Mass spectrometric assays monitoring the deubiquitinase activity of the SARS-CoV-2 papain-like protease inform on the basis of substrate selectivity and have utility for substrate identification

Graphical abstract


Introduction
The SARS-CoV-2 genome encodes two nucleophilic cysteine proteases that catalyze hydrolysis of the viral polyproteins pp1a/1ab into functional non-structural proteins (nsps), i.e. the papain-like protease (PL pro , a domain of nsp3) and the main protease (M pro , nsp5). 1,2PL pro catalyzes the hydrolysis of peptide bonds in pp1a/1ab C-terminal to three LXGG motifs (X represents a non-conserved residue) to release nsps1-3.M pro catalyzes the hydrolysis of peptide bonds in pp1a/1ab Cterminal to eleven glutamine residues in (L/F/V)Q(S/A/N) motifs to release nsps 4-16. 1 Catalysis by the two SARS-CoV-2 proteases is essential for viral replication and their inhibition is of therapeutic relevance, [3][4][5][6][7][8] i.e. the small-molecules nirmatrelvir and ensitrelvir, which are clinically used to treat COVID-19, inhibit M pro . 9,102][13][14][15] Interestingly, only a few of the human proteins that contain (L/F/V)Q motifs in their coding sequence are validated substrates of isolated recombinant SARS-CoV-2 M pro , e.g.human nuclear factor (NF)-κB essential modulator (NEMO), the M pro -catalyzed hydrolysis of which has been reported to induce the death of brain endothelial cells. 16,179][20] Although proteomic studies with samples from COVID-19 patients indicate that PL pro may accept human proteins as substrates in vivo, 21 both the extent and the clinical significance of the PL pro -catalyzed degradation of human host proteins during SARS-CoV-2 infections are incompletely understood.Such knowledge is important with respect to understanding the consequences of therapeutic PL pro inhibition.
LXGG motifs are conserved in human proteins which are posttranslationally modified with ubiquitin (Ub) or ubiquitin-like modifiers (UBL), e.g.interferon stimulated gene 15 (ISG15), [22][23][24][25] neural precursor cell expressed, developmentally down-regulated 8 (NEDD8), 26,27 and ubiquitin-related modifier-1 (URM1). 28,29Ub and UBLs are linked to one or multiple N ε -amino groups of a protein lysine residue via an isopeptide amide bond formed with their C-terminal residue, which forms a LXGG motif in the case of Ub, ISG15, NEDD8, and URM1. 30,31][44] The apparent selectivity of SARS-CoV-2 PL pro for catalyzing deISG15ylations is precedented by the substrate selectivity of the human DUB ubiquitin specific peptidase 18 (USP18) which also employs a nucleophilic cysteine to selectively catalyze protein deISG15ylations over protein deubiquitinylations. 45 Notably, the DUB activity of SARS-CoV-2 PL pro is reported to attenuate the host innate immune response via catalyzing the deISG15ylation of ISG15ylated interferon regulatory factor 3 (IRF3), thus potentially enhancing virulence. 39,46eported assays which monitor the DUB activity of PL pro typically employ Ub/UBL derivatives conjugated to a fluorescent group via a Cterminal amide (e.g.Ub-AMC or ISG15-AMC; AMC: 7-amino-4-methylcoumarin), the hydrolysis of which can be spectroscopically monitored. 39,40,48,49Matrix-assisted laser desorption/ionization (MALDI) mass spectrometry (MS)-based assays, originally developed to monitor the activity of human DUBs, 50 have also been employed to monitor the DUB activity of PL pro ; however, the MALDI-MS assay uses Ub 2 as substrate which is apparently not the preferred PL pro substrate in cells.The use of substrates which do not bind to the PL pro S′ sites, the relatively high costs associated with use of the Ub-AMC, ISG15-AMC, and Ub substrates, and/or the use of artificial C-terminal fluorophores limit potential applications of these assays.
We have reported solid-phase extraction coupled to mass spectrometry (SPE-MS)-based assays which directly monitor PL pro -catalyzed hydrolysis of oligopeptide fragments derived from the sequence of the Wuhan-Hu-1 strain 52 of SARS-CoV-2 pp1a/1ab (Fig. 1b).The assay results revealed that PL pro preferentially catalyzes hydrolysis of an oligopeptide which was based on the LXGG motif separating nsp2 and nsp3, i. e. nsp2/3 808-827 (2; Fig. 1). 47Here we report the results of SPE-MS assays employing N ε -lysine branched synthetic oligopeptides as substrates, that bind to both the S and S′ sites of PL pro , to monitor the DUB activity of isolated recombinant SARS-CoV-2 PL pro .The results inform on the substrate requirements for efficient PL pro DUB catalysis and reveal that not only the presence of an LXGG motif but also the entire coding sequence of the substrate which binds in proximity of the active site, in particular that of the UBL, determines the catalytic efficiency of PL pro .

Assaying the DUB activity of SARS-CoV-2 PL pro using mass spectrometry
SARS-CoV-2 PL pro is reported to attenuate the human innate immune system by catalyzing the deISG15ylation of post-translationally ISG15ylated IRF3; 39 however, to our knowledge, there are no reports on validation of the DUB activity of isolated recombinant PL pro with oligopeptide substrates.We thus synthesized SPE-MS-compatible N εlysine-branched oligopeptides based on the three reported ISG15ylation sites of IRF3, i.e.K193, K360, and K366, 53 using solid phase peptide synthesis (SPPS).The IRF3-derived peptide fragment was synthesized on a Rink amide resin from the C-to the N-terminus employing Fmoc strategy, N-Fmoc lysine with a 4-methyltrityl (Mtt) protected N ε -amine at the site of the lysine branching, and an N-terminal amino acid with a Boc-protected N α -amine.Branching was introduced via selective deprotection of the lysine N ε -Mtt group using 1% v/v trifluoroacetic acid (TFA) and 2% v/v triisopropylsilane (TIPS) in dichloromethane, followed by SPPS to build the ISG15 fragment from C-to N-termini starting with the N ε -amino group of the IRF3 lysine at the C-terminus.The purified branched oligopeptides were obtained following global deprotection and cleavage from the resin with TFA and HPLC purification (Supporting Figure S1).
The K 193 -branched IRF3 189-197 -ISG15 oligopeptide 4 was incubated with the isolated recombinant PL pro domain of SARS-CoV-2 nsp3 (0.2 µM; enzyme/substrate ratio: 1/10) employing our reported conditions for the PL pro -catalyzed hydrolysis of the linear pp1a/1ab-derived oligopeptide nsp2/3 808-827 (2) (50 mM Tris, pH 8.0, 37 • C). 47 Analysis of the reaction mixture after 14 h incubation by SPE-MS indicated apparent quantitative isopeptide amide bond hydrolysis, as supported by the observed masses of the resultant N-and C-terminal product fragments.Thus, the results confirm the DUB activity of PL pro , which has been assigned on the basis of cellular studies and on studies with ISG15 containing a C-terminal fluorophore, 39,40,48 with an N ε -lysine-branched oligopeptide substrate.
L. Brewitz et al.  conditions were subsequently varied, with respect to, e.g., buffer composition, pH, and salt additives; the tested variations did not, however, increase the rates of isopeptide amide bond hydrolysis substantially.Note that the corresponding N-terminally N-acetylated ISG15-derived hydrolysis product peptide, i.e.Ac-LSTVFMNLRLRGG-NH 2 (5), was added as an inert internal standard to the reaction mixture to enable quantification of PL pro -catalyzed product formation; the presence of 5 in the reaction mixture did not affect PL pro catalysis, at least substantially (Supporting Figure S2).
Derivatives of the K 193 -branched IRF3 189-197 -ISG15 oligopeptide 4 were synthesized to investigate whether varying the length of the IRF3or the ISG15-derived fragments affects PL pro catalysis (Fig. 2a and 2b).Interestingly, SPE-MS analysis of the PL pro -catalyzed hydrolysis of the oligopeptides 4, 6, 7, and 8 revealed that varying the length of the IRF3derived fragment did not affect catalysis substantially, though the hydrolysis of the K 193 -branched IRF3 185-200 -ISG15 oligopeptide 8, which has the relatively largest IRF3-derived fragment of the tested oligopeptides, appeared to be slightly favored (Fig. 2a).By contrast, varying the length of the ISG15-derived fragment had a substantial effect on PL pro catalysis; the oligopeptide 9 with the relatively largest ISG15-derived fragment bearing 18 C-terminal amino acids was hydrolyzed substantially more efficiently than oligopeptide 11 with the relatively smallest ISG15-derived fragment bearing only the essential C-terminal LRGG tetrad (Fig. 2b).Levels of PL pro -catalyzed substrate hydrolysis appear to increase with increasing length of the ISG15-derived fragment peptide; the rank order of PL pro substrate preference was determined by quantifying product formation relative to Ac-ENPLKRLLV-NH 2 (12), which was used as an internal standard in the reactions.The combined oligopeptide results indicate that the binding of the N ε -lysine-branched oligopeptides to the S′ sites is less important for catalysis than binding to the S sites.The situation may, however, be different with full-length folded protein substrates.Note that either the corresponding N-terminally N-acetylated ISG15-derived hydrolysis product peptide, i.e.Ac-LSTVFMNLRLRGG-NH 2 (5), or the corresponding N-terminally N-acetylated IRF3 189-197 -derived hydrolysis product peptide, i.e.Ac-ENPLKRLLV-NH 2 (12), was added as an internal standard to the reaction mixtures to enable quantification of PL pro -catalyzed product formation; the presence of 5 or 12 in the reaction mixture did not affect PL pro catalysis, at least substantially (Supporting Figures S2 and S3).
The K 360 -branched IRF3 357-364 -ISG15 peptide 13 and the K 366branched IRF3 362-370 -ISG15 peptide 14 were synthesized and tested to investigate the potential of isolated recombinant PL pro to catalyze regioselective deISG15ylations of post-translationally modified IRF3.Oligopeptides 4, 13, and 14 were incubated with PL pro in the presence of the internal standard 5, and isopeptide amide bond hydrolysis was monitored using SPE-MS (Fig. 2c).The results reveal that the sequences of the three tested IRF3-derived peptide fragments, which likely bind to the S′ sites of PL pro , do not affect the rate of peptide hydrolysis substantially, in accord with prior results which showed that the length of the IRF3-derived peptide fragment affects PL pro catalysis less substantially than the length of the ISG15-derived peptide fragment (Fig. 2a and  2b).The observation that PL pro may catalyze the deISG15ylation of oligopeptide 13 slightly more efficiently than that of 4 and 14 might be a result of the comparatively lower purity of 13 compared to that of all other tested peptides (Supporting Figure S1).Note that PL pro may catalyze the regioselective deISG15ylation of full-length folded IRF3 in a cellular context.

The primary sequence of the PL pro substrates affects catalysis
To investigate whether the substrate preference of SARS-CoV-2 PL pro depends on the primary sequence of the UBLs, we synthesized the oligopeptide derivatives of K 193 -branched IRF3 189-197 -ISG15 4, in which the ISG15-fragment was substituted for the corresponding first 13 Cterminal amino acids of Ub (IRF3 189-197 -Ub, 15) or the UBLs NEDD8 (IRF3 189-197 -NEDD8, 16), URM1 (IRF3 189-197 -URM1, 17), and small ubiquitin-related modifier 1 (SUMO1; IRF3 189-197 -SUMO1, 18), the latter of which does not contain a C-terminal LXGG motif and should hence not be a substrate of PL pro (Supporting Figure S1).K 193 -branched IRF3 189-197 derivatives were attractive synthesis targets because the ubiquitinylation at K193 had been reported 54 ; the corresponding NEDD8, URM1, and SUMO1 derivatives of 4 were also synthesized despite their unclear biological relevance.The length of the UBL fragment was kept constant to help enable comparison, as PL pro catalyzes the hydrolysis of derivatives of 4, in which the length of the ISG15 fragment was varied, with different efficiencies (Fig. 2b).
Interestingly, the results with 4, 15, and 16 reflect reported SARS-CoV-2 PL pro substrate preferences observed in cellular studies which also showed that PL pro -catalyzed protein deISG15ylations are more efficient than the corresponding deubiquitinylations, which are both more efficient than deNEDD8ylations. 39,40Thus, the combined evidence suggests that not only the substrate fold affects PL pro catalysis, inter alia via binding to allosteric sites, but also, in particular, the sequence identity of the amino acid fragment N-terminal to the LXGG motif of the substrate.
In general, the ISG15ylation sites of human proteins are poorly characterized, in part, likely because the standard workflow to localize protein ubiquitinylation sites, i.e. protein denaturation followed by tryptic digestion and analysis of the resultant peptide fragments by MS/ MS, affords the same lysine-branched peptide fragments for both ubiquitinylated and ISG15ylated proteins.Thus, a different workflow is required, e.g.6][57][58] Apart from ISG15ylation sites of IRF3, the ISG15ylation sites of the eIF4Ehomologous protein (4EHP) have been characterized, i.e. at K134 and K222. 594EHP binds to the mRNA 5′-cap structure and suppresses translation by competing with the eukaryotic initiation factor 4E (eIF4E) for binding to the cap structure.The ISG15ylation of 4EHP is proposed to control translation during immune responses; 59 note that the translation initiation factor eIF4G is a reported substrate of rhinovirus 2A pro and of the foot-and-mouth-disease virus (FMDV) leader protease (L pro ), suggesting that modulation of protein translation may be a general strategy of Riboviria to evade the host immune system. 60he K 134 -branched 4EHP 132-139 -ISG15 peptide 19 and the K 222branched 4EHP 218-227 -ISG15 peptide 20 were synthesized to investigate the potential of isolated recombinant SARS-CoV-2 PL pro to catalyze selective deISG15ylations of post-translationally modified 4EHP.Oligopeptides 4, 19, and 20 were incubated with PL pro in the presence of the inert internal standard 5, and isopeptide amide bond hydrolysis was monitored using SPE-MS (Fig. 3b).The results reveal that the rates of PL pro -catalyzed peptide hydrolyses for 4, 19, and 20 are similar, within experimental error (Fig. 3b).This observation is in accord with the previous results showing that PL pro catalyzes the deISG15ylation of IRF3-derived fragment peptides regardless of the site of ISG15ylation (Fig. 2c) and thus supports the proposal that the substrate binding in the S sites of PL pro is more important for efficient catalysis than substrate binding in the S′ sites of PL pro , in accord with reported studies on the substrate efficiency of oligopeptides based on SARS-CoV-2 nsps. 61Note that the observed variation in the absolute efficiency of PL pro catalysis appeared to depend on the batch of PL pro ; hence, we determined the rank order of substrate preference by quantifying product formation relative to internal standards, the rank order did not depend on the batch of PL pro .The results suggest that ISG15ylated 4EHP may be a substrate of PL pro in cells.(24) in (c)) in buffer (50 mM Tris, pH 8.0, ambient temperature).Measurement times were normalized to the first sample injection analyzed after the addition of PL pro to the Substrate Mixture (t = 0 s), by which time low levels of substrate hydrolysis were manifest.The inert N-acetylated hydrolysis product 5, 12 or 24 was used as an internal standard to quantify hydrolysis (Supporting Figures S2-S4); SPE-MS assay results are a mean of independent triplicates (n = 3; mean ± SD).

N ε -Lysine-branched oligopeptides may help enable the identification of PL pro substrates
To further dissect the effect of the UBL fragment of the tested N εlysine-branched oligopeptides, which binds to the PL pro S sites, on the rate of isopeptide amide bond hydrolysis from that of the fragment which binds to the PL pro S′ sites, and to investigate the potential of synthetic N ε -lysine-branched oligopeptides for the identification of SARS-CoV-2 PL pro substrates, we synthesized a set of N ε -lysine-branched oligopeptides which mimic protein ubiquitinylation, i.e. via employing the 13 C-terminal residues of Ub.In general, the sites of protein (poly) ubiquitinylations have been characterized in greater detail than the sites of protein ISG15ylations.Potential SARS-CoV-2 PL pro substrates were chosen on the basis of reported proteomic studies which showed, e.g., that K338 of the SARS-CoV-2 nucleocapsid (N) protein can be ubiquitinylated. 62Reduced levels of ubiquitinylation of the autophagyrelated human transmembrane protein TMEM59 (at e.g.K287 and K302) have been associated with SARS-CoV-2 infections, which may suggest that TMEM59 could be a substrate of SARS-CoV-2 PL pro . 62We thus synthesized the corresponding K 287 -branched TMEM59 285-291 -Ub peptide 21, the K 302 -branched TMEM59 298-305 -Ub peptide 22, and the K 338 -branched SARS-CoV-2_N 335-341 -Ub peptide 23 (Supporting Figure S1).
The oligopeptides 15, 21, 22, and 23 were incubated with isolated recombinant SARS-CoV-2 PL pro and hydrolysis of their isopeptide amide bond was monitored using SPE-MS (Fig. 3c).Note that the corresponding N-terminally N-acetylated Ub-derived hydrolysis product peptide, i. e. Ac-ESTLHLVLRLRGG-NH 2 (24; Supporting Figure S1), was added as an inert internal standard to the reaction mixtures to enable quantification of PL pro -catalyzed product formation; the presence of 24 in the reaction mixture did not affect PL pro catalysis substantially (Supporting Figure S4).The results reveal that SARS-CoV-2 PL pro catalyzes the hydrolysis of the IRF3 189-197 -Ub (15) isopeptide amide bond more efficiently than those of 21, 22, and 23 (Fig. 3c).The PL pro -catalyzed isopeptide hydrolysis in IRF3 189-197 -Ub (15) was ~2-fold more efficient as in SARS-CoV-2_N 335-341 -Ub (23), which was the least efficient substrate of the four substrates tested.This observation indicates that PL pro catalysis can be affected by substrate binding to the S′ sites and, by implication, also by the substrate fold in proximity of the S′ sites, by contrast to previous results with peptide fragments mimicking the ISG15ylation of IRF3 and 4EHP (Fig. 2).
The combined results suggest that PL pro may be promiscuous (including compared to M pro ) with respect to its in vivo protein substrates, and that it likely has the capability to catalyze a broad range of deISG15ylation and deubiquitinylation reactions of both host and viral proteins.The results also highlight the potential of N ε -lysine-branched oligopeptides to help enable identification of PL pro substrates, including post-translationally modified viral proteins.

Single amino acid substitutions in PL pro substrates can affect catalysis
The observation that IRF3 189-197 -ISG15 (4), IRF3 189-197 -Ub (15), and IRF3 189-197 -NEDD8 (16) all have an LRGG motif, whereas IRF3 189-197 -URM1 ( 17) has an LHGG motif and, relative to 4, 15, and 16, is a substantially poorer PL pro substrate raises the question as to whether PL pro catalyzes the hydrolysis of the corresponding IRF3 189-197 -URM1-R (25) variant of 17, in which the LHGG histidine residue has been substituted for an arginine residue, more efficiently than that of 17 (Fig. 3a).25 was  12) in (a), Ac-ESTLHLVLRLRGG-NH 2 (24) in (b)) in buffer (50 mM Tris, pH 8.0, ambient temperature).Measurement times were normalized to the first sample injection analyzed after the addition of PL pro to the Substrate Mixture (t = 0 s), by which time low levels of substrate hydrolysis were manifest.The inert N-acetylated hydrolysis products were used as internal standard to quantify hydrolysis (Supporting Figures S3 and S4); SPE-MS assay results are a mean of independent triplicates (n = 3; mean ± SD). thus synthesized (Supporting Figure S1) and incubated with isolated recombinant PL pro , and its hydrolysis was monitored using SPE-MS (Fig. 4a).
Although the PL pro -catalyzed hydrolysis of 25 was ~4-fold more efficient than that of 17 (after 4 h), it remained ~7-fold less efficient compared to the PL pro -catalyzed hydrolysis of IRF3 189-197 -ISG15 (4) (after 4 h), indicating that peptides with an LRGG motif are, in principle, better substrates than those with an LHGG motif (Fig. 4a).However, considering that 25 and 4 are based on the identical IRF3 189-197 fragment C-terminal to their LRGG motif which likely binds to the S′ sites of PL pro in an identical manner, the results support the proposal that the primary sequence of the substrate N-terminal to the LXGG motif can substantially affect catalysis.This observation may reflect the importance of substrate binding to S sites other than S1-S4, including at allosteric positions, and/or the presence of secondary structural elements in the substrate that favor PL pro catalysis.][44] The PL pro -catalyzed hydrolysis of TMEM59 298-305 -Ub ( 22) and SARS-CoV-2_N 335-341 -Ub (23), which bear an acidic ED or DD sequence, respectively, in the sequence C-terminal to their LXGG motif, is less efficient compared to that of IRF3 189-197 -Ub (15) (Fig. 3c), which bears a hydrophobic LV sequence at the corresponding position.This observation raises the possibility that acidic residues proximate to the substrate P1′ lysine residue involved in the isopeptide amide bond, are detrimental for PL pro catalysis.To test this proposal, we synthesized the corresponding SARS-CoV-2_N 335-341 -Ub_L 340 V 341 variant (26; Supporting Figure S1), in which the DD sequence was substituted for an LV sequence as present in 15, and monitored its PL pro -catalyzed hydrolysis using SPE-MS (Fig. 4b).The results reveal that PL pro catalyzes the hydrolysis of 26 with similar efficiency as that of 15 and ~2.5-fold more efficiently than that of 23 (after 4 h), supporting the proposal that hydrophobic residues may be preferred in proximity of the substrate lysine residue involved in the isopeptide amide bond, potentially due to improved binding to the PL pro S′ sites.
The combined results clearly indicate that the primary sequence of oligopeptides affects SARS-CoV-2 PL pro catalysis and that apparently minor changes in their sequence, which may be distal to the LXGG motif, can have a pronounced effect on catalysis.It thus appears that substrate binding to both the S and S′ sites of PL pro has potential to modulate catalysis.

Substrate competition studies
To investigate whether N ε -lysine-branched oligopeptides or the reported pp1a/1ab-derived linear oligopeptide nsp2/3 808-827 (2) are more efficient substrates of SARS-CoV-2 PL pro , we attempted to determine k cat /K m -values.These efforts were, however, unsuccessful due to technical limitations of the SPE-MS assay, i.e. high peptide concentrations saturated the sensor of the mass spectrometer impeding the quantification of PL pro catalysis.As an alternative, direct substrate competition studies were performed using the linear pp1a/1ab-derived oligopeptide nsp2/3 808-827 (2) and the two lysine-branched oligopeptide PL pro substrates IRF3 189-197 -ISG15 (4) or IRF3 189-197 -Ub (15).Note that nsp2/ 3 808-827 was employed as a pp1a/1ab-derived linear substrate because it was a more efficient substrate compared to linear peptides based on the nsp1/2 or nsp3/4 cleavage site, 47 in accord with studies that employed a LCMS assay to investigate the substrate preference of SARS-CoV-2 PL pro . 63quimolar amounts of nsp2/3 808-827 (2) and either IRF3 189-197 -ISG15 (4) or IRF3 189-197 -Ub (15) were incubated with isolated recombinant SARS-CoV-2 PL pro in the same reaction vessel, together with the corresponding inert N-acetylated hydrolysis product peptides of 2 and 4 or 15 (as internal standards).PL pro -catalyzed peptide hydrolysis was monitored using SPE-MS, which was suitable for these experiments provided that substrates and products have different masses. 64The combined results reveal that PL pro -catalyzed hydrolysis of 2 and 4/15 does not affect each other under the tested conditions, i.e. the PL pro -catalyzed hydrolysis of 2 proceeds as efficiently in the presence of 4 or 15 as in its absence, and vice versa (Fig. 5).This observation may indicate that, under the tested conditions, the initial binding of the peptides to PL pro is not rate-limiting regardless of whether peptides are branched or linear, i. e. the on/off-rates for peptide binding are faster than the peptide hydrolysis rates, potentially because the peptides lack appropriate folding.
The results also reveal that the PL pro -catalyzed hydrolysis of nsp2/ 3 808-827 (2) is less efficient than that of 4 or 15, i.e. ~10 % of 2 was observed to be hydrolyzed after 5 h whereas ~35-40 % of 4 and 15 were observed to be hydrolyzed after 5 h which may indicate that the k cat /K m -values of the N ε -lysine-branched oligopeptides are higher than that of 2. Notably, it appears that the absolute levels of PL pro catalysis depended on the batch of PL pro and substrate sequence used.
Analysis of the backbone RMSFs of the PL pro -bound N ε -lysinebranched oligopeptides reveals that the P6-P1 residues of the UBLderived peptide fragment are stable, with values <3 Å (Supporting Figure S14); an exception was IRF3 189-197 -URM1 (17) with a neutral P3 histidine imidazole, an observation suggesting that the protonation state of this group may affect complex stability.RMSF analysis of the oligopeptide sidechains reveals that the conformation of the P4 residue is rigid relative to its adjacent P5 and P3 residues (Supporting Figure S15), likely reflecting the conserved nature of the LXGG motif leucine residue which binds to the PL pro S4 site.Notably, the conformation of the P4 glutamine residue of the IRF3 189-197 -SUMO1 oligopeptide (18), which is not a substrate of isolated recombinant PL pro and which lacks the LXGG motif (Fig. 3a), is also rigid, suggesting that complex stability itself may not necessarily indicate productive substrate recognition.
Conserved hydrogen bonding interactions appear to stabilize binding of the oligopeptide backbone to the PL pro S4-S1 sites (Fig. 6, Supporting Figures S18 and S19), reminiscent of the interactions observed in the modelled PL pro :nsp2/3 808-827 (2) complex. 61Hydrogen bonding interactions involving residues of the UBL-derived peptide fragment Nterminal to P4-P1 may also contribute to binding, e.g. with T75 and Q174; however, they appear to be less conserved than those with the P4-P1 LXGG motif (Supporting Table S1, Fig. 6, Supporting Figures S18 and  S19), consistent with previous observations on the modelled interactions of pp1a/1ab-derived linear oligopeptides with PL pro . 61Notably, the IRF3 189-197 -SUMO1 oligopeptide (18) can apparently engage in similar interactions with PL pro as the other modelled N ε -lysine-branched oligopeptides; however, it appears that its P4 glutamine sidechain binds to the S4 site less efficiently than a leucine sidechain (Supporting Figures S20-S24).
In all the modelled PL pro :substrate complexes, the residues C-terminal to the substrate LXGG motif manifest high conformational flexibility, with backbone RMSFs exceeding 4 Å (Supporting Figure S16).The Vshaped RMSF plots indicate that the peptide backbone is flexible at both its N-and C-termini (Supporting Figure S16).Nonetheless, transient hydrogen bonding and dispersion interactions were observed between the substrates and e.g.8][69][70] In the apo form of PL pro , the BL2 loop is present in an open conformation; substrate binding likely triggers an induced fit mechanism resulting in closure of the BL2 loop, in a manner helping to orient the C-terminus of the UBL productively towards the active site. 70he conformational changes in the BL2 loop associated with substrate binding to PL pro are conserved in, at least, some human DUBs, suggesting that both viral and human DUBs employ similar mechanisms to regulate substrate selectivity. 71,72Note that SARS-CoV and SARS-CoV-2 PL pro inhibitors can also bind proximate to BL2 and alter or stabilize its conformation, 39,67,70,[73][74][75][76] as precedented by the inhibitor-induced conformational changes in BL2 of human DUBs, such as USP7 and USP14. 77ur previous modelling studies have shown that the BL2 conformation is stable in its closed form in the PL pro :nsp2/3 808-827 (2) substrate complex, though alternating between open and closed forms with apo PL pro , or when the active site is occupied by a peptide that is not efficiently hydrolyzed. 61We analyzed BL2 dynamics in all the nine modelled PL pro :substrate complexes by backbone RMSF of the most flexible residue in BL2 (i.e.Y268) (Fig. 7, Supporting Figures S25 and   S26) and the Y268-P248 Cα-Cα distance (Supporting Figures S27 and  S28); in the latter case, a distance of >12 Å was considered to be indicative of an open BL2 conformation.The results reveal that BL2 remains stably closed in the PL pro :IRF3 189-197 -ISG15 (4) and PL pro : IRF3 189-197 -Ub (15) complexes (Fig. 7), which are the most efficient substrates among the modelled peptides (Fig. 3).By contrast, BL2 is flexible in the modelled PL pro complex with the relatively inefficient substrate IRF3 189-197 -URM1 (17), regardless of the charge state of the P3 histidine imidazole group.Similarly, in the modelled PL pro :IRF3 189-197 -SUMO1 (18) complex, BL2 frequently opens and closes (Supporting Figure S28); note that 18 is not a PL pro substrate (Fig. 3).The latter observation is likely a result of the P4 leucine to glutamine substitution in 18; consistent with this proposal, previous reports have shown that the binding of hydrophobic sidechains to the PL pro S4 site induces BL2 closure. 65,70,78,79n the modelled PL pro complexes with the N ε -lysine-branched oligopeptides that mimic post-translational protein ubiquitinylation (i.e. 15, 21, 22, and 23), BL2 remains rigidly closed for the relatively more efficient substrates (i.e. 15 and 21), as observed for the pp1a/1abderived nsp2/3 808-827 peptide 2, 61 but not for the less efficient substrates (i.e.22 and 23) (Fig. 7, Supporting Figures S27 and S28).These observations are in accord with the results on the stability of the complexes of PL pro and N ε -lysine-branched oligopeptides mimicking posttranslational protein ISG15ylation.In particular, it appears that hydrophobic residues C-terminal to the P1′ lysine residue can affect BL2 flexibility (Supporting Figures S13, S23, S24, and S29-S31), in accord with the experimentally observed preference of PL pro for hydrophobic residues C-terminal to the substrate P1′ lysine residue involved in the  4) is not affected by the presence of nsp2/3 808-827 (2) and vice versa; 4 appears to be a more efficient PL pro substrate than 2; (b) the SARS-CoV-2 PL pro -catalyzed hydrolysis of IRF3 189-197 -Ub ( 15) is not affected by the presence of nsp2/3 808-827 (2) and vice versa; 15 appears to be a more efficient PL pro substrate than 2. Conditions: SARS-CoV-2 PL pro (0.2 μM), substrate peptide(s) (2.0 μM), inert N-acetylated standard peptides (0.2 μM, as indicated) in buffer (50 mM Tris, pH 8.0, ambient temperature).Measurement times were normalized to the first sample injection analyzed after the addition of PL pro to the Substrate Mixture (t = 0 s), by which time low levels of substrate hydrolysis were manifest.The inert N-acetylated hydrolysis products 27 47 and 5 or 24 were used as internal standard to quantify hydrolysis (Supporting Figures S2 and S4); SPE-MS assay results are a mean of independent triplicates (n = 3; mean ± SD). isopeptide amide bond (Fig. 4).
The combined modelling and experimental results reveal the importance of the conformational dynamics of the BL2 loop; it not only enables productive substrate binding in the active site, but also governs substrate selectivity.Importantly, the BL2 flexibility and opening tendency observed in the MD simulations correlate with the experimentally observed substrate selectivity, i.e. conformational BL2 flexibility in the PL pro :substrate complex appears to result in inefficient catalysis (Fig. 7).Although substrate peptide dissociation was not observed over the course of the MD simulations, the observations suggest that a sustained closed state of BL2 may be necessary for initiating productive catalysis, potentially because BL2 is proximal to H272 which is part of the catalytic triad. 68Thus, the observation that the PL pro -catalyzed hydrolysis of the linear pp1a/1ab-derived oligopeptide nsp2/3 808-827 (2) and the N εlysine-branched oligopeptide substrates do not appear to affect each other when incubated in the same reaction vessel (Fig. 5), implies that inefficient substrate binding to PL pro results in dissociation being faster than substrate hydrolysis.

Conclusions
SARS-CoV-2 PL pro catalyzes the hydrolysis of peptide amide bonds that are C-terminal to three LXGG motifs in pp1a/1ab to release functional nsps1-3. 1,2By contrast with M pro , PL pro is also a deubiquitinase (DUB) that catalyzed the hydrolysis of isopeptide amide bonds of human proteins that are post-translationally modified with Ub/UBLs, in particular ISG15. 39,40Thus, the DUB substrate preference of PL pro is apparently similar to that of the human ubiquitin specific peptidase 18 (USP18), a DUB which also employs a nucleophilic cysteine and which selectively catalyzes protein deISG15ylations. 35,45It has been proposed that the DUB activity of PL pro modulates virulence via catalyzing the deISG15ylation of post-translationally ISG15ylated IRF3. 39However, to our knowledge, the DUB activity of isolated recombinant PL pro has not yet been validated with oligopeptide substrates in vitro.
We developed SPE-MS assays that directly monitor the hydrolysis of the isopeptide amide bond of synthetic N ε -lysine-branched oligopeptides, which mimic post-translational protein modification by Ub/UBLs, catalyzed by isolated recombinant SARS-CoV-2 PL pro , extending the scope of our reported PL pro SPE-MS assays with linear oligopeptides as substrates to N ε -lysine-branched oligopeptide substrates. 47The SPE-MS assays compare favorably to reported spectroscopic-or MS-based DUB assays, because these do not employ substrates that bind to the PL pro S′ sites, 39,40,48,49 and are associated with comparatively high costs of the substrates. 51Due to variations in the catalytic efficiency of different PL pro batches, we thus determined the rank order of substrate preference by quantifying product formation relative to internal standards.
Importantly, our results with N ε -lysine-branched oligopeptides support cellular studies concerning the DUB activities of PL pro , including its reported preference for catalyzing the hydrolysis of isopeptide amide bonds C-terminal to the LRGG motif of ISG15 rather than Ub/Ub 2 and, in particular, NEDD8. 39,40The rank order observed for PL pro -catalyzed hydrolysis of K 193 -branched IRF3-derived oligopeptides mimicking post-Fig.6. Modelled interactions of PL pro with lysine-branched oligopeptides.Views of clustering-derived representative PL pro :oligopeptide complex structures exemplifying the overall binding mode of the oligopeptides and their interactions with PL pro N-terminal and C-terminal to the S4-S1 sites.Peptides are colored as in Fig. 3, with the P1 scissile amide carbons as green spheres; the N-terminus of the UBL-derived fragment and the N-and C-termini of the S′ binding fragment are labelled N, N′, and C′, respectively.PL pro residues interacting via hydrogen bonding with the oligopeptide are in lime (the calculated occurrence is ≥25%; orange dashes; not necessarily present in the frame).PL pro residues within 4 Å of the peptide residues that are calculated to contribute ≥0.5 kcal mol − 1 binding energy, are in white.18), the latter of which was not a substrate of PL pro (Fig. 3a).The results thus imply that the substrate preference of PL pro not only depends on the fold of the substrate, interactions of the substrate at the PL pro S1-S4 sites, 42,65 and on allosteric interactions with PL pro , 42-44 but also on the sequence identity of the region of the substrate which binds to the active site beyond S1-S4.This proposal is further supported by the pronounced effect of amino acid substitutions in the N ε -lysine-branched oligopeptide substrates on PL pro catalysis (Fig. 4), as well as by reported work on the hydrolysis of linear oligopeptides catalyzed by SARS-CoV-2 PL pro and PL pro s of other coronaviruses. 80,81he combined results show that the primary sequence of both the peptide fragment derived from the C-terminus of the UBL and that of the post-translationally modified human protein affects the efficiency of PL pro catalysis, the latter potentially by binding to the S′ sites of PL pro , possibly in a substrate context-dependent manner.Hence, SPE-MS assays with oligopeptides have potential to predict the efficacy with which PL pro catalyzes the hydrolysis of post-translationally modified (human) proteins, although factors other than the primary sequence also contribute to catalysis efficiency (see above).It will be useful to perform studies with the N ε -lysine-branched oligopeptides and isolated recombinant SARS-CoV PL pro and MERS-CoV PL pro , as well as reported variants of SARS-CoV-2 PL pro , to inform on different substrate preferences and on how these differences manifest in altered virulence.Similarly, the substrate preferences of viral proteases other than coronavirus PL pro s, that have been reported to catalyze protein deISG15ylation and/ or deubiquitinylation, [82][83][84] should be examined.
The results indicate that PL pro may catalyze the deISG15ylation of IRF3 regardless of the ISG15ylation site (Fig. 2c); studies with SARS-CoV-2 infected human cells are required to investigate the significance of this observation in a cellular context.Note that the ISG15ylation of IRF3 at the N ε -amino group of K193, K366, and, by implication, K360 may compete with other post-translational modifications, e.g.ubiquitinylation at K193, 54 and/or acetylation 85 and methylation 86 at K366, in a manner potentially affecting PL pro catalysis.SPE-MS assay results with N ε -lysine-branched oligopeptides mimicking the ISG15ylation of 4EHP, the ISG15ylation of which has been proposed to control translation during immune response, 59 suggest that ISG15ylated 4EHP may be a substrate of SARS-CoV-2 PL pro in cells (Fig. 3b).The relevance of the PL pro -catalyzed deISG15ylation of ISG15ylated 4EHP for virulence should thus also be probed in cellular studies.
It is likely that PL pro not only catalyzes the deISG15ylation of IRF3 and 4EHP (Fig. 3), but also of other human proteins. 87][57][58] Note that evidence suggests that residues other than lysine residues may be ISG15ylated, e.g.cysteine residues 88,89 ; thus, the ability of SARS-CoV-2 PL pro to catalyze the deISG15ylation of other residues than lysine should be subject of future work.Considering that protein ubiquitinylation is apparently more widespread than ISG15ylation, it is possible that human ubiquitinylated proteins may be identified which are more efficient substrates than ISG15ylated IRF3 in vitro and in cells.
1][92] PL pro may catalyze the deISG15ylation of ISG15ylated viral proteins to counteract the host innate immune response.Although, to our knowledge, little information is currently available on SARS-CoV-2 protein ISG15ylation, multiple ubiquitinylation sites have been identified in SARS-CoV-2 proteins in proteomic MS studies. 62Our work provides MS evidence that isolated recombinant PL pro catalyzes the hydrolysis of the isopeptide amide bond of an N ε -lysine-branched oligopeptide based on a reported ubiquitinylation site of the SARS-CoV-2 N protein, i.e.SARS-CoV-2_N 335-341 -Ub 23 (Fig. 3c).It is thus possible that PL pro catalyzes the deubiquitinylation of at least some of the (poly)ubiquitinylated lysine residues of viral proteins to e.g.regulate protein function or to prevent protein degradation via the host proteasome, as (poly)ubiquitinylation can be a signal for proteasomal degradation and the removal of the (poly)ubiquitin may hence stabilize the viral protein; conversely, PL pro has potential to also catalyze the deISG15ylation of ISG15ylated viral proteins.
3][94] Interestingly, viral proteases catalyze deISG15ylation reactions at different sites, e.g. the leader protease of foot-and-mouth-disease virus (FMDV) is reported to catalyze the hydrolysis of ISG15 N-terminal to diglycine of the LRGG motif, 82 opposed to SARS-CoV-2 PL pro which catalyzes the hydrolysis of ISG15 C-terminal to diglycine of the LRGG motif.Viral proteases have also been reported to catalyze the direct hydrolysis of the main chain of interferon regulatory factors, [95][96][97] including that of IRF3 by SARS-CoV-2 PL pro , 18 further highlighting the importance of modulating (post-translationally modified) interferon regulatory factors such as IRF3 during viral infections.
At least in most cases, the identity of the substrate residues binding to the S sites affects PL pro catalysis to a relatively greater extent than the identity of the substrate residues binding to the S′ sites (Fig. 2).This observation may reflect the different substrate selectivities of PL pro and M pro .M pro apparently requires the presence of specific residues (i.e., S/ A/N) at the S1′ site for efficient catalysis, though its substrate requirements for S2′-S4′ appear to be less stringent. 98Defining the full substrate scope of M pro is, however, the subject of ongoing investigations. 99,100Notably, NEMO, the M pro -catalyzed hydrolysis of which has been reported to induce the death of brain endothelial cells, 16,17 has a valine residue at the P1′ position, indicating that the presence of (L/F/V)Q(S/A/N) motifs in human proteins may not be sufficient to predict the efficiency of M pro catalysis.
Eleven sites in pp1a/1ab fulfil the apparently somewhat narrower substrate requirements for M pro , whereas only three sites in pp1a/1ab fulfil the substrate requirements for the apparently more promiscuous PL pro . 1,2Human proteases that accept (L/F/V)Q(S/A/N) motifs as substrates are currently unknown, suggesting that M pro may have evolved to exclusively catalyze the release of those nsps from pp1a/1ab that are directly involved in replication and transcription (e.g.nsp12, which has a RNA-dependent RNA polymerase domain, and the nsp13 helicase), under strict spatiotemporal control.By contrast, in principle, human DUBs could catalyze the hydrolysis of pp1a/1ab at three sites, releasing inter alia nsp1 and nsp2 which have been proposed to help the virus evade the host immune system, 101 and thus compete with PL pro for viral substrates.The apparent lower substrate specificity of PL pro may reflect its functions in enabling the virus to counteract the host immune system, with respect to catalyzing both the release of nsp1 and nsp2 and the deISG15ylation of post-translationally modified human and, potentially, viral proteins, a process which might not require strict spatiotemporal control and thus tolerates reduced substrate specificity.
The ability of PL pro , but not M pro (as far as is known), to catalyze hydrolysis of isopeptide as well as peptide amide bonds is striking.The comparison of substrate and non-substrate binding modes for the two proteases is thus of interest, because the precise nature of the dynamic interactions that regulate the different substrate selectivities of PL pro and M pro is not defined.Reported PL pro structures reveal that substrate binding can induce conformational changes of flexible regions proximate to the active site, including the PL pro blocking loop 2 (BL2) which is a key factor in regulating the accessibility of potential substrates to the active site. 70The role of the PL pro BL2 in regulating substrate recognition is precedented by BL2 in, at least some, human DUBs, including e.g.USP7 and USP14; 71,72 however, in human DUBs, post-translational modification of BL2 residues may further alter catalysis. 102It appears that the PL pro substrate selectivity is not only a result of the affinity of the substrate to bind to S sites and, to a lesser extent, S′ sites, but also of the ability of a potential substrate to stabilize the closed conformation of BL2, so enabling a catalytically productive enzyme-substrate conformation which may be achieved via interactions with PL pro that may be relatively remote from the active site, including by specific interactions of hydrophobic substrate residues with the S′ sites (Fig. 6).
Movement of active site bordering loops to bind, enclose, and/or orientate substrates (and reject non-substrates) are common in enzyme catalysis.4][105][106] The presence of a BL2-type loop in PL pro , but not M pro , is interesting.Whether or not the presence of this loop affects the ability of PL pro to accept different types of substrates (peptide and isopeptide amide bonds), whilst manifesting an apparently narrower selectivity with respect to polyprotein pp1a/1ab hydrolysis, is unclear.In this regard, it will be of interest to investigate how the substrate selectivities of PL pro homologues have changed during the course of viral evolution, both with respect to pp1a/1ab and proteins in different host organisms, e.g.SARS-CoV-2 PL pro prefers ISG15 as a substrate, whereas SARS-CoV PL pro prefers diubiquitin. 39,40Notably, the efficiency and selectivity of MERS-CoV PL pro catalysis has been proposed to be distinct from that of SARS-CoV and SARS-CoV-2 PL pro , in part as a result of amino acid variations in the BL2 loop. 48,748][109][110][111][112][113] Our SPE-MS assays employing synthetic N ε -lysinebranched oligopeptides as substrates of isolated recombinant PL pro have enabled studies on the substrate scope and substrate selectivity of PL pro and thus helped to characterize the DUB activity of PL pro .The combination of analogous MS-based assays with synthetic N ε -lysine-branched oligopeptides as substrates of isolated recombinant human DUBs, together with proteomic studies, will enable studies on the substrate preferences of human DUBs, such as USP18 which also catalyzes protein deISG15ylation, and which is a current medicinal chemistry target. 35,45

Production and purification of isolated recombinant SARS-CoV-2 PL pro
The PL pro domain of the SARS-CoV-2 nsp3 (E746-T1063) was produced using E. coli Lemo21(DE3) cells and purified as reported previously. 47

Peptide synthesis
Linear oligopeptides were prepared by solid phase peptide synthesis (SPPS) using a Liberty Blue peptide synthesizer (CEM Microwave Technology Ltd.), as reported for the synthesis of the oligopeptides 1-3. 47he N ε -lysine-branched oligopeptides mimicking posttranslationally modified proteins were synthesized by microwaveassisted SPPS using the Fmoc-protection strategy from the C-to N-terminus on Rink Amide MBHA resin (AGTC Bioproducts Ltd.; loading: 0.6-0.8mmol/g) similar to reported procedures. 47Initially, a linear oligopeptide was synthesized; N-Fmoc lysine with a 4-methyltrityl (Mtt) protected N ε -amine was used at the site of the lysine branching, and an amino acid with a Boc-protected N α -amine was used at the N-terminus.Branching was introduced via selective deprotection of the lysine N ε -Mtt group of the resin-bound peptides using 1% v/v trifluoroacetic acid (TFA) and 2% v/v triisopropylsilane (TIPS) in dichloromethane at ambient temperature while shaking (300 rpm) (5 cycles; after each cycle, the deprotection mixture was removed and the resin was washed with dichloromethane).Following Mtt deprotection, the UBL-derived peptide fragment was synthesized by SPPS from the C-to N-terminus starting with the free N ε -amino group of the resin-bound peptides at the Cterminus.
After completion of the synthesis, the resin-bound peptides were washed with dichloromethane and subsequently cleaved from the resin and simultaneously deprotected using a mixture of trifluoroacetic acid, triisopropylsilane, 1,3-dimethoxybenzene, and water (92.5/2.5/2.5/2.5% v/v , respectively).Solids were separated; the remaining clear solution was diluted with diethyl ether (45 mL/0.1 mmol resin).After incubation for 30 min at 0 • C, the mixture was centrifuged for 10 min using a Beckman Coulter Allegra X-30R centrifuge equipped with a SX4400 rotor (4500 rpm); the supernatant was discarded.The solid residue was dissolved in a water/acetonitrile mixture, frozen using liquid N 2 , and then lyophilized.The dried crude product was dissolved in a water/acetonitrile mixture, filtered, and purified using a semipreparative HPLC machine (Shimadzu UK Ltd.) equipped with a reverse phase column (Gemini 00G-4454-U0-AX; phase: NX-C18).A linear gradient (typically 2-47% v/v over 38 min) of acetonitrile in milli-Q grade water (each containing 0.1% v/v trifluoroacetic acid) was used as eluent.Fractions were analyzed by SPE-MS and those containing the pure peptide were combined and lyophilized.Sequences, mass spectra, and purification characteristics of the synthetic oligopeptides are shown in Supporting Figure S1.

SPE-MS assays
PL pro assays for turnover and competition experiments were performed in 96-well polypropylene assay plates (Greiner), either with a 1.0 or 0.5 mL final reaction volume, using isolated recombinant SARS-CoV-2 PL pro (0.2 μM), substrate peptide(s) (2.0 μM), inert N-acetylated standard peptides (0.2 μM, as indicated in the individual experiments) in buffer (50 mM Tris, pH 8.0) at ambient temperature.PL pro catalysis was directly monitored using SPE-MS. 47The RapidFire RF 365 highthroughput sampling robot used was programmed to aspirate samples from the reaction mixture at the time intervals indicated in the individual experiments.
MS-analyses were performed using a RapidFire RF 365 highthroughput sampling robot (Agilent) attached to an iFunnel Agilent 6550 accurate mass quadrupole time-of-flight (Q-TOF) mass spectrometer operated in the positive ionization mode. 47Assay samples were aspirated under vacuum for 0.6 s and loaded onto a C4 solid phase extraction (SPE) cartridge.After loading, the C4 SPE cartridge was washed with 0.1% v/v aqueous formic acid to remove non-volatile buffer salts (5.5 s, 1.5 mL/min).The peptide was eluted from the SPE cartridge with 0.1% v/v aqueous formic acid in 85/15 v/v acetonitrile/water into the mass spectrometer (5.5 s, 1.25 mL/min) and the SPE cartridge reequilibrated with 0.1% v/v aqueous formic acid (0.5 s, 1.25 mL/min).The mass spectrometer parameters were: capillary voltage: 4000 V; nozzle voltage: 1000 V; fragmentor voltage: 365 V; gas temperature: 280 • C; gas flow: 13 L/min; sheath gas temperature: 350 • C; sheath gas flow: 12 L/min.
For data analysis and to quantify product formation, the charge states of both the C-terminal and N-terminal product peptides and the corresponding N-acetylated C-terminal and N-terminal product peptides (i.e. 5, 12, and 24), which were used as internal standards, were used to extract ion chromatogram data (m/z +1 for 12 and the corresponding product peptide; m/z +2 for 5 and 24, as well as for the corresponding product peptides); peak areas were integrated using the RapidFire Integrator software (Agilent).Data were exported into Microsoft Excel and used to calculate the product peptide concentrations using the equation: peptide concentration = 0.2 μM × (integral C-or N-terminal product peptide) / (integral N-acetylated C-or N-terminal product peptide).

Preparation of PL pro :oligopeptide models
A reported SARS-CoV-2 PL pro structure (PDB ID: 6WX4 65 ) was prepared for modelling as described (Supporting Table S2). 61Based on reported quantum mechanics/molecular mechanics-umbrella sampling (QM/MM-US) calculations on proton transfer processes in the PL pro catalytic triad, 61 C111 was modelled in its deprotonated form, H272 as doubly protonated, and D286 as deprotonated.
The N ε -lysine-branched oligopeptides complexed with PL pro were constructed using reported AutoDock CrankPep (ADCP) 66 -docked conformations of the linear oligopeptide nsp2/3 808-827 (2). 61The residues N-terminal to the scissile amide in the top ranked pose (d2_01, reported nomenclature 61 ), which was successful in placing P4-P1 LKGG in the respective S4-S1 subsites, were used to build the N-terminal ubiquitin (Ub) or Ub-like modifier (UBL) derived fragment of the N ε -lysinebranched oligopeptides.The C-terminal fragment of the N ε -lysinebranched oligopeptides was built based on the 48th-ranked pose of (d2_48), which did not pass through S4-S1, but in which the P4′ lysine sidechain N ε amine was proximate to C111 (Supporting Figure S5).The lysine side chain branching was modelled by linking the C ε atom of the d2_48 P4′ lysine sidechain to the carboxamide N atom of the 2_01 P1′ alanine.To prepare each of the eight branched peptides in IRF3 189-197 -(ISG15/Ub/NEDD8/URM1/SUMO1) and (TMEM59 285-291 /TMEM59 298- 305 /SARS-CoV-2_N 335-341 )-Ub (Fig. 3), the two poses of 2 were modified to the lengths of the N-and C-terminal fragments.The residues were modified to the target sequences using the mutagenesis tool of PyMOL (open source, v. 2.3.0), 114selecting the least sterically clashing backbone-dependent rotamer in each case. 115The N-termini of both fragments were uncapped, while the C-terminus of the IRF3 189-197 / TMEM59 285-291 /TMEM59 298-305 /SARS-CoV-2_N 335-341 fragment was NH 2 -capped.Peptide histidine residues were modelled in their neutral state, with N ε protonated (HIE in AMBER nomenclature), 116 except for IRF3 189-197 -URM1 where both the neutral HIE and the positively charged, doubly protonated (HIP) states of the P3 histidine residue were considered, denoted as URM1 N and URM1 + respectively.
Each of the nine PL pro :oligopeptide complexes (eight unique sequences, with URM1 N and URM1 + for two possible charge states of the P3 histidine group in 17) was centered in a rhombic dodecahedral box with at least 1.0 nm separation from box edges, solvated, neutralized with sodium/chloride ions (110,788-110,847 atoms in total), and minimized until the maximum force was below 1000 kJ mol − 1 nm − 1 .From the minimized system, three replicas were initiated using random velocities at 298.15 K, subjected to 200 ps (1 fs step) restrained NVT equilibration at 298.15 K, followed by 200 ps (1 fs step) NPT equilibration at 298.15 K and 1.0 bar.The equilibrated complexes were subjected to 200 ns production MD (2 fs step).A velocity-rescaling thermostat with a stochastic term (time constant 0.1 ps; protein and non-protein coupled separately) 125 and a Parrinello-Rahman barostat (time constant 2 ps) were used. 126,127Long-range electrostatic interactions were calculated by smooth Particle-mesh Ewald (1 nm cutoff). 128,129Van der Waals interactions were cut off at 1 nm.
The 3 × 200 ns MD trajectories were fitted based on the PL pro backbone and analyzed using GROMACS tools (v 2019.2). 117 A hydrogen bond was defined on observation of a donor-acceptor distance <3.5 Å and hydrogen-donor-acceptor angle <30 • .To obtain representative structures, clustering was performed with a 3 Å RMSD cut-off of the peptide backbone, using the gromos algorithm. 130Per residue decomposition of PL pro :oligopeptide binding energies calculated by molecular mechanics/generalized Born surface area (MM/ GBSA) [131][132][133][134] was performed on frames in 5 ns intervals, using MMPBSA.py (AMBER18) with an ionic strength of 0.15 M, mbondi2 radii, and igb = 5. 124,135,136

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1.PL pro catalysis.(a) PL pro catalyzes substrate hydrolysis C-terminal to LXGG motifs; (b) sequences of the reported SARS-CoV-2 pp1a/1ab-derived linear oligopeptides nsp1/2 169-188 (1), nsp2/3 808-827 (2), and nsp3/4 2753-2771 (3), previously used in SPE-MS assays; PL pro catalyzed the hydrolysis of 2 substantially more efficiently than that of 1 and 3. 47 Residues of the LXGG motif are in red; (c) SARS-CoV-2 PL pro (C111S variant, orange) in complex with ISG15 (PDB ID: 7RBS 44 ), which is composed of a C-and N-terminal Ub-like domain (light and dark green, respectively), that indicates binding to both the PL pro active site via the C-terminus and the C-terminal Ub-like domain of ISG15 and binding to allosteric sites of PL pro via the N-terminal Ub-like domain (dark green).