A desthiobiotin labelled NAD+ analogue to uncover Poly(ADP‐ribose) polymerase 1 protein targets

ADP‐ribosylation is a post‐translational modification catalyzed by the enzyme family of polyadenosine diphosphate (ADP)‐ribose) polymerases (PARPs). This enzymatic process involves the transfer of single or multiple ADP‐ribose molecules onto proteins, utilizing nicotinamide adenine dinucleotide (NAD+) as a substrate. It, thus, plays a pivotal role in regulating various biological processes. Unveiling PARP‐selective protein targets is crucial for a better understanding of their biological functions. Nonetheless, this task proves challenging due to overlapping targets shared among PARP family members. Therefore, we applied the “bump‐and‐hole” strategy to modify the nicotinamide binding site of PARP1 by introducing a hydrophobic pocket (“hole”). This PARP1‐mutant binds an orthogonal NAD+ (Et‐DTB‐NAD+) containing an ethyl group (“bump”) at the nicotinamide moiety. Furthermore, we added a desthiobiotin (DTB) tag directly to the adenosine moiety, enabling affinity enrichment of ADP‐ribosylated proteins. Employing this approach, we successfully identified protein targets modified by PARP1 in cell lysate. This strategy expands the arsenal of chemically modified NAD+ analogs available for studying ADP‐ribosylation, providing a powerful tool to study these critical post‐translational modifications.


Introduction
1b,2] It involves either the addition of a single ADP-ribose unit to a target protein, termed mono(ADP-ribos)ylation (MARylation), or multiple ADP-ribose units, termed poly(ADP-ribos)ylation (PARylation).It is catalyzed by a family of enzymes called poly(ADP-ribose)-polymerase (PARPs, also referred to as ADP-ribose transferases diphtheria toxin-like (ARTDs)), comprising 17 members that use NAD + as a substrate. [3]Despite intensive studies and notable progress, gaining additional knowledge on PARP-specific protein targets is indispensable for a comprehensive understanding of their biological functions.Methods for identifying target proteins based on mass spectrometry combined with affinity purification using antibodies, [4] macro domains, [5] or boronates [6] have been developed. Furthermore, the antibody approach [4] solely enables the detection of proteins with long and branched PAR chains.Thus proteins, modified with shorter PAR chains or MARylated proteins remain undetected.An alternative approach is based on chemically modified NAD + s.These NAD + analogs have small bioorthogonal groups such as alkyne [7] or azide [8] tags for functionalizing or affinity tags such as biotin [9] or desthiobiotin [10] and are useful tools for identifying direct protein targets for ADP-ribosylation.Importantly, these NAD + analogs are not sufficient to identify PARP member-specific targets, because multiple or all PARP enzymes may use these NAD + analogs.Recently, an azidemodified NAD + was reported to show some degree of selectivity for PARP1. [11]Carter-O'Connell et al., [12] Rodriguez et al. [13] and Gibson et al. [14] developed chemical genetics strategies to address these challenges.To identify PARP1specific protein targets, Carter-O'Connell et al. [12a] modified the nicotinamide moiety of NAD + (5-Et-6-a-NAD + , Figure 1A) by introducing an ethyl group ('bump') at the C5 position.Thus, wild-type (wtÀ )PARP1 and other PARPs are incapable of using this NAD + analog as a substrate due to steric hindrance.By mutating amino acid residues in the nicotinamide binding site of PARP1, a hydrophobic pocket was introduced, allowing the PARP1 mutant to process an NAD + analog containing an ethyl group at the C5 position.Furthermore, the NAD + analog was modified with an alkyne tag at the N6 position of the adenine moiety for functionalization with a biotin tag for affinity enrichment.Gibson et al. [14] chose to modify the NAD + with a but-3-yne-1-thioether at C8 of the adenine moiety (8-Bu(3-yne)T-NAD + , Figure 1A).The introduced group at C8 acts as a 'bump' and allows functionalization with a dye or biotin.Both approaches successfully identified PARP1-specific ADP-ribosylation, but rely on the application of copper(I)-catalyzed azidealkyne cycloaddition (CuAAC) [15] for functionalization.
CuAAC is widely used in chemical biology for rapid and efficient conjugation of a substrate to a biomolecule of interest, [16] but has also some limitations. [17]16c,18] The commonly used reducing reagent ascorbate, in its oxidized form, can react with lysines, arginines, [19] and cysteines. [20]dditionally, the alkyne group, particularly when directly attached to electron-withdrawing groups, can undergo side reactions with cysteines presumably via a thiol-yne reaction. [21]oreover, it is noteworthy that CuAAC reactions are concentration dependent (second-order rate constants of 10-200 M À 1 s À 1 [16c] ) and thus fail to trap less abundant substrates.
Here, we report a novel NAD + analog (Et-DTB-NAD + , Figure 1A) that allows affinity enrichment of PARP1-specific targets without the need for CuAAC chemistry.We chose to modify the adenine side of the NAD + analog according to Lehner et al. [10] with a desthiobiotin tag directly attached to the C2 position, since it has been shown by Wallrodt et al. [22] that modifications at C2 are generally better accepted compared to N6.This finding was further corroborated by Langelier et al. [23] who solved the crystal structure of PARP1 bound to benzamide adenine dinucleotide (BAD), a non-hydrolyzable NAD + analog showing that C2 is oriented outward from the ADP-ribose binding pocket. [13]Since the 'bump-and-hole strategy' had previously been applied for PARP1, [12a,14] we decided to use the same enzyme as a proof of principle.We employed the engineered enzyme/substrate pair developed by Carter-O'Connell et al., [12a] because this approach introduced bulky groups on the nicotinamide mononucleotide (NMN) moiety (Figure 1B).Therefore, this strategy facilitated the synthesis of various NAD + analogs with different alkyl or benzyl groups at the C5 position of the nicotinamide moiety and enabled the independent modification of the adenine moiety with a DTB tag for affinity enrichment, making this approach broadly applicable to other PARP mutants.

Result and Discussion
First, we synthesized the NAD + analog Et-DTB-NAD + (Scheme 1).To that end, the β-nicotinamide mononucleotide analog was synthesized according to Carter-O'Connell et al. [12a] and the AMP building block was synthesized in three steps according to Lehner et al. [10] Upon phosphate activation of the AMP analog as a morpholidate, it was coupled with the βnicotinamide mononucleotides analog resulting in Et-DTB-NAD + .Having Et-DTB-NAD + in hand, we proceeded with the in vitro validation using an auto-ADP-ribosylation assay.
We initially expressed KA-PARP1, [12a] a PARP1-mutant, where a lysine residue in the nicotinamide binding site was substituted with a sterically less demanding alanine.KA-PARP1 was selected because it can accommodate an NAD + analog bearing an ethyl group at the C5 position of the nicotinamide Figure 1.A) NAD + analogs applied in a chemical genetics approach for PARP1 target identification.12a] B) Schematic overview of the bump-hole-strategy using the engineered enzyme-substrate pair KA-PARP1 and ET-DTB-NAD + .12a] Next, we performed ADP-ribosylation assays with Et-DTB-NAD + and KA-PARP1 using wt-PARP1 and DTB-NAD + [10] and natural NAD + as controls.DTB-NAD + is an analog previously synthesized by our group that exclusively bears a DTB tag on the adenine moiety (Figure S1).ADP-ribosylation was detected via western blot with ExtrAvidin®-peroxidase binding the DTB-tag.Figure 2A illustrates the results, indicating high selectivity of Et-DTB-NAD + for KA-PARP1 over wt-PARP1 with Et-NAD-DTB + showing excellent acceptance by KA-PARP1, and only little product formation detected for the wild-type enzyme.It is noteworthy that wt-PARP1 produced poly-ADPribose (PAR), leading to heterogeneous product formation in the presence of DTB-NAD + .This resulted in a shift of PARP1 towards higher molecular weights (Figure 2A, line 4).12a] Additionally, the modified KA-PARP1 activity has the benefit of simplifying sample complexity for LC-MS/MS analysis [12a] allowing direct sample submission without additional pre-treatments like PARG treatment for MARylated protein isolation proteins [10] To evaluate enzyme specificity in the presence of other PARPs, we investigated if Et-DTB-NAD + is accepted by other PARPs in vitro.Therefore, the ADP-ribosylation activity of PARP2, 3, 5a, and 5b were investigated and compared to KA-PARP1.Little to no ADP-ribosylation was observed for PARP2, 3 and 5a in the presence of Et-DTB-NAD + compared to DTB-NAD + .(Figure 2B, line 3, line 5, line 7) demonstrating the high specificity of Et-DTB-NAD + for KA-PARP1.Only little product formation was observed for PARP5b after incubation with Et-DTB-NAD + , however, the intensity was lower than the band observed for KA-PARP1.
After successfully validating Et-DTB-NAD + in vitro, we applied the tool to determine the direct protein targets of PARP1 liquid chromatography-tandem mass spectrometry (LC-MS/MS).First, KA-PARP1 was expressed in KO-PARP1 HeLa cells [24] followed by cell lysis and incubation with Et-DTB-NAD + for the ADP-ribosylation of proteins.Figure 3A presents a Scheme 1. Synthetic scheme for the synthesis of Et-DTB-NAD + .A) 2 was synthesized according to Lehner et al. [10] in 3 steps.B) Reaction: conditions: PPh 3 , morpholine, 2,2-dithiodipyridne, DMSO, 1.5 h, 79 %.C) 5 was synthesized according to Carter-O'Connell et al. [12a] in 5 steps.D) Reaction conditions: Compound 3, MgSO 4 , MnCl 2 , formamide, 2 d, 32 %.schematic workflow.As controls, we also analyzed cell lysate without Et-DTB-NAD + , cell lysate without KA-PARP1, and cell lysate without KA-PARP1 and Et-DTB-NAD + to ensure that the identified proteins are KA-PARP1 protein targets.Then, ADPribosylated proteins were immobilized using Streptavidin beads, washed, and proteins bound to the beads were competitively eluted with biotin.Eluted proteins were separated by SDS-PAGE and detected by Western blotting against the DTB tag (Fig- ure 3B).Notably, labeling was solely detected when Et-DTB-NAD + and KA-PARP1 were present, whereas no labeling was observed without the presence of either Et-DTB-NAD + or KA-PARP1.The remaining eluted proteins were proteolyzed by trypsinization and analyzed by LC-MS/MS.
Label-free quantification (LFQ) was performed using Max Quant [25] resulting in the identification of 995 proteins (4 measurements per condition, 2 biological replicates and 2 technical replicates).Subsequently, Perseus [26] was used for further analysis resulting in the identification of 437 proteins postfiltering (i.e. proteins had to be reliably identified and quantified in at least 2 out of 4 measurements).Differences in protein enrichment among conditions were assessed using analysis of variance (ANOVA) as multiple-sample test (FDR � 0.05, S0 = 0.1) followed by pairwise comparison to negative controls via Tukey's honestly significant difference (HSD) test (FDR � 0.05).This yielded 45 significantly enriched proteins that are likely ADP-ribosylated by KA-PARP1.The heatmap with the significantly enriched proteins is shown in Figure 3C.To evaluate the identified proteins, we conducted a functional analysis using the DAVID app (Database for Annotation Visualization and Integrated Discovery). [27]5b,10, 28] These findings also support a biological relevance for the identified proteins.1b] Furthermore, some of the proteins found have already been shown to be ADPribosylated by PARP1 in vitro, such as TOP1 [29] and MPG [5b] demonstrating that specific protein targets were identified with this approach.Since our primary objective was to identify specific protein targets for PARP1, we performed a comparison of our data to the bump-hole strategies (Figure 4) reported previously.First, we compared our identified proteins with those identified by Carter-O'Connell et al., [12a] who used the same PARP1 mutant (KA-PARP1) together with 5-Et-6-a-NAD + .Our approach yielded a similar number of proteins (45 proteins in this work vs. 42 proteins by Carter-O'Connell et al.), with a subset of 9 proteins (MTHDF2, WARS2, RPA1, SSBP1, TMPO, EIF5B, ABCF1, XRCC6, PARP1) present in both studies, representing an overlap of 20 %.A comparison with Gibson et al. [14] revealed a subset of 13 proteins (TMPO, EIF5B, ABCF1, XRCC6, PARP1, HLTF, GTF2I, DNMT1, PPP1R10, UHRF1, MPG, VCP, TOP1).Since Gibson and colleagues provided only the total number of proteins found for PARP1 without specifying which proteins were considered as a hit, we only compared the proteins identified in this work to the 395 most abundant proteins in their paper (cut-off was made for proteins identified with less than ten peptides) similar to a review article recently published. [30]If all three studies are taken into account, only a subset of 5 proteins could be identified namely TMPO, EIF5B, ABCF1, XRCC6, and PARP1.
One potential reason for the different results between these three publications may be that Carter-O'Connell et al. [12a] used HEK cell lysate, whereas Gibson et al. [14] and the present study used HeLa cell lysate, which may result in a different target pattern.Furthermore, prior studies utilized a recombinantly expressed PARP1 mutant in the lysate, while our method involved transiently expressed KA-PARP1 in HeLa cells.This approach may improve target specificity in situations where eukaryotic posttranslational modifications are required for particular protein interactions or if the recombinantly expressed protein folds incorrectly. [31]n addition, the positioning of the affinity tag varied in all three approaches, potentially resulting in differences in the acceptance of the NAD + molecules by PARPs.This observation has previously been reported by Wallrodt et al. [22] who found that modifications at the C2 position of the adenine moiety were better accepted than those at the N6 position.Finally, it is important to highlight the advantages of evading click chemistry.It increases sensitivity in detecting low-abundant  [12a] (black) and Gibson et al. [14] (grey) Off note: Proteins targets of Gibson et al. for PARP1 were used similar to the review of Rodriguez et al. [30]  proteins and prevents side reactions with amino acids.Additionally, our newly described method facilitates the procedure and therefore reduces the risk of errors.

Conclusions
In conclusion, we synthesized the novel NAD + molecule Et-DTB-NAD + and used it in chemical proteomics which enabled the identification of PARP1-specific targets without the necessity for click chemistry.This study provides additional insights into the interaction network of PARP1 and confirms its role in DNA damage repair.Consequently, the desthiobiotin-modified NAD + offers a valuable tool for pinpointing PARP-specific protein targets.It should also be combined easily with other PARP mutants and will therefore represent an important step in elucidating the biological role of PARP1 and other PARP members.

Figure 3 .
Figure 3. A) Identified proteins in the affinity enrichment experiment using KA-PARP1 and Et-DTB-NAD + .A) Schematic overview of the workflow.Et-DTB-NAD + was incubated with cell lysate containing KA-PARP1 resulting in ADP-ribosylation of proteins.KA-PARP1 targets were then isolated via affinity enrichment, digested, and analyzed via LC-MS/MS.B) Western blot visualization of the elution fraction of the affinity enrichment using ExtrAvidin®-peroxidase.As controls lysates containing no KA-PARP1, no Et-DTB-NAD + and no PARP1-KA and Et-DTB-NAD + were used.C) Heat map representation of identified KA-PARP1 protein targets.Each row represents one protein (gene names).Applied conditions are indicated in the respective column.Identified proteins were quantified in at least 2 out of 4 measurements per condition.Significantly bait enriched proteins were analyzed by an ANOVA-based multiple-sample test (false discovery rate (FDR) � 0.05, S0 = 0.1), followed by post hoc Tukey's honestly significant difference (HSD) test (FDR � 0.05).(D) Functional annotations for biological process and cellular component of the significant protein hits using DAVID.

Figure 4 .
Figure 4. Comparison of proteins found in this experiment compared to literature.A) Venn diagram showing the overlap of proteins found in this publication (green) vs. Carter-O'Connell et al.[12a]  (black) and Gibson et al.[14] (grey) Off note: Proteins targets of Gibson et al. for PARP1 were used similar to the review of Rodriguez et al.[30] B) Table showing protein hits (gene names) in this publication and Carter-O'Connell et al. (row 1), this publication, and Gibson et al. (row 2) and proteins found in all 3 publications (row 3).
Figure 4. Comparison of proteins found in this experiment compared to literature.A) Venn diagram showing the overlap of proteins found in this publication (green) vs. Carter-O'Connell et al.[12a]  (black) and Gibson et al.[14] (grey) Off note: Proteins targets of Gibson et al. for PARP1 were used similar to the review of Rodriguez et al.[30] B) Table showing protein hits (gene names) in this publication and Carter-O'Connell et al. (row 1), this publication, and Gibson et al. (row 2) and proteins found in all 3 publications (row 3).