Cyclotriphosphate: A Brief History, Recent Developments, and Perspectives in Synthesis

Abstract There has been a recent upsurge in the study and application of approaches utilizing cyclotriphosphate 1 (cyclo‐TP, also known as trimetaphosphate, TMP) and/or proceeding through its analogues in synthetic chemistry to access modified oligo‐ and polyphosphates. This is especially useful in the area of chemical nucleotide synthesis, but by no means restricted to it. Enabled by new high yielding and easy‐to‐implement methodologies, these approaches promise to open up an area of research that has previously been underappreciated. Additionally, refinements of concepts of prebiotic phosphorylation chemistry have been disclosed that ultimately rely on cyclo‐TP 1 as a precursor, placing it as a potentially central compound in the emergence of life. Given the importance of such concepts for our understanding of prebiotic chemistry in combination with the need to readily access modified polyphosphates for structural and biological studies, this paper will discuss selected recent developments in the field of cyclo‐TP chemistry, briefly touch on ultraphosphate chemistry, and highlight areas in which further developments can be expected.


Introduction
Condensed phosphates-molecules containing phosphoric anhydride bonds-can occur in linear,c yclic, and branched forms, [1] giving rise to polyphosphates, [2] cyclo-polyphosphates, and ultrapolyphosphates, respectively ( Figure 1). [3] Also, combinations of such substructures have been described or proposed. [3b] The potentiald iversity of the condensed phosphates has yet to be explored and arguably the synthetic challenges associated with such an endeavor remainp rofound given the difficulties in handling such compounds. [3c] Recent results, however, indicate that many defined condensed phosphates could be readily available following optimized protocols and thus, new types of compounds are becoming available. This will facilitate the design of new reagents [4] and polyelectrolytes [5] ,b ut also probes andt ools to answer fundamental ques-tions of chemistry and biology. [6] It is, therefore, an exciting time to work in the interdisciplinary field of the condensed phosphates.
Water-soluble polyphosphates (linear and cyclic) are formed under prebiotic conditions as ar esult of volcanic activity. [7] Consequently,t hese compounds could have been available as soluble phosphate sources in prebiotic phosphorylation chemistry and much research has been dedicated to understand phosphorylation reactions based on polyphosphate and cyclo-TP 1. [8] As the direct transformation of nucleosides and sugarsw ith 1 is sluggish under severalc onditions, it may simply serve as the precursor for other more potent alcoholphosphorylating agentso btainedb ys equential aminolysis, such as amidotriphosphate, [9] or diamidophosphate [10] (Figure 2A).
It is noteworthy that the presence of an aminen ucleophile and another nucleophile in one molecule can trigger phosphorylation cascades, such as found in the condensation of amino acids using cyclo-TP 1 ( Figure 2B). This supports the argument that in prebiotic phosphorylation reactions relyingo n cyclo-TP 1,a mine nucleophiles potentially played an important role. [11] These examples generally assume phosphorous derivatives in the oxidation state P V ,y et other possibilities include P III derivatives as hydrolysis products of schreibersite, which is delivered by meteorites. [12] Apart from the potential role of cyclo-TP 1 in prebiotic phosphorylation chemistry,n ot much is known about its metabolism in living organisms. Clearly,i th as been identified in yeast extracts in the 1950s [13] but follow-up studies, mainly dealingw ith algae,h ave been scarce. [14] Also the excitingf inding that the single celledg reen algae Chlorella containsc yclic polyphosphates including imido nitrogen atoms [15] has received only very limited attention ( Figure 2C shows some proposed structures still awaiting validation). Thus, much research remains to be done-noto nly regarding prebioticc hemistry of cyclo-TP 1 and relateda nalogues, such as its adenosine ester, [16] but also about its potential biological functions.
Modified cyclo-TP 1 has inspired the minds of synthetic and structuralchemists for along time. The potential of 1 to directly generatem odified triphosphate esters using alcohols is a highly appealing strategy. [17] The reactions should proceed through am odifiedc yclo-TP monoester that is then linearized with water (or other nucleophiles). Despite initial efforts in the 1960s [18] and 70s, [19] this reaction has only been transformed into au seful application severald ecades later.T he activation of cyclo-TP 1 with mesitylene chloride and DABCO [20] is the first example of as ynthetically broadly useful application. Recently,   Cummins et al. conducted al andmarks tudy using the peptidecoupling reagent PyAOP giving stable and isolatable salts of cyclo-TP esters. Moreover,X -ray crystal structures of the modified cyclo-TPs were obtained,a nd intriguingly,a lso of the activated species. [21] These approaches now enable the direct customization of cyclo-TP 1 with different nucleophilesf ollowed by hydrolysis to give linear modified triphosphates.O fn ote, modified cyclo-TPs have been generated synthetically by other successful approaches previously,b ut not by direct modifica-tion of intact cyclo-TP 1.A no verview of the strategies can be found in Scheme1,w hich have also been the subject of comprehensive reviews. [22] This Conceptp aper aims to provide an update of how cyclo-TP 1 can be chemically generated or directly activatedfocusingo nr ecent approaches. We have arbitrarily grouped the approaches into three categories:1 )generation of cyclo-TP esters, in which the P-anhydrides are constructed subsequent to R-OH modification;2 )generation of cyclo-TP esters, which rely on direct functionalization of cyclo-TP 1;3 )generation of activated cyclo-TP esters, in which the P-anhydrides are constructed prior to R-OHm odification.E specially the second and third strategy will be discussed in more detail, because strategy 1h as been known and appliedf or al ong time. We will then discuss under 4) how cyclo-TP analoguesc an be used to construct P-anhydrides as well.
These approaches have been used in their majority to generate modified nucleoside oligophosphates that serve as versatile tools to monitor,p erturb, and understand biological processes. Especially,n on-hydrolyzablea nalogues of nucleoside triphosphates belong to the standardr epertoire in structural biology. [23] This focus is not surprising, given the centralr oles of nucleosidet riphosphates in life. [24] Yet, the approaches discussed herein are highly useful to access other structures as well, which will be discussed later.S everala pplications of NTP analogues,s uch as monitoring enzyme activity using spectroscopic/fluorescent probeso ra nalyzing the (m)RNA cap structure, just to provide two examples of how useful such analogues are, have recently been the subjecto ft wo in-depthR eviews, which are highly recommended to the interested reader. [25] 1. Synthesiso fCyclo-TP Esters Using Phosphoric Anhydride Forming Reactions Especially the Ludwig-Eckstein protocol has found widespread application. [26] It proceeds by using the commercial P III reagent 2-chloro-1,3,2-benzodioxaphosphorin-4-one, which is initially Scheme1.Overview of different strategies to generate modified (deoxy)-cyclotriphosphates. P i = orthophosphate;PP i = pyrophosphate;DCC = dicyclohexylcarbodiimide. added to the nucleoside followed by pyrophosphate to generate ad eoxy-cyclo-TP ester.T his intermediate is then oxidized to the cyclo-TP ester and linearized (Scheme 2). [22b,e] Another approachg enerating deoxy-cyclo-TPs as intermediates was published by Fischer. [27] In this case, ap hosphordiamidite or dichloridite is used as precursor (Scheme 2). The widely-applied Ludwig-Eckstein approach was preceded by P V chemistry following the Yoshikawa and Ludwig strategies, and earlier protocols by Khorana using dehydrating agents, such as DCC. [28] The Khorana method likely leads to accumulationo fn ucleoside triphosphates by ac yclo-TP ester,w hich was later demonstrated by 18 Oi sotope exchange experiments. [29] When using P V chemistry,t he oxidation step can be omitted, which can be beneficial in cases were labile nucleosides are used. However, the oxidations tep can also be highly valuable to introduce additional modificationsinthe a-position. Moreover,coordination of BH 3 enables the generation of boranophosphates, which are useful nucleotide analogues anda ntivirals. [30] Also, non-hydrolyzable analogues can be generated using this approach, for example, bearing CF 2 replacements of the b-g oxygen atom (Scheme 2). [31]

Direct Modification of Cyclo-TP
In 1949, Thilo reported that after hydrolysis of cyclo-TP 1 with sodiumh ydroxide only linear triphosphate 2 was obtained. [32] In the following years, variousn ew strategies were published using different aminen ucleophiles for the linearization. The first phenolysis of 1 in basic solution (pH 9) was reported by Feldmann in 1966. [18] The reaction was very slow and after 10 days only 7% of the linear product 3 was obtained. In 1971, Trowbridge reported the pH dependency of the ring-opening of 1 using methanol as nucleophile. [19] If no base is added to the solution of cyclo-TP 1 in anhydrous methanol, only methylated monophosphate was isolated as product. When adding a base, such as lithium methoxide, up to 10 %o fm ethyl triphosphate 4 was obtained( Scheme 3A). Additionally,i tw as possible to synthesize an ATPa nalogue 6 containing ap hosphoramidate at the 5'-carbon of the ribose by treating the 5'-amino nucleoside 5 with 1.T his was the first example of applying cyclo-TP 1 as areagent in nucleotide chemistry( Scheme 3B).
Mohamady and Ta ylor,w ho published an ovel synthesis of nucleoside oligophosphates using cyclo-TP 1 as triphosphorylation reagent, revived its application in nucleotides ynthesis in the 2010s. [20,33] Af irst successfuls ynthesis of protected nucleo-side triphosphates by esterification was reported by the same group in 2016. [20b] The activation of 1 is split into two parts. First, it is sulfonylated with 2-mesitylensulfonyl chloride (MstCl). Subsequently,t hey propose that the sulfonyl group is substituted in less than one minute by an ucleophilic attack of an amine base, such as 1,4-diazabicyclo[2.2.2]octane (DABCO) or N-methyl imidazole (NMI) to generate the reactive intermediate 7 that readily reacts with 2',3'-protected nucleosides (Scheme 4). Linearization with hydroxide is achieved in 100 mm triethylammonium acetate (TEAA) buffer (pH 7). After removalo fp rotecting groups, followed by reverse phase HPLC (RP-HPLC)p urification, the target nucleoside triphosphates 8-13 were isolated in yields between 70 and 79 %.
In 2019, Cummins published ag roundbreakinga pproach utilizingc yclo-TP activation that enables reactions with C, N, and On ucleophiles. [21] The bis(triphenylphosphine)iminium (PPN) salt of cyclo-TP 1 reacts on agram scale with the peptide coupling reagent PyAOP (Scheme 5) under ambient conditions to intermediate 14 in 70 %y ield. In as econd step, an ucleophile substitutes the tripyrrolidinophosphine oxide leaving group to give 15-19 in yields of 40 to 74 %. The modified cyclo-TP is then hydrolyzed with tetrabutylammonium hydroxide (TBA-OH), giving modified linear triphosphates 20-22 in yields between 54 and 82 %. Although this reagent has not yet been used with nucleosides, its reactivitywith MeOH and EtOH (17,18)h as been studied.
Intriguingly, 14 is obtained as as table solid that was studied by X-ray crystal analysis ( Figure 3) providing insight into its molecular structure.I nt he 31 PNMR spectra,apeak at around À30 ppm is observed that corresponds to the central "ultraphosphate"-like phosphorous atom of the reagent. The thorough characterization of the activated species will greatly facilitate the design of other reagents in the future and holds promise for further exciting discoveries.
The Cummins group goes on to demonstrate an extension of the work by using the Wittig reagent CH 2 PPh 3 ,t hereby including C-nucleophiles in the reaction scope (Scheme 6). The resultinga nalogue 23 of cyclo-TP was obtained in 61 %y ield after 24 hours. With this product in hand, many different new manipulations can be envisioned. Simple hydrolysis yields a methyl phosphonate analogue 24 of cyclo-TP.R eactions with aldehydes, however,e nable the synthesis of alkenyl phosphonates 25 and 26 in yields between 63 and 85 %. Furthermore, the cyclic alkenylp hosphonate 25,g enerated with formaldehyde, wasl inearizeda sd escribed above to yield the first linear triphosphate analogue 27 synthesized from cyclo-TP 1 with a PÀCbond in 54 %y ield.  In this context,t he bismethylenea nalogueo fc yclo-TP 30 has been studied previously regarding its reactions with nucleophiles (Scheme 7). In 1969, Trowbridge reported the synthesis of an on-hydrolyzable analogue 31 of ATPu sing the modified bismethylene cyclo-TP 30 in 16 %y ield. [12] 29 wass ynthesized according to ap rocedure of Maier. [34] These bismethylene analogues of cyclo-TP have also been studied by Overhand [35] as potentials qualene synthase inhibitors. In these cases,h owever, the PCP bond was part of the ring structure, different to the compounds 23-27 reported by Cummins.

Synthesis of Cyclo-TP Esters Using Phosphate-Ester Forming Reactions
There have been two reportso nt he formation of cyclo-TP 1 from fragments of it that result in the generation of an activated electrophilicd eoxy-cyclo-TP.T he first approachb yH uang is related to the Ludwig-Eckstein procedure, butr everses the order of addition of alcohol and pyrophosphate to reagent 32. As ar esult, an activated deoxy-cyclo-TP 33 is formed (Scheme 8) that bears ap henolic leaving group on the P III atom. This can then be replaced with nucleophilesi natwo-step mechanism proposed by Huang:I nitiallyb ya ttack on the P III atom andc leavage of the anhydride, to give 34,f ollowed by regenerationo ft he deoxy-cyclo-TP ester 35 by expulsion of salicylate. This approach is applicable to unprotected nucleosides as demonstrated in their seminal study,r esultingp referentiallyi nt riphosphorylation on the 5'-OH (5':3' ratios ca. 85:15) giving products 8-11. [36] The second approach was published recently by our group in an effort to generate phosphoramidite (P-amidite) analogues of 1.T he rationaleb ehind this project was that phosphoramidites [37] provide several benefits in terms of reactivitya nd selectivity,w hich we have exploited over the years in the selective construction of modified phosphoric anhydrides. [38] Thus, it seemed promising to develop ande xplore such reagents in the context of tri-and polyphosphorylations. These reagents look similar to the activated deoxy-cyclo-TP 33 proposed by Huang( the PÀOl eaving group is exchanged foraP ÀNl eaving group), but may have advantages regarding long-term storability,h igh and tunable reactivity after activation (depending on the activator), and ease of handling. Intriguingly,a lso pyrophosphate analogues,s uch as several phosphonates can be used to generate af amily of reagents, which are summarized Scheme6.Reagent 14 is aversatile precursor to obtain triphosphate analogues containingaP ÀCbond.
Scheme8.Synthesis of deoxy-cyclo-TP by P-ester synthesis according to Huang.  Figure 4a lso shows the 31 PNMR data of the reagents 36-39 generated in acetonitrile under dry conditions, underlining their stability,p urity,a nd ease of preparation.
The reagents 36-39 weret hen applied to the synthesis of modified nucleoside triphosphates and non-hydrolyzable analogues. [39b] After phosphitylation of alcohols with c-PyPAr eagents, the oxidation step enables the introduction of O, S, and Se in the a-position. Detailed 31 PNMR studies indicated very clean conversions, so that the linearization step could be conducted from almost pure cyclo-TP esters 40.N ucleophiles used in the linearization step included N-nucleophiles (ammo-nia, azide, primary andsecondary amines, imidazole), O-nucleophiles (hydroxide, alcohols, fluorophosphate), and fluoride. In all cases, ac lean linearization was observed, providing the products 41 in high yields, even in the case of unprotected nucleosides, such as Tand A. In these cases partial phosphorylation also on the 3'-OH was observed (ca. 10-15 %). Scheme 9 gives ageneral overview on the versatility of the approach.
DFT calculations provided insight into the mechanism of the ring-opening reaction suggesting an in-line attack of the nucleophile on the sterically least hindered phosphate followed by expulsion of the least charged leaving group. The proposed transition state is in line with the empiricalr eaction energetics leadingt ol inearization and placement of the nucleophile in the terminal position.

Extension of the Concept:S ynthesis of Phosphoric Anhydridesw ith Cyclo-TP Analogues
Activated cyclo-TP analogues can also be subjected to reactions with phosphate nucleophiles, thereby generating phosphorylated cyclo-TPs like intermediate 42 that can be characterized as cyclic ultraphosphates,a st hey contain at rifurcation on at least one central phosphate subunit (Scheme10, Pa tt rifurcationh ighlighted in orange). In analogyt ot he previous chapters, such structures can also be linearized to give, for example,nucleoside tetraphosphates or linear polyphosphates.
Ta ylor has reported on the reaction of nucleoside monophosphates with their activated cyclo-TP 41 (MstCl, followed by NMI), with subsequent linearization with hydroxide or other nucleotides to give nucleoside tetraphosphates 43 and dinucleoside pentaphosphates 44,r espectively (Scheme 10). [20a] Additionally, fluorescent labels werea ttached to nucleoside tetraphosphates using this strategy. [33] Recently,K ool has made use of Ta ylor's approacht os ynthesize dicaptides, whicha re mixed dinucleoside pentaphosphates. These are substrates of the Klenow fragment of DNA polymerase Ia nd other polymerases. Interestingly,t wo of these heterodimeric nucleotides sufficef or full four-base primer extension on DNA template strands. [40] This approach by Ta ylor was also used in our laboratory to construct the first monodisperse linear octaphosphate analogue with terminal phosphoramidate labels for polyphosphate transfer studies to proteins.T he study demonstrated that cyclo-TP 1 is an excellent precursor to build up longer polyphosphate chainsb yb idirectionals ynthesis. [41] It proceeds by reactingt wo equivalents of the activated cyclo-TP with pyrophosphate 45 giving the pyrophosphate bridged cyclo-TP 46 as intermediate (Scheme 11)t hat is then linearized with amine nucleophiles. The symmetric intermediate was identified by 31 PNMR in solutiona nd the product 47 with terminal propargyl phosphoramidates was obtainedi n1 3% isolated yield after ion exchange chromatography.
The need to access polyphosphate probes with defined chain length and modifications for studies into its biological functions has also motivated af ollow-up study using c-PyPA 36 as triphosphorylating reagent. [39a] This application has led to significantly improved yields (40 %o f47)a nd also enabled the generation of polyphosphates with different labels at the chain termini as shown in Scheme 12. In brief, c-PyPA 36 was reactedi ne xcessw ith pyrophosphate 45 to achiever eaction on both terminal phosphates. The P III intermediate was oxidizedw ith mCPBA to give the same intermediate 46 as previously achievedw ith Taylor's reagent. Linearization with propargylamine gave an eight-units polyphosphate 47 with identical phosphoramidateso nt he termini. On the other hand, the use of c-PyPA 36 in limiting amountse nabled the selective reaction with pyrophosphate 45 on one terminus only,g iving access to intermediate 48,a sd emonstrated by 31 PNMR spectroscopy. After linearization with propargylamine, the five-units polyphosphate 49 with one terminal phosphoramidate was obtained. The unreacted terminus of that compound easily engaged in as econd round of coupling with c-PyPA 36,f ollowed by oxidation andr ing-opening with as econd nucleophile. This iterative triphosphorylations trategy using 36 provides access to polyphosphates with different labels at the termini( 50, Scheme 12). Such compounds will be useful in studying polyP turnover in cells [42] and could potentially help in identifying the elusivem ammalian polyP kinase.
Cummins has reported on the applicationo ft heir PyAOP-activatedc yclo-TP reagent 14 in reactions with phosphate as a nucleophile to generate monophosphorylatedc yclo-TP 19 (see Concept Scheme 5). The identityo ft his intriguing structure was confirmed by X-ray crystallography. [21] Thus, 14 can also be used to construct efficiently phosphoric anhydrides.
To extend the applications of cyclo-TP 1 and improveo ur generalu nderstanding of cyclophosphate chemistry,l arger cyclic phosphates and their reactions are under investigation. Cummins, for example, reported am odification of cyclotetraphosphate 51 (tetrametaphosphate) to synthesize the first fully characterized esters of this family as shown in Scheme 13. [43] Te trametaphosphateg ets protonated in presence of as trong acid like trifluoroacetic acid (TFA) to generate dihydrogent etrametaphosphate 52 in 94 %y ield on ag ram scale. Further treatment with N,N'-dicyclohexylcarbodiimide (DCC) resulted in ac ondensed intermediate 53 that is in equilibrium with dihydrogen tetrametaphosphate 52.T his intermediate contains an "ultraphosphate-type" structure with characteristic shifts in the 31 PNMR at around À35 ppm, featuring ab icyclic trimetaphosphate with ap hosphorica nhydride bridge. [44] Its structure was established by X-ray diffractiona nalysis.T he forward reaction was carriedo ut in acetonitrile with ay ield of 82 %. The backward reactiono ccurred in wet acetonew ithin one minute to regenerate the dihydrogen form in 68 %i solated yield. Methanol was used as nucleophile to cleave the anhydride and form an ovel methyl cyclotetraphosphate ester 54 in 96 %y ield.

Outlook
The generation of "ultraphosphate-type" intermediates shown in Scheme 5, Scheme 10, Scheme1 1a nd Scheme 12 by an ucleophilic attack on activated cyclo-TP promises efficient entry to the underexplored field of branched condensed phosphates devoid of cyclic substructures. Currently,u ltraphosphates are generally known as polydisperse glass-forming and crystalline substances with av ariety of substructures. [45] Alreadyi nt he 1950s, cross-links within phosphate chains were detected after heatingN aHPO 4 ·H 2 Ot o9 70 8Cf or 14 h [46] and initial data on the hydrolysis of the branching points were collected by viscosity and pH measurements. [47] Although crystalline ultraphosphates have been used in ar ange of different applications by high-temperature processes, [48] no rational synthesis of defined molecular non-cyclic ultraphosphates has been reporteds of ar. Moreover,c onsidering that all possible classes of condensed phosphates have been identified in biology,t he absence of ultraphosphates in this list is surprising. Yet, the idea of their potential involvementi nb iological processes has been discarded and will be experimentally demanding to study "because they are unusually rapidlyh ydrolyzed in aqueous solution". [14c] Particularly in light of the lack of data on monodisperse ultraphosphates regarding their hydrolysis half-life andt he fact that unstable structures occur as intermediates in biology,asynthetic access to well-defined ultraphosphates to betteru nderstand their properties is desirable and may lead to interesting new findings.
The synthesis of defined structures could, for example, commence from cyclo-TP analogues as described above (Scheme 14), [21,41] if the regioselectivityo ft he ring-opening on the ultraphosphate-like intermediate 19 could be controlled. Up to now,o nly linearized products 55 were isolated after the attack of an ucleophile on cyclo-TP anhydrides [39a] and DFT calculations suggest ad ifference of about 8kcal mol À1 in the transition states preferring linearization over branching. [39b] Nevertheless, variousc onditions and metal coordination may be exploited to direct the ring-openingt owards branching. In doing so, several modificationsm ay be introduced by using different nucleophiles to give 56 facilitating in vitro or in vivo studies with novel ultraphosphate tools to study the occurrence of ultraphosphates in biology.M oreover,ultraphosphates could-as high-energye ntities-represent an underappreciated phos-phorylating agent in prebiotic chemistry-much more reactive than for example, cyclo-TP 1 itself.
The synthesis of larger cyclophosphate rings and their transformationi nto novel reagents is also ah ighlyp romising area of research. One recent example is the cyclic ultraphosphate disclosedb yC ummins (see Scheme 13). Many more reagents of this type can be envisioned-as structurally beautiful and at the same time highly useful compounds for manifold applications.

Conclusions
The direct customization of cyclo-TP 1 or its assembly from precursors as reagents in the synthesis of diverse condensed phosphates is setting up significant possibilities in the field of nucleotide synthesis, but also beyond. These compounds have interesting implications and applicationsi nt he fields of prebiotic chemistry, polyelectrolyte research, and as reagents in organic and inorganic synthesis. Modified (cyclic)polyphosphates are now becoming accessible, and the potentialg eneration and study of defined ultraphosphates and their interactions with metals holds promise to open up aw hole new area of research that haso nly received very limited attentioni nr ecent years. Such studies could furtherl ead to new concepts in prebiotic phosphorylation chemistry and will in general provide new insights into the field of the condensed phosphates.
2019 is the year of the periodic table and also the 350th anniversary of the discovery of elemental phosphorous by HennigB rand in Hamburg. We hope that with this perspective, we have not only informed aboutr ecent developments in cyclo-TP chemistry,b ut also conferred to the reader-at least in part-ourf ascination of the condensed phosphates-a substance class that provides so many challenges and opportunities still waiting to be explored.