Differentiation of Plant and Animal‐Derived Cholesterol Using irm‐¹³C NMR and IRMS

ABSTRACT

In the European Union, animal products for human consumption are regulated as products of animal origin (POAO), while those not intended for human consumption are considered animal by‐products (ABPs). Both POAO and ABP are subject to strict regulations to ensure animal and public health, and they are controlled through various EU directives and regulations. Consumer demand for plant‐derived cosmetics, botanicals, and nutraceuticals has also increased sharply over the past decade. Ensuring the origin of ingredients, starting materials, and excipients used in these complex consumer products can be challenging, even when certified reference standards are available. Herein, we report a reliable approach for identifying the origin of cholesterol, a common ingredient in cosmetics and beauty products, using irm‐¹³C NMR and IRMS to differentiate plant versus animal‐derived cholesterol. Further, we show that we can also discriminate semi‐synthetic cholesterol from plant and animal‐derived varieties.

Determining and certifying the origin of ingredients, starting materials, and excipients used in manufactured goods like cosmetics and medicines can be difficult. In this report, we describe a robust approach for identifying the origin of cholesterol, a component of myriad consumer products using irm‐¹³C NMR and IRMS to differentiate plant versus animal‐derived cholesterol. We also establish that we can discriminate semi‐synthetic cholesterol from plant and animal‐derived varieties.

1. Introduction

Cholesterol (1) is biosynthesized by animals and in much lower levels in plants [1]. Plants contain much higher amounts of substances with chemical structures emblematic of cholesterol but with a slightly different molecular constitution called phytosterols or plant sterols [2]. Beyond the myriad biochemical pathways and structures that cholesterol modulates or contributes to in animals, it is also a key ingredient in a wide array of cosmetics, nutraceuticals, and botanical consumer products [3, 4, 5]. In cosmetics, cholesterol is used as a moisturizer and emollient to help soften and smooth the skin, as well as to improve the appearance of fine lines and wrinkles. It is also believed to help strengthen the skin's barrier function and protect against environmental damage [6, 7]. Of serious concern, there is evidence to show that animal‐derived ingredients could lead to transmissible spongiform encephalopathies (TSE), a class of deadly prion diseases that can lead to morbidity within a few months to a few years of onset [8, 9, 10, 11, 12]. As a result, some manufacturers utilize plant sterols to synthesize cholesterol, which is sometimes used in pharmaceuticals [13, 14]. The monitoring of products of animal origin (POAO) and animal by‐products (ABPs) is closely monitored in the European Union (EU) with strict regulations intended to ensure animal and public health [15]. While less closely controlled in the United States, there is still a strong consumer appetite for products free from animal‐derived ingredients [16, 17]. These items, often advertised as being manufactured from “pure” plant ingredients, typically command higher prices and larger sales volumes than their conventional counterparts with ingredients of unknown origin [18, 19, 20].

The analysis of key starting materials and/or excipients and active ingredients in these plant‐only derived products provides a formidable challenge. In general, the only valid analytical tests to confirm the origin of the ingredients are derived on analysis of isotopic ratios of ²H or ¹³C. These analyses can be used to determine the average isotope ratios present in the material with isotope ratio mass spectrometry (IRMS) [21, 22, 23, 24] or the specific molar fraction of each isotopomer with irm‐¹³C NMR (isotopic ratio monitoring) [25, 26]. These two techniques, especially irm‐¹³C NMR, provide a unique isotopic fingerprint of a molecule, which can reveal detailed information about its history and origin. This technique has been famously proven as a robust methodology to track the true origin of basic ingredients like anetholes [27], benzaldehyde [28, 29], vanillin [29, 30], glycerol [31, 32], and others [33, 34]. It has also been an invaluable tool for the authentication of juices [35], wines [36], olive oil [37], and other high value foodstuffs [38]. For cholesterol, the determination of origin is especially challenging since the biosynthetic pathway leading to squalene, a unified advanced biosynthetic precursor, follows the same primary steps in animal and plants. This biosynthesis begins with the production of mevalonate from acetyl‐CoA, followed by the formation of isoprenoid units from mevalonate by loss of CO₂ and the condensation of six isoprenoid units to form the linear intermediate squalene. Squalene is cyclized to form the primary precursor lanosterol in animals and further modified through 19 steps of oxidation and the removal or migration of methyl groups to form cholesterol (Figure 1) [39]. In plants, squalene is cyclized to cycloartenol that is modified through several enzymatic steps, including side‐chain reduction and demethylation, diverging from phytosterol pathways to form cholesterol (Figure 1) [40]. As a result, any divergence in the ²H or ¹³C isotopic composition of plant versus animal‐derived cholesterol is due to subtle differences in specific enzyme structure and kinetics [41, 42], geographic region, environmental conditions [43], extraction procedures, and/or myriad other mitigating factors [44]. Interestingly, the origin of various squalene samples has been interrogated by irm‐¹³C NMR using ²H as the isotopic probe of choice. In that study, the site‐specific relative isotopic abundance of ²H was compared directly and by principal component analysis (PCA) methods [45]. The resulting data were able to clearly differentiate squalene from shark liver oil, olive oil, and synthetic sources. More recently, Bejjani and coworkers used adiabatic refocused INEPT to analyze a variety of cholesterol samples derived from animal sources [46]. In that report, they successfully used PCA to determine the origin of cholesterol from various hen egg sources and farming types. In another report, optimized HSQC data were used to analyze cholesterol from cheese samples as a proof‐of‐concept study, but no specific conclusions were drawn as to the origin of the different cheese samples [47]. Utilizing a simplified experimental design, compared to alternative NMR experiments that require complex pulse sequences, aligns with the practical application of developing a suitable approach for compliance with federal regulatory policies. The direct ¹³C NMR method employed here prioritizes reproducibility and straightforward implementation, making it more suitable for routine and regulatory applications.

FIGURE 1.

Structure of cholesterol depicting the origin of each carbon from the “head” or “tail” of acetyl‐CoA (A) and structure of semi‐synthetic plant‐derived cholesterol highlighting (open red oval) synthetically added carbons (B).

2. Experimental

2.1. Samples

In this study, nine replicates of animal‐derived cholesterol (MilliporeSigma Inc.), eight replicates of plant (ChemHub), and eight replicates of a proprietary semi‐synthetic form (Figure 1) that was synthetically derived (Chemo Dynamics Inc.) from a plant‐derived cholesterol intermediate were analyzed.

2.2. HRMS Analysis of Cholesterol Samples 1–4

All samples were dissolved in MS‐grade 2:1 CHCl₃:MeOH and diluted to a final concentration of 0.01 mg/mL via serial dilution in the same solvent. Mass spectrometry analysis was conducted using a Waters Xevo G2‐XS QToF mass spectrometer coupled to a Waters I‐Class Acquity UPLC system. A 3‐μL injection was performed on a Waters Acquity BEH C18 MS column (2.1 × 50 mm I.D.; S‐1.7 μm particle size; Part No. 186002350; Ser. No. 0457340222514) with UV detection at 214 and 238 nm. All solvents were LC–MS grade (Honeywell).

Chromatography was conducted with a flow rate of 0.45 mL/min and a solvent composition of Solvent A being 50:50 (v/v) H₂O:MeOH amended with 0.1% formic acid, 10‐mM ammonium acetate and that of Solvent B being 80:20 (v/v) IPA:MeOH amended with 0.1% formic acid, 10‐mM ammonium acetate. Separation was achieved following Chandramouli and Kamat [48]. Briefly, a 30‐min gradient elution included the following gradients: 0–4 min, 40% solvent B; 4–6 min, 40%–60% solvent B; 6–16 min, 60%–100% solvent B; 16–22 min, 100% solvent B; 22–24 min, 100%–40% solvent B; and reequilibration at the initial conditions for the final 6 min. Samples and column were maintained at 10°C and 40°C, respectively.

Mass spectrometric acquisition was performed using an electrospray ionization source in positive ion mode. Sodium formate was used to calibrate the instrument between 100.00 and 1800.00 amu and Leu‐Enkaphalin was used as an internal lock mass in positive mode with a m/z = 556.2771 amu. Proprietary Waters MS^E acquisition mode was utilized with function 1 utilizing no collision energy as a proxy for a full scan (MS¹) and function 2 utilizing a high‐collision energy ramp from 35 to 45 V depending on analyte size as a proxy for MS².

2.3. NMR Spectroscopy

NMR samples were prepared in 5‐mm NMR tubes (Norrell Inc.) by dissolving 200 mg of animal‐derived, plant‐derived, or semi‐synthetic cholesterol in 600‐μL 99.9% CDCl₃ (Cambridge Isotope Laboratories Inc.). The NMR data were recorded at 298 K on a Bruker Biospin 600‐MHz AV‐III Series Digital NMR equipped with a 5‐mm TXI inverse probe. The following experiments were acquired for analysis of the sample: 1D ¹H NMR data were acquired with 32 scans; SW = 12,000 Hz; o1p = 6.175 ppm; relaxation delay = 1.0 s. 1D ¹³C data were acquired with inverse gated decoupling to suppress NOE, 512 scans; SW = 36231.9; o2p = 70 ppm; acquisition time = 904.4 ms; relaxation delay = 30 s. WALTZ‐16 was used for heteronuclear decoupling with a power level of 4.5 W [30]. Parameter optimization was not further pursued, as the method was developed to be compatible with conventional NMR instrumentation without the need for specialized decoupling sequences. Additional heteronuclear decoupling performance enhancement was realized through the application of the inverse detected coil, which places the ¹H coil closest to the sample with a concomitant decrease in ¹³C S/N. Data were processed with 2‐Hz LB and baseline corrected with a third‐order polynomial function. Individualized integration regions were utilized for all samples (see details in Supporting Information), and integration was performed with no bias correction. No deconvolution of ¹³C signals was applied, as overlapping resonances remained consistent across all samples regardless of origin, enabling reliable comparative analysis of integrations without introducing a model‐dependent bias. Samples were acquired with four replicates for integration and statistical analysis on each sample. Total 1D ¹³C data acquisition required ~5 h for each replicate. All NMR data processing and chemical shift predictions were conducted in MestReNova version 14.2.3‐29241.

2.4. Statistical Approaches

Multivariate statistics were conducted using Solo version 9.5 (Eigenvector Research Inc.) and Matlab version 9.9.0.2037887 (R2020b) Update 8. An orthogonal partial least squares‐discriminant analysis (OPLS‐DA) model was generated using relative isotope fractions of animal‐ and plant‐derived cholesterol. Data were autoscaled only as a priori normalization of relative isotope fractions was conducted during NMR processing. Cross‐validation was conducted using the Venetian blinds method with 13% data left out per blind. Relative isotope fractions of semi‐synthetic cholesterol were applied to the OPLS‐DA model as a test set to determine the predictive accuracy of the model.

3. Results and Discussion

Our analysis of plant‐ and animal‐derived cholesterol began with IRMS. Given the samples provided, mass spectrometry‐derived isotopic abundance analysis could differentiate between animal‐, plant‐, and semi‐synthetic‐derived cholesterol. Semi‐synthetic cholesterol, which contains a 22‐carbon plant‐derived component, more closely resembled plant‐derived cholesterol standards (Figure 2). However, large variability prevented statistically significant differentiation. irm‐¹³C NMR was subsequently performed to provide additional confirmatory evidence for the origin of these samples.

FIGURE 2.

Comparison of relative isotope fractions for animal‐derived, plant‐derived, and semi‐synthetic cholesterol indicating differentiation of three sources through comparison of the relative amount of cholesterol molecules containing (A) zero, (B) one, and (C) two ¹³C atoms. (D) Representative isotopic envelope of ¹³C containing cholesterol from IRMS. Highlighted blue features represent incorporation of zero, one, and two ¹³C atoms from left to right. Masses and ion counts were measured for the [M‐H₂O + H]⁺ species following Chandramouli and Kamat [48].

Although the isotopic variation in ¹³C specific incorporation generally varies less than 40‰ (40 permil) as compared to more than 500‰ (500 permil) for ²H incorporation, irm‐¹³C NMR provides a superior alternative to ²H for directly comparing isotope contents at different molecular positions in a single experiment [30]. In addition, ¹³C has the advantage that it is much higher in natural abundance and overall receptivity than deuterium, requiring less sample and shorter analysis times. However, unlike ²H, ¹³C is not quadrupolar, so care must be taken to account for increased relaxation times and a higher probability of complications arising from heteronuclear NOE during relaxation delays and data acquisition [24, 49, 50]. On a relative basis, the site‐specific abundances, typically much smaller than ²H [49], A _i, can be calculated from the molar fractions, f _i , of the ¹³C isotopomers, mono‐labeled at position i. The actual relative molar fractions, f _i , are determined from the signal area, S _i of the mono‐labeled isotopomers [23]:

$$f_i=S_i/{{\mathrm{\Sigma }}_i}^{*}{S}_i$$

Relative abundance of isotopomers can be effectively evaluated by calculating molar fractions f _i from the equation above, assuming permil precision in ¹³C measurements. For authentication purposes, the absence of strict trueness requirements allows the molar fraction profile to be used as an isotopic fingerprint [26]. Also, in the case of overlap or molecular degeneracy, it is not essential to measure the isotope content of all the positions of the molecule under study to obtain a valid result. As can be seen in Figure 3, a clear visual distinction of plant vs. animal‐derived material could be made with 11 of the 27 cholesterol carbons derived on the signature of relative mole fraction of ¹³C for each specific carbon atom in cholesterol.

FIGURE 3.

(A) Comparison of relative isotope fractions for animal‐derived and plant‐derived cholesterol indicating clear divergence in the two sets of samples for 11 out of 27 individual carbons. (B) Four isochronous carbons at 42.0 and 31.8 ppm respectively.

Incorporation of a semi‐synthetic plant‐derived cholesterol sample highlights a subset of carbon positions where both the plant and semi‐synthetic samples show consistent and reproducible divergence from the animal‐derived counterpart (Figure 4). When compared to the initial analysis where 11 carbon positions exhibit clear separation between plant‐ and animal‐derived cholesterol, inclusion of the additional plant sample shows 8 of the 11 site‐specific carbons remain in strong agreement between the two plant datasets, showing distinct divergence from the animal sample. At these positions, the semi‐synthetic cholesterol aligns closer with the original plant material, demonstrating reproducibility within plant‐derived cholesterol and confirming that these isotopic differences are not sample‐specific anomalies. These 8 carbon positions therefore represent the most reliable markers for distinguishing plant‐ and animal‐derived cholesterols.

FIGURE 4.

(A) Comparison of relative isotope fractions for animal‐derived, plant‐derived, and semi‐synthetic cholesterol, indicating clear divergence of the plant‐derived and semi‐synthetic materials from the animal‐derived materials. Error bars for the semi‐synthetic sample omitted for clarity but can be visualized in the Supporting Information. (B) Four isochronous carbons at 42.0 and 31.8 ppm, respectively.

Using the relative isotope fractions of plant‐ and animal‐derived cholesterol, a cross‐validated, orthogonal partial least squares‐discriminant analysis (OPLS‐DA) model was generated (Figure 5). Using the relative isotope fractions of semi‐synthetic cholesterol as a test set, the model classified all eight samples as plant‐derived cholesterol with a cross‐validated class error rate of 0%. This result confirms the ability of irm‐¹³C NMR to accurately differentiate between animal‐ and plant‐derived cholesterol standards and the ability to accurately classify cholesterol of unknown derivation to the correct source.

FIGURE 5.

Orthogonal partial least squares‐discriminant analysis (OPLS‐DA) of ¹³C NMR data for animal‐derived, plant‐derived, and semi‐synthetic cholesterol. (A) OPLS‐DA model of animal‐derived cholesterol in red versus plant‐derived cholesterol in green (N = 9 and 8 respectively) with shaded regions representing the 95% confidence region for each class. The model was cross‐validated using Venetian blinds (13% data left out per blind), giving a cross‐validated class error average of 0%. (B) Class prediction model of semi‐synthetic cholesterol samples using OPLS‐DA model in panel (A). Test samples are shown in blue (N = 8) with all semi‐synthetic samples accurately predicted as being predominantly derived from plant‐derived cholesterol.

4. Conclusions

The application of IRMS and irm‐¹³C NMR provided a robust and reproducible methodology for differentiating plant‐, animal‐, and semi‐synthetic‐derived cholesterol. The bulk isotopic measurements obtained by IRMS revealed measurable differences in the overall ¹³C incorporation among the cholesterol sources, and irm‐¹³C NMR enabled a deeper site‐specific interrogation of isotopic composition, thus yielding a distinctive molecular fingerprint that directly reflects biosynthetic origin. The eight site‐specific carbon positions conserved across the plant‐, animal‐, and semi‐synthetic cholesterol samples represent the carbon atoms that are most susceptible to minute isotopic differences arising from divergent biochemical and enzymatic processes. Although the limit of quantification for site‐specific ¹³C measurements is influenced by the reduced sensitivity of inverse probe configurations, as observed in TXI probes, the sample quantities used in this study provide sufficient precision for classification and did not limit the applicability of the method. Multivariate analysis further reinforced these findings, with OPLS‐DA models providing complete distinction between animal‐ and plant‐derived cholesterol and correctly classifying semi‐synthetic cholesterol as predominantly plant‐derived. Using the observed differences in ¹³C distribution in cholesterol from plant and animal sources may be well suited for regulatory compliance, quality control, and fraud prevention in cosmetics, nutraceuticals, and pharmaceuticals, where verification of ingredient origin is increasingly critical.

Author Contributions

Anika M. Singh: conceptualization, writing – original draft, methodology, investigation. Subir Chakraborty: conceptualization, supervision, funding acquisition. Remington X. Poulin: conceptualization, writing – original draft, writing – review and editing, funding acquisition, methodology, investigation, supervision, software. R. Thomas Williamson: conceptualization, writing – original draft, supervision, investigation, methodology, project administration, writing – review and editing, resources, funding acquisition, software.

Conflicts of Interest

Subir Chakraborty declares that he is CEO of Chemo Dynamics Inc. The other authors declare no conflicts of interest.

Supporting information

Data S1: Supporting information.

Data S2: Supporting information.

Data Availability Statement

The tabulated NMR data and assignments that support the findings of this study are available in the Supporting Information of this article as files named “NMR and MS data for model_01262026” and “Plant vs Animal Cholesterol_supporting information_01262026.” The raw ¹³C NMR data generated from this study are available upon request to the corresponding author.