DNA Barcode Authentication of Devil’s Claw Herbal Dietary Supplements

Devil’s claw is the vernacular name for a genus of medicinal plants that occur in the Kalahari Desert and Namibia Steppes. The genus comprises two distinct species: Harpagophytum procumbens and H. zeyheri. Although the European pharmacopeia considers the species interchangeable, recent studies have demonstrated that H. procumbens and H. zeyheri are chemically distinct and should not be treated as the same species. Further, the sale of H. zeyheri as an herbal supplement is not legal in the United States. Four markers were tested for their ability to distinguish H. procumbens from H. zeyheri: rbcL, matK, nrITS2, and psbA-trnH. Of these, only psbA-trnH was successful. A novel DNA mini-barcode assay that produces a 178-base amplicon in Harpagophytum (specificity = 1.00 [95% confidence interval = 0.80–1.00]; sensitivity = 1.00 [95% confidence interval = 0.75–1.00]) was used to estimate mislabeling frequency in a sample of 23 devil’s claw supplements purchased in the United States. PCR amplification failed in 13% of cases. Among the 20 fully-analyzable supplements: H. procumbens was not detected in 75%; 25% contained both H. procumbens and H. zeyheri; none contained only H. procumbens. We recommend this novel mini-barcode region as a standard method of quality control in the manufacture of devil’s claw supplements.


Introduction
Harpagophytum (Pedaliaceae) is a genus of tuberous plants from the Kalahari Desert and Namibia Steppes that is commonly known as devil's claw due to its hooked fruits [1]. The genus comprises two distinct species-H. procumbens and H. zeyheri-that have been separated on the basis of morphology [2][3][4] and chemistry [5]. Harpagophytum procumbens consists of two subspecies [3], H. procumbens subsp. procumbens, which occurs across Namibia, Botswana, and Northern South Africa, and H. procumbens subsp. transvaalense, which occurs only in the Limpopo region of South Africa. Harpagophytum zeyheri comprises three subspecies [3], H. zeyheri subsp. zeyheri, which is restricted in distribution to northeastern South Africa, and H. zeyheri subsp. schijffii and H. zeyheri subsp. sublobatum, which are both widely distributed across regions of Angola, Zambia, and Zimbabwe and the northern regions of Namibia and Botswana.
There are unsubstantiated reports of possible hybridization in the few places where H. procumbens and H. zeyheri are sympatric [4,6,7]. Although purporting to demonstrate hybridization, RAPD and ISSR data [7] are, at best, inconclusive: no species-specific genotype groups were detected [7], thus a definitive pattern of hybridization cannot possibly be observed; the published Principal Component Analysis [7]-which is inappropriate for detecting hybridization [8,9]-identifies five putative hybrids, but only one individual is truly intermediate while several non-hybrid samples are equally or more intermediate than the putative hybrids; and the published UPGMA dendrogram [7] refutes the hypothesis of hybridization because it nests the putative hybrids well within the two parental clusters rather than at the cluster base where hybrids are expected to appear [10].
In addition, morphological data [4] purportedly demonstrate hybridization, but they are not statistically significant: the published Discriminant Function Analysis (DFA) [4] improperly implemented DFA such that hybrids were assumed to be present rather than using DFA to test that supposition. In addition, measurements that violate the Gaussian distribution assumed by DFA [11] were included. If DFA is conducted on the five characteristics that do not deviate [12] significantly (p > 0.01) from the Gaussian distribution (arm width, seed column height, fruit length, fruit width, and fruit circumference), the putative hybrids [4] are classified without evidence of intermediacy (pp ≥ 0.99999). Independent of the improperly implemented DFA, no statistical test was conducted to determine if the putative hybrids were truly intermediate [4]: the character count procedure [9] employing the sign [13] and Scheffé [14] tests (p = 0.05) does not indicate intermediacy for any characters and thus no trace of hybridity was detected (p = 1.0).
Given this critical review, there are no published data showing evidence of hybridization between Harpagophytum species and further study of additional specimens and characteristics is needed to determine if hybridization does indeed occur.
Devil's claw has traditionally been used to treat dyspepsia, fever, constipation, hypertension, and venereal disease [1]. Commercial preparations of H. procumbens are sold to treat arthritis in both the European and United States markets [15]. Harpagophytum zeyheri cannot be legally sold as an herbal supplement in the United States [16] but it was appended to the European Pharmacopeia [17]. Both species are wild sourced-primarily from Namibia [18].
Commercial herbal supplements are most frequently sold as dry fragments or powders. As a result, the authentication of these materials has traditionally relied upon macro-and microscopic morphological examination along with chemical assays for specific compounds or classes of compounds [34]. In the last two decades, DNA-based assays have become more common with assays for specific plants (e.g., molecular marker-based methods that utilize simple sequence repeats (SSR) or single nucleotide polymorphisms (SNP)) and general untargeted analysis techniques (e.g., short fragment sequencing methods such as whole metagenome analysis and metabarcoding) now being prominently used [35][36][37]. DNA barcoding has emerged as a preferred method of herbal supplement authentication due to the fact that it generally works well with highly fragmented DNA from high-copy regions (e.g., plastid), can detect multiple species at once, and is relatively inexpensive. These characteristics make the method ideal for assaying the DNA in highly degraded herbal products.
Devil's claw supplements are sold mainly in capsule or tablet form [38]. Thus, it is impossible to determine which species they contain without additional analysis. A reliable identification method to ensure correct labeling is needed. We aim to create and test a DNA mini-barcode assay for both Harpagophytum species.

An Analysis of Herbal Supplements
Amplifiable DNA was extracted from 20 of 23 (87%) herbal supplements. Amplification success was significantly correlated with the reports of root extract on product labels (McNemar test [45]; p = 0.04331; Table 2). The failure rate for samples labeled as having root extract (17%) was nearly double that of samples without root extract (9%; Table 2). Table 2. Herbal dietary supplement label ingredients and psbA-trnH mini-barcode determination. Supplement sequence type corresponds to those in Figure 1. If Latin names were not provided on the product label, the Latin name was determined using [16]. Despite being noted on some labels, the sale of supplements containing H. zeyheri is not legal in the United States. PCR products were successfully sequenced for all 20 amplifiable supplements: minibarcode median sequence quality was 0.561 (IQR 0.451-0.587)-very similar to the quality of the validation samples.
Types A, B, and C contain H. zeyheri haplotypes that exhibit the same variation found in the reference samples. Type A is composed of samples that contain only one H. zeyheri haplotype, while types B and C are mixtures of H. zeyheri haplotypes (e.g., Figure 2). In contrast, types D, E, and F are mixtures of H. procumbens and H. zeyheri haplotypes. Type E contains one H. procumbens haplotype and two H. zeyheri haplotypes (a "D" ["A", "G" and "T"] at alignment position 77; Figure 1).

Discussion
The psbA-trnH mini-barcode absolutely differentiates Harpagophytum from all other Pedaliaceae (Figure 1: blue highlighted positions 16, 64, and 116) and in turn H. procumbens and H. zeyheri from one another (Figure 1: orange highlighted positions 76 and 107). Thus, the species have consistent character state differences and can be considered distinct phylogenetic species [46]. The absolute consistency of psbA-trnH mini-barcode alignment positions 16, 64, 76, 107, and 116 prevent infraspecific variation from having any bearing on repeatable Harpagophytum species identification (specificity = 1.00 [95% confidence interval = 0.80-1.00]; sensitivity = 1.00 [95% confidence interval = 0.75-1.00]). Although there are reports of possible interbreeding between the two Harpagophytum species [4,6,7], the pattern observed here is inconsistent with hybridization because the morphological and molecular species identifications exactly match. No intermediate morphological phenotypes have been confirmed either in the literature or in our research, suggesting that hybrids, if they exist, have retained strong morphological similarity to one of the parental species. Therefore, absolute rejection of the hybridization hypothesis would require the investigation of multiple biparentally inherited molecular markers. Given the lack of support for the supposition of hybridization in the data, the regulatory distinction between H. procumbens and H. zeyheri in the United States [16] can be enforced.
The variation within the psbA-trnH mini-barcode used to differentiate between the two Harpagophytum species could be assayed using molecular techniques other than the Sanger sequencing method demonstrated here. For instance, one could use PCR-RFLP with AseI (5 -ATTAAT-3 ) to assay alignment position 107 (H. procumbens will cut, but H. zeyheri will not); RT-PCR with specific primers and/or probes targeted to alignment positions 16, 64, 76, 107, and/or 116; or short read genome skimming (e.g., Illumina) with appropriate bioinformatic postprocessing to find alignment positions 16,64,76,107, and 116 in the output sequences. Depending upon the needs of the user, each of these techniques could be conducted in such a way as to quantify the relative or absolute amounts of DNA from each species present in the sample.
The observed mini-barcode PCR amplification failure rate from herbal supplements of 13% is a bit high compared to the 3-10% reported for similar studies [47][48][49]. Although the processing of plant materials for herbal supplement manufacturing frequently results in DNA fragmentation and destruction  that can prevent amplification, the processing techniques used for devil's claw may be more damaging than those used for other herbal supplements studied thus far-which is supported by the significant correlation between reports of root extract (a relatively damaging technique [70]) on product labels and PCR failure (McNemar test [45]; p = 0.04331; Table 2). It is also possible that some, or all, of the high rate of PCR failure can be attributed to the amount of recoverable DNA in devil's claw tap roots being low and/or less enzymatically accessible in comparison to aerial parts as is the case in carrot (Daucus carota) tap roots [71,72].
Labels of only two of the 20 analyzable supplements ( Table 2) list Harpagophytum zeyheri, but H. zeyheri was found in all 20 fully-analyzable samples. Somehow the two, predominantly allopatric [1], species were mixed. Although H. zeyheri can be legally sold in the European Union [17], it cannot be sold in the United States [16].
Bulk materials of devil's claw are usually sold in a morphologically unidentifiable state [1,5]. Thus, a chemical test that measures the relative quantity of harpagoside and 8-p-coumaroyl-harpagide is often used to distinguish between bulk materials from the two species [73]. The data that purport to validate the assay were not analyzed statistically [73]. Unfortunately, the data do not statistically differentiate between the Harpagophytum species (Mann-Whitney test [74]; p = 0.1386)-perhaps due to the miniscule sample size (n = 5). Therefore, this chemical assay cannot be considered reliable. Revalidation with additional, morphologically identifiable and vouchered samples may redeem this assay for harpagoside and 8-p-coumaroyl-harpagide.
Due to the legal status of H. zeyheri in the United States, it is imperative that supplement manufacturers employ a robust method of quality control to evaluate all devil's claw supplements sold. Because the mini-barcode presented here is reliable, cost-efficient, and simple to use, we recommend it as a standard method of quality control instead of the relative quantity of harpagoside and 8-p-coumaroyl-harpagide.
Validation samples were chosen arbitrarily (n = 30; Table 1). Herbal supplements (capsules and compression tablets) were purchased online.
A psbA-trnH mini-barcode was designed from all Pedaliaceae reference sequences. The mini-barcode is anchored within the intergenic spacer (alignment positions 1-122) and extends into trnH (alignment positions 123-147; Figure A1). This region was selected for its compactness and discriminatory power.