N-donor Base Adducts of P2O5 as Diphosphorylation Reagents

Diphosphorylation of select nucleophiles was achieved by treatment with neutral, doubly zwitterionic adducts of nucleophilic N-donor bases (pyridine, DABCO) and P2O5. Treatment of the PPN salt of TriMP with TsCl and triethylamine yielded a mixed PPN and triethylammonium salt of tosyl TriMP ([PPN][HNEt3][P3O9Ts], 1), isolated in 55% yield and crystallographically characterized. Treatment of 1 with two equivalents of DABCO led in situ to an unisolated ring opened intermediate, an adduct of two DABCO molecules and (P3O8) – ([DABCO-P3O8-DABCO], 2). Reaction of 2 with additional DABCO results in precipitation of a neutral, doubly zwitterionic adduct of DABCO and P2O5 (DABCO-P2O5-DABCO, 3). Compound 3 is more conveniently prepared by treatment of polymeric phosphorus pentoxide ((P2O5)∞) with DABCO in acetonitrile. Both syntheses also result in an inseparable neutral triply zwitterionic impurity, an adduct of three molecules of DABCO and P4O10 (P4O10(DABCO)3, 4). Both 3 and 4 were crystallographically characterized by careful selection of crystals for X-ray diffraction. A similar neutral, doubly zwitterionic adduct of pyridine and P2O5 (Pyr-P2O5-Pyr, 5) was prepared by treatment of (P2O5)∞ with pyridine in 77% yield. Due to higher purity, 5 was used for further reactivity studies. Compound 5 reacts with DBU (6, 65% yield), diethylamine (7, 77% yield), and methanol (8 and 9) to producing the corresponding difunctionalized pyrophosphate derivatives. Compound 5 also reacts with the TBA salt of AMP to produce adenosinesubstituted TriMP (10) as an intermediate that was ring opened with water, diethylamine, and the TBA salt of UMP, yielding ammonium salts of ATP (11, 89% yield), γ-diethylamino-ATP (12, 58% yield), and adenosine uridine tetraphosphate (13, 74% yield) after purification by AX-HPLC. Similarly, adenosine tetraphosphate (14) was prepared in 40% yield by treatment of 5 with ADP and hydrolysis of the resulting intermediate. The last several years have seen an explosion of new reagents and methodologies for the synthesis of oligophosphorylated molecules, due to their prevalence in biological systems. This field has slowly developed since Khorana’s pioneering first synthesis of adenosine triphosphate (ATP) in 1954, and an improved synthesis of ATP and of other nucleoside triphosphate (NTP) derivatives has been a major area of focus in synthetic oligophosphorylation. Many of the recent developments have focused on reagents for the simultaneous addition of three phosphate units, via trimetaphosphate ((P3O9) , TriMP) in order to minimize the number of synthetic steps to convert a nucleoside into an NTP. The first effective TriMP based methodology was that of Taylor, utilizing the activation of TriMP with sulfonyl chloride reagents. These seminal papers have been followed up more recently with our own group’s as well as Jessen’s isolable triphosphorylation reagents. By utilizing substrates P O O O P O O O P O O O TsCl, NEt3 P O O O P O O O P O O O S O O

The last several years have seen an explosion of new reagents and methodologies for the synthesis of oligophosphorylated molecules, due to their prevalence in biological systems. [1][2][3][4][5][6] This field has slowly developed since Khorana's pioneering first synthesis of adenosine triphosphate (ATP) in 1954, 7 and an improved synthesis of ATP and of other nucleoside triphosphate (NTP) derivatives has been a major area of focus in synthetic oligophosphorylation. 8 Many of the recent developments have focused on reagents for the simultaneous addition of three phosphate units, via trimetaphosphate ((P 3 O 9 ) 3-, TriMP) in order to minimize the number of synthetic steps to convert a nucleoside into an NTP. 9 The first effective TriMP based methodology was that of Taylor, utilizing the activation of TriMP with sulfonyl chloride reagents. 4,10-13 These seminal papers have been followed up more recently with our own group's 1 as well as Jessen's 3,14 isolable triphosphorylation reagents. By utilizing substrates B. X-ray crystal structure of 1 with thermal ellipsoids set at 50% probability and cations omitted (red for oxygen, orange for phosphorus, yellow for sulfur, grey for carbon). C. X-ray crystal structure of 3 with thermal ellipsoids set at 50% probability and cations omitted (red for oxygen, orange for phosphorus, blue for nitrogen, grey for carbon).
such as nucleosides, nucleoside monophosphates, and nucleoside diphosphates, these reagents are able to add three phosphate units to form nucleoside triphosphates, tetraphosphates, and pentaphosphates, respectively. Our group has sought to further expand this methodology of well defined, isolable oligophosphorylation reagents to introduce a varying number of phosphate units in one transformation. To that end, we have reported a tetraphosphorylation reagent 15 and its utility to synthesize nucleoside tetra-and pentaphosphates. 2,16 Here we further develop this strategy by reporting a potent diphosphorylation reagent and subsequent synthesis of NTP derivatives, a dinucleoside tetraphosphate, and a nucleoside pentaphosphate. The current investigation began by repeating the earlier triphosphorylation work of Taylor where the TBA (tetrabutylammonium) salt of TriMP is activated with mesitylenesulfonyl chloride and nucleophilic N-donor bases, such as DABCO (1,4-diazabicyclo[2.2.2]octane), pyridine, and N -methylimidazole. 4,[10][11][12][13] The putative mesitylenesulfonyl TriMP species was unisolated in Taylor's reports. However, in an analogous reaction of the PPN (bis(triphenylphosphine)iminium) salt of TriMP and tosyl chloride we were able to isolate and crystallographically characterize tosyl TriMP (1, Figure 1). Performing this reaction in dried acetone under anhydrous conditions results in precipitation of anhydrous crystals of a mixed PPN and triethylammonium salt of 1.
Taylor's phosphorylation methodology additionally involves several amine additives, notably DABCO. [11][12][13] Treating TriMP with tosyl chloride and DABCO in a one pot procedure gives a complex mixture of products (see S2.3), but isolated 1 reacts cleanly with two equivalents of DABCO to form a ring opened species, an adduct of (P 3 O 8 ) 3and two DABCO molecules (2, Figure 1A), as determined by 31 P{ 1 H} NMR. The utility of 1 and 2 as phosphorylation reagents is however hindered by their instability. Even in the crystalline state at −35 • C, 1 shows significant decomposition to a complex mixture after two weeks, and we were unable to isolate 2 without decomposition to a complex mixture of species (See SI).
Serendipitously, an acetonitrile solution of 2 in the presence of excess DABCO formed colorless crystals upon standing overnight. An X-ray diffraction study of these crystals revealed a neutral, doubly zwitterionic adduct of P2O5 and two molecules of DABCO (3, Figure 1C). The structure of 3 shows a P-N bond distance of 1.84 Å, in close agreement with an expected P-N single bond length of 1.82 Å. 19 Formation of 3 by reaction of 2 with DABCO formally produces an equivalent of the monoanionic adduct of DABCO and [PO3] − . However, 31 P{ 1 H} NMR of the supernatant solution after precipitation of 3 shows a complex mixture of oligophosphate species. The corresponding quinuclidine adduct of metaphosphate has been transiently observed in aqueous solution by 31 P{ 1 H} NMR and was found to rapidly convert to TriMP. 20 However, TriMP is not a significant component of the mixture of phosphates we observed. Additionally, 3 is only poorly soluble in dichloromethane and chloroform, but 31 P{ 1 H} NMR spectroscopy indicated the  presence of an impurity which we assigned as 4 but were unable to separate (see below, Figure 3). Interested in the potential phosphorylation reactivity of 3, we sought to develop an easy and large scale synthesis of of this compound for further reactivity studies.
A dianionic bis-phosphorimidazolide has been reported by Wright and is facilely prepared by treatment of pyrophosphate with carbonyldimidazole (CDI) (Figure 2). 17 This compound has been investigated for diphosphorylation activity, but imidazole groups are only moderate leaving groups. Nevertheless, the imidazole adduct reacts slowly with nucleoside monophosphates, a reaction that is greatly accelerated in the presence of promoters such as ZnCl2, to give symmetric dinucleoside tetraphosphates ( Figure 2). The only previous report of a neutral doubly zwitterionic adduct of P2O5 and an N-donor base is the pyridine adduct reported by Meisel, 18 (5) generated by refluxing phosphorus pentoxide ((P2O5)∞) in pyridine ( Figure 2). However, this putative pyridine adduct was not characterized or investigated for phosphorylation reactivity. We therefore sought to characterize this pyridine adduct and adapt this synthetic method to 3. Notably, Wright patented their bisphosphorimidazolide compound and included the pyridine adduct, 5, as well as a proposed N-methylimidazole adduct in their application. 21 However, they do not appear to have actually synthesized or explored the reactivity of either of 6   Treating a mixture of 3 and 4 with methanol yields a mixture of 8 and 9. B. X-ray crystal structure of 7 with thermal ellipsoids set at 50% probability and cations omitted (red for oxygen, orange for phosphorus, blue for nitrogen, grey for carbon).
these compound. Treatment of a suspension of (P2O5)∞ in acetonitrile with excess DABCO leads to rapid formation of a precipitate consisting of a mixture of 3 and 4 ( Figure 3A). Conducting this reaction at suitable dilution results in slow precipitation of crystalline solids and the ability to filter out any insoluble impurities. However, the resulting product was always found to be a mixture of 3 and 4. Extensive attempts to vary conditions or effect purification of the mixture resulted in limited impact on the ratio of the two products. However, careful selection of a crystal for X-ray diffraction provided structural analysis of 4 ( Figure 3B), validating our initial assignment of this impurity. Both 3 and 4 can be viewed as two different neutral building blocks of (P2O5)∞ that have been cut out of the polymeric structure 22 by DABCO.
Due to our inability to purify 3, we turned our investigation to Meisel's 18 analogous pyridine adduct, 5 ( Figure  4). Heating (P2O5)∞ and pyridine in acetonitrile to near reflux for 24 hours results in a fine white precipitate. The product is very poorly soluble in pyridine, chloroform, and dichloromethane, but multinuclear NMR spectroscopy gives no evidence for the pyridine analog of 4. However, a disparity between expected and observed elemental analysis data for 5 suggests the possibility of impurities analogous to 4 that are too insoluble to observe. The synthesis of 5 is easily performed on a 5 gram scale in 77% yield from two very inexpensive and commercially available reagents, providing trivial access to this diphosphorylation reagent. In the pursuit of a more soluble and spectroscopically pure N-donor base adduct of P2O5, we similarly investigated the reaction of (P2O5)∞ with triethylamine, tributylamine, 4-tertbutylpyridine, and DMF (dimethylformamide). A rapid reaction was observed in all cases, yielding more soluble products than in the case of DABCO and pyridine (see SI). However, these reactions all produced significant side products, and we chose 5 to explore further phosphorylation chemistry. Treatment of 5 with DABCO led to rapid formation of a mixture of 3 and 4. To investigate whether 4 is formed in this reaction or if an insoluble pyridine analog was present in the starting 5, we also treated 5 with DBU. This reaction resulted in clean formation of the DBU adduct, 6 ( Figure  4A). Compound 6 is freely soluble in DCM and contains no significant impurities. However, we observed no reactivity between 6 and simple nucleophiles such as methanol and O-phenylenediamine.
Combining a suspension of 5 in DCM with a slight excess of diethylamine yields 7 in less than five minutes ( Figure  4A). A simple workup provides the diethylammonium salt of 7, and we were able to obtain a crystal structure of this compound ( Figure 4B). We were also able to prepare 7 by treatment of a mixture of 3 and 4 with diethylamine, although in a lower isolated yield (see SI).
Phosphorylation of alcohols with 5 was less facile. Treatment with an excess of methanol gives rise to the expected dimethyl pyrophosphate compound, 8, as well as a trifunc-tionalized tetraphosphate derivative 9 ( Figure 4A). The corresponding diethylamino derivative of 9 was also observed as a minor impurity in the crude reaction of 5 with diethylamine, but this impurity was easily removed due to poor solubility in cold acetonitrile. However, we were unable to develop conditions for purification of 8 from 9. Treatment of 5 with more sterically demanding alcohols such as neopentanol and adenosine yielded only complex mixtures of products.
Compound 5 reacts rapidly with phosphate nucleophiles, and treatment with the TBA salt of adenosine monophosphate (AMP) leads to adenosine-substituted TriMP (10) in less than five minutes with only minor impurities observed in the 31 P{ 1 H} NMR spectrum. This transformation is reminiscent of the Khorana 7 and Ludwig-Eckstein 23 reactions which append pyrophosphate to form a nucleoside substituted TriMP intermediate. However, the present reaction is markedly cleaner and more selective than those canonical reactions, whose difficult chromatographic separations have spawned decades of research. 6 Additionally, this reactivity of neutral doubly zwitterionic 5 is in stark contrast to the slow reaction of the dianonic imidazole adduct of P2O5 with AMP to form diadenosine tetraphosphate. 17 Intermediate 10 is the same compound observed in Taylor 13 and Jessen's 3,14 syntheses of NTPs and is therefore amenable to a variety of ring opening reactions, which were briefly investigated in one-pot procedures starting from 5. Treatment of 10 with water results in ring opening to ATP in seconds; the ATP was purified by anion exchange HPLC (AX-HPLC) yielding the ammonium salt of ATP(11, Figure  4B). Similarly, treatment of 10 with diethylamine resulted in γ-diethylamino-ATP, which was purified by AX-HPLC (12, Figure 4B). Additionally, 10 can be ring opened by treatment with UMP in the presence of MgCl2. This results in selective formation of adenosine uridine tetraphosphate (13) isolated as its ammonium salt after purification by AX-HPLC. A number of syntheses of dinucleoside tetraphosphates have been reported, 24 and this synthesis, as well as Wright's 17 are convenient due to their high yields and utilization of NMP substrates rather than more expensive and difficult to synthesize NTPs. Furthermore, the present synthesis provides convenient access to unsymmetrical dinucleoside tetraphosphates, which is not possible with Wright's reagent. Lastly, longer nucleoside oligophosphates are competent phosphorylation substrates, albeit with reduced yield, and treatment of 5 with the TBA salt of ADP and subsequent hydrolysis yields adenosine tetraphosphate (14).
Acknowledgement We thank Dr. Peter Müller for help collecting X-ray crystal data. Research reported in this publication was supported by the National Institutes of Health under award number R01GM130936.

Supporting Information Available
See the supporting information for experimental details, further characterization data, and X-ray crystallographic information. X-ray crystallographic data for C49H53N2O11P5S (CCDC 2004821) X-ray crystallographic data for C12H24N4O5P2 (CCDC 2071686) X-ray crystallographic data for C22.50H46.50N7O10P4 (CCDC 2059218) X-ray crystallographic data for C16H44N4O5P2 (CCDC 2071687)

General Considerations
All chemical syntheses were performed in a Vacuum Atmospheres model MO-40M glovebox under an inert atmosphere of purified nitrogen unless otherwise stated. All solvents used in the glovebox were obtained anhydrous and oxygen-free by bubble degassing with argon and purification through columns of alumina and Q5 by the method of Grubbs, except for water and acetone. 1 HPLC grade water was purchased from Sigma Aldrich and used as received. Dried acetone was prepared by stirring overnight over anhydrous calcium sulfate followed by distillation under an anhydrous nitrogen atmosphere from fresh calcium sulfate. 2 The tetrabutylammonium (TBA) salts of 5'-UMP, 5'-AMP, and 5'-ADP were prepared according to the procedure of Taylor, 3 by ion exchange on Dowex 50WX8 (H+) followed by neutralization with TBA hydroxide and subsequent lyophilization. The resulting TBA nucleotide salts were stored as either DMF or acetonitrile stock solutions (0.10 g/mL) over 4Å molecular sieves (dried under vacuum at 200 • C for 24 hours) for at least 24 hours prior to use. 4 was prepared according to the literature procedure. All other reagents were purchased and used as received. Deuterated solvents were purchased from Cambridge Isotope Labs and used as received. NMR spectra were obtained at ambient temperature (ca. 25 • C) on a Bruker Avance 400 instrument. 1 H and 13 C NMR spectra were referenced internally to residual solvent signals. 31 P NMR spectra were referenced internally to the PPN signal or externally to 85% phosphoric acid. Electrospray ionization mass spectra (ESI-MS(−)) were acquired on a Micromass Q-TOF ESI spectrometer. Samples were prepared in acetonitrile or acetonitrile/water at an approximate concentration of 20 ng/µL, and a source temperature of 80 • C and desolvation gas temperature of 400 • C were used. Elemental analyses were performed by Midwest Microlab. HPLC was performed with Waters 515 HPLC pumps coupled to a Waters 2487 UV absorbance detector and a PRP-X100 anion exchange column. HPLC purifications were carried out with a gradient method from 100% A, 0% B to 0% A, 100% B (A = 100% water; B = 1M (NH 4 )(HCO 3 ) in water) with a 2.5 mL/min flow rate. The HPLC purifications were monitored by absorbance at 260 nm. S1 2 Synthesis and Characterization of Compounds  In a glovebox, 4 (1.36 g, 0.72 mmol) was suspended in 10 mL of anhydrous acetone. (Note: At a concentration less than 1 g/mL, precipitation does not occur in the following step.) To this stirring suspension was added triethylamine (0.20 mL, 1.43 mmol) followed by a solution of TsCl (0.27 g, 1.43 mmol) in anhydrous acetone (2 mL). The solution was stirred vigorously for 15 min. After the addition of reagents, the mixture briefly became homogeneous before a thick precipitate formed. The mixture was then filtered through Celite® to remove the precipitate. The resulting colorless solution was left undisturbed at ambient temperature for 12 hours, resulting in precipitation of colorless crystals. The supernatant solution was decanted off the crystals, which were then rinsed three times with 1 mL of acetone. The resulting crystals were dried under vacuum, yielding the product as a white, crystalline solid (0.41 g, 0.39 mmol, 55% yield). The product was stored in the glovebox freezer at −35 • C to prevent decomposition.
Elem    Figure S8: 31 P{ 1 H} NMR spectrum of the solid obtained after precipitating the crude reaction mixture containing 2 with diethyl ether (acetonitrile-d 3 , 202 MHz).

Attempted formation of 2 in one pot from TriMP
The reagents used for the synthesis of 2 from TriMP were combined in one pot rather than proceeding through isolated 1. This is analogous to the procedure employed by Taylor. 5 In a glovebox, a stock solution was prepared of [TBA] 3 [P 3 O 9 ] · H 2 O in acetonitrile at a concentration of 0.10 g/mL. This solution was stored over activated 4Å sieves for at least 24 hours. To this solution of [TBA] 3 [P 3 O 9 ] · H 2 O (0.50 mL, 0.052 mmol) was added triethylamine (0.015 mL, 0.10 mmol). To this mixture was then added a solution of MstCl (0.011 g, 0.052 mmol) in acetonitrile (0.5 mL) followed by a solution of DABCO (0.023 g, 0.21 mmol) in acetonitrile (0.5 mL). The resulting mixture was mixed and analyzed by 31 P{ 1 H} NMR. 31 P{ 1 H} NMR analysis reveals a complex mixture of phosphates ( Figure S9). Dominant, assignable species are 2 and TriMP. The majority of the other phosphorus environments correspond to two or more species with chemical shifts similar to that of 1. One of these species may correspond to the Mst analog of 1. Another possible species is µ-oxo-bis- There are also numerous difficult to assign species with small integrations. While this mixture appears to contain 2, the numerous other species likely result in side products when this mixture is used for phosphorylation reactions.

Synthesis of 3 and 4
The following three synthetic methods give spectroscopically identical mixtures of 3 and 4, differing only in the ratio of the two products.          In the glovebox, a Schlenk tube was charged with (P 2 O 5 ) ∞ (0.19 g, 1.4 mmol) and acetonitrile (10 mL). To this suspension was then added pyridine (1.1 mL, 14 mmol). The vessel was then sealed and removed from the glovebox. The vessel was then placed in an oil bath and the mixture stirred at 70 • C overnight. The vessel was then cooled to room temperature and returned to the glovebox. The fine white precipitate was collected by filtration and rinsed thoroughly with acetonitrile. Drying the solid under vacuum yielded the product as a free flowing white powder (0.36 g, 1.1 mmol, 82% yield). The resulting product is poorly soluble in dichloromethane, chloroform, and pyridine (approximately 1 to 5 mg/mL) and not appreciably soluble in acetonitrile.

Treatment of (P 2 O 5 ) ∞ with Various N-donor Bases
In a general experiment, (P 2 O 5 ) ∞ (0.050 g, 0.35 mmol) was suspended in acetonitrile (5 mL), in the glovebox. To this stirring suspension was added a solution of an N-donor base (1.4 mmol) in acetonitrile (2 mL) and the resulting mixture stirred for 5 min. During this time, most of the suspended material dissolved. The resulting solution was then analyzed by 31 P{ 1 H} NMR. The N-donor bases tested were triethylamine, 4-tert-butylpyridine, tributylamine, and dimethylformamide.       In the glovebox, 5 (0.11 g, 0.36 mmol) was suspended in dichloromethane (4 mL). To this stirring suspension was added DBU (0.22 g, 1.5 mmol) in dichloromethane (2 mL). The resulting suspension was stirred for 30 min, resulting in dissolution of most of the solid material. The resulting mixture was filtered through Celite® to remove insoluble material. All volatile materials were then evaporated under reduced pressure, leaving a pale yellow solid. The solid residue was then dissolved in acetonitrile (2 mL) and diethyl ether (15 ml) was added, resulting in a cloudy suspension. The next day, the mixture had settled into a polycrystalline solid. The supernatant was decanted off the crystals, and the solids were rinsed with diethyl ether. The resulting solid was dried under vacuum, giving the product as a white powder (0.10 g, 0.23 mmol, 65% yield). 1 Figure S25: 13 C NMR spectrum of 6 (chloroform-d, 100.6 MHz).  Treatment of either a mixture of 3 and 4 or 5 with diethylamine leads immediately to 7 which is purified with a simple workup.  In the glovebox, 5 (0.11 g, 0.36 mmol) was suspended in dichloromethane (5 mL) in a Schlenk flask. The flask was then removed from the glovebox and attached to a Schlenk line. Against a flow of nitrogen, methanol (0.74 mL, 18.2 mmol) was added by syringe all at once. The mixture was stirred for 15 min and then filtered to remove insoluble material. Removing volatiles under reduced pressure left a colorless oil (0.066 g, 2:1 ratio of 8 to 9 by 1 H NMR integration).         In a glovebox, 5 (0.0097 g, 0.032 mmol) was suspended in acetonitrile (1 mL). To this stirring suspension was added a 0.10 g/mL solution of (TBA) 2 pA in acetonitrile (0.27 mL, 0.032 mmol). After stirring for 5 min, all of the solid material had dissolved. Volatiles were then evaporated under reduced pressure, leaving an oily solid. This residue was then removed from the glovebox and dissolved in deionized water (2 mL). The resulting solution was then filtered through an Acrodisc 0.2 µm wwPTFE syringe filter and purified by AX-HPLC according to the procedure in the General Considerations. The fractions containing the product were pooled and lyophilized. The resulting solid was again dissolved in deionized water and lyophilized giving the product as a fluffy white solid (0.016 g, 0.029 mmol, 89% yield).
The experimental characterization data for [NH 4 ] 4 [11] is in good agreement with the literature data for ATP.   In a glovebox, 5 (0.012 g, 0.038 mmol) was suspended in acetonitrile (1 mL). To this stirring suspension was added a 0.10 g/mL solution of (TBA) 2 pA in acetonitrile (0.31 mL, 0.038 mmol). After stirring for 5 min, all of the solid material had dissolved. To this solution was then added diethylamine (0.014 g, 0.19 mmol) in acetonitrile (1 mL). The resulting mixture was stirred for 6 hours, then volatiles were evaporated under reduced pressure, leaving an oily solid. This residue was then removed from the glovebox and dissolved in deionized water (2 mL). The resulting solution was then filtered through an Acrodisc 0.2 µm wwPTFE syringe filter and purified by AX-HPLC according to the procedure in the General Considerations. The fractions containing the product were pooled and lyophilized. The resulting solid was again dissolved in deionized water and lyophilized giving the product as a fluffy white solid (0.013 g, 0.022 mmol, 58% yield). After repeated lyophilizations of the product, the P-N bond was found to undergo hydrolysis, giving ATP as the primary phosphorus containing species.   In a glovebox, 5 (0.011 g, 0.036 mmol) was suspended in acetonitrile (1 mL). To this stirring suspension was added a 0.10 g/mL solution of (TBA) 2 pA in acetonitrile (0.30 mL, 0.036 mmol). After stirring for 5 min, all of the solid material had dissolved. Volatiles were then evaporated under reduced pressure, leaving an oily solid. This residue was then dissolved in DMF (1 mL). To this solution was added MgCl 2 (0.0068 g, 0.072 mmol) and a 0.10 g/mL solution of (TBA) 2 pU in DMF (0.35 mL, 0.043 mmol). The resulting solution was stirred for 24 hours. Volatiles were then removed under reduced pressure at 40°C leaving an oily solid. This residue was then removed from the glovebox and dissolved in deionized water (2 mL). The resulting solution was then filtered through an Acrodisc 0.2 µm wwPTFE syringe filter and purified by AX-HPLC according to the procedure in the General Considerations. The fractions containing the product were pooled and lyophilized. The resulting solid was again dissolved in deionized water and lyophilized giving the product as a fluffy white solid (0.023 g, 0.027 mmol, 74% yield

General Considerations
Crystals were mounted in hydrocarbon oil on a nylon fiber. Low-temperature (100 K) data were collected on a Bruker-AXS X8 Kappa Duo diffractometer coupled to a Smart Apex2 CCD detector with Cu κ α radiation (λ = 1.54178 Å) or Mo κ α radiation with ω-and φ-scans.

S41
A semi-empirical absorption correction was applied to the diffraction data using SADABS. 8 All structures were solved by intrinsic phasing using SHELXT 9 and refined against F 2 on all data by full-matrix least squares with SHELXL-2015 10 using established methods. All non-hydrogen atoms were refined anisotropically. All hydrogen atoms were included in the model at geometrically calculated positions and refined using a riding model unless otherwise noted. The isotropic displacement parameters of all hydrogen atoms were fixed to 1.2 times the U eq value of the atoms they are linked to (1.5 times for methyl groups). Descriptions of the individual refinements follow below and details of the data quality and a summary of the residual values of the refinements for all structures are given. Further details can be found in the form of .cif files available from the CCDC. were grown by spontaneous precipitation from the acetone reaction mixture, as described in the synthetic section. The crystal was found to be significantly twinned. For the refinement, only the primary unit cell was used. There is one anomalous Q peak of about 1.4 electrons, near a phenyl ring of PPN. This was ignored. Additionally, checkcif notes an unusually low C-C bond precision. Therefore, this structure is best used to illustrate connectivity rather than to provide high precision bond metrics. The tosyl group was found to be disordered over two positions. The unusual Q peak could theoretically be a partially occupied water molecule.  Figure S63: Single crystal X-ray Structure of 3 with thermal ellipsoids set at 50%.

X-Ray Diffraction Study of 3
Colorless crystals of a mixture of 3 and 4 were prepared by spontaneous precipitation from the reaction medium as described above. Carefully selecting crystals from the mixture allowed for an X-ray diffraction study of each component of the mixture. 3.4 X-Ray Diffraction Study of 4 Figure S64: Single crystal X-ray Structure of 4 with thermal ellipsoids set at 50% probability and hydrogen atoms omitted for clarity.

S44
Colorless crystals of a mixture of 3 and 4 were prepared by spontaneous precipitation from the reaction medium as described above. The structure of 4 contains a large amount of highly disordered acetonitrile molecules in the asymmetric unit. This density could not be satisfactorily modeled with discrete acetonitrile molecules and instead SQUEEZE was implemented in PLATON. 11  [6] in the glovebox. The data quality is not great due to poor crystal quality. There is also significant disorder in the crystal. Every atom was modeled over two positions except the nitrogen and one of the terminal oxygens, all of which seem well behaved. In the structure, each diethylammonium counterion is hydrogen bonded to the terminal oxygen of one phosphate as well as the other terminal oxygen of another molecule resulting in a infinite hydrogen bonding network.