488 ER peptides from 38 proteins identified in whole-cell lysates (Figure 1C). In addition to Uniprot, the Human Protein Atlas database³⁷ contains annotation for proteins whose localization to the ER has been experimentally detected. Of all protein encoding genes, 443 genes have been experimentally shown to localize to the ER. Using this database as a measure of ER coverage, only ~10% of all experimentally detected ER proteins are identified in whole-cell lysates compared to ~40% in the ER fraction, representing a substantial increase in coverage of ER proteins using differential centrifugation to enrich the ER prior to MS analysis (Supplemental Figure S2B and Supplemental Table S4).

In addition to ER-annotated proteins, UniProt annotation were used to identify the number of secretory pathway proteins (cell surface, extracellular, intracellular vesicle, and golgi) included in the dataset. These proteins would be expected to pass through the ER toward their final cellular destination. In total, 719 proteins were identified as localized to the ER and secretory pathway in the ER fraction, compared to 234 in whole-cell lysates (Figures 1D). In subsequent analyses, we use the term ER proteins to refer to both ER-localized and secretory pathway proteins identified in our ER-enriched fraction.

IsoTOP-ABPP analysis uncovers reactive and functional cysteines in the ER.

Reactive-cysteine profiling by isoTOP-ABPP¹⁸ enables the quantification of cysteine reactivity across two different samples. This platform relies on the use of a cysteine-reactive iodoacetamide-alkyne (IA) probe to modify highly nucleophilic cysteines within complex proteomes. Concentration-dependent analysis of cysteine labeling by the IA probe can be used to rank cysteines by reactivity. Here, we applied isoTOP-ABPP to ER-enriched samples, in order to rank cysteines in the ER by reactivity. Isotopically encoded IA-alkyne (IA-light and IA-heavy) probes were utilized for this analysis³³. ER proteomes were treated with either 10 µM IA-heavy or 100 µM IA-light prior to copper(I)-catalyzed azide-alkyne cycloaddition (CuAAC) of a photo-cleavable biotin azide tag (Figure 2A and Supplemental Figure S3). Following CuAAC, light and heavy samples were combined, labeled cysteines were enriched on streptavidin beads, and subjected to on-bead trypsin digestion and UV exposure to release IA-tagged peptides for MS analysis. LC/LC-MS/MS analysis generated parent ion (MS1) spectra for peptide quantitation, and fragmentation (MS2) spectra for peptide identification. For each identified peptide, the light to heavy ratio (L:H) reports on the reactivity of the referenced cysteine residues. A highly reactive cysteine is expected to saturate labeling at low IA concentrations to generate low L:H values of ~1, while less reactive cysteines will show increased labeling in the 100 uM (IA-light) sample resulting in L:H ratios >>1. isoTOP-ABPP analysis of ER-enriched samples, determined L:H ratios for 923 cysteine-containing peptides from ER proteins (Figure 2B and Supplemental Table S5).

Figure 2.

Quantitative MS analysis of reactive and functional cysteines in isolated ER fractions. A) Schematic representation of the isoTOP-ABPP platform used to rank cysteine residues by reactivity. B) L:H ratio plot ranking secretory pathway cysteines by reactivity. Inset 1 is composed of cysteine residues with a ratio <2.5 (highly reactive cysteines) while inset 2 consists of cysteine residues with a ratio >7.5 (cysteines with low reactivity). Cysteines with no functional annotation are grey. Cysteine functional annotation was obtained from Uniprot and SwissPalm. Protein localization was determined using UniprotKB and GO databases.

Highly reactive cysteines (L:H <2.5) identified in the ER fraction were searched against a previous whole-cell analysis¹⁸ to evaluate whether subcellular fractionation reveals a novel subset of highly reactive cysteines. This comparative analysis revealed that ~85% (332 out of 388) of these highly reactive cysteines were unique to the ER. To further interrogate the relationship between cysteine reactivity and function in the ER, cysteines annotated as functional in Uniprot and SwissPalm databases were binned by isoTOP-ABPP ratio (L:H <2.5, 2.5-5.0, 5.0-7.5, and >7.5) (Supplemental Figure S4A and Supplemental Table S5). This analysis revealed that ~30% of all highly reactive secretory pathway cysteines were functionally annotated, which is comparable to the percentage of highly reactive cysteines identified as functional in the whole-cell proteome (Supplemental Figure S4B). Interestingly, the functions attributed to reactive cysteines in the ER differed significantly from total cellular proteins. Specifically, highly reactive cysteines (L:H <2.5) in the ER were enriched for disulfides and S-palmitoylated residues, compared to unreactive cysteines (L:H >7.5) (Figure 2B Insets 1 and 2). These cysteines are likely from partially matured proteins in the ER where disulfide bonds and S-palmitoyl units have yet to be installed. In contrast, disulfide and palmitoylated cysteines were significantly underrepresented in whole-cell datasets, likely because these cysteines are fully oxidized or S-palmitoylated, rendering them unreactive. Intriguingly, our data show that cysteines destined for disulfide-bond formation or S-palmitoylation display significantly elevated reactivity relative to other non-functional cysteines within the ER proteome. In addition to disulfide and S-palmitoylated residues, cysteines located within nucleotide-binding regions of proteins were identified in the ER but not whole-cell fractions, including cysteines in the highly conserved G-domain of several secretory pathway localized Ras superfamily members (Supplemental Figure S5). The majority of these G-domain cysteines belong to a G4 motif containing a conserved GNKCD motif that interacts with guanidine³⁹, and is known to be sensitive to oxidation^{40, 41}.

Adaptation of the OxICAT platform to interrogate cysteine oxidation in the ER.

The significant presence of reactive cysteines annotated to be sites of disulfide-bond formation triggered our interest in more directly interrogating the oxidation states of cysteines, and thereby differentiating fully oxidized ER cysteines from those that are fully or partially reduced. To achieve this, we implemented a proteomic platform combining ER enrichment with an adapted OxICAT method³⁴. For this analysis, ER fractions were isolated from SKOV-3 cells, and these microsomal fractions were immediately lysed in a denaturing buffer containing 10 mM IA-light. Reversibly oxidized cysteines were then reduced with tris(2-carboxyethyl)phosphine (TCEP) and alkylated with 10 mM IA-heavy. A photocleavable biotin-azide tag was appended to labeled cysteines via CuAAC, followed by streptavidin enrichment, trypsin digestion, photocleavage, and LC/LC-MS/MS analysis (Figure 3A and Supplemental Figure S3). Cysteine oxidation was determined by converting L:H and H:L ratios into a calculated value of percent oxidation (Supplemental Methods) for each individual cysteine.

Figure 3.

OxICAT analysis to determine cysteine oxidation states in the ER. A) Schematic representation of the OxICAT platform designed to quantify oxidation states of cysteines from the ER fraction. B) Plot of cysteine oxidation states in ER and secretory pathway proteins. Uniprot annotated cysteine residues that partake in disulfide bonds are colored blue. SwissPalm annotated sites of S-palmitoylation are colored red. Oxidation states are presented as an average percent oxidized, calculated directly from light:heavy and heavy:light ratios of cysteine residues, n = 2. C) Extracted ion chromatograms from OxICAT analysis for cysteine residues in CALR and ERp44. Uniprot annotated disulfide bonds are represented by a bridge. Displayed chromatograms are representative of a single replicate. Protein localization was determined using UniprotKB and GO databases.

Across two replicates, oxidation states were determined for 734 cysteines on ER proteins (Figure 3B and Supplemental Table S6). Of these cysteines, the majority (499, 68%) were highly reduced (<25% oxidation), and ~16% (116 peptides) were highly oxidized (>75% oxidized). Of these 116 highly oxidized cysteines, 70 cysteines (60%) were functionally annotated as structural disulfides. The remaining 46 cysteines have no functional annotation, however, it is likely that many of these residues are unannotated structural disulfides. For example, C192 and C208 (95% oxidized) in the CD151 antigen are part of a large extracellular loop (EC2) conserved in the tetraspanin superfamily. These residues are predicted to form intramolecular disulfides based on structural similarities with other tetraspanins^{42, 43}. Additionally, several unannotated oxidized cysteines belonging to proteins that are highly disulfide rich were identified, including six cysteine residues (C531, C1042, C1270, C1282, C1461, and C1476; 79-95% oxidized) in cation-independent mannose-6-phosphate receptor (IGF2R) and C615 (93% oxidized) in plexin-B2 (PLXNB2). Therefore, our data help to identify previously unannotated structural disulfides on ER proteins.

Notably, only three cysteines annotated as disulfides are found in a highly reduced state. These include C13 in S100A11 (5% oxidized), C343 in Annexin A1 (ANXA1) (7% oxidized), and C166 in ERO1A (21% oxidized). Of these, C13 in S100A11 has been shown to form an interchain disulfide upon S100A11 dimerization in the peroxisomes⁴⁴. Therefore, the ER-resident pool of S100A11 would be expected to be in the monomeric, cysteine reduced form as indicated by our data. In general, cysteines annotated as metal binding, S-nitrosated, or within nucleotide-binding domains were mostly reduced. Notably, S-palmitoylated cysteines were also generally highly reduced (91 residues) (Figure 3B and Supplemental Table S6), agreeing with the established observation that plasma membrane proteins primarily undergo S-palmitoylation in the Golgi prior to trafficking to the membrane⁴⁵, and are therefore found in the reduced and unmodified state within the ER.

The ability of the OxICAT platform to provide residue-specific oxidation information is underscored by several proteins where multiple cysteines with disparate oxidation states were detected. For example, the oxidation states for 3 out of 4 cysteines in CALR were quantified (Figure 3C), where two of these cysteines (C105 and C137) form a disulfide bond and were highly oxidized (86.17% and 92.65% oxidized, respectively), while the third cysteine (C163) was mostly reduced (27.57% oxidized) and is proposed to participate in a disulfide-bridged homodimer under physiological stress⁴⁶. Similarly, ERp44 is a multifunctional chaperone protein responsible for quality control in the early secretory pathway⁴⁷. Three cysteine oxidation states were quantified in ERp44. C301 and C318, which are annotated as a structural disulfide, were found to be highly oxidized (84.76% and 95.24% oxidized, respectively). C92 was identified in our data as mostly reduced (18.38% oxidized). These findings are in agreement with the available structural data on ERp44 (PDB 2R2J⁴⁸), where C301 and C318 are present as a disulfide, whereas C92 is present in the reduced thiol state. ERp44 is known to retain ERO1 in the ER through an interchain disulfide bond, but the cysteine involved in disulfide linkages with ERO1 (C58) was not identified in our analyses.

Upregulation of the UPR has minimal effects on cysteine oxidation state.

We applied the OxICAT analysis to interrogate changes in cysteine oxidation upon induction of the unfolded protein response (UPR). The accumulation of unfolded proteins in the ER activates the UPR by triggering three different signaling pathways: activating transcription factor-6α (ATF6), inositol-requiring 1α (IRE1α), and PKR-like ER kinase (PERK)^{7, 49, 50}. Several small molecules, including thapsigargin (Tg) and tunicamycin (Tn) (Figure 4A), target different homeostatic mechanisms in the ER to activate the UPR. Tg is a sesquiterpene lactone that non-competitively inhibits the sarco/endoplasmic reticulum Ca²⁺-ATPase (SERCA). Inhibiting SERCA-mediated Ca²⁺ import decreases the activity of Ca²⁺ binding chaperones such as CALR, CALX, and BiP^{51, 52}, leading to an accumulation of unfolded proteins. Tn is a nucleoside antibiotic that inhibits GlcNAc phosphotransferase (GPT), preventing the formation of preassembled oligosaccharides required for N-linked glycosylation in the ER. Inhibition of N-glycosylation hinders proper protein folding, and additionally prevents lectin chaperones⁵³ from recognizing misfolded proteins, leading to activation of the UPR. Although none of these chemical UPR modulators directly target disulfide-bond forming pathways in the ER, we were interested to study the effects on cysteine oxidation since dysregulation of protein homeostasis could potentially indirectly alter disulfide-bond formation within the ER.

Figure 4.

Quantification of changes to ER cysteine oxidation states following induction of the unfolded protein response (UPR). A) Structures of UPR activating agents, Tg and Tn. B) Activation of UPR target genes after 5 hours of treatment was evaluated by qRT-PCR following treatments with DMSO, Tg (5 µM), or Tn (3 µM). Data was plotted as a fold change in mRNA expression, Tg or Tn relative to DMSO, and represents the mean ± s.d. of two replicates. * P value <0.05, *** P value <0.001, unpaired t test. C,D) ReDiMe MS analyses of unenriched ER fractions upon 5 hours of Tg and Tn treatment. Proteins above (>1) or below (<-1) dotted lines are more than 2-fold upregulated or downregulated, respectively. C) Log₂ ratio plot (Tg:DMSO) of identified ER and secretory pathway proteins. D) Log₂ ratio plot (Tn:DMSO) of identified ER and secretory pathway proteins. E) Plot comparing secretory pathway cysteine oxidation states in DMSO and Tg (5 hr) treated samples. F) Plot comparing ER cysteine oxidation states in DMSO and Tg (24 hr) treated samples. Light colored regions indicate cysteine residues that show >10% change in cysteine oxidation state, and dark colored regions indicate cysteine residues that show >50% change in oxidation state. Protein localization was determined using UniprotKB and GO databases.

UPR activation upon administration of Tg and Tn was confirmed using quantitative polymerase chain reaction (qPCR). Briefly, SKOV-3 cells were treated with DMSO, Tg (5 µM), or Tn (3 µM) for 5 hours and qPCR analysis was performed for known downstream targets from each arm of the UPR, including GRP94 (ATF6-regulated), ERDJ4 (IRE1α-regulated), and CHOP (PERK-regulated). As expected, all three arms of the UPR were activated following Tg and Tn treatments (Figure 4B).

Prior to examining cysteine-oxidation changes upon Tg and Tn treatments, unenriched proteomic analyses were performed to determine changes in protein abundance in ER fractions from UPR-upregulated cells. Quantitative proteomic analyses were performed upon isotopic tagging of peptides using reductive dimethylation (ReDiMe) of primary amines⁵⁴. Peptides resulting from the DMSO-treated sample were dimethylated with light ReDiMe reagents while Tg or Tn samples were dimethylated with heavy ReDiMe reagents for LC/LC-MS/MS analysis. Across two replicates, we quantified 656 and 632 ER proteins in DMSO:Tg (Figure 4C and Supplemental Table S7) and DMSO:Tn samples (Figure 4D and Supplemental Table S8), respectively. While we were able to confirm activation of the UPR at the transcriptomic level for both treatments, only Tg treatments led to increased expression of proteins known to be associated with UPR activation. PDIs (e.g. PDIA3, PDIA4 and PDIA6), HSPs (e.g. HSPA5, HSPA13, and HYOU1), and N-linked glycosylation and lectin-assisted folding chaperones (e.g. STT3A and CALR) showed a greater than two-fold increase following Tg treatment (Figure 4C). Furthermore, the protein abundance changes following Tg treatment were similar to those reported in a previous proteomic study after independent activation of the IRE1α and ATF6 arms of the UPR⁵⁵. Surprisingly, very few changes in protein abundance were observed for Tn-treated samples (Figure 4D and Supplemental Table S8), where folding chaperones were relatively unchanged. This general proteome profile was similar to a recent study using stable isotope labeling with amino acids in cell culture (SILAC) to profile changes in protein abundance following Tn treatments²¹. Additionally, the differences between Tn-treated transcriptomic and proteomic data could be attributed to known discrepancies between mRNA expression and protein expression or a temporal delay between transcriptional and translational machineries or both^{56, 57}.

To explore perturbations to disulfide-bond formation upon UPR activation, the OxICAT platform was implemented in Tg and Tn treated cells. As before, SKOV-3 cells were treated with DMSO, Tg (5 µM), or Tn (3 µM) for 5 hours and subjected to the OxICAT analysis (Figure 3A) (Supplemental Table S9 and S10). Percent oxidation for each cysteine upon Tn and Tg treatment was compared to the DMSO control (Figure 4E and Supplemental Figure S6). The majority of identified cysteine residues showed minimal changes in oxidation state upon Tg and Tn treatment (grey circles, <10% change in oxidation state). However, 105 cysteines (21.0%) and 68 cysteines (19.9%) in Tg and Tn treated samples, respectively, exhibited a change larger than 10% (black circles). For example, C163 in CALR become >10% more oxidized following Tg treatments. This observation is supported by a previous study that found that Tg-mediated ER Ca²⁺ depletion resulted in the formation of a CALR disulfide bound homodimer at C163⁵⁸. Only three cysteines displayed large (>50%) changes in oxidation. These include C312 in neural cell adhesion molecule L1 precursor (L1CAM) (76% decrease in oxidation), C1257 in myopalladin (MYPN) (65% decrease in oxidation), and C657/C663 in lysyl oxidase (LOXL2) (58% decrease in oxidation). C312 in L1CAM is annotated as a structural disulfide. It has been shown that L1CAM expression is Ca²⁺ dependent, whereby Tg treatment leads to increased expression⁵⁹. This Tg-induced increase in L1CAM expression may overwhelm folding chaperones producing a misfolded population where the C312 disulfide is not properly installed. The identified cysteines (C657 and C663) on LOXL2 are also annotated as a structural disulfide. Interestingly, overexpression of LOXL2 is known to induce the UPR by interacting with BiP⁶⁰, but the effect of UPR activation on LOXL2 function are unknown. Our data suggests that UPR activation by Tg results in partial misfolding of LOXL2 due to improper disulfide-bond formation. C1257 in MYPN is not annotated as a structural disulfide, and this protein has not previously been implicated in the UPR.

Since the changes observed after 5hr treatments with Tg and Tn were relatively minimal (only 3 cysteines with >50% change in oxidation state), we sought to determine if longer incubation times with Tg would show more dramatic perturbations to disulfide-bond formation. We collected OxICAT data after 24hrs of incubation with Tg (Figure 4F) (Supplemental Table S11). Surprisingly, fewer changes to cysteine oxidation were observed at the longer time point, suggesting that both acute and sustained UPR activation has minimal effects on disulfide-bonding capacity of the ER, likely due to upregulation of the PDIs and folding chaperones that ensure proper disulfide-bond formation on ER proteins.

In conclusion, we have developed a proteomic platform to quantify cysteine reactivity and oxidation in secretory pathway proteins localized in the ER. Previous studies on ER cysteines relied on isoTOP-ABPP analysis of whole-cell lysates¹⁸ or the use of fluorinated-rhodol tagged alkylating reagents (ERMs)²¹, but both strategies were plagued by low ER cysteine coverage. Using subcellular fractionation by differential centrifugation prior to proteomic analysis, we enrich for ~40% of the Human Protein Atlas annotated ER proteome, a 4-fold enrichment compared to whole-cell lysates. Combining ER enrichment with isoTOP-ABPP, we ranked 923 cysteine residues from 523 ER proteins by reactivity, representing a significant increase in identified cysteine residues relative to previous studies^{18, 21}. Highly reactive cysteines were annotated as destined for S-palmitoylation or disulfide-bond formation, highlighting a subset of cysteine functionality typically not identified in whole-cell cysteine reactivity studies.

To investigate cysteine oxidation in the ER, we combined ER isolation with OxICAT to quantify the percentage oxidation for 734 cysteine residues on ER proteins. Highly oxidized cysteines were enriched in annotated structural disulfide bonds while S-palmitoylated residues were predominantly present in the reduced state. These results were expected, given the spatiotemporal introduction of each cysteine modification within the cell, whereby disulfide-bond formation occurs in the ER, resulting in annotated disulfides identified as primarily oxidized, whereas S-palmitoylation occurs primarily in the Golgi, and thereby S-palmitoylated cysteines in the ER would be expected to be in the reduced, unmodified state. The OxICAT platform was applied to evaluate changes in cysteine oxidation following chemical induction of the UPR with Tg or Tn. In general, these treatments resulted in minimal changes to cysteine oxidation, with a few proteins, such as L1CAM, displaying dysregulated disulfide-bond formation after 5 hrs of Tg treatment. One limitation of the OxICAT approach is that artifactual cysteine oxidation on ER proteins could occur during organelle fractionation. Capping reduced cysteines prior to organelle fractionation would better preserve endogenous oxidation states through the lengthy fractionation protocol, however the requirement for complete protein denaturation prior to labeling with the IA-light probe precludes cysteine labeling pre-fractionation. As a means to minimize artifactual oxidation, we ensure that the ER membrane is preserved during the fractionation steps, and cysteine capping is performed concurrent to ER membrane lysis.

Together, these studies demonstrate the advantages of combining organelle isolation with MS-proteomic platforms such as isoTOP-ABPP and OxICAT to improve coverage of low-abundant proteins in organelles such as the ER. These chemoproteomic platforms can be applied to investigate proteomic changes induced by direct perturbation of ER disulfide-bond forming chaperones, and in misfolding diseases, such as prion-related disorders and ALS, where changes in disulfide-bond formation are proposed to induce protein aggregation⁶¹. Further improvements in fractionation methodology, at the organelle, protein and/or peptide level, together with improved sensitivity of new MS instrumentation, can generate even more improved coverage of the ER cysteinome.

Methods

See Supporting Information for detailed experimental protocols for ER isolation, qPCR, mass spectrometry preparation, and data analysis.