A trader's sack labelled with the name of a medicinal plant may hold the species named, a close relative, an unrelated plant that shares a vernacular name, or a blend of these. The causes are well known to pharmacognosists: collectors identify by local names that cover several species, scarce species are replaced by common ones, and material passes through many hands before it reaches a manufacturer. Conventional quality control relies on macroscopic and microscopic examination and on chromatographic fingerprints (WHO, 2011). These methods are indispensable, yet they struggle with fragmented material from closely related species whose anatomy and chemistry overlap.

DNA barcoding was introduced as a way of identifying organisms from a short, standardised gene region compared against a library of sequences from identified specimens (Hebert et al., 2003). Its application to herbal materials has grown quickly, and reviews agree both that it works for raw botanicals and that it has been reported with very uneven rigour (de Boer et al., 2015; Parveen et al., 2016; Mishra et al., 2016). The dispute that followed the market survey of Newmaster et al. (2013) showed how much depends on method: critics noted that processed products may contain little intact DNA, so that a failure to detect a species need not mean that it is absent.

This article sets out a protocol for the simpler and more tractable case of crude raw material, meaning dried plant parts that have not been extracted. It is a proposal, built from published methods, and it has not been applied to samples in the course of this work. The aim is a procedure definite enough to be tested, criticised and improved.

No single region of the plant genome does for plants what the mitochondrial gene COI does for animals. After comparing candidates, the CBOL Plant Working Group recommended two plastid coding regions, rbcL and matK, as the core barcode for land plants (CBOL Plant Working Group, 2009). The pair is robust and well represented in databases, but it often fails to separate closely related species. Kress et al. (2005) had earlier proposed the plastid spacer trnH-psbA together with the nuclear ribosomal internal transcribed spacer (ITS). A large comparative study later concluded that ITS should be added to the core barcode for seed plants (China Plant BOL Group, 2011).

For medicinal plants the most influential proposal has been that of Chen et al. (2010), who tested several regions on a large set of medicinal species and recommended ITS2, the second spacer of the ITS region, as a standard. ITS2 is short, which matters greatly for dried and stored material in which DNA is broken into small fragments. Hollingsworth et al. (2011) review the trade-offs between markers in practical terms. Table 1 condenses the properties that bear on the present protocol.

Two consequences follow. A protocol for traded material should lead with a short nuclear region and support it with a plastid region of independent inheritance, so that hybridisation or contamination affecting one locus is exposed by the other. And the choice of marker cannot be separated from the reference library, since a region is useful only if verified sequences for the target and its substitutes exist.

The protocol is arranged in eight steps, shown in order in Figure 1. Steps 1 and 2 are done once for each commodity and then maintained. Steps 3 to 8 are repeated for each consignment.

The protocol is meant to sit beside pharmacognostic and chemical tests. Morphology and microscopy confirm the plant part and detect foreign matter; chromatography addresses constituents; the barcode addresses species identity. An inconclusive barcode result is referred back to these methods, and where they disagree with the sequence the conflict is investigated before any conclusion is issued. A laboratory without its own sequencer can perform Steps 1 to 5 and send products to a sequencing service, which keeps the equipment needed to that of an ordinary molecular biology bench.

The heaviest cost is the reference library. It is incurred once per commodity and grows in value as more consignments are tested against it. Laboratories serving the same trade would gain from pooling vouchered sequences, and public deposition makes that possible without formal agreements.

The protocol has not been validated here, and several of its claims are expectations drawn from the literature. Validation would need to establish, for each commodity, that the chosen loci separate the target from every species on the substitute list, using reference specimens that were not used to set the threshold; that the procedure gives the same answer when repeated by different analysts on different days; and how often dried material of realistic age yields a usable sequence. Blind testing with deliberately substituted samples would give the most direct measure of error rates.

Some limits are inherent. Recently diverged species and hybrids may share sequences at any practical number of loci. Heavily processed or very old material may yield no DNA. Taxonomic disagreement about species boundaries carries through to the reference library. In these situations the protocol is designed to return an ambiguous or null result, which is preferable to a confident error.

Authentication of crude medicinal plant material by DNA barcoding is within reach of an ordinary quality-control laboratory if the work is organised around a defined list of substitutes, a vouchered reference library, two independent loci, routine controls and a decision rule set down beforehand. The eight steps proposed here supply that organisation. Their performance on real consignments is an empirical matter that remains to be established, commodity by commodity.

To be declared by the author(s) before publication.

To be declared by the author(s) before publication.