Quantitative Analysis of Salidroside and p-Tyrosol in the Traditional Tibetan Medicine Rhodiola crenulata by Fourier Transform Near-Infrared Spectroscopy

Abstract

A nondestructive, efficient, and rapid method for quantitative analysis of two bioactive components (salidroside and p-tyrosol) in Rhodiola crenulata, a traditional Tibetan medicine, by Fourier transform near-infrared (FT-NIR) spectroscopy was developed. Near-infrared diffuse reflectance spectra in the range of 4000 to 10000 cm⁻¹ of 50 samples of Rhodiola crenulata with different sources were measured. To get a satisfying result, partial least squares regression (PLSR) was used to establish NIR models for salidroside and p-tyrosol content determination. Different preprocessing methods, including smoothing, taking a second derivative, standard normal variate (SNV) transformation, and multiplicative scatter correction (MSC), were investigated to improve the model accuracy of PLSR. The performance of the two final models (salidroside model and p-tyrosol model) was evaluated by factors such as the values of correlation coefficient (R²), root mean square error of prediction (RMSEP), and root mean square error of calibration (RMSEC). The optimal results of the PLSR model of salidroside showed that R², RMSEP and RMSEC were 0.99572, 0.0294 and 0.0309, respectively. Meanwhile, in the optimization model of p-tyrosol, the R², RMSEP and RMSEC were 0.99714, 0.0154 and 0.0168, respectively. These results demonstrate that FT-NIR spectroscopy not only provides a precise, rapid method for quantitative analysis of major effective constituents in Rhodiola crenulata, but can also be applied to the quality control of Rhodiola crenulata.

Rhodiola crenulata is a kind of alpine perennial herb which belongs to the Rhodiola L. genus, Crassulaceae family. Various Rhodiola species have been extensively utilized in traditional Tibetan medicine in order to keep body healthy and to cure sickness for over a millennium.^1,2) Rhodiola is often used to enhance physical performance, vitalize energy, moderate psychological stress, and prevent diseases by local people.^3,4) The Rhodiola species was recorded in the Tibetan medicine literature “Somaratsa,” “Four-Volume Medical Code” and “Jing Zhu Materia Medica” at the earliest. As the only Rhodiola species which is included in Chinese Pharmacopoeia (2015), Rhodiola crenulata has very high medicinal value. It has been used for the treatment of different situations such as eliminating toxins from the body, clearing heat in the lungs, treating various epidemic diseases, edema of limbs, traumatic injuries and burns.⁵⁾ Nowadays, some researches also indicate that Rhodiola crenulata is available to treat anti-hypoxia, antifatigue, principally, and it is able to improve work efficiency, defer senescence of body, prevent and treat diseases about ageing.^6,7) Therefore, its economic value and broad development prospect cheer people up. Salidroside and p-tyrosol, as the main bioactive components in Rhodiola crenulata, have the similar effects for antifatigue, anticancer, anti-inflammatory and antioxidation.^8–11) Moreover, salidroside is regarded as the index for estimation of the quality of Rhodiola crenulata. The chemical structures of those two compounds were shown in Fig. 1.

Fig. 1. The Chemical Structures of Salidroside (a) and p-Tyrosol (b)

Although many analytical methods such as HPLC,¹²⁾ TLC,¹³⁾ GC,¹⁴⁾ and UV spectrophotometry¹⁵⁾ have been reported for the quantitative analysis of salidroside and p-tyrosol, they are all costly, destructive, time-consuming, and need preliminary treatments of the samples. Instead, the Fourier transform near-infrared (FT-NIR) technology is an apposite alternative method. The merits of this method are mainly as follows: rapid speed, less time spent, no need for samples destruction, little or no sample preparation.^16–19) Compared to the traditional methods of analysis, NIR spectroscopy also has been proven to be a reliable tool in the study of quantitative analysis, quality control, and qualitative analysis.²⁰⁾

In this study, two bioactive components salidroside and p-tyrosol in traditional Tibetan medicine of Rhodiola crenulata were quantificationally determined by using the FT-NIR analytical method with partial least squares regression (PLSR), and HPLC method was used as a reference.

Experimental

Sample Preparation

In this study, 50 samples of the roots and rhizomes of Rhodiola crenulata that were collected during flowering and fruiting time from the western Sichuan province plateau areas and Tibet of China in 2013 to 2014 for the content determination. Samples were uniformly numbered and the detailed information about the source of Rhodiola crenulata was shown in Table 1. All of the 50 samples were pulverized into powder and heated for 12 h at 50°C. The 50 samples were identified by associate Professor Tao Li (West China School of Pharmacy, Sichuan University, Chengdu, Sichuan, P. R. China). Voucher specimens were deposited in the Herbarium of Pharmacognosy, West China School of Pharmacy, Sichuan University (WCU). Two reference compounds of salidroside and p-tyrosol were purchased from the National Institute for the Control of Pharmaceutical and Biological Products (Beijing, China).

Table 1. Source of the 50 Samples of Rhodiola crenulata

Sample No.	Locality	Date of collection
1–5	Wenchuan, Sichuan	August, 2014
6–10	Xiaojin, Sichuan	July, 2014
11–15	Baoxing, Sichuan	July, 2014
16–20	Jiulong, Sichuan	September, 2013
21–25	Songpan, Sichuan	August, 2013
26–30	Heishui, Sichuan	July, 2013
31–35	Linzhi, Tibet	September, 2013
36–40	Hailuogou, Sichuan	July, 2014
41–45	Danba, Sichuan	August, 2014
46–48	Kangding, Sichuan	July, 2014
49–50	Hongyuan, Sichuan	August, 2013

NIR Spectra Collection

The NIR spectra were recorded using a Thermo Antaris II FT-NIR spectrometer (Thermo Fisher Scientific Inc., U.S.A.) with an InGaAs detector and sample cup. The instrument control and data processing were operated by the software of TQ Analyst 8.0 (Thermo Fisher Scientific Inc.). Each spectrum was the average of 32 scans and the spectral range was 4000–10000 cm⁻¹. Each sample was measured for 3 times and a mean spectrum of the 3 spectra of each sample was used in the further analysis.

Determination of Salidroside and p-Tyrosol in Rhodiola crenulata by HPLC

The values which determined by HPLC were regarded as actual values to establish the NIR calibration models. As to samples, the concentration ranges of the salidroside and p-tyrosol were presented in Fig. 2. For each sample, 0.5 g of powder was precisely weighed and extracted by 10 mL of methanol in an Erlenmeyer flask with a stopper via ultrasonic wave for 30 min. The weight which included powder, 10 mL of methanol, and Erlenmeyer flask with a stopper was recorded accurately before and after extraction. Then, the lost weight was filled by methanol. After extracting, this solution was filtered through a 0.45 µm membrane filter, centrifuged for 10 min, and then injected the final solution into the HPLC system. Representative HPLC chromatograms of standard solution (a) and sample solution (b) were shown in Fig. 3.

Fig. 2. The Concentration Ranges of Salidroside and p-Tyrosol

Fig. 3. Representative HPLC Chromatograms of Standard Solution (a) and Sample Solution (b)

1. Salidroside; 2. p-Tyrosol.

The contents of salidroside and p-tyrosol in Rhodiola crenulata were measured using HPLC (Shimadzu LC-10AT, Japan). The chromatographic separation was performed on a Shim-pack VP-ODS analytical column (5 µm, 4.6 mm×150 mm) with a guard column (C₁₈, 5 µm, 4.6 mm×7.5 mm) used at 35°C. A UV detector at the wavelength of 278 nm was used in this study. The mobile phase of solvent system consisted of methanol–water (15 : 85, v/v). The sample injection volume was 10 µL and the flow rate was set at 1.0 mL/min.

NIR Data Processing

In all quantitative analysis types, the measurements of spectral information at the specified regions can be correlated to the relevant component amounts or concentrations in the sample. These quantitative analysis types may use different algorithms to build the NIR calibration models such as stepwise multiple linear regression (SMLR), principal component regression (PCR), classical least squares, PLS, and so on. PLS was the most continually utilized in modeling. Therefore, the two calibration models of NIR were constructed via using PLS with TQ Analyst 8.0 software.

Results and Discussion

Evaluation of the HPLC Method

Linearity and Limit of Detection (LOD)

Linearity was determined using five standard solutions of different concentrations. Calibration curves were established by the value of peak area (Y) and the concentration of reference solutions (X mg/mL). For the two constituents, a good linearity with R²>0.999 (R² of salidroside is 0.9993, R² of p-tyrosol is 0.9997) was obtained. And the regression equations of salidroside and p-tyrosol were y=5087840x+536.8, y=9520200x+1066.5, respectively. The LOD of salidroside and p-tyrosol in samples was measured based on visual evaluation with a signal-to-noise ratio of 3 : 1. The LOD were determined to be 1.152 and 0.766 mg/L. In addition, the quantitation limits were measured according to a signal-to-noise ratio of 10 : 1 for five replicated analyses of spiked matrix blank. The quantitation limits of salidroside and p-tyrosol were estimated to be 1.728 and 1.149 mg/L.

Precision

The intra- and inter-day assays which were used to evaluate the precision were examined by five consecutive injections of the standard solutions during a day and four consecutive days, respectively. The relative standard deviation (RSD) of salidroside and p-tyrosol were 0.7333 and 1.2377%, respectively for intra-day assays and 0.2902, 0.4207%, respectively, for inter-day assays. The results indicated clearly that the samples had good precision during this period.

Accuracy

In HPLC experiments, this factor is usually expressed by recovery test. The recovery rate was measured by spiked samples with different concentration levels of 80, 100, and 120% of salidroside and p-tyrosol in the samples, respectively. The recovery was 99.2694 and 99.3829% for salidroside and p-tyrosol, respectively, at the spiked level of 80%; 98.8977 and 99.1196% for salidroside and p-tyrosol, respectively, at the spiked level of 100%; 98.6636 and 98.5760% for salidroside and p-tyrosol, respectively, at the spiked level of 120%.

All the parameters indicated that as the HPLC reference method was reliable.

Spectroscopic Investigation

According to the raw NIR spectra (Fig. 4) of 50 Rhodiola crenulata samples at wavenumbers ranging from 4000 to 10000 cm⁻¹, several characteristic absorption peaks can be seen. For example, 4263 cm⁻¹ is the C–H stretch/C–H deformation in phenyl rings, 4672 cm⁻¹ is the performance of stretching vibration of C–C and C=H bonds, 5164 cm⁻¹ due to the first overtones of C–H bond and maybe also the stretching and deformation of O–H bonds, 5937 cm⁻¹ results from first overtone of stretching C–H bonds in aromatic rings and perhaps also of C–H bonds in vinylene groups, and 6872 cm⁻¹ is the O–H stretching first overtone.^21–24) These signals could be caused by aromatic compounds in Rhodiola crenulata like salidroside and p-tyrosol, indicating that NIR spectra can reflect the chemical information of Rhodiola crenulata.

Fig. 4. Raw NIR Spectra of 50 Rhodiola crenulata Samples for Modeling

Selection of Validation Set

The raw NIR spectra of 50 samples of Rhodiola crenulata were divided into the validation set and calibration set at the approximately ratio of 1 : 4 to establish the NIR models. In order to avoid bias in subset division, 4 plans were considered to select the optimal validation set. According to the sample number in NIR model for salidroside, 4 plans were made as follows: in plan A, the first ten samples were chosen; in plan B, the last 10 samples were chosen; in plan C, the 10 samples in the middle were chosen; in plan D, one spectrum of every five samples was selected into the validation set and a total of 10 samples chosen finally. Instead of 10 samples in each validation set plan, the model of p-tyrosol had 9 in order to get good performance.

Spectral Region Choosing

In the complicated operating system, the option of spectral region was an important part. As shown in Fig. 4, the spectral regions of 50 samples, 10000–8500 cm⁻¹ and 4400–4000 cm⁻¹, have the noise of high level. While choosing the regions for modeling, all the spectral absorption information must not be overlooked. Fortunately, the PLS method of TQ Analyst 8.0 is a powerful multivariate analytical method which can recommend the appropriate spectral region for the NIR models. The region, 8419.69–4574.32 cm⁻¹, was provided to establish the two calibration models.

Spectral Pretreatment Methods

When the spectral data was collected by the NIR instrument; the scatter effects, baseline drift, and noise have appeared synchronously in the spectra. It is vitally necessary to preprocess the spectra before PLS modeling. In fact, there are many spectral pretreatment methods, including: first and second derivative, smoothing, multiplicative scatter correction (MSC), standard normal variate (SNV) transformation, and so on.

In the NIR calibration models for salidroside and p-tyrosol, all spectra preprocessing methods, related parameters, and the corresponding plans of selecting optimal validation sets which were mentioned above were shown in Tables 2(a) and (b).

Table 2. (a). PLS Results for Salidroside Obtained Using Different Spectra Processing Methods

Preprocessing	Plan A			Plan B			Plan C			Plan D
	R2	RMSEC	RMSEP	R2	RMSEC	RMSEP	R2	RMSEC	RMSEP	R2	RMSEC	RMSEP
MSC+SG	0.95656	0.0996	0.2490	0.96137	0.0810	0.0619	0.96519	0.0786	0.1110	0.98325	0.0610	0.0611
MSC+SG+1st derivative	0.99490	0.0344	0.1510	0.99615	0.0258	0.0940	0.99130	0.0396	0.1180	0.99138	0.0438	0.0582
MSC+SG+2nd derivative	0.99400	0.0374	0.1690	0.99616	0.0258	0.0718	0.98826	0.0459	0.1660	0.99572	0.0309	0.0294
MSC+ND+1st derivative	0.97441	0.0768	0.2360	0.99273	0.0354	0.0874	0.98914	0.0442	0.1410	0.99168	0.0431	0.0533
MSC+ND+2nd derivative	0.98171	0.0650	0.2190	0.99439	0.0311	0.0931	0.99507	0.0298	0.1060	0.99121	0.0443	0.0472
SNV+SG	0.96183	0.0935	0.1950	0.96051	0.0818	0.1140	0.97213	0.0705	0.1490	0.98593	0.0559	0.0572
SNV+SG+1st derivative	0.99410	0.0370	0.1210	0.99619	0.0256	0.1050	0.99451	0.0315	0.1340	0.99443	0.0353	0.0438
SNV+SG+2nd derivative	0.99512	0.0337	0.1480	0.99493	0.0296	0.0842	0.99655	0.0249	0.1110	0.99526	0.0325	0.0317
SNV+ND+1st derivative	0.97276	0.0792	0.2170	0.99363	0.0332	0.0940	0.97828	0.0623	0.1100	0.99182	0.0427	0.0528
SNV+ND+2nd derivative	0.98008	0.0678	0.2000	0.99449	0.0308	0.0994	0.98691	0.0485	0.1560	0.98957	0.0482	0.0622

Table 2. (b). PLS Results for p-Tyrosol Obtained Using Different Spectra Processing Methods

Preprocessing	Plan A			Plan B			Plan C			Plan D
	R2	RMSEC	RMSEP	R2	RMSEC	RMSEP	R2	RMSEC	RMSEP	R2	RMSEC	RMSEP
MSC+SG	0.94882	0.0700	0.0610	0.99082	0.0298	0.0544	0.94718	0.0711	0.0987	0.99097	0.0297	0.0395
MSC+SG+1st derivative	0.99785	0.0145	0.0529	0.98694	0.0355	0.0618	0.98202	0.0418	0.0765	0.98631	0.0366	0.0276
MSC+SG+2nd derivative	0.99904	0.00969	0.0681	0.99616	0.0193	0.0873	0.97529	0.0488	0.0539	0.99784	0.0146	0.0194
MSC+ND+1st derivative	0.99192	0.0281	0.0503	0.99229	0.0274	0.0351	0.99189	0.0282	0.0281	0.99069	0.0302	0.0173
MSC+ND+2nd derivative	0.97609	0.0482	0.1210	0.99401	0.0241	0.0438	0.98650	0.0363	0.0262	0.99198	0.0280	0.0213
SNV+SG	0.95043	0.0689	0.0788	0.99173	0.0283	0.0569	0.99307	0.0261	0.0444	0.99245	0.0272	0.0311
SNV+SG+1st derivative	0.99788	0.0144	0.0425	0.99507	0.0218	0.0937	0.99050	0.0305	0.0164	0.98554	0.0376	0.0281
SNV+SG+2nd derivative	0.99934	0.00806	0.0465	0.99736	0.0160	0.0604	0.97496	0.0493	0.0541	0.99714	0.0168	0.0154
SNV+ND+1st derivative	0.98872	0.0332	0.0509	0.99261	0.0267	0.0669	0.99237	0.0273	0.0275	0.99055	0.0304	0.0172
SNV+ND+2nd derivative	0.98076	0.0433	0.0829	0.99426	0.0236	0.0903	0.98917	0.0325	0.0239	0.99019	0.0310	0.0252

Optimization Model

The value of root mean square error of calibration (RMSEC), root mean square error of prediction (RMSEP), and correlation coefficient (R²) were used to evaluate the NIR calibration models. The optimal calibration models had the highest R² as well as lowest RMSEP.²⁵⁾

According to the above standards and Table 2(a), the best calibration model for salidroside was generated based on plan D of choosing the validation set with following spectral pretreatment methods: MSC, the Savitzky–Golay (SG) filter, and second derivative of spectra; as for p-tyrosol, the optimized model was plan D in Table 2(b) with SNV, SG filter and the spectral second derivative, too. As can be seen from Figs. 5(a) and 6(a), MSC was well used to remove the scatter effects and slope variation,²⁶⁾ meanwhile, SNV mathematical method was able to reduce the spectral variations produced by scattering and different particle size distribution²⁷⁾; and the SG filter was frequently utilized to smooth the spectra.²⁸⁾ And, according to Figs. 5(b) and 6(b), the processing by second spectral derivative was performed good to remove linear baseline shifts and enhance resolution of spectra.

Fig. 5. NIR Spectra Processed by MSC+SG (a) and MSC+SG+2nd Derivative (b) for Salidroside

Fig. 6. NIR Spectra Processed by SNV+SG (a) and SNV+SG+2nd Derivative (b) for p-Tyrosol

As can be seen from Table 3, in NIR model of salidroside the R² was 0.99572, as well as RMSEC and RMSEP were 0.0309, 0.0294, respectively. Meanwhile, the R², RMSEC and RMSEP in NIR model of p-tyrosol were 0.99714, 0.0168 and 0.0154, respectively. These results demonstrated that the two models had good performance, and the model for p-tyrosol showed slightly better than salidroside. Additionally, the R² in both Figs. 7(a) and (b) presented a good correlation between the calculated values of NIR model and the actual values of HPLC for salidroside and p-tyrosol, respectively.

Table 3. Parameters of Optimal Calibration Models by PLS Analysis

Model	Spectral pretreatment method	R2	RMSEC	RMSEP	Spectrum region for measurement (cm−1)
NIR model of salidroside	Plan D+MSC+SG+2nd derivative	0.99572	0.0309	0.0294	8419.69–4574.32
NIR model of p-tyrosol	Plan D+SNV+SG+2nd derivative	0.99714	0.0168	0.0154	8419.69–4574.32

Fig. 7. Correlation Diagrams of Salidroside (a) and p-Tyrosol (b) between the Calculated Values by NIR Model and the Actual Values by HPLC

The Optimum Number of Latent Variables (LV)

It is important to find the optimal number of LV in the application of the PLSR algorithm. If the LV numbers applied in the calibration model is more or less than the optimal one, the situation of ‘under-fitting’ or ‘over-fitting’ will appear, which shows the performance of the model may not be good enough.²⁹⁾ In general, the leave-one-out (LOO) cross-validation was used to calculate the optimum LV numbers. If the LV numbers come to the best, the curve will tend toward stability. Whereas, if the LV numbers exceed the optimal one, the PRESS values will almost remain unchanged or increase lightly.³⁰⁾ As shown in Fig. 8, the relation diagrams of PRESS values and LV numbers were used to find the best LV numbers in salidroside and p-tyrosol NIR models. Both in the two NIR calibration models, the optimum LV numbers were 8, and the PRESS values decreased radically while the LV numbers increasing.

Fig. 8. The LV Numbers for PLSR-Calibration of Salidroside (a) and p-Tyrosol (b)

Validation and Evaluation of the Optimization Model

For both salidroside (Fig. 9(a)) and p-tyrosol (Fig. 9(b)), the concentrations determined by HPLC and NIR method were closely parallel. Precision and accuracy were taken into account in this study so as to validate and evaluate the prediction ability of the two NIR spectroscopic models.

Fig. 9. The Concentration of Salidroside (a) and p-Tyrosol (b) via HPLC Method and NIR Method

Precision

This parameter stands for the degree of scatter between a series of measured values of the same sample. The RSD and relative error (RE) of a series of measurements are generally utilized to express the precision. Its standard evaluated in this part was intra- and inter-day assays. The intra-day assay was examined by six samples of three levels of concentration using the optimal models within a day, whereas the inter-day variability assay was tested in four consecutive days respectively. It can be observed from Table 4, that the intra-day assay had good performance with RSD less than 2.0% for the both salidroside and p-tyrosol NIR model. And the inter-day variability was less than 3.5% for the salidroside NIR model, meanwhile, it was less than 5.0% for the p-tyrosol NIR model. So these studies showed that the two models had good precision.

Table 4. The Precision and Recovery Test of Salidroside and p-Tyrosol Assay by NIR Spectroscopy

	Precision test		Recovery test
	Intra-day RSD (%)	Inter-day RSD (%)	Recovery (%)	RSD (%)	RE (%)
NIR model of salidroside
Low-concentration	1.7703	1.9627	96.9557	3.3328	−5.9959
Mid-concentration	0.9190	1.2089	105.4944	3.9863	5.4944
High-concentration	1.7076	3.4860	103.4279	2.5968	3.4279
NIR model of p-tyrosol
Low-concentration	0.4794	4.9919	100.1429	4.7077	0.1429
Mid-concentration	1.2188	2.4493	96.0427	3.6798	−3.9573
High-concentration	1.0042	2.2882	98.6245	3.5723	−1.3755

Accuracy

It represents the level of similarity between the actual value (or reference value) and the value calculated by the NIR calibration model. This parameter is usually expressed as RMSEP, RSD and RE of samples in validation set. Moreover, recovery test and some statistical testing methods were carried through to validate and evaluate the optimized models.

The two developed optimization models above were utilized to respectively calculate the concentration of samples in the validation sets of salidroside and p-tyrosol calibration model. According to the previous part of this study, the RMSEP of salidroside model and p-tyrosol model was 0.0249 and 0.0154, respectively. And a list of values of validation set was made with NIR models in Table 5.

Table 5. Results of Validation Set for Estimation by Optimization Model

Sample No.	Actual values by HPLC (mg/mL)	Calculated values by NIR (mg/mL)	RSD (%)	RE (%)
NIR model of salidroside
5	0.2376	0.2304	2.1836	−3.0443
10	0.3745	0.3765	4.9971	0.5251
15	0.3230	0.3184	0.9521	−1.4241
20	0.1597	0.1502	0.0769	−5.9695
25	0.5622	0.5912	6.4164	5.1583
30	0.8516	0.8479	4.1392	−0.4345
35	0.9910	1.0582	0.3166	6.7777
40	0.7652	0.7867	0.7302	2.8141
45	0.4109	0.3982	2.8996	−3.0827
50	0.7606	0.7851	1.1252	3.2211
NIR model of p-tyrosol
5	0.0247	0.0250	6.1150	1.3495
10	0.3856	0.3773	0.9826	−2.1611
15	0.5897	0.5896	0.8627	−0.0170
20	0.0704	0.0711	0.2813	0.9943
25	0.2129	0.2077	2.3494	−2.4581
30	0.1749	0.1719	3.5537	−1.7343
35	0.1355	0.1449	0.5786	6.9619
45	0.5398	0.5415	1.1571	0.3211
50	0.5010	0.4902	1.8904	−2.1557

The recovery test in three known concentration levels also reflected the accuracy. Six samples of the three levels in a validation set were tested by the optimized model, respectively, and then the average recovery was obtained. As shown in Table 4, the recovery test results of the salidroside and p-tyrosol NIR model were between 95 and 105.5% with the RSD less than 5.0% and RE less than 6.0% for the targets.

The accuracy for the predicted results of the validation set was also tested with t and F test, which demonstrated that the accuracy was satisfactory with a significant level of 0.05. Therefore, these works in this part proved that the accuracy of the two models was good.

As can be seen in Tables 4 and 5, the validation and evaluation results indicate that the two NIR spectroscopic models of PLS are precise, accurate, and stable and have good properties for quantitative analysis of salidroside and p-tyrosol.

Conclusion

According to the results, the two NIR spectroscopic models of PLS provided precise, accurate, and repeatable quantitative analysis for salidroside and p-tyrosol in Rhodiola crenulata which is a significant traditional Tibetan medicine and traditional Chinese medicine. Compared with traditional methods such as HPLC, TLC, and GC, NIR spectroscopy is nondestructive, rapid and efficient. Additionally, the results can provide technical support for the further analysis of salidroside and p-tyrosol in Rhodiola crenulata. Moreover, this method can be utilized to control the quality of traditional Tibetan medicine and traditional Chinese medicine without sample preparation.

Acknowledgments

This research was supported by the Science and Technology Department of Sichuan Province Research Fund (Grant No.: 2016JY0247). We are also grateful to Dr. Carden, kelly A. (The University of Iowa, America) for proofreading an early version of the manuscript.

Conflict of Interest

The authors declare no conflict of interest.

References

• 1) Han X., Zhang T. Y., Wei Y., Cao X. L., Ito Y., J. Chromatogr. A, 971, 237–241 (2002).
• 2) Zhang Z. H., Liu J. X., Shang X. H., Yang J. H., Chu J. N., Wang Z. G., Yao Z. Y., Ma H. L., Li Q. Y., Wang Y. S., China J. Chin. Mater. Med., 23, 104–106 (1998).
• 3) Zhang Z. J., Tong Y., Zou J., Chen P. J., Yu D. H., Chin. J. Integr. Med., 15, 177–183 (2009).
• 4) Bucci L. R., Am. J. Clin. Nutr., 72 (Suppl.), 624S–636S (2000).
• 5) Chen S. L., Zhang D. J., Chen S. Y., Xia T., Gao Q. B., Duan Y. Z., Zhang F. Q., Chromatographia, 68, 299–302 (2008).
• 6) Zhang X., Chin. Wild Plant Resour., 2, 33–35 (1993).
• 7) Xu J. F., Ying P. Q., Su Z. G., Chin. Tradit. Herbal Drugs, 29, 202–205 (1998).
• 8) Iaremiǐ I. N., Grigor’eva N. F., Eksp. Klin. Farmakol., 65, 57–59 (2002).
• 9) Kanupriya, Prasad D., Sai Ram M., Kumar R., Sawhney R. C., Sharma S. K., Ilavazhagan G., Kumar D., Banerjee P. K., Mol. Cell. Biochem., 275, 1–6 (2005).
• 10) Kucinskaite A., Briedis V., Savickas A., Medicina (Kaunas), 40, 614–619 (2004).
• 11) Díaz Lanza A. M., Abad Martinez M. J., Fernández Matellano L., Recuero Carretero C., Villaescusa Castillo L., Silvan Sen A. M., Bermejo Benito P., Planta Med., 67, 219–223 (2001).
• 12) Li T., Zhang H., Chem. Pharm. Bull., 56, 807–814 (2008).
• 13) An F., Yue S. J., Guo D. A., Zheng J. H., China J. of Chin. Mater. Med., 23, 43–44 (1998).
• 14) Jefferson A., Stacey C. L., Scheinmann F., J. Chromatogr. A, 57, 247–254 (1971).
• 15) Pothitirat W., Gritsanapan W., J. Health Res., 22, 161–166 (2008).
• 16) Otsuka M., Chemom. Intell. Lab. Syst., 82, 109–114 (2006).
• 17) Mark H., Ritchie G. E., Roller R. W., Ciurczak E. W., Tso C., Mac Donald S. A., J. Pharm. Biomed. Anal., 28, 251–260 (2002).
• 18) Ritchie G. E., Roller R. W., Ciurczak E. W., Mark H., Tso C., Mac Donald S. A., J. Pharm. Biomed. Anal., 29, 159–171 (2002).
• 19) Kreft K., Kozamernik B., Urleb U., Int. J. Pharm., 177, 1–6 (1999).
• 20) Chen Y., Xie M. Y., Zhang H., Wang Y. X., Nie S. P., Li C., Food Chem., 135, 268–275 (2012).
• 21) Li W. L., Xing L. H., Cai Y., Qu H. B., Vib. Spectrosc., 55, 58–64 (2011).
• 22) Li W., Wang Y., Qu H., Vib. Spectrosc., 62, 159–164 (2012).
• 23) Goddu R. F., Delker D. A., Anal. Chem., 32, 140–141 (1960).
• 24) Ying X. H., Pei Y., Liu M. Y., Ding G. Y., Jiang M., Liang Q. L., Wang Y. M., Bai G., Luo G. A., Anal. Merhodss-UK, 6, 4212–4218 (2014).
• 25) Liew C. V., Karande A. D., Heng P. W., Int. J. Pharm., 386, 138–148 (2010).
• 26) Maleki M. R., Mouazen A. M., Ramon H., De Baerdemaeker J., Biosystems Eng., 96, 427–433 (2007).
• 27) Barnes R. J., Dhanoa M. S., Lister S. J., Appl. Spectrosc., 43, 772–777 (1989).
• 28) Zimmermann B., Kohler A., Appl. Spectrosc., 67, 892–902 (2013).
• 29) Liu W., Wang Z. Z., Qing J. P., Li H. J., Xiao W., Pharmacogn. Mag., 10, 441–448 (2014).
• 30) Kuriakose S., Thankappan X., Joe H., Venkataraman V., Analyst, 135, 2676–2681 (2010).