Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
J Org Chem. Author manuscript; available in PMC 2011 May 21.
Published in final edited form as:
PMCID: PMC2874677

Versatile 1H-31P-31P COSY 2D NMR Techniques for the Characterization of Polyphosphorylated Small Molecules


An external file that holds a picture, illustration, etc.
Object name is nihms-198934-f0001.jpg

Di- and triphosphorylated small molecules represent key intermediates in a wide range of biological and chemical processes. The importance of polyphosphorylated species in biology and medicine underscores the need to develop methods for the detection and characterization of this compound class. We have reported two-dimensional HPP-COSY spectroscopy techniques to identify diphosphate-containing metabolic intermediates at sub-millimolar concentrations in the methylerythritol phosphate (MEP) isoprenoid biosynthetic pathway.1 In this work, we explore the scope of HPP-COSY based techniques to characterize a diverse group of small organic molecules bearing di- and tri-phosphorylated moieties. These include molecules containing P–O–P and P–C–P connectivities, multivalent P(III)–O–P(V) phosphorus nuclei with widely separated chemical shifts, as well as virtually overlapping 31P resonances exhibiting strong coupling effects. We also demonstrate the utility of these experiments to rapidly distinguish between mono- and diphosphates. A detailed protocol for optimizing these experiments to achieve best performance is presented.


Polyphosphorylated compounds are ubiquitous in Nature, and are essential for numerous biological processes. Nucleotides, the most abundant polyphosphorylated natural products, serve as building blocks in DNA and RNA biosynthesis. In addition, these naturally occurring polyphosphorylated compounds play key roles in metabolism, acting as energy sources (ATP and GTP) and performing important roles as cofactors in many biochemical processes (CoA, NADH, etc.). In medicine, nucleoside diphosphates and triphosphates are the active metabolites of a variety of cytotoxic nucleoside anticancer and antiviral agents.2,3 Bisphosphonates are stable diphosphate analogs with applications in organometallic chemistry and medicine4,5,6-9,10,11 and as chemical probes in biological processes.12,13

The detection and/or characterization of polyphosphorylated molecules poses particular challenges in chemistry and biology. The characterization of 31P nuclei in these compounds by NMR has traditionally been accomplished by direct, one-dimensional 31P NMR, often at low magnetic field strengths on relatively concentrated samples (typically > 50 mM). Complex systems containing multiple 31P sites or a mix of mono- and diphosphate centers require more detailed and unambiguous characterization, which is hampered at low concentrations. Furthermore, reaction mixtures containing multiple phosphorylated compounds are difficult to characterize given the difficulties in making accurate assignments to 31P resonances in individual polyphosphorylated species. These difficulties can be addressed by using two-dimensional, indirect 1H detected techniques, which correlate 31P centers in the molecule to covalently linked protons. Not only does this increase the information context of the spectrum, but also potentially benefits from the high sensitivity of 1H excitation and detection experiments. These experiments may be adapted from the repertoire of HSQC14,15, HMQC16 or HMBC17 techniques, that are available for 13C and 15N characterization. In contrast to 1H-13C and 1H-15N experiments, which require either isotope labeling or, for natural abundance studies, highly sensitive probes and high sample concentration, 1H-31P experiments benefit from the nearly 100% natural abundance of 31P.

Despite these potential advantages, 1H-31P experiments are underutilized. Since the sensitivity of 1HX (X = 13C, 15N, etc) correlated experiments is proportional to the strength of the appropriate JHX coupling constant, the large and relatively uniform 1JHC and 1JHN coupling constants result in high sensitivity in 1H-13C and 1H-15N correlation spectra under conditions of isotopic enrichment (or high concentrations at natural abundance). However, most 1H-31P experiments do not offer this advantage. Although one-bond 1JHP coupling constants are large (> 600 Hz), typical phosphorus containing compounds only possess multiple-bond nJHP couplings (often through H-C-O-P linkages) which are significantly smaller (< 20 Hz) and highly variable due to dependence on local geometry. As a result, 1H-31P HSQC and related experiments have not been widely used for the characterization of small molecules containing 31P nuclei.

A growing need for detecting 31P containing compounds at low concentrations revives interest in 1H-detected, multi-dimensional 31P NMR. This has been facilitated by the ever increasing sensitivity of NMR probes for 1H detection, including the availability of cryogenic probes. On modern spectrometers, even on room-temperature probes, low millimolar to sub-millimolar levels of 31P-containing low molecular weight compounds may be detected using 1H-detected techniques, as long as the JHP ≥ 5 Hz18-21.

We have found the sensitivity of 1H-detected 31P experiments to be especially advantageous when applied to polyphosphorylated compounds.1 In 1D 31P NMR spectra, di- and triphosphates with distinct chemical shifts are usually identified by splittings due to nJPP coupling constants. However, more detailed and unambiguous structural characterization is often required and may be obtained from 1H31P-31P COSY spectra, in which 31P-31P COSY cross peaks are correlated to relevant protons through JHP couplings. These experiments are particularly useful for distinguishing “open” (1H-31P-31P) species relative to “bounded” (1H-31P-31P-1H) “coupling networks”, or for characterizing structural (conformational) rearrangements in which the identities of the coupled 1H or 31P partners are changed. The large (> 20 Hz) and fairly uniform 31P-31P coupling constants exhibited by polyphosphorylated compounds facilitates high sensitivity in 1H detected, 1H-31P-31P 2D NMR experiments for detailed structural information at low concentrations. Recently, we have used this technique very effectively to characterize and discriminate a series of diphosphate-containing intermediates produced enzymatically in the non-mammalian MEP pathway for isoprenoid biosynthesis. Using HPP-COSY techniques, we demonstrated these enzymatic reactions could be studied in situ, at low substrate concentrations (< 3 mM) without the need for isotopic enrichment or purification1. The success of this approach for the characterization of MEP pathway intermediates highlights the potential value of HPP-COSY techniques as structure determination tools for the organic chemist and has prompted us to explore the scope and limitations of these techniques for characterization of other structurally diverse phosphorylated small molecules.

Here, we report and discuss applications of HP-HSQC and HPP-COSY based techniques to a variety of phosphorylated small molecules to illustrate the applicability of these experiments. We have highlighted seven compounds (1-7, Figure 1) in order to demonstrate HP-NMR techniques for the characterization of biologically and chemically relevant molecules exhibiting various distinguishing features, such as the presence of a P-C-P linkage (2), large differences in 31P chemical shift (3), overlapping, strongly coupled 31P nuclei (4), triphosphate (6, 7) linkages. Finally, we provide a practical guide for the execution of experiments, such that HPP-COSY experiments can be used successfully at low millimolar or even sub-millimolar concentrations relevant in many chemical and biological systems.

Figure 1
Structures of compounds highlighted for the HP NMR studies: UDP-galactose (1), P,P’-[[3,5-bis(1,1-dimethylethyl)-4-hydroxyphenyl]ethenylidene]bis-,P,P,P’,P’-tetraethyl ester (2), Bis(diphenylphosphine)methane monooxide (dppmO) ...



The 1H-31P and 1H-31P-31P pulse sequences used in this work have been published elsewhere1, but have been reproduced in Figure S1 (see Supporting Information) for the sake of clarity. The 1H-31P correlation experiments are essentially HSQC and Constant-time (CT) HSQC experiments (see Supporting Information for theoretical and experimental details). We refer to the relevant network of J-coupled nuclei as HA–PA–PB–HB, where HA:PA and HB:PB are correlated by nJHP coupling constants (n may be different for HA:PA and HB:PB) and PA:PB are coupled via a 2JPP coupling constant (see Figure 2). Note that PA and PB refer to 31P nuclei whose chemical shifts are distinct and magnetically and chemically non-equivalent. In an 1H-31P HSQC spectrum of an HA–PA–PB–HB spin system, the presence of a coupling constant between HA and PA and between HB and PB, results in the observation of cross peaks HA:PA (correlation between the chemical shift frequencies of HA along the 1H dimension and PA along the 31P dimension) and HB:PB (Figure 2a). The intensity of the cross peaks (and hence, the sensitivity of the spectrum) is proportional to the strength of the appropriate coupling constant nJHP (where n is the number of intervening bonds between 1H and 31P). When the 31P dimension is acquired in high resolution, the HA:PA and HB:PB cross peaks are split into doublets along the 31P axis by 2JPP. In comparison, in an isolated HC–PC spin system (monophosphate) where PC is not coupled to a second 31P nucleus, the observed cross peak HC:PC remains a singlet. This is shown schematically in Figure 2a.

Figure 2
Schematic representations of HP-HSQC and HPP-COSY spectra for the HA–PA–PB–HB network and an independent, HC–PC monophosphate species. Generic structures depicting HP networks are shown on the left. Figures (a)-(c) depict ...

The observed splitting of the PA and PB resonances into doublets indicates that they belong to a diphosphate species. However, establishing that PA and PB are coupled to each other requires a 31P-31P 2D COSY experiment, where the 2JPP coupling constant between the two nuclei results in the observation of PA:PB cross peaks. A considerably more detailed characterization of the entire HA–PA–PB–HB system may be achieved by combining the features of an 1H-31P HSQC with a homonuclear 31P-31P COSY experiment, which effectively integrates the PA:PB COSY connectivity into the HA,B:PA,B HSQC spectrum. Furthermore, a gain in sensitivity is achieved from the intrinsically high sensitivity of 1H detection. This integration of 1H-31P HSQC and homonuclear 31P-31P COSY experiments is achieved by the HPP-COSY technique, whose spectral features are shown schematically in Figure 2b, c. The HPP-COSY spectrum in Figure 2b consists of the three HA:PA, HB:PB and HC:PC HSQC peaks, which will be referred to as “auto” peaks, in this context. The monophosphate species (HC:PC) shows no additional features in the spectrum. However, the HA–PA–PB–HB diphosphate species exhibits two additional, HA:PB and HB:PA “cosy” peaks, which represent a COSY pattern between the coupled PA and PB nuclei. Essentially, the HPP-COSY experiment facilitates magnetization transfer first between coupled 1H and 31P nuclei (HSQC), and subsequently between coupled 31P nuclei (COSY). Eventually, magnetization is returned to the originating 1H for detection. When this magnetization transfer process is achieved from either end of a HA–PA–PB–HB network, the entire network may be established unambiguously.

The intensities and phases of the HA:PB and HB:PA “cosy” cross peaks are dictated by the experimental delay parameter tpp (See Figure S1 in Supporting Information). When tpp = 1/(4JPP), magnetization is transferred partially (50%) between 31P nuclei in the 31P-31P COSY step, and the spectrum consists of both HA:PA /HB:PB (“auto”) and HA:PB/HB:PA (“cosy”) cross peaks, which are equal in intensity and opposite in phase (Figure 2b). We will refer to these spectra as HPP-4 COSY to denote the pattern of four cross peaks (two auto and two cosy) observed in an HA–PA–PB–HB network (and also to reflect the tpp = 1/4JPP delay). For tpp = 1/(2JPP), we obtain an HPP-2 COSY spectrum, which consists of only the two HA:PB and HB:PA cosy cross peaks in an HA–PA–PB–HB network (Figure 2c). In this case, complete magnetization transfer (100%) from one 31P nucleus to the other has been achieved in the 31P-31P COSY step. HPP-2 spectra are beneficial in cases where the PA and PB chemical shifts (and/or HA and HB chemical shifts) are close, leading to overcrowded HPP-4 spectra.

In many situations, there exists only a very slight difference in the chemical environments of PA and PB, causing their chemical shifts to be extremely close (typically, < 0.4 ppm). Typically, for the same reasons, the proton chemical shifts, HA and HB are also in close proximity. Under these circumstances, the diphosphate peaks are extensively overlapped leading to cancellation of auto and cosy cross peaks. In addition, strong coupling effects between PA and PB, are manifested as undesirable phase distortions in the peaks. As a result, the HPP-4 spectra are normally unsuitable for analysis. Although the HPP-2 spectra often perform better in terms of alleviating distortions due to strong coupling effects, these experiments do not always provide adequate resolution, owing to peak broadening as a result of the large 2JPP splittings. To address these situations, additional efforts are required to maximize the resolution in the 31P dimension as well as to ameliorate the deleterious impact of strong coupling effects. This is achieved through the constant-time (CT) modification of HP-HSQC and HPP-COSY experiments. In CT experiments, 31P doublets are effectively “homo-decoupled” and “collapse” into singlets in the 31P dimension22,23. In this process, strong coupling effects between PA and PB and associated phase distortions are also suppressed significantly. The combined effects of homo-decoupling of a relatively large 31P-31P coupling (typically 20 Hz) and the minimization of phase distortions result in significantly improved peak resolution and spectral quality. The schematic of a CT HP-HSQC spectrum is shown in Figure 2d. Without the CT modification, the 2JPP doublet splittings will prevent proper resolution of the PA and PB auto peaks and distortions due to strong coupling effects will lead to poor spectral quality. This advantage is carried over to CT HPP-COSY spectra, which possess the same dependence on the tpp parameter as HPP-COSY spectra. Schematics of CT HPP-4 and CT HPP-2 COSY spectra are shown in Figures 2e and 2f, respectively. It is evident from Figure 2e that the collapsing of the 31P doublets in a CT HPP-4 spectrum presents a clear improvement over a simple HPP-4 COSY spectrum, in similar fashion to the CT-HSQC spectrum. However, if PA and PB are in extremely close proximity, then even the constant time experiment cannot avoid cancellation of auto and cosy peaks. Under these circumstances, the CT HPP-2 spectrum (Figure 2f), which consists only of cosy peaks, is the most desirable since it provides all the necessary cosy information, attenuates strong coupling artifacts, and minimizes spectral overlap. Complete connectivity in an HA-PA-PB-HB network may be obtained by overlaying independently acquired CT HSQC (Figure 2d) and CT HPP-2 (Figure 2f) spectra, without suffering from the intrinsic problems of HPP-4 or CT HPP-4 spectra.

Although CT spectroscopy is the experiment of choice for HPP spectroscopy to characterize diphosphates with closely spaced 31P resonances, there is another feature of CT spectra which may be applied advantageously to all diphosphate containing compounds. The relative phases (signs) of mono- and diphosphate signals are dictated by the experimental delay parameter TC in HSQC experiment (see Figure S1 in Supporting Information). When TC = 1/JPP, diphosphate resonances appear with inverted phases relative to monophosphate resonances (Figure 2d), making this a useful tool for distinguishing between the two species. This is true regardless of the magnitude of the chemical shift difference between PA and PB resonances, as long as they are distinct (i.e., magnetically and chemically non-equivalent). Note that it is not necessary to acquire a 2D spectrum to identify which protons are coupled to mono- or diphosphate species. If the detected protons are well-resolved, the phase relationship is evident from a 1D version of the CT-HSQC experiment (see Figure S2 in Supporting Information). In two-dimensional experiments, even a low resolution 2D spectrum can distinguish mono- and diphosphate species because the information is contained in the phases of the peaks as opposed to an HSQC spectrum, which needs to be acquired in high resolution in order to observe 2JPP doublets for identifying diphosphate peaks.

1H-31P-31P COSY to characterize polyphosphorylated small molecules

UDP-galactose (1)

The naturally occurring diphosphate UDP-galactose is a precursor in polysaccharide biosynthesis and is structurally related to the NDP-sugar substrates of glycosyltransferases in natural product biosynthesis. The need to synthesize and characterize NDP-sugar substrates for the study of glycosyltransferases and for glycorandomization approaches in glycosylated natural product biosynthesis26-28 highlights NDP-sugars as important diphosphate-bearing compounds. Here, UDP-galactose represents our model for demonstrating the basic elements of HPP spectra. The HA–PA–PB–HB system is highlighted in Figure 1. Figure 3a shows an HP-HSQC spectrum of UDP-galactose, demonstrating correlations between the HA and PA nuclei, as well as correlations between HB and PB nuclei. The spectrum acquired at high resolution in the 31P dimension (Figure 3b) shows that each of the HA:PA and HB:PB cross peaks are split into doublets along the 31P axis, due to the 2JPP (21 Hz) coupling constant between PA and PB. The fact that PA and PB are indeed coupled to each other may also be established from an HPP-4 COSY spectrum (Figure 3c) acquired with tpp = 1/(4JPP) (13.0 ms). The spectrum contains both HA:PA, HB:PB (“auto”) peaks as well as phase-inverted HA:PB and HB:PA (“cosy”) peaks. Note that in this example, the HPP-COSY spectrum need not be acquired in high resolution along the 31P dimension. Since PA and PB exhibit a substantial chemical shift difference, “auto” and “cosy” peaks may be adequately and rapidly resolved, even in low resolution HPP-COSY spectra. From these spectra, the HA–PA–PB–HB diphosphate linkage between the uridine and galactose components can be unambiguously established.

Figure 3
1H-31P HSQC and 1H-31P-31P COSY spectra of a 5 mM UDP-galactose (1) sample in D2O at 30 °C. Only one of the two diastereotopic protons, HB, is shown in the structure. (a) Low-resolution 1H-31P HSQC showing the HA:PA and HB:PB correlations. (b) ...

P,P’-[[3,5-bis(1,1-dimethylethyl)-4-hydroxyphenyl]ethenylidene]bis-,P,P,P’,P’-tetraethyl ester (2)

Also known as SR-12813, this bisphosphonate analog is a potent cholesterol lowering agent29 and is distinct from UDP-galactose in that it bears a P–C–P linkage. The HA–PA–PB–HB J-coupling network and HPP-4 COSY spectrum are shown in Figure 4. “Auto” peaks (HA:PA, HB:PB) and “cosy” peaks (HA:PB, HB:PA) are clearly observable, and establish the presence of the HA(...)PA–C–PB(...)HB connectivity. This example demonstrates the application of HPP spectroscopy for characterization of the bisphosphonate compound class bearing P-C-P linkages.

Figure 4
HPP-4 COSY spectrum of 5 mM 2 in C6D6 at 30 °C with tpp = 1/(4JPP) = 2.5 ms, illustrating the application of HPP NMR to compounds with P-C-P linkage. The “auto” (HA:PA and HB:PB) and “cosy” (HA:PB and HB:PA) peaks ...

Bis(diphenylphosphine)methane monooxide (dppmO) (3)

DppmO and its analogs are versatile, widely-used hemilabile ligands in organometallic chemistry.30,31 Their effectiveness as metal ligands in catalysis arises from the presence of two phosphorus nuclei, one in the P(III) oxidation state and the other in the P(V) oxidation state, connected through a P-C-P linkage. This structural feature also renders them interesting candidates for HPP-COSY experiments, as the trivalent and pentavalent 31P nuclei exhibit a large chemical shift difference. The coupling network, as outlined in Figure 1, consists of two groups of protons (HA1, HA2) coupled to PA (III) and HB coupled to PB (V). The chemical shift difference between PA and PB is 52.3 ppm, corresponding to 10583 Hz at 200 MHz (31P). This difference in chemical shift represents a challenge for HPP experiments, since the 31P pulses in the sequence (shown in Figure S1 Supplementary Information) are inadequate to cover the entire range. A modification of the pulse sequence, using composite 31P pulses, proved to be effective for adequate excitation of the wider bandwidth. A comparison of the HP-HSQC and HPP-COSY spectra is shown in Figures 5a and 5b. Acquisition parameters are such that the 31P spectral width has been reduced by a factor of ~9, so that the two chemical shifts only appear to differ by only 6 ppm. This “folding” in the 31P dimension achieves higher resolution in a shorter time period. In the HSQC spectrum, we observe HA1,2:PA and HB:PB cross peaks. Interestingly, the direct correlation between HA1,2 and PB is not observed, likely due to unfavorable geometry between these nuclei. However, these peaks can be observed through PA–PB J-coupling using HPP-COSY spectroscopy. In the HPP-4 COSY spectrum, additional HA1,2:PB and HB:PA “cosy” crosspeaks establish the connectivity between PA(III) and PB(V).

Figure 5
Illustration of HPP NMR applied to dppmO (3), a compound with a large chemical shift difference in the 31P dimension. Both the 1H-31P HSQC (a) and HPP-4 COSY (b) spectra were taken with a 5 mM dppmO sample in C6D6 at 30 °C. The spectra were folded ...

Nicotinamide adenine dinucleotide (NAD+) (4)

NAD+ is a common cofactor found in all living cells, and is involved in a large number of biochemical redox reactions. For our studies, NAD+ represents a particular challenge in HPP NMR: the chemical shift difference between 31P nuclei in NAD+ is small (Δppm ~ 0.3) and causes a number of complications arising from resonance overlap and strong coupling effects between the two 31P nuclei in the 2D spectra. The importance of employing CT-HSQC and CT-HPP-2 COSY experiments to address these complicating factors is illustrated here. In addition, we demonstrate the additional utility of CT experiments to rapidly distinguish between mono- and diphosphate moieties without the need to acquire high resolution 2D spectra.

HPP vs CT HPP spectra

Figure 6a shows a low resolution H-P HSQC spectrum of a mixture of NAD+ and the monophosphate PEP. The diphosphate linkage in NAD+ consists of HA1,2–PA–PB–HB1,2 as shown in Figure 6, where the two HA1,2 protons are associated with the nicotinamide moiety and the two HB1,2 protons with the adenosine moiety. The closely spaced HA1,2:PA (δPA = −36.9 ppm) and HB1,2:PB peaks (δPB = −36.6 ppm) are almost overlapping, due to the combined effects of the proximity of the PA and PB chemical shifts, the HB and HA proton resonances and the broadening of peaks in 31P due to the 20 Hz JPP coupling constant. Figure 6b shows an HPP-4 COSY spectrum of NAD+, acquired with the same resolution in the 31P dimension as the HSQC spectrum in Figure 6a. Both “auto” and “cosy” peaks are severely distorted due to strong coupling effects, which are most dominant when tpp is set to 1/(4JPP). In addition, the close chemical shifts in the 31P dimension cause mutual extinction of the HA:PB and HA:PA cross peaks, which are opposite in phase. As a result, this experiment is clearly unsuitable for identifying HPP-COSY peaks to establish H-P-P-H connectivity in this compound.

Figure 6
Mixture of NAD+ (4) and monophosphate PEP (5) demonstrating the importance of constant time (CT) HSQC and CT HPP NMR spectra: (a) 1H-31P HSQC spectrum of NAD+ (4) and PEP (5) showing the HA1,2:PA, HB1,2:PB correlations and peak overlap due to proximity ...

Figure 6c shows the CT-HSQC spectrum of a mixture of NAD+ (diphosphate) and PEP (phosphoenol pyruvate, a monophosphate) (5) (Figure 1) obtained with TC = 1/JPP. The relevance of the PEP spectrum (shown as an inset) is explained below. Here, we focus on the NAD+ diphosphate peaks. The constant time experiment collapses the 2JPP doublet into a singlet, thereby significantly improving the resolution of the spectrum. This is depicted in a comparison of the one-dimensional traces parallel to the 31P dimension, between Figures Figures6a6a (HSQC) and and6c6c (CT HSQC). In order to correlate PA and PB unambiguously, the CT HPP-2 COSY spectrum was acquired (Figure 6d). As expected, it provides higher spectral quality, combining all the benefits of resolution enhancement due to the homo decoupling effect, reduced strong coupling artifacts and elimination of auto and cosy peak overlap.

Distinction between mono- and diphosphate using CT HP-HSQC

As described above, CT HSQC spectra were acquired on a mixture of NAD+ (diphosphate) and PEP (phosphoenol pyruvate, a monophosphate) (5, Figure 1). PEP bears two protons HC1,2 coupled to PC. Figure 6e shows a CT-HSQC spectrum of the NAD+/PEP mixture. The HC1,2:PC cross peaks appear significantly downfield of the NAD+ peaks in the 31P dimension (δPC = −22 ppm) and are presented as an inset in Figure 6c. In the CT-HSQC spectrum, the diphosphate (NAD+) peaks are of opposite phase relative to the monophosphate (PEP) peaks. Thus, this method permits rapid, unequivocal discrimination of mono- and diphosphates (as long as the diphosphate 31P nuclei are magnetically non-equivalent and their chemical shifts are distinct). Importantly, acquisition of a 2D spectrum is not always necessary to make this distinction. If the proton resonances of the mono- and diphosphate species are distinct, a 1D CT-HSQC spectrum acquired with TC = 1/JPP will yield opposite phases for the protons associated with mono- and diphosphates (Figure S2, Supporting Information). Even in the two-dimensional mode, where the mono- and diphosphate distinction is contained in the phases of the peaks, this information can be obtained far more quickly than an HSQC spectrum, where relatively high resolution spectra need to be acquired in order to resolve 2JPP doublets for identification.

Expanding HPP-COSY techniques for the characterization of triphosphate analogs

The HPP-COSY concept is easily applied to identify triphosphate moieties defined by HA–PA–PB–PC–HC J-coupled networks. A simple extension involves the addition of a second, relayed COSY step to the HPP-COSY pulse sequence. The most straightforward example of a triphosphate linkage is shown in the case of dUTP (6, Figure 7), which exemplifies a HA–PA–PB–PC system. In this HPPP-COSY experiment, adjusting the duration (tpp) of two back-to-back PA → PB and PB → PC COSY steps results in different cross peak patterns (Figure 7), which can be used to identify the triphosphate moeity. Although this was easily achievable in the case of dUTP, relayed COSY steps are not efficient when 31P magnetization decays rapidly during the COSY steps. In small organic molecules, one of the most common reasons for rapid 31P signal decay is exchange broadening in the presence of excess salt (as opposed to nuclear relaxation, which is more severe in large macromolecules). An example of this is demonstrated in the case of (7), a capped ATP analog which is used for photolysis studies, bioimaging experiments and time-resolved studies of ATP-requiring biological systems32,33. The HA–PA–PB–PC–HC network in this molecule is shown in Figure 1. 1H-31P-31P-31P COSY experiments attempted on the ammonium salt of this sample failed to achieve multiple HA→PA→PB→PC and HC→PC→PB→PA COSY steps. However, standard HPP-4 COSY experiments were successful in observing HA:PB and HC:PB cosy cross peaks (Figure 8), thus establishing the presence of the HA–PA–PB–PC–HC network.

Figure 7
Characterization of a nucleoside triphosphate, dUTP (6), with the 1H-31P-31P-31P COSY experiment. The 1H-31P-31P-31P COSY spectra were acquired with two successive, “relayed” 31P-31P COSY elements, each with individual tpp parameters. ...
Figure 8
HPP-4 COSY spectrum of 7 (tpp = 1/(4JPP) = 11.0 ms) showing auto peaks HA:PA and HC:PC and cosy peaks HA:PB and HC:PB to demonstrate the HA–PA–PB–PC–HC connectivity. The spectrum was acquired with a sample containing 5 ...

A practical guide for acquiring HP and HPP spectra

To effectively adapt HPP NMR techniques to diverse situations, we have developed a systematic protocol for optimizing these experiments. An illustrated guide (Figures (Figures99 and S2) to optimize HP-HSQC, CT-HP-HSQC, HPP-COSY and CT-HPP-COSY experiments is presented below and in Figure S2.

Figure 9
General guide for the setup of the HPP NMR experiment. The shaded peaks have opposite phase relative to the unshaded ones.
  1. If possible, obtain a 31P 1D spectrum with and without 1H decoupling. This yields information about 31P chemical shifts and identifies 31P resonances that are coupled to 1H, as well as provides estimates of the JHP coupling. 31P-31P splittings also provide information about the JPP coupling constants and strong coupling effects. However, acquiring 31P spectra at low concentrations (< 1 mM), is often not possible. In that case, 31P chemical shifts may be inferred from typical literature values.
  2. Set up a 1D HP-HSQC experiment using the pulse sequence in Figure S1 (Supporting Information). Array thp over a range of values (Figure 9b), and choose the spectrum with maximum intensity and minimum phase distortions (these arise from competing JHH coupling constants, especially for geminal protons). Typical values of thp range from 10 to 40 ms.
  3. Acquire a low-resolution 2D HP-HSQC spectrum (Figure 9c) and adjust the spectral width along the 31P dimension if necessary. When the spectral width is very large, folding the 31P dimension is recommended.
  4. Set up a 1D HPP-COSY experiment to optimize the tpp value (Figure 9d). Acquire 1D spectra with different values of tpp and note the null (tpp = 1/4JPP, typically 10 to 12 ms) and the first negative maximum (tpp = 1/2JPP, typically 20 to 24 ms). (a) Acquire a 2D spectrum with tpp = 1/(4JPP) for compounds with large chemical shift difference in PA and PB. The final spectrum will have “auto” and “cosy” peaks of opposite phase. (b) Acquire a 2D spectrum with tpp = 1/(2JPP) for compounds with small chemical shift difference in PA and PB. The final spectrum will show only the “cosy” peaks.

If the spectra display severely overlapped peaks, use CT HP-HSQC and CT HPP-COSY experiments. If the relaxation losses are severe, regular HPP-COSY (without CT) should be obtained, using tpp = 1/(2JPP). The parameter TC can be obtained by running arrays of TC to look for the maximum inversion signal (Figure S2, Supporting Information).


In this work, we have shown the applicability of HPP-COSY to a wide range of structurally diverse phosphorus-containing small molecules. The compounds highlighted in this study demonstrate HP-NMR characterization techniques on biologically and chemically interesting scaffolds. The basic technique was demonstrated using the model UDP-galactose, in which a characteristic HP-HSQC spectrum depicts correlations between the HA and PA nuclei as well as between the HB and PB nuclei (“auto” peaks). To establish the connectivity of uridine and galactose components through a P-O-P diphosphate linkage, the basic HSQC concept was combined with a 31P-31P COSY step, to generate the HPP-4 and HPP-2 experiments. In an HPP-4 COSY, a relevant experimental parameter tpp is set to 1/(4 JPP), resulting in the observation of a new set of “cosy” peaks correlating HA:PB and HB:PA that are opposite in phase compared to the auto peaks. The combination of the four peaks namely; HA:PA (auto)/ HA:PB (cosy) and HB:PB (auto)/ HB:PA (cosy) cross peaks permit characterization of the entire HA-PAPB-HB J-coupled network. These experiments were also successfully applied to the bisphosphonate SR-12813, which bears a P-C-P linkage, and DppmO, a small molecule containing two 31P nuclei with drastically different 31P chemical shifts. In this case, 31P pulses with broad excitation profiles were applied to cover the wider 31P bandwidth. These latter experiments demonstrate the applicability of HPP-COSY spectroscopy to the important bisphosphonate compound class.

In contrast to UDP-galactose, SR-12813 and DppmO, close proximity of phosphorus and proton resonances in NAD+, along with artifacts due to strong coupling between the coupled 31P nuclei, resulted in near overlap of HA:PA and HB:PB auto peaks in the HSQC spectrum. This led to mutual extinction of auto and cosy cross peaks in the HPP-4 COSY. In this case, a constant time (CT) modification of HSQC spectra improved resolution by collapsing the PA:PB doublets into singlets and also minimized strong coupling artifacts. The CT modification was then incorporated into an HPP-2 COSY experiment (tpp = 1/(2JPP), yielding only cosy cross peaks in the spectrum, thereby resulting in considerable overall improvement in spectral quality. In addition to improving the resolution in the 31P dimension by effectively homo-decoupling the P-P doublets, constant time experiments permit the rapid discrimination between mono- and diphosphate resonances as a result of these species displaying resonances of opposite phase in the CT-HSQC spectrum. Thus, a mixture containing NAD+ (diphosphate) and PEP (monophosphate) could be easily distinguished.

We have also demonstrated extensions of the HPP-COSY techniques for the characterization of triphosphate linkages in the biologically relevant triphosphate dUTP and the related masked ATP analog 7. In the case of dUTP, the connectivity of deoxyuridine through the PA–O–PB–O–PC triphosphate linkage was easily established using the HPPP-COSY experiment which contains two successive PA-PB and PB-PC COSY steps. However, the efficiency of HPPP-COSY to establish the triphosphate linkage in the caged ATP analog 7 was significantly reduced, likely a result of 31P signal loss during the extended COSY steps, due to exchange broadening from high salt concentration in the sample. In this case, identification of the HA-PA-PB-PC-HC network was instead accomplished through HA:PB and HC:PB obtained using a simple HPP-4 COSY experiment. Establishing connectivity to PB from opposing molecular components of 7 effectively characterizes this triphosphate moiety.

Although the HPP NMR provides an effective way to distinguish diphosphates and monophosphates in general by showing opposite phase in the CT HP-HSQC and HPP-COSY spectra of these compounds, these experiments are not suitable for detection of a bisphosphonate bearing two magnetically equivalent 31P nuclei. For bisphosphonates possessing high symmetry, a simple 1D 1H NMR, acquired with or without 31P decoupling, can be used to identify the number of 31P nuclei in the structure (Figure S4, Supporting Information).

Finally, we have presented a practical guide for the execution of HP and HPP NMR experiments such that these techniques can be easily implemented and effectively used as structure determination tools for the organic chemist. The examples we have discussed here, in conjunction with our previously reported application of HPP-COSY techniques to the characterization of metabolic intermediates in the MEP biosynthetic pathway, illustrate the versatility of HPP-COSY spectroscopy. When experiments are properly optimized and favorable 31P relaxation conditions typical of small molecules exist, HP NMR is highly sensitive and can be applied to samples at low millimolar or sub-millimolar concentrations to selectively detect and characterize phosphorylated species even in crude reaction mixtures.

Experimental Section

NMR Spectroscopy

All data were acquired on a 500 MHz (1H) NMR spectrometer at 30 °C, on an “inverse detection” penta-probe (1H,31P,15N,31P,2H) equipped with actively-shielded z-gradient coils. Data were processed with nmrPipe and analyzed using nmrDraw software.34 All spectra were referenced with respect to triphenylphosphine oxide (TPPO) as an external standard.

Sample Preparation

The compounds used in this study are shown in Figure 1. Compounds 17 were purchased from commercial suppliers and dissolved in appropriate solvents at a final concentration of ~5 mM.

Supplementary Material



We acknowledge the Johns Hopkins University Biomolecular NMR center for the use of their 500-MHz NMR spectrometer. This work was supported by funding from the NIH (T32CA009243 for M.S.) and from Johns Hopkins Malaria Research Institute Pilot Grant (for C.F.M. and M.S.).


Supporting Information Available: Pulse sequences and some theoretical details for HPP experiments; Detailed acquisition parameters for the HPP spectra shown in Figures Figures33--8;8; A figure illustrating the procedure for the optimization of the constant time HP-HSQC and HPP-COSY spectroscopy; Overlay of Figure 6c and 6d; HPP and 1H 1D spectra for PEP (5), methylenediphosphonic acid and zoledronic acid. This material is available free of charge via the Internet at


(1) Majumdar A, Shah MH, Kipchirchir Bitok J, Hassis-LeBeau ME, Meyers CLF. Mol. BioSyst. 2009;5:935–944. [PMC free article] [PubMed]
(2) Salas-Alvarez LM. Curr. Topics Med. Chem. 2008;8:1379–1404. [PubMed]
(3) De Clercq E. Rev. Med. Virol. 2009;19:287–299. [PubMed]
(4) Reid IR. Seminars in Cell and Developmental Biology. 2008;19:473–478. [PubMed]
(5) Russell RG. Ann. N. Y. Acad. Sci. 2006;1068:367–401. [PubMed]
(6) Winter MC, Coleman RE. Curr. Opin. Oncol. 2009;21:499–506. [PubMed]
(7) Husaini HA, Wheatly-Price P, Clemons M, Shepherd FA. J. Thoracic Oncol. 2009;4:251–259. [PubMed]
(8) Li Y-Y, Chang JW, Chou W-C, Liaw C-C, Wang H-M, Huang J-S, Wang C-H, Yeh K.-Y. s. Lung Cancer. 2008;59:180–191. [PubMed]
(9) Zhang Y, Leon A, Song Y, Studer D, Haase C, Koscielski LA, Oldfield E. J. Med. Chem. 2006;49:5804–5814. [PubMed]
(10) Martin MB, Grimley JS, Lewis v. C., Heath HT, III, Bailey BN, Kendrick H, Yardley V, Caldera A, Lira R, Urbina JA, Moreno SNJ, Docampo R, Croft SL, Oldfield E. J. Med. Chem. 2001;44:909–916. [PubMed]
(11) Leon A, Liu L, Yang v., Hudock MP, Hall P, Yin F, Studer D, Puan K-J, Morita CT, Oldfield E. J. Med. Chem. 2006;49:7331–7341. [PubMed]
(12) Joseph SM, Pifer MA, Przybylski RJ, Dubyak GR. British J. Pharmacology. 2004;142:1002–1014. [PMC free article] [PubMed]
(13) Zhang Y, el Kouni MH, Ealick SE. Acta. Cryst. 2007;D63:126–134. [PubMed]
(14) Bax A, Griffey RH, Hawkins BLJ. J. Magn. Reson. 1983;55:301–315.
(15) Bax A, Subramanian S. J. Magn. Reson. 1986;67:565–569.
(16) Bodenhausen G, Ruben DJ. Chem. Phys. Lett. 1980;69:185–189.
(17) Bax A, Summers MF. J. Am. Chem. Soc. 1986;108:2093–2094.
(18) Albaret C, Loeillet D, Auge P, Pierre-Louis F. Anal. Chem. 1997;69:2694–2700.
(19) Gradwell MJ, Fan TW, Lane AN. Anal. Biochem. 1998;263:139–149. [PubMed]
(20) Larijani B, Poccia DL, Dickinson LC. Lipids. 2000;35:1289–1297. [PubMed]
(21) Petzold K, Olofsson A, Arnqvist A, Gröbner G, Schleucher J. J. Am. Chem. Soc. 2009;131:14150–14151. [PubMed]
(22) Santoro J, King GC. J. Magn. Reson. 1992;97:202–207.
(23) Vuister GW, Bax A. J. Magn. Reson. 1992;98:428–435.
(24) Wu Z, Bax A. J. Magn. Reson. 2001;151:242–252. [PubMed]
(25) Thrippleton MJ, Edden RAE, Keeler J. J. Magn. Reson. 2005;174:97–109. [PubMed]
(26) Blanchard S, Thorson JS. Curr. Opin. Chem. Biol. 2006;10:263–271. [PubMed]
(27) Luzhetskyy A, Bechthold A. Appl. Microbiol. Biotechnol. 2008;80:945–952. [PubMed]
(28) Timmons SC, Thorson JS. Curr. Opin. Chem. Biol. 2008;12:297–305. [PMC free article] [PubMed]
(29) Berkhout TA, Simon HM, Pate DD, Bentzen C, Niesor E, Jackson B, Suckling KE. J. Biol. Chem. 1996;271:14376–14382. [PubMed]
(30) Grim SO, Satek LC, Tolman CA, Jesson JP. 1975. pp. 656–660.
(31) Grushin VV. Chem. Rev. 2004;104:1629–1662. [PubMed]
(32) Hartung K, Froehlich JP, Fendler K. Biophys. J. 1997;72:2503–2514. [PubMed]
(33) Gropp T, Cornelius F, Fendler K. Biochimica et Biophysica Acta. 1998;1368:184–200. [PubMed]
(34) Delaglio F, Grzesiek S, Vuister GW, Zhu G, J. P, Bax A. J. Biomol. NMR. 1995;6:277–293. [PubMed]