Introduction

With the increasing use of complementary and alternative medicines (CAMs), the quality and safety of such products have been called into question. The 2017 global market for herbal medicines alone is currently estimated to be US$107 billion [1] and expected to reach US$115 billion by 2020 [2]. The reasons behind this increased use are numerous, but the popular belief that CAM products are “natural” and therefore less harmful, with fewer side effects than conventional pharmaceuticals, has been suggested as a key reason for the observed growth [2,3,4]. There is now considerable data to challenge this assumption of safety, and CAMs are estimated to cause 23,000 emergency department visits per year in the USA [5]. Adulteration and contamination with pharmaceuticals, heavy metals, and lower quality ingredients are not uncommon and should be a serious safety concern for consumers [6].

Despite these concerns, regulation of CAM is often trust-based so there is a pressing need for new tools to rapidly assess the safety of herbal CAM both pre- and post-market. Rather than using a multi-target analysis, the screening and detection of pharmaceutical adulterants in complex herbal matrices have mostly been focused on specific pharmaceuticals being added to herbal products with similar indications [6, 7] (e.g., the adulteration of male sexual health products with sildenafil [8]). This focused screening can miss contamination of products with pharmaceuticals of unrelated indications [6], drug isomers or novel compounds [9,10,11]. The data [6, 7, 9,10,11] gathered to date clearly demonstrates the need for a rapid, multi-target or untargeted screening method. Until such a method is developed and a comprehensive screen of products on the market is undertaken, the scope of adulteration and contamination of herbal CAM products will remain unquantified.

In this study, we explore the use of a rapid, high-throughput screen for pharmaceutical adulterants using the AxION 2 Direct Sample Analysis–time of flight (DSA-TOF) mass spectrometer. This analysis needs minimal sample preparation and employs rapid, ambient ionization mass spectrometry via the DSA source coupled to a TOF for high-resolution spectral data [12]. This allows for tentative identification of possible adulterated or contaminated products through accurate mass and isotopic pattern matching without the requirement for separation by chromatography. The resulting matches can then be rated in confidence, with subsequent confirmation and quantification of positive results by Liquid chromatography–mass spectrometry (LC-MS), such as employed in our previous work [6] and that by Bogusz, et al. [7]. The benefits of this two-stage approach are that the more complex and time-consuming confirmatory analysis has a smaller sample load, the stratified scoring system allows for prioritizing of samples, and the analysis is carried out over two platforms, giving greater confidence in the results. We used NIST standard reference materials (SRMs) and five previously analyzed CAM products as a proof of concept for testing the efficacy of DSA-TOF to detect adulteration and contamination in herbal CAMs and provide a rapid method for pharmacovigilance.

Methods

Materials

Drug standards amoxicillin, aspirin, brucine, caffeine, chlorpheniramine, cyproheptadine, dexamethasone, diphenhydramine, ephedrine, ibuprofen, lignocaine, mianserin, paracetamol, prednisolone, ranitidine, salicylic acid, sildenafil, simvastatin, strychnine, tadalafil, and warfarin were purchased from Sigma-Aldrich (Sydney, Australia) in the highest purity available. Codeine, diazepam, digoxin, sibutramine, streptomycin, tamoxifen, vardenafil, verapamil, yohimbine, and the internal standard diazepam-D5 were purchased from Novachem (Melbourne, Australia) in the highest purity available. LC-MS-grade methanol and water were purchased from Fisher Scientific (Hampton, USA).

The NIST standard reference materials (SRMs) used were Camellia sinensis (green tea) powder (SRM 3254), multivitamin tablets (SRM 3280), and Ginkgo-containing powder (SRM 3248). The five CAM products (tablets, capsules, and tea) were purchased in Australia in 2014 from pharmacies, health food stores, and traditional herbal retailers. These products had been previously analyzed (manuscript in review) for heavy metal content, pharmaceutical adulteration, and undeclared plant content via next-generation DNA sequencing. The samples were indicated for various uses such as anxiety and depression, blood circulation, stress, “energy,” “brain function,” “mental clarity,” and “mood.”

Sample Preparation

The NIST SRM and CAM samples were ground to a fine homogenized powder using a Precellys lysing kit consisting of CKMix 50-R 2-mL tubes (Bertin Technologies, France). The sample extraction was based on our previous work [6]. The powdered samples were transferred to 2-mL lysis tubes and stored at – 80 °C until use. For methanol extraction, approximately 5 mg of each sample was weighed out in triplicate 2-mL lysis tubes. One milliliter of methanol was added, and samples were then shaken for 80 s at 6500 rpm in a Precellys tissue lyser (Bertin Technologies, France). After shaking, the extracts were centrifuged for 15 min at 13.2 × 104 × g. The supernatant was transferred to fresh 1.5-mL tubes and dried by evaporation using an Eppendorf Concentrator Plus rotary vacuum concentrator (Eppendorf South Pacific Pty. Ltd., Sydney, Australia). Dry extracts were stored at − 80 °C until use. For analysis, 1 mL of 10% methanol with 0.1% formic acid was added to the dried extracts and the samples shaken for 10 min. Ten microliters of reconstituted sample was then spotted for analysis.

A standard mix of 30 drugs from various pharmaceutical classes (listed in section 2.1 above) was used as a positive control and run in triplicate at the beginning of each day. The concentration of each compound was 100 ng/mL with a mass range of 138 to 780 Da.

DSA-TOF Analysis

The DSA is a combined ambient ionization source and sampling platform previously described in Botch-Jones, et al. [13] and Winter, et al. [12]. DSA-TOF (PerkinElmer, Waltham, USA) conditions were as follows: positive acquisition mode, 350 °C probe temperature, 60 psi nebulizer gas (N2) pressure, 4 L/min auxiliary gas flow and drying gas at 3 L/min and 25 °C. The corona current was 4 μA, scan rate was 10 spectra/s, capillary entrance was set to 800 V, endplate heater was 200 V, and the capillary exit was 100 V. The total method run time was 0.55 min, with a sample spot time of 30 s. The Agilent APCI-L low concentration tuning mix (Agilent Technologies, Santa Clara, USA) was used to calibrate the DSA-TOF before analysis each day and was also used as the internal “lock-mass” calibrant during sample acquisition.

Data Analysis

Spectral data were viewed and analyzed using AxION SOLO software, and R. Raw spectra were scanned by AxION SOLO for matches to exact monoisotopic masses and isotope ratios of targeted compounds based on compound formula. Over 350 pharmaceuticals, analogs, and phytochemicals were screened against an in-house library in this first step (Supplementary Table 1). After initial analysis in SOLO, the individual spectra of each sample were processed and analyzed in R using locally developed code. The enviPat R package [14] (v2.2) was used to compile the isotopic patterns of the targeted compounds to compare with the mass spectral data.

Results

Standard Compound Mix

Of the 30 pharmaceuticals in the standard mix, 21 were consistently detected (Figure 1): brucine, caffeine, chlorpheniramine, codeine, cyproheptadine, diazepam, diphenhydramine, ephedrine, lignocaine, mianserin, paracetamol, ranitidine, sibutramine, sildenafil, strychnine, tadalafil, tamoxifen, vardenafil, verapamil, warfarin, and yohimbine. Compounds not detected were amoxicillin, aspirin, dexamethasone, digoxin, ibuprofen, prednisolone, salicylic acid, simvastatin, and streptomycin; reasons for these non-detections are discussed below.

Figure 1
figure 1

DSA spectra and ppm error for 18 of 21 detected pharmaceutical standards. From left to right: ephedrine, paracetamol, caffeine, lignocaine, diphenhydramine, mianserin, chlorpheniramine, diazepam, codeine, rantidine, strychnine, yohimbine, tamoxifen, brucine, verapamil, sildenafil, vardenafil. Not highlighted in this spectra are cyproheptadine, sibutramine, tadalafil, and warfarin

Full size image

NIST Standard Reference Materials

Table 1 shows the data from DSA-TOF analysis of the three NIST SRMs: Camellia sinensis (green tea) powder, multivitamin tablet, and ginkgo powder. The certified composition is compared with that detected by the analysis.

Table 1 Data from DSA-TOF Analysis of the Three NIST SRMs. Ingredients Detected by the DSA Analysis Are Listed in Bold. #Compounds That Were Detected with Error Rates Above 10 ppm
Full size table

Camellia sinensis (Green Tea) Powder

The expected content of the SRM was caffeine, catechin, epicatechin, epicatechin gallate, epigallocatechin, epigallocatechin gallate, gallic acid, gallocatechin, gallocatechin gallate, L-theanine, and theobromine. Caffeine, catechin/epicatechin, and epigallocatechin/gallocatechin were all detected with errors less than 5 ppm for the respective molecular ions [M + H]+. The diastereoisomers catechin/epicatechin and epigallocatechin/gallocatechin cannot be distinguished using this method, so are referred to collectively. Gallic acid and theanine were also putatively identified, but with higher errors of 10.52 and 9.71 ppm, respectively. Kaempferol and quercetin were also detected but are not listed on the certificate of analysis for this NIST standard. These compounds, however, are known to occur in green tea, and several other plants [15]. Theobromine was also detected, but with a high ppm error (> 10 ppm). The gallic acid esters, epicatechin gallate, epigallocatechin gallate, and gallocatechin gallate were not detected.

Multivitamin Tablet

The NIST multivitamin tablet was certified to contain the following vitamins and carotenoids: α-tocopherol, ascorbic acid, biotin, cis-β-carotene, cyanocobalamin, ergocalciferol, folic acid, lutein, nicotinamide, pantothenic acid, phylloquinone, pyridoxine hydrochloride, retinol, riboflavin, thiamine hydrochloride, and trans-β-carotene. In the DSA analysis, pantothenic acid and pyridoxine were identified with 4.69 and 4.51-ppm error, respectively. Nicotinamide, thiamine, and riboflavin were also identified but had errors > 10 ppm. None of the fat-soluble vitamins were detected; however, these compounds were not expected to be extracted or detected using the methodologies employed here.

Ginkgo Powder

The ginkgo powder was listed to contain bilobalide; ginkgolides A, B, C, and J; isorhamnetin; kaempferol; and quercetin. Isorhamnetin (5.05-ppm error), kaempferol (1.98-ppm error), and quercetin (5.17-ppm error) were all detected by the DSA-TOF analysis. However, none of the terpene lactones (ginkgolides and bilobalide) were identified. There was also an alkaloid-like compound that was detected but not definitively identified.

CAM Products

Table 2 shows the data from the DSA-TOF analysis compared with a previous investigation using next-generation DNA sequencing (NGS) and LC-MS (manuscript in review). The previous analysis of the samples had shown no contamination or adulteration with heavy metals or pharmaceuticals. Due to the processing of the products, which are mostly extracts, DNA material was not expected to be recovered in 3 of the 5 samples. These Ginkgo biloba products, samples 92 and 157, showed the same contents as the ginkgo SRM with isorhamnetin, kaempferol, and quercetin all detected. The Skullcap tea, sample 175, also contained isorhamnetin, kaempferol, and the flavonoid wogonin. All the compounds in samples 92, 157, and 175 had errors ≤ 5 ppm.

Table 2 Comparison of Analytes Detected in Five CAM Samples Using DSA-TOF Analysis Against That Found Using a Combination of LC-MS and Next-Generation DNA Sequencing. Ingredients and Previously Detected Compounds and DNA Which Agree with the DSA Analysis Are Listed in Bold. *Catecholamine and Alkaloid-Like Compounds Were Detected but Not Definitively Identified
Full size table

The ingredients listed for sample 73 were vitamins ascorbic acid, biotin, calcium pantothenate, cyanocobalamin, nicotinamide, pyridoxine hydrochloride, riboflavin sodium phosphate, thiamine hydrochloride, Eleutherococcus senticosus (Siberian ginseng), and Paullinia cupana (guarana). Caffeine was detected in both analyses of sample 73; nicotinamide, pantothenic acid, pyridoxine, and trigonelline were also detected. All the compounds detected in sample 73 had errors < 5 ppm.

The multivitamin, sample 100, was listed to contain several B vitamins including calcium pantothenate, nicotinamide, and pyridoxine hydrochloride, along with Rhodiola rosea, calcium hydrogen phosphate, choline bitartrate, chromium picolinate, inositol, potassium sulphate, and zinc sulphate monohydrate. Similar to the SRM multivitamin, only nicotinamide, pantothenic acid, and pyridoxine were detected, with errors of approximately 9 ppm.

Discussion

This paper aimed at assessing the appropriateness of the DSA-TOF as a rapid, high-throughput screening method for herbal medicines/CAMs. A high-throughput approach has been advocated by the authors [16] to aid the implementation of a more thorough approach to pre- and post-market auditing and overall regulation of herbal CAMs. Until such a method is developed and a comprehensive screen of products on the market is undertaken, we will not have a clear picture of the scope of contamination and adulteration of herbal CAM products. The obvious benefits of the DSA-TOF method are rapid analysis and minimal sample preparation, as seen in this study where simple methanol extracts were each analyzed in less than 1 min.

Another benefit of this method is the collection of untargeted data, with the method only becoming “targeted” when a specific library is applied during data analysis. In this study, the target database consisted of over 300 pharmaceuticals, vitamins, phytochemicals, and analogs of these compounds. This broad coverage, coupled with the ability for retrospective analysis beyond the original target database, is a substantial benefit of the DSA-TOF analysis. In some cases, the DSA-TOF was unable to detect certain standards: amoxicillin, aspirin, dexamethasone, digoxin, ibuprofen, prednisolone, salicylic acid, simvastatin, and streptomycin (see Supplementary Table 2 for predicted detections). As the analysis was only carried out in positive ion mode, detection of aspirin, ibuprofen, and salicylic acid was not expected, as these compounds are typically analyzed in negative ion mode [17]. As the DSA is a modified APCI source [12], the inability to detect amoxicillin, dexamethasone, prednisolone, and streptomycin was possibly due to the low volatility or high polarity of these compounds. However, this method still has the potential to give a good indication of what a product does or does not contain with 70% of the target pharmaceutical standards detected with less than a 1-min analysis. With the use of negative mode, the coverage of this method could be improved further. More crucially, the common adulterants sibutramine, sildenafil, and tadalafil can be detected. In the first 6 months of 2018, these three drugs have been responsible for 14 regulator health alerts in Australia after being detected in counterfeit products [18].

The absence of chromatographic separation reduces the level of assurance and confidence in the identification of compounds. Analysis of the multivitamin SRM detected only 25% of the contents within an acceptable ppm range (Figure 2), highlighting the challenges posed by ion suppression or saturation, due to the wide dynamic range of the complex sample matrices and analyte concentrations. For example, the apparent detection of paracetamol in the multivitamin SRM was based on a false-positive match to the [M + H − H2O]+ fragment of pyridoxine (Figure 2). Isobaric compounds can also not be distinguished using this method. Accurate mass was the primary basis of identification for this study. As discussed by De Vijlder, et al. [19], the accuracy of an isotopic pattern is dependent on instrument type and resolving power, the mass of the ion of interest and potential inference from other chemicals and noise in the spectra. Subsequently, isotopic patterns have not been used as a primary basis for identification; instead, the calculated pattern of the suspected compound was used to help rule out possible matches or to substantiate the potential identification.

Figure 2
figure 2

DSA spectra of the NIST multivitamin SRM. The false-positive match for paracetamol, [M + H]+ 152.0706 m/z, is based on the [M + H − H2O]+ fragment of pyridoxine (also labeled as norepinephrine). The [M + H + methanol]+ adduct of pyridoxine and [M + H]+ of pantothenic acid are also highlighted

Full size image

Overall, analysis of the CAM samples confirmed the samples’ listed ingredients. For example, B vitamins listed as ingredients in sample 100 were detected by this analysis. The B vitamins and the alkaloid trigonelline found in sample 73 were also consistent with both listed ingredients and the DNA data. The DSA analysis correlated well with the previous toxicological LC-MS screen; for example, caffeine in sample 73 was detected in both the previous and current analyses.

It is important to highlight that this method does not intend to confirm the presence/absence or quantity of compounds within these products, but rather to serve as a rapid screening tool for pre- and post-market monitoring of herbal products, to give a snapshot of the composition. For example, the flavonoid wogonin is known to occur specifically in Scutellaria baicalensis [20] and its detection is in agreement with the DNA analysis of sample 175, which showed the product contained Scutellaria sp., as well as undeclared grass species and human and dog contamination. We continue to advocate that a combination of mass spectrometry compound detection coupled with DNA analysis provides key insights into ingredients, substitutions, and adulterations.

The analogous results between the Ginkgo biloba samples and the NIST Ginkgo powder further demonstrate the potential for the DSA-TOF technique. Both the samples and the NIST standard contained isorhamnetin, kaempferol, and quercetin, as well as other catecholamine and alkaloid-like compounds. While no specific Ginkgo phytochemicals were detected, the detection of other phytochemicals does suggest that there was plant material in the products. As no DNA material was detected and the previous analysis did not screen for such phytochemicals, the DSA results are the first to support the products’ claim of plant-based contents.

The correlation between the previous LC-MS and DNA analyses and the current DSA-TOF analysis, as well as the detection of non-specific plant alkaloids and flavonoids, gives strength to this rapid analysis for its ability to profile relevant phytochemicals and possible undeclared ingredients for quality control purposes. While there are limitations to this approach, DSA analysis allows screening of hundreds of compounds with minimal sample preparation, generating an outline profile of a product’s composition. For products where DNA cannot be extracted, or where interest in the composition is beyond a standard toxicological screen, such analysis can aid in a zero-tolerance approach to contaminated and adulterated products. Given that many regulatory agencies face the challenge of limited resources, with thousands of samples to be screened, but only a proportion needing comprehensive confirmation and quantification of possible adulterants and contaminants, DSA analysis has the potential to substantially improve efficiency.