Non-Hydrolysable Analogues of Cyclic and Branched Condensed Phosphates: Chemistry and Chemical Proteomics

Studies into the biology of condensed phosphates almost exclusively cover linear polyphosphates. However, there is evidence for the presence of cyclic polyphosphates (metaphos-phates) in organisms and for enzymatic digestion of branched phosphates (ultraphosphates) with alkaline phosphatase. Further research of non-linear condensed phosphates in biology would profit from interactome data of such molecules, however, their stability in biological media is limited. Here we present syntheses of modified, non-hydrolysable analogues of cyclic and branched condensed phosphates, called meta-and ultra-phosphonates, and their application in a chemical proteomics approach using yeast cell extracts. We identify putative interactors with overlapping hits for structurally related capture compounds underlining the quality of our results. The datasets serve as starting point to study the biological relevance and functions of meta-and ultraphosphates. In addition, we examine the reactivity of meta-and ultraphosphonates with implications for their “hydrolysable” analogues: Efforts to increase the ring-sizes of meta-or cyclic ultraphosphonates revealed a strong preference to form trimetaphosphate-ana-logue structures by cyclization and/or ring-contraction. Using carbodiimides for condensation, the so far inaccessible dianhy-dro product of ultraphosphonate, corresponding to P 4 O 11 2 � , was selectively obtained and then ring-opened by different nucleo-philes yielding modified cyclic ultraphosphonates.


Introduction
For a long time, cellular condensed phosphates were defined as exclusively linear structures [1,2] until evidence was recently given for the presence of cyclic polyphosphates (metaphosphates) in bacterial phosphate granules. [3][5][6][7] While several biological functions [8] -depending on abundance, chain length and subcellular location -are known for linear polyphosphates and the enzymology of (poly)phosphate homeostasis has been well described for different pro-and eukaryotes, [9,10] the biology of metaphosphates remains understudied.However, evidence is available for their existence and a role in the origins of life has been discussed. [3,11]It was, for example, shown that hexametaphosphate is metabolized by Xanthobacter autotrophicus extracts but the catabolic enzyme(s) are unknown. [3]Studies on the interaction of metaphosphates with proteins are hampered by a dearth of water-stable, functionalized analogues, which allow the identification and enrichment of interactors.
[20] Avoiding the need for an additional activation step, aryne chemistry enables direct modification of metaphosphates with different ring-sizes representing a broadly applicable method. [21]Yet, metaphosphate esters are susceptible to nucleophiles, including water, and such reactions lead to linearization usually within minutes to hours, depending on the nucleophile.This reactivity is detrimental to their use as pull-down probes in proteomics studies. [12,13,15,22]hosphoric anhydride bonds can be stabilized via replacement of oxygen by CH 2 or CF 2 groups, which gives rise to phosphonates. [23,24,25]Taking advantage of the increased stability of such non-hydrolysable analogues, affinity reagents can be constructed to investigate the interactome of a substrate, as, for example, recently shown for inositol pyrophosphates. [26]A substitution of the ester function of modified trimetaphosphate by a stable P C bond was recently reported by Cummins. [27]owever, these trimetaphosphate analogues were usually linearized in water within 24 h, highlighting the need to stabilize phosphoanhydride bonds.[30] Condensation of unmodified bismethylene triphosphate using N,N'-dicyclohexylcarbodiimide (DCC) yields the trimetaphosphate analogue, from here on called trimetaphosphonate, which Kenyon used to synthesize a nonhydrolysable adenosine triphosphate (ATP) analogue. [29]Since no simple coherent nomenclature is defined for differently CH 2substituted analogues of condensed phosphates, we decided to generally indicate such modifications by naming structures as phosphonates, although they might contain a phosphinate (as in 14) or phosphine oxide (as in 16) substructure as well.
Based on the described results, we envisioned the synthesis of trimetaphosphonates modified at the phosphinate function rather than at the phosphonate subunits (Scheme 1b) and study the hydrolytic stability of this unexplored structural motif.Showing sufficient stability, trimetaphosphonate should be developed into capture compounds for pull-down experiments.
[33] Due to the lability of the branching phosphate, [34] these structures can easily be linearized by different nucleophiles. [12,35]n the context of the hydrolysis of branched condensed phosphates (ultraphosphates), Van Wazer coined the antibranching-rule in 1950, describing ultraphosphates as exceedingly unstable and more labile than cyclic or linear polyphosphates. [1,36]Over the decades, the rule persisted and led to the widespread perception that ultraphosphates will have no role to play in biology. [1,37]Even so, we know that many unstable modifications such as phosphohistidine [38] do exist in nature and therefore experimental approaches to discover a potential ultraphosphate biology are warranted.In this context, hydrolysis half-lives up to days were recently determined for synthetic monodisperse ultraphosphates and evidence was given for their hydrolysis by alkaline phosphatase. [39]The sensitivity of ultraphosphates towards acidic pH, nucleophilic reagents, divalent cations and drying however hamper their enrichment and analysis in efforts to detect branched phosphates in biological samples.Thus, we aimed to functionalize a non-hydrolysable analogue, from here on called ultraphosphonate (15), to serve as a capture compound in a chemical proteomics approach to investigate the interactome of branched condensed phosphates for future studies into their potential biology.
Herein, we describe the syntheses of trifunctional pull-down probes with non-hydrolysable analogues of trimetaphosphate, different ultraphosphates as well as inorganic tetraphosphate as selection function.These capture compounds were applied in photoaffinity pull-down experiments with yeast cell extracts and putative interactors were identified.These hits may aid in the identification of enzymes that process or make metaphosphates or branched condensed phosphates.

Synthesis of modified trimetaphosphonate
The aryne phosphate reaction provides direct access to arylated metaphosphates of controllable ring-size without the need for an additional activation step and was therefore initially tested as method to modify non-hydrolysable analogues of metaphosphates. [21]Since substitution of the bridging oxygen of a phosphoanhydride bond by a CF 2 -group retains the correct polarity, while it is reversed with CH 2 as the bridging unit, bis(difluoromethylene)triphosphate ( 17) was synthesized following a procedure of Olah [24] and the corresponding tetrabutylammonium (TBA) salt of 17 cyclized in 64 % yield using DCC for condensation (Scheme 2a).To avoid linearization of the esterified metaphosphate-analogue in accordance with previous reports, [12,13,15,21,27] arylation at the phosphinate oxygen was desired (Scheme 2a, marked in green).CF 2 -trimetaphosphonate 18 was reacted with the alkynylated Kobayashi-type aryne precursor 19 to enable subsequent further functionalization by copper-catalysed 1,3-dipolar cycloaddition (CuAAC) resulting in a complex mixture instead of monoarylated 20.The products were separated and characterized using HRMS and NMR analysis, revealing overreaction to bisarylated products, ring-opening by water or fluoride, products of a side reaction of the aryne with THF [21,40] and combinations thereof (see Supporting Information, Scheme S1).We therefore decided to first selectively introduce an alkyne moiety at the phosphinate function and then construct the bismethylene triphosphate using an Arbuzov reaction (Scheme 2b).The latter are usually performed with trialkylphosphites and the resulting phosphonate esters can be hydrolyzed by either refluxing with conc.HCl or treatment with TMSBr. [28,41]While strongly acidic conditions would likely hydrolyze phosphinate esters as well, reports on the selective removal of benzyl protecting groups in arylphosphate esters using TMSBr [42] and selective dealkylation of arylated bisphosphonates using TMSCl and NaI [43] suggested compatibility of these deprotection strategies.
Esterification of bis(chloromethyl)phosphinic chloride (22) with the alkyne-modified phenol 19 gave the arylphosphinate 23 in 68 % yield.The Arbuzov reaction of 23 with P(OEt) 3 at 170 °C gave a mixture of aryl-modified 25 and pentaethyl bismethylene triphosphate (24) in 3 : 1 ratio overnight, as indicated by 31 P{ 1 H}-NMR reaction control.This ratio changed to 2 : 3 after two days at 170 °C suggesting that 25 can arylate P(OEt) 3 , similar to reports on alkylation of P(III) structures by phosphinate alkylesters. [44]This reactivity was not observed when the Arbuzov reaction was carried out at 155 °C, giving 25 in 60 % yield.Of note, attempts to react bis(chloromethyl)phosphinic acid in an Arbuzov reaction without alkylation of the phosphinate function failed (see Supporting Information, Scheme S2).We found that the phosphinate is already alkylated after the first Arbuzov reaction and analyzed the corresponding product crystallographically (see Supporting Information, compound S11).Following the procedure of Guillaumet, [42] the ethyl esters of 25 were cleaved by treatment with a mixture of TMSBr and pyridine followed by hydrolysis of the silylated intermediate.Corresponding 31 P{ 1 H}-NMR spectra of the reaction are shown in the Supporting Information, Scheme S3. 31 P-HMBC confirmed that the phosphinate still carried the aryl modification.Partial cleavage of the alkyne-TMS group was observed at pH 8.5, but albeit desilylation was desired for subsequent clickreactions, the pH was not further increased to avoid hydrolysis of the aryl ester under more basic conditions. [45]After strong anion exchange (SAX) purification, the aryl ester was even hydrolyzed at almost neutral pH with 80 % of the isolated product being decomposed after 40 h in water.We therefore abandoned this synthetic strategy.
To avoid hydrolytic cleavage of the alkyne function, we reacted 22 with the Grignard reagent 30 [46] to access the phosphine oxide 31 in 41 % yield (Scheme 2c). [47,48]The Arbuzov reaction of 31 with P(OEt) 3 gave 32 in 89 % yield.The ethyl esters were cleaved by silylation using a mixture of TMSBr and pyridine [42] followed by methanolysis giving 33.After treatment with NaOH solution (pH 10) to remove the alkyne-TMS group, 34 was obtained in 81 % yield over two steps.After conversion to the TBA salt, 34 was condensed using N,N'-diisopropylcarbodiimide (DIC).NaClO 4 in acetone was used for precipitation giving product 35 in quantitative yield as sodium salt, which formed single crystals during slow evaporation.X-ray analysis confirmed the structure shown in Scheme 2c allowing comparison of bond lengths and angles of the non-hydrolysable analogue 35 with crystallographic data of trimetaphosphate tetramethylammonium salt. [49]While the P O bond lengths in these two structures only differ by 1-2 %, the P C bond is markedly elongated to 1.81 Å compared with 1.65 Å for the P O bonds in the trimetaphosphate salt.For the bond angles, deviations in the range of 5-13 % were determined, corresponding to results of Yount for the comparison of pyrophosphate and its methylene-bridged analogue. [50]e monitored the hydrolytic stability of the phosphoanhydride bond of 35 in water (neutral pH) at room temperature by 31 P{ 1 H}-NMR.We did not detect any decomposition over the course of 24 h and found approximately 1 % hydrolysis after ca. 6 days allowing further development of 35 into a capture compound for pull-down experiments.35 was reacted with amino-PEG 3 -azide in a CuAAC click reaction giving 36 in 51 % yield.The introduced amine moiety enabled ready modification with commercially available sulfo-SBED, containing biotin and a photoreactive phenyl azide group.Recently, this linker was successfully used by us to determine the (p)ppGpp interactome. [51]The trifunctional pull-down probe 37 was obtained in 70 % yield for the last step.

Phosphoramidite chemistry with linear triphosphonate
Since larger cyclic phosphates than trimetaphosphate have also been identified in biology, [3,11,52] we intended to take advantage of the hydrolytically stable alkyne modification in 34 and 35 and increase the ring-size using phosphoramidite chemistry (Scheme 3).The reaction of 34 with fluorenylmethyl (Fm) protected phosphoramidite 44 gave the tetraphosphonate/ phosphate analogue 38 only as an intermediate, which cyclized within minutes to trimetaphosphonate 35.Therefore, we reacted 34 with phosphordiamidite 45 to directly obtain the cyclic tetrametaphosphonate product.However, after twofold coupling, the resulting intermediate 39 underwent an intra-molecular nucleophilic attack of one phosphonate onto another, resulting in ring contraction and formation of 41, reminiscent of a cyclic ultraphosphonate containing a P(III)-P(V)anhydride (Scheme 3).Next, 41 was hydrolyzed giving again trimetaphosphonate 35 as the product.Changing to basic conditions, 34 was treated with methyl dichlorophosphite (46)  in presence of NEt 3 .After oxidation, signals at δ = 24 ppm indicated successful formation of 40.However, trimetaphosphonate 35 was also already present as judged by 31 P{ 1 H}-NMR and found to be the main product after precipitation while only traces of 40 were left.Attempts to isolate 40 failed but linearized 43 -as another decomposition product of 42 -was purified and characterized (see Supporting Information).
In analogy to the ring-opening of cyclic ultraphosphates and functionalized trimetaphosphates, one would have expected 43 as the single hydrolysis product. [12,14,19,32,53]Also the rate of trimetaphosphate formation was unexpectedly high, suggesting that geometrical differences caused by methylene substitution of oxygen influence the reactivity.Considering the "hydrolysable" analogues, Glonek and Myers already reported slow conversion of larger metaphosphates to trimetaphosphate with intermediary formation of cyclic ultraphosphates but details of the mechanism remained elusive. [17,54]Computational results by Cummins suggest a plausible mechanism, [20] supporting the pathway shown in Scheme 3. In addition, Cummins recently showed that the P 2 O 5 (pyridine) 2 adduct does not form tetrametaphosphate with the bis(triphenylphosphine)iminium (PPN) salt of pyrophosphate but orthophosphoryl trimetaphosphate, underlining the high propensity to form trimetaphosphate rings. [27]

Ultraphosphonate aryne reaction
To study the potential interactome of ultraphosphates, the nonhydrolysable analogue 15, called ultraphosphonate, was chosen as selection function for a pull-down probe.15 was synthesized according to a procedure of Maier. [48,55]The corresponding TBA salt was obtained as an oil, which slowly crystallized during storage at 4 °C allowing for X-ray analysis (Scheme 4).Regarding the functionalization of 15, we were interested to study whether an aryne reaction would result in O-arylated products as in the aryne phosphate reaction [21] or in P C bond formation as observed for phosphinates and phosphonate esters by insertion of arynes into P O bonds. [56]sing Kobayashi-type o-silylphenyltriflate aryne precursors, we found O-arylation and mono-to trisarylated products with formation of phosphonate monoester preferred over diester formation under the applied conditions (Scheme 4).Reactions with 2 equiv. of aryne precursor typically showed ca. 25 % starting material left and ca.40 % of the desired monoarylated product as indicated by 31 P{ 1 H}-NMR reaction control.Around 25 % were found to be bisarylated.The products were readily separated using a C18 AQ column and isolated with the yields indicated in Scheme 4. Due to the propensity of 15 to act as a nucleophile in additional arylations, but at the same time to intramolecularly condense to metaphosphonate substructures, these yields are acceptable.The aryne reaction with 19 results in positional isomers and a ratio of paravs.meta-substitution of 2 : 1 was found for monoarylated 53, while bis-and trisarylated 54 and 55 were isolated as more complex isomeric mixtures, which we did not separate and assign.Synthetically, the formation of regioisomers might appear disadvantageous but for the preparation and successful application of pull-down probes higher structural variety can actually be helpful.

Ultraphosphonate condensation chemistry
Since cyclic ultraphosphates are structurally closely related to non-cyclic ultra-as well as metaphosphates and show constitutional isomerism with the latter, we were interested to also synthesize and functionalize the non-hydrolysable analogue 56 (Scheme 5).Preparation of the fully dehydrated species 58representing an analogue of P 4 O 10 -by intramolecular dehydration of 15 using carboxylic anhydrides or acyl chlorides at elevated temperatures was reported by Maier and Kerst. [57,58]he "full" anhydride is rapidly hydrolysed to anhydride 56, whereas further hydrolysis to non-cyclic ultraphosphonate is only achieved after 8 h at 80 °C. [57]We studied the chemistry of 15 in presence of DIC as condensing agent and found that the dianhydro product 57 is already formed prior to consumption of non-cyclic ultraphosphonate 15. 25 equiv.DIC were sufficient for full turnover to 57 at 80 °C overnight.The 31 P{ 1 H}-NMR spectrum of 57 in DMF-d 7 is shown in red in Scheme 5. Notably, we did not observe formation of the adamantane structure 58 even at higher excess of DIC.While no rational synthesis of dianhydro ultraphosphonate (57) was known yet, our finding is in accordance with reports on carbodiimide-mediated condensation reactions of phosphoric acid, linear and cyclic polyphosphates as well as mixtures of these, leading essentially to only 1,5-μ-oxo-tetrametaphosphoric acid. [59]Dianhydro ultraphosphonate (57) was precipitated as its TBA salt with minor decomposition (< 2 %) using Et 2 O.It contained residual urea byproduct from the condensing agent.Remarkably, precipitated 57 was not only soluble in DMF, but also in acetone and chloroform and showed no decay in further analyses of the solutions several days later.
The yield for 57 was determined by addition of a defined volume of aqueous phosphonoacetic acid solution and integration against the hydrolysis product anhydro ultraphosphonate (56).Using this method, we determined a yield of 85 % for precipitated 57.The latter could be stored at 20 °C with only minimal decomposition detected after two weeks.Using NaClO 4 solution in acetone for precipitation of 57 gives the corresponding sodium salt, which is only soluble in water. 31P{ 1 H}-NMR spectra immediately after dissolution of 57 sodium salt in water showed around 50 % of 57, which was consumed after 90 min.However, 56 was not the single hydrolysis product but dimerized anhydro ultraphosphonate 59 -as the product of a nucleophilic attack of anhydro ultraphosphonate 56 on dianhydro ultraphosphonate 57 -was formed as well with a ratio of 4 : 1 56 vs. 59 (Scheme 5).The rate of 59 formation was increased to ca. one third using a wet NaClO 4 solution in acetone (5 % water) for precipitation.After SAX purification, 59 was thus obtained in 48 % yield.The 31 P{ 1 H}-NMR spectrum is shown in dark blue in Scheme 5. To establish a protocol for the synthesis of anhydro ultraphosphonate 56 starting form 57, we attempted to avoid the formation of 59 but could only reduce its amount to 5-10 % by adding an acetone solution of 57 dropwise to water.Following this procedure, we obtained 65 % of 56 after SAX-purification (Scheme 5, in blue).In addition to water, propargylamine and AMP were found to be appropriate nucleophiles for ring-opening of 57.Of note, trials adding propargylamine together with the condensing agent and prior to completed formation of 57 suggest that ultraphosphonate first undergoes intramolecular dehydration followed by nucleophilic ring-opening rather than direct condensation with the amine.In accordance, we found no propargylamino-modified, non-cyclic ultraphosphonate in these reactions.Ring-opening of 57 to 60 was finished after two to three days when 50 equiv.propargylamine were used and the product 60 isolated in 58 % yield (Scheme 5, in green).Strikingly, 57 only underwent ringopening at the Q 2 phosphorus.This is in contrast to computational results for the "hydrolysable" analogue 1,5-μ-oxo-tetrametaphosphoric acid revealing a ca. 5 kcal/mol higher thermodynamic barrier for nucleophilic attack of a Q 2 vs. a Q 3 phosphorus and tetrametaphosphate structures as the kinetic products. [20]Formation of a modified tetrametaphosphonate by nucleophilic attack of the Q 3 phosphorus in 57 is however not possible due to the non-hydrolysable CH 2 bridge between the Scheme 4. Ultraphosphonate aryne reactions using Kobayashi-type aryne precursors.Crystal structure of 15 with ellipsoids drawn at 50 % probability level.two Q 3 phosphorus atoms.The enforced reactivity may be energetically unfavourable and could rationalize the required large excess of nucleophile.Further ring-opening of 60 to noncyclic bispropargylamino ultraphosphonate was not observed even with 100 equiv.propargylamine or at 80 °C.Taking AMP TBA salt as nucleophile gave only traces of 61, but addition of 1.5 equiv.MgCl 2 promoted the ring-opening in accordance with reports on similar reactions with cyclic ultraphosphates. [33]With Mg 2 + as additive, 2 equiv.AMP were sufficient to consume 57 within two to three days.After SAX-purification, the product 61 was obtained with 54 % yield (Scheme 5, in orange).
As already reported for ultraphosphonate, [60] such structures can act as potent chelating agents and the ability to complex different metals can be expected to be further increased for 63.Considering trisphospho ultraphosphonate (63) as model structure for ultraphosphates with elongated phosphate chains, this so far uncharacterized structural motif may have sufficient stability to become accessible as well. [39]s we observed a high tendency to form trimetaphosphonate (sub)structures (Scheme 3), we were interested to see whether a cyclic ultraphosphonate, containing a tetrametaphosphonate subunit, can be synthesized.We thus treated ultraphosphonate with 1.1 equiv. of phosphordiamidite 45 and detected dianhydro ultraphosphonate (57) and Fm-phosphonate after the coupling step (Scheme 6).Formation of 57 requires two condensation steps suggesting that a tetraphosphonate-phosphite 65 was successfully formed as intermediate.We propose that this mixed P(III)-P(V) anhydride undergoes intramolecular nucleophilic attack of one phosphonate of the ring onto another resulting in ring contracted 67 with a noncyclic P(III)-P(V)-anhydride.The resulting phosphite diester (or isomerized phosphonate) then acts as leaving group for a

Pull-down probes of linear and branched condensed phosph(on)ates
For the construction of a pull-down probe, anhydro ultraphosphonate (56) was reacted with alkyne-tagged aryne precursor 19 to enable further modifications.The product 68 (Scheme 7) was obtained in 14 % yield with a regioisomeric ratio of 2 : 1 (para vs. meta).As observed for metaphosphonate 18 as well (Scheme 2a), fluoride acted as nucleophile to partially open the ring-structure, explaining the low yield of the reaction.No further optimisation was conducted at this point.
Since ultraphosphate and linear tetraphosphate are constitutional isomers, the latter is an important negative control to determine, whether proteins were bound in a pull-down experiment due to a constitutional preference or just because of similar ionic interactions.The sufficient stability of tetraphosphate [61] rendered the synthesis of a non-hydrolysable analogue unnecessary.Propargyl phosphate (69) was triphosphorylated using cyclic pyrophosphoryl phosphoramidite ( 70, c-PyPA) [33,53] giving 59 % 71.

Pull-down experiments with yeast cell lysates
Photoaffinity capture compounds are typically used in combination with proteomics, which enables identification of interactors from complex mixtures such as cell lysates with high specificity.][64][65] The methodology described in those references is now adapted herein to pull down proteins binding to different condensed (cyclic or branched) phosphate analogues to delineate the elusive interactome of non-linear polyphosphates.The pull-down experiments were conducted with yeast cell extracts, since inorganic polyphosphate is highly abundant in yeast [66] and the enzymology is well described. [10,67]6][7]61] To study interactomes related to phosphate supply, yeast may be cultured under phosphate depletion, resupply conditions or in phosphate-rich media.We decided to prepare lysates from BJ3505 WT with no phosphate present in the culture medium.Future studies could be designed to investigate the interactome under alternative conditions.
The experimental procedure of the photoaffinity pull-down experiments is shown in Scheme 8a: each capture compound was initially incubated with streptavidin-coated magnetic beads to ensure high loading (step 1).This procedure was developed using SDS-PAGE control experiments that showed in general higher protein enrichment as the reverse process: incubating the probe first with the lysate and then with the beads.Therefore, in the second step, the beads were incubated with the yeast cell lysate either in absence (pull-down experiment) or presence of ca.300-fold excess of competitor (competition control).The competition control should reduce false-positive hits as the high excess of the unmodified structure of interest (competitor) should preferentially occupy the active site of interactors suppressing their interaction with the capture compound while unspecific interactions are not prevented.False-positive hits of the pull-down experiment can thus be reduced by comparison with the results obtained in absence of the competitor.The mixture was irradiated with UV light in step 3 to induce photo cross-linking by transformation of the phenyl azide into a reactive nitrene.The latter can undergo C H insertion forming covalent bonds with other molecules in direct vicinity -usually an interactor of the selection function.This stable linkage then enables stringent washing of the beads, reducing false-positive hits (step 4).Each experiment was performed in duplicate and one sample was used for trypsination and LC-MS/MS analysis while the other sample was subjected to SDS-PAGE analysis to validate the procedure by comparison of the number and intensity of obtained bands.
Without competitor, distinct bands should show increased intensity indicating enrichment of putative interactors whereas the competition control reveals unselectively bound proteins.
SDS-PAGE analyses of the initial experiments indicated strong unspecific binding since the pull-down experiment and competition control only showed slight differences and similar bands were obtained for all capture compounds.Importantly, the negative control containing no capture compound showed the same bands as the pull-down experiment pointing towards a loss of the selection function.We assumed that high concentrations of reduced glutathione [68] caused cleavage of the disulfide bridge in the capture compounds.Reductive cleavage of the commercial BED linker was undesired in our experiments.We therefore performed further pull-down experiments under oxidative conditions by incubation of the lysate with 0.1 % H 2 O 2 , which should convert glutathione to its oxidized form and thus ensure stability of the capture compound.Since only the lysate is treated with H 2 O 2 , the proteome remains unaffected by upregulation or repression of gene transcription associated with oxidative stress response. [69]o test our hypothesis, we performed initial pull-down experiments under oxidative conditions with 37. Scheme 8b shows the result for the corresponding SDS-PAGE analysis.Additional bands for the pull-down experiment (lane 2) compared with the negative control containing no 37 (lane 5) now suggested retention of the selection function.The competition control (lane 3) showed less bands indicating selective protein binding and strong intensity differences point towards successful enrichment of proteins.The overall lower intensity of the bands in lane 4 -corresponding to a sample, which was not UVirradiated -furthermore suggest successful photo cross-linking.We thus adapted the optimized procedure to the further capture compounds 75-77.SDS-PAGE analysis indicated selective binding of proteins albeit differences between number and intensity of bands were weak (Figure S1).Similar results for the different capture compounds may arise from their structural relationship but could also indicate general binding of proteins to densely charged structures.Since interacting proteins may only show weak bands, their determination by PAGE analysis is hampered impeding further evaluation by this analytical method.Therefore, we analysed the samples by on-bead trypsin digestion and LC-MS/MS analysis.The data are available via ProteomeXchange with the dataset identifier PXD043919.We chose log 2 (enrichment) > 1.5, corresponding to a ca.3-fold increase of protein in the pull-down compared to the competition control experiment, and razor unique peptides 3 as threshold for considering a protein as hit.258 proteins were identified for 37 matching this criteria, 35 for 75, 115 for 76 and 11 for 77.The proteins are listed in the Supporting Information.Exemplarily, Scheme 8c shows selected hits for capture compound 37. Results, which were obtained for both 37 and 76 are marked in orange in Scheme 8c; common results for 37, 76 and 75 are indicated in green.Due to the close structural relation, overlapping hits were expected and support the quality of the data.In line, the results for linear tetraphosphate differed markedly.However, it is unclear whether this can be assigned to structural preferences or if the low number of captured proteins is a result of partial degradation of the "hydrolysable" capture compound 77 in the cell extract.We did not use phosphatase inhibitors to avoid blocking interaction sites our probes could bind to.
Although methylene-substitution is generally accepted as a good bioisosteric replacement for phosphoanhydride bonds, [25] different bond lengths and angles -as determined from crystallographic data for 35 as well (Scheme 2c) -, reversed polarity of the CH 2 -modification [24] or lacking hydrogen bonds may lower binding affinities.This is exemplified by a report of Blackburn that yeast exopolyphosphatase (PPX) only bound adenosine tetraphosphate with native P O P bridges. [70]Yet, in our example, PPX was captured by pull-down probe 37 that contains methylene bridges.The susceptibility of metaphosphate esters to linearization and the sensitivity of ultraphosphates towards several conditions obviously require stabilization of anhydride bonds for development into affinity reagents.The pull-down data provided here serve as valuable starting point for further investigations, while additional efforts should be made to establish new methods, compatible with the reactivity and properties of non-linear condensed phosphates.

Conclusions
In the present paper, we disclosed syntheses of capture compounds containing non-hydrolysable analogues of cyclic and branched condensed phosphates as selection function and applied them in photoaffinity pull-down experiments with yeast cell extracts.Along the way, we developed new phosphonate chemistry and studied the transformations that occur in these systems.
We found that methylene-bridged trimetaphosphate analogues are poor substrates for aryne chemistry due to several side reactions and that arylesters at the phosphinate function of a bismethylene trisphosphate are prone to hydrolysis.Therefore, an alkyne tagged bis(chloromethyl)phosphine oxide was prepared and further reacted to the bismethylene triphosphate analogue using an Arbuzov reaction.After condensation to the corresponding trimetaphosphonate, we detected only little hydrolytic ring-opening after several days in solution allowing the development into pull-down probes.Efforts to prepare a tetrametaphosphonate sample revealed a strong preference to form trimetaphosphonate structures by either cyclization or ring contraction.Similar results were obtained for ultraphosphonate, which only formed a transient tetrametaphosphonate subunit and underwent fast consecutive reactions to form dianhydro ultraphosphonate.The latter was selectively accessible using DIC for condensation of ultraphosphonate and could be ring-opened by water to yield anhydro ultraphosphonate or amine and phosphate nucleophiles.Ultraphosphonate and anhydro ultraphosphonate were successfully functionalized using aryne chemistry and further transformed into pull-down probes.
Pull-down experiments with yeast cell extracts were performed under oxidative conditions.With the chemical proteomics approach, we could identify numerous putative interactors of non-linear polyphosphates providing a first entry into their elusive interactomes.Our data serve as a starting point for further investigations giving prospects to not only identify cyclic and branched phosphates as enzyme substrates but also their non-hydrolysable analogues as potential inhibitors.Enzymatic reactions including four phosphates may be of special interest as they could potentially proceed via a phosphate walk-like reaction [27,39] -a so far unconsidered reactivity, which would involve branched condensed phosphates as intermediates.Concerning metaphosphates, the formation of aminoacyladenylate analogues in presence of trimetaphosphate has already been reported under prebiotically plausible conditions. [71]This supports the idea that trimetaphosphate could indeed act as substrate for tRNA ligases, of which we identified several in this pull-down (Scheme 8c and Table S1).While these examples only illustrate considerations about the plausibility of the obtained data, the latter may aid to comprehensively study the potential biology of non-linear polyphosphates.Based on recent reports on the presence of metaphosphates in organisms and the interaction of synthetic ultraphosphates with alkaline phosphatase, we are confident that a rich biology of cyclic and branched condensed phosphates awaits its discovery.

Experimental Section
General experimental remarks, detailed synthetic procedures and analyses by NMR and mass spectrometry are described in the Supporting Information.
General procedure for dealkylation using TMSBr: A solution of a triphosphonate ethyl ester (1.0 equiv.) in CH 2 Cl 2 (100 mm) was treated with a solution of TMSBr (10 equiv.)and pyridine (25.0 equiv.) in CH 2 Cl 2 (ca. 3 m referring to pyridine) at 0 °C for 5 h.The mixture was allowed to slowly reach room temp.and further stirred overnight.Water was added, the aqueous layer separated and evaporated to dryness.The residue was dissolved in water and the pH adjusted to 8.5 using NaOH solution (1 m).It was purified by AIEX chromatography (Q Sepharose® Fast Flow, increasing concentrations of NaClO 4 ).The product was precipitated using ice-cooled acetone (product fraction/acetone: 1 : 9 v/v), washed with acetone twice and dried in vacuo.
Crystallographic data: Deposition number(s) 2127320 (for 35), 2100905 (for 15), and 2114836 (for S11) contain(s) the supplementary crystallographic data for this paper.These data are provided free of charge by the joint Cambridge Crystallographic Data Centre and Fachinformationszentrum Karlsruhe Access Structures service www.ccdc.cam.ac.uk/structures..

Pull-down experiments:
The procedure was adapted from Jenal. [62,64,65]Cell extracts from yeast BJ3505 WT were grown with no P i present in the medium.Experiments performed in duplicates.Details on the used buffer solutions are described in the Supporting Information.

Capture solutions
Prepare solutions of protein (ca.18 mg protein) with 300 μl 5×capture buffer, with and without 15 μl competitor (40 mm stock).Dilute with water to 1403 μl.Incubate at 4 °C for 30 min on a rotating wheel.

Mix and bind
Add the protein/competitor mixture to the beads/capture compound suspension.
Incubate at 4 °C for 2 h on a rotating wheel.Centrifuge briefly, resuspend and transfer samples to a 12-well plate.

Cross-link
Cross link in 12-well plate on ice under xenon light for 2 min.
Transfer sample to Eppendorf tube on magnetic rack.Collect the supernatant.Rinse the well with 1 ml 1×washing buffer and add solution to the beads.

Scheme 1 .
Scheme 1. Approaches for modified metaphosphates and non-hydrolysable analogues of cyclic and branched condensed phosphates.a) Modification of metaphosphates using aryne chemistry or by activation with different leaving groups and subsequent nucleophilic reaction.Hydrolytic linearization of modified metaphosphates.b) Reaction design for modified meta-and ultraphosphonates employing Arbuzov and condensation reactions respectively aryne chemistry.