347 stainless steel coiled tubing chemical component ,Identification of novel interferon-responsive human leukocyte antigen-A (HLA-A) chaperone proteins using cross-linked mass spectrometry (CLMS)

Thank you for visiting Nature.com. You are using a browser version with limited CSS support. For the best experience, we recommend that you use an updated browser (or disable Compatibility Mode in Internet Explorer). In addition, to ensure ongoing support, we show the site without styles and JavaScript.
Sliders showing three articles per slide. Use the back and next buttons to move through the slides, or the slide controller buttons at the end to move through each slide.

Product Description

Stainless Steel 347L Coil Tubes, Steel Grade: SS347L

The interferon signaling system induces a strong cytokine response to a wide range of pathogenic and intrinsic pathological signals from the environment, resulting in the induction of subsets of interferon-inducible proteins. We applied DSS-mediated cross-link mass spectrometry (CLMS) to detect new protein-protein interactions in the domain of interferon-induced proteins. In addition to the expected interferon-inducible proteins, we also identified novel intermolecular and intramolecular cross-linked adducts of canonical interferon-inducible proteins such as MX1, USP18, OAS3, and STAT1. We focused on orthogonal validation of a novel set of interferon-inducible protein networks formed by HLA-A proteins (H2BFS-HLA-A-HMGA1) using co-immunoprecipitation and their further study using molecular dynamics modeling. Modeling of the conformational dynamics of the protein complex revealed several interaction sites that reflected the interactions identified in the CLMS findings. Together, we present a pilot study of CLMS to identify new signaling complexes induced by interferon, and look forward to the wider use of CLMS to identify new dynamics of protein interactions in the tumor microenvironment.
Before an adaptive immune response begins, the host’s innate defense system mounts an antimicrobial response mediated by a family of secreted alpha-helical cytokines called interferons (IFNs). The type I IFN classes IFNα and IFNβ activate cellular responses, including antiviral, proapoptotic, proinflammatory, and antiproliferative states. In humans, 13 subtypes of IFNα are known, all clustered on chromosome 91. Surprisingly, only IFNα2 has been studied for clinical use. Recently, special attention has been paid to research on other subtypes of IFNα. A recent study showed that IFNα14 is one of the most effective isoforms in restricting HBV2 and HIV-13,4 replication compared to the canonical IFNα2 subtype.
It has been established that activated type I interferon receptor complexes (IFNAR1 and IFNAR2) trigger a signal transduction cascade mediated by Janus kinases TYK2 and JAK15,6. These Janus kinases phosphorylate signal transducers and transcriptional protein activators (STAT1 and STAT2) on tyrosine residues to initiate SH2 domain-mediated heterodimerization6. Subsequently, IRF9 binds STAT heterodimers to form a trimeric complex of the IFN-stimulated factor 3 gene (ISGF3), which translocates to the nucleus and induces the transcription of over 2000 interferon-stimulated genes (ISGs)5,6,7,8.
ISGs form the backbone of the innate immune system, especially in response to viral attack. As a first line of defense against viral infection, cells rapidly deploy extensive interactions of cellular proteins with a wide range of biological activities. These proteins include pattern recognition receptors, signaling molecules, transcription factors, and proteins with direct antiviral functions, as well as negative regulators of immune responses9. Much of the information on ISG activity comes from functional screens using overexpression screens10,11 or gene silencing techniques (siRNA, RNAi and CRISPR)12,13 in which individual ISGs are expressed or inhibited and their activity is tested on various viruses. Although these studies have determined the antiviral properties of individual ISGs, the underlying molecular mechanisms of each ISG remain largely unknown. It is generally accepted that many proteins interact with one or more cytokines to ensure full activity, so either ISGs interact directly or their interactions are mediated by cellular proteins. For example, a recent photocrosslinked proteomics study identified ATPase VCP/p97 as a major IFITM3 interaction partner, whose inhibition leads to defects in lysosomal sorting, turnover, and cotransport of IFITM3 with viral particles 14 . Using immunoprecipitation, we identified VAPA, a vesicle-associated protein, as an interaction partner with IFITM1/2/3 that mediates cholesterol-mediated viral maturation, and this was confirmed by another study using a yeast two-hybrid system. Scientific Support 15 , 16 .
A fundamental biological process involved in the suppression of infection and malignant transformation is antigen presentation, which is mediated by major histocompatibility complex (MHC) molecules. Peptides (8-12 amino acids long) from cleaved, prematurely terminated or misfolded proteins are loaded into the MHC-I heterodimer (consisting of MHC-I heavy and light chains, called β-2-microglobulin; β2M) 17,18 . The resulting stable MHC-I trimers are transported to the cell surface, where they present intracellular peptides to CD8+ T cells (cytotoxic T cells)17. T cells recognize and destroy these pathogens and cells carrying a tumor-specific antigen. Consequently, pathogens and tumor cells often suppress the antigen presentation process to avoid immune surveillance. In addition, MHC-I is downregulated in 40-90% of human tumors and is often associated with a poorer prognosis19.
Genes involved in responding to pathogens must quickly switch between a state of rest and a state of active transcription. Therefore, several cellular proteins are hypothesized to be involved in the response to high IFN demand over short periods of time, including remodeling and modification of promoter chromatin 20,21 . Most studies have focused on the identification of individual ISG protein partners in the presence of IFN. Several proteomic and transcriptomic studies in model cell systems have elucidated the effect of IFN on the cellular landscape. However, despite a growing understanding of the dynamics induced by interferons, we still know little about the involvement of ISGs. When considering the complexity and time-dependent dynamics of interferon signaling, two questions arise: (i) is it possible to stabilize and trap the multiprotein complexes involved in fast signaling, and (ii) can these interactions be mapped into 3D space?
To address these issues, we implemented disuccinimide suberate-mediated chemical cross-linking (DSS) coupled with mass spectrometry (CLMS) to study the IFNα-induced protein interaction network and its dynamics. DSS adds covalent bonds between proximal residues of proteins and/or protein complexes in vivo. Subsequent MS analysis reveals specific crosslinking sites that reflect the spatial proximity of regions within a particular protein, called internal linkages, or subunits in protein complexes, called interrelationships. Using this approach, we have identified several novel protein-protein complexes as well as interferon-induced multiprotein interaction networks. By further testing a subset of these new interactions, we demonstrate that H2BFS (H2B histone-type FS; hereinafter referred to as H2B) and MDN1 act as binding partners for HLA-A.
Flo-1 cells are one of the best known in vitro models of esophageal adenocarcinoma as they mimic key features of esophageal tumors22,23. However, not all tumors are immunogenic, and to determine if Flo-1 cells respond to interferon treatment, we treated Flo-1 cells with 10 ng/ml IFNα for 72 hours. Flo-1 cells showed early induction of pSTAT1 and IRF1, starting 2 hours after treatment and continuing for 72 hours, with a time-dependent decrease in stationary levels of IRF1 (Figure 1A). ISGs (MX1, IFITM1, OAS1/2, and ISG15) were found to be strongly induced after 6 hours, mimicking the classic mid and late phase responses to IFNα (Figure 1A). Together, these data suggest that this cellular model can be used to study interferon responses.
Differential protein expression responses in Flo-1 cells after IFNα treatment. (A) Protein expression in Flo-1 cells treated with 10 ng/ml IFNα for 2, 6, 24, 48 and 72 hours was analyzed by immunoblot using the indicated ISG antibodies. (B) Coomassie blue stained SDS-PAGE gels of whole cell extracts after cross-linking with DSS for indicated times and concentrations. (C) Representative immunoblot examined with p53(DO-1) antibody from the same samples to assess the degree of protein cross-linking.
To capture the in situ protein interaction landscape, we used DSS, a widely used cross-linking agent due to its high membrane permeability and relatively short reaction time. The shorter reaction time helps to prevent the formation of large aggregates of crosslinked proteins, thereby maintaining the stability of the crosslinker. To determine the optimal DSS concentration and avoid over-crosslinking, we first exposed the cells to 5, 2.5, and 1 mM DSS for 5, 10, 5, and 30 minutes, respectively, and analyzed the lysates by Coomassie-stained SDS-PAGE (data not shown) . Cell lysates appear to be highly cross-linked at the lowest concentration and at the shortest time point. Therefore, DSS was titrated to 1, 0.5, and 0.1 mM over 5 minutes (Figure 1B). Optimal crosslinking was observed with 0.5 mM DSS for 5 minutes, and these conditions were chosen for cells treated with IFNα. In addition, Figure 1C shows a Western blot performed using the p53 (DO-1) antibody to assess the degree of protein cross-linking.
Flo-1 cells were treated with 10 ng/ml IFNα for 24 hours before adding the crosslinker. Cross-linked cells were subsequently lysed by two-step proteolysis and proteins were processed by FASP (Fig. 2)24,25. Cross-linked tryptic peptides were analyzed by mass spectrometry (Fig. 2). The MS/MS spectra are then matched to the protein sequence and quantified with MaxQuant26,27. Cross-linked peptides were identified from the obtained spectra using the SIM-XL program, and individual compounds were combined into a complex network using the xQuest28 and SIM-XL29 open source computing software pipelines (Fig. 2). SIM-XL identifies protein-protein interactions, internal chains and individual chains in simple or complex protein mixtures and provides scripts for visualizing interactions in protein structures. In addition, it ranks each cross-reference as an ID score according to the MS/MS29 spectrum quality. Several highly reliable protein-protein interactions and complexes have been identified, and a new set of interactions has been further investigated using co-immunoprecipitation and conformational changes of complexes using molecular dynamics (MD) modeling (Fig. 2) 30, 31.
Schematic overview of the CLMS method. Flo-1 cells were treated with 10 ng/ml IFNα for 24 hours followed by in situ protein cross-linking using DSS followed by cell lysis and trypsinization. Cross-linked samples were analyzed using an Orbitrap mass spectrometer and further sampled for fragmentation of peptide precursors during LC-MS/MS. Two linked peptides were identified from the obtained spectra using the Spectrum Recognition Machine of the Crosslinked Peptides (SIM-XL) program, and all compounds were combined into a complex network using computational pipelines. Filter out low confidence interactions based on false positive rate (FDR) scores. Several new high-fidelity protein-protein interactions were further confirmed using co-immunoprecipitation, and conformational changes in the complexes were examined using molecular dynamics (MD) modeling.
A total of ~30,500 and ~28,500 peptides were detected using MaxQuant in unstimulated and stimulated IFNα samples, respectively (Supplementary Table S1, Fig. 3A). The peptide length distribution in both cases showed a higher proportion of larger peptides, indicating the presence of cross-linked peptides (Fig. 3B,C). In addition, a larger proportion of larger peptides were present in the 40–55 range in IFNα-treated samples (Fig. 3C). Protein mapping against log2 intensity showed that classic interferon-stimulated proteins were the most abundant compared to untreated samples, including MX1, IFIT1/3, OAS2/3, DDX58, and HLA-F (Figure 3D). Analysis of pathways for proteins more than three-fold enriched in response to IFNα treatment using the Reactome pathway database showed that MHC-I-mediated antigen presentation and processing was the most dominant pathway (Figure 3E). Consistent with earlier reports, antiviral responses mediated by OAS and ISG15 as well as IFNα/β and cytokine signaling were among the pathways activated. In addition, lysine- and serine-specific protein cross-links were identified from the originally acquired MS/MS spectra using SIM-XL. A recent study reported 104 ISGs spanning 20 viruses from 9 virus classes by meta-analysis of individual ISG overexpression studies in 5 cell types9. However, to overcome the computational limitations of screening large datasets, we started with a smaller dataset to explore possible interactions between the list of IRDS genes reported by Padaria et al., most of which are ISGs.
Identification of differentially expressed cross-linked proteins in response to IFNα (data obtained from MaxQuant). (A) Venn diagram representing the number of common and exclusive peptides identified in IFNα14 treated and untreated Flo-1 samples. Peptide length distribution of untreated (B) and IFNα treated (C) crosslinked samples. (D) Heat map representing log2 (LFQ intensity) between untreated and IFNα14 treated Flo-1 cells. The left panel shows the proteins most actively activated in the presence of IFNα. (E) Histogram representing the 20 major enrichment pathways after IFNα treatment. The Reactome pathway database analyzed more than four-fold changes in upregulated IFNα-responsive proteins.
Interferon-mediated ISG stimulation is well documented, but at the molecular level it is poorly understood how these proteins culminate in a wide range of biological functions. We investigated protein interactions with a high degree of confidence between known ISGs. Interestingly, we identified a network including MX1, USP18, ROBO1, OAS3, and STAT1 proteins that form a large complex in response to IFNα treatment (Figure 4, Table S2) 32,33,34. Most importantly, these interactions were found in all triplicates treated with IFNα and were not found in untreated samples, suggesting that they were formed specifically in response to IFNα treatment. It is known that STAT1 transcriptionally regulates the expression of these ISGs, but its interaction with ISGs at the protein level has not been studied. The crystal structure of STAT1 showed that its helical domain (CCD) is not involved in the interaction with DNA or protomers during the formation of dimers35. These α-helices form a helical helix structure that provides a predominantly hydrophilic surface area for interactions to occur 35 . In our CLMS data, we observed that most of the interactions with STAT1 occurred in the SH2 domain preceding the CCD, the linker domain, or the C-terminal tail (residues 700-708) (Figure 4A). A previous study reported that USP18 binds to the CCD and DNA-binding domain (DBD) of STAT2 and is recruited to the subunit of the type I interferon receptor IFNAR2 to mediate inhibition of type I interferon signaling 24 . Our data also showed that the USP18 catalytic domain interacts with STAT1 DBD (Figure 4A,D), suggesting that both STAT1 and STAT2 may play a role in attracting USP18 to IFNAR2.
Protein-protein ISG network identified in cross-linked cells treated with IFNα. (A) 2D interaction plot showing protein-protein interactions (generated in the SIM-XL program), with lines representing intermolecular interactions (crosslink cutoff set to 3.5). Domains of different identities are marked by their color32: MX1 domain, Dynamin_N (73–249), Dynamin_M (259–547), and GED (569–660). OAS3 domains: OAS1_C (160-344), OAS1_C (559-745), NTP_transf_2 (780-872), and OAS1_C (903-108). Domain ROBO1, Ig_3 (67–151), I-set (170–258), I-set (262–347), Ig_3 (350–432), Ig_3 (454–529), fn3 (562–646), fn3 (678–758) and fn3 (777–864). STAT1 fields: STAT_int (2–120), STAT_alpha (143–309), STAT_bind (321–458), SH2 (573–657), and STAT1_TAZ2bind (715–739). (B) Circular viewer of cross-linked proteins (MX1, UBP18, OAS3, ROBO1, and STAT1) with interactions and interactions labeled in blue and red, respectively. The cross-link threshold was set at 3.5. Dot plots indicate STAT1 interaction sites with MX1 (C), USP18 (D), ROBO1 (E), and OAS3 (F), as well as K or S interaction sites between the two peptides. In the figure, the cross-link score threshold is set to 3.0. (G) Various interaction sites between STAT1 and OAS3 DI domains superimposed on their protein structures in PyMol (PyMOL molecular graphics system, version 2.0 Schrödinger, LLC.); STAT1 (pdb id: 1bf533) and OAS3 (pdb id: 4s3n34). ) program.
Two isoforms of USP18 have been described in humans, a full-length protein that is predominantly located in the nucleus, and an isoform without an N-terminal domain, USP18-sf, which is evenly distributed in the cytoplasm and nucleus 36 . In addition, the N-terminus was predicted to be unstructured and does not require isopeptidase activity or ISG1537 binding. Most of the interactions identified in our study were located at the N-terminus of the protein, suggesting that these interactions involve full-length USP18 (Figure 4A,D) and thus likely occur in the nucleus. Moreover, our data also indicate that the N-terminus is specialized for protein-to-protein interactions. The IFNAR2 binding site is located between residues 312-368, and notably, none of the proteins in the complex bind to this region (Fig. 4A) 37,38 . These data taken together indicate that the IFNAR2 binding domain is exclusively used by the receptor protein. In addition, only OAS3 and ROBO1 were found to be associated with domains upstream of the N-terminus and IFNAR2 binding site (Figure 4A).
ROBO1 belongs to the immunoglobulin (Ig) superfamily of transmembrane signaling molecules and consists of five Ig domains and three fibronectin (Fn) domains in the extracellular region. These extracellular domains are followed by a membrane-proximal region and a single transmembrane helix 39. An unstructured intracellular region is located at the C-terminus and contains conserved sequence motifs that mediate effector protein binding39. The region extending from amino acids ~1100 to 1600 is mostly disordered. We found that MX1 interacts with ROBO1 through Ig, Fn, and intracellular domains, while most interactions with STAT1 occur between its CCD, linker domain, and the C-terminus of ROBO1 (Fig. 4A,E). On the other hand, interactions with DI, DIII, and OAS3 linker regions were distributed throughout the ROBO1 protein (Fig. 4A).
The oligoadenylate synthase (OAS) protein family accepts and binds intracellular double-stranded RNA (dsRNA), undergoes conformational changes, and synthesizes 2′,5′-linked oligoadenylates (2-5 As) 40 . It was found that among the three OASs, OAS3 exhibits the highest affinity for dsRNA and synthesizes the least amount of 2-5 As, which can activate RNase L and thereby limit viral replication 41 . The OAS family consists of polymerase beta (pol-β)-like nucleotide transferase domains. Previous research has shown that the catalytic activity of the C-terminal domain (DIII) is dependent on the dsRNA-binding domain (DI), which is required for the activation of OAS342. We observed that the DI and DII domains of OAS3 interact with CCD and a small junction region between SH2 and STAT1 TAD (Figure 4A,F). Overlaying different crosslinking sites on the protein structure revealed an interaction between the β-sheet and DBD STAT1 loop and an open pocket or cavity formed by residues 60–75 in the DI domain of OAS3 (Fig. 4G). The orientation of the proteins in the complex also indicated that none of the interactions with OAS3 interfered with the DNA-binding ability of its DI domain (Fig. S1A). In addition, the N-terminal domain of GTPase MX1 interacts extensively with the DI and DIII domains of OAS3 (Fig. 4A). We also observed an interaction between OAS1 and MX1 in all three IFNα-treated repeats, where a single OAS1 domain (also catalytically active) interacted with all three MX1 domains (Figure S2A,B).
MX proteins are part of a large family of dynein-like GTPases that contain an N-terminal GTPase domain that binds and hydrolyzes GTP, an intermediate domain that mediates self-assembly, and a C-terminal leucine zipper that acts as a GTPase (LZ). domain effector domain25,43. MX1 binds to subunits of viral polymerases to block transcription of the viral gene43. A previously reported yeast two-hybrid screen showed that PIAS1-associated MX1 inhibits STAT1-mediated gene activation by blocking DNA-binding activity and also has SUMO E344,45 ligase activity. Here, we demonstrate that MX1 binds to STAT1 (Figure 4C,D), however how this interaction affects STAT1-mediated gene activation in response to IFNα needs further study. In addition, we also found that MX1 interacted with IFIT3 and DDX60 in all three IFNα-treated repeats (Fig. S2C).
DDX60 is an IFN-induced cytoplasmic helicase that has previously been reported to play a role in RIG-I-independent degradation of viral RNA46. It interacts with RIG-I and activates its signaling in a ligand-specific manner 46. DDX60 consists of a DEXD/H-Box helicase domain and a C-terminal helicase domain that bind viral RNA and DNA47. Most of its interactions with MX1 and IFIT3 occur within long N- and C-terminal regions without canonical domains or motifs (Fig. S2E,F). However, MX1 is also associated with the DEXD/H-Box helicase domain (Fig. S2E). Proteins of the IFIT family have tandem copies of a distinctive helix-turn-helix motif called the tetrapeptide repeat (TPR). IFIT3 was found to be a positive modulator of RIG-I signaling and hence a component of the MAVS complex. Taken together, our data suggest that IFIT3 and DDX60 interact primarily in the region between TPR 3–6 of IFIT3 and may play a role in RIG-I/MAVS signaling (Fig. S2F).
Given that screening the entire proteome is computationally intensive, we then screened the entire human UniProt database for the presence of one of the IFNα-treated repeats. In this replica, we found several highly reliable interaction networks for HLA-A. Analysis of protein pathways identified by MS/MS spectra showed that MHC-I-based antigen processing and presentation is the main pathway induced by interferon (Fig. 3D). Therefore, we focused on studying the protein interactions of MHC-I molecules with a high degree of confidence in all cross-linked samples. HLA consists of α1, α2 and α3 domains and light chains, and microglobulin β2 (β2m) is a constant chaperone protein49. Once assembled in the endoplasmic reticulum, HLA is unstable in the absence of peptide ligands50. The peptide-binding groove is formed by the highly polymorphic and unstructured α1 and α2 domains in non-peptide form and the relatively less polymorphic α351 domain. In the presence of IFNα, we detected two HLA-A complexes: one interacts with HMGA1 and H2B (Figure 5, Table S3) and the other interacts with MDN1, LRCH4 and H2B (Figure 6).
IFNα induces an HLA-A interaction network with H2B (H2BFS) and HMGA1. (A) 2D plot (generated in SIM-XL software) depicting different types of interactions in the H2B-HLA-A-HMGA1 complex: interlink (blue), interlink (red) and single link (black). . Domains of different identities are color coded32: H2B (histone; 2–102) and MHC-I (MHC_1; 25–203, group C1; 210–290 and MHC_I_C; 337–364). The cross-link threshold was set at 3.5. Dot plots indicate HLA-A interaction sites with H2B (B) and HMGA1 (C), as well as K or S interaction sites between the two peptides. In the figure, the cross-link score threshold is set to 3.0. (D) Relationships between proteins shown in the structures of the H2B, HLA-A, and HMGA1 proteins in the PyMOL program. These structures were modeled using the Phyre2 server (http://www.sbg.bio.ic.ac.uk/phyre2) and the template structures for the H2B, HLA-A and HMGA1 proteins were 1kx552, 1kj349 and 2eze55, respectively.
IFNα induces an HLA-A interaction network with H2B (H2BFS), MDN1 and LRCH4. (A) Intramolecular (red) and intermolecular (blue) crosslinks presented on a 2D interactive map (generated in SIM-XL software) with MDN1 represented as a circle. The cross-link threshold was set at 3.5. Domains of different identities are color coded32: H2B (histone; 2–102), MHC-I (MHC_1; 25–203, group C1; 210–290 and MHC_I_C; 337–364) and LRCH4 (LRR_8 (68–126), LRR_8 (137–194) and CH (535–641)). (B) Relationships between proteins shown in the structures of the H2B, HLA-A, LRCH4, and MDN1 proteins in the PyMOL program. These structures were modeled using the Phyre2 server (http://www.sbg.bio.ic.ac.uk/phyre2) with template structures 1kx552, 1kj349, 6hlu62 and 6i2665 for the H2B, HLA-A, LRCH4 and MDN1 proteins, respectively. Dot plots showing K or S interaction sites for HLA-A with H2B (C), LRCH4 (D), and MDN1 (E). For plots, the cross-link score threshold was set to 3.0.
In addition to maintaining the integrity of the genome, histone H2B is also involved in the regulation of transcription. The H2B protein consists of a central histone domain (HFD) formed by three α-helices separated by loops and a C-terminal tail 41,52. Most of the interaction with H2B occurs in the α1 helix, which provides trimerization with the HFD heterodimer (Fig. 5A,B). Although lysins are involved in DNA binding, some lysins are also alternative acetylation or methylation sites. For example, residues K43, K46, and K57 from H2B are not involved in direct DNA binding, but are targets of various post-transcriptional modifications53. Similarly, residues K44, K47, and K57 in H2B may play an alternative role in the presence of IFNα, including interactions with other proteins (Fig. 5A, B). In addition, the extrachromosomal histone H2B activates the immune response in various cell types, acting as a cytosolic sensor to detect double-stranded DNA (dsDNA) fragments derived from infectious agents or damaged cells54. In the presence of DNA viruses, H2B depletion inhibited IFN-β production and STAT154 phosphorylation. H2B is also known to move in and out of the nucleus faster than other core histones54. H2B interactions with MDN1 and LRCH4 were also observed in selected untreated samples. We found that HLA-A interacted with H2B in all three IFNα-treated samples and in one untreated repeat sample. These data reflect the role of H2B in an alternative physiological function independent of transcriptional regulation.
HMGA1 (high mobility group AT-Hook 1), a small nucleoprotein rich in disease-promoting amino acids, has been identified in association with HLA-A. It has an acidic C-terminal tail and three distinct DBDs called AT hooks because they bind to the minor groove of the AT-rich region in dsDNA55,56. This binding causes the DNA to bend or straighten, allowing canonical transcription factors to access its consensus sequence. The C-terminal tail is believed to be involved in protein-protein interactions and the recruitment of transcription factors, since C-terminal deletion mutants are unable to initiate transcription57. Moreover, this domain contains several conserved phosphorylation sites that are known substrates for kinases 58 . We observed HLA-A and H2B interactions with HMGA1 outside the C-terminal domain, suggesting that the C-terminal domain is mainly used for transcription factor binding (Fig. 5A, C). HMGA proteins compete with histone H1 for binding to adapter DNA, thereby increasing accessibility57. Similarly, it seems likely that HMGA interacts with histone H2B along the linker DNA in competition with histone H1. HMGB1 induces the expression of HLA-A, -B, and -C in dendritic cells, leading to their activation59, but an interaction between HMG and HLA has not been previously reported. We found that HMGA1 interacts with the α1 and α3 domains of HLA-A, with most of the interactions outside of its 3 DBD (Figure 5A,C). In our hands, HLA-A was found to be localized in the nucleus (data not shown), and given that H2B and HMGA1 are also present in the nucleus, this interaction likely occurs in the nucleus. Specific adducts measured between H2B, HLA-A, and HMGA1 are shown in Figure 5D.
Most interactions of HLA-A with other proteins occur within its α1 and α2 domains and the disordered C-terminal domain (Fig. 6). In one of these examples, we found that HLA-A interacts with the disordered N-terminal tail of LRCH4 (Figure 6A,D). LRCH4 regulates TLR4 activation and LPS cytokine induction, thereby modulating the innate immune response60,61. It is a membrane protein with nine leucine-rich repeats (LRRs) and a calmodulin (CH) homology motif in its ectodomain, followed by a transmembrane domain (TMD) 60 , 62 . CH domains have been reported to mediate protein-protein interactions 60 . A stretch of about 300 amino acids between the LRR and CH domains is relatively accessible but disordered. Based on the function of disordered regions as mediators of protein-protein networks and vesicular transport 63 , we found that most protein interactions occur in disordered regions. Interactions with MDN1 were distributed throughout the length of the protein, including the LRR1, LRR6, CH domains, and random regions, while H2B mainly bound to the CH domain (Fig. 6A, B). Notably, none of the interactions included the TMJ, suggesting the specificity of the CLMS approach (Figure 6A, B).
MDN1 has also been identified as part of the HLA-A protein network (Figure 6A). It belongs to the AAA family of proteins (ATPases associated with different activities). This is the same N-terminal AAA domain that organizes into a hexameric ring and removes the assembly factor from the 60S 64 ribosomal subunit. appears to be similar to dynein64,65,66. In addition, the Asp/Glu rich region is followed by the MIDAS domain (metal ion dependent site). Due to the large size of MDN1 (approximately 5600 amino acids) and its limited homology with well-studied proteins, little is known about its structure and function in humans. We identified HLA-A, H2B, and LRCH4 as MDN1 binding partners and revealed their orientation as protein complexes in PyMol (Fig. 6A,B). These three proteins interact with the AAA domain, the dynein-like linker domain, and possibly the MIDAS MDN1 domain. In a previous report, affinity purification of bait proteins identified MDN1 as a protein associated with histone H2B67. In addition, a recent study also reported an interaction between MDN and HLA-B in HCT116 cells using affinity-purified mass spectrometry, supporting our findings68. The identification of this complex in IFNα-treated samples suggests a role for MDN1 in interferon signaling.
Because HLA genes are highly polymorphic, we extracted sequencing reads mapping HLA-A, -B, and -C from RNA sequencing data of Flo-1 cells (data not shown). Peptide sequences consistent with the sequencing reading revealed significant differences between HLA-A, -B, and -C in regions where cross-linked peptides were located in HLA-A (Figure S3). In addition, we did not observe protein-to-protein cross-linking of HLA-B/C molecules with H2B/HMGA1/MDN1/LRCH4 proteins. This suggests that the protein interaction found between HLA-A, MDN1, LRCH1 and HMGA1 is HLA-A specific. In addition, proteomic analysis of non-crosslinked samples (Table S4) showed that HLA-A has higher sequence coverage compared to HLA-B or HLA-C. The peptides identified for HLA-A were high in intensity in both IFNα-treated and untreated samples.
To ensure that the interactions identified here were not due to non-specific cross-linking of two proteins in close spatial proximity, we further confirmed two new HLA-A interacting factors by performing co-immunoprecipitation assays. HLA-A interactions with endogenous MDN1 and H2B were detected in both IFNα-treated and untreated Flo-1 cells (Figure 7, Figure S4). We confirmed that HLA-A was captured by H2B in the immunoprecipitates and that this association was due to IFNα treatment since HLA-A was absent in the immunoprecipitate samples from untreated cells (Figure 7A). However, our data suggest that IFNα differentially regulates HLA-A binding to H2B and MDN1. IFNα induces association between H2B and HLA-A, but reduces its association with MDN1. We found that MDN1 was associated with HLA-A in controls, and the addition of IFNα reduced this interaction independent of MDN1 induction by IFNα (Figure 7B,C). In addition, HLA-A immunoprecipitation captured H2B in A549 cells (Fig. S4), suggesting that this interaction is independent of cell type. Taken together, these results support interferon-mediated interactions of HLA-A with H2B and MDN1.
HLA-A co-purifies H2B and MDN1. Representative endogenous H2B (A) and MDN1 (B) immunoblots were immunoprecipitated from IFNα-treated Flo-1 cells and probed for the indicated antibodies. Mouse and rabbit IgG were used as a negative control. (C) Relative amounts (input) of different antigens are depicted by immunoblots probed against indicated antibodies, β-actin was used as a loading control.
The structural properties of one of the interferon-induced highly reliable cross-linked networks, H2B-HLA-A-HMGA1, were investigated. We used molecular dynamics modeling as an alternative approach to understand the conformational dynamics of the proteins involved in this complex (Figure 8). Inferences from the CLMS data suggest the possibility of different conformations of the H2B, HLA-A, and HMGA1 proteins. Therefore, the following potential complexes were modeled in a solvent medium: H2B-HLA-A, HMGA1-HLA-A, and H2B-HLA-A-HMGA1. An initial protein-protein docking screen using the MOE (Molecular Operating Environment; Chemical Computing Group Inc., Montreal, Quebec, Canada) package suggested possible conformations that differ between these proteins (Fig. 8A). Visualization of the docking protein complex revealed several interactions and possible conformations (Figure 5A, 8). Thus, one possible conformation is shown in Figure 8A (with labeled cross-links) and it was further evaluated using the MD modeling pipeline. In addition, binding of H2B or HMGA1 to HLA-A highlights the higher affinity of H2B for HLA-A (Fig. 8A).
Conformational dynamics of possible networks between the H2B (H2BFS)-HLA-A, HMGA1-HLA-A, and H2B-HLA-A-HMGA1 complexes. (A) Left panel is a 2D map (generated in SIM-XL software) of intramolecular (red) and intermolecular (blue) crosslinks (crosslink cutoff set to 3.5). In addition, cross-linking residues identified are labeled on the structures of the H2B, HLA-A, and HMGA1 proteins. The associated conformations of these proteins were extracted using the docking pipeline implemented in the MOE package. The lower left panel shows the various possible conformations of the H2B-HLA-A and HMGA1-HLA-A complexes with different protein-protein binding affinities (GBVI/WSA dG; kcal/mol). (B) Standard deviation (RMSD) of atomic positions (excluding hydrogen atoms) for each protein structure. (C) Intermolecular protein-protein hydrogen bond interactions from various simulated complexes considering specific interactions of duration ≥ 10 ns. The h-bond donor-acceptor cutoff distance was set to 3.5 Å, and the donor-H-acceptor cutoff angle was set to ≥ 160°–180°. (D) Labeled residues forming HLA-A protein-protein interactions with their respective partners, spanning ≥ 20 ns, extracted from dummy HLA-A-H2B and HLA-A-HMGA1 complexes. Protein structures represent an average structure of 100 ns MDS. (E) Interactions between HLA-A-H2B and HLA-A-HMGA1 complexes compared to interactions tracked by H2B-HLA simulation over 100 ns based on the K or S interaction site between the two peptides. Complexes /HMGA1-HLA-A/H2B-HLA-A-HMGA1. The threshold value for the evaluation of cross-links was set to 3.0, and specific interactions from MDS taking ≥ 10 ns were taken into account. Protein structures were visualized using BIOVIA Discovery Studio (Dassault Systèmes, BIOVIA Corp., San Diego, CA, USA) and Molecular Operating Environment (MOE; Chemical Computing Group Inc., Montreal, Quebec, Canada) packages.
The stability of HLA-A molecules over time (standard deviation; RMSD or standard deviation; RMSF) indicated that the presence of H2B or HMGA1 proteins in the complexes stabilized HLA-A (Figure 8B, Figure S5). The HMGA1 protein binds tightly to the B2M site of HLA-A, inducing the stability of the HLA-A amino acids in the HLA-A-HMGA1 or H2B-HLA-A-HMGA1 complex (Figure 8B, Figure S5). in particular, HLA residues ~60-90 and ~180-210 were found to be less flexible in the presence of H2B (FIG. 8B). H2B and HMGA1 showed better binding to HLA-A in the H2B-HLA-A-HMGA1 complex compared to HLA-A binding to H2B or HMGA1 alone (Figure 8C,D; Table S5). Residues involved in hydrogen bonding (MD modeled high occupancy ≥ 10 ns) coincide with CLMS interaction sites (K or S residues) in the complex, suggesting that the interactions identified by CLMS are very reliable. Reliability (Fig. 8E ). In CLMS and MD modeling, HLA-A residues between about 190-210 and about 200-220 amino acids were found to bind H2B and HMGA1, respectively (FIG. 8E).
Protein-protein interactions form dynamic structural networks that mediate intracellular communication in response to certain stimuli. Because many proteomics approaches detect changes in the overall steady state level of a protein, protein-protein interaction dynamics require additional tools to capture binding interfaces, and CLMS is one such tool. The interferon signaling system is a cytokine network that allows cells to respond to a range of environmental pathogenic and intrinsic pathological signals, culminating in the induction of subsets of interferon-inducible proteins. We applied CLMS to determine if novel protein-protein interactions could be identified among a panel of interferon-induced proteins. Global protein cross-linking analysis in an interferon-responsive Flo-1 cell model was used to capture protein complexes. Extraction of tryptic peptides from non-cross-linked and cross-linked cells allows for peptide counting, pathway enrichment, and peptide length distribution with defined LFQ intensity. Canonical interferon-inducible proteins were identified as a positive internal control, while new intermolecular and intramolecular cross-linked adducts of canonical interferon-inducible proteins such as MX1, UP18, OAS3 and STAT1 were observed. Various structural features and interactions in functional areas have been investigated.
An interaction between HLA-A, MDN1 and H2B was detected by immunoblotting in Flo-1 and A549 cells treated and untreated with IFNα. Our results highlight that HLA-A complexes with H2B in an IFNα-dependent manner. Our work represents an interesting avenue for further exploration of the co-localization of these two complexes. It would also be interesting to expand the CLMS approach to a panel of cell lines to identify cell-type-independent interferon-mediated protein interactions. Finally, we used MD modeling as an alternative approach to understand the conformational dynamics of proteins involved in the H2BFS-HLA-A-HMGA1 complex, which tracked intramolecular and intermolecular cross-talks. Inferences from the CLMS data suggest the possibility of different conformations of the H2BFS, HLA-A, and HMGA1 proteins. The possible different conformations between these docking protein complexes revealed several interactions similar to those observed in the CLMS dataset. One of the main strengths of our method is that it allows easy identification of interacting highly polymorphic genes such as HLA, so it will be interesting to study interactions of HLA haplotype-specific proteins that are otherwise difficult to study. Taken together, our data demonstrate that CLMS can be used to expand our understanding of interferon-induced signaling networks and provide a basis for studying more complex intercellular systems in the tumor microenvironment.
Flo-1 cells were obtained from ATCC and maintained in DMEM (Gibco) supplemented with 1% penicillin/streptomycin (Invitrogen), 10% fetal bovine serum (Gibco) and stored at 37°C and 5% CO2. Incubation. Cells were grown to 70-80% confluence before being treated with IFNα14 (manufactured by Edinburgh Protein Production Facility). All other chemicals and reagents were purchased from Sigma Aldrich unless otherwise noted.
Flo-1 cells were cultured in 6-well plates and the next day the cells were treated with 10 ng/ml IFNα14 for 24 hours to approximately 80% confluence. Cells were washed three times with PBS and ligated with freshly prepared DSS (Thermo Fisher Scientific) (dissolved in DMSO) in PBS for 5 min at 37° C. to a final concentration of 0.5 mM. The DSS crosslinking reaction was replaced with PBS and residual DSS was quenched by adding 20 mM Tris (pH 8.0) in PBS for 15 min at 37°C. Cells were collected by scraping and collected in low binding tubes (Axygen).
The cell pellet was lysed with 300 µl of urea lysis buffer (8 M urea, 0.1 M Tris, pH 8.5) for 30 min at room temperature with occasional shaking. All centrifugation steps were performed at 14,000 xg at 8°C. Centrifuge the lysate for 10 minutes and transfer the supernatant to a new tube. The remaining clear particles were dissolved in 150 μl of the second lysis buffer (2 M urea, 2% (w/v) SDS (sodium dodecyl sulfate)) for 30 minutes or more until a homogeneous aqueous solution was obtained. The lysate was centrifuged for 20 minutes and the supernatant was mixed with the lysate obtained in the previous step. Protein concentrations were assessed using the Micro BCA assay (Thermo Fisher Scientific) according to the manufacturer’s instructions for microplate procedures. The samples were quickly frozen in liquid nitrogen and stored at -80°C.
Approximately 100 μg of soluble cross-linked protein was processed using a modified filtration sample preparation protocol (FASP) as described by Wisniewski et al. 69 Briefly, the protein is crosslinked with 200 µl of urea buffer (8 M urea in 0.1 M Tris, pH 8.5), vortexed and halved. All centrifugation steps were performed at 14,000 xg at 25°C. The first half of the cross-linked protein lysate was transferred to a 10 kDa Microcon centrifugal filter device equipped with an Ultracel-10 membrane (Merck), followed by centrifugation on the filter for 25 minutes. Then add the second half of the protein to the filter and repeat the same steps. Protein recovery was performed by adding 100 μl of 17 mM tris(2-carboxyethyl)phosphine hydrochloride (TCEP) in urea buffer. Recovery was stirred on a thermomixer at 600 rpm for 30 min at 37°C. In addition, the column was centrifuged and the reduced cross-linked protein was alkylated using 100 μl of 50 mM iodoacetamide in urea buffer. The alkylation reaction was carried out at room temperature for 20 minutes in the dark. Rotate the column, wash the column walls 3 times with 100 µl urea buffer, and then centrifuge. The same operation was performed 3 times using 100 μl of 100 mM ammonium bicarbonate. Before trypsinization, replace the collection tube with a new one. Add digestion buffer containing 50 mM ammonium bicarbonate and 1 µl trypsin diluted in trypsin buffer (Promega). The ratio of trypsin to protein was maintained at about 1:33, and digestion reactions were incubated overnight at 37° C. in a humid chamber. The crosslinked peptide was eluted from the filter by centrifugation for 25 minutes. Peptide recovery was improved by adding 50 μl of 0.5 M NaCl to the filter, followed by centrifugation for 25 minutes.
C18 Micro Spin columns (Harvard Apparatus) were used to desalt cross-linked tryptic peptides following the protocol described by Bouchal et al.70 with minor modifications. Briefly, C18 spin columns were activated with three washes of 0.1% formic acid (FA) in acetonitrile (AcN) (Merck) and two washes of 0.1% FA. The column was hydrated with 0.1% FA for 15 minutes. Load samples into spin columns and wash 3 times with 0.1% FA. The desalted peptides were sequentially eluted with a stepwise gradient using 50%, 80% and 100% AcN in 0.1% FA. The samples were dried in a SpeedVac Plus concentrator (Eppendorf) until the residual liquid completely disappeared. The eluted peptides were dissolved in 100 μl of 0.08% trifluoroacetic acid in 2.5% AcN and the concentrations were measured on a NanoDrop 2000 (Thermo Scientific). Approximately 1 μg of crosslinked peptide per sample was injected into the LC-MS/MS system.
Cross-linked peptides were separated on an UltiMate 3000 RSLCnano LC system (Thermo Scientific) connected to an Orbitrap Exploris 480 mass spectrometer (Thermo Scientific). Cross-linked peptides were collected on a 300 µm ID, 5 mm long µ-pre-column C18 capture column packed with C18 PepMap100 sorbent and 5 µm PepMap sorbent (Thermo Scientific). Load the pump flow set at 5 µl/min 0.08% trifluoroacetic acid dissolved in 2.5% AcN. Cross-linked peptides were separated on an analytical fused silica column with an inner diameter of 75 μm and a length of 150 mm, filled with a 2 μm PepMap sorbent (Thermo Scientific). Mobile phases A and B consisted of 0.1% FA in water and 0.1% FA in acetonitrile, respectively. The gradient starts at 2.5% B and increases linearly to 40% B over 90 minutes, then to 90% B over the next 2 minutes. The mobile phase composition was maintained at 90% B for 10 minutes and then decreased linearly to 2.5% B over 2 minutes. The column was equilibrated at 2.5% B for 8 minutes before the next cycle. Cross-linked peptides eluted from the analytical column were ionized in a nanoelectrospray ionization (NSI) source and injected into an Exploris 480 mass spectrometer (Thermo Scientific).
The Orbitrap Exploris 480 mass spectrometer operated in the positive data correlation mode. A full scan was performed in section mode at a resolution of 120,000 with range settings from m/z 350 Th to m/z 2000 Th. The normalized AGC target was set at 300% with a maximum input time of 50ms. Monoisotopic peak detection has been established for peptides. The constraint relaxation parameter is set to true if too few precursors are found. The minimum ionic strength of the precursor was set to 5.0e3 and precursor charge states up to +8 were included in the experiments.
The cycle time between major scans in data correlation mode was set to 2.5 seconds. Dynamic mass exclusion was set to 20 s after the first fragmentation of the precursor ion. The precursor isolation window was set to 2 Th. The type of normalized collision energy with a fixed collision energy mode was chosen in a data dependent MS/MS scan. Collision energy set to 30%. The Orbitrap resolution was set to 15,000 and the AGC target to 100%. The custom maximum injection time is set to 60 milliseconds.
Before tracking the protein-protein network in cross-linked samples, we processed the raw files using the MaxQuant package (version 1.6.12.0)26,27 to identify traceable peptides/proteins in the samples. In addition, similar proteomic analyzes were performed on uncrosslinked Flo-1 samples treated and untreated with IFNα. MS/MS data were searched in the UniProt human database (www.uniprot.org) (uploaded August 12, 2020, contains 75,093 entries) using the built-in search engine Andromeda27. The search was carried out without indicating the specificity of the enzyme and various modifications of deamidation (N, Q) and oxidation (M). Precursor mass tolerances were set at 20 ppm and product ions at 0.02 Da. The initial and maximum mass deviation was set to 10 ppm. The maximum mass of the peptide was set at 4600 Da and the sequence similarity was set between 7 and 25 amino acids (a.a.). Further statistical analysis was performed using the Perseus program (version 1.6.10.45). Protein content was calculated by normalizing the spectral intensity of the protein (LFQ intensity; unlabeled quantification)27 and the intensity values ​​were converted to Log2. A hierarchical clustering of proteins identified by their peptide intensity was built using the pheatmap (v1.0.12) package in R (v 4.1.2). Pathway enrichment analysis was performed using the Reactome pathway database for IFNα-treated proteins that were more than four times activated compared to untreated samples.
Identification of lysine (K) or serine (S) specific chemical crosslinks of protein complexes monitored by LC-MS/MS was performed using a spectroscopic identification machine (SIM-XL) for cross-linked peptides (SIM-XL)29. First, possible interactions between interferon-associated (IFN) DNA damage resistance signature (IRDS) genes were investigated using the IRDS protein dataset described in Padariya et al.28. Screening all conditions and repeats of the entire human UniProt is computationally intensive, so the entire human UniProt database (www.uniprot.org) (downloaded 12 August 2020, contains 75,093 entries) against IFNα-treated repeats. One of the filters for high trust interactions. These high-significance interactions obtained were expanded and tested in all repetitions and conditions.
In SIM-XL, DSS was used for the crosslinker (XL) and the XL weight shift and modification weight shift were set to 138.06 and 156.07, respectively. The following crosslinking reaction sites are considered: KK, KS and KN-TERM, without reporter ions. Both precursor and fragment ppm were set to 20 and the Xrea threshold was set to 0.15. Trypsin was considered to be completely specific, and a high-energy C-trap (HCD) fragmentation method was implemented. The XCorr dynamic DB reduction threshold and the minimum number of peptides for dynamic DB reduction were set to 2.5 and 2, respectively. Other parameters are: monoisotope probability and peak coincidence cutoff, minimum 4 AA residues per strand and maximum strand charge, and 3 maxima of missed splits. The resulting stitched 2D maps were analyzed in (SIM-XL) and the xQuest28 graphical representation was used to build the 2D maps. Protein crosslinks on protein structures are provided in PyMol (PyMOL Molecular Graphics System, version 2.0 Schrödinger, LLC).
Protein model structures were created using the Phyre2 server (http://www.sbg.bio.ic.ac.uk/phyre2)11 using the principles of homology modeling and implementation of the “Hidden Markov Method”. Phyre2 generates model structures based on sequence alignment with known protein structures. For H2BFS, HLA-A, HMGA1, LRCH4, and MDN1 proteins, template structures 1kx552, 1kj349, 2eze55, 6hlu62, and 6i2665 were used. In addition, the structure of AlphaFold71 MX1, UBP18 and ROBO1 was also considered. The protein structure was visualized using the BIOVIA Discovery Studio Visualizer package (Dassault Systèmes, BIOVIA, San Diego, CA, USA) and the Molecular Operating Environment package (MOE; Chemical Computing Group Inc., Montreal, Quebec, Canada).