Poly(ADP-ribose): the structure and the synthesis

Poly(ADP-ribose) belongs to a momentous post-translational modification of eukaryotic nuclear protein. Protein poly(ADP-ribosyl)ation mediates numerous cell signalings and is related to some diseases. Poly(ADP-ribose) is currently available through enzymatic or chemical synthesis. Herein, we summarize the structure characteristics and recent developments in the synthesis of poly(ADP-ribose) from a chemical point of view.


Introduction
Post-translational modification is very important for the functions of proteins. However, abnormal modifications of proteins will lead to serious cell dysfunction. Alzheimer's disease (AD) and Parkinson's disease (PD) are considered as two common neurodegenerative diseases, which are generally caused by protein aggregates. 1 Some investigations have found that post-translational modifications are able to regulate protein aggregation processes, such as O-GlcNAcylation of AD-associated tau protein at Ser356 2 and phosphorylation of PD-associated alpha-synuclein (α-Syn) at Ser129. 3 Aberrant truncated glycosylation of tumor-associated mucin-1 (MUC1) caused by glycosyltransferases with abnormal activity induces MUC1-specific immune responses, showing the potential application in antitumor MUC1 vaccines. 4,5 Poly(adenosine diphosphate ribose) (abbreviated as Poly(ADP-ribose) or PAR), which can covalently bind to amino acid residues (such as Glu, Asp, Arg and Lys), is also a distinctive post-translational modification of eukaryotic nuclear protein. [6][7][8][9] PAR regulates extensive cellular processes, such as ATP-dependent energy metabolism, 10, 11 chromatin remodelling, [12][13][14][15][16] DNA-damage repair, 15 cell death, 17 gene expression 18 and cell division. 19 Protein PARylation, in mechanism, influences protein-protein interactions, which modulates enzyme activity and thus may be related to some diseases. [20][21][22][23] Mono(ADP-ribose)(MAR) or PAR is biosynthesized both in vivo and in vitro under the catalysis of PAR polymerase family (PARPs) which includes 17 enzyme members. Since intracellular protein PARylation initiates, length of PAR increases in the main chain direction by adding ADP one by one from nicotinamide adenine dinucleotide (NAD + ). [6][7][8][9][10] In addition, PARylation of DNA at 5' terminal phosphate group improves its stability against phosphatases. Modification with MAR at 3' hydroxyl group and 3' terminal phosphate group of DNA was also reported to produce a ribose-ribose glycosidic bond and a phosphodiester bond, respectively. 24 As mentioned above, PAR widely participates in biochemical, physiological and pathological processes. Based on these researches, tri(ADP-ribose), a chemically synthesized PAR-mimic molecule, is a more potent effector than di(ADP-ribose) for Amplified in Liver Cancer 1 (ALC1) activation with a nanomolar affinity. Tri(ADP-ribose) was found to abolish the macrodomain-ATPase interaction and to recover the activity of remodeler ALC1, indicating the significance of PAR synthesis in vitro. 16 Figure 3. Hydrolysis of MARylated or PARylated proteins. PARG has both exo-and endo-glycohydrolase activities, which cleaves ribose-ribose α(2',1'')-glycosidic bond in the terminal position of PAR or glycosidic bond closed to acceptor, producing ADPR units and intact free form PARs, respectively (A). There are many identified enzymes (B) that hydrolyze different bonds of MARylated or PARylated proteins. Colored arrows represent the enzymatic hydrolysis sites.

Poly(ADP-ribose) recognition of protein
It is now well known that PAR influences many life processes of cell from two aspects. Firstly, PAR can be covalently modified onto nuclear proteins by linking with carboxylic or other nucleophilic amino acid residues. [6][7][8][9] Secondly, free PAR released from PARylated proteins, catalyzed by PARG, conversely regulates functions of acceptor proteins and non-covalently interacts with other proteins, 44,[51][52][53][54][55] resulting in the balance of PARylated protein/PAR ratio. However, disequilibrium of PARylated protein/PAR ratio may cause cell death. 49 Some reports have revealed that the non-covalent interaction between PAR and proteins is in a PAR chain length-dependent manner. 56 It is obvious that basic and hydrophobic amino acid residues are key factors for proteins to non-covalently bind to PAR. 57 For the non-covalent PAR binding, several PAR recognition regions of protein have been found, including PAR-binding zinc finger (PBZ), PAR-binding motif (PBM), Trp-Trp-Glu (WWE) domain and macrodomain ( Figure  4). 9,47,49,58,59 These regions are most likely to bind PAR through electrostatic interactions between the positively charged domains of acceptor proteins and the negatively charged phosphate groups of PAR, except for PBZ which may also bind PAR by zinc-adenine coordination bond. 7

Enzymatic synthesis of poly(ADP-ribose)
NAD + is an adenylate derivative with a small molecular weight of 662.42, which functions as a coenzyme that is ubiquitous in many intracellular redox reactions. 6, 60 On the other hand, NAD + acts as the substrate of DNA damage-initiated protein PARylation. 6-9 17 members have been found in PARP family, however, not all members have enzymatic activities. Kraus et al. classified PARPs into four subfamilies in their reviews, 7 including: the DNA-dependent PARPs, such as PARP1, PARP2 and PARP3; the tankyrases, such as PARP5A and PARP5B; the Cys-Cys-Cys-His contained PARPs, such as PARP7, PARP12, PARP13.1 and PARP13.2; the macrodomian contained PARPs, such as PARP9, PARP14 and PARP15.
Enzymatically active PARPs have a NAD + -binding domain that binds the nicotinamide and N-ribose fragments of NAD + in different orientations. 7,61 As DNA damage occurs, nuclear localizing PARPs are recruited to the damage sites and NAD + is consumed as a donor of ADPR unit, releasing a nicotinamide and a proton and resulting in PARPs auto-PARylation ( Figure 5). 6,26,61 The first ADPR unit is covalently attached to protein by linking with carboxylic or nucleophilic amino acid residues. [6][7][8][9] Subsequently, repeatedly adding ADPR units to the former synthesized PAR promotes the main chain to elongate. PAR chain that is elongating may branch every 20 to 50 ADPR units, catalyzing by PARPs with branching enzyme activities. 6,7,62 The linkage between ADPR units both in main chain elongated and branched directions ( Figure 1B) adopts α-selective ribofuranosylation of adenosine (namely, the ribose-ribose α(2',1'')-glycosidic bond) ( Figure 5). 34 Figure 5. Process of enzymatic synthesis of PAR, in which NAD + is used as a donor of ADPR unit. The route indicated by filled arrows represents the total reaction and the route indicated by hollow arrows represents stepwise reactions. α-selective ribofuranosylation is adopted to elongate the chain of PAR, producing abundant dendritic PARs with high molecular weights.
Though it is the main enzyme for PAR synthesis, PAPR1 is non-specific to its substrate NAD + , making it possible to enzymatically synthesize PAR derivatives 31 by utilizing three forms of modified NAD + that include base-modified derivative, 29, 63-65 nicotinamide-modified derivative 65,66 and ribose-modified derivative. 64 Marx et al. found that 2'-hydroxyl modified NAD + analogues, picked from a series of modified NAD + compounds, could be promising to visualize the processes of PARP1 catalyzed intracellular PARylation. 67 A small library of compounds that containined alkynyl modification, azide modification and alkenyl modification of NAD + , had been designed (1-10, Figure 6). Among these compounds, 4-10 were obtained according to the synthetic route given by Wallrodt's literature. 67 Bioorthogonal chemical reactions, such as Copper(I)-mediated click reactions (CuAAC), strain-promoted azide-alkyne cycloaddition (SPAAC) and inverse electron demand Diels-Alder reactions (DAinv), were thereby utilized for fluorescent labeling of PAR derivatives which were synthesized by PARPs through incorporating these NAD + analogues into PAR skeleton in vitro.

Chemical synthesis of poly(ADP-ribose)
It ought to be noted that enzymatic synthesis of PAR always generates a pile of compounds with different molecular weights and even with severe branching, waning the structural homogeneity of PAR. Multiple chromatographic purification steps are thus indispensable for enzymatic synthesis to prepare PAR. Therefore, chemical synthesis strategies may afford solutions to this issue because well-defined structure of PAR is available by exquisitely designing synthesis steps. Still, there are two key problems for chemical synthesis to be addressed: formation of disaccharide nucleoside with a ribose moiety connected to adenine by α(2',1'')glycosidic bond and the formation of rather labile pyrophosphate bond. 30 68 and marine organisms. 70,71 Disaccharide nucleosides exist not only in free state, but also as the components of bioacitve molecules such as tRNA 72 , antibiotics and PAR. 68 The common structural feature of disaccharide nucleosides is that they all have extra pyranose or furanose moiety connected to nucleoside at 2'-, 3'-or 5'-OH. In general, there are two routes that can be used to synthesize disaccharide nucleosides (Scheme 1). 38,68 The first route is the coupling of the protected disaccharide with purine, pyrimidine or purine/pyrimidine derivatives. Another route is O-glycosylation of nucleoside, resulting in the generation of O-glycosidic bond between nucleoside containing a free hydroxyl group and pre-activated monosaccharide. Both of the two routes undergo an intermediate of oxonium ion after Xleaving, that determines the transconfiguration of nucleophilic fragment relative to the OR group at C1 site (Scheme 1). It is important that proper route must be selected to prepare disaccharide nucleosides because different reaction sequences or conditions may engender distinct configurations of products. Besides, much attention should be paid to the protective groups under various coupling conditions, so that stable estimation of hydroxyl protection and stereoselective formation of glycosidic bonds can be successfully achieved.

Scheme 1.
There are two routes for the synthesis of disaccharide nucleosides.
Two challenging aspects for disaccharide nucleoside synthesis are inevitable. Firstly, the configuration of O-glycosidic bonds including α-and β-anomers should be controllable. Secondly, selective protection and deprotection of hydroxyl groups must be taken into consideration. Some strategies have been taken to solve these posers. Mikhailov et al. utilized the Markiewicz blocking group (namely, the (ClSi i Pr2)2O reagent was used) to concurrently protect 3'-and 5'-OH. [73][74][75] They also adopted Robins oxidation-reduction sequence to inverse the configuration of 2''-OH. Filippov et al. also used the Markiewicz blocking group for 3'-and 5'protection in the synthesis of disaccharide nucleoside building block of PAR oligomers. 76 Marx et al. synthesized a pool of disaccharide nucleoside analogues by Mikhailov's method. 77 Then they found that some of these analogues showed inhibitory effects on human PARP1 in vitro. 77 Boronic acid has a high affinity with hydroxyl groups, Aoki et al., therefore, realized the regioselective 5'-O-glycosylation of nucleosides by transient protection of 2',3'-diol using a boronic ester. 78,79 In general, benzyl protection of hydroxyl groups on the ribofuranosyl donor in glycosylation reaction enables production of α-glycosidic bond. This α-selective fashion can be explained by nucleophilic attack on the five-membered oxocarbenium ion which may be an active intermediate in the nucleophilic substitution reaction. In contrast, acyl protection of hydroxyl groups exclusively forms β-glycosidic bond owing to neighboring group participation. Besides, 2,3-isopropylidene or 2,3-benzylidene protection of hydroxyl groups is likely to generate a mixture of α-and β-anomers. 76 Many synthetic methods of disaccharide nucleosides have been summarized in some reviews. [38][39][40]68 Here, we mainly focus on the synthetic tactics of protected disaccharide nucleosides which are the key building blocks for the chemical synthesis of PAR. Currently, route 2 (Scheme 1) has been increasingly applied to prepare protected 2'-O-α-D-ribofuranosyladenosine (11, Scheme 2), and the construction of this building block determines PAR synthesis efficiency. The chemical synthesis of 2'-O-α-D-ribofuranosyladenosine was firstly developed by Mikhailov et al. in 2008, using 3',5'-O-protected adenosine (12) and 1-O-acetyl-2,3,5-tri-Obenzoyl-D-arabinofuranose (13) as starting materials. 73 The approach was later opted to prepare 2'-O-Dribofuranosyluridine. 75  Considering the fact that disaccharide nucleoside fragments in PAR uniquely bear α-glycosidic bonds, a suitable and orthogonal protecting group tactic that generates products with accurate configuration via Oglycosylation reaction between ribofuranosyl donors and acceptors, is urgently demanded. Filippov et al. found that imidate compound 21 (Scheme 3) was an efficient ribofuranosyl donor which was nucleophilically attacked by the unprotected hydroxyl group of 22 in the presence of TMSOTf through the mechanism of bimolecular nucleophilic substitution reaction (SN2), forming the expected product 23 containing α-glycosidic linkage. 76 From the reaction of 2,3,5-tri-O-Bn-ribose (20) with Cl(C=NPh)CF3, 21 was prepared with a yield of 92%. In order to avert N-glycosylation, amino group of adenosine 22 was caged by two benzoyl groups. For deprotective reactions, palladium or BCl3 was selected to catalyze the benzyl removal. NH3/MeOH was utilized to completely remove the benzoyl protection. Nevertheless, MeNH2 could only partially deprotect adenosine, generating 25 with single benzoyl protected amino group. TBAF could solely eliminate organosilicon protection and was used to remove the Markiewicz blocking group. Unfortunately, the overall yield of the deprotective reactions was undesired. Compound 24-26 were obtained after deprotecting hydroxyl or amino groups. Finally, fully orthogonal protected building block 28 was prepared with a moderate yield for further synthesis of PARylated peptides. 76 Scheme 3. Imidate 21 was utilized as ribofuranosyl donor for α-selective O-glycosylation of adenosine. The final product 28 was introduced with orthogonal protective groups.
To simulate the authentic structure, construction of the core motif of branched PAR, trisaccharide nucleosides, should hence be taken into consideration. This is a more challenging task because it includes two α-selective glycosylation sites. Gratifyingly, many attempts have been made by Filippov lab. 80, 81 Their strategy has proclaimed success for chemical construction of trisaccharide nucleosides of branched PAR.

Synthesis of the pyrophosphate linkage.
Mono-or multi-phosphorylation (including di-and triphosphorylation) is required for intracellular synthesis of nucleotides, nucleic acids and their derivatives which are converted from nucleosides in the presence of kinases. 82 To date, many approaches of converting alcohols into their phosphate esters have been fabricated by exerting inorganic phosphorylation reagents, such as white phosphorus, 83 phosphoric acid 83 and POCl3. 84,85 These reactions need amines as organic bases or organic ammonium salts as catalysts, and some reactions even take place under hostile conditions, such as rather high temperature. On the other hand, yields of these reactions are always undesired. Nevertheless, very few methods have been developed for multi-phosphorylation of alcohols that contains more than three phosphorus atoms. 86,87 Besides, organic phosphorylation reagents, including 2-cyanoethyldiisopropylphosphoramidochloridite and cyanoethyl P-imidazolides, are also utilized to form mono-or multi-phosphate esters. 88,89 The use of 2cyanoethyldiisopropylphosphoramidochloridite should avoid humid environment and oxidants. A striking feature for multi-phosphate compounds is that phosphate bonds with high energy are quite easy to hydrolyze, resulting in low yields and difficulties in purification and storage. Conventional methods for phosphate ester formation have suffered from some problems. For example, the poor solubility of intermediates and products brings about a need of unamiable solvents such as dichloromethane, tetrahydrofuran and N,Ndimethylformamide; phosphorylation reagents should be prepared in advance under oxygen-free and dry environment; multitudinous protective and deprotective reactions with extensive purification procedures generate low overall yields.
For multi-phosphorylation, such as diphosphorylation and triphosphorylation, two strategies have been developed and are used frequently (Scheme 4). The first strategy is the coupling of alcohol with pre-prepared multi-phosphitylation reagent, followed by oxidation reaction that converts the phosphorus species from P(III) to P(V). Parang et al. reported a solid-phase strategy for multi-phosphitylation reagents construction which were used for multi-phosphorylation of nucleosides and saccharides (Scheme 4C). 88,90 Compared to liquidphase strategy (Scheme 4B), solid-phase strategy led to simplified purification processes of intermediates and products. First of all, compounds 32 and 33 were prepared in the absence of any base via method illustrated in Scheme 4A. 88 PCl3 was selected as the source of phosphorus and reacted with 3-hydroxypropionitrile forming the water-labile compound 29. Diisopropylamino group was used to temporarily protect trivalent phosphorus (compound 30), which could be removed by nucleophilic substitution of alcohols under the catalysis of 1Htetrazole. Then, solid-phase synthesis of multi-phosphitylation reagents (Scheme 4C) was performed onto aminomethyl polystyrene resin in multiple steps. The hindered resin-bound multi-phosphitylation reagents selectively reacted with the most reactive hydroxyl group of nucleosides or saccharides. Trivalent phosphorus was oxidized by either tBuOOH or 3H-1,2-benzotrithiole-3-one-1,1-dioxide (Beaucage's reagent), affording multi-phosphate (46a-h and 47a-h) or multi-thiophosphate (46'a-h and 47'a-h). 1,8-diazabicyclo[5.4.0] undec-7-ene (DBU) was used to remove the protective groups of P-O bonds and final products were collected by cleaving it from resin beads with TFA/DCM/H2O solution. The liquid-phase synthesis illustrated in Scheme 4B is more complicated though synthesis of multi-phosphitylation reagents can also be realized. 90 In 2007, Parang et al. used their solid-phase method (Scheme 4C) to successfully synthesize a group of modified oligodeoxynucleotides bearing pyrophosphate linkages. 91 The second strategy for diphosphorylation of alcohols is the coupling of unprotected hydroxyl group of monophosphates with trivalent phosphitylation reagents, 34 which can be referenced from a work of Filippov lab 92 and a review paper 93 and will be described in part 3.2.3.

Synthesis of poly(ADP-ribose).
PARylation is a universal post-translational modification of nuclear proteins and mediates numerous intracellular processes. Thus, chemical synthesis may be a powerful strategy for construction of PARylated peptides and free oligo(ADP-ribose) molecules which are conducive to the research of cellular signalings involving PARylation. Since there have been some breakthroughs for construction of disaccharide nucleosides and pyrophosphate linkages, chemical synthesis of PAR is no longer out of reach. We here do not emphasize the synthesis of MARylated or PARylated peptides, though there are exactly some instances. 30,92,93 We intend to focus on the synthesis of free PAR molecules by giving several examples.
In 2015, Filippov et al. reported their solid-phase synthetic strategy of ADP-ribose oligomers, which was considered as the first case of chemical synthesis of di-and tri(ADP-ribose). 34 By inverse synthesis analysis (Scheme 5), the skeleton of PAR with repeated ADPR units could be divided into three parts: (1) the initial ribose 5''-phosphate fragment; (2) the disaccharide nucleoside 5',5''-diphosphate fragment; (3) the terminal adenosine 5'-phosphate. Each fragment should be separately synthesized in advance, and then these synthesized fragments (68, 71 and 72) could be further integrated into an entire PAR molecule (Scheme 6C). Thus, a pressing matter of the synthesis is to prepare these fragments with just correct protective groups.  94 However, the most challenging work was the preparation of compound 68 (Scheme 5 and 6B) which was a key building block for PAR synthesis. The coupling of 59 with 1,3,5-tri-O-benzoylribose (60) at low temperature yielded protected disaccharide 61 (Scheme 6B). With benzyl group replaced by acetyl group, benzoyl protected adenine was added to the five-membered ribose (compound 63), catalyzed by HClO4-SiO2. Subsequently, disaccharide adenosine 65 was generated after protecting group replacement and removal of triisopropylsilyl group. (tBuO)2PN i Pr2 reagent was used to phosphorylate the 5''-hydroxyl group of 65, followed by oxidation with tBuOOH, yielding compound 66. Finally, phosphoramidite group was fixed on the 5'hydroxyl group after selectively eliminating DMTr group under a solution of 1% (v/v) trifluoroacetic acid (TFA) in CH2Cl2. Thus far, the building block 68 was acquired by 10 reaction steps in an overall yield of 19%. The resin adenosine 87. Phosphorylation of 5'-hydroxyl group of 87 produced compound 88 which reacted with chlorophosphate 89, generating the building block 80. Compared to solid-phase synthetic strategy described by Filippov et al., 34 the liquid-phase synthetic strategy reported by Hergenrother et al. 95 for forming ADPribose dimer had two obvious differences: (1) the chain elongation was designed in the opposite direction over that of the solid-phase synthetic strategy, so compound 81 was used as the initiator; (2) the pyrophosphate bond between ribose and disaccharide adenosine fragments was formed through the coupling of 78 with 91 which both had tetravalent phosphates, rather than the coupling between one trivalent phosphorimide and one tetravalent phosphate. Besides, the biggest advantage of Hergenrother's liqiud-phase synthetic strategy was that the final purified product 76 had been gained by HPLC in a high yield of 40%.
Even though well-defined and homogeneous products can be obtained, length of chemically synthesized PAR is currently limited to 2 to 3 ADPR units because of a markable hindrance during solid-or liquid-phase synthetic process. It is frustrating that the more complex branched PAR which is closer to natural structure has not been obtained by chemical synthesis. On the other hand, too many reaction steps and complicated postprocessing lower yields of final products. Fortunately, this kind of short chain ADP-ribose oligomers can cover a demand for researches. 16

Conclusions
This short review offers a summary of PAR from its structure to its enzymatic and chemical syntheses. PAR, a negatively charged biomacromolecule, is structurally similar to nucleic acids. Natural PAR can be synthesized by PAR polymerase families as a dendritic polymer with a size up to 400 ADPR units. PAR cannot only be covalently modified to proteins, resulting in the changes of protein-protein interactions, but also, in its free state, influence functions of proteins via non-covalent interactions. Compared to enzymatic synthesis, chemical synthesis is more likely to provide homogeneous PAR products and can reduce the difficulty of purification. However, limitations for chemical synthesis of PAR still exist because, at present, only short chain oligomers (2 to 3 ADPR units) can be obtained. On the other hand, branched ADP-ribose oligomers which are closer to natural structure of PAR, has not been constructed by chemical synthesis strategies. Therefore, efforts have to be paid to overcome these challenges.