key: cord-0721969-r9e2l1r2
authors: Osipiuk, Jerzy; Wydorski, Pawel M.; Lanham, Benjamin T.; Tesar, Christine; Endres, Michael; Engle, Elizabeth; Jedrzejczak, Robert; Mullapudi, Vishruth; Michalska, Karolina; Fidelis, Krzysztof; Fushman, David; Joachimiak, Andrzej; Joachimiak, Lukasz A.
title: Dual domain recognition determines SARS-CoV-2 PLpro selectivity for human ISG15 and K48-linked di-ubiquitin
date: 2022-04-25
journal: bioRxiv
DOI: 10.1101/2021.09.15.460543
sha: b0023326d5bcfd7d96957656bdabdefa2f7b662d
doc_id: 721969
cord_uid: r9e2l1r2

The Papain-like protease (PLpro) is a domain of a multi-functional, non-structural protein 3 of coronaviruses. PLpro cleaves viral polyproteins and posttranslational conjugates with poly-ubiquitin and protective ISG15, composed of two ubiquitin-like (UBL) domains. Across coronaviruses, PLpro showed divergent selectivity for recognition and cleavage of posttranslational conjugates despite sequence conservation. We show that SARS-CoV-2 PLpro binds human ISG15 and K48-linked di-ubiquitin (K48-Ub2) with nanomolar affinity and detect alternate weaker-binding modes. Crystal structures of untethered PLpro complexes with ISG15 and K48-Ub2 combined with solution NMR and cross-linking mass spectrometry revealed how the two domains of ISG15 or K48-Ub2 are differently utilized in interactions with PLpro. Analysis of protein interface energetics uncovered differential binding stabilities of the two UBL/Ub domains. We emphasize how substrate recognition can be tuned to cleave specifically ISG15 or K48-Ub2 modifications while retaining capacity to cleave mono-Ub conjugates. These results highlight alternative druggable surfaces that would inhibit PLpro function.

The COVID-19 pandemic is caused by Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) belonging to the Coronaviridae family. SARS-CoV-2 is a spherical, enveloped, nonsegmented, (+) sense RNA virion with a ~30 kbs genome 1,2 . The genome is used for translation of the replication-transcription complex (RTC) comprised of two polyproteins (Pp1a/Pp1ab), which are processed by two viral proteases: papain-like protease (PLpro) and 3C-like protease. PLpro is a domain of non-structural protein 3 (Nsp3). It cleaves three sites in SARS-CoV-2 polyproteins yielding Nsp1, Nsp2 and Nsp3. The "LXGG↓XX" motif found in the polyproteins is essential for protease recognition and cleavage. PLpro has been shown to have additional functions 3 , including deubiquitinating 4-6 and deISGylating activities 7, 8 , occurring by the cleavage of the "LRGG" sequence motif found at the C-terminus of human ubiquitin (Ub) and ISG15. PLpro from SARS-CoV-1 (PLpro CoV-1 ) preferentially removes K48-polyUb modification 9 . Removing these post-translational modifications (PTM) interferes with interferon (IFN) expression and blocking NF-kappaB signaling 10 . By cleaving off ISG15 from STAT3 it induces up-regulation of TGF-β1 11 . Some PLpro functions involve direct cleavage of host proteins influencing broad processes from blood coagulation to nuclear transport [12] [13] [14] . PLpro may also play roles beyond its proteolytic activity 6 , illustrating its diverse and complex functions 3 . The SARS-CoV-2 PLpro (PLpro CoV-2 ) sequence and structural fold are conserved among SARS-CoV-1 (83% identical), MERS-CoV (30% identical) and other coronaviruses. Despite low sequence identities (~10%) 7 , PLpros also share common structural architecture and catalytic site with the human ubiquitin specific proteases (USPs), one of the five distinct deubiquitinating enzyme (DUB) families.

PLpro from SARS-and MERS-CoVs have deubiquitinating 4-6 and deISGylating activities 7,8 , but they show somewhat different substrate selectivity, with PLpro CoV-1 preferentially removing K48-polyUb modification 9 and MERS-CoV PLpro being the least specific. It is not fully understood how this activity influences interaction of the virus with the host immune system. Ubiquitination is an essential PTM engaged in multiple functions in humans, including signaling proteasome-dependent protein degradation. The modification is mediated by the Ub-conjugating system and could be reversed by DUBs 8 . ISG15 is an IFNa-stimulated gene that is a critical component of the antiviral response 15, 16 . It is not precisely clear how ISG15 interferes with viral processes but it is believed that tagging newly translated viral proteins with ISG15 sterically prevents their folding, assembly or interactions 17 . Removal of Ub and ISG15 conjugates from specific substrates in host cells may also have a diverse impact on numerous cellular processes and specifically may frustrate the host response to viral infection 15, 16, 18 . ISG15 is processed and subsequently activated in a manner similar to Ub using interferon-induced factors that follow the ubiquitination-like E1, E2 and E3 enzyme cascade to mediate co-translational ISGylation -an addition of ISG15, via its C-terminal LRGG motif, to substrate lysine residues 19 . PLpro can recognize and cleave off both appendages as they share PLpro recognition motif at their C-termini and have common other structural features: ISG15 comprises two Ub-like (UBL) domains and mimics a head-to-tail linked Ub 2 . While K48-linked Ub 2 (K48-Ub 2 ) and ISG15 are homologous both in sequence and fold, the topologies of how the two domains are linked are distinct. It remains unknown how PLpro discriminates between the Ub 2 and ISG15 substrates.

Recently published work suggested that mutations in PLpro CoV-1 to PLpro CoV-2 changed its binding preference from K48-linked Ub 2 to human ISG15 (hISG15) 17, 20 . Inspired and attracted by these results we investigated interaction of PLpro CoV-2 with mono-, di-, and tri-ubiquitins (Ub 1 , Ub 2 , Ub 3 ) and hISG15. We employed complementary biochemical, structural x-ray, NMR and computational approaches, to understand how PLpro CoV-2 can differentiate between Ub 1 , K48-Ub 2 and hISG15 substrates. We find that PLpro CoV-2 binds both hISG15 and K48-Ub 2 with high and similar affinity but shows weaker interactions with Ub 1 . We also find lower affinity alternate binding modes for hISG15, K48-Ub 2 and Ub 1 which can be explained by non-stoichiometric binding model and sequence preference outside the conserved recognition motif. To reveal details of substrate recognition, we determined structures of non-covalent complexes of PLpro CoV-2 (C111S proteolytically inactive mutant) with hISG15 and K48-Ub 2 . These are the first structures of non-modified, complete complexes and they uncover that hISG15 binding is determined by recognition of both UBL domains while K48-Ub 2 is recognized mainly through the proximal Ub. We further examined this PLpro binding to K48-Ub 2 , Ub 1 and hISG15 using NMR experiments. These data, together with crosslinking mass spectrometry (XL-MS) suggest that PLpro CoV-2 interacts with both UBL domains of ISG15 whereas K48-Ub 2 is recognized largely through the proximal Ub, with distal Ub contributing less to binding through different interactions. We used modelling to predict alternative modes of PLpro binding to substrates that are consistent with cross-linking data. Finally, we tested our binding models by performing an in silico DDG alanine scan on PLpro CoV-2 in complex with K48-Ub 2 /hISG15 substrates and show differential domain utilization by the PLpro for the two substrates. Our findings uncover binding heterogeneity in PLpro interactions with hISG15 and ubiquitin substrates that decouples binding affinity from proteolytic activity.

Sequence and topological differences between hISG15 and K48-Ub 2

Recent biochemical binding and cleavage assays have shown that PLpro CoV-1 prefers K48-Ub 2 while the related PLpro CoV-2 binds more tightly to both human and mouse ISG15 (hISG15 and mISG15). A nearly 20-fold stronger affinity compared to Ub 2 17,20 suggests that the sequence variation at the substrate binding interface between PLpro CoV-1 and PLpro CoV-2 may dictate substrate specificity 17, 20 . Importantly, these previous studies on PLpro CoV-1 /PLpro CoV-2 binding to Ub 2 used a nonhydrolysable synthetic triazole linker between the Ubs rather than a native isopeptide K48 linkage, raising questions how linker geometry and rigidity may influence binding to PLpro 17, 20, 21 . When considering both domains in K48-Ub 2 and hISG15, they are 33% identical in sequence ( Supplementary Fig. 1a ) while the distal (N-terminal) UBL domain of hISG15 is 29% identical to Ub, and the proximal (C-terminal) UBL domain of hISG15 has a slightly higher sequence identity of 37%. hISG15 and mISG15 are 63% sequence identical, and both have similar sequence identities to Ub ( Supplementary Fig. 1a,b) . Intriguingly, however, a recent study reported that hISG15 binding to PLpro CoV-2 has an order of magnitude higher on-and off-rates than mISG15 binding 17 . Importantly, Ub and UBLs of hISG15 and mISG15 also vary in the active binding surfaces ( Supplementary Fig.   1b ,c). The protein domains in hISG15 and K48-Ub 2 have homologous folds but their sequences and topologies of how the two domains are linked are different ( Fig. 1a and Supplementary Fig. 1d ).

The earlier reported structure of the PLpro CoV-1 :K48-Ub2 complex shows the proximal Ub bound to the Zn finger and palm domains via its surface hydrophobic patch 21 (comprising residues L8, I44 and V70) placing the C-terminal tail modified with allylamine in a groove that is covalently linked to active site C111 ( Supplementary Fig. 1d) 22 . A recent structure of full-length mISG15 bound to PLpro CoV-2 20 revealed a distinct binding mode of the proximal and distal UBL domains of mISG15.

The proximal UBL is shifted away from the finger domain compared to the proximal-Ub binding mode while still placing the C-terminal LRGG tail into the active site of the protease ( Supplementary   Fig. 1d ). Comparison of mISG15 and Ub active surfaces reveals that the hydrophobic patch centered on I44 in Ub ( Supplementary Fig. 1b,c) is more polar in hISG15. Given that the prior studies used a triazole-linked Ub dimer we wanted to test the influence of the linker composition on binding to PLpro CoV-2 . We used microscale thermophoresis (MST) binding experiments to quantify affinity between PLpro CoV-2 and three substrates: hISG15, K48-Ub 2 and Ub 1 (Fig. 1a ). Additionally, we tested PLpro CoV-2 binding to K48-Ub 2 and Ub 1 containing a C-terminal aspartic acid (D77) after the "LRGG" PLpro recognition site (Fig. 1a) , which is typically used for controlled enzymatic synthesis of ubiquitin chains 23 . Fitting our binding data to a 1:1 binding model ( Supplementary Fig. 2a ) resulted in abnormally high c 2 and systematic deviation in the residuals ( Supplementary Fig. 2b ). Improved fits were observed using a model that assumes two binding events with different K d s ( Supplementary   Fig. 2a ) yielding statistically significant reductions in the c 2 compared to one binding event for all datasets except Ub 1 -D77 ( Supplementary Fig. 2b ). We find that PLpro CoV-2 binds both hISG15 and K48-Ub 2 with high affinity (90 ± 30 nM, 80 ± 30 nM, respectively) and more strongly than Ub 1 (apparent K d , 160 ± 70 nM) (Fig. 1b) although the actual microscopic K d of Ub 1 could be even higher as it may be able to bind to multiple sites on PLpro CoV-2 (see below). Interestingly, K48-Ub 2 with a Cterminal aspartic acid (K48-Ub 2 -D77) binds tenfold weaker (790 ± 540 nM) compared to Ub 2 ( Fig.   1b and Supplementary Fig. 2c) , and Ub 1 -D77 exhibited weak binding (146.67 ± 90.74 µM) ( Fig. 1b and Supplementary Fig. 2c ). This is consistent with a lack of reported protease substrates with acidic residues at the C-terminus of the "LRGG(X)" motif 7 . Our analysis also suggests the presence of secondary binding events for hISG15 and K48-Ub 2 with µM affinities (Fig. 1b , gray insert). This is supported by cross-linking data where we observed covalent adducts with molecular weight corresponding to heterodimers (PLpro:substrate) but also heterotrimers ((PLpro) 2 :substrate) ( Fig. 1c and Supplementary Fig. 2d ,e). The observed heterogeneity of the species formed suggests that ISG15 may bind in a more defined orientation to PLpro while Ub 2 appears to binds in several arrangements.

These experiments indicate that there is a dominant binding mode between the substrate and PLpro but higher order complexes are also possible, thus justifying the need to fit our binding data with more complex binding models.

Derived from published structures of mISG15:PLpro CoV-2 and Ub 2 :PLpro CoV-1 17,20,21 , we anticipated that hISG15 and Ub 2 bind PLpro utilizing both UBL/Ub domains. However, the difference in Ub 2 and Ub 1 affinities suggests that the second Ub contributes modestly to the binding. Additionally, Ub 1 -D77 binds nearly 1000-fold more weakly compared to WT Ub 1 while Ub 2 -D77 yields an affinity more similar to Ub 1 , perhaps indicative of a change in binding mode primarily utilizing a single Ub.

To explore how affinity relates to PLpro proteolytic activity, we conducted PLpro CoV-2 cleavage assays with hISG15 with modified C-terminal tails mimicking natural SARSC-CoV-2 substrates or K48linked Ub 3 /Ub 2 . We found that PLpro CoV-2 can efficiently cleave peptides containing the "LRGG motif", but the rate of cleavage varies. The hISG15-Nsp2 fusion is cut faster than fusions with Nsp3 and Nsp4 peptides ( Fig. 1d and Supplementary Fig. 2f ). We investigated how amino acid X (position 158) at the C-terminus of "LRGG(X)" motif can impact cleavage. When Ala of Nsp2 peptide is substituted with Glu or Asp, the fusion peptide is being cut the slowest ( Fig. 1e and Supplementary   Fig. 2g ). We also found that PLpro CoV-2 hydrolyzes K48-Ub 3 to Ub 2 and Ub 1 rapidly, but the subsequent cleavage of Ub 2 to Ub monomers is slow (Fig. 1f,g) . Finally, we found that Ub 2 -D77 is cleaved more rapidly compared to Ub 2 , which suggests that despite Ub 2 -D77 binding with a notably lower affinity to PLpro CoV-2 it must be bound differently than Ub 2 to enable the hydrolysis ( Supplementary Fig. 2h ). If Ub 3 binds predominantly with two Ub units, there are two possible binding modes of Ub 3 on PLpro. To test which binding mode is dominant we used a K48-Ub 3 that contains three "distinct" Ub units: 1 (mutant Ub-K48R), 2 (U-15 N-labeled Ub) and 3 (Ub-D77) (Fig.   1g ), allowing mass spectrometry-based identification of each cleavage product. Analysis by mass spectrometry of a cleavage time course of this Ub 3 construct reveals that the first cleavage occurs between Ub units 2 and 3, releasing the C-terminal Ub 1 -D77 with only minor products for the other Ubs (Fig. 1g ). This is consistent with units 1 and 2 bound to PLpro with the C-terminal tail of unit 2 fitting the active site for rapid hydrolysis (Fig. 1g) . Three Ub binding sites (S2, S1, and S1') were proposed for PLpro 20 . In this model, unit 1 would bind to S2, unit 2 to S1 and unit 3 to S1' (Supplementary Fig. 1d ). Interestingly, Ub 2 can bind to PLpro in two modes. In the high-affinity mode, as in Ub 3 binding, where Ub unit 1 (distal) binds to S2 and unit 2 to S1, results in no cleavage.

Only the second, lower-affinity mode is productive, wherein Ub unit 1 is bound to S1 and unit 2 (proximal Ub) located at S1'. In this arrangement the C-terminal tail of unit 1 connecting the two Ubs is placed in the active site of PLpro, and Ub 2 is cut into monomers. These two modes of binding are competitive and because there is a difference in affinity, the rate of cleavage is reduced, but eventually resulting in complete disassembly of Ub 2 . Thus, our experiments uncover more complex alternate binding modes and sequence dependence of PLpro for two related substrates that have not been described to date. Our data also provide alternative interpretation of recently published work 17,20 as the composition and flexibility of the Ub-Ub linker can significantly impact the binding or cleavage or both for Ub 2 and other protein substrates. This may explain the previously observed difference in affinity of PLpro CoV-2 between Ub 2 and ISG15 which is likely attributed to changes in mode of binding and/or conformational flexibility of the substrate linkage rather than mutations in the PLpro enzyme.

We further investigated details of interaction between the PLpro CoV-2 , hISG15 and K48-Ub 2 to reveal similarities and differences. We determined crystal structures of PLpro CoV-2 with an active site C111S mutation, which inactivates PLpro, in complex with hISG15 at 2.98 Å resolution (Fig. 2a) and with K48-Ub 2 at 1.88 Å resolution (Fig. 2b) . For the PLpro CoV-2 :hISG15 complex we observe well resolved electron density for the proximal and distal UBL domains bound to S1 and S2 sites respectively (Fig. 2a) . By contrast, for the PLpro CoV-2 :K48-Ub 2 structures determined at much higher resolution we observe strong electron density for the proximal Ub bound to S1 site with only weak signal for the distal Ub ( Fig. 2b and Supplementary Fig. 3b ). Despite low electron density for the distal Ub, some regions of densities resemble a-helix and amino acid chains corresponding to portions of the distal Ub from superposed structure of PLpro CoV-1 bound to a K48-Ub 2 (PDB id: 5E6J) upon only slight adjustment ( Supplementary Fig. 3c ). Clearly, there is sufficient room in our crystals to fit the distal Ub at site S2 in a binding mode alike PLpro CoV-1 bound to a K48-Ub 2 ( Supplementary   Fig. 3d ) 21 . An alternative explanation is that PLpro CoV-2 (C111S) can hydrolyze the K48 linkage slowly, leading to a mixture of Ub 1 and Ub 2 . To test this directly, we ran SDS-PAGE gels of our crystals and observed predominantly Ub 2 with only a minor Ub 1 species ( Supplementary Fig. 3e ).

Our structures revealed how the protease differentially recognizes hISG15 using both UBL domains while K48-Ub 2 is predominantly recognized using the proximal Ub. We additionally determined structures of the free hISG15 and K48-Ub 2 to 2.15 Å and 1.25 Å resolution, respectively. The bound and free hISG15 conformations are similar with a root-mean-square deviation (rmsd) for Ca atoms We also compared our structure to the previously published PLpro CoV-2 :mISG15 17 complex.

Overlay of the two structures (PDB ids: 7RBS and 6YVA) reveals good structural similarity with the overall rmsd of 0.70 Å for PLpro and 1.40 Å for hISG15 and mISG15 ( Supplementary Fig. 5 ). The proximal UBL domains of both ISG15s are well aligned and make several conserved interactions with the PLpro but interaction with the distal UBL domain shows important differences (see below)

with the largest conformational deviation in the distal UBL domain ( Supplementary Fig. 5a ). We compared the contacts from the distal domain of mISG15 and hISG15 to previously determined hotspot residues (F69 and V66) on PLpro CoV-2 17 . We find that in the PLpro CoV-2 :mISG15 structure K30 and M23 of the distal UBL of mISG15 interact with F69 of PLpro CoV-2 , while V66 of PLpro CoV-2 interacts with A2 of the substrate (Supplementary Fig. 5b ). By contrast, in our new PLpro CoV-2 :hISG15 structure residue 30 of hISG15 is an alanine, thus leaving M23 alone to stabilize the interaction with F69 of PLpro CoV-2 , while the N-terminus of hISG15 interacts with V66 of PLpro CoV-2 ( Supplementary   Fig. 5c ). Residue 20 in ISG15 makes similar nonpolar contacts with V66 but it varies between the mouse (T20) and human (S20) protein ( Supplementary Fig. 5b, c) . We additionally compared the interactions between the proximal UBL domains of the mISG15 and hISG15 where the UBL binds in a similar binding mode ( Supplementary Fig. 5d ). We find that overall, the two ISG15 proteins make similar, but not identical contacts determined by the sequence variation between mISG15 and hISG15. This suggests that the virus may have a different impact if infects distinctive species. The central interacting residues on PLpro CoV-2 are Y171, E167 and M208 which interact with conserved R153/R151, W123/W121 and P130/P128 on the proximal domains of hISG15 and mISG15, respectively ( Supplementary Fig. 5e,f) . The interaction is centered on a salt bridge between E167 of PLpro CoV-2 and R153 of hISG15 while the equivalent arginine (R151) in mISG15 is not oriented properly to form a salt bridge. Nonetheless, this core interaction is stabilized by nonpolar interactions of the surrounding residues from both sides of the interface including Y171 of PLpro CoV-2 and W123/W121 and P130/P128 from hISG15 and mISG15, respectively ( Supplementary Fig. 5e, f) . By contrast, interactions with R166 of PLpro CoV-2 vary more significantly between mISG15 and hISG15.

In the mISG15 structure the side chain of M208 is not resolved while in the hISG15 structure M208 packs against R166 ( Supplementary Fig. 5f ). Interestingly, R166 forms a salt bridge with E87 of mISG15 which is changed to asparagine (N89) in hISG15 ( Supplementary Fig. 5e , f). To compensate for this loss of interaction, N151 of hISG15 makes a hydrogen bond with R166 ( Supplementary Fig.   5e ,f). This highlights subtle sequence changes between mISG15 and hISG15 that allow interface rearrangements while preserving the binding mode. Our structures show that hISG15 binds PLpro CoV-2 utilizing both proximal and distal domains (Fig. 2a) , while binding of K48-Ub 2 is primarily driven by interaction with the proximal domain with only weak density observed for the distal domain ( Fig.   2b and Supplementary Fig. 3b ). We also compared our PLpro CoV-2 :hISG15 structure to a recent structure of PLpro CoV-2 bound to only the proximal domain of hISG15 20 (Supplementary Fig. 6a ; PDB id: 6XA9, 2.9 Å resolution). As in the PLpro CoV-2 :mISG15 complex, the structural similarity is high, with an overall Ca rmsd of 1.0 Å. A comparison of the interface contacts reveals nearly identical interactions, even preserving side-chain rotamers between the proximal hISG15 and PLpro in the two structures ( Supplementary Fig. 6b ). Finally, our new structure of PLpro CoV-2 :K48-Ub 2 is nearly identical in binding mode to the previously published structure of PLpro CoV-2 :Ub 1 ( Supplementary   Fig. 6c ; PDB id: 6XAA, 2.7 Å resolution) with a Ca rmsd of 0.32 Å and nearly identical side chain rotamers at the interface ( Supplementary Fig. 6d ). Interestingly, our structure was determined to higher resolution and without the introduction of a covalent linkage of Ub 2 to PLpro CoV-2 suggesting that the covalent linkage does not alter the physiological binding of the substrate. In contrast, however, introduction of a synthetic linker between the Ubs in Ub 2 does influence binding of the substrate to PLpro CoV-2 .

We then used NMR to further characterize PLpro CoV-2 binding to hISG15 and K48-Ub 2 and to examine if the contacts observed in crystals also occur in solution. The addition of unlabeled PLpro CoV-2 (C111S) caused substantial perturbations in the NMR spectra of 15 N-labeled hISG15 (Fig.   3a) . We observed disappearance of signals of free hISG15 and emergence of new ones; this indicates slow-exchange binding regime 26, 27 , consistent with the sub-μM K d values measured by MST. The strongly attenuated signals of hISG15 residues, including the C-terminal G157, are consistent with our crystal structure of PLpro CoV-2 :hISG15 complex (Fig. 3a, h) . A similar behavior was observed for K48-Ub 2 upon addition of PLpro CoV-2 (Fig. 3b-c) , where both the distal and proximal Ubs exhibited strong attenuation or disappearance of NMR signals and emergence of new signals, primarily for residues in and around the hydrophobic patch as well as the C-termini. The affected residues mapped to the binding interface in our PLpro CoV-2 :K48-Ub 2 crystal structure (Fig. 3i) , and the slow-exchange behavior is also consistent with the sub-μM K d values. The slow-exchange binding regime for both hISG15 and K48-Ub 2 is also generally consistent with the reported slow off-rates (0.2, 0.4 s -1 ) 17 .

PLpro CoV-2 also caused noticeable perturbations in the NMR spectra of Ub 1 , although these were weaker than in Ub 2 , and several residues showed gradual signal shifts indicative of fast-exchange, consistent with our MST measurements (Fig. 3d) .

We also performed reverse-titration NMR experiments where unlabeled ISG15, K48-Ub 2 , or Ub 1 was added to 15 N-labeled PLpro CoV-2 . Both ISG15 and K48-Ub 2 caused substantial perturbations in the 15 N-PLpro CoV-2 spectra (Fig. 3e-f ). Particularly noticeable was the change in the indole NH signals of W93 and W106 located in close proximity to the active site of PLpro, as well as of imidazole NH signal that might be attributed to the active-site H272 (Fig. 3e-f, Supplementary Fig. 7 ), in agreement with the C-termini of ISG15 and Ub 2 entering the active site of PLpro in our crystal structures ( Supplementary Fig. 7f ). The addition of Ub 1 caused significantly lesser overall 15 N-PLpro CoV-2 signal perturbations, although the We and Hd signal shifts were clearly visible when Ub 1 was in significant excess (Fig. 3g, Supplementary Fig. 7c,e) . Even at 8-molar excess of Ub 1 both free and bound We/Hd signals were present, consistent with weaker binding. Taken together, the NMR data suggest that the apparent strength of PLpro CoV-2 binding is: ISG15 ≈ K48-Ub 2 > Ub 1 , consistent with our MST data.

These NMR data indicate that binding to PLpro involves both UBLs of hISG15 and both Ubs of K48-Ub 2 . Interestingly, despite being identical and having very similar chemical shifts in the unbound state ( Supplementary Fig. 8a) , the distal and proximal Ubs show markedly different signal perturbations indicative of distinct contacts with PLpro (Fig. 3b,c, and Supplementary Fig. 8b) . While the perturbed residues in the proximal Ub agree well with our PLpro CoV-2 :K48-Ub 2 crystal structure, where this Ub occupies the S1 site, several perturbations observed in the distal Ub (most notably for Fig. 8b ), suggesting that Ub 1 might be sampling both S1 and S2 sites on PLpro.

In agreement with the MST results, our NMR data demonstrate that placement of an aspartate at the C-terminus of ISG15, K48-Ub 2 , and Ub 1 reduced substantially their affinity for PLpro CoV-2 , as evident from noticeably weaker NMR signal perturbations observed in both the D-extended substrates and PLpro ( Supplementary Figs. 8e and 9 ). It should be mentioned that in all the NMR studies presented here the addition of PLpro CoV-2 resulted in the overall NMR signal broadening/attenuation reflecting an increase in the size (hence slower molecular tumbling) upon complexation with a ~36 kDa protein. The finding that hISG15 and K48-Ub 2 bind to the same sites on PLpro CoV-2 enabled us to directly compare their affinities for PLpro CoV-2 in a competition assay where hISG15 was added to a preformed PLpro CoV-2 :K48-Ub 2 complex, and the bound state of Ub 2 was monitored by 1 H-15 N NMR signals of 15 N-labeled proximal Ub (Fig. 3j) . Titration of unlabeled hISG15 into a 1:1.5 mixture of K48-Ub 2 and PLpro CoV-2 resulted in gradual disappearance of PLprobound signals of Ub and concomitant emergence of free K48-Ub 2 signals at their unbound positions in the spectra (Fig. 3j, Supplementary Fig.8c ). The observed decrease in the intensity of the bound signals agrees with the prediction based on the K d1 values for K48-Ub 2 and ISG15 derived from our MST experiments (Fig. 3j) but not with the K d values reported previously 17 (Supplementary Fig. 8d ).

Since ISG15 binds to the same PLpro CoV-2 surface as Ub 2 and contains an uncleavable linkage, we then examined if ISG15 can inhibit polyUb cleavage by PLpro. When hISG15 was added to cleavage reaction of Ub 3 or Ub 2 by PLpro CoV-2 it did not interfere with Ub 3 cleavage to Ub 2 (Fig. 3k, left gel) but it blocked hydrolysis of Ub 2 to monomers (Fig. 3k , right gel, also Supplementary Fig. 8e) . A similar effect was observed on cleavage of Ub 2 in the presence of Ub 1 (Supplementary Fig. 8e ). This can be explained by different options for productive cleavage of Ub 3 and Ub 2 . The productive cleavage of Ub 3 is predominantly accomplished by binding of two Ub units (2 and 1) to S1 and S2 sites on PLpro CoV-2 , respectively and the unit 3 of Ub 3 occupying the S1' site, thus placing the isopeptide bond on its K48 in the active site of PLpro CoV-2 . This is a high affinity Ub 3 binding, and ISG15 and particularly Ub 1 cannot easily compete for binding. As discussed earlier, K48-Ub 2 can bind in two different modes, one with two Ub domains binding to S1 and S2 sites on PLpro, but this binding cannot not result in cleavage. In order to break the isopeptide bond between two Ubs the distal Ub must bind to the S1 site on PLpro such that the proximal Ub will then occupy the S1' site.

The LRGG motif can then be recognized, and the isopeptide bond is cleaved. But Ub 2 binding through a single Ub unit to S1 site is of low affinity, thus both ISG15 and Ub 1 can compete with Ub 2 and inhibit its cleavage.

Our structural experiments indicate differences in how ISG15 and Ub 2 are recognized by PLpro CoV-2 . To gain more insight into the proposed dynamics of the interactions we employed a XL-MS approach (Fig. 4a ) using three different cross-linking chemistries. We found that the 4-(4,6 dimethoxy-1,3,5-triazin-2-yl)-4-methyl-morpholinium chloride (DMTMM) cross-linker produced robust heterodimers of PLpro CoV-2 with hISG15, K48-Ub 2 , Ub 1 or K48-Ub 2 -D77 ( Fig. 4b and Supplementary Fig. 10a ). We identified two contacts between D61 and D62 on PLpro CoV-2 to K35 on the distal UBL of hISG15 (Fig. 4c, 19 and 43 contacts) which map well onto our structure with the distances between carboxylates (D61 and D62) and Nz (K35) of 14.9 and 8.6 Å, respectively ( Fig.   4c ) with Cb-Cb distances below 30 Å consistent with the cross-linker geometry 28 . By contrast, we detected 19 cross-links between PLpro and K48-Ub 2 , however, due to the sequence degeneracy between the two Ubs we interpreted the data based on shortest distance (Supplementary Fig. 10b ).

Using this strategy, 12 of the 19 observed contacts fall below a 30 Å threshold (Fig. 4d ). Of these 12 contacts, 7 involve K6 from the distal Ub to the N-terminal thumb domain of PLpro CoV-2 consistent with the distal Ub binding mode seen in the PLpro CoV-1 :Ub 2 structure 21 . Additionally, of the 12 contacts, two between K190 on the fingers domain of PLpro CoV-2 to E64 and E18 of Ub have Cb-Cb 16 Å and 28.6 Å distances which are compatible with the proximal Ub geometry (Fig. 4d) . These data suggest contacts compatible with placement of the Ubs in the S1 (proximal) and S2 (distal) sites.

Similarly, for Ub 1 we found 23 cross-links that localize to both S1 and S2 sites ( Supplementary Fig.   10c ), consistent with our NMR data. For the PLpro CoV-2 :K48-Ub 2 cross-links we can explain 12 of 19 identified pairs based on Cb-Cb based on distances and geometry alone. To find alternate binding sites, we designed a modeling strategy in which the distal Ub is modeled by docking a Ub monomer to PLpro:Ub complex containing Ub in the S1 (proximal) binding site and utilizing constraints that place the docked (distal) Ub with its C-terminus in proximity to K48 of the proximal Ub (Fig. 4e, g, blue spheres). We produced over 5000 models and compared the energy of the assembly as a function of the sum of distances between 7 cross-linked atoms pairs that were unexplained in the first model ( Supplementary Fig. 10d ). We found a low energy model that explains an additional 4 of the 7 crosslinks (Fig. 4f ) that localize the Ub to the PLpro's UBL domain and the lower part of the N-terminal thumb domain (Fig. 4g) . In a parallel docking approach, we assumed an alternative binding mode where the distal Ub was placed in the S1 (proximal) binding site ( Supplementary Fig. 10e ), and applied a constraint from the C-terminus of that distal Ub to K48 of the docked (proximal) Ub.

Similarly, we compared the ensemble of structures as a function of total energy and the sum of distances between cross-link pairs ( Supplementary Fig. 10e ). We found a low energy model in which the docked Ub is placed near the PLpro's UBL that explains 4 of 7 contacts (Supplementary Fig. 10f ).

These analyses indicate that the 4 contacts can be explained with two different binding modes. To resolve this discrepancy, we performed XL-MS on PLpro CoV-2 :Ub 2 in which only the proximal Ub was uniformly 15 N labeled. Analysis of cross-links containing the unlabeled (distal) Ub uncovered 14 cross-links ( Supplementary Fig. 10g ) of which 8 are compatible with placement of this Ub in the S2 (distal) site and 7 of these again involve K6 interacting with the N-terminal thumb domain similar to the data collected with unlabeled Ub 2 (Fig. 4d) . This allowed us to propose that K48-Ub 2 can bind to PLpro in two different binding modes (see Fig. 4e and Supplementary Fig. 10e ).

Having more confidence that the distal Ub is bound in the S2 site, we again mapped the remaining 7 unexplained contacts onto the docked model in which we sampled movement of the distal Ub (Fig.   4e) ; this model can explain 4 additional contacts (Fig. 4f, g) . Finally, we also interpreted XL-MS data on PLpro CoV-2 :Ub 2 -D77 and using a model derived from the sampling of the S1 site ( Supplementary   Fig. 10e) we can explain the two contacts consistent with alternate binding modes of Ub 2 -D77 ( Supplementary Fig. 10i ) with one contact comprising the distal Ub bound to the S1 site and the second involving the proximal Ub bound to the UBL domain of PLpro (S1' site). This change in binding mode may explain faster cleavage kinetics of Ub 2 -D77 compared to Ub 2 (Supplementary Fig.   2h ). Our combined experiments not only reaffirm the dominant binding modes between PLpro and ISG15 and K48-Ub 2 (to S1 and S2 sites) but also begin to clarify alternate binding modes that explain the heterogeneity of binding to S1 and S1' sites.

In silico alanine scan of PLpro CoV-2 in complex with substrates reveals binding hotspots for Ub 2

To better understand the energetic contribution of the residues at the PLpro:substrate interfaces, we applied an in silico alanine scan approach 29 to PLpro CoV-2 in complexes with hISG15 and K48-Ub 2 (Fig. 4h) . We first identified PLpro interface residues that contact the substrates and employed Rosetta 29 to calculate DDG binding comparing WT and alanine mutants. Analysis of DDG binding for PLpro CoV-2 with the two substrates revealed interaction hotspots, most notably PLpro CoV-2 residues E167/Y264 and F69, for stabilizing S1 and S2 Ub/UBL binding sites in K48-Ub 2 and hISG15 substrates (Fig. 4i) . Additionally, we find that for PLpro CoV-2 , the distal site (S2) has a preference for ISG15 ( Fig. 4i ,j, colored in red) but the proximal site (S1) has an overall preference for Ub (Fig. 4i ,j, colored in blue). For K48-Ub 2 the primary interaction is with the proximal Ub domain in S1 site but additional interaction comes also from the distal Ub interacting with S2 site contributing to stronger binding. In protein complexes, residues that surround protein interaction hotspots typically play important roles in determining specificity 30 . Indeed, in the S1 binding site, Y171 provides more stabilization for hISG15 compared to Ub 2 (Fig. 4i) . These analyses highlight how prediction of binding energetics combined with structural data can help interpret dynamics of domain binding.

The literature highlights that PLpro CoV-1 prefers binding to K48-linked polyubiquitin over ISG15 21 .

Our work incited by recent studies 17,20 reveals that PLpro CoV-2 recognizes ISG15 and K48-Ub 2 with very similar affinities. Sequence analysis suggests that the PLpro from these two viruses only vary at 8 amino acid positions at the substrate binding interface implicating only minor sequence changes responsible for improving binding of ISG15. Interestingly, both K48-Ub 2 and ISG15 utilize two Ub/UBL domains to recognize and bind PLpro CoV-2 but our data show they do it differently. In ISG15 the two UBL domains are connected through a relatively short, likely more rigid peptide linker (DKCDEP in hISG15), and the C-terminal (proximal) UBL binds to S1 site while the N-terminal (distal) UBL binds to S2 site on PLpro CoV-2 . The amino acid sequences of the proximal and distal UBLs are somewhat different and show distinct contacts that are required for productive binding. In K48-Ub 2 two identical Ub units are connected through a flexible linker (RLRGG-K48) 31,32 and contribute differently to binding. In the high affinity complex with PLpro the proximal Ub binds to the S1 site and the distal Ub binds to S2 site. The proximal Ub is very well ordered and shows multiple interactions with PLpro. The distal Ub interacts with PLpro differently. In the crystal structure this Ub domain is less ordered, suggesting multiple possible states. NMR data clearly show interactions of the distal Ub with PLpro, but presence of less occupied states cannot be excluded.

K48-Ub 2 also seems to bind to PLpro with an altered register where the distal Ub binds to the S1 site while the proximal Ub occupies the S1' site. This is consistent with how PLpro disassembles Ub 3 by cleaving off the unit 3 and the fact that the binding curves of Ub 2 and Ub 1 to PLpro CoV-2 are explained better by presence of two binding sites. Both hISG15 and the K48-Ub 2 should be sensitive to mutations of S1 and S2 sites in PLpro, but because of different interaction modes, binding of the hISG15 and the K48-Ub 2 substrates is likely to have different sensitivity to such mutations.

Furthermore, evolutionary analysis of SARS-CoV-2 variants (Supplementary Fig. 11 ) highlights sequence variation at key sites suggesting the occurrence of PLpro variants that may have altered substrate specificity with unknown disease outcomes.

The structure of PLpro CoV-1 :K48-Ub 2 complex 21 revealed both Ubs bound to the protease S1 and S2 binding sites, with proximal and distal Ubs connected via a non-hydrolysable triazole linker (in lieu of the native isopeptide linkage) and the C-terminal tail covalently attached to the protease. This may rigidify the K48-Ub 2 concealing the true interactions. There is no structure of the ISG15 bound to PLpro CoV-1 that could reveal their interactions. Our structure of PLpro CoV-2 :hISG15 as well as the previous structure of PLpro CoV-2 :mISG15 reveal a dual UBL domain recognition binding mode despite surprising species sequence variation at the UBL active surfaces (Supplementary Fig. 1a-c) . Our in silico alanine scan of the interfaces uncovered residues that may play central roles in stabilizing the ISG15 binding mode for PLpro CoV-2 and implicated V66 and F69 as being important for stabilizing the distal UBL domain of ISG15.

Much of the focus on understanding how sequence variation impacts pathogenicity, infectivity and virulence of SARS-CoV-2 has been centered on sequence changes in surface proteins such as the receptor-binding domain (RBD) of the spike protein which are essential for recognition of ACE2, virus entry into host cell and thus infectivity. Furthermore, there is concern that mutations in the viral receptors may overcome vaccines which were designed against an engineered prefusion "stabilized" conformation of RBD of the spike protein, particularly worrisome with emerging new variants like the Omicron BA.2 and Ontario WTD clade 33 . Therefore, additional SARS-CoV-2 life cycle steps must be explored, and appropriate key drug targets identified to expand treatment options.

Viral interference with host innate immune response is one of these steps of which ISG15 is integral. Several coronavirus Nsps have been shown to contribute to diminishing this complex response mechanism. Modeling of the protein interfaces suggests that the sequence variation between PLpro from SARS-CoV-1 and SARS-CoV-2 plays a role in recognition specificity of host factors.

Furthermore, we also show that sequence variation within PLpro from 2.3 million SARS-CoV-2 isolates is overall distributed with some hotspots that mimic sequence variation observed between SARS-CoV-1 and SARS-CoV-2. While we do not understand how differential recognition of Ub compared to ISG15 impacts pathogenicity and virulence of SARS coronaviruses, balance between dysregulation of the protective interferon response and ubiquitin-proteasome systems likely influences virus interaction with the host defense mechanisms. Future work must be focused on understanding how protease specificity impacts pathogenicity. Furthermore, it remains unknown whether PLpro encodes additional specificity for the substrates that are linked to Ub or ISG15 modifications.

PLpro CoV-2 must recognize and process multiple substrates: polyproteins 1a and 1ab, polyUb and ISG15. It is also known to cleave several human host proteins. All these substrates have common sequence recognition motif "LRGG", however they differ in several ways. In coronavirus polyproteins and several host proteins PLpro cleaves a regular peptide bond. In K48-polyUb and ISG15-modified protein substrates the cleaved isopeptide bond is between the C-terminal carboxylate of Ub/UBL and a lysine side chain of Ub or other protein, but these latter substrates differ in how the Ubs or UBLs are linked. It is interesting that conformation of PLpro in complexes with Ub, hISG15 and mISG15 is very similar (0.7 Å rmsd). However, the substrates conformations differ.

What may be the biological implication of single vs dual domain recognition of Ub/UBL for hydrolysis of polyUb/ISG15 modifications? ISG15 is gene-coded fused dimer, it functions as a di-UBL and is attached covalently to proteins as such. Its specific removal by viral PLpro is also hardwired to its dimer structure. Ubiquitin is different as it exists as a monomer and is added to a polyUb chain or other proteins in units of monomer. However, PLpro shows the highest affinity for Ub 2 (or longer chains) vs Ub 1 and it most efficiently removes the proximal Ub (unit 3) from Ub 3 . Because PLpro binds single Ub less strongly, this suggests that in the cell proteins tagged with polyUb chains containing an odd number of Ubs may accumulate Ub 1 -substrate adducts, as shown in our cleavage studies.

In summary, we find that PLpro CoV-2 binds both hISG15 and K48-Ub 2 with high and similar affinity but shows weaker interactions with Ub 1 . Our data also suggest the presence of lower affinity secondary binding sites for hISG15 and K48-Ub 2 which can be explained by alternate binding modes.

We determined the first structures of untethered PLpro CoV-2 complexes with hISG15 and K48-Ub 2 .

We combine these data with solution NMR, cross-linking mass spectrometry, binding and computational modeling. Our collective data showed that high affinity binding to hISG15 is determined by dual domain recognition while K48-Ub 2 is recognized mainly through proximal Ub, with distal Ub contributing to binding using different binding modes. Our findings pave the way to understand the interaction of PLpro with hISG15 and (poly)ubiquitin substrates and uncover binding heterogeneity that appears to decouple binding affinity from protease activity. Future experiments will focus on how sequence changes in PLpro influence the distribution of primary and secondary binding sites and how this modifies proteolytic activity. Finally, it remains unknown how changes in the protease sequence dictate substrate binding modes and viral pathogenesis outcomes.

WT and mutant PLpro CoV-2 , hISG15, ISG15-D158, Ub, and Ub-D77 were expressed and purified as previously described 17, 23, 34 . K48-linked Ub 2 , Ub 2 -D77 and Ub 3 were enzymatically synthesized as Samples were analyzed on a Thermo Orbitrap Fusion Tribrid system at the UTSW and the data processed by xQuest to identify high confidence contacts. Cross-links were visualized on the structures using PyMOL and the distances between cross-linked atoms were calculated using an inhouse script. Ub was docked onto PLpro:Ub proximal complexes to identify alternate binding sites. Over 5000 models were produced using the docking module in Rosetta 46 using two different geometric restraints: 1) between K48 of proximal Ub and C-term of docked Ub to identify alternate S2 sites or 2) between C-term of distal Ub and K48 of docked Ub to identify alternate S1' sites. The sum of distances between all cross-link pairs were evaluated as a function of energy of the complex. Low energy docked models that explained the cross-links were further evaluated. The Flex ddG protocol 29 was combined with an in silico alanine scan to evaluate the per residue energetics of PLpro CoV-in supplementary information.

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

The results included in this manuscript can be reproduced by following protocols and using materials described in Materials and Methods.

The structural datasets generated during the current study are available in the Protein Data Bank 

The authors declare that they have no competing interests. 

Ub units are shown as in (d). (f) Overlay of bound conformation of K48-linked Ub 2

Proteins are shown in cartoon and colored grey (PLpro), magenta (hISG15) and orange (Ub 2 ). (h) Zoom in of the boxed area in (g) representing overlay of the proximal domain of hISG15 and Ub. Proximal domain of hISG15 and Ub are represented as in (g). Rotation of the active surface indicated with arrows. (i,j) Comparison of the active surfaces of hISG15 (i) and Ub 1 (j)

Schematic illustration of cross-linking mass spectrometry experiments for heterodimer complexes of PLpro CoV-2 with hISG15, K48-Ub 2 , and Ub 1 . Active site of PLpro CoV-2 is indicated in yellow. (b) Cross-linked samples reveal formation of covalent heterodimer complex bands (red, DMTMM cross-linker) compared to untreated reactions (control) by SDS-PAGE. (c) Interactions between PLpro CoV-2 (D61/D62) and the distal hISG15 domain (K35) identified by XL-MS mapped on the structure of heterocomplex. Both identified contacts are shorter than 30 Å. PLpro (grey) and hISG15 (magenta) are

Cross-links between distal Ub and thumb and UBL domains of PLpro CoV-2 are colored red. Constraint used in docking (e) is indicated as blue spheres. PLpro CoV-2 and K48-Ub 2 are shown as in (d). (h) Schematic illustration of identification of substrate binding surfaces on PLpro CoV-2 and in silico mutagenesis for heterodimer complexes with hISG15 and K48-Ub 2 . (i) Heat-map results of ΔΔG calculations of in silico alanine scan for PLpro CoV-2 in complex with hISG15 or K48-Ub 2 . Row titles represent positions of residues on the interaction surface. The last column represents results of calculations of difference between REU (Rosetta energy units) for PLpro CoV-2 :hISG15 and PLpro CoV-2 :K48-Ub 2 . (j) Results of in silico mutagenesis for complexes of PLpro CoV-2 with hISG15 and K48-Ub 2 mapped on surface representation of

A high-resolution temporal atlas of the SARS-CoV-2 translatome and transcriptome

The Architecture of SARS-CoV-2 Transcriptome

Interactions between SARS coronavirus 2 papain-like protease and immune system: A potential drug target for the treatment of COVID-19

The autocatalytic release of a putative RNA virus transcription factor from its polyprotein precursor involves two paralogous papain-like proteases that cleave the same peptide bond

Nidovirus papain-like proteases: multifunctional enzymes with protease, deubiquitinating and deISGylating activities

Decoupling deISGylating and deubiquitinating activities of the MERS virus papain-like protease

Binding site-based classification of coronaviral papain-like proteases

Deubiquitinating enzymes (DUBs): Regulation, homeostasis, and oxidative stress response

Structural Basis for the Ubiquitin-Linkage Specificity and deISGylating activity of SARS-CoV papain-like protease

SARS coronavirus papain-like protease inhibits the type I interferon signaling pathway through interaction with the STING-TRAF3-TBK1 complex

SARS coronavirus papain-like protease induces Egr-1-dependent upregulation of TGF-β1 via ROS/p38 MAPK/STAT3 pathway

SARS-CoV-2 proteases PLpro and 3CLpro cleave IRF3 and critical modulators of inflammatory pathways (NLRP12 and TAB1): implications for disease presentation across species

The SARS-CoV-2 SSHHPS Recognized by the Papain-like Protease

The papain-like protease of coronaviruses cleaves ULK1 to disrupt host autophagy

Modulation of Extracellular ISG15 Signaling by Pathogens and Viral Effector Proteins

ISG15-dependent activation of the sensor MDA5 is antagonized by the SARS-CoV-2 papain-like protease to evade host innate immunity

Papain-like protease regulates SARS-CoV-2 viral spread and innate immunity

SARS-CoV-2 infection of human iPSC-derived cardiac cells reflects cytopathic features in hearts of patients with COVID-19

Structural Insights into the Interaction of Coronavirus Papain-Like Proteases and Interferon-Stimulated Gene Product 15 from Different Species

Mechanism and inhibition of the papain-like protease, PLpro, of SARS-CoV-2

Recognition of Lys48-Linked Di-ubiquitin and Deubiquitinating Activities of the SARS Coronavirus Papain-like Protease

Surface hydrophobic residues of multiubiquitin chains essential for proteolytic targeting

Controlled synthesis of polyubiquitin chains

Structural and biochemical studies of the open state of Lys48-linked diubiquitin

Structural properties of polyubiquitin chains in solution

Polyubiquitin and ubiquitin-like signals share common recognition sites on proteasomal subunit Rpn1

Chemical cross-linking/mass spectrometry targeting acidic residues in proteins and protein complexes

Flex ddG: Rosetta Ensemble-Based Estimation of Changes in Protein-Protein Binding Affinity upon Mutation

A De Novo Protein Binding Pair By Computational Design and Directed Evolution

Linkage-specific conformational ensembles of non-canonical polyubiquitin chains

Recovering a representative conformational ensemble from underdetermined macromolecular structural data

Highly divergent white-tailed deer SARS-CoV-2 with potential deer-tohuman transmission

Structure of papain-like protease from SARS-CoV-2 and its complexes with non-covalent inhibitors

On the acquisition and analysis of microscale thermophoresis data

Using two-site binding models to analyze microscale thermophoresis data

HKL-3000: the integration of data reduction and structure solution--from diffraction images to an initial model in minutes

The PDB_REDO server for macromolecular structure model optimization

Overview of the CCP4 suite and current developments

Features and development of Coot

NMRFAM-SPARKY: enhanced software for biomolecular NMR spectroscopy

Lysine-specific chemical cross-linking of protein complexes and identification of cross-linking sites using LC-MS/MS and the xQuest/xProphet software pipeline

Chemical cross-linking/mass spectometry targeting acidic residues in proteins and protein complexes

Proximity-enhanced SuFEx chemical cross-linker for specific and multitargeting cross-linking mass spectrometry

Identification of cross-linked peptides from large sequence databases

Protein-protein docking with backbone flexibility

An exact mathematical expression for describing competitive binding of two different ligands to a protein molecule

We thank the members of the SBC at Argonne National Laboratory, especially Darren Sherrell, Krzysztof Lazarski and Alex Lavens for their help with setting beamline and data collection at