The Untargeted Capability of NMR Recognizes Nefarious Adulteration in Natural Products

Abstract

Curcuma longa (turmeric) has a long ethnomedical background for common ailments, and “Curcumin”-containing Dietary Supplements (CDS) are a highly visible portion of today’s self-medication market. Owing to raw material cost pressure, CDS products are affected by economically motivated, nefarious adulteration with synthetic curcumin (“syncumin”), leading to unexpected toxicological issues due to “residual” impurities. Using a combination of targeted and untargeted (phyto)chemical analysis, this study investigated the botanical integrity of two commercial “turmeric” CDS with vitamin and other additives that were associated with reported clinical cases of hepatotoxicity. Analyzing multi-solvent extracts of the CDS by 100% quantitative ¹H NMR (qHNMR), alone and in combination with countercurrent separation (CCS), provided chemical fingerprints that allowed both the targeted identification and quantification of declared components and the untargeted recognition of adulteration. While confirming the presence of curcumin as a major constituent, the universal detection capability of NMR spectroscopy identified significant residual impurities, including potentially toxic components. While the loss-free nature of CCS captured a wide polarity range of declared and unwanted chemical components and increased dynamic range, (q)HNMR determined their mass proportions and chemical constitutions. The results demonstrate that NMR spectroscopy can recognize undeclared constituents even if they represent only a fraction of the mass balance of a dietary supplement product. The chemical information associated with the missing 4.8 and 7.4% (m/m) in the two commercial samples, exhibiting an otherwise adequate curcumin content of 95.2 and 92.6%, respectively, pointed to a product integrity issue and adulteration with undeclared synthetic curcumin. Impurities from synthesis are most plausibly the cause of the observed adverse clinical effects. The study exemplifies how the simultaneously targeted and untargeted analytical principle of 100% qHNMR method, performed with entry-level high-field instrumentation (400 MHz), can enhance the safety of dietary supplements by identifying adulterated, non-natural “natural” products.

GRAPHICAL ABSTRACT

Representing a continuously growing part of the global natural product goods including foods, dietary supplements are an extensive marketplace currently valued at ca. $8+B in U.S. sales and $200+B globally. It is well established that plant-derived dietary supplements such as crude plant materials, their extracts, and products derived from them are chemically complex. Their composition reflects the metabolome(s) of the producing organism(s), i.e., the source plant(s), their associated microorganisms, and the processing methods used. In contrast to this widely acknowledged complexity, reductionist approaches continue to prevail in both the chemical analysis and biological evaluation of dietary supplements. In fact, as evident from dietary supplement labels (see ODS/NIH Dietary Supplement Label Database [DSLD] at dsld.od.nih.gov), the majority of contemporary standardization protocols and (botanical) product integrity analyses, including compendial and regulatory methods, are limited to only one or very few readily accessible phytochemicals. These compounds often function as chemotaxonomic rather than pharmacologically relevant markers, and the standardization approaches tend to neglect bioactive and metabolically variable components of the dietary supplements.

This already concerning (phyto)chemical reductionism translates into a potentially even more problematic biological reductionism: the majority of biological studies with dietary supplements and their major constituents focus on a relatively small number of “prominent” compounds. This leads to overemphasis of the perceived supplemental or therapeutic potential of relatively few natural molecules. “The literature” describes these natural products as having a myriad of purported beneficial biological properties, which leads to health and related claims for which rigorous substantiation is frequently lacking. More plausibly, the majority of the overly prominent molecules could be considered as being IMPs, i.e., Invalid/Improbable/Interfering Metabolic Panaceas,^1-3 which are highly abundant in their source organisms while tending to be elusive as rational bioactive principles.

The Importance of “Yellow” Terminology - Label vs. Content.

The term “curcumin” has been utilized - often confusingly - to describe five different types of materials that have a striking deep yellow to orange color. They are all marketed as “curcumin”-containing dietary supplements (CDS):³ (i) turmeric (T), the raw turmeric (Curcuma longa) rhizome; (ii) turmeric extract (TE), the solvent extract of dried or fresh C. longa rhizomes; (iii) curcuminoid-enriched turmeric extract (CTE), typically obtained via precipitation at lower temperatures;⁴ (iv) curcuminoids-enriched materials (CEM), obtained from CTEs via chromatographic purification; (v) curcumin (1) as a single component, developed as highly purified, validated reference material and/or used for botanical standardization. As these five materials have distinct chemical compositions, with complexity decreasing from (i) to (v), their biological activity profiles have to be different.

Existing confusion about the term “curcumin” and the labels given to CDS has already led to many incomparable and/or irreproducible reported biological results.² It is widely accepted that describing the phytochemical composition of any crude dietary supplement material is a requirement for the accurate biological evaluation of its potential therapeutic value.⁵ However, this basic tenet does not apply to the majority of CDS: their popularity as both a marketed product and a research target, along with a frequent lack of specificity including in the documentation, has blurred the lines between the single constituent, the crude plant extract, and all intermediate stages [see (i)-(v) above] of CDS products.

Pharmacokinetic properties complicate the matter further: curcumin (1; Scheme 1), the single compound, is very poorly absorbed, as are its congeners, demethoxycurcumin (2) and bisdemethoxycurcumin (3). Only traces of degradation compounds of 1 appear in the blood during the time when most of the curcumin is excreted in the feces.^6,7 Owing to poor absorption of 1 into the bloodstream, piperine (4) has been recommended to enhance the curcumin absorption by 20-fold.⁸ However, this approach of metabolic inhibition is neither specific for 1, nor does it enhance 1 titers enough to reach pharmacologically relevant levels.

Scheme 1.

Chemical Structures of the Ingredients of the Studied Curcumin Dietary Supplement (CDS).

The Consequences of Using a Single Compound to Represent a Whole Plant.

In the contemporary literature, the compound curcumin (1) is widely used as a sole “representative” of the plant, C. longa. This error has escalated to a point where products are sold in the marketplace that contain synthetic 1 (“syncumin”) rather than naturally derived 1 of pharmaceutical-grade purity (“purcumin”³). Moreover, such products are sometimes even labeled as being or containing C. longa extracts, which are allegedly standardized to a certain content of 1 and its congeneric curcuminoids 2 and 3. The discovery of this form of fraud has previously led to the investigation of its adverse effects in humans.^9-11

Liver toxicity has been associated with CDS products labeled as preparations that contain turmeric in a number of case reports.^12-14 Moreover, between May and June of 2019, at least 21 cases of acute non-viral cholestatic hepatitis were reported in Italy and associated by authorities with over 20 different turmeric products. In a subset of products for which information was available, daily dosage recommendations ranged between 50 and 1,200 mg of “turmeric extract”, labeled as being standardized to 95% curcuminoids, and containing 5 to 160 mg of pepper (Piper nigrum, Piperaceae) extract, labeled as containing 95% piperine. To prevent further cases, in July 2019, the Italian Ministry of Health eventually requested manufacturers of turmeric food supplements to include a warning about avoiding turmeric in patients with pre-existing liver conditions, in line with a recommendation by the European Medicines Agency (EMA) in 2018.¹⁵ Moreover, in cases of concomitant use of API-based medicines,¹⁶ the Italian Ministry of Health, through testing, had excluded contaminants, reaction by-products, solvent residues, and substitution with other plants as possible causes of the observed hepatotoxicities. However, the authorities also stated that there was no correlation between turmeric [not, e.g., curcumin! note implications from the terminology] itself and liver injury.¹⁷

The Focus on “Yellow” Inspires Nefarious Adulteration Practices.

Another major complicating factor in the CDS supply chain is the occurrence of adulteration. The intense yellow to orange color of CDS is both visually pleasant and a cover for inappropriate additives. The addition of undeclared lower-cost ingredients to powdered turmeric root/rhizome and turmeric extracts is a major concern among regulators, healthcare professionals, and responsible members of the dietary supplement and herbal medicine industry. Reports of turmeric adulteration include substitution with other Curcuma species (e.g., C. zedoaria, C. xanthorrhiza, or C. aromatica), or the undeclared admixture of starches or yellow/orange dyes. Pigments such as Metanil Yellow, lead chromate, or Sudan Red have been added even to crude turmeric to produce a visually more appealing product.¹⁸ These practices appear to be more frequent in rural areas of India and Bangladesh. Particularly disturbing is a recent publication documenting elevated blood levels of lead in pregnant women living in rural Bangladesh due to the ingestion of turmeric adulterated with lead chromate.^19,20

Synthetic curcumin can be made for about one-third of the cost of natural curcumin (note the difference between pure 1, isolated from the plant [“purcumin”],³ and products marketed as being or containing “curcumin”), providing a financial incentive to unethical suppliers for diluting or replacing turmeric extracts or refined turmeric products with synthetic materials. A 2019 study used ¹H NMR spectroscopy and HPTLC to investigate the quality of 56 turmeric food supplements sold in Europe and the United States. It found that at least three of these supplements contained synthetic curcumin or synthetic curcuminoids based on the absence of the typically co-occurring turmeric compounds.²¹ In addition, a number of products claiming to contain piperine did not show the characteristic signals of the aromatic hydrogens in the ¹H NMR spectrum or the yellow band for the alkaloid in the HPTLC traces.²²

In order to develop a more untargeted approach for recognizing “curcumin fraud”, the present study considered known targeted analytical methods that have employed NMR, MS, and countercurrent separation methodologies.^3,23-28 The only currently accepted proof that “curcumin” is synthetic utilizes isotope ratio mass spectrometry (IRMS).^29,30 For the detection of synthetic 1, the determination of ¹⁴C concentrations has proven useful as the synthesis of 1 uses petroleum-based starting materials, which are devoid of ¹⁴C.¹⁸ Contrarily, an equilibrium exists between ¹⁴C intake via photosynthesis and its decay in living plants. Therefore, natural 1 and other natural turmeric compounds contain measurable amounts of ¹⁴C.

The use of HPLC or HPTLC in the assessment of turmeric extract quality usually infers the presence of synthetic curcumin(oids) by the absence of the natural congeners or other compounds typical for turmeric extracts, e.g., ar-turmerone. The frequently observed admixture of other constituents such as (multi) vitamins and other plant extracts (e.g., those containing 4, or 4 itself) can produce HPLC peak overlap and, thereby, challenge even validated QC methods. This provides an additional rationale for the need to consider orthogonal analytical methods such as NMR spectroscopy. Regardless of the analytical challenges associated with the untargeted detection of adulteration, economic calculations clearly show the (misguided) incentive of chemical synthesis over the purification of these compounds from a natural source.

RESULTS AND DISCUSSION

Extraction for CDS Adulteration Analysis.

For non-polar extraction, a multi-solvent process was chosen (Figure S1, Supporting Information). Ten tablets were taken from each of the two batches A (7.98 g) and B (8.07 g) of the identically labeled commercial CDS from the same manufacturer. First, each batch was extracted twice, accompanied by ultrasonic treatment for 20 min, with 160 mL of CH₂Cl₂, yielding 1.15 g (14.4%, w/w) for batch A and 1.29 g (16.0%) for batch B. This was followed by MeOH extraction under the same conditions to yield 2.63 g (33.0%) and 3.22 g (39.9%), respectively (Figure S1, Supporting Information). The combined CH₂Cl₂ and MeOH extracts of each batch were chemically profiled by (U)HPLC-UV and qHNMR. The nonpolar extraction yield of 3.78 g (47.6%) and 4.51 g (52.9%), respectively, showed that the two CDS batches from the same manufacturer had different constitutions, thereby revealing a potential quality (control) issue. For extraction of the declared polar constituents, namely vitamins, five tablets of each batch A (4.01 g) and B (4.05 g) were extracted twice, accompanied by ultrasonic treatment for 20 min, with 80 mL of 50% MeOH/H₂O, yielding 110 mg (2.74%) for batch A and 470 mg (11.60%) for batch B with 50% MeOH/H₂O (20:1, v/w), again using the ultrasonic treatment for 20 min (Figure S2, Supporting Information). Qualitative and quantitative analysis of the extracts showed differences in composition for batches A and B, as also reflected by the significant differences in yield, color, and chemical profiles (Figures S1-S3, Supporting Information).

Choice of Analytical Methodologies.

Reflecting the abundance of HPLC instrumentation in laboratories worldwide, LC is employed widely for qualitative and quantitative including purity analysis via relative referencing. The HPLC chemical profile for the non-polar extract (Figures S4 and S5, Supporting Information) is dominated by the curcumin (1) peak. The HPLC chromatogram also showed a small amount of piperine (4) and vitamin B2 (riboflavin; 6) as well as unidentified peaks of more polar compounds eluting around 6.5 min before 1. Polar extracts were characterized by the presence of 1, 4, and 6. Vitamins B1 (thiamine; 5) and B6 (pyridoxine; 7) were too polar under the chosen conditions to be sufficiently retained for definitive identification (Figure S4, Supporting Information). The LC-based chemical profiles provided useful information relative to the available reference standards, but un-referenced peaks give little added information. Therefore, while LC-based targeted analysis can tentatively identify known chemical adulterants, its untargeted ability to recognize residual complexity is limited by the relatively high selectivity of the hyphenated detection method and/or the captured chromatographic polarity window.

Analysis by ¹H NMR can produce metabolomic fingerprints for chemometric evaluation and detection of adulteration of C. longa powders,^21,31 and MS has been utilized to confirm the presence of compounds following structure elucidation by NMR data.³² NMR spectroscopy offers a meaningful advantage due to its simple sample preparation, high reproducibility, and ability to detect diverse plant metabolites simultaneously. Counter-current separation (CCS) has been utilized for fractionation and isolation of 1 and its analogues from turmeric extracts.^3,33-36 Notably, CCS can recover essentially 100% of loaded sample,³⁷ because the stationary phase is liquid and, thus, all retained compounds can be readily recovered during an extrusion step or by washing.³ The present study employed a combination of CCS for sample preparation and quantitative ¹H NMR (qHNMR) to investigate the commercial CDS connected with the recently observed clinical hepatotoxicities.

NMR Identification of the Chemical Compositions.

The CDS extracts were analyzed by ¹H NMR spectroscopy as an orthogonal method to (U)HPLC-UV (Figure S6, Supporting Information). The results from the non-polar extracts showed 1 as a major component, a trace amount of 4, and signals at 1.2 ppm (aliphatic chain of fatty acid) indicated the presence of fatty material, likely adjuvants used as anti-caking agents (Figure 2 and Figure S6, Supporting Information). From the polar solvent extracts, only the presence of 4, 6, and polysaccharides exhibiting diagnostic signals in the range 2.8 - 5.8 ppm along with impurities was observed. However, neither 5 claimed to be present at 2.7 mg/2 tablets nor 7 claimed to be present at 3.3 mg/2 tablets could be detected in the CDS extracts (Figure S6, Supporting Information). Thus, either these two vitamins were difficult to extract from the tablets with the chosen solvents, or their contents in the tablets were below the LODs of the two analytical methods.

Figure 2. Residual complexity of non-natural “curcumin”.

The ¹H NMR spectra, processed with Lorentzian-Gaussian resolution enhancement, of the CDS extracts, batches A and B, revealed the absence of the natural curcuminoid congeners, demethoxycurcumin (2) and bisdemethoxycurcumin (3). Instead, several non-natural aromatic components were present that gave rise to AMX-type signal patterns similar to 1: ~8 Hz doublets (impurity signal, e.g., at δ_H 8.58); ~2 and ~8 Hz doublets of doublets (e.g., impurity signal centered δ_H 7.00); ~2 Hz doublets (e.g., impurity signal centered δ_H 7.19); ~8Hz doublets (e.g., impurity signal centered at δ_H 7.68). The residual complexity patterns of the commercial samples were distinct from those of natural 1 and its natural precursors.

The results demonstrate that (U)HPLC-UV and ¹H NMR spectroscopy are complementary methods for creating chemical profiles. (U)HPLC-UV detected compounds with high UV absorption at 263 nm as chosen wavelength, but was blind to the presence of fatty and saccharide components. On the other hand, ¹H NMR spectroscopy not only observed the resonances belonging to the compounds detected via (U)HPLC-UV, but provided additional characteristic signals attributed to unknown components that were undetectable and/or unidentifiable by (U)HPLC-UV without the use of identical reference standards.

The labels of the investigated CDS declared that both products contained “95% curcuminoids [plural! note implications from this terminology] from dried turmeric rhizome extracts (500 mg/tablet)”, suggesting a natural source that contains multiple constituent curcuminoids. The labels also declared that 1 accounted for 76% of the content, while the remaining 19% of the 95% were due to 2 and 3. The ¹H NMR and (U)HPLC-UV analyses of the non-polar extracts confirmed the presence of the major component, 1. However, when the ¹H NMR spectra of the non-polar extract are compared with the spectra of 2 and 3, it is evident that the extracts failed to show the characteristic signals of 2 and 3 in the 6.65 - 6.75 ppm interval (Figure 1). Even when the baselines from the same spectra were enlarged in Figure 2, there was no evidence of signals that define 2 and 3 in the 6.65 - 6.75 ppm interval. The NMR spectra indicated, therefore, that the source of the declared “curcuminoids” in the investigated commercial products lacked the residual complexity associated with nature-derived 1. The dihydro derivatives indicative of a natural CDS³ were also absent. Inevitable residual complexity (go.uic.edu/residualcomplexity) results from the biosynthetic origin of the material and reflects the fact that a certain (“residual”) amount of close analogues always accompany the major component as impurities, even after multiple large-scale enrichment or high-resolution laboratory purification steps.

Figure 1. Comparison of the aromatic and olefinic range of the ¹H NMR profiles.

The extracts of the “curcumin”-containing dietary supplement (CDS), batches A and B. with those of authentic curcumin (1), demethoxycurcumin (2), and bisdemethoxycurcumin (3) showed the lack of detectable natural congeners, thereby indicating adulteration with synthetic 1.

Fingerprint Analysis of “Curcumin”-containing Dietary Supplements (CDS).

Chemical fingerprints capture the complexity of a botanical dietary supplement, which is a result of their diverse chemical composition and as the diagnostically different abundance levels of their constituents. Chromatographic and spectroscopic fingerprints reveal known and/or unknown constituents by being targeted and/or untargeted, respectively. Of the most suitable untargeted techniques, NMR and FT-IR,^39-41 NMR spectroscopy provides the more dispersed spectra with less signal overlap and, thereby, offers better capabilities for both targeted and untargeted analysis. This dual capability of (q)NMR, in combination with simultaneous quantification, was highlighted by the present outcomes of the present CDS analyses.

Mass Balance without Weighing and Free of Weighing Error.

To further characterize the minor components of the residual complexity in the non-polar extracts, 100% qHNMR was applied. This method is highly suitable for adulteration analysis as it forces the analyst to interpret as many signals as possible, ideally all observed signals (“100%”). While this approach remains asymptotic by nature, it is still a powerful yet generally underappreciated.

The fact that the 100% qHNMR method does not consider sample weight might appear to be counterintuitive for a quantitative method, but in fact, it bears a major advantage in quantitation as it is independent of any weighing errors, including deviations from humidity effects. The main error of the 100% qHNMR method is systematic and arises from chemical species which cannot be accounted for due to the lack of detectable ¹H nuclei and/or major disproportions in relative molecular weights. While the analyst must bear this in mind, it is a relatively minor limitation in the practice of natural product and pharmaceutical analysis, at least in the authors’ long-term experience. Notably, the unity response factor of qNMR remains a major intrinsic advantage for mass-balance and adulteration analysis, which outweighs the systematic error.

Interestingly, 100% methodology is ubiquitous in hyphenated LC-based quantitation. Thus, there is no reason why the well-established and easy-to-use 100% “mass balance” quantitative evaluation method, consisting of integrating all visible peaks and determining their ratio on a 100% basis, can be the gold standard in (U)HPLC and its many derivative methods, but could not be used equally widely in qHNMR. In fact, the 100% qHNMR method should be considered more seriously as a very practical method that can be readily applied to almost any routine qualitative ¹H NMR data set.³⁸ Notably, as a metrological method, qHNMR has the advantage of being a metrological method with validation characteristics superior to LC-based methods (see ref ³⁸ and refs therein).

Practicability of 100% qHNMR Fingerprint Analysis.

NMR spectroscopy addresses one major limitation of HPLC quantitation, i.e., its need for calibration curves that require the availability of identical reference standards for each target analyte. As an orthogonal method, qNMR methods offer both absolute (EC and IC qHNMR; abs-qNMR) and relative (100% qHNMR; rel-qNMR) quantification for purity and integrity analysis. The advantage of the 100% qHNMR method is that it is both untargeted and independent of identical reference standards. It also does not necessarily require knowledge of the mass of a detected molecular species as it measures relative molarity. Moreover, 100% qHNMR does measurements with a unity response factor, regardless of the mass or identity of the impurities.

In the present study, data processing was performed using a previously optimized approach for qHNMR spectra,⁴² with the following parameters: a Lorentzian-Gaussian resolution enhancement (LG) with a Gaussian factor of 0.8 Hz and a line broadening factor of −1.5 Hz was applied; all 64k-sized time domain spectra were zero-filled to 512k data points to increase the digital resolution; a 5th order polynomial fit was applied for the baseline correction.

Table 1 provides a summary of the findings from applying the 100% qNMR method to the two non-polar extracts. The curcumin content was 95.2 and 92.6% in batches A and B, respectively, of the curcumin dietary supplement. In addition, batches A and B contained some 4 (0.26 and 0.76%, respectively) when signals that coincided with the signals produced by 4 reference material were integrated (Figure 4). The presence of curcumin(oid) degradation products in the non-polar extracts was confirmed by comparing their ¹H NMR spectra to those of purified natural curcumin (Figure 4). NMR spectra of natural 1 inevitably show the presence of degradation compounds as minor constituents and were utilized as reference for the same degradation products in the spectra of the investigated CDS. After the presence of 4 as well as likely degradation products was accounted for, unidentified aromatic impurities were observed and quantitated (Figure 4). In line with pharmacopoeial practices for unidentified impurities, the molecular weight of non-curcuminoid impurities was considered as the molecular weight of 1. Four different types of aromatic impurities were distinguished, and their relative abundance measured via integration. Impurity 1 was associated with signals at 8.084, 7.905, and 7.870 ppm that had identical integration values, impurity 2 was associated with the signal at 7.893 ppm, impurity 3 was associated with the signal at 7.642 ppm, and finally, impurity 4 was associated with signals at 7.205 and 7.003 ppm. It is very likely that there are other signals from the compounds that are obscured by overlapping absorbances. The origin of these signals is unknown. It is possible that these impurities may have been left behind by the starting materials and the reagents used in a synthetic procedure.

Table 1.

The Content of Curcumin (1) in the Commercial Samples Was Determined Using the 100% qHNMR Approach, Capturing Both Known and Unknown Constituents on a Relative Molarity Basis and Converting Them to % Values by Calculating the Unknown as 1. The Non-natural Curcumin Content in the Curcumin Dietary Supplements (CDS) was 95.2 and 92.6%, respectively, which Appear Unusually High Compared to Typical Plant-derived Curcuminoid Preparations.

	quantities of curcumin in the dietary supplements
component	δH	batch A (ppm, %)		average integral value
curcumin	7.323 (97.78)	7.151 (100.00)		98.89
impurity 1	8.084 (0.13)	7.905 (0.13)	7.870 (0.14)	0.13
impurity 2	7.893 (0.25)			0.25
impurity 3	7.642 (0.10)			0.10
impurity 4	7.205 (4.51)	7.003 (4.58)		4.55
curcumin content (%): 95.2
		batch B (ppm, %)		average integral value
curcumin	7.323 (97.40)	7.151 (100.00)		98.70
impurity 1	8.084 (0.11)	7.905 (0.11)	7.870 (0.12)	0.11
impurity 2	7.893 (0.24)			0.24
impurity 3	7.642 (0.08)			0.08
impurity 4	7.205 (7.49)	7.003 (7.48)		7.49
curcumin content (%): 92.6

Figure 4. Assignment of the ¹H NMR Aromatic Fingerprints in CDS.

Both batches A [D] and B [C] of the CDS contained the expected analytical targets: observation of the same major and minor signals as present in the spectrum of “pure” reference material (B) indicated the presence of an inevitable degradation compound in the main constituent, curcumin (1). The NMR fingerprints also contained the diagnostic signals of piperine (4; reference spectrum in A), despite its relatively low abundance at 0.54% and 2.01% for batch A and B, respectively. In addition, four aromatic impurities were recognized in both CDS batches. Among the impurities compounds, impurity 4 revealed an AMX spin system which may synthetic precursor has the partial structure as the aromatic part of curcumin. For the relative qHNMR analysis, the integrals of the impurities were selected as indicated by the boxes (see Table 1).

While the values measured for 1 alone fell within the range of declaration (“95%”), the findings were in conflict with regard to two other declared properties: (i) the specific content of 1 was much higher than declared, which is in line with the observation that 1 was the only curcuminoid present; and, (ii) the demethoxy (2 and 3) and dihydro curcuminoids, which constitute the inevitable residual complexity of C. longa products derived from nature, were below the detection limit of the qNMR method, which was reached at a signal to noise ratio <1.5:1. This absence indicated that the material could not have been produced from plant material. Overall, the quantification results for 1, the absence of natural curcuminoid congeners, and the identification of unknown aromatic impurities implied that the investigated material was adulterated.

Countercurrent Fractionation of “Curcumin”-containing Dietary Supplements.

To understand the chemical constituents more fully, chromatography was employed to distribute chemical species with overlapping NMR signals into separate fractions. In a previous study, curcumin knock-out methodology was established using countercurrent separation (CCS) methodology.³ Specifically, centrifugal partition chromatography (CPC) was employed to fractionate the non-polar extract, using the two-phase solvent system composed of HDiMWat (3:7:7:3, hexanes-CH₂Cl₂-MeOH-water). The separation was performed in normal phase (ascending) mode, employing both elution and extrusion steps. The weight distribution of the resulting fractions was as follows (Figure 3): Fr. 1 8.39 mg (7.1%), Fr. 2 90.49 mg (76.0%), Fr. 3 6.43 mg (5.4%), Fr. 4 12.06 mg (10.1%), Fr. 5 1.71 mg (1.4%), Fr. 6 < 1mg (<1%). The curcumin knock-out materials obtained from both CDS were then compared with reported curcuminoid-enriched turmeric extracts. Using authentic reference compounds for comparison, Fr. 1 showed the presence of 4, Fr. 2 was identified as 1, and Frs. 3 and 4 represent minor amounts of 1 eluting at the tail-end of the chromatographic peak of 1. Observed NMR signal shifts are likely due to differences in residual water, concentration, and interaction/complexation with metal ions. However, the congeners 2 and 3, which are more highly retained in this solvent system, could not be detected at all, indicating the absence of a natural mixture of natural curcuminoids in the CDS. The polar fractions 5 and 6 showed the typical signals of saccharides (3.0 - 5.2 ppm) (Figure 3). CCS-based curcumin knock-out methodology is orthogonal to NMR, widens the dynamic range of subsequent analysis (see also Figure 3), and, by virtue of its liquid-liquid partition mechanism and 100% recovery capability, can help confirm or refute the phytochemical integrity of a metabolome.

Figure 3. Aromatic range of the ¹H NMR spectra of the CPC fractions of one of the “curcumin”-containing dietary supplements.

CPC was employed as an orthogonal and pre-separation method for two reasons: first, to widen the chemical window and dynamic range for the targeted identification of the declared curcuminoids, vitamins, and piperine alkaloids; second, for the untargeted detection of unwanted constituents. The separation was performed as in normal phase (ascending) mode with hexanes-CH₂Cl₂-MeOH-water (3:7:7:3) as a two-phase solvent system. The components eluted and eventually extruded (EECCC) in order of increasing polarity and were divided into seven fractions (see main text for weight distribution). The vertical scale of the spectra was chosen to show the major components. The intensity scaling factors reflect the vastly different abundance of the various components and highlight the achievable dynamic range enhancement of CCS pre-fractionation.

Breadth of Expected Chemical Variability.

Plant metabolites exhibit wide structural variability, reflecting their wide range of functions for growth, survival, and reproduction.^43-45 Plants also produce metabolites ahead of their death and to protect themselves from abiotic stress, harmful environmental pollution, and other external danger such as herbivores or for interspecies defense.^46-48 Each plant produces specific chemical markers with minor phytochemical derivatives that are involved directly or indirectly in their survival. ^46,49,50 The production of all these constituents is affected by the condition of soil, cultivation, and time of harvest.^51-53 While a single, ideally highly taxon-specific metabolite can be used as a reference standard for ID and dietary supplement fraud detection under specific circumstances,^54,55 this approach does not capture the chemical variability of plants. This points to the contemptuous question how natural chemical variation can be distinguished from adulteration.

As targeted analyses typically test for the presence or abundance levels of “marker” metabolites, they are blind to adulteration outside of the probed very narrow (i.e., targeted) chemical window. In contrast, untargeted analysis has the capability of reaching outside this window. Distinguishing between acceptable and unacceptable differences can utilize detailed knowledge of natural variation, e.g., by utilizing a body of reference data from authentic materials, with visual or chemometric evaluation. Such a distinction can also involve the recognition of more or less prominent analytical features that are either missing or extraneous or both. Examples of potentially missing features are the total absence of the signals for Curcuma oleoresins in CDS; examples of extraneous features are the presence of signals of aromatic compounds that are incompatible with the expected natural chemistry, such as those of synthetic precursors.

In addition to the consideration of targeted vs. untargeted analytical methods, regulatory aspects impact the potential answers to the above question. In the United States and the European Union, the rhizome of C. longa L. and preparations derived from it are permitted products as set forth in the 1994 Dietary Supplements, Health, and Education Act (DSHEA 1994) and the Traditional Herbal Medicinal Products, a pan-EU harmonized legislation, respectively. Neither provision has specifications or requirements to prevent adulteration. In the EU, such preparations can also be used in food supplements, the regulation of which is largely the responsibility of individual member states. Countries such as Belgium, France, and Italy have developed lists of plants, preparations permitted in food supplements. However, such lists do not address identification, authenticity, or prevention of adulteration. Specifically, before 2019, Italy had listed C. longa L. as a permitted plant, with no conditions of use. Since the 2019 adverse reports, mandatory warnings have been introduced, albeit with no further restrictions.⁵⁶

However, no rules were introduced in reference to the potential adulteration of C. longa products. This applies to Belgium as well, which introduced further mandatory warnings in 2019, but also imposed a maximum limit of 500 mg of curcuminoids per day for turmeric extracts with natural bioavailability, or equivalent amounts in case of enhanced bioavailability.⁵⁷ The limit is based on food additive approval of “curcumin” (see the Introduction regarding general issues with the specificity of the nomenclature) as a food color and related uses.⁵⁸ Other EU member states such as Ireland⁵⁹ or France had or have also relied on maximum limits derived from food additive studies. Unfortunately, none of these regulatory actions took into account the authenticity of the plant source including the natural nature of the source. In terms of specific and new preparations, an application to market tetrahydro curcuminoids from turmeric has been recently submitted to the European Commission.⁶⁰ With no apparent evidence of significant human consumption in the EU before 1997, synthetic 1 is considered an unauthorized novel food that cannot be marketed.

Methods for Targeted vs. Untargeted Detection of CDS Adulteration.

Detecting the many different types of adulteration demands the use of orthogonal analytical methods. As the most commonly used, HPTLC and HPLC-UV/Vis can detect other Curcuma species and some of the colorants.^24,61-65 The detection of lead chromate is challenging: it escapes routine testing methods due to its poor solubility in organic solvents and water, thus requiring targeted methods such as ICP-MS or flame atomic absorption spectroscopy.^20,66 The Food Safety and Standards Authority of India also proposed an ashing/sulfuric acid-diphenylcarbazide test as a rapid screening for the lead but required confirmation by quantitative assays to confirm the level of contamination.⁶⁷ Admixture of starches to powdered turmeric is readily detected by microscopic powder examination using iodine stain or a xylene mount with full polarization. In the protocol here, starches are recognized as insoluble residues of extraction.

So far, the identification of undeclared synthetic curcumin has been achieved via MS experiments measuring the concentrations of ¹⁴C isotopes. While a strict proof remains difficult, the present ¹H NMR method strongly inferred adulteration with synthetic curcumin and mislabeling of the product as turmeric extract based on the lack of residually complex, natural curcuminoid patterns, as well as the presence of several % of unidentified aromatic impurities that can be assigned to synthetic reagents and/or by products. To date, ¹H NMR spectroscopy has not been widely explored as a means for quality control of turmeric-based ingredients.⁶⁸ However, its untargeted capabilities position the NMR method uniquely to solve many of the quality issues related to turmeric, its extracts and CDS in general.

Conceptually, NMR spectroscopy can readily detect admixtures of and substitution with other Curcuma species, especially when the sample size is sufficient to allow multivariate statistics. Organic impurities, such as Metanil Yellow and Sudan Red dyes, can be identified by their characteristic signals between 7.0 and 8.5 ppm using ¹H NMR data and were not detected in the investigated samples.⁶⁹ As exemplified here, ¹H NMR data can reveal a wealth of information about sample composition and, to some extent, the underlying manufacturing process by close examination of the residual complexity of the sample,^3,69 thus favoring the 100% qNMR approach.

More Comprehensive Coverage of Adulterants.

The present work introduces the combined use of two methods with “100% properties”:

• CCS permits quantitative separation with 100% sample recovery, due to its liquid-only nature; and

• 100% qHNMR captures the entirety of hydrogen-containing chemical species in a (soluble) sample, due to its nuclear mechanism of detection.

In addition to being loss free, CCS can partition the sample into a few, yet chemically highly distinct fractions, which are amenable to 100% qHNMR, separately or after adequate recombination. Figure 3 shows the power of this off-line CCS-qHNMR combination: (a) it captures both major and minor components by assessing the mass proportions of all sample components via the masses of CCS fractions; (b) it provides information about the relative polarity of the constituents via the CCS K-values; and (c) it resolves details of chemical constituents via the NMR δ scale and dispersion, depending on the magnetic field strength used.

Moreover, the initial CCS step enhances the capability of recognizing adulteration with metal salts: while beyond the scope of the present study, insoluble metal salt additives would remain as precipitates for further analysis, whereas soluble additives would be enriched in the aqueous high polarity fraction and amenable to heteronuclear NMR spectroscopic analysis. Considering the documented occurrence of yellow-colored Pb salts as adulterants, this could inspire the development of ²⁰⁷Pb qNMR methods (spin ½ nucleus, ~1% of ¹H sensitivity at 23% natural abundance) for CDS adulteration analysis.

The study also shows that the combination of CCS and qNMR, two highly orthogonal analytical methods, enables concurrent identity and adulteration analysis that is targeted and untargeted simultaneously. Accordingly, while representing a two-step process, such a combination widens the dynamic range of the subsequent NMR analysis by virtue of the ability of CCS to distribute distinct chemical entities into separate fractions. The CCS-based grouping of analytes not only reduces signal overlap, from which any subsequent analysis can benefit, but is also fully quantitative due to the loss-free nature of CCS. Combined with the nuclear view of NMR analysis, this explains the versatility of CCS-plus-qNMR for targeted and untargeted analysis. However, as shown above, qHNMR alone is still capable of making key conclusions about adulteration. As shown in Figure 4, even without CCS pre-fractionation, qHNMR detected the presence of aromatic impurities assigned to synthetic by-products, as well as the minor content of 4. While the latter matched the declared constituent from P. nigrum, the CPC fractions enriched in 4 did not indicate the presence of additional alkaloids. Considering that 4 vastly dominates the alkaloid spectrum of natural P. nigrum extracts, distinction of natural extract from synthetic 4 is challenging at the low level of the declared ingredient in the investigated preparations.

Putting Dietary Supplement Analysis and 100% qHNMR Methodology in Context.

Challenges in the quality control (QC) of botanical dietary supplements are associated with discerning adulterations, contaminants, and impurities as representing “chemical overlays” of an already daunting variety of natural metabolites produced by the plants. Considering the chemical complexity of botanicals, the adulteration of dietary supplements enhances existing difficulties in producing reproducible biological results thereof. The reductionist approach of using single or very few metabolites as biological placeholders and analytical reference standards exacerbates the overall challenge of detecting adulteration and fraudulent products sold in the marketplace.

This study shows how the universal and quantitative detection capabilities of (q)NMR, by itself or in combination with countercurrent pre-separation, provide an advanced foundation for concurrent targeted and untargeted dietary supplement adulteration assays. Demonstrated for two “curcumin”-containing dietary supplements that compromised the liver functions of several consumers in 2019, the approach is fit for recognizing adulteration of non-natural natural product ingredients (here: synthetic 1) in the presence of analytically confounding factors such as multi-vitamin and piperine fortification. These findings underscore the assertion that materials with limited supply, high demand, and/or high value drive economic motivations behind adulteration,^28,70,71 which frequently are produced with additives (here: multivitamins) that complicate the adulteration analysis rather than serving well-developed dietary benefits, or even implied or claimed therapeutic value in the context of the main component. Considering the modest sample preparation requirements and ability to directly analyze extracts, the wealth of qualitative information achievable from 1D ¹H NMR spectra in tandem with their intrinsic quantitative capabilities via 100% qHNMR evaluation make NMR spectroscopy a highly versatile adulteration assay that combines targeted and untargeted in a single method. The approach is suitable for adoption to further applications in dietary supplement, food, herbal medicine, and drug analysis.

EXPERIMENTAL SECTION

General Experimental Procedures.

Curcumin dietary supplements (Lot. A: 19B914, Lot. B: 18L823, Italy) claiming to be turmeric rhizome dry extract contained 95% curcuminoid and 76% curcumin, black pepper fruit dry extract contained 95% piperine, vitamins; vitamin B1 (thiamine hydrochloride), B2 (riboflavin), B6 (pyridoxine hydrochloride), bulking agent; CaCO₃, cellulose (pregelatinized corn starch) and anticaking agent; magnesium salts of fatty acids, silicon dioxide as ingredients. Reference compounds of natural curcumin (part# 81025/lot 191793-106, Cayman Chemical Company, Ann Arbor, MI, USA), thiamine hydrochloride (part# PHR1037-1G/lot LRAB2930, Sigma-Aldrich, St. Louis, MO, USA), riboflavin (part# PHR1054-1G/lot LRAB3710, Sigma-Aldrich), pyridoxine hydrochloride (part# PHR1036-500MG/lot LRAB3720, Sigma-Aldrich) were purchased for identification of chemical profile. Extraction solvents of CDS used were; CH₂Cl₂ (part# 313000ACS/lot C17G20DRM-0000DCM, Greenfield Global, Shelbyville, KY, USA), MeOH (part# 339000000/lot K18H15K10, Greenfield Global), water (Millipore Milli-Q gradient water purification system). UHPLC analyses were using a Shimadzu UFLC (Shimadzu Corp., Kyoto, Japan) Nexera UHPLC system equipped with a Diode Array Detector (SPD-M20-A), and fluorescence detector (RF-20A/20Axs), performed on a Kinetex 1.7 μm XB-C₁₈ 100Å column (50 × 2.1 mm, Phenomenex, Torrance, CA, USA). UHPLC-UV data analyses were processed with the Shimadzu Labsolution software package. HPLC grade solvents, MeCN (part# 75-05-8/lot 183919) and water (part# 7732-18-5/lot 163613) were purchased from Sigma-Aldrich with formic acid (part# 64-18-6/lot B0538111B, Acros organic, NJ, USA). The sonicator, B-32H (Branson, CT, USA), was used for the extraction tablets of CDS. The ¹H NMR analyses (qHNMR) were performed on JEOL ECZ 400 MHz equipped with a Super COOL probe (NM-Z161331TH5SC, JEOL Resonance Inc., Peabody, MA, USA). The NMR samples were diluted in 200 μL of DMSO-d₆ (part# DML-10-10X1/lot 12G-464, Cambridge Isotope Laboratories, Inc., Tewksbury, MA, USA) transferred in 3 mm NMR tubes (part# S-3-HT-7/lot D030915B, Norell, Morganton, NC, USA). The acquired data were processed using the Mnova NMR software package (v.14.1.0, MestreLab Research S.L., A Corunña, Spain).

Extracts Preparation.

For extraction of the non-polar constituents, 10 tablets (#A: 7.979 g; #B: 8.073 g) of each CDS of batch A or B were extracted successively twice with 160 mL of CH₂Cl₂, subsequently extracted twice with 160 mL of MeOH by sonication. The CH₂Cl₂ and the MeOH extracts were combined to analyze the chemical profile by UHPLC-UV and ¹H NMR. For extracting the polar constituents, every five tablets (#A: 4.009 g; #B: 4.051 g) of curcumin dietary supplement for batch A or B were extracted successively with 80 mL of 50% MeOH/water by sonication to analyze the chemical profile. The non-polar and polar extracts were analyzed for the chemical profile throughout UHPLC-UV and ¹H NMR. The non-polar extraction yielded 1.15 g (14.4%) of batch A and 1.29 g (16.0%) of batch B extracted by CH₂Cl₂ and obtained 2.63 g (33.0%) for batch A and 3.22 g (39.9%) for batch B extracted by MeOH. The total extracts yield of non-polar extracts were 47.4% for batch A and 55.9% for batch B. The polar extracts yielded 2.7% for batch A and 11.6% for batch B extracted by 50% MeOH/water. For the quantitative NMR analyses, batch A (12.20 mg) and B (12.32 mg) as non-polar extracts and batch A (12.59 mg) and B (10.83 mg) as polar extracts, reference compounds; curcumin (7.77 mg), piperine (9.00 mg), thiamine hydrochloride (7.59 mg), riboflavin (2.33 mg) and pyridoxine (7.10 mg) were each separately dissolved in 200 μL of DMSO-d₆ and transferred into 3 mm NMR tubes.

Acquisition of Chemical Profile of the Extracts.

UHPLC-UV chemical profile analysis was performed with the Kinetex UHPLC column with solvents composed of (A) water and (B) MeCN both with 0.1 % formic acid: from 5% B linearly increasing to 100% in 25 min at 0.7 mL/min. All extracts and reference compounds were prepared at 0.1 mg/mL with MeOH. 2 μL of the solution was injected for UHPLC-UV analyses under 190 nm of UV absorption.

The qHNMR data were acquired at 25°C under quantitative conditions using a 90° single pulse experiment with a relaxation delay of 60 sec, a receiver gain of 46, and 64 scans. Post-acquisition data processing was performed using zero-filling of the 64k FID to 512k data points, a mild Lorentzian-Gaussian window function (Exponential factor −1.8, Gaussian factor 0.9 in GF mode), and baseline correction with a fifth-order polynomial function. The residual DMSO-d₅ signal was used for chemical shift referencing (2.5000 ppm). The NMR spectra were exported as a jdx (JCAMP-DX) file for quantitative processing in MNova.