Speciation of Phosphorus in Pet Foods by Solid-State ³¹P-MAS-NMR Spectroscopy

Abstract

Solid-state magic angle spinning ³¹P NMR spectroscopy is used to identify and quantify phosphorus-containing species in pet foods. The measurement is challenging due to the long spin–lattice relaxation times (T₁s). Data acquisition times are shortened by acquiring data with a tip angle smaller than 90° and shortening the repetition time. However, the spin–lattice relaxation times (T₁s) of the different ³¹P compounds are quite different, necessitating a separate measurement for each compound in the pet food. Knowledge of T₁ is used to calculate the relative amount of ³¹P in the samples. Samples of known concentration are also measured, enabling the quantitative measurement of total phosphorus content.

1. Introduction

Phosphorus is an essential nutrient and is ubiquitous in foods as a naturally occurring element in the molecular formula of various chemical components of food ingredients.¹⁻⁵ Phosphorus is also a common food additive, and phosphorus-containing compounds are used in foods to enhance flavor, thicken mixtures, control texture, and act as a preservative.^6,7

Combined with other elements like calcium, phosphorus supports and maintains healthy bones in animals. It also acts as a catalyst for various bodily functions. On the other hand, excess consumption of some forms of phosphorus can have adverse effects on health. For example, excessive dietary phosphorus can lead to hyperphosphatemia, which can have an adverse impact on kidney function.⁶ Thus, there is a safe range for the phosphorus content of food within which health is supported and promoted.

Pet food is regulated animal feed,^8,9 and most pet foods are complete and balanced, which means each portion must meet the complete nutritional requirements of the animal consumer. Complete and balanced nutrition encompasses macro- and micronutrients, including phosphorus, which plays a key role in maintaining the systemic health of pets.

Whether derived from raw materials or added as a functional ingredient, phosphorus in pet food must be balanced and within the regulated range; however, not all forms of phosphorus are equal. Phosphorus derived from animal and plant-based food ingredients such as hydroxyapatite in bones and phytate in plants may be less bioavailable than some soluble phosphorus-containing salts, which have been shown to acutely affect calcium and bone metabolism in humans^1,2 and cause dose-dependent increases in postprandial serum phosphate in feline and canine models.³⁻⁵

A recent study demonstrated that soluble P in foods, determined using an in vitro digestion assay and modified to take into account gastric and small intestine conditions in felines, was able to estimate in vivo P digestibility for wet-format, commercially formulated feline diets.¹⁰ However, Ca/P ratios in diets and other factors, such as the solubility of different P-containing compounds, were found to impact the strength of the association between the in vitro and in vivo data.¹⁰ Inductively-coupled plasma atomic emission and optical emission techniques have been developed using sample ashing preparation followed by acid reconstitution and analysis to yield total phosphorus content.^11,12 Ion chromatographic methods speciate some low molecular weight n-mers and can be coupled to an electron light scattering detector or ICP–MS. Organic phosphorus compounds, like ADP and ATP, can be analyzed by HPLC and an electrochemical detector.¹³ These methods are limited in their ability to differentiate between phosphorus species. Although it is important to be able to determine the total phosphorus concentration in foods, an analytical determination of the proportion of different forms of phosphorus in foods would be much more valuable for predicting health benefits and food safety. The challenge is to identify different forms of phosphorus in a food. Those different forms of phosphorus must be resolved by sample preparation, analytical method, or both.

Both liquid and solid states ³¹P nuclear magnetic resonance (NMR) has been used to study phosphorus in foods as well as other organic matter. Thomsen et al. (1995) studied native casein micelles using magic angle spinning (MAS) ³¹P NMR.¹⁴ Their studies revealed that the micellar inorganic calcium phosphates exhibit structural similarities to hydroxyapatite and hence resemble mineralized bone tissue. Hydroxyapatite was readily identified in human mineralized bone tissue by ³¹P NMR analysis.^15,16 Frolich et al. (1988) used ³¹P NMR to follow the breakdown of phytate in oats and wheat.¹⁷ Spyros and Dais (2000) determined that ³¹P NMR spectroscopy is a facile quantitative technique for the analysis of mono- and diglyceride content in virgin olive oils.¹⁸ Spyros and Dais (2009) utilized ³¹P NMR to obtain information on the composition of phospholipid membranes, lipid headgroup orientation relative to the bilayer normal, and the phase state of phospholipid systems.¹⁹³¹P NMR T₁ relaxation studies were performed on muscle cells in various animals and human being.²⁰³¹P solid-state NMR was also applied to studies of natural organic materials. Mathew and co-workers (2011) studied orthophosphate structural environments in biomimetic amorphous calcium phosphate and hydroxyapatite grown in vitro at the surface of a mesoporous bioactive glass.²¹

One of the significant advantages of NMR spectroscopy is its capacity to provide quantitative information for systems from molecular to macroscopic scales. Quantitative solid-state NMR allows measuring the populations of different sites and thus the relative concentrations of the components in a solid sample without dissolving them into solution. Thus, chemical analysis can be performed nondestructively. These experiments can be performed using a single pulse MAS scheme with either an adequate repetition time (TR > 5 × T₁) to measure relative concentrations directly or knowledge of the T₁ of each component to calculate the concentrations. In this study, the latter method is used to measure the composition (speciation and quantification) of phosphorus compounds in pet food.

2. Experiments

2.1. Sample Preparation

2.1.1. Wet Pet Food Samples

Commercially available wet pet food was directly obtained from the foil tray pack in wet “loaf” form and was lyophilized for 5 days to remove water from the sample. The dried chunks were ground into a fine powder and mixed well so that the sample was homogeneous for solid-state NMR. This step is necessary for stable MAS spinning. The weight loss due to lyophilization was 78.5%, leaving dry components to constitute 21.5% of the pet food from the original container.

2.1.2. Dry Pet Food Samples

Dry pet food samples consisted of kibbles of multiple shapes and colors. Each type of kibble was ground into fine particles with an agate mortar and pestle set and then mixed according to the weight fractions of the whole pellets. The fine particles were ground into fine powders and loaded into a 4 mm zirconia rotor for solids MAS NMR.

2.1.3. Standards

Inorganic phosphorus-containing compounds that might be present in the pet food were prepared as standards for comparison of the NMR spectra of these standards to those of pet food samples. Detailed information is summarized in Table 1.

Table 1. Standards for ³¹P NMR Measurement and Spiking Experiments.

symbol	formula	name	F.W. (g)	CAS #	purity	manufacturer/vendor	catalogue #	country
bone meal	unknown	bone meal powder	N/A	N/A	N/A	KAL	21245 556299	USA
MCP	Ca[H2PO4]2	calcium phosphate monobasic, calcium bis(dihydrogen phosphate)	234.05	N/A	95%	Aldrich	C8017-100G	Japan
MSP	Na[H2PO4]	sodium phosphate monobasic, sodium dihydrogen phosphate	119.88	7558-80-7	N/A	Sigma-Aldrich	S2554-100G	Switzerland
MKP	K[H2PO4]	potassium phosphate monobasic, potassium dihydrogen phosphate	136.09	7778-77-0	99%	SIGMA	P5655-100G	USA
STPP	Na5[P3O10]	sodium triphosphate pentabasic	367.86	7758-29-4	98%	Sigma-Aldrich	72061-100G	Germany
KTPP	K5[P3O10]	potassium triphosphate pentabasic	448.41	13845-36-8	N/A	Spectrum Chemical MFG. Corp.	P1669-500G	USA
TSPP	Na4[P2O7]	sodium pyrophosphate tetrabasic	265.9	7722-88-5	95%	Aldrich	P8010-500G	USA
TKPP	K4[P2O7]	potassium pyrophosphate tetrabasic	330.34	7320-45-5	97%	Aldrich	322431-500G	USA

The original bottle containing each standard was opened in a plastic glove bag filled with flowing dry nitrogen. The dry powder was ground with an agate mortar and pestle and packed in a 4 mm zirconia rotor and Kel-F cap. All sample handling is performed inside the glove bag filled with flowing dry nitrogen to prevent from possible hydration by the moisture in the air. After taking out appropriate amount of the sample the original sample containers were tightly caped and sealed with paraffin film.

2.1.4. Spiked Samples

Spiked samples were prepared by adding known amounts of standards at various levels to pet food samples for quantitative studies. This was done in a blind manner, i.e., the amounts added were not known to the person analyzing the data. Dry products were ground, and the spiked compounds were added to aliquots. Wet products were first homogenized/lyophilized and then ground and spiked in the same manner. In this study, we labeled the spiked samples as A, B, C, D, E, and F and the un-spiked sample as the control (also called lyophilized wet food in Section 2.1.1).

2.2. Solid-State NMR Measurements

³¹P solid-state NMR spectra were collected with a Bruker Avance 500 MHz NMR spectrometer equipped with an 11.7 tesla wide-bore magnet and a 4 mm triple resonance Bruker MAS probe. The ³¹P Larmor frequency was 202.445 MHz. ³¹P direct polarization (DP) and proton-decoupled pulse sequences with a 60° tip angle were used. The recycle delay time was 300 s. Weights were measured before and after a sample was loaded into a 4 mm NMR rotor. The number of scans was in the range of 8–96, as needed for signal averaging. 4 dummy scans were used before data acquisition. The total spectral acquisition for each sample ran from 1 to 8.3 h. All data were recorded with a 16,384-time domain size and processed with an 8192 block size and 20 Hz of exponential line broadening. The signal-to-noise ratio values for the standards were in the range of 200–250. For the pet food samples, the signal-to-noise ratio values were in the range of 65–90. The MAS speed was 15 kHz. The ³¹P chemical shifts were externally referenced to 85% H₃PO₄. Na₂HPO₄·7H₂O was used as the secondary standard for ³¹P MAS NMR measurements, and its chemical shift is 4.875 ppm.

Proton-decoupled ³¹P MAS NMR data were collected with a single ³¹P pulse and a two-pulse-phase-modulated (TPPM) decoupling pulse sequence. ³¹P–¹H cross-polarization(CP) MAS NMR data were acquired for lyophilized wet sample, the contact time was 2 ms, and the recycle delay time was 5 s.

The spin lattice relaxation time (T₁) was measured using the saturation recovery method. A pulse train of eight 90-degree pulses with 5 ms of inter-pulse delay was used for saturation, and a delay time (TD) between saturation and inspection was varied. The integral (M(TD)) over each recognizable peak was determined at each delay time, and the integrals were fit to the equation

1

The fit yields both T₁ and the equilibrium magnetization, M₀. Note that in samples with well-separated peaks, there is a T₁ relaxation time for each peak due to differential relaxation phenomena. In some samples, the NMR signals are heavily overlapped; therefore, it is very difficult to separate these peaks and to measure the T₁ relaxation time for each component. In this case, the integrals over the entire overlapped range were used to fit eq 1 to determine a largely averaged T₁ relaxation time. We denote this value as bulk T₁.

3. Results

3.1. ³¹P MAS NMR Data of Standards

All ³¹P NMR spectra of the standards have isotropic signals in the 10 to −20 ppm range with spinning side bands out to 158 ppm away from the isotropic peaks. The isotropic peaks constitute 65% ≤ x ≤ 96% of the total intensity. The intensity fraction [(the integral of the isotropic region)/(the total integral)] of the isotropic peak varied depending on these compounds’ composition and structure. For example, the isotropic peak constitutes about 96.25% in MKP (Figure 1) and 65.65% in STPP (Figure 2). This is because the sideband distribution is related to the chemical shift tensor, which is dependent on the coordination environment of the phosphorus atom. We performed solid NMR line shape analysis with Bruker’s Topspin software and obtained the isotropic chemical shift, chemical shift anisotropy, asymmetry parameter, and fraction for each component for the standards.

Figure 1.

³¹P MAS NMR spectrum of MKP (H₂KPO₄). The spinning sidebands are very small in this case, but not in every case. This necessitates including the side bands in the integrals. Compare with Figure 2 where the spinning side bands are much larger relative to isotropic peaks.

Figure 2.

³¹P MAS NMR spectrum of STPP. (a) full spectrum, (b) isotropic region. The spinning sidebands are much larger relative to the isotropic peaks than in Figure 1.

We included all the spinning sidebands in the total NMR signal intensity when performing quantitative analysis. Ideally, the samples can be spun fast enough to eliminate the spinning sidebands by placing the sidebands outside of the extent of the chemical shift tensor and thus eliminating this complication, but that was not possible here. Furthermore, the ³¹P T₁ relaxation times are different in different structural environments, even within the same compound. Therefore, it is necessary to back-calculate the real NMR signal intensity (M₀ in eq 1) from the experimentally observed integrals.

The ³¹P MAS NMR spectral line shape of all the standards was analyzed with Bruker’s Topspin 4.1.4. and is presented in Table 2. The deconvolution results, including isotropic chemical shift, chemical shift anisotropy, asymmetry parameter, and fraction of each individual peak, reflect the structural features of these phosphates. Without other complementary structural characterization, e.g., XRD, we fitted the spectra with the apparent number of peaks. Repeated acquisition of a given sample indicates an error of 1.5%. All other sources of errors are significantly less than 1.5%. Thus, we conservatively estimate the error of the integral at 2% or less. Table 2 also lists the T₁ relaxation times of the isotropic peaks. The T₁ relaxation times are quite long compared to the experimental repetition time (TR = 300 s). When performing quantitative analysis based on ³¹P NMR signal intensities, it is necessary to calculate the T₁ relaxation back-calculation factor (vide infra, Discussion, eq 3)

Table 2. ³¹P MAS NMR Isotropic Chemical Shift, Chemical Shift Anisotropy, Asymmetry Parameter, Composition, Fitting Quality, T₁ Relaxation Times, and Back-Calculation Factors of Phosphate Standards.

sample	formula	δ(iso)a	δ(CSA)b	η(CSA)c	fractiond	best overlap (%)e	T1 (s)f	T1 back-calculation factorg
MCP	Ca[H2PO4]2	0.69	–84.8	0.05	0.49	69.2%	877	2.57
		–3.77	–96.9	0.14	0.51		1099	2.99
MSP	Na[H2PO4]	3.37	–93.7	0.46	0.46	93.1%	1388	3.55
		2.11	–64.7	0.00	0.54		1492	3.75
MKP	K[H2PO4]	4.93	–46.4	0.07	1.00	88.0%	546	1.94
STPP	Na5[P3O10]	5.85	109.8	0.08	0.38	90.0%	1031	2.86
		2.21	97.7	0.10	0.30		631	2.10
		–4.93	–123.1	0.11	0.18		1408	3.59
		–6.54	–127.8	0.10	0.14		1036	2.87
KTPP	K5[P3O10]	0.11	101.3	0.06	0.36	89.9%	801	2.43
		–2.83	104.8	0.06	0.31		1009	2.82
		–15.68	–117.0	0.08	0.029		271	1.44
		–18.0	–134.9	0.08	0.30		1543	3.84
TSPP	Na4[P2O7]	3.95	107.4	0.11	0.15	73.9%	753	2.33
		3.47	83.1	0.02	0.44
		3.04	83.3	0.03	0.28
		2.65	98.1	0.02	0.13
TKPP	K4[P2O7]	2.04	94.0	0.10	0.15	93.2%	523	1.90
		1.61	91.8	0.10	0.072		605	2.05
		1.24	88.3	0.08	0.15		536	1.92
		0.58	79.8	0.09	0.093		612	2.07
		–0.07	86.1	0.10	0.24		408	1.69
		–1.85	83.9	0.11	0.082		644	2.13
		–3.11	86.4	0.12	0.19		608	2.06
		–3.67	89.0	0.11	0.027		537	1.93
bone meal	unknown	3.56	–60.5	0.06	0.03	95.1%	23	1.15
		1.68	–62.2	0.01	0.12		25	1.15
		0.16	–52.2	0.00	0.28		27	1.15
		–1.08	–64.3	0.00	0.57		27	1.15

^aIsotropic chemical shift in ppm.

^bChemical shift anisotropy in ppm.

^cAsymmetry parameter.

^dFraction = (integral of one component)/(total integral from all components). The relative uncertainties (standard deviation/mean) from repeated iterations of deconvolution are less than 0.3%.

^eBest overlap (%): the simplex iteration minimizes the least squares difference between the experimental spectrum and the superimposed simulated spectra. overlap (%) > 70–90, good agreement. A single Lorentzian–Gaussian line width was used for all the peaks.

^fT₁ relaxation time (fitted with eq 1) for each individual isotropic peak. The components for TSPP are overlapped; therefore, a bulk T₁ relaxation time is measured from the entire isotropic chemical shift region (5.2–1.7 ppm). The relative uncertainties (standard deviation/mean) from multiple measurements of T₁ for each signal are less than 0.8%.

^gT₁ relaxation back-calculation factor is defined in eq 3 in the Discussion session (vide infra, Discussion, eq 3).

Most of the standards contain multiple peaks in the isotropic chemical shift region. This is due to multiple ³¹P atoms’ different local structural environments. For example, there are multiple phosphorus sites with different P–O bond distances and different local symmetries (bond angles) in P–O–P linkage (the bridging oxygens) in polymeric tri-phosphate. This leads to varying next nearest neighbor distances. Figure 3 displays the isotropic regions of the ³¹P MAS NMR spectra of all the standards. These patterns are the signatures for identifying the phosphorus speciation in the pet food samples.

Figure 3.

Isotropic regions of the ³¹P MAS NMR spectra of all the standards. Each compound has its own signature pattern. Note that there are small impurities in KTPP and STPP that are on the order of 5%.

The T₁ relaxation times of phosphorus compounds can be quite long. Thus, waiting 5 × T₁s can be impractical—especially when signal averaging is necessary. Data acquisition time can be decreased by shortening the repetition time (TR) and S/N maximized by lowering the tip angle to the Ernst angle.²² However, the Ernst angle depends on the ratio TR/T₁, and there is a large spread of T₁’s in our samples. For our T₁’s the Ernst angle ranged from about 45–65°. Fortunately, the S/N vs tip angle is relatively flat with respect to tip angle. We used 60° tip angle and 300 s of recycle delay time. This combination approximately maximizes acquisition of signal to noise per unit time. However, this requires a knowledge of the spin lattice relaxation time (T₁) of each peak and the tip angle to correct the integrals and thus obtain quantitative data. T₁s were measured via saturation recovery and T₁ values for each peak and bulk T₁ values for overlapped peaks are listed in Table 2. To estimate the uncertainties of T₁ calculation, we used the manual integral of a signal and repeated the measurement multiple times with slightly different integration regions. An example of a T₁ measurement for MCP is displayed in Figure 4.

Figure 4.

Saturation recovery T₁ measurement of MCP. Stacked spectra with variable inter-pulse delay times. T₁ relaxation time of the peak at −3.77 ppm is 1099 ± 2 s, the T₁ relaxation time of the peak at 0.69 ppm is 877 ± 2 s.

3.2. ³¹P MAS NMR Spectrum of Bone Meal

The ³¹P NMR spectrum of bone meal is displayed in Figure 5. It has four overlapping components with isotropic signals at −1.08, 0.16, 1.68, and 3.56 ppm. The deconvolution of the peaks and their fractions is shown in Figure 5 and summarized in Table 2. The individual peak patterns were calculated using the line fitting program DMFIT (http://nmr.cemhti.cnrs-orleans.fr/dmfit/help/models/default.aspx). Comparing the ³¹P MAS NMR spectrum of bone meal with Table 2 shows no match. Thus, bone meal does not contain any of the phosphate species listed in Table 2. Other phosphates, for example, organic or bio-phosphorus species, must be present in this sample.

Figure 5.

³¹P MAS NMR spectrum of bone meal. The spectrum can be fitted with four Gaussian–Lorentzian lines.

3.3. ³¹P MAS NMR Spectra of Dry Pet Food

The ³¹P MAS NMR spectra for a dry pet food are displayed in Figure 6. Figure 6a,b presents the single pulse DP spectrum and the zoomed in isotropic region, respectively. Figure 6c is the DP ³¹P NMR data acquired with ¹H TPPM decoupling. Proton decoupling narrows the ³¹P lines, and thus, they are better resolved. This suggests that the phosphate species in the dry pellet mixture are either protonated or hydrated.

Figure 6.

³¹P MAS NMR spectra of dry pet food. (a) DP ³¹P MAS NMR full spectrum, including spinning sidebands. (b) Magnification of the isotropic peaks. (c) ¹H-decoupled ³¹P MAS NMR spectrum. Proton decoupling removes the dipolar coupling between protons and phosphorus atoms, thus narrowing the ³¹P lines. (d) Deconvolution results of the ¹H-decoupled ³¹P MAS NMR spectrum. The solid line is the sum of all six components (dash lines).

Spectral deconvolution of the ¹H-decoupled ³¹P data (see Figure 6d) (without cross polarization) shows five narrow components and one broad signal with a linewidth of 1500 Hz. The presence of multiple narrow peaks indicates that there are crystalline phases in the dry pellet mixture. For simplicity, only one Gaussian/Lorentzian line is used to simulate the broad signal centered at 2.41 ppm. However, the fitted broad component is probably from a distribution of multiple wide lines. This distribution may be due to amorphous organic or biological phosphorus-containing species.

Just as quantitative measurement of single pulse ³¹P is affected by the NMR repetition time when TR is not long enough, ¹H-decoupled ³¹P spectrum is also affected for quantitative ³¹P NMR measurements because of insufficient relaxation delay time. However, deconvolution of these spectra does yield speciation. The deconvolution result is given in Table 3. Thus, the integrals of the lines in Figure 6d and listed in Table 3 do not necessarily represent the accurate populations of different types of phosphorus in the structure. Never-the-less, ³¹P NMR distinguishes different species of phosphorus, which is beneficial to the study of phosphorus content in foods.

Table 3. ¹H-Decoupled ³¹P MAS NMR Spectral Deconvolution of the Dry Pellet Mixture.

	SITES
	1	2	3	4	5	6
δ(iso) (ppm)	7.06	6.54	4.95	4.44	3.84	2.41
δ(CSA) (ppm)	62.33	80.54	59.06	62.28	23.65	67.84
η(CSA)	0.121	0.122	0.086	0.109	0.1	0.136
width (Hz)	104	156	114	125	187	1500
integrala	1.02 × 109	1.12 × 109	5.36 × 109	4.41 × 109	6.00 × 109	2.72 × 109
fraction	0.049	0.054	0.260	0.214	0.291	0.132

^aBest overlap 91.65%.

3.4. ³¹P MAS NMR Data of the Lyophilized Canned Sample

³¹P DP with ¹H decoupling and CP MAS spectra of lyophilized canned wet sample are shown in Figure 7. A broad signal centered at about 3.9 ppm and a shoulder at about 0 ppm are present. Cross polarization is highly dependent on ³¹P–¹H inter-nuclear distance. The signal at 0 ppm becomes a greater fraction of the total signal (CP vs DP), indicating that those ³¹P atoms are more tightly dipolar-coupled to proton atoms than the signal at 3.9 ppm. The preferential enhancement occurs because there are more protons near the 0 ppm phosphorus atoms. This, in turn, indicates that lyophilization does not remove all the water molecules from the original wet pet food sample. The degree of lyophilization determines the residual water content and mass normalized ³¹P NMR signal intensity (see Discussion).

Figure 7.

³¹P DP with ¹H decoupling (red) and cross polarization (blue) MAS NMR spectra of lyophilized wet food. The DP signal in red and the cross-polarization signal in blue are deconvoluted into two Lorentzian lines at 3.9 and 0 ppm. Cross polarization conditions preferentially polarize the ³¹P atoms at 0 ppm. This indicates that these ³¹P atoms are relatively closer to protons.

3.5. ³¹P MAS NMR Data of Spiked Samples

Shop bought commercially available pet food samples were spiked with varying amounts of standards in Table 1 and packed into MAS rotors. The weight of each rotor with and without the sample was recorded so that ³¹P MAS NMR signal intensity could be scaled to determine phosphorus content. The relative uncertainties of weight measurement are less than 0.1%. The spiking matrix is available in Supporting Information section.

The ³¹P MAS NMR spectra for the spiked samples (A–F) are displayed in Figures 8 and 9. The ³¹P NMR signals from spiked standards are more obvious in samples D, E, and F when compared with samples A, B, and C. Obviously, samples D, E, and F contain added STPP, MKP, and bone meal, respectively. The quantity of these recognizable standards can be determined by spectral fitting and deconvolution. The spectra of the pure compounds (in Figure 3) are used for quantification of the components in the mixture.

Figure 8.

³¹P MAS NMR spectra of spiked samples A, B, C, and un-spiked Control and known added standards MSP, TSPP, BTKPP, and MCP. The spiking matrix is available in the Supporting Information section.

Figure 9.

³¹P MAS NMR spectra of spiked samples D, E, F, and un-spiked Control. The ³¹P MAS NMR spectra for spiked standards MKP, STPP, and bone meal is displayed at the bottom. Spiking matrix is available in Supporting Information section.

In samples A, B, and C, the spiked MSP, MCP, TKPP, and TSPP standards cannot be easily identified from the ³¹P NMR spectra (Figure 8). Spectral deconvolution is not possible because of the broad signal from the un-spiked control sample dominates the spectrum.

4. Discussion

³¹P NMR spectroscopy is a powerful tool to analyze the local structural environments of phosphorus atoms, and as such, inorganic phosphates have been extensively studied by ³¹P NMR spectroscopy. The fundamental building block of phosphates is the PO₄ tetrahedron. The nomenclature used here is that orthophosphate, a single PO₄ tetrahedron not linked to another PO₄, is Q₀. Two PO₄ tetrahedra linked together by a bridging oxygen to form a P₂O₇ dimer (pyro-phosphate) is defined as Q₁. A linear tripoly-phosphate, P₃O₁₀ (three PO₄ tetrahedra linked together by two bridging oxygens), forms a three-membered chain. The two end members are similar to the Q₁ in a dimer. The middle ³¹P has two bridging oxygens and is termed Q₂.

Typical ³¹P NMR data for common ortho-phosphates, pyro-phosphates, and tripoly-phosphates are summarized in Table 4. Table 4 exhibits the general trend of ³¹P NMR peaks shifting towards higher fields (becoming more shielded due to higher electron density) when:

• (1)PO₄ tetrahedra are polymerized. Q₂ and Q₁ have more negative chemical shifts than Q₀, e.g., Na₅P₃O₁₀, (Q₂ middle site, −8.9 ppm and Q₁ end site, −0.2 ppm) vs Na₃PO₄ (13 ppm),
• (2)cations are stronger, e.g., Ca₃(PO₄)₂ (3 ppm) a vs Na₃PO₄, (13 ppm).
• (3)PO₄ is protonated (forming HPO₄ linkage) when molecular water (or structural water) is present in the structure.

Table 4. ³¹P NMR Chemical Shift of Common Phosphorus-Containing Materials.

compound	chemical Shift (ppm)	reference
ortho-Phosphate (Q0)
85% H3PO4	0.0	universal
Na3PO4	13, 14.1, 14	Dupree et al. (1988),25 Ducel et al. (1994),26 Hayashi and Hayamizu (1989)33
Na3PO4·10H2O	7.8	Hayashi and Hayamizu (1989)33
Na3PO4·12H2O	7.8	Turner et al. (1986a)24
Na2HPO4	6.6	Turner et al. (1986a)24
Na2HPO4·7H2O	4.58	this study
Na[H2PO4]	3.37, 2.11	this study
NaH2PO4·H2O	2.3	Turner et al. (1986a)24
K3PO4	11.7, 12,5	Grimmer and Haubenreisser (1983),32 Hayashi and Hayamizu (1989)33
K3PO4·3H2O	11.3	Hayashi and Hayamizu (1989)33
K[H2PO4]	4.93	this study
KH2PO4	4.3	Turner et al. (1986a)24
K2HPO4·3H2O	2.1	Turner et al. (1986a)24
Ca3(PO4)2	3.0	Turner et al. (1986a)24
CaH(PO4)	–0.6	Turner et al. (1986a)24
CaH(PO4)	–1.6, −0.4, 1.4	Louati et al. (2005)27
CaH(PO4)·2H2O	1.6	Cheetham et al. (1986)23
CaH(PO4)·2H2O	1.35	Bak et al. (2001)29
Ca(H2PO4)2, site 1	–0.9	Cheetham et al. (1986)23
Ca(H2PO4)2, site 2	0.8	Cheetham et al. (1986)23
Ca[H2PO4]2	0.69, −3.77	this study
Ca(H2PO4)2·H2O, site 1	–5.3	Cheetham et al. (1986)23
Ca(H2PO4)2·H2O, site 2	–0.6	Cheetham et al. (1986)23
Pyro-Phosphate (Q1)
Na4P2O7, site 1	1.6, 1.3	Cheetham et al. (1986),23 Ducel et al. (1994)26
Na4P2O7, site 2	2.5	Cheetham et al. (1986)23
Na4[P2O7]	3.95, 3.47, 3.04, 2.65	this study
K4P2O7	0.9	Dupree et al. (1988)25
K4[P2O7]	2.04, 1.61, 1.24, 0.58, −0.07, −1.85, −3.11, −3.67	this study
a-Ca2P2O7, site 1	–7.8	Cheetham et al. (1986)23
a-Ca2P2O7, site 2	–10.7	Cheetham et al. (1986)23
b-Ca2P2O7, site 1	–6.3	Cheetham et al. (1986)23
b-Ca2P2O7, site 2	–7.8	Cheetham et al. (1986)23
b-Ca2P2O7, site 3	–9.1	Cheetham et al. (1986)23
b-Ca2P2O7, site 4	–10.0	Cheetham et al. (1986)23
Tripoly-Phosphate (Q1, Q2)
Na5P3O10, middle unit (Q2)	–8.9	Ducel et al. (1994)26
Na5P3O10, end unit (Q1)	–0.2	Ducel et al. (1994)26
Na5[P3O10] (Q1, Q2)	5.85, 2.21, −4.93, −6.54	this study
K5[P3O10] (Q1, Q2)	0.11, −2.83, −15.68, −18.0	this study
P3O8, middle unit (Q2)	∼−19	Maki et al. (2013)31
P3O8, end unit (Q1)	∼−6	Maki et al. (2013)31
Organic/Bio-Phosphorus
hydroxyapatite (Q1)	∼3.1 (broad signal)	Thomsen et al. (1995)14
Ca hydroxyapatite (Q1)	2.81	Bak et al. (2001)29
phytates	4.4, 5.6	Frolich et al. (1988)17
bovine bone	2.57, 3.13	Bak et al. (2001)29
sugar phosphate	5.5	Belton et al. (1987)30
orthophosphate	4	Belton et al. (1987)30
creatine phosphate	0	Belton et al. (1987)30
g-phosphates of ATP	–3	Belton et al. (1987)30
a-phosphates of ATP	–7.5	Belton et al. (1987)30
b-phosphates of ATP	–17	Belton et al. (1987)30
organic phosphorus compounds	wide range	Lara and Barron (2019)35

This is a very important factor that changes the ³¹P chemical shifts. The spectra discrepancy of our Ca[H₂PO₄]₂, with signals at 0.69 and −3.77 ppm compared with Cheetham et al. (1986)²³ results at 0.8 and −0.9 ppm, could be due to different hydration states during sample handling and NMR measurement.

The ³¹P NMR spectrum for bone meal (Figure 5) contains signals at 3.56, 1.68, 0.16, and −1.08 ppm with chemical shift anisotropy values of −60 to −65 ppm (Table 2), which suggests the presence of CaHPO₄ type of structural sites.²⁷ It is also possible that some bio-orthophosphates, including hydroxyapatite and brushite, are present in the bone meal sample. Hydroxyapatite [Ca₁₀(OH)₂(PO₄)₆] and brushite [CaHPO₄·2H₂O] are common minerals found in nature in bones and are well studied by NMR.^14,29,34 Orthophosphates are also commonly seen in ATP of animal muscle cells.³⁰

A unique feature of bone meal is its short T₁ relaxation time (longest being 27 s) compared with that of the inorganic standards (600–1800 s), see Table 2. Cohen and Burt (1977) studied ³¹P NMR T₁ relaxation behavior of phosphocreatine in intact muscle, and they reported a T₁ value of 5 s in solution state NMR.³⁶ The mobility of the phosphocreatine in intact muscle and the ³¹P coupling to ¹H and ¹³C facilitate the faster T₁ relaxation time compared with inorganic crystalline phosphates. In vivo measurements of T₁ relaxation times of ³¹P-metabolites in human skeletal muscle are also in the range of 3–5 s.²⁰

The ³¹P NMR spectrum for dry pet food (Figure 6) has peaks at 7.06, 6.54, 4.95, 4.44, and 3.84 ppm. Based on the references in Table 4, we can identify the presence of sodium ortho-phosphates with various protonation and hydration states and probably hydroxyapatite. The broad component ranging from 10 to −5 ppm could be due to the presence of organo-phosphorus species (e.g., sugar phosphates and nucleotides).

The broad feature of the ³¹P NMR spectrum of lyophilized wet sample (Figure 7) makes it difficult to identify any specific inorganic standards. But, the signals centered at 3.9 and 0 ppm strongly suggest the presence of hydroxyapatite and protonated/hydrated calcium orthophosphates.²⁸ The fact that the signal at 0 ppm becomes relatively stronger in the CP-NMR spectrum compared with DP NMR spectrum indicates stronger ³¹P–¹H dipolar coupling. The preferential polarization of the ³¹P NMR signal at 0 ppm can be ascribed to the presence of protonated HPO₄ sites in the structure.²⁹ We observed that some compounds become unrecognizable after being mixed with the pet food, as in the spiked samples A, B, and C (Figure 8).

Usually, when performing quantitative NMR measurements, the repetition time of the experiment is set to 5 × T₁ to ensure the integration of the peaks is proportional to the content. The ³¹P relaxation times measured here, however, are very long. The experiment can be shortened with knowledge of the T₁ relaxation times of both the standard compounds and the pet food samples and still determine the relative content. In general, an experiment with multiple scans is running at a steady state. The steady-state magnetization is given by the equation (see ref (22), page 125)²²

2

where θ is the excitation flip angle, M_ss is the detected NMR signal intensity from the sample, and M₀ is the thermal equilibrium magnetization–proportional to the total number of phosphorus atoms (³¹P content). E₁ = exp(−TR/T₁), where TR is the NMR repetition time, and T₁ has been measured. A back-calculation factor, C, is needed to obtain information about the actual ³¹P content from experimental ³¹P NMR signal intensity. C is defined by

3

In this study, the back-calculation factor, C, derived from eq 3 was used to calculate the measured P content against the actual P content. Different structural sites within a pure compound have different T₁ values, for example MCP in Figure 4 the T₁ values are 877 s for the peak at 0.69 ppm and 1099 s for the peak at −3.77 ppm. If we use the longest T₁ relaxation time to calibrate the signal intensity with eq 2 the signal intensities will not be reflected according to the true scale due to differential relaxation. Therefore, it is necessary to compute a composite back-calculation factor, which includes all the individual peaks (for example Table 2 for bone meal) and their T₁ relaxation times. The composite back-calculation factor C_a is defined by

4

where F_c is the fraction of a peak relative to total ³¹P NMR intensity, and

5

and

6

and T_1c is the T₁ relaxation time of a component peak, and TR is the repetition time of the experiment. In cases where the individual peaks could not be distinguished, the Bulk T₁ was used. This worked surprisingly well. Both procedures were performed for MCP and KTPP, and the error introduced was less than 5%. The error estimation is based on the possible sources of errors that are related to the ³¹P NMR signal quantitation. The relative uncertainty from sample weight measurement is less than 0.1%; the ³¹P NMR signal integral from repeated acquisition on a given sample indicates an error of 1.5%; repeated fitting of T₁ relaxation time reveals a relative uncertainty of less than 0.8%; the relative uncertainties of spectral deconvolution for multiple components (either completely separated or partially overlapped) are all under 0.4%.

Figure 10 displays a linear relationship between the ³¹P NMR signal intensity (after T₁ back calculation) and the formula-predicted ³¹P content for the known standard samples with above 90% of confidence band. The deviation from the linear regression could be from the presence of impurities (see Figure 3), moisture adsorption on the sample surface, and uncertainties during NMR data acquisition and processing (for example phase correction, baseline correction, and integration) and curve-fitting. The chemical formulas of the anhydrous compounds listed in Table 1 were used. However, the standards contain up to 5% of impurities. Without other structural characterization, e.g., XRD analysis, we had no information about the phase composition. On the other hand, all the standards have been exposed to air and, inevitably, hygroscopic compounds. For example, KH₂PO₄ (https://www.sigmaaldrich.com/US/en/sds/sigald/p0662) will adsorb some moisture on the surface. The actual structure formula would have contained some amount of water, which increases the formula weight and decreases the ³¹P content. Uncertainties during NMR measurement, including sample weight measurement, NMR data processing, and deconvolution, also populate errors in the final calculation.

Figure 10.

³¹P NMR signal intensity (calibrated relative to weight, number of scans, and T₁ back-calculation factor) vs P content (based on the formula weight in Table 1). The ³¹P content (wt %) of lyophilized wet food samples is calculated based on the linear regression from the standards. The legend is on the right. Int = a × x + y0; where x = ³¹P content based on the formula; R = 0.871; R² = 0.758.

From this linear regression, we can measure the ³¹P content of pet food samples. For the lyophilized wet food, the ³¹P content is 1.85% by weight. With a 78.5% of lyophilization rate, the ³¹P content in the original wet food is 0.40% (1.85% × [1 – 0.785] = 0.40%) by weight.

The back-calculated ³¹P MAS NMR signal intensities of the spiked standards are linear with respect to the total added amount (Figure 11, red squares, R² ∼ 0.99). It is evident that with both STPP (top) and bone meal, without the back calculation, there would be a systematic deviation of the measured phosphorus content from the actual phosphorus content. This is because the observed NMR signal intensities are affected by the experimental polarization tip angle and the ratio of repetition time to T₁ relaxation times (TR/T₁) of each standard. This is a direct result of the need to use TRs on the order of the long phosphorus T₁s as a practical matter to collect data in a timely fashion.

Figure 11.

Fraction of ³¹P MAS NMR signal intensity of the spiked standard vs the actual spiked quantity, top: STPP; bottom: bone meal. Black circles represent experimentally measured NMR intensity; the fraction of ³¹P from a standard is defined by the integral of the signal from the standard against the entire NMR signal intensity from the spiked samples. It reflects the portion of ³¹P from that specific standard in the spiked sample. Because the observed NMR signal intensities are affected by the experimental polarization tip angle and the ratio of repetition time to T₁ relaxation times (TR/T₁) of each standard, it is necessary to calibrate and re-normalize these ³¹P NMR intensities. The red squares are ³¹P NMR signal intensities after T₁ relaxation calibration and re-normalization. A linear regression is performed and yields R² ∼ 0.99.

Using the linear relationship given in Figure 10, we can correlate the NMR-measured ³¹P quantity with the actual spiked standards, and the ³¹P NMR results agree with the spike experiments. As shown in Figure 12, the fitted linear correlation equation, Y = 0.972 × X + 14.9 and R² = 0.901, clearly indicates that ³¹P MAS NMR measurement with T₁ back calculation agrees with the true P content. We also note that the intercept is not 0. This reflects the fact that the control sample contains some phosphorus at the chemical shifts of the target compound. MKP has a broad peak that substantially overlaps with the control (see Figure 9), which makes it difficult to measure the integral of MKP, much like the data in Figure 8 and thus the linear fit fails for MKP.

Figure 12.

Phosphorus content measured by ³¹P NMR signal intensity vs actual spiked known amount of phosphorus. Red circle, spiked bone meal, blue triangle, spiked STPP, green triangle, and spiked MKP. The amounts of ³¹P calculated from the NMR signal are calibrated by a number of scans and the T₁ back-calculation factor. Linear regression yields the relationship Y = 0.972 × X + 14.91, R = 0.949, and R² = 0.901.

5. Conclusions

Phosphorus content in pet foods can be determined with ³¹P NMR. In some cases, specific phosphorus containing compounds can be identified. Systematic acquisition of ³¹P NMR spectra and measurement of T₁ relaxation times of ³¹P standards are necessary. Knowledge of ³¹P T₁ permits the acquisition of quantitative solid-state NMR data despite long spin–lattice relaxation times and thus determines phosphorus content within the 90% confidence level. These results demonstrate that when impurities are in sufficient quantity such that the unique chemical shift peaks are evident, ³¹P content can be determined. Thus, the ³¹P NMR-MAS technique shows promise for quantitative phosphorus measurement in foods. This method works for dry pet food and for lyophilized wet pet food.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jafc.2c07339.

• Phosphate compounds spiking matrix, ³¹P MAS NMR spectra of phosphate standards, and deconvolution components with Bruker’s Topspin software (PDF)

The authors declare the following competing financial interest(s): This work was funded by Mars Petcare. J.W.M. and P.S. were employees of Mars at the time that the study was conducted.