A supercritical fluid workflow for the quality assessment of herbal drugs and commercial preparations from Rhodiola rosea

Abstract Introduction Preparations from the Rhodiola rosea are experiencing an increase in popularity: extracts of dried roots and rhizomes are used as adaptogen to treat stress, fatigue, and weakness. To meet high pharmaceutical standards, fast and reliable methods to assess phytochemical variations in respect of quality control are needed. Objective The aim of this study was to extract and quantify seven characteristic secondary metabolites of R. rosea , namely p‐tyrosol (1), rosin (2), rosiridin (3), salidroside (4), rosarin (5), rosavin (6), and tricin‐5‐O‐β‐d‐glucopyranoside (7) in 24 herbal drugs and seven commercial preparations using a newly established supercritical fluid workflow. Methods The developed protocol allowed for an exhaustive extraction of compounds 1–7 using 60% carbon dioxide (CO2) and 40% methanol. The constituents were analysed on an ultra‐high‐performance supercritical fluid chromatography (UHPSFC) instrument using a charged surface hybrid fluoro‐phenyl (CSH FP) column (3.0 mm × 100 mm, 1.7 μm; mobile phase: CO2 and methanol). Results The seven compounds were separated in a remarkably short time (< 3.5 minutes). For their quantitation, good results in terms of selectivity, linearity (R 2 ≥ 0.99), precision (intraday ≤ 3.03%, interday ≤ 5.17%) and accuracy (recovery rates 96.6–102.4%) were achieved using selected ion recording on a Quadrupole Dalton (QDa) mass detector. Conclusion The quantitative analysis of the investigated herbal drugs showed a highly differing metabolite pattern which was also observed in the investigated commercial products. None of the commercial dietary products met the declared content of rosavins and salidroside. The developed and validated protocol offers a novel and reliable method to assess the quantitative composition of Rhodiola herbal drugs and preparations.


| INTRODUCTION
Rhodiola rosea L. (rose root, Arctic root or golden root), a species mainly growing in Arctic regions of Europe and Asia, has been a valuable medicinal plant used as adaptogen for centuries. 1 The scientific evidence for the health beneficial properties such as anti-depressive, anti-fatigue, anxiolytic, cardioprotective, central nervous system (CNS) stimulating, neuroprotective, and nootropic effects is increasing continuously. 1,2 For commercial products including herbal medicinal products and dietary supplements in Europe, Asia, and the United States, 3  About 25% of these products were either adulterated or did not conform to their label specification regarding rosavin levels. 6 The United States Pharmacopoeia (USP) contains a monograph for R. rosea as well as for the. Asian species R. crenulata (Hook.f. & Thomson) H.Ohba, [7][8][9] whereas the Chinese Pharmacopoeia only records R. crenulata 10 and the Russian Pharmacopoeia only R. rosea. 11,12 The elaboration of a monograph for the European Pharmacopoeia is still pending. 13 Roots and rhizomes of rose root mainly contain phenylethanoids, phenylpropanoids, monoterpene alcohols, flavonoids, and their respective glycosides as well as proanthocyanidins and gallic acid derivatives. 1 In dried roots and rhizomes, the USP monograph requires not less than 0.08% salidroside and not less than 0.3% phenylpropanoids (comprising rosavin, rosarin, and rosin), 7 whereas the Russian Pharmacopoeia demands a more than three-times higher content of phenylpropanoids (> 1.0%) and a 10-times higher content of phenylethanoids calculated as salidroside (> 0.8%) ( Table 1). 12 Regarding quality and sustainability, Peschel et al. recently analysed the influence of geographic provenance, harvest season, plant sex, plant part (root or rhizome), and processing on marker compounds of Rhodiola species. 14 Especially, marker ratios, e.g. salidroside vs. total rosavin content or rosarin vs. rosavin vs. rosin, turned out to be useful for quality control. Whether the plant is male or female showed no influence on the phenylpropanoid content 15 and also drying temperature and cutting conditions are less important. 14 However, origin, harvest season, plant part (rhizomes contain more rosavins than roots), and processing have a major influence on the quantity of Rhodiola constituents. 15 Commonly applied extraction procedures for rose root samples comprise sonication and maceration using solvents such as methanol or ethanol or hydroethanolic/methanolic mixtures ranging from 38% to 75% alcohol. [16][17][18][19][20][21][22] Moreover, accelerated solvent extraction with 85% methanol and 15% water was reported. 19 Classic chromatographic approaches to study phytochemical variations in different Rhodiola samples most commonly involve a high-performance liquid chromatography (HPLC) instrument hyphenated to an ultraviolet-visible (UV-vis) detector. [16][17][18]20,23 However, up until now, published protocols are suffering from drawbacks, e.g. long analysis times of more than 30 minutes 16,19,[21][22][23] or they are limited to the analysis of only one or few compound classes. 18,20,24 Supercritical fluid-based (SFx) technologies have many advantages for the extraction and separation of plant constituents, e.g. little solvent consumption and remarkable short analysis times for high efficiency separations due to the high diffusivity and low viscosity of the mobile phase. Due to the non-polar character of supercritical carbon dioxide (CO 2 ), the primary focus of separation lies on non-polar analytes like carotenoids, fatty acids or terpenes. 25 However, the adjustment of the mobile phase polarity with organic modifiers and the availability of new stationary phase materials with sub-2 μm particles extend the spectrum of this technology. Thus, also polar compounds like glycosides can be extracted and separated successfully. 26 For instance, Gibitz-Eisath et al.
recently achieved the separation of seven glycosides to establish a quantitation method for common vervain. 27 The aim of this study was to overcome disadvantages of current standard protocols for the extraction and quantitation of characteristic R. rosea constituents. This goal was implemented with a fast, ecofriendly and efficient workflow taking advantage of CO 2 based technologies. Whilst therapeutic effects of Rhodiola constituents are still the subject of intense research, phenylpropenoid glycosides, i.e. rosin (2), rosarin (5), and rosavin (6), the monoterpene glucoside rosiridin (3) as well as the phenylethanoid p-tyrosol (1) and its glucoside salidroside (4) are well established marker compounds ( Figure 1). Although tricin-5-O-β-D-glucopyranoside (7) has not been considered in previous standardisation studies of Rhodiola, it was included in the present study for two reasons: (i) for its importance as anti-influenza A virus active, 28 α-amylase 29 and NO production inhibiting 30 Table S2).
Before extraction, crude root samples were ground to powder using a household grinder. All powder samples were kept in paper bags at room temperature until use. All HPLC grade solvents were purchased from VWR Chemicals. Compressed 4.5 grade CO 2 (purity ≥ 99.995%) was purchased from Messer.

| Supercritical fluid extraction
Extractions were performed using a Waters MV-10 supercritical fluid extraction (SFE) instrumentation consisting of a fluid delivery module (cooled down by a Thermo Scientific Accel 500 LC chiller), a 10 vessel column oven, an automated backpressure regulator, a heat exchanger and an extraction collector connected to a make-up pump. The instrument is controlled via ChromScope 1.6 software. The extraction F I G U R E 1 Chemical structures of compounds 1-7 vessels hold a volume of 5 mL each. Regarding the commercial rose root products, the content of one capsule or one mortared tablet (except for RR06: two capsules) was placed into the extraction vessel and filled up with glass beads. Concerning the herbal drugs, 1.00 g of ground and dried sample was placed in the extraction vessel and filled up with glass beads. The optimised extraction method is presented in Table 2. The obtained extracts were transferred into a volumetric flask and filled up with methanol to 250.0 mL. The established protocol provides an extraction efficiency of more than 96.0% for all reference compounds. Samples were stored at 8 C until analysis.

| Analytical UHPSFC
An analytical method, the ultra-high-performance supercritical fluid chromatography (UHPSFC) instrument Acquity UPC 2 (ultraperformance convergence chromatography) comprising a sample-, binary solvent-, column-, isocratic solvent-and convergence-manager with a photodiode array (PDA) detector and a Quadrupole Dalton (QDa) detector was used. Nitrogen served as nebulising gas for QDa operation. The instrument was controlled via Empower 3 software.
The parameters resulting in the best separation are given in Table 2.
For method validation and quantitation experiments, data were collected by selected ion recording (SIR) in accordance with the specific masses of target compounds.

| Method validation
The optimised methods for the extraction and analysis were validated level was injected in triplicate and the peaks were integrated using Empower 3 software. The limit of detection (LOD) and the limit of quantitation (LOQ) were determined visually by the concentration showing a signal-to-noise ratio of at least 3 and 10 times, respectively.
Precision was determined by intraday (evaluation within 1 day) and interday (evaluation over 3 days) experiments with sample RR29. For accuracy, recovery rates were measured by spiking sample RR05 with high (125%), medium (100%) and low (75%) amounts of the respective standard compound. Spiked samples were then extracted using the MV-10 device and analysed on the UPC 2 as described earlier. The validation parameters are presented in Table 3.
In six herbal drug samples, namely RR19, RR27 and RR28-RR31, compound 3 was outside the linearity range due to its high content.
Therefore, the respective samples were diluted with methanol in a ratio of 1:1 32 in order to be inside the linear range of the validated method (Table S1).
Assessment of global uncertainty was carried out on the basis of Konieczka and Namiésnik 33 and Ratola et al. 34 Combined uncertainty (U) was calculated from following the expression for each compound, respectively: U = √ (U1 2 +U2 2 +U3 2 +U4 2 +U5 2 ) where U1 is uncertainty associated with sample preparation, U2 is uncertainty associated with calibration, U3 is uncertainty associated with precision, U4 is uncertainty associated with accuracy (not included for compounds 3 and 7) and U5 is uncertainty associated with analyte concentration. Expanded uncertainty (Uexp) is expressed as twice U (k = 2) ( Table 3).

| Method development for extraction
With the aim to establish a SFx protocol, supercritical CO 2 was used for both the extraction and analysis of compounds 1 to 7. Since methanol is known as optimum solvent to extract secondary metabolites of rose root, it was used as modifier to adjust the polarity of supercritical CO 2 . Initially, a stepwise extraction using 100%, 90%, 80%, 70%, 60% and 50% CO 2 was selected. Extraction with a mixture of 60% CO 2 and 40% methanol as modifier revealed to be the best suitable combination to cover a broad polarity spectrum of constitu-   Figure S1).

| UHPSFC method development
The test of four different co-solvents (methanol, ethanol, isopropyl alcohol and acetonitrile) revealed that methanol without any additives was the best choice regarding peak shape, retention time, and resolution ( Figure S2). Neither the addition of acid (0.1% formic acid) nor a mixture of methanol and acetonitrile (50:50) as co-solvent improved the result (data not shown).
A parameter unique to SFx technologies is the backpressure controlled by the ABPR. In comparison to the standard setting of 2000 psi, an improved peak shape was observed for the ABPR set to 2100 psi. A further increase of backpressure (up to 2500 psi) did not improve resolution or peak shape but only led to a retention time shift. Furthermore, the best separation was achieved with a flowrate of 1.0 mL/min and a column temperature of 40 C.
The gradient of the binary mobile phase comprising CO 2 and methanol as co-solvent was tested with and without isocratic intermediate steps. With the finally optimised method conditions, a separation of rose root compounds in less than 3.5 minutes was achieved (Table 2).
To ensure optimal ionisation for mass analysis using a QDa detec- probe temperature (500 C), sampling rate (five) and gain (three) gave the best results with default settings ( Table 2).
The parameters for the final methods for extraction and separation are given in Table 2. The chromatogram of the extract of sample RR29 and extracted SIR channels of analytes 1-7 are given in

| Method validation
To ensure the suitability of the generated protocols for the quantitation of compounds 1-7 in R. rosea samples using mass detection, a validation according to ICH guidelines was performed. 31 Results shown in Table 3  Precision was determined as standard deviation based on peak area within 1 day (intraday) and 3 days (interday) using sample RR29.
Intraday and interday variations are acceptable and typical for plant material showing some inhomogeneity, however suggesting a F I G U R E 2 UHPSFC analysis of the herbal drug sample RR29 (PDA 220 nm) and extracted SIR chromatograms of compounds 1-7 with optimized parameters on a CSH FP column (1.7 μm, 100 mm × 3 mm) good precision in comparison to other published methods for phenylethanoids and phenylpropanoids. 16,17 For the determination of accuracy, sample RR05 was spiked with 125%, 100% and 75% of two representative standard compounds, namely p-tyrosol (1) and rosin (2). The spiked samples were extracted and analysed as described in the established SFx protocol. Good recovery percentages ranging from 96.6% to 102.4% were found, which are in agreement with published chromatographic methods. 18 To ensure measurement reliability, combined and expanded uncertainties (U and Uexp) were calculated. In fact, U resulted in a value below 5% for compounds 1 and 3-7. The highest U value was However, these directions give room for a broad range of metabolites present in the extract, as can be seen from our quantitative analysis of sample RR01 and RR02 (Figure 3). Although both products fulfil the HMPC requirement, p-tyrosol (1), rosin (2) and salidroside (4) could not be detected in RR01, whereas RR02 contains 6.95% of the quantified compounds in total. RR03 only contains 0.15% p-tyrosol (1) and none of the other six standard compounds was detected, while the other four samples (RR04-RR07) are similar to the herbal medicinal product RR02 or even higher in total secondary metabolite content.
Some manufacturers declare a minimum content of salidroside (4) and/or total rosavins (the sum of 2, 5 and 6): RR04 claims to contain 5.27% rosavins and 2.36% salidroside; RR05 should contain 3% rosavins and RR06 declares 3% rosavins and 1% salidroside. The USP monograph requires a range between 90% and 110% of the declared content of rosavins and salidroside. 35 Only one sample -RR06with a quantified amount of 2.91% of rosavins is in accordance with these USP requirements. However, the salidroside content in RR06 exceeds its declaration of 1% salidroside (1.66% quantified) and is therefore outside the limit given by the USP monograph. In conclusion, none of the analysed commercial products complies with its declared content of constituents.
The results of the quantitation of the investigated herbal drugs are given in Figure 3, ordered by origin from West to East. In general, there is a high variability of the quantified amount of the individual compounds 1-7 as well as in their total content. The high total metab- Tricin-5-O-β-D-glucopyranoside (7) was detected only in three samples derived from Poland (RR19) and Russia (RR27 and RR29).
Although the first applications of SFx technologies were already reported more than 50 years ago, this is the first time, an SFx protocol was established for rose root constituents. A reproducible, precise, and accurate protocol for the fast extraction and quantitation of seven secondary metabolites was generated. This proves once more the high potential and applicability of environmentally friendly and robust CO 2 based instruments not only for non-polar but also for polar constituents. Using supercritical fluids for both, chromatography and targeted extraction (SFE), allows for a very focused and straightforward procedure without the need for any specific sample clean-up prior to analysis. In addition, the selectivity, reliability, and separation speed of the UHPSFC technique even exceeds classical methods like HPLC, UPLC, and gas chromatography (GC). 16,18,19,21 As presented here, the hyphenation of a UHPSFC device to a mass detector (QDa) provides additional benefits such lower detection limits than conventional set-ups.
Considering the rising popularity of Rhodiola commodities, fast and reliable methods to analyse the content of key constituents are of great importance to guarantee the supply of high-quality products.
Therefore, and to test the versatility of SFx techniques, we aimed not Compared to the USP, the Russian Pharmacopoeia requires higher minimum contents for roots and rhizomes of R. rosea, i.e. 1.0% rosavins and 0.8% salidroside (Table 1). Taking this into account when comparing the herbal drug samples RR08-RR31, only RR09 and RR10 originating from Switzerland as well as RR26 from Finland meet these requirements. Surprisingly, three out of the five food supplement samples RR03-RR07, i.e. RR05, RR06, and RR07, also meet these comparably high contents required by the Russian Pharmacopoeia for rosavins and salidroside. 12 When considering the USP monograph for R. rosea, all herbal samples except for RR15 and RR17 fulfil the requirements (> 0.3% rosavins and > 0.08% salidroside). 9 The same accounts for commercial products except for the herbal medicinal product RR01, where no salidroside was detected and the food supplement RR03, where neither salidroside nor rosavins were found.
Comparing the quantities of R. rosea constituents determined by the SFx workflow established in the present study with results from classic quantitation methods reported in the scientific literature, similar concentration ranges were found (Table S1). The high variations of constituent concentrations in samples from different origins have also been observed in previous quantitative analyses. 16,18,19 In sum, we have demonstrated that SFx technologies are suitable for the extraction and quantitation of Rhodiola constituents in various samples. Whether the here developed SFx protocol is also suitable for the authentication of Rhodiola raw material and if it can be applied to distinguish between different Rhodiola species and potential adulterations needs to be investigated in future studies.