Development of ADPribosyl Ubiquitin Analogues to Study Enzymes Involved in Legionella Infection

Abstract Legionnaires’ disease is caused by infection with the intracellularly replicating Gram‐negative bacterium Legionella pneumophila. This pathogen uses an unconventional way of ubiquitinating host proteins by generating a phosphoribosyl linkage between substrate proteins and ubiquitin by making use of an ADPribosylated ubiquitin (UbADPr) intermediate. The family of SidE effector enzymes that catalyze this reaction is counteracted by Legionella hydrolases, which are called Dups. This unusual ubiquitination process is important for Legionella proliferation and understanding these processes on a molecular level might prove invaluable in finding new treatments. Herein, a modular approach is used for the synthesis of triazole‐linked UbADPr, and analogues thereof, and their affinity towards the hydrolase DupA is determined and hydrolysis rates are compared to natively linked UbADPr. The inhibitory effects of modified Ub on the canonical eukaryotic E1‐enzyme Uba1 are investigated and rationalized in the context of a high‐resolution crystal structure reported herein. Finally, it is shown that synthetic UbADPr analogues can be used to effectively pull‐down overexpressed DupA from cell lysate.


Introduction
The dogma in post-translationalm odification by ubiquitin (Ub) is that Ub-activatinge nzymes (E1), Ub-conjugating enzymes (E2), and Ub ligases( E3) are required to work together to activate the C-terminal carboxylate of Ub,i na na denosine triphosphate (ATP)-dependent manner,a nd subsequently ligate it to predominantly the e-amino group of al ysine in as ubstrate protein. Discovery of ac lass of Legionella pneumophila effector proteins that can conjugate Ub to substrate proteins,i ndependento ft he canonical machinery andw ithout the need for ATP, has gained much interest. [1] These multidomainb acterial enzymes are able to ADP-ribosylate the d-guanidinium group of arginine 42 (Arg42) of Ub at the expenseo fn icotinamide adenine dinucleotide (NAD + )b yu sing their mono-ADP-transferase (mART)d omain in the first step, followed by the action of their phosphodiesterase( PDE) domain, whichc atalyzes the transfer of phosphoribose-Ub (Ub Pr )t ot he serine of asubstrate protein, while expelling adenosine monophosphate (AMP; Figure1). [2] Legionella has its own regulatory mechanism in place to control the temporal activityo ft hese SidEl igasesb y blockingt heir active-site glutamateu sing the glutamylase SidJ. [3] The recently identified deubiquitinases for phosphoribosyl ubiquitination (Dups), DupA and DupB, also known as LaiE and LaiF,c ounteract the SidE-mediated attachment of phosphoribosyl-linked Ub to substrates. [4] DupA and DupB were identified on the basis of their structuralh omology to the SidE PDE domains,b ut lack the ability to Pr-ubiquitinate the substrate protein Rab33bu pon incubation with Arg42 Ub ADPr .T hese DUPs,h owever,w eres hown to release proteins that were Pr-ubiquitinated by SidE ligasesb yc leaving the phosphodiester bond between the substrate serinea nd Arg42 Ub Pr . [4a] Although SidE ligases and Dups have oppositef unctions, they are structurally very similar and, even more so, the ligase SdeA is shown to mediate hydrolysis of the pyrophosphateb ond in Ub ADPr if no suitable substrate protein is present.T he ligase effectively mediates transfer of aw ater moleculei nstead of a serine residue to the activated pyrophosphate bond, thereby expelling AMP. [1b] By using ac atalytically inactive version of DupA to enrich for Pr-ubiquitinated substrates in HEK293T cells infected with Legionella,1 80 host proteins were identified based on their affinity for DupA. [4a] Most of these proteins are involved in endoplasmicr eticulum membrane recruitment to Legionella-containing vacuoles (LCVs). This highlights the importance of Prubiquitination upon Legionella infection because maintaining LCV integrity is essential for Legionella proliferationa nd the onset of Legionnaires' disease.
In the canonical ubiquitination pathway,t he use of chemically preparedt ools, such as substrate reagents and activitybased probes, has been aw idely applied and successful approach to allow the study of kinetic parameters, as well as capturing and identifying both ligasesa nd proteases. [5] The recent development of fluorescent polarization based assay reagents and inhibitors to study enzymesi nvolved in the Pr-ubiquitination pathway highlights the applicability of chemically synthesized tools to study  Hence, the construction of probes targeting the ADPr-mediated ubiquitination machinery will be as imilarly useful asset in studying the enzymes involved. We set out to prepare a-O-propargyl ADPr (1; Figure 2) and its stabilized methylene bisphosphonate analogue, a-O-propargyl me-ADPr (2), in whicho xygen in the pyrophosphate linkagei sr eplaced with am ethylene group. [7] Facile copper-catalyzed Huisgen azide-to-alkyne 1,3-dipolar cycloaddition (CuAAC, or click reaction) of these propargyl-con-tainingA DPr analogues to azide-modified Ub allowed the gen-  eration of probes 4 and 5 to investigate Legionella enzyme activity.T he rationale behind the oxygen-to-carbon substitution in 5 is that the PDE activity in SidE enzymes relies on expelling AMP.R eplacingt he diphosphate with am ethylene bisphosphonate prevents this step from occurring, thereby blockingS idE-mediatedc onjugation to substrate proteins. [8] This stabilized Ub me-ADPr conjugate 5 would thus be able to capture the Legionella enzyme and function asasuitable nonhydrolyzable probe to target such enzymes.A dditionally,l ittle is known about the role of the phosphoribosyl residue that remains on the Ub moiety after Dup-mediated hydrolysis of the targeted substrate protein, and we envision Ub Pr -based tools, such as 6,t obee ssential to decipher the role of Ub Pr .

Results and Discussion
The inherenti ncompatibility of ADPr and other nucleotidebased structures with strongly acidic conditions routinelyu sed in fluorenylmethoxycarbonyl (Fmoc)/tert-butyloxycarbonyl (Boc) solid-phase peptide synthesis (SPPS) prohibits the total chemicals ynthesiso fl argeA DPr peptides or proteins and only allows for the construction of relatively short ADPr peptides by adaptingp rotecting-group schemes. [9] This has triggered the development of modulars ynthetic approaches towards such structures, [7a, 10] in which the polypeptide can be treated with a strong acid to remove protecting groups and be released from ap eptides ynthesis resin followed by HPLC purification,b efore it is attached to the delicate ADPr moiety.T oa llow this final conjugation step to be executed under mild conditions, we envision click chemistry to be the most effective strategy. [7a] Upon substituting Arg42 of Ub with azidohomoalanine through SPPS, conjugation can be achieveda tp hysiological pH with am inimum of chemical additives (3 mm CuSO 4 , 20 mm sodium ascorbate, and 3mm tris[(1-benzyl-4-triazolyl)methyl]amine (TBTA) ligand) to the a-oriented propargyl ether on the anomericp osition of the riboside in ADPr (1),m e-ADPr (2),o rP r( 3) ( Figure 2A). The Ub ADPr conjugate formed in such aC uAAC reaction carries at riazole linkage between the ribose and peptidep art, from now on referredt oa s triazole Ub ADPr ,t hus slightly deviating from the native arginine guanidinium linkage ( Figure 2B).
After the successful CuAAC reactions of 1, 2,a nd 3 to Ub carrying an azidohomoalanine mutation on position 42, triazole-linked triazole42 Ub ADPr (4), triazole42 Ub me-ADPr (5), and triazole42 Ub Pr (6)w ere obtained.W es et out to compare these triazole-linked conjugates,a nd nativelyl inked Arg42 Ub ADPr ,w hich was prepared enzymatically by using aS deA mutant, for their affinity towards DupA. [1b] To this end, we used biolayer interferometry (BLI), and repeated the assay that was described earlier,b yi mmobilizing the different Ub analogues on streptavidin (SA) biosensor tips through the biotin handle attached on the Nt erminus of Ub, and using glutathione S-transferase (GST)-tagged DupA-H67A as the analyte. With this setup,c onjugates 4 and 5 show very high affinitieso f1 1.2 and 10.6 nm,r espectively,w hich are comparable to the K d value of 5.7 nm observed for native Arg42 Ub ADPr ( Figure S3 Ai nt he Supporting Information). [4a] However,t he observed nanomolar affinity for unmodified Ub (54.5 nm) would render all DupA inside ah uman cell bound to unmodified Ub (product-like) and unavailable for catalysis. We repeated the experiment with the catalytically inactive mutant, DupA-H67A, lacking the GST tag ( Figure 3A). The results obtained show biologically plausible K d values of 2.2 and 1.2 mm for 4 and 5,r espectively,w hereas 6 and unmodified Ub have at least a1 5-fold reduced affinity (Figure3Ba nd Figure S3 Bi n the Supporting Information). The discrepancy betweent he two assays couldpotentially be attributed to an artefact arising from dimerization of the GST-taggeda nalyte that we cannot fully explain at this point (see Figure S3 Di nt he Supporting Information).
The resulting K d values in the absence of the GST tag make biological sensea nd would fit with the mechanism of the hydrolase, which accepts substrates linked through ap hosphodiesterb ond to ribosylated Ub, with micromolar affinity,a nd releases the phosphomonoester Ub Pr product due to the lower affinity of the latter.
Next, we wondered whether DupA could hydrolyze 4 to form triazole Ub Pr ,a sr eported previously for native Arg42 Ub ADPr . [4a] We indeed observed robust hydrolysis of nativelyl inked Arg42 Ub ADPr (1 mm)b y5 00 nm DupA after incubation for 1h at 37 8C( Figure 4A). Upon applying the same conditions to triazole-linked 4,w eo bserved as imilar hydrolysis reaction and formation of phosphoribosyl Ub 6,a sm onitored by meanso f mass spectrometry ( Figure 4B), whereas DupA was not able to mediate hydrolysis of stabilized 5 (see Figure S4 in the Supporting Information). In control experiments on both native Arg42 Ub ADPr and triazole-linked 4 in the absence of DupA, only a minor amount of hydrolysis of the pyrophosphate bond is observed, which is most likely due to the acidic conditions employed duringm ass spectrometry.D upA-mediated hydrolysis can be attributed to the catalytic specificity of the enzyme because control experiments with triazole-linked triazole54 Ub ADPr , triazole72 Ub ADPr ,a nd triazole74 Ub ADPr showedn either hydrolysis nor formation of the corresponding Ub Pr s. To investigate this further,w ea ssessed thesec ontrol compounds for DupA affinity using our BLI setup ( Figure S3 Ci nt he SupportingI nformation). We could not detects ignificant binding of triazole54 Ub ADPr or triazole74 Ub ADPr to DupA H67A, giving ac lue to why they are not processed by DupA. For triazole72 Ub ADPr ,h owever,w ec ould detect bindingt oD upA H67A with a K d of 9.3 mm,s uggesting that the adenosine moiety could be positioned in am anner resembling the configuration present in 4,b ut so that the diphosphate linkage is not oriented appropriately for hydrolysis towards triazole72 Ub Pr .T oi nvestigatea ny differences in catalysis of DupA on 4 or native Arg42 Ub ADPr ,wefollowed DupA-mediated Ub Pr formationo ver time by mass spectrometry using al ower enzyme concentration of DupA (30 nm)o n3mm of both hydrolyzable substrates. We observed ac lear reduction in velocity (3.5-fold), when comparing relative V max for DupA-mediated hydrolysiso ft riazole-linked 4 to that of native Arg42 Ub ADPr (Figure 4C). It is apparent that, although accepted by DupA, triazole-linked 4 is hydrolyzed at ar educed rate relative to that of native Arg42 Ub ADPr .M ost likely,t his reduced cleavage rate is caused by the more sterically demanding and rigid triazole linkage.
ADPribosylation or phosphoribosylationo fA rg42 in Ub impairs the conventionalu biquitination machinery because activation by E1, trans-thioesterification to E2, andE 3-mediated discharge from the E2 were shown to be compromised upon the introduction of the modification by Legionella ligase SdeA. [1b] From the crystal structure of Arg42 Ub Pr ,i tb ecomes apparentt hat any modification of Arg42 or Arg72 will interfere with Ub bindingt oE 1, which could explain the inability of E1 to activate the Ub Pr molecule. [1b] These two arginine residues are reported to be critical in the interaction with the E1 enzyme Uba1, as in ap reviouss tudy mutations of Arg42 or Arg72 to leucine were shown to result in ad ramatically lower affinity between the E1 enzyme and Ub adenylate. [11] In addition, residue 72 is crucial for determining Ub-like specific rec-ognition by E1, where for Ub this residue is an arginine, for Nedd8i ti sa na lanine, and for SUMO-family members it is either ag lutamate or glutamine residue. [12] We managed to improve the resolution of our previously reported X-ray structure of Saccharomyces cerevisiae Uba1 in complex with Ub from crystalsd iffractinga nisotropically to 2.03 (Figure5A), which showst he C-terminal tail of Ub reaching towards the adenylation site of Uba1. This yeast homolog of Uba1 has ac onserved overall structure with high sequence identity (68 %) in the active adenylation domain compared to human Uba1. [13] Figure 5A shows the crossover loop connecting the adenylation domain to the catalytic cysteine domaine ncompassing the Cterminal tail of Ub just above the Arg42 and Arg72 guanidinium groups of Ub. The close spatial positioning of these residues could explain our observation that Ub ADPribosylated at Arg72 can still bind to DupA.
Furthermore,w eo bserve aw eak electron density for the guanidinium groups of Arg54 and Arg74 in this structure, indicating flexibility andt he possibility fort hese residues to adopt  multiple conformations. Both the s-weighted 2 F o ÀF c electron density map and the B factors of the guanidinium groups of the arginine residues suggest that Arg42 and Arg72 remaini n am ore rigid conformation,a sp art of the bindingi nterface with the Uba1 adenylation domain, compared to Arg74 and Arg54.T he B factors of the CZ atom of the guanidiumg roups of Arg42, Arg54, Arg72,and Arg74 of Ub are 38. 5, 63.7, 30, and 55.9 2 ,r espectively.T he guanidinium groups of Arg42 and Arg72 show well-defined electron densities, indicative of their fixed placement in as ingle conformation, necessary for binding to the adenylation domain of Uba1. To validate whether the triazole-linked Ub ADPr analoguesw ould interferew ithU ba1mediated activation of Ub, we incubated triazole42 Ub me-ADPr 5, triazole54 Ub me-ADPr , triazole72 Ub me-ADPr , triazole74 Ub me-ADPr , triazole42 Ub Pr 6, triazole54 Ub Pr , triazole72 Ub Pr ,a nd triazole74 Ub Pr with human Uba1 (E1) in the presence of sodium2 -sulfanylethanesulfonate (MESNa) and ATP, and monitored thioester formation using mass spectrometry ( Figure 5B). It became apparent that both me-ADPr and Pr modification of positions 72 and 42 completely abolished formation of the Ub-Gly76-MESNa thioester,w hereas the same modifications at positions 54 and 74 had no effect since efficient thioester formationw as observed. When using Arg-to-azidohomoalanine Ub mutants, the precursorsu sed for click chemistry,a ll azido-containing mutants were accepteda nd processed by the E1 enzymet of orm Ub-MESNa thioesters (bars labeled N 3 in Figure 5C). Notably,t he Arg72-Aham utant was significantly slower and Arg42-Aha was moderately slower in thioester formation as only 43(AE 1.6) %a nd 96(AE 0.1) %, respectively,o ft he thioester was formedi nt he same time frame that the 54 and 74 mutantsn eededt or each full conversion. Upon longer incubation, the 72 mutant also reached complete conversion to the MESNa thioester( see Figure S5 in the Supporting Information). It becamea pparent that changes in the chemicalp roperties of the Arg42 and Arg72 guanidinium groups were tolerated since changing them to azides only slowed down E1-mediated thioester formation, butd id not completely abolish the activity.T he introduction of the larger Pr or me-ADPr modification on either Arg42 or Arg72, however,d oes lead to ac omplete loss of thioester formation. This can potentially be explained by the stericb ulk of the modification clashing with the E1 crossover loop and/or the negative charge present on the modification, which might play ar ole in electrostatic repulsion by the negatively charged pocket of E1 that normally accommodates the positively charged Arg72 guanidinium group of Ub. [12] These results again reflected similar behavior of native arginine-linked Arg42 Ub ADPr and the modified Ub analogues,w hich, although carrying the triazole linkage, show ac omparable affinity and biochemical functioning. We were eager to see if these tools could indeed be used as probes. For this purpose, we decided to test whether DupA-WT could be pulled down from cell lysates using biotinylated 5 as bait. HEK293T cells were transfected with mCherryo r mCherry-DupA-WT and lysates were prepared to perform a pull-downe xperiment under nondenaturing conditions. mCherryi sa no ptimized fluorescent protein tag that allows for visualization of the tagged protein of interesti nc ells, as well as their pull-down and visualization in Western blottingw ith the anti-mCherry antibody. [14] It is important to bear in mind that, compared with other Ub activity-basedp robes, in which ac ovalentc omplex is formed upon action of the targeted enzyme on the probe (e.g.,aDUB capturing aU b-VS, Ub-VME, or Ub-Prg probe by means of cysteine catalysis), [15] in our case, the interaction with biotinylated 5 does not lead to ac ovalent complex and solely relies on its intrinsic affinity.W ef ound that probe 5 was able to bind ande nrich DupA-WT ( Figure 6, lane 4), whereas controls with either mCherry( lane 3) or anonspecific interaction with the SA beads (lane 2) showedn oo r only minimal DupAr ecovery,r espectively.S imilarly,p ull-down with biotin-Ub only showed marginal enrichment for DupA ( Figure 6, lane 6) to ac omparable extentt ot hat in the beadsonly control. We repeated this experiment with as light excess of biotin-Ub and quantified theser esultsu sing densitometry, showing > 10-folde nrichment of mCherry-DupA recovery by biotinylated 5 compared with biotinylated wild-type Ub or beads ( Figure 6). We then performed pull-downs from HEKT293T cell lysate using nonhydrolyzable probes 5 and triazole72 Ub me-ADPr and subjected the interacting proteins to trypsin digestiona nd MS/MS analysistoc omparet heir interactome versusn ative Ub ( Figure S7 in the Supporting Information).
Intriguingly,b oth sites of ADPribosylation lead to increased interaction with distinct proteins compared with that of unmodified Ub,s uch as the Ub ligase MYCBP for triazole72 Ub me-ADPr and Ub ligase TRIM28 for 5,a sw ell as the deubiquitinating enzymeO TUD4 for 5.Adecreased interaction with the deubiquitinating enzymeU SP5 is observed for both sites of modification in comparison with unmodified Ub. The change of interaction partners for Ub me-ADPr contains, among others, deubiquitinatinge nzymes,U bl igases, and proteins involved in intracellular (endosomal) trafficking or endoplasmic reticulum-Golgi maintenance. These initial results need further validation and are aw orthy subjecto ff urther research, to define the underlying cellular pathways whereinU bA DPribosylationp lays ar ole.

Conclusion
The preparation of ADP-ribose, adenosine methylenebisphosphonate ribose, and phosphoribose carrying an a-oriented alkyne on the anomeric positiona llowed us to conjugate azidohomoalanine-modified biotin-Ub through CuAACcycloadditions on all four arginine positions in Ub (42,54,72,and 74). This modular approach ensured the construction of Ub ADPr analogues that were used to study ad eubiquitinating enzyme from the Legionella bacterium and am ammalian canonical Ub activating enzyme, the activity of which was shown to be affected by modifications of Ub caused by Legionella infection. We found that the Legionella effector enzyme DupA had a high affinity for chemically prepared triazole-linked 4,w hich was comparable to that of natively linked Arg42 Ub ADPr ,a lthough hydrolysise xperiments showed that the rate of cleavage was reducedf or triazole-linked Ub ADPr .W ef urthermore demonstrate that DupA was as ite-specific hydrolase since 54-, 72-, and 74-Ub ADPr weren ot converted into the corresponding Ub Pr s. The thioester-forming activity at the C-terminal Gly76 of Ub by Uba1 was fully abrogated if positions 42 or 72 carriedm e-ADPr or Pr modifications, buto nly reducedi ns peed if azidohomoalaninew as introduced at those positions. Neitherm e-ADPr nor Pr modifications at positions 54 or 74 had any influence on the E1-mediated reaction, whereas positions 42 and 72 were found to be critical. These experiments, in combination with detailedi nsights from the high-resolution structure, further establishedt hat, by modifying Arg42, Legionella was able to block activation of Ub mediated by the canonical ubiquitination cascade.I na ddition, the affinity of 5 for DupA allowed such tools to be used as an oncovalent probe to enrich overexpressed mCherry-tagged DupA from HEK293T cell lysates. Because structural homology betweenb acterial enzymes is often poor and similarity searches have not yet identified any other bacterial or mammalian enzymes involved in the Pr-ubiquitination pathway,w ee nvision the molecular tools prepared herein to be of great value in answering, in the near future, the question whethert his unusuall igase and hydrolase machinery plays arole in other organisms besides Legionella.