**Investigation of Organic Acids in Sa**ff**ron Stigmas (***Crocus sativus* **L.) Extract by Derivatization Method and Determination by GC**/**MS**

**Laurynas Jarukas <sup>1</sup> , Olga Mykhailenko <sup>2</sup> , Juste Baranauskaite <sup>3</sup> , Mindaugas Marksa 1,\* and Liudas Ivanauskas <sup>1</sup>**


Received: 25 May 2020; Accepted: 25 July 2020; Published: 28 July 2020

**Abstract:** The beneficial health properties of organic acids make them target compounds in multiple studies. This is the reason why developing a simple and sensitive determination and investigation method of organic acids is a priority. In this study, an effective method has been established for the determination of organic (lactic, glycolic, and malic) acids in saffron stigmas. *N*-(*tert*-butyldimethylsilyl)-*N*-methyltrifluoroacetamide (MTBSTFA) was used as a derivatization reagent in gas chromatography combined with mass spectrometric detection (GC/MS). The saffron stigmas extract was evaporated to dryness with a stream of nitrogen gas. The derivatization procedure: 0.1 g of dried extract was diluted into 0.1 mL of tetrahydrofuran, then 0.1 mL MTBSTFA was orderly and successively added into a vial. Two different techniques were used to obtain the highest amount of organic acid derivatives from saffron stigmas. To the best of our knowledge, this is the first report of the quantitative and qualitative GC/MS detection of organic acids in saffron stigmas using MTBSTFA reagent, also comparing different derivatization conditions, such as time, temperature and the effect of reagent amount on derivatization process. The identification of these derivatives was performed via GC-electron impact ionization mass spectrometry in positive-ion detection mode. Under optimal conditions, excellent linearity for all organic acids was obtained with determination coefficients of R <sup>2</sup> > 0.9955. The detection limits (LODs) and quantitation limits (LOQs) ranged from 0.317 to 0.410 µg/mL and 0.085 to 1.53 µg/mL, respectively. The results showed that the highest yield of organic acids was conducted by using 0.1 mL of MTBSTFA and derivatization method with a conventional heating process at 130 ◦C for 90 min. This method has been successfully applied to the quantitative analysis of organic acids in saffron stigmas.

**Keywords:** lactic acid; malic acid; glycolic acid; GC-MS/EI

### **1. Introduction**

Saffron is considered the world's most expensive spice and medicinal plant. Besides uses in food, saffron has attracted interest because of its health-promoting properties [1,2]. In addition, saffron stigmas have been proven to have anti-inflammatory, antioxidant, anti-allergic, and antidepressant biological functions [3,4]. These specific properties are considered to be connected to the presence of diverse compounds such as proteins, fats, minerals, sugars, and organic acids [2]. Among these ingredients,

organic acids are in a prominent position because not only do they affect the flavor of saffron but also, they have various pharmacological actions [5,6].

Moreover, there is increasing interest in studies examining characteristics of organic acids, searching for positive effects of given compounds on the human body. According to the literature, the short-chain fatty acids, medium-chain fatty acids, and other organic acids have more or less pronounced antimicrobial activity, depending on the concentration of the acids and the bacterial species exposed to the acids [5,6]. It is well known that benzoic and salicylic acids exhibit antibacterial activity, hydroxycinnamic acid, and their derivatives–anti-inflammatory activity, gallic acid is an antimutagenic, anticarcinogenic, and anti-inflammatory agent [7]. Furthermore, succinic acid, acetic acid, citric acid, lactic acid, malic acid, glutamic acid, and their salts promote gastrointestinal absorption of iron [7]. Moreover, citric and malic acids have significant protective effects on the myocardium and act on ischemic lesions, according to a study by Tang et al., where supplying a patients' diet with these compounds gave significant positive results [8]. Moreover, organic acids play a principal role in maintaining the quality and nutritional value of food. These compounds can be added as acidulants or stabilizers (e.g., citric, ascorbic, benzoic, fumaric, and malic acids).

The quantitative determination of organic acids in such type of samples is of high interest in many industrial and research institutes. However, as the attention on the health benefits of organic acids is increasing, a simple and sensitive method for determining and investigating organic acids is needed [9–12]. According to the literature, organic acids are weak acids and are only partly dissociated [6]. To increase the stability and solubility of organic acids, different extraction or other chemical processing techniques are needed. The scientific literature describes that one of the most commonly used processes is derivatization, a chemical process for modifying compounds in order to generate new products with better chromatographic properties [13]. Different derivatization techniques can be used; chemical derivatization is usually used for amino acid detection, and it becomes a necessary procedure to transform analytes into derivatives that can be easily isolated, separated, and detected [14–17]. Gas chromatography is the most widely used and accepted technique for quantitative analysis of derivatization products due to its high resolution, sensitivity, great versatility, and simple sample treatment.

The derivatization parameters were systematically studied. In this study, a simple and sensitive GC/MS method for determination and investigation of organic acids in saffron stigmas after derivatization with *N*-(*tert*-butyldimethylsilyl)-*N*-methyltrifluoroacetamide reagent was presented. The adequacy of the proposed method was estimated in terms of accuracy, linearity, precision, and detection limit. To the best of our knowledge, this is the first report of quantitative and qualitative organic acid GC/MS detection in saffron stigmas with MTBSTFA derivatization reagent, and comparison of different conditions.

#### **2. Results and Discussion**

#### *2.1. Derivatization Solvents and Reagents*

When it comes to dealing with highly complex matrices, such as organic acids, it is advisable to use a derivatization process in order to improve parameters of separation, such as volatility, thermal stability, resolution, as well as detection parameters, when gas chromatography is used [18]. During this process, the derivatization reagent plays an important role in the separation and resolution of the analytes [18]. Among the different derivatization reagents for organic acids, silylating agents are the most popular ones, which, moreover, have been proven as excellent reagents for derivatization after extraction. Hence, two silylating agents were tested: *N*-methyl-*N*-(trimethylsilyl) trifluoroacetamide (MSTFA) and MTBSTFA [13,19,20]. This study is intended for increasing knowledge about the behavior and interest of both reagents on the efficiency of organic acid derivatization yield. However, the use of MSTFA yielded an incomplete crocus stigmas extract derivatization of the organic acid, and thus, further studies were performed with MTBSTFA. According to results, significantly 1.2 times higher

amount of organic acids (expressed in percentage of the total amount of lactic, malic, and glycolic acids) has been found while using MTBSTFA reagent, and it was selected as a derivatization reagent for further study. Similar results were reported in previous studies [21,22]. Morville et al. evaluated the efficiency of the derivatization process on organic acids' (glutaric, adipic, and suberic acids) yield when MTBSTFA and MSTFA reagents were used. Results showed that glutaric, adipic and suberic acid yields significantly increased, 1.6, 1.8, and 1.3 times respectively, after using MTBSTFA reagent [21]. This might be explained by the high volatility of MTBSTFA that it did not coelute in the GC system with other peaks and improved parameters of separation, thermal stability and resolution. The MTBSTFA produces dimethyl-*tert*-butylsilyl (TBDMS) derivatives. MTBSTFA-derivatives produce characteristic fragmentation patterns presenting mainly the fragments of [M <sup>−</sup> 15]<sup>+</sup> (cleavage of methyl from the molecular ion) and [M <sup>−</sup> 57]<sup>+</sup> (cleavage of the *<sup>t</sup>*-butyl moiety), of which [M <sup>−</sup> 57]<sup>+</sup> is generally dominant in the mass spectrum. MSTFA-derivatives yielding trimethylsilyl (TMS) derivatives, mainly show the fragments [M <sup>−</sup> 15]<sup>+</sup> (cleavage of a methyl from the molecular ion) and [M <sup>−</sup> 31]<sup>+</sup> (cleavage of the trimethylsilyl ether moiety followed by cyclization involving the silyl group). The TBDMS are more stable to hydrolysis than the corresponding TMS derivatives. As demonstrated in previous studies, TBDMS derivatives are formed more easily and have, thus, higher sensitivities (10–100 times) as well as repeatabilities than the corresponding TMS derivatives [22]. and it was selected as a derivatization reagent for further study. Similar results were reported in previous studies [21,22]. Morville et al. evaluated the efficiency of the derivatization process on organic acids' (glutaric, adipic, and suberic acids) yield when MTBSTFA and MSTFA reagents were used. Results showed that glutaric, adipic and suberic acid yields significantly increased, 1.6, 1.8, and 1.3 times respectively, after using MTBSTFA reagent [21]. This might be explained by the high volatility of MTBSTFA that it did not coelute in the GC system with other peaks and improved parameters of separation, thermal stability and resolution. The MTBSTFA produces dimethyl-*tert*butylsilyl (TBDMS) derivatives. MTBSTFA-derivatives produce characteristic fragmentation patterns presenting mainly the fragments of [M − 15]+ (cleavage of methyl from the molecular ion) and [M − 57]+ (cleavage of the *t*-butyl moiety), of which [M − 57]+ is generally dominant in the mass spectrum. MSTFA-derivatives yielding trimethylsilyl (TMS) derivatives, mainly show the fragments [M − 15]+ (cleavage of a methyl from the molecular ion) and [M − 31]+ (cleavage of the trimethylsilyl ether moiety followed by cyclization involving the silyl group). The TBDMS are more stable to hydrolysis than the corresponding TMS derivatives. As demonstrated in previous studies, TBDMS derivatives are formed more easily and have, thus, higher sensitivities (10–100 times) as well as repeatabilities than the corresponding TMS derivatives [22]. The ultimate goal of extraction is the maximization of the yield and coverage of metabolites in a

*Molecules* **2020**, *25*, x FOR PEER REVIEW 3 of 10

The ultimate goal of extraction is the maximization of the yield and coverage of metabolites in a rapid and reproducible way while minimizing enzymatic, chemical, and physical degradation [20]. The derivatization yield of carboxylic acids with MTBSTFA depends on factors including the nature of the solvent in which the analytes are dissolved. The main factors contributing to the increase of the efficiency and the rate of the silylation reaction are the silyl donor ability of the reagent and the ease of silylation of different functional groups in the analyte. The solvent used as a medium and the compounds present or added in the silylation medium may also play a role in derivatization efficiency. The reagent excess is sometimes important for displacing the equilibrium in the desired direction, and usually, an excess up to ten times larger than stoichiometrically needed is used for silylation. The primary purpose was to determine the effect of different solvents (tetrahydrofuran (THF) and acetonitrile (ACN)) on derivatization's yield of organic acid. During this part of the study, two samples were produced by using derivatization procedure, mentioned above (sample of 0.1 g dried Saffron stigmas extract was diluted into 0.1 mL of extraction solvent, and 0.1 mL derivatization agent (MTBSTFA) was added in sequence; the vial was sealed and oscillated by vortex-mixer for 1 min, then the mixture was placed in glycerol bath allowing it to react at 50 ◦C for 60 min). Primary investigations revealed that between the ACN and THF samples, significant differences in the organic acid derivate yields (expressed in percentage of the total amount of lactic, malic, and glycolic acids) were obtained (*p* < 0.05; Figure 1). rapid and reproducible way while minimizing enzymatic, chemical, and physical degradation [20]. The derivatization yield of carboxylic acids with MTBSTFA depends on factors including the nature of the solvent in which the analytes are dissolved. The main factors contributing to the increase of the efficiency and the rate of the silylation reaction are the silyl donor ability of the reagent and the ease of silylation of different functional groups in the analyte. The solvent used as a medium and the compounds present or added in the silylation medium may also play a role in derivatization efficiency. The reagent excess is sometimes important for displacing the equilibrium in the desired direction, and usually, an excess up to ten times larger than stoichiometrically needed is used for silylation. The primary purpose was to determine the effect of different solvents (tetrahydrofuran (THF) and acetonitrile (ACN)) on derivatization's yield of organic acid. During this part of the study, two samples were produced by using derivatization procedure, mentioned above (sample of 0.1 g dried Saffron stigmas extract was diluted into 0.1 mL of extraction solvent, and 0.1mL derivatization agent (MTBSTFA) was added in sequence; the vial was sealed and oscillated by vortex-mixer for 1 min, then the mixture was placed in glycerol bath allowing it to react at 50 °C for 60 min). Primary investigations revealed that between the ACN and THF samples, significant differences in the organic acid derivate yields (expressed in percentage of the total amount of lactic, malic, and glycolic acids) were obtained (*p* < 0.05; Figure 1).

**Figure 1.** The effect of extraction solvents on the yield of organic acid derivatives (expressed in percentage of the total amount of lactic, malic, and glycolic acids) from Saffron stigmas extract, *n* = 6. **Figure 1.** The effect of extraction solvents on the yield of organic acid derivatives (expressed in percentage of the total amount of lactic, malic, and glycolic acids) from Saffron stigmas extract, *n* = 6. Values within columns followed by the same lowercase letter (a, b) differed statistically at *p* < 0.05 (Tukey's test). Results are expressed as means ± standard error.

Moreover, the results showed 1.2 times higher yield of organic acids derivate by using THF as the derivatization solvent in comparison to ACN (*p* < 0.05) (Figure 1). The results could be explained by the sensitivity of the analysis. According to Wittmann et al., by using the THF as extraction solvent increased analysis sensitivity of more polar compounds (lactic acid, oxalic acid, methylcitric acid, 3-hydroxypropionic acid, 3-hydroxyisovaleric acid, kynurenic acid, glycolic acid, orotic acid and quinolinic acid) in comparison with less polar compounds (glycine conjugates) [23]. Hence, the THF was chosen for future experiments. Values within columns followed by the same lowercase letter (**a**,**b**) differed statistically at *p* < 0.05 (Tukey's test). Results are expressed as means ± standard error. Moreover, the results showed 1.2 times higher yield of organic acids derivate by using THF as the derivatization solvent in comparison to ACN (*p* < 0.05) (Figure 1). The results could be explained by the sensitivity of the analysis. According to Wittmann et al., by using the THF as extraction solvent increased analysis sensitivity of more polar compounds (lactic acid, oxalic acid, methylcitric acid, 3 hydroxypropionic acid, 3-hydroxyisovaleric acid, kynurenic acid, glycolic acid, orotic acid and quinolinic acid) in comparison with less polar compounds (glycine conjugates) [23]. Hence, the THF was chosen for

During this study we compared MSTFA and MTBSTFA in the derivatization efficiency of organic acids. This study is intended for increasing knowledge about the behavior and interest of both reagents. These results clearly demonstrate that solvent plays a significant role in the derivatization procedure. As a matter of fact, MTBSTFA and THF possess the most appropriate derivatization efficiency of the above-mentioned compounds, and they were selected for further studies. future experiments. During this study we compared MSTFA and MTBSTFA in the derivatization efficiency of organic acids. This study is intended for increasing knowledge about the behavior and interest of both reagents. These results clearly demonstrate that solvent plays a significant role in the derivatization procedure. As a matter of fact, MTBSTFA and THF possess the most appropriate derivatization efficiency of the abovementioned compounds, and they were selected for further studies.

#### *2.2. Comparison of the Derivatization Parameters 2.2. Comparison of the Derivatization Parameters*

Sample preparation is a critical part of every analytical procedure. The increasing demand to determine compounds at low concentrations in complex matrices requires a preliminary step. To achieve the best derivatization efficiency, a variety of important parameters, such as a derivatization temperature, time and amount of MTBSTFA, were optimized. In this study, the saffron stigmas extract was employed to optimize derivatization conditions. The concentration trends of three representative organic acids (lactic, glycolic, and malic acids) relative to different parameters are shown in Figure 2. Sample preparation is a critical part of every analytical procedure. The increasing demand to determine compounds at low concentrations in complex matrices requires a preliminary step. To achieve the best derivatization efficiency, a variety of important parameters, such as a derivatization temperature, time and amount of MTBSTFA, were optimized. In this study, the saffron stigmas extract was employed to optimize derivatization conditions. The concentration trends of three representative organic acids (lactic, glycolic, and malic acids) relative to different parameters are shown in Figure 2.

**Figure 2.** The effect of operation parameters on derivatization reaction: (the yield of lactic acid—62 µg/g, glycolic acid—40 µg/g and malic acid—30 µg/g in the example sample). Temperature (**a**), Time (**b**), derivatization agent (MTBSTFA) (**c**). Results are expressed as means ± standard error (*n* = 6). **Figure 2.** The effect of operation parameters on derivatization reaction: (the yield of lactic acid—62 µg/g, glycolic acid—40 µg/g and malic acid—30 µg/g in the example sample). Temperature (**a**), Time (**b**), derivatization agent (MTBSTFA) (**c**). Results are expressed as means ± standard error (*n* = 6).

The primary purpose was to determine the influence of different temperatures and extraction time on the derivatization yield of investigated organic acids. The results showed that increasing temperature significantly increased the yields of organic acids (Figure 2a; *p* < 0.05). At a derivatization temperature of 130 ◦C, lactic, glycolic, and malic acid derivative yields significantly increased, 7.7, 16.1, and 5 times, respectively, in comparison with samples prepared at a temperature of 25 ◦C

(Figure 2a). The explanation of such results could be that higher temperature is speculated to enhance derivatization efficiency, by increasing solubility of derivatization reagents and organic acid metabolites [24]. Similarly, Gulberg et al. found that increasing temperatures had an appreciable effect on derivatization efficiency [24]. Moreover, the effect of extraction time (30, 60, 90, 120, 150, 210, and 240 min) on the yield of organic acids derivatization was investigated. The derivatization procedure was carried in the same way as mentioned above. Primary investigations revealed that a prolonged extraction time of 90 min, had no significant influence on derivatization yields of lactic, glycolic and malic acids, extracted from saffron stigmas (*p* > 0.05; Figure 2b). As the reaction time was prolonged, the signal response of organic acid derivatives remained constant. Similarly, Elias and co-authors found that long term silylation-derivatization process was beneficial to stearic acid and glucose-6-phosphate [25]. Therefore, the derivatization reaction between MTBSTFA and organic acids was carried out for 90 min at 130 ◦C.

To ensure the complete and repeatable derivatization, the desired amount of MTBSTFA reagent was required. The influence of different volumes (50, 100, 150 µL) of MTBSTFA reagent on organic acid yields was optimized. As indicated in Figure 2c, the highest derivatization yield of organic acids was obtained when the reagent amount was 100 µL (*p* < 0.05). However, a decrease in derivatization efficiency was observed when the MTBSTFA reagent amount in the extract increased to 150 µL. According to literature, the most reported amounts of silylation agents for the silylation of polar plant extracts range between 30–125 µL [26]. This is in line with the findings of Koek, who showed that organic acids and sugars need relatively low volumes of the silylating agent [27]. It could be explained as molecular interaction because the reactivity of TMS groups is low to oxygen in organic acids, having a lower number of unshared electrons, higher steric hindrance, and transition state energy [25].

As a final conclusion of this study, when the derivatization temperature was 130 ◦C, silylation time was 90 min, and the MTBSTFA reagent amount was 100 µL, the highest yield of organic acid derivates in the sample, extracted from the Saffron stigmas was obtained (Table 1). Such a significant shortening of time was achieved by applying high temperature, which allowed to avoid time-consuming sorption of derivatization reagent and time-consuming desorption of analytes, which allowed to reach the high derivatization efficiency. Moreover, the rapid derivatization procedure improved parameters of separation, such as volatility, thermal stability, resolution, as well as detection parameters. For the standard (lactic, glycolic, and malic acids) and saffron stigmas extract, produced by using optimal conditions, chromatograms are shown in Figure 3.


**Table 1.** Linearity and sensitivity data for organic acids (lactic, glycolic, malic acids) used as a standard.

Experimental conditions as in Section 3.4. \* For R<sup>2</sup> the correlation coefficient. The *p* value was <0.0001 for all calibration curves. <sup>a</sup> *Molecules*  LOD were based on S/N = 3; LOQ were based on S/N = 10. **2020**, *25*, x FOR PEER REVIEW 6 of 10

**Figure 3.** *Cont.*

**Figure 3.** GC/MS chromatograms of standards (1-lactic, 2-glycolic, and 3-malic acids) and organic acid

The GC method was validated by following the ICH Q2 (R1) guidelines [28]. The developed method was evaluated via the correlation coefficient (R2), linear range, detection limit (LOD), quantitative limit (LOQ), accuracy, and precision. The electron impact ionization of lactic, glycolic

derivatives extracted from Saffron stigmas.

*2.3. GC/MS Method Validation* 

**Figure 3.** GC/MS chromatograms of standards (1-lactic, 2-glycolic, and 3-malic acids) and organic acid derivatives extracted from Saffron stigmas. **Figure 3.** GC/MS chromatograms of standards (1-lactic, 2-glycolic, and 3-malic acids) and organic acid derivatives extracted from Saffron stigmas.

#### *2.3. GC/MS Method Validation*  The GC method was validated by following the ICH Q2 (R1) guidelines [28]. The developed *2.3. GC*/*MS Method Validation*

method was evaluated via the correlation coefficient (R2), linear range, detection limit (LOD), quantitative limit (LOQ), accuracy, and precision. The electron impact ionization of lactic, glycolic The GC method was validated by following the ICH Q2 (R1) guidelines [28]. The developed method was evaluated via the correlation coefficient (R<sup>2</sup> ), linear range, detection limit (LOD), quantitative limit (LOQ), accuracy, and precision. The electron impact ionization of lactic, glycolic and malic acids, produced the [M]<sup>+</sup> ions at 156, 83 and 84 under positive ionization conditions. The product ion spectra ions at *m*/*z* 147, *m*/*z* 73, and *m*/*z* 73 were produced as the prominent product ions for lactic, glycolic and malic acids (Table 1). The calibration curves of the three organic acids (lactic, glycolic, and malic acids) were established by injecting the standard solutions in the range of 15–242 µg·L −1 , 12–379 µg·L <sup>−</sup><sup>1</sup> and 12–758 µg·L −1 , respectively and plotting the average peak areas versus the average concentrations of organic acids based on the data of triplicate measurements. The good linearity response over the tested concentration range was obtained with the developed method for the compounds used as lactic acid standards, having R<sup>2</sup> > 0.997, as shown in Table 1. The LOD value was 0.153 µg/mL, while the LOQ was 0.317 µg/mL (Table 1), which indicates that the method is sensitive. Moreover, R<sup>2</sup> values of the glycolic acid standard were higher than 0.996, thus confirming the linearity of the method (Table 1). Thus, the LOD value was 0.101 µg/mL, while the LOQ was 0.41 µg/mL (Table 1), which suggested full capacity for the quantification of the glycolic acid investigated. Furthermore, R<sup>2</sup> values of the malic acid standard were higher than 0.999, and the LOD value was 0.085 µg/mL, while the LOQ was 0.339 µg/mL (Table 1). According to the described data above, it can be concluded that this method is a reliable tool for the identification and quantification of organic acid in saffron stigmas, conforming to the ICH guidelines.

#### **3. Materials and Methods**

#### *3.1. Materials and Methods*

Ultrapure water was obtained in the laboratory using a Milli-Q water purification system (Millipore, Billerica, MA, USA). *N*-(t-butyldimethylsilyl)-*N*-methyltrifluoroacetamide (MTBSTFA) (>99%), *N*-methyl-*N*-(trimethylsilyl) trifluoroacetamide (MSTFA) (>98.5%) tetrahydrofuran, acetonitrile were

purchased from Sigma–Aldrich (St. Louis, MO, USA). The GC-equipment was run with helium (purity 5.0) as the carrier gas was purchased from Gazchema (Lithuania). Ethanol (96%) for extraction was purchased from Vilniaus degtine (Vilnius, Lithuania). Lactic acid ( ˙ >98%), glycolic acid (>99%), and malic acid (>99%) standards were purchased from Sigma–Aldrich (Co., Birkenhead, UK).

#### *3.2. Sample Preparation*

Saffron stigma was purchased from Novin Saffron Company, Mashhad, Iran. Prior to the extract preparation, the saffron stigma was dried with a stream of nitrogen gas, then was grounded in a cross beater mill IKA A11 Basic Grinder (IKA Works, Guangzhou, China) and sieved using vibratory sieve shaker AS 200 basic (Retch, UK) equipped with a 125 µm sieve. Then, the powdered sample (1 g) was extracted with 10 mL of 70% (*v*/*v*) methanol-aqua solution in a volumetric flask using an ultrasound bath for 20 min and filtered through a 0.45 µm nylon filter.

#### *3.3. Derivatization Procedure*

0.1 g of prepared extract solution was evaporated to dryness with a stream of nitrogen gas. Briefly, to a 2 mL ampoule bottle, 0.1 g of dried extract sample was diluted into 0.1 mL of extraction solvent (tetrahydrofuran), and 0.1 mL derivatization agent (MTBSTFA) was added in sequence. The vial was sealed and oscillated by vortex-mixer for 1 min, and then, to allow the mixture to react at room temperature (25 ◦C) for 60 min, it was placed in glycerol bath. The subsequent solution was transferred to 200 µL insert placed autosampler vials, and 1 µL aliquot was injected into GC-MS system for analysis. Efficiency extraction parameters were evaluated and optimized, including derivatization time, extraction temperature and reagent amount on derivatization. The comparison of chromatographic responses was used to evaluate the extraction efficiency. A similar procedure was used for lactic, glycolic, and malic acid standards. Standards were diluted in the THF, producing a mixture of 150 µmol L−<sup>1</sup> . The same derivatization reaction was applied to this standard mixture, with the exception of the addition of 0.1 mL of THF, because it was already present in the standard mixture.

### *3.4. GC*/*MS Method*

Analyses were performed using a SHIMADZU GC/MS-QP2010nc Ultra chromatography system (coupled to an Electron Ionization (EI) ion source and a single quadrupole MS (Shimadzu Technologies, Kyoto, Japan). A robotic autosampler and a split/splitless injection port were used. Injection port temperature was kept at 250 ◦C until the end of the analysis. The separation of analytes was carried out on a with Rxi-5 ms (Restek Corporation, Bellefonte, PA, USA, capillary column (30 m long, 0.25 mm outer diameter and 0.25 µm liquid stationary phase thickness) with a liquid stationary phase) 5% diphenyl and 95% polydimethylsiloxane) with helium at a purity of 99.999% as the carrier gas in a constant flow of 1.49 mL/min. The oven temperature was programmed at 75 ◦C for 5 min, then increased to 290 ◦C at 10 ◦C/min and increased to 320 ◦C at 20 ◦C/min and kept for 10 min. The total time was 41 min. The temperatures of the MS interface and ion source were set at 280, 200 ◦C, respectively. The MS was operated in positive mode (electron energy 70 eV). The full-scan acquisition was performed with the mass detection range set at 35–500 *m*/*z* to determine retention times of analytes, optimize oven temperature gradient, and to observe characteristic mass fragments for each compound. Data acquisition and analysis were executed by LabSolution GC/MS (version 5.71) (Shimadzu Corporation). For the identification and quantification of the analytes, single-ion monitoring (SIM) mode was used.

### *3.5. GC*/*MS Method Validation*

The validation of the GC/MS method was performed according to the international guidelines on analytical techniques for the quality control of pharmaceuticals (ICH guidelines) [28]. Method validation was performed to assess linearity, LOD, LOQ, and precision. The calibration curves of the organic acids (lactic, malic and glycolic acids) were established by injecting the standard solutions in the range of 12–758 µg/mL) and plotting the average peak areas versus the average concentrations of organic acids based on the data of triplicate measurements (Table 1). Analytes stock solution was prepared in THF by diluting of the analytical standards to reach a concentration of lactic acid 242 µg/mL, glycolic acid 379 µg/mL, and malic acid 758 µg/mL. Then, the subsequent dilutions were prepared with MilliQ water. The standard solutions (*n* = 3) were prepared at approximate concentrations of lactic acid: 242, 121, 60.5, 30.25, 15.13 µg/mL; glycolic acid: 379, 189.5, 94.75, 47.38, 23.69, 11.84 µg/mL, and malic acid: 758, 379, 189.5, 94.75, 47.38, 23.69, 11.84 µg/mL, due to the wide levels found in saffron samples. The concentrations of the lactic, glycolic and malic acids in each solution was maintained arranged as follows 15.13, 11.84, 11.84 µg/mL, respectively. The LOD and LOQ were calculated at signal-to-noise (S/N) ratios of 3 and 10, respectively. The precision of the method was evaluated by calculating the repeatability (r). The precision of the extraction technique was validated by repeating the extraction procedure with the standard mix solutions six times. An aliquot of each extract was then injected and quantified. The precision of the chromatographic system was tested by checking the %RSD of retention times and peak areas. Six injections were performed each day for three consecutive days.

#### *3.6. Statistical Analysis*

We used between five and six biological replicates for saffron stigmas extract samples and six technical replicates for standard samples. Each biological replicate consisted of a stigma of 10 plants, resulting in the isolation of 4–8 mg of stigmas. Raw data was assessed using ANOVA statistical testing (specifically one-way analysis of variance) and Tukey's multiple comparison test. For this purpose, a software package was utilized (Prism v. 5.04, GraphPad Software Inc., La Jolla, CA, USA) with statistical significance being defined as *p* < 0.05. Results were expressed as average ± standard error.

#### **4. Conclusions**

The method for organic acid analysis in saffron was developed and validated. The method consisted of sample preparation, derivatization, and chromatographic analysis. All steps were extensively studied and optimized for the derivatization procedure. To the best of our knowledge, this is the first report for the GC/MS detection of the amount and types of organic acids in saffron stigmas with MTBSTFA derivatization reagent and comparison of different conditions. The derivatization reaction was rapid, and the maximum yields of organic acids (lactic, glycolic, and malic acids) were observed by using optimal derivatization conditions. The major advantages of optimal conditions led to reach the highest derivatization efficiency of organic acids in only 90 min by using a conventional heating process. The developed method has been successfully applied to the quantification of organic acids in saffron stigmas. This research also shows the interesting agricultural potential of saffron stigmas, in relation to the preparation of certified extracts with a high content of organic acid to be used in the pharmaceutical and nutraceutical area.

**Author Contributions:** L.J. contributed to investigation, data analysis, and original draft preparation. M.M. and J.B. contributed to methodology, data analysis, visualization, review, and editing. L.I. and O.M. contributed to conceptualization, resources, original draft preparation, review and editing, project administration, and supervision. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors are thankful for the financial support provided by the Science Foundation of Lithuanian University of Health Sciences.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

1. Guo, M.; Shi, T.; Duan, Y.; Zhu, J.; Li, J.; Cao, Y. Investigation of amino acids in wolfberry fruit (*Lycium barbarum*) by solid-phase extraction and liquid chromatography with precolumn derivatization. *J. Food Compos. Anal.* **2015**, *42*, 84–90. [CrossRef]


**Sample Availability:** Samples of the Saffron stigmas extract are available from the authors.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Cyanogenic Glycoside Analysis in American Elderberry**

**Michael K. Appenteng <sup>1</sup> , Ritter Krueger <sup>1</sup> , Mitch C. Johnson <sup>1</sup> , Harrison Ingold <sup>1</sup> , Richard Bell <sup>2</sup> , Andrew L. Thomas <sup>3</sup> and C. Michael Greenlief 1,\***


**Abstract:** Cyanogenic glycosides (CNGs) are naturally occurring plant molecules (nitrogenous plant secondary metabolites) which consist of an aglycone and a sugar moiety. Hydrogen cyanide (HCN) is released from these compounds following enzymatic hydrolysis causing potential toxicity issues. The presence of CNGs in American elderberry (AE) fruit, *Sambucus nigra* (subsp. *canadensis*), is uncertain. A sensitive, reproducible and robust LC-MS/MS method was developed and optimized for accurate identification and quantification of the intact glycoside. A complimentary picrate paper test method was modified to determine the total cyanogenic potential (TCP). TCP analysis was performed using a camera-phone and UV-Vis spectrophotometry. A method validation was conducted and the developed methods were successfully applied to the assessment of TCP and quantification of intact CNGs in different tissues of AE samples. Results showed no quantifiable trace of CNGs in commercial AE juice. Levels of CNGs found in various fruit tissues of AE cultivars studied ranged from between 0.12 and 6.38 µg/g. In pressed juice samples, the concentration range measured was 0.29–2.36 µg/mL and in seeds the levels were 0.12–2.38 µg/g. TCP was highest in the stems and green berries. Concentration levels in all tissues were generally low and at a level that poses no threat to consumers of fresh and processed AE products.

**Keywords:** American elderberry; total cyanogenic potential; cyanogenic glycosides; picrate method; solid phase extraction; UHPLC-MS/MS

#### **1. Introduction**

American elderberry (AE), *Sambucus nigra* (subsp. *canadensis*) is a rapidly growing specialty crop in the United States [1]. Native to eastern and midwestern North America, AE is increasingly cultivated for its fruits and flowers that are used in a variety of foods, jellies, syrups, wines, and more importantly, dietary supplement products [2]. Elderberry is known for its nutritional and medicinal health benefits [3–5]. The fruit is rich in carbohydrates, fatty acids, organic acids, minerals, vitamins (A, B6 and C), essential oils, and is high in fiber [6,7]. Researchers have linked elderberry products to anti-inflammatory, antioxidant, anti-carcinogenic, anti-viral, anti-influenza, and antibacterial activities [3,8–13]. Whereas little scientific research has been conducted on AE as compared to its close relative, the European elderberry (EE), *Sambucus nigra* (subsp. *nigra*), both species are excellent sources of flavonoids, polyphenols and anthocyanins [3,8,9,14,15]. The elderberry industry is poised for major expansion and has increased significantly in sales (~0.8 to 107.6 M dollars) between 2011 and 2019 [16]. However, its competitiveness with other herbal dietary supplements [16] may by hampered in part due to uncertainty regarding the presence of cyanogenic glycosides (CNGs) and/or their putative toxicity.

Cyanogenic glycosides are naturally occurring molecules in plants (nitrogenous secondary plant metabolites) which consist of aglycone and a sugar moiety [17,18]. A gen-

**Citation:** Appenteng, M.K.; Krueger, R.; Johnson, M.C.; Ingold, H.; Bell, R.; Thomas, A.L.; Greenlief, C.M. Cyanogenic Glycoside Analysis in American Elderberry. *Molecules* **2021**, *26*, 1384. https://doi.org/10.3390/ molecules26051384

Academic Editor: Wilfried Rozhon

Received: 28 January 2021 Accepted: 1 March 2021 Published: 4 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

eralized structure is shown in Figure 1. There are about 60 CNGs widely distributed in the plant kingdom, occurring in over 2600 plant species representing more than 130 families [19–22]. CNGs are stored in vacuoles within plant cells, separating them from plant hydrolyzing endogenous enzymes (β-1,6-glycosidases and hydroxynitrile lyases) [17,23]. Although intact CNGs are nontoxic, endogenous plant enzymes can react with CNGs, and release hydrogen cyanide (HCN) causing potential toxicity issues [17–19,24,25]. Most CNG-containing plants also produce these endogenous enzymes so when their tissues are disrupted, for example by crushing, herbivory, or disease, CNGs can react with endogenous enzymes resulting in the release of HCN [23,26]. In plants, CNGs serve as important chemical defense compounds against herbivores and pathogens [19,21,27]. Clinical trials have shown mixed results regarding the potential of amygdalin (a CNG found in *Sambucus*) in cancer treatment and as a cough suppressant in various preparations [28,29]. In humans, consumption of cyanogenic plants can cause sub-acute cyanide poisoning (depending on dose) with symptoms including anxiety, headache, vomiting, nausea, abdominal cramps diarrhea, dizziness, weakness and mental confusion. Acute cyanide toxicity in humans (0.5–3.5 mg kg−<sup>1</sup> body weight) [18,19] can result in decreased consciousness, hypotension, paralysis, coma and even death [17–19,24,25,30,31]. Acute cyanide poisoning has been reported from the ingestion of apricot kernels [32], bitter almonds [33], and cassava [34]. in the plant kingdom, occurring in over 2600 plant species representing more than 130 families [19**–**22]. CNGs are stored in vacuoles within plant cells, separating them from plant hydrolyzing endogenous enzymes (β‐1,6‐glycosidases and hydroxynitrile lyases) [17,23]. Although intact CNGs are nontoxic, endogenous plant enzymes can react with CNGs, and release hydrogen cyanide (HCN) causing potential toxicity issues [17**–** 19,24,25]. Most CNG‐containing plants also produce these endogenous enzymes so when their tissues are disrupted, for example by crushing, herbivory, or disease, CNGs can react with endogenous enzymes resulting in the release of HCN [23,26]. In plants, CNGs serve as important chemical defense compounds against herbivores and pathogens [19,21,27]. Clinical trials have shown mixed results regarding the potential of amygdalin (a CNG found in *Sambucus*) in cancer treatment and as a cough suppressant in various preparations [28,29]. In humans, consumption of cyanogenic plants can cause sub‐acute cyanide poisoning (depending on dose) with symptoms including anxiety, headache, vomiting, nausea, abdominal cramps diarrhea, dizziness, weakness and mental confusion. Acute cyanide toxicity in humans (0.5**–**3.5 mg kg−<sup>1</sup> body weight) [18,19] can result in decreased consciousness, hypotension, paralysis, coma and even death [17**–** 19,24,25,30,31]. Acute cyanide poisoning has been reported from the ingestion of apricot kernels [32], bitter almonds [33], and cassava [34].

Cyanogenic glycosides are naturally occurring molecules in plants (nitrogenous secondary plant metabolites) which consist of aglycone and a sugar moiety [17,18]. A generalized structure is shown in Figure 1. There are about 60 CNGs widely distributed

*Molecules* **2021**, *26*, x FOR PEER REVIEW 2 of 18

**Figure 1.** Generic structure for a cyanogenic glycoside, where R1 is often methyl or a proton and R2 is a variable organic group. **Figure 1.** Generic structure for a cyanogenic glycoside, where R<sup>1</sup> is often methyl or a proton and R<sup>2</sup> is a variable organic group.

The only study previously made on *Sambucus canadensis* (American elderberry) is by Buhrmester et al. [20]. They examined the presence or absence of cyanogenic glycosides for individuals from nine populations of *Sambucus canadensis L*. (elderberry) in east‐ central Illinois. The study tested for cyanogenic glycosides in the leaves. Of the nine elderberry populations examined, only one population had a test positive for HCN production each of the three times tested. In another population the production of HCN was highly variable. The cyanogenic glycoside was determined to be (*S*)‐sambunigrin by gas chromatographic separation of the TMS‐derivative. The only study previously made on *Sambucus canadensis* (American elderberry) is by Buhrmester et al. [20]. They examined the presence or absence of cyanogenic glycosides for individuals from nine populations of *Sambucus canadensis* L. (elderberry) in east-central Illinois. The study tested for cyanogenic glycosides in the leaves. Of the nine elderberry populations examined, only one population had a test positive for HCN production each of the three times tested. In another population the production of HCN was highly variable. The cyanogenic glycoside was determined to be (*S*)-sambunigrin by gas chromatographic separation of the TMS-derivative.

A review of the medical literature revealed no reports of elderberry juice poisoning in the past 30 years. The Centers for Disease Control and Prevention [35] did issue a bulletin about a poisoning incident on August 26, 1983 involving a group in California attributed to consumption of juice prepared from fresh wild elderberries along with leaves and stems (most likely blue elderberry, *Sambucus cerulea*) [35]. Cyanide was initially implicated in the incident, but was subsequently disproven. There remains uncertainty as to the presence of CNGs in elderberry juice and its products. Recent studies of European elderberry by Senica et al. [25,36] reported average levels of sambunigrin in fresh and processed berry products ranging between 0.8 and 18.8 μg/g [25] and higher amounts in elder leaves (27.68 and 209.61μg/g FW) [36]. Koss‐Mikolajczyk et al. [37] in similar work A review of the medical literature revealed no reports of elderberry juice poisoning in the past 30 years. The Centers for Disease Control and Prevention [35] did issue a bulletin about a poisoning incident on 26 August 1983 involving a group in California attributed to consumption of juice prepared from fresh wild elderberries along with leaves and stems (most likely blue elderberry, *Sambucus cerulea*) [35]. Cyanide was initially implicated in the incident, but was subsequently disproven. There remains uncertainty as to the presence of CNGs in elderberry juice and its products. Recent studies of European elderberry by Senica et al. [25,36] reported average levels of sambunigrin in fresh and processed berry products ranging between 0.8 and 18.8 µg/g [25] and higher amounts in elder leaves (27.68 and 209.61 µg/g FW) [36]. Koss-Mikolajczyk et al. [37] in similar work however recorded no quantifiable amounts of CNGs. To date, no exhaustive work has been completed on AE to conclusively ascertain the presence, forms, and levels of CNGs in ripe and unripe berries.

Traditional and modern food-processing techniques such as chopping, grinding, and heating are used to reduce the potential toxicity of plants containing CNGs [38–40]. However, the effectiveness of these techniques depends on the processing method [40], the plant tissue, and the intended processed forms. Soaking, for instance, may be effective when CNGs are soluble in the solution (discarded later) without enzymatic degradation [39–41]. Boiling can inhibit the activity of endogenous β-glucosidase due to high temperatures and halt the production of HCN [39]. However, this is only partially effective in reducing HCN because some CNGs are relatively heat stable [39] and HCN is water soluble [42]. Therefore, if CNGs do not hydrolyze due to enzyme inactivation, toxicity may still result from catabolism of these compounds in the gastrointestinal tract [43,44].

Quantification of CNGs can be made either indirectly (by determining HCN released after hydrolysis) or directly (by determining the intact glycoside) [17,19]. A very sensitive, reproducible and robust liquid chromatography/mass spectrometry-based method (LC-MS/MS) was developed and optimized for accurate identification and quantification of intact CNGs. Ultrahigh-performance liquid chromatography triple-quadrupole mass spectrometry (UHPLC-MS/MS) was used for this purpose [45,46]. A complimentary picrate paper method was modified to assess the total cyanogenic potential (TCP) by determining the total cyanide concentration following action of endogenous enzymes with CNGs. Analysis was performed using a camera-phone and UV-Vis spectrophotometry. In this study, we examine different elderberry fruit tissues. This study provides definitive and much needed answers to lingering questions regarding the safety of AE.

#### **2. Results and Discussion**

#### *2.1. Picrate Paper Method*

The TCP was first determined by adapting a picrate paper method originally developed by Bradbury and co-workers [47]. This is a colorimetric method where picrate paper changes color in the presence of HCN. It is based on the reaction of picric acid with HCN. The method is described in more detail in Section 3.5. Amygdalin was used as a CNG standard to generate HCN. Cyanide equivalent (CN− eq.) solutions were prepared from a 1000 µg/mL KCN/NaOH stock solution to develop a calibration curve over the range of 1 to 100 µg/mL. The observed color change of the picrate paper for amygdalin improved significantly and became consistent when the adapted method [47] was modified by using minimal liquid (<0.5 mL water). This was to enhance the HCN reaction with picric acid since HCN is highly water-soluble [42]. A standard calibration curve showing the amount of CN− eq. with its corresponding absorbance is shown in Figure 2, using amygdalin as the cyanide source. Supplemental Figure S1 shows the corresponding standard curve using a camera-phone as a detector. Supplemental Figure S2 shows the expanded UV-Vis data from 0 to 10 µg. Table 1 summarizes the LLOD, upper limit of quantification (ULOQ), and the regression coefficients (*R* 2 -values) for camera-phone and UV-Vis analysis. UV-Vis showed better linearity compared to camera-phone method in repeated analysis.

**Table 1.** LLOD, ULOQ, and Pearson coefficients (*R* 2 -values) for calibration curves from camera-phone and UV-Vis analysis using amygdalin as a CNG standard.


Qualitative inspection of the picrate paper strips showed no visible color change for the commercial elderberry juice sample. UV-Vis analysis of the picrate paper test strips detected no quantifiable amount of cyanide (<0.14 µg CN− eq.). Two different AE genotypes, Ozark and Ozone, were then analyzed using the picrate paper method. Sample tissues (juice, skin, stem, seeds) for each genotype showed no visible color change on qualitative assessment (Supplemental Figure S3). Generally, results obtained for lyophilized

samples were comparable to fresh samples. Quantitative determination by UV-Vis revealed low levels of cyanide with average amounts ranging from (2.60–9.20 µg CN− eq.)/g of sample. TCP levels obtained were comparable for both AE genotypes and for all tissue types of Ozone and Ozark, respectively. TCP amounts increased in the order juice < seeds < skin < stem for both genotypes as shown in Figure 3 [48]. *Molecules* **2021**, *26*, x FOR PEER REVIEW 4 of 18

**Figure 3.** Total cyanogenic potential for different types of tissue of Ozone and Ozark AE genotypes. The amounts of CNGs in these genotypes were determined using UV‐Vis spectrophotometry. The error bars represent the standard deviation of at least three replicate **Figure 3.** Total cyanogenic potential for different types of tissue of Ozone and Ozark AE genotypes. The amounts of CNGs in these genotypes were determined using UV-Vis spectrophotometry. The error bars represent the standard deviation of at least three replicate samples.

A set of pooled AE samples was generated using five AE genotypes (Ozark, York,

ripe berries. Qualitative inspection of picrate paper test strips for pooled AE samples showed a visible faint color change for the green berries and stems (Supplemental Figure S4). UV‐Vis analysis showed the highest CN‐ levels for stems and green berries with lower amounts for the other tissue types (Figure 4). TCP levels in analyzed tissues increased in the order: whole ripe berries < whole red berries < juice < seeds < skin < pulp < whole green berries < stem, with highest average amounts in the stems (37.43 ± 9.19 μg CN‐ eq./g) and whole green berries (25.6 ± 5.07μg CN‐ eq./g). Koss‐Mikolajczyk et al. [37] in a recent EE study observed a weak and unstable signal for a peak corresponding to sambunigrin which decreased with advancing stage of ripeness in elderberry fruit. In another study, Zahmanov et al. [49] reported metabolic differences in mature and immature fruits, and plant leaves of *Sambucus ebulus*. These observations may account for the slightly higher levels recorded in green berries. Although the CNG amounts in the stems and green berries are not sufficient to pose a threat of toxicity, it is nevertheless advisable to carefully

exclude green elderberries and stems during juice preparation.

samples.

A set of pooled AE samples was generated using five AE genotypes (Ozark, York, Wyldewood, Ocoee, and Bob Gordon). The pooled samples were divided into different types of tissue. These included seeds, skin, pulp, stems, juice, and whole green, red, and ripe berries. Qualitative inspection of picrate paper test strips for pooled AE samples showed a visible faint color change for the green berries and stems (Supplemental Figure S4). UV-Vis analysis showed the highest CN− levels for stems and green berries with lower amounts for the other tissue types (Figure 4). TCP levels in analyzed tissues increased in the order: whole ripe berries < whole red berries < juice < seeds < skin < pulp < whole green berries < stem, with highest average amounts in the stems (37.43 ± 9.19 µg CN<sup>−</sup> eq./g) and whole green berries (25.6 ± 5.07 µg CN<sup>−</sup> eq./g). Koss-Mikolajczyk et al. [37] in a recent EE study observed a weak and unstable signal for a peak corresponding to sambunigrin which decreased with advancing stage of ripeness in elderberry fruit. In another study, Zahmanov et al. [49] reported metabolic differences in mature and immature fruits, and plant leaves of *Sambucus ebulus*. These observations may account for the slightly higher levels recorded in green berries. Although the CNG amounts in the stems and green berries are not sufficient to pose a threat of toxicity, it is nevertheless advisable to carefully exclude green elderberries and stems during juice preparation. *Molecules* **2021**, *26*, x FOR PEER REVIEW 6 of 18

Pooled AE samples

**Figure 4.** Total cyanogenic potential for different types of AE tissue of pooled samples made up of five different genotypes. The amounts of CNGs in pooled samples were determined using UV‐Vis **Figure 4.** Total cyanogenic potential for different types of AE tissue of pooled samples made up of five different genotypes. The amounts of CNGs in pooled samples were determined using UV-Vis spectrophotometry. The error bars represent the standard deviation of at least three replicate samples.

spectrophotometry. The error bars represent the standard deviation of at least three replicate samples. Two different types of seeds from Gala and Granny Smith apples were obtained and prepared for analysis as discussed in Section 3.5. Apple seeds were chosen as their TCP levels are known and should be readable using the picrate paper method. Color change on the picrate paper test strip for the apple seeds occurred swiftly at room temperatures even before test strips were transferred into the oven for overnight heating (30**–**40 °C). A deep red color change was observed on inspection for both fresh and lyophilized samples Two different types of seeds from Gala and Granny Smith apples were obtained and prepared for analysis as discussed in Section 3.5. Apple seeds were chosen as their TCP levels are known and should be readable using the picrate paper method. Color change on the picrate paper test strip for the apple seeds occurred swiftly at room temperatures even before test strips were transferred into the oven for overnight heating (30–40 ◦C). A deep red color change was observed on inspection for both fresh and lyophilized samples (Supplemental Figure S5). UV-Vis analysis showed high average cyanide amounts (TCP) ranging from (497.50–603.20 µg CN− eq.)/g of apple seeds. TCP levels in analyzed seeds

(Supplemental Figure S5). UV‐Vis analysis showed high average cyanide amounts (TCP) ranging from (497.50**–**603.20 μg CN<sup>−</sup> eq.)/g of apple seeds. TCP levels in analyzed seeds were higher in Granny Smith as compared to Gala apple varieties. These results were

Results from the endogenous enzymes test made using pooled AE stems and green berries revealed higher cyanide levels in samples with added β‐glucosidase than those without added β‐glucosidase (Supplemental Figure S6). Approximately 77% and 33% more cyanide were measured in pooled AE stems and whole green berries, respectively, with added β‐glucosidase. These findings indicated that while AE contains endogenous β‐glucosidase enzymes sufficient to initiate self‐hydrolysis of CNGs, it may not be sufficient for complete hydrolysis of all CNGs (55**–**75%) when the tissues are disrupted. This implies that not all available CNGs in elderberry may necessarily be able to transform into HCN. These observations are supported in a similar analysis by Miller et al. [51] using foliage of the tropical trees *Beilschmiedia collina* and *Mischocarpus spp*. Apple seeds however showed no appreciable change in cyanide concentration with or without addition of β‐glucosidase enzymes (Supplemental Figure S7), thus indicating that the seeds of the apple varieties used contain sufficient endogenous β‐glucosidase for complete hydrolysis of all CNGs when the tissues are disrupted. The picrate paper test method is quick and simple and could serve as an effective field test for elderberry producers.

were higher in Granny Smith as compared to Gala apple varieties. These results were comparable to available literature [17,50].

Results from the endogenous enzymes test made using pooled AE stems and green berries revealed higher cyanide levels in samples with added β-glucosidase than those without added β-glucosidase (Supplemental Figure S6). Approximately 77% and 33% more cyanide were measured in pooled AE stems and whole green berries, respectively, with added β-glucosidase. These findings indicated that while AE contains endogenous β-glucosidase enzymes sufficient to initiate self-hydrolysis of CNGs, it may not be sufficient for complete hydrolysis of all CNGs (55–75%) when the tissues are disrupted. This implies that not all available CNGs in elderberry may necessarily be able to transform into HCN. These observations are supported in a similar analysis by Miller et al. [51] using foliage of the tropical trees *Beilschmiedia collina* and *Mischocarpus spp*. Apple seeds however showed no appreciable change in cyanide concentration with or without addition of β-glucosidase enzymes (Supplemental Figure S7), thus indicating that the seeds of the apple varieties used contain sufficient endogenous β-glucosidase for complete hydrolysis of all CNGs when the tissues are disrupted. The picrate paper test method is quick and simple and could serve as an effective field test for elderberry producers. *Molecules* **2021**, *26*, x FOR PEER REVIEW 7 of 18

#### *2.2. UHPLC MS/MS Method of Analysis* 2.2.1. Method Development and Optimization

*2.2. UHPLC MS/MS Method of Analysis*

precursor ion).

#### 2.2.1. Method Development and Optimization An attempt was made to find multiple reaction monitoring (MRM) transitions for

An attempt was made to find multiple reaction monitoring (MRM) transitions for four cyanogenic standards (CNS) using both electrospray ionization (ESI) and atmospheric pressure chemical ionization (APCI) sources. Positive and negative ionization modes were performed for each standard with both ionization techniques. The only successful MRM transition identified was for amygdalin in ESI positive mode. Figure 5 shows the positive mode product ion (296 *m*/*z*, product) spectrum for amygdalin (465 *m*/*z*, precursor). All other standards readily formed sodium adducts, which did not sufficiently fragment due to their high stability. Alternative mass spectrometry scans were investigated to overcome this problem. Quantification for all four CNS's were performed using selected ion recording (SIR) mode. The developed UHPLC and MS method displayed excellent separation of the four standards and exhibited retention time repeatability and good peak shape. A chromatogram for the separation with retention times (RT) and scanning modes is shown in Figure 6. It took less than 6 min to separate and elute all 4 CNS. four cyanogenic standards (CNS) using both electrospray ionization (ESI) and atmospheric pressure chemical ionization (APCI) sources. Positive and negative ionization modes were performed for each standard with both ionization techniques. The only successful MRM transition identified was for amygdalin in ESI positive mode. Figure 5 shows the positive mode product ion (296 *m*/*z*, product) spectrum for amygdalin (465 *m*/*z*, precursor). All other standards readily formed sodium adducts, which did not sufficiently fragment due to their high stability. Alternative mass spectrometry scans were investigated to overcome this problem. Quantification for all four CNS's were performed using selected ion recording (SIR) mode. The developed UHPLC and MS method displayed excellent separation of the four standards and exhibited retention time repeatability and good peak shape. A chromatogram for the separation with retention times (RT) and scanning modes is shown in Figure 6. It took less than 6 min to separate and elute all 4 CNS.

**Figure 5.** ESI positive mode product ion (296 *m*/*z*, product ion) spectrum for amygdalin (465 *m*/*z*, **Figure 5.** ESI positive mode product ion (296 *m*/*z*, product ion) spectrum for amygdalin (465 *m*/*z*, precursor ion).

curve beyond which the linearity breaks. Details are summarized in Table 2.

of variation [52] (CV% ≤ 20%, for seven repeated injections) for confirmation. A linear range with lower (LLOQ, S/N =10, CV% ≤ 20%) and upper (ULOQ) limit of quantification was determined. The ULOQ was determined as the highest concentration of the linear

Standard calibration curves showed good linear correlations (*R*<sup>2</sup> values) between

**Figure 6.** Ion chromatograms for (**A**) amygdalin (MRM), (**B**) amygdalin, (**C**) dhurrin, (**D**) prunasin, and (**E**) linamarin (SIR). Retention times in min. are: 4.61, 4.61, 2.54, 5.37, and 1.18, respectively. **Figure 6.** Ion chromatograms for (**A**) amygdalin (MRM), (**B**) amygdalin, (**C**) dhurrin, (**D**) prunasin, and (**E**) linamarin (SIR). Retention times in min. are: 4.61, 4.61, 2.54, 5.37, and 1.18, respectively.

> **Table 2.** Summary of Pearson coefficient, detection and quantification limits (ng/mL) for CNGs. **Parameters ng/mL MRM Amygdalin SIR Amygdalin Dhurrin Prunasin Linamari n LLOD** 0.3 3 3 3 1 **LLOQ** 1 10 10 5 5 Standard calibration curves showed good linear correlations (*R* <sup>2</sup> values) between integrated peak areas and known CNS concentrations. The lower limit of detection (LLOD) was determined based on a signal to noise ratio of three and a targeted coefficient of variation [52] (CV% ≤ 20%, for seven repeated injections) for confirmation. A linear range with lower (LLOQ, S/N =10, CV% ≤ 20%) and upper (ULOQ) limit of quantification was determined. The ULOQ was determined as the highest concentration of the linear curve beyond which the linearity breaks. Details are summarized in Table 2.


**R2** 0.9998 0.9998 0.9983 0.9984 0.9910 **Table 2.** Summary of Pearson coefficient, detection and quantification limits (ng/mL) for CNGs.

**ULOQ** 8000 8000 6000 6000 2000

(ME) were evaluated by an approach based on responses from pre‐extraction spike matrix 2.2.2. Optimized Extraction, Recovery and Matrix Effect

(a), post‐extraction spike matrix (b) and a neat spike standard (c). RE and ME were calculated using Equations (1) and (2) [53] (where +ME implies ion enhancement, −ME implies ion suppression). Table 3 below compares the mean recoveries and standard deviations for 30 min sonication at 30°C to overnight shaking (16**–**24 h) at room temperature with intended spike concentration of 1000 ng/mL and 100 ng/mL CNS mixture. ME estimation was found to range between 10.20 and 18.34%. This was a negative estimation and as such indicated some degree of ion suppression. Although further dilution of sample matrix from 10 to 1000‐fold reduced this value appreciably, it also decreased the sensitivity of sample detection. Hence a 10‐fold dilution was used. ൌ a <sup>b</sup> ൈ 100 (1) Selecting the most appropriate extraction solvent was key to development of the extraction methodology. Recoveries from aqueous ethanol or methanol combinations were evaluated. Higher recoveries were obtained with 75% methanol extraction with overnight shaking (16–24 h) at room temperature and 30 min sonication at 30 ◦C as compared to other extraction methods and conditions. The recovery (RE) and matrix effect (ME) were evaluated by an approach based on responses from pre-extraction spike matrix (a), postextraction spike matrix (b) and a neat spike standard (c). RE and ME were calculated using Equations (1) and (2) [53] (where +ME implies ion enhancement, −ME implies ion suppression). Table 3 below compares the mean recoveries and standard deviations for 30 min sonication at 30 ◦C to overnight shaking (16–24 h) at room temperature with intended spike concentration of 1000 ng/mL and 100 ng/mL CNS mixture. ME estimation

was found to range between 10.20 and 18.34%. This was a negative estimation and as such indicated some degree of ion suppression. Although further dilution of sample matrix from 10 to 1000-fold reduced this value appreciably, it also decreased the sensitivity of sample detection. Hence a 10-fold dilution was used.

$$\mathbf{RE} = \frac{\mathbf{a}}{\mathbf{b}} \times 100\tag{1}$$

$$\mathbf{ME} = \left[ \left( \frac{\mathbf{b}}{\mathbf{c}} \right) - 1 \right] \times 100 \tag{2}$$


*Molecules* **2021**, *26*, x FOR PEER REVIEW 9 of 18


#### 2.2.3. Optimized SPE Method 2.2.3. Optimized SPE Method

To assess and evaluate the elution solvent strength in the SPE method, an elution profile showing methanol and water proportions from 0:100 to 100:0 (*v*/*v*) versus peak area was made for all four standards. Figure 7 shows the elution profile for amygdalin. Evaluation of these profile diagrams revealed 30–40% methanol as the best solvent strength for elution of all four CNS. Confirmation using water methanol proportions from 0:40 to 40:0 (*v*/*v*) was made to determine the best elution solvent as 35% methanol. To assess and evaluate the elution solvent strength in the SPE method, an elution profile showing methanol and water proportions from 0:100 to 100:0 (*v*/*v*) versus peak area was made for all four standards. Figure 7 shows the elution profile for amygdalin. Evaluation of these profile diagrams revealed 30**–**40% methanol as the best solvent strength for elution of all four CNS. Confirmation using water methanol proportions from 0:40 to 40:0 (*v*/*v*) was made to determine the best elution solvent as 35% methanol.

**Figure 7.** Elution profile for amygdalin as a function of methanol content in the extraction solvent. (AUC is the area under the curve). **Figure 7.** Elution profile for amygdalin as a function of methanol content in the extraction solvent. (AUC is the area under the curve).

#### 2.2.4. Sample Test 2.2.4. Sample Test

The developed UHPLC‐MS/MS method, as detailed in Section 3.6, was used to determine the levels of intact CNGs in different AE samples. Analysis of commercial elderberry juice showed no quantifiable amounts of CNGs. However, extracts of Ozark The developed UHPLC-MS/MS method, as detailed in Section 3.6, was used to determine the levels of intact CNGs in different AE samples. Analysis of commercial elderberry juice showed no quantifiable amounts of CNGs. However, extracts of Ozark

and Ozone elderberry tissues (seeds, juice, skin and stem) for both lyophilized and fresh

amounts (μg/g) in tissues were generally higher in Ozone compared to Ozark. Higher levels (μg/g) were recorded in the stems and skin tissues as compared to levels in the seeds and juice, respectively for the Ozone and Ozark samples. A detailed summary of amounts for each detected CNGs in AE tissues are shown in Table 4. Figures 8 and 9 show the

and Ozone elderberry tissues (seeds, juice, skin and stem) for both lyophilized and fresh samples showed low traces of CNGs (amygdalin, dhurrin, linamarin and prunasin). Levels of CNGs detected in lyophilized samples were comparable to fresh samples. The amounts (µg/g) in tissues were generally higher in Ozone compared to Ozark. Higher levels (µg/g) were recorded in the stems and skin tissues as compared to levels in the seeds and juice, respectively for the Ozone and Ozark samples. A detailed summary of amounts for each detected CNGs in AE tissues are shown in Table 4. Figures 8 and 9 show the amounts of these cyanogens in tissues for Ozone and Ozark, respectively. The levels (µg/g) of CNGs in analyzed tissues increased in the order: linamarin < dhurrin < prunasin < amygdalin, respectively, for Ozone and Ozark samples tissues. In contrast to this trend, prunasin levels were highest in the juice and stems of Ozark AE. *Molecules* **2021**, *26*, x FOR PEER REVIEW 10 of 18 amounts of these cyanogens in tissues for Ozone and Ozark, respectively. The levels (μg/g) of CNGs in analyzed tissues increased in the order: linamarin < dhurrin < prunasin < amygdalin, respectively, for Ozone and Ozark samples tissues. In contrast to this trend, prunasin levels were highest in the juice and stems of Ozark AE.


**Table 4.** Amounts (µg/g) of CNGs found in tissues of Ozone and Ozark AE samples. **Table 4.** Amounts (μg/g) of CNGs found in tissues of Ozone and Ozark AE samples.

**Figure 8.** Amounts of CNGs (μg/g) in tissues (seeds, juice, skin and stem) of Ozone elderberry samples as measured by UHPLC‐MS/MS. **Figure 8.** Amounts of CNGs (µg/g) in tissues (seeds, juice, skin and stem) of Ozone elderberry samples as measured by UHPLC-MS/MS.

2.5 3 3.5 4 µg/g) In our UHPLC MS/MS, we are not able to distinguish between the two diastereomers, (*R*)-prunasin and (*S*)-sambunigrin. The two compounds were not uniquely separated by UHPLC using a C18 column. Further, their fragmentation patterns are very similar. Therefore, the prunasin concentrations should be viewed as a sum of the prunasin and sambunigrin concentrations.

0 0.5 1 1.5 2 CNG amount (A review of literature in similar areas of study found comparable results, but also revealed an interesting trend of observation. A recent study by Senica et al. [40] on the EE (subsp. *nigra*) reported average levels of sambunigrin (µg/g) in fresh berries (18.8 ± 4.3), processed juice (10.6 ± 0.7), tea (3.8 ± 1.7), spread (0.8 ± 0.19) and liqueur (0.8 ± 0.21). Our measured levels of (prunasin + sambunigrin) for AE are lower for fresh berries compared to EE. Senica et al. [36] in a similar work reported highest amounts of sambunigrin in elder leaves (27.68–209.61 µg/g FW), lower amounts in flowers (1.23–18.88 µg/g FW) and

**Figure 9.** Amounts of CNGs (μg/g) in tissues (seeds, juice, skin and stem) of Ozark elderberry

Seeds Juice Skin Stem

CNG's

samples as measured by UHPLC‐MS/MS.

6

µg/g)

8

lowest amounts in berries (0.08–0.77 µg/g FW). In the work by Buhrmester et al. [20], also observed similar sambunigrin concentrations for AE leaves. Senica and co-workers concluded that the content of sambunigrin in elderberry changes depending on the growing altitudes (higher content on hill tops and lower in foothills) [36]. Another study by Koss-Mikolajczyk et al. [37] on EE (subsp. *nigra*) observed the highest signal for a peak detected as sambunigrin in the elder leaves although this peak became undetectable after one day of cold storage of extracts. It was also reported that the level of cyanogens in cassava leaves are 10 times more than in the roots [39]. Deductions from this trend of results suggest that the leaves of most cyanogenic plants may accumulate larger amounts of CNGs, with elder as no exception. The trend of results also corroborates the fact that elderberry juice, being it, AE or EE showed very low levels of CNGs. **Figure 8.** Amounts of CNGs (μg/g) in tissues (seeds, juice, skin and stem) of Ozone elderberry samples as measured by UHPLC‐MS/MS. 0 2 4 Seeds Juice Skin Stem CNG amount (CNG's Amygdalin Dhurrin Prunasin Linamarin

*Molecules* **2021**, *26*, x FOR PEER REVIEW 10 of 18

prunasin levels were highest in the juice and stems of Ozark AE.

**Table 4.** Amounts (μg/g) of CNGs found in tissues of Ozone and Ozark AE samples.

**Elderberry Samples Concentration <sup>±</sup> Standard Deviation (μg/g)**

amounts of these cyanogens in tissues for Ozone and Ozark, respectively. The levels (μg/g) of CNGs in analyzed tissues increased in the order: linamarin < dhurrin < prunasin < amygdalin, respectively, for Ozone and Ozark samples tissues. In contrast to this trend,

**Seeds** Ozone 2.38 ± 0.09 0.27 ± 0.05 0.58 ± 0.04 0.12 ± 0.06

**Juice** Ozone 1.57 ± 0.08 0.70 ± 0.12 1.45 ± 0.06 0.29 ± 0.03

**Skin** Ozone 6.38 ± 0.40 0.12 ± 0.08 2.39 ± 0.04 0.75 ± 0.06

**Stem** Ozone 5.42 ± 0.12 0.94 ± 0.06 2.84 ± 0.02 0.48 ± 0.04

Ozark 0.68 ± 0.12 0.22 ± 0.03 0.36 ± 0.05 0.13 ± 0.05

Ozark 0.36 ± 0.03 0.63 ± 0.04 2.36 ± 0.08 0.31 ± 0.01

Ozark 3.48 ± 0.14 1.46 ± 0.20 2.53 ± 0.08 0.90 ± 0.11

Ozark 2.15 ± 0.17 1.91 ± 0.03 3.07 ± 0.06 0.57 ± 0.06

**Amygdalin Dhurrin Prunasin Linamarin**

**Figure 9.** Amounts of CNGs (μg/g) in tissues (seeds, juice, skin and stem) of Ozark elderberry samples as measured by UHPLC‐MS/MS. **Figure 9.** Amounts of CNGs (µg/g) in tissues (seeds, juice, skin and stem) of Ozark elderberry samples as measured by UHPLC-MS/MS.

The levels of CNGs detected in all tissues of AE samples were extremely low compared to levels of amygdalin detected in apple seeds (950–3910) µg/g, pressed apple juice (10–39) µg/mL and commercially available apple juice (1–7) µg/mL for 15 apple varieties [17]. Acute CN toxicity occurs at a concentration of 0.5–3.5 mg/kg of body weight. For cyanide in blood, the toxicity threshold for cyanide alone ranges from 0.5 to 1.0 mg/L, and the lethal threshold ranges from 2.5 to 3.0 mg/L [54]. Despite the high cyanide levels in apple seeds as revealed in the control picrate test (497.50–603.20 µg CN− eq./g), signs of cyanide toxicity may occur in an average adult male of weight 82 kg, only after consuming about 14 or more apples including mastication of all associated seeds. This estimation was made considering the threshold value of 0.50 mg/kg body weight for cyanide toxicity, average TCP per seed of 550 µg/g (or 0.55 mg/g), the average weight of an apple seed (0.75 g), the average number of seeds per apple (7 or 8), and assuming maximum enzyme activity.

Different processing techniques such as chopping, grinding, soaking, fermentation, drying, roasting, boiling, and steaming have been used to remove or reduce the potential toxicity of cyanogens in plants [38–40]. The effectiveness of these processes is dependent on the specific processing method [40], the plant tissues and the intended processed forms. Boiling of juice for instance may have a different effect compared with boiling or soaking cassava chips where the associated water can easily be discarded [39–41]. This may be due to enzyme inactivation and solubilization of CNGs in discarded water [39]. A study by Montagnac et al. further indicated that the effectiveness of these techniques depends on the processing steps, the sequence utilized, and is often time-dependent [39]. They proposed that to increase the efficiency of cyanogen removal from cassava, efficient processing techniques should be combined [39]. For example, soaking, fermenting and roasting removes about 98% of cyanogens [39]. A recent study by Senica et al. [40] also

showed that thermal processing, time and type of extraction solution greatly affected phenolics and cyanogenic glycosides in different elderberry products. They showed that higher processing temperatures decreased the levels of cyanogenic glycosides by 44% in elderberry juice, 80% in tea and as much as 96% in elderberry liqueur and spread [40]. It has been confirmed that pasteurization effectively decreases the levels of harmful compounds, such as cyanogenic glycosides [17,40]. Bolarinwa et al. [17] moreover observed that holding apple juice at room temperature for 120 min either before or after pasteurizing decreased the amygdalin content by about 19% compared to the original juice. These methods are very effective and can be applied to remove or further reduce the levels of CNGs in elderberry product. It is however important to establish that the types and levels of CNGs observed in AE are very low and pose no threat to consumers in the use of fresh or processed AE products.

#### **3. Materials and Methods**

#### *3.1. Chemicals and Reagents*

Water, acetonitrile (ACN), methanol, ethanol and formic acid were purchased from Fisher Scientific (Fair Lawn, NJ, USA, HPLC grade). β-glucosidase enzymes (250 mg lyophilized powder ≥ 6U/mg), amygdalin (1 g, ≥99%), dhurrin (1 mg, ≥98%), prunasin (1 mg, ≥95%) and linamarin standards (1 mg, ≥95%), together with picric acid (100 g moisten with water ≥ 98%) and potassium cyanide (≥98%) were purchased from Sigma Aldrich Chemical Co. (St. Louis, MO, USA). Whatman no.1 filter paper, sodium carbonate, sodium hydroxide and pH 8 phosphate buffer (500 mL) were also purchased from Fisher Scientific (Fair Lawn, NJ, USA). Plastic backing and hobby glue (adhesive neutral pH) were purchased from the Mizzou Store (Columbia, MO, USA).

#### *3.2. American Elderberry Samples*

Plant Material. Elderberry fruit samples were harvested from experimental field plots that were previously described in detail [2,45]. Briefly, a replicated evaluation of eight American elderberry genotypes was established in Missouri (USA) in 2008. Fruits from six genotypes (Bob Gordon, Ocoee, Ozark, Ozone, Wyldewood, and York) were harvested from one of the study sites (Mt. Vernon, MO, USA) at peak ripeness in August 2016, and promptly frozen. Frozen, de-stemmed, whole berries (>400 g) from the five genotypes were provided to the laboratory, along with frozen unripe and almost-ripe berries (green and red-colored, respectively). Additionally, hundreds of individual berries from each genotype were thawed and painstakingly separated into skins (epicarp), pulp (mesocarp), seeds, juice, and small green stems (pedicels) that connect the berry to the infructescence. After dissection, samples were re-frozen. For detailed CNG analysis, tissue and juice samples from the genotypes Ozone and Ozark were analyzed separately. Further, tissue and juice from five genotypes (Bob Gordon, Ocoee, Ozark, Wyldewood, and York) were combined into pooled samples for a broader analysis. Sufficient material was dissected to produce samples exceeding 10 mg.

Commercially processed elderberry juice was purchased from River Hills Harvest, Hartsburg, MO, USA.

#### *3.3. Sample Preparation and Extraction*

Berries were transferred into small zippered plastic bags, thawed, and gently pressed. The juice was transferred to a clean vial. Seeds were separated from skin and placed into different vials. 100 g of berries produced about 60 g of juice, 20 g of seeds, 12 g of skin, and some left-over stems. Between 5–10 g of each sample tissue (excluding juice) was transferred into 15 mL Eppendorf tubes, flash frozen for about 5 min in liquid nitrogen and freeze-dried for 24 h using a Labconco FreeZone 4.5 Liter Benchtop Lyophilizer (Labconco Corp., Kansas City, MO, USA) at −105 ◦C. The lyophilized samples were ground using a clean mortar and pestle to obtain about 3–5 g of homogenized seeds, stem and skin samples for extraction.

To obtain an optimized sample pretreatment and extraction, equal volumes of commercially processed elderberry juice, in replicates of 4, were spiked with varying amounts of 10 µg/mL cyanogenic standard (CNS) stock mixture (amygdalin, dhurrin, prunasin and linamarin, Figure 10) to obtain intended spike concentrations of 1000, 100, and 10 ng/mL. The solutions were extracted with 1 mL of different ethanol/methanol and water proportions from 60:40 to 80:20 (*v*/*v*). Extraction was performed via sonication (10–60 min) at 30 ◦C, overnight shaking (16–24 h) and 2 min vortexing at room temperature on a Genie 2 Shaker Mixer (Scientific Industries, Inc., Bohemia, NY, USA) at 600 rpm. Extracts were centrifuged for 15 min, dried under nitrogen gas and reconstituted in 1 mL of mobile phase (0.1% FA in ACN) for SPE clean up. Sample extraction was performed with both fresh and lyophilized samples. *Molecules* **2021**, *26*, x FOR PEER REVIEW 13 of 18 water proportions from 60:40 to 80:20 (*v*/*v*). Extraction was performed via sonication (10**–** 60 min) at 30°C, overnight shaking (16**–**24 h) and 2 min vortexing at room temperature on a Genie 2 Shaker Mixer (Scientific Industries, Inc., Bohemia, NY, USA) at 600 rpm. Extracts were centrifuged for 15 min, dried under nitrogen gas and reconstituted in 1 mL of mobile phase (0.1% FA in ACN) for SPE clean up. Sample extraction was performed with both

**Figure 10.** Cyanogenic glycosides standards used in this study: amygdalin, dhurrin, prunasin, and linamarin. **Figure 10.** Cyanogenic glycosides standards used in this study: amygdalin, dhurrin, prunasin, and linamarin.

#### *3.4. Solid Phase Extraction (SPE) 3.4. Solid Phase Extraction (SPE)*

fresh and lyophilized samples.

An SPE method [24] previously used for the determination of amygdalin in almonds was adapted and optimized for our use. A Supelco Visiprep™ SPE vacuum manifold (Sigma‐Aldrich, St. Louis, MO, USA) was used for this purpose. A vacuum of 10 in Hg (35 kPa) and a flowrate of about 1**–**2 drops/s was maintained throughout the process. A SPE cartridge (Sep‐Pak Vac 3 cc (500 mg) C18 cartridge, (Waters, Milford, MA, USA) was conditioned with 2 mL of methanol and equilibrated with 2 mL of water. 1 mL of the sample was loaded onto the column. An additional 1 mL of 0.1% FA in water was used to remove remaining residue in the extraction tube. The column was flushed with 2 mL of 0.1% FA in water. CNGs were finally eluted with 2 mL of varying proportions of methanol/water (0, 10, 20, 30 to 100% *v*/*v*). The extracts were dried under nitrogen gas, reconstituted into 0.1% FA in water and filtered through a 0.45 μm filter prior to UHPLC‐ An SPE method [24] previously used for the determination of amygdalin in almonds was adapted and optimized for our use. A Supelco Visiprep™ SPE vacuum manifold (Sigma-Aldrich, St. Louis, MO, USA) was used for this purpose. A vacuum of 10 in Hg (35 kPa) and a flowrate of about 1–2 drops/s was maintained throughout the process. A SPE cartridge (Sep-Pak Vac 3 cc (500 mg) C18 cartridge, Waters, Milford, MA, USA) was conditioned with 2 mL of methanol and equilibrated with 2 mL of water. 1 mL of the sample was loaded onto the column. An additional 1 mL of 0.1% FA in water was used to remove remaining residue in the extraction tube. The column was flushed with 2 mL of 0.1% FA in water. CNGs were finally eluted with 2 mL of varying proportions of methanol/water (0, 10, 20, 30 to 100% *v*/*v*). The extracts were dried under nitrogen gas, reconstituted into 0.1% FA in water and filtered through a 0.45 µm filter prior to UHPLC-MS/MS analysis.

#### *3.5. Picrate Paper Method of Analysis*

MS/MS analysis.

*3.5. Picrate Paper Method of Analysis* The picrate paper method is based on the reaction of enzymes that catalyze the release of HCN gas, which reacts with picric acid on a paper test strip producing 2,6‐ dinitro‐5‐hydroxy‐4‐hydroxylamino‐1,3 dicyclobenzene inducing a color change The picrate paper method is based on the reaction of enzymes that catalyze the release of HCN gas, which reacts with picric acid on a paper test strip producing 2,6-dinitro-5-hydroxy-4-hydroxylamino-1,3 dicyclobenzene inducing a color change (Supplemental Figure S8) [46].

(Supplemental Figure S8) [46]. A previously published picrate method described by Bradbury et al. [47] for the determination of the total cyanogenic content in cassava roots was adapted and modified for use. Briefly, the picrate paper was prepared beforehand by dipping a sheet of Whatman 3 MM filter paper in a picrate solution (1.4% *w*/*v* moist picric acid diluted in 2.5% *w*/*v* Na2CO3 solution), allowing the paper to air dry and cutting it into 3 cm × 1 cm A previously published picrate method described by Bradbury et al. [47] for the determination of the total cyanogenic content in cassava roots was adapted and modified for use. Briefly, the picrate paper was prepared beforehand by dipping a sheet of Whatman 3 MM filter paper in a picrate solution (1.4% *w*/*v* moist picric acid diluted in 2.5% *w*/*v* Na2CO<sup>3</sup> solution), allowing the paper to air dry and cutting it into 3 cm × 1 cm strips. The strips were attached using a drop of PVA hobby glue to 5 cm × 1.2 cm clear plastic

strips. The strips were attached using a drop of PVA hobby glue to 5 cm × 1.2 cm clear plastic strips to keep the paper clear of the liquid. They were stored at 4 °C prior to use.

solution. The stock solution was prepared by dissolving KCN in 0.1M NaOH as the solvent. The calibration curve covered the range of 1 to 100 μg CN<sup>−</sup> eq. and the method was verified using amygdalin as a positive control. One of the most complicated portions

strips to keep the paper clear of the liquid. They were stored at 4 ◦C prior to use. Cyanide equivalent (CN− eq.) solutions were prepared from a 1000 µg/mL KCN stock solution. The stock solution was prepared by dissolving KCN in 0.1M NaOH as the solvent. The calibration curve covered the range of 1 to 100 µg CN− eq. and the method was verified using amygdalin as a positive control. One of the most complicated portions of this analysis is the enzymatic hydrolysis of amygdalin. Enzymes are macromolecular biological catalysts whose amount for a specific enzymatic activity is measured in Units (U). One U is defined as the amount of enzyme needed to catalyze the conversion of 1 micromole of substrate per minute [48]. Enzymatic degradation of amygdalin was achieved by adding 50 µL of 3U/mL β-glucosidase.

The commercially processed elderberry juice was tested for TCP along with the AE samples. 100 µL/100 mg each of lyophilized and fresh tissues of Ozone, Ozark and pooled AE samples were measured/weighed into clean 20 mL scintillation vials. 50 µL of βglucosidase solution (3U/mL) in pH 8 phosphate buffer was added alongside the picrate paper and the vial was immediately closed with a screw stopper. Each sample analysis was made in replicates of four. Similar set-ups were made for amygdalin standards and blank (no amygdalin added). These were left overnight (16–24 h) in an oven at 30–40 ◦C.

Two control experiments were performed. The first was to confirm the effectiveness of the picrate paper test method to known literature. Seeds from two apple varieties, Granny Smith (GS) and Gala (G), were prepared and tested for TCP using the same protocol for the AE samples. The second control experiment used seeds (ground) from the two apple varieties and stems and green berries from AE pooled samples. This was to test for the presence of endogenous enzymes in the samples to assess the extent of self-hydrolysis of CNGs. To accomplish this, two different picrate paper set-ups were made, a control (without an external β-glucosidase solution) and a second typical picrate set-up (with an external β-glucosidase solution). Four replicates for each set-up were performed.

A simple and quick method of analyzing the reacted picric acid is by qualitative inspection. This appears to be an effective method for quantifying CN− equivalents ranging between 1 and 100 µg. However, as the amount of CN− eq. increases, the ability to differentiate between the colors decreases. A color chart shown in Figure 11 can be used to relate the color change of the paper to total amount of CN− evolved. A semi-quantitative approach using a camera-phone as a detector was used. An image of a concentration from Figure 11 was converted from color to greyscale. This was done using Image J software (https://imagej.nih.gov/ij/index.html (accessed on 22 December 2020), version 1.46r, National Institutes of Health, Bethesda, MD, USA). The method generated mean intensity values corresponding to each CN− eq. and was used to generate a calibration curve. Quantification was confirmed using a UV-Vis spectrometer (Agilent 8454 photodiode array, Agilent Technologies, Santa Clara, CA, USA). The reacted picrate paper test strip was extracted in 3.5 mL of water in cuvettes and the resulting solution analyzed at a wavelength (λmax) of 510 nm after standing for 30 min.

#### *3.6. UHPLC-MS/MS Method of Analysis*

Separation and analysis of cyanogenic glycosides were performed with a C18 column (Acquity BEH, 1.7 µm, 50 × 2.1 mm, Waters, Milford, MA, USA) using a Waters Acquity UHPLC coupled to a Xevo TQ-S triple quadrupole mass spectrometer (UHPLC-MS/MS). A previously published gradient [24] for the quantification of amygdalin in almonds was reduced from 20 min down to 10 min. The mobile phase included 0.1% formic acid in water (mobile phase A) and 0.1% formic acid in acetonitrile (mobile phase B). The gradient used was 95% A, 0−1 min; 95−80% A, 1−3 min; 80−40% A, 3−7 min; 40% B, 7−8 min and 95% A 8.1−10 min re-equilibration. The flow rate was 200 <sup>µ</sup>L min−<sup>1</sup> and the following conditions were used for the electrospray ionization (ESI) source: source temperature 150 ◦C, desolvation temperature 350 ◦C, capillary voltage 3.07 kV, cone voltage 21, and nebulizer gas 500 L h−<sup>1</sup> N2. Argon was used as the collision gas. The collision energies were optimized and ranged from 17 to 30 eV for individual analytes. The column and

sample temperatures were 40 ◦ and 10 ◦ C, respectively. The ESI source was operated in the positive ion mode. Instrument control and data processing were performed by using MassLynx software (version 4.1, Waters, Milford, MA, USA). Cyanogenic standard solutions were prepared with concentrations ranging from 1 ng mL−<sup>1</sup> to 10 µg mL−<sup>1</sup> . All analyses were done in triplicate along with a blank. The interday and intraday precisions of the method had a CV% of less than 5%. National Institutes of Health, Bethesda, MD, USA). The method generated mean intensity values corresponding to each CN<sup>−</sup> eq. and was used to generate a calibration curve. Quantification was confirmed using a UV‐Vis spectrometer (Agilent 8454 photodiode array, Agilent Technologies, Santa Clara, CA, USA). The reacted picrate paper test strip was extracted in 3.5 mL of water in cuvettes and the resulting solution analyzed at a wavelength (λmax) of 510 nm after standing for 30 min.

*Molecules* **2021**, *26*, x FOR PEER REVIEW 14 of 18

achieved by adding 50 μL of 3U/mL β‐glucosidase.

of this analysis is the enzymatic hydrolysis of amygdalin. Enzymes are macromolecular biological catalysts whose amount for a specific enzymatic activity is measured in Units (U). One U is defined as the amount of enzyme needed to catalyze the conversion of 1 micromole of substrate per minute [48]. Enzymatic degradation of amygdalin was

The commercially processed elderberry juice was tested for TCP along with the AE samples. 100 μL/100 mg each of lyophilized and fresh tissues of Ozone, Ozark and pooled AE samples were measured/weighed into clean 20 mL scintillation vials. 50 μL of β‐ glucosidase solution (3U/mL) in pH 8 phosphate buffer was added alongside the picrate paper and the vial was immediately closed with a screw stopper. Each sample analysis was made in replicates of four. Similar set‐ups were made for amygdalin standards and blank (no amygdalin added). These were left overnight (16**–**24 h) in an oven at 30**–**40 °C. Two control experiments were performed. The first was to confirm the effectiveness of the picrate paper test method to known literature. Seeds from two apple varieties, Granny Smith (GS) and Gala (G), were prepared and tested for TCP using the same protocol for the AE samples. The second control experiment used seeds (ground) from the two apple varieties and stems and green berries from AE pooled samples. This was to test for the presence of endogenous enzymes in the samples to assess the extent of self‐ hydrolysis of CNGs. To accomplish this, two different picrate paper set‐ups were made, a control (without an external β‐glucosidase solution) and a second typical picrate set‐up (with an external β‐glucosidase solution). Four replicates for each set‐up were performed. A simple and quick method of analyzing the reacted picric acid is by qualitative inspection. This appears to be an effective method for quantifying CN<sup>−</sup> equivalents ranging between 1 and 100 μg. However, as the amount of CN<sup>−</sup> eq. increases, the ability to differentiate between the colors decreases. A color chart shown in Figure 11 can be used to relate the color change of the paper to total amount of CN<sup>−</sup> evolved. A semi‐quantitative approach using a camera‐phone as a detector was used. An image of a concentration from Figure 11 was converted from color to greyscale. This was done using Image J software (https://imagej.nih.gov/ij/index.html (accessed on 22 December 2020), version 1.46r,

**Figure 11.** A picrate‐paper cyanide color chart for qualitative analysis of CNGs for the range of 0**–** 100 μg CN<sup>−</sup> eq. **Figure 11.** A picrate-paper cyanide color chart for qualitative analysis of CNGs for the range of 0–100 µg CN− eq.

#### *3.7. Statistical Analysis*

For the determination of cyanide by UV-Vis and cyanogenic glycosides by LC-MS/MS, samples were prepared in three biological and three analytical replicates for each sample. Statistical analyses were performed in Excel (Microsoft Office 2016). Results are expressed as the mean ± standard error of mean (SEM). Additionally, the coefficient of variation for six (6) repeated injections (CV% ≤ 20%) was used to confirm candidate concentrations for LLOQ and LLOD.

#### **4. Conclusions**

The UHPLC-MS/MS and picrate paper methods developed were used to reliably determine the intact CNGs and assess the TCP in various AE fruit tissue. No quantifiable trace of cyanide or CNG was detected in commercial elderberry juice. Moreover, traces of CNGs (amygdalin, dhurrin, (prunasin + sambunigrin), and linamarin) detected in tissues of AE samples were generally low with lower levels in the juice and seeds as compared to stems and skin. TCP assessed in both pure and pooled AE sample tissues were generally low with higher concentrations recorded in pooled stems and unripe (green) berries. The picrate paper method can also be used to help detect the presence of CNGs. A camera-phone and UV-Vis spectrophotometer can both be used as a detector. The cameraphone can give results easily with limits of detection that are useful for CNG analysis. Although the TCP and CNGs levels in tissues of AE pose no threat to consumers, it is advisable to separate out the stems, green berries and leaves [36] from AE ripe berries during product preparation.

**Supplementary Materials:** The following are available online: Figure S1: Picrate-paper results for the using a camera-phone to detect the presence of CN− using amygdalin. Figure S2: Picrate-paper results for the using a UV-Vis spectrophotometer in the low concentration range (0–10 µg) to detect the presence of CN− using amygdalin. Figure S3: Picrate Paper results for tissues of Ozone and Ozark AE samples, Figure S4: Picrate paper results for pooled AE samples, Figure S5: Picrate paper results for apple seeds and juice, Figure S6: Picrate paper results for endogenous enzymes test for pooled AE tissues, Figure S7: Picrate paper results for endogenous enzymes test for fresh and lyophilized apple seeds, Figure S8: Cyanide reaction with picric acid.

**Author Contributions:** Conceptualization, C.M.G. and A.L.T.; methodology, M.K.A., M.C.J., H.I., R.B., R.K.; formal analysis, M.K.A. and C.M.G.; investigation M.K.A., M.C.J., H.I., R.B., R.K.; Writing original draft preparation, M.K.A., C.M.G. and A.L.T.; writing—review and editing, M.K.A., C.M.G. and A.L.T.; project administration, C.M.G. and A.L.T.; funding acquisition, C.M.G. and A.L.T. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by the Missouri Department of Agriculture through the USDA's Specialty Crop Block Grant Program; by Hatch project accession no. 1011521 from the USDA National Institute of Food and Agriculture; and by the University of Missouri Center for Agroforestry under a Cooperative Agreement (58-6020-6-001) with the USDA/ARS Dale Bumpers Small Farms Research Center.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Original data is available upon request to the corresponding author.

**Acknowledgments:** We thank Jillian Boydston and Samuel Sergent for their excellent technical assistance.

**Conflicts of Interest:** The authors declare no conflict of interest.

**Sample Availability:** Not available.

#### **Abbreviations**


#### **References**


## *Article* **Subcritical Water Extraction of** *Salvia miltiorrhiza*

**Brahmam Kapalavavi <sup>1</sup> , Ninad Doctor <sup>1</sup> , Baohong Zhang <sup>2</sup> and Yu Yang 1,\***


**Abstract:** In this work, a green extraction technique, subcritical water extraction (SBWE), was employed to extract active pharmaceutical ingredients (APIs) from an important Chinese medicinal herb, *Salvia miltiorrhiza* (danshen), at various temperatures. The APIs included tanshinone I, tanshinone IIA, protocatechualdehyde, caffeic acid, and ferulic acid. Traditional herbal decoction (THD) of *Salvia miltiorrhiza* was also carried out for comparison purposes. Reproduction assay of herbal extracts obtained by both SBWE and THD were then conducted on *Caenorhabditis elegans* so that SBWE conditions could be optimized for the purpose of developing efficacious herbal medicine from *Salvia miltiorrhiza.* The extraction efficiency was mostly enhanced with increasing extraction temperature. The quantity of tanshinone I in the herbal extract obtained by SBWE at 150 ◦C was 370-fold higher than that achieved by THD extraction. Reproduction evaluation revealed that the worm reproduction rate decreased and the reproduction inhibition rate increased with elevated SBWE temperatures. Most importantly, the reproduction inhibition rate of the SBWE herbal extracts obtained at all four temperatures investigated was higher than that of traditional herbal decoction extracts. The results of this work show that there are several benefits of subcritical water extraction of medicinal herbs over other existing herbal medicine preparation techniques. Compared to THD, the thousand-year-old and yet still popular herbal preparation method used in herbal medicine, subcritical water extraction is conducted in a closed system where no loss of volatile active pharmaceutical ingredients occurs, although analyte degradation may happen at higher temperatures. Temperature optimization in SBWE makes it possible to be more efficient in extracting APIs from medicinal herbs than the THD method. Compared to other industrial processes of producing herbal medicine, subcritical water extraction eliminates toxic organic solvents. Thus, subcritical water extraction is not only environmentally friendly but also produces safer herbal medicine for patients.

**Keywords:** active pharmaceutical ingredients; reproduction; medicinal herbs; *Salvia miltiorrhiza*; subcritical water extraction

#### **1. Introduction**

Due to its green nature and low side effects, herbal medicine has gained greater attention in the Western world nowadays [1–3]. Both raw and preprepared herbal medicines are available in many developed countries [4,5].

The traditional way for patients to take the herbal medicine prescribed by doctors is to cook the medicinal herbs in boiling water for 60 to 90 min and then drink the "soup medicine". This herbal medicine preparation method is called traditional herbal decoction (THD). Although this herbal decoction method has been used since ancient times, there are several major drawbacks associated with it. Firstly, a large portion of the volatile active pharmaceutical ingredients (APIs) contained in medicinal herbs are lost during the cooking process with boiling water. This is because the decoction process is an open system, and volatile APIs are thus lost to the atmosphere as vapor. Secondly, some APIs contained in medicinal herbs may be degraded due to the prolonged cooking time of 60 to 90 min. In both cases, the effectiveness of THD extracts, the soup medicine, in treating diseases may

**Citation:** Kapalavavi, B.; Doctor, N.; Zhang, B.; Yang, Y. Subcritical Water Extraction of *Salvia miltiorrhiza*. *Molecules* **2021**, *26*, 1634. https:// doi.org/10.3390/molecules26061634

Academic Editor: Gavino Sanna

Received: 17 February 2021 Accepted: 12 March 2021 Published: 15 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

be decreased due to the reduced quantity of APIs in the herbal medicine as a result of them being lost to vapor or by degradation. Concurrently, even as APIs are lost, compounds with detrimental health effects may be extracted during the lengthy THD process. The presence of such toxicants in medicinal herbal extracts may not be safe for patient use. Lastly, it would be a rare coincidence for 100 ◦C to be the best temperature for effective extraction of all APIs from medicinal herbs. Proper scientific investigation of other temperatures may yield more potent yet safer herbal medicine.

Several other methods have been used for extraction of herbs and plants, including Soxhlet extraction, sonication, pressurized liquid extraction, accelerated solvent extraction, microwave-assisted extraction, and sub- and supercritical fluid extraction [6–10]. Because organic solvents are used in most of these extraction techniques, such as Soxhlet and sonication extractions, they are not suitable for preparing herbal medicine due to the toxicity of organic solvents.

Herbal extracts, such as small bags of medicinal herb extracts, are prepared by largescale THD for patients so that they can take them directly without having to cook the herbs. This preprepared herbal medicine has gained popularity due to its convenience. Other forms of preprepared herbal medicines, such as tablets, capsules, and instant beverages, are also available commercially. While these products provide convenience to the consumer, their production via commonly used industrial extraction techniques is taking its toll on the environment and perhaps even on the patients. These techniques include maceration, vertical or turbo extraction, ultrasonic extraction, percolation, and counter current extraction. Many of the organic solvents required for use in these herbal extraction methods are toxic, and some are even carcinogenic [11]. The solvents required in these herbal preparation processes are costly not only to purchase but also for its waste disposal. Overall, such harsh extraction methods carry risks for the consumer and the environment, making them principally at odds with the perceived desire of the consumer who is likely looking for natural remedies rather than pollution-causing industrial processes and persistent trace carcinogens.

A scientifically rigorous path for modernization of herbal medicine preparation techniques is of great interest. It is important to not simply mimic THD but also to improve the efficacy of herbal medicines than those prepared by THD. This leads to this research, subcritical water extraction (SBWE) of medicinal herbs. Subcritical water refers to hightemperature and high-pressure water under conditions lower than the critical point of water: 374 ◦C and 218 atm. Water at elevated temperatures acts like an organic solvent due to its weakened hydrogen bonds and decreased polarity [12,13]. The solubility of organic compounds such as APIs in medicinal herbs is dramatically enhanced by simply increasing the water temperature. This unique characteristic of high-temperature water makes it an alternative mobile phase solvent for reversed-phase liquid chromatography [13–16] and an excellent extraction fluid for efficient removal of organics from various sample matrices, including plants and medicinal herbs [17–23]. Because different temperatures can be employed to carry out subcritical water extractions, there will be an optimized temperature that yields the highest quantity of APIs and in turn produces the most potent herbal medicine. Ideally, the solvent for extraction of medicinal herbs should be nontoxic, and the extraction technique should be more efficient in extracting active pharmaceutical ingredients and not cause their significant loss during the extraction process. Thus, subcritical water is an excellent choice for preparing herbal medicines.

In order to evaluate and optimize the SBWE technique, *Salvia miltiorrhiza* (also known as danshen in Chinese), a popular and important herb prescribed in traditional Chinese medicine (TCM), was used in this study. *Salvia miltiorrhiza* is a perennial plant in the genus *Salvia* of the mint family. Its roots are highly valued in traditional Chinese medicine and used in the treatment of various diseases, such as blood circulation, cardiovascular, and hepatic diseases [24–26]. Researchers have isolated about 70 compounds from the extract of *Salvia miltiorrhiza* [27]. Some of the identified anticancer compounds present in *Salvia miltiorrhiza* include tanshinone I, tanshinone IIA, protocatechualdehyde, caffeic acid, and

ferulic acid. These APIs have already been found to demonstrate antiproliferative effect on various cancer cells, such as colon, leukemia, lung, and breast cancers, at either pre-clinical or clinical level [28–31]. Therefore, these five APIs were investigated in this study.

The main goal of this work was to investigate a potential herbal medicine preparation technique using subcritical water to yield efficacious herbal medicine. Therefore, subcritical water extraction of *Salvia miltiorrhiza* roots was carried out at four different temperatures (75, 100, 125, and 150 ◦C). For comparison and evaluation purposes, traditional herbal decoction of *Salvia miltiorrhiza* was also conducted. Then, these herbal extracts were characterized using GC/MS and HPLC to identify and quantify various anticancer agents. In order to evaluate the efficacy of the SBWE herbal extracts at various temperature conditions, the reproduction assay of SBWE and THD herbal extracts were conducted on *Caenorhabditis elegans*.

Despite being a simple multicellular organism, *Caenorhabditis elegans* has been widely employed to study complex behavior and syndromes. It has been used in many recent studies to understand human diseases, including cancer, ageing, development, addiction, and neurodegenerative diseases, as well as in pharmaceutical and toxicity studies [32–36]. Research on the worm bridges the gap between in vitro systems and preclinical studies in mammalian models. Experiments using cell lines often do not represent organism-level responses. On the other hand, *Caenorhabditis elegans* is particularly useful for reverse genetic approaches due to its short life cycle, availability of strains and feasibility of customized mutants, ability to perform complex behavior, and transparent cuticle for imaging assays. In this work, we employed *Caenorhabditis elegans* as a model system to investigate the drug potency of the extracted APIs from *Salvia miltiorrhiza* using reproduction analysis.

It is a novel approach to employ subcritical water for extraction of active pharmaceutical ingredients from medicinal herbs. Compared to THD, the thousand-year-old and yet still popular herbal preparation method, subcritical water extraction is conducted in a closed system. Therefore, no loss of volatile active pharmaceutical ingredients occurs due to loss of APIs to the open environment. However, loss of analytes can still occur due to degradation at elevated temperatures. Under optimized temperature, SBWE is more efficient in extracting APIs from medicinal herbs than the THD method. Compared to other industrial processes of producing herbal medicine, subcritical water extraction eliminates toxic organic solvents. Therefore, subcritical water extraction is not only more efficient and cheaper but also environment friendly because of its green nature. Many researchers have made efforts in recent years to develop greener analytical chemistry techniques. For example, an analytical Eco-Scale has been proposed as a tool for green analysis evaluation [37]. Another new tool introduced for assessment of the green character of analytical procedures is the Green Analytical Procedure Index [38]. The work reported in this paper also contributes to the field of green analytical chemistry.

#### **2. Results and Discussion**

#### *2.1. Subcritical Water Extraction of Salvia Miltiorrhiza*

As stated later in the Materials and Methods section, the quantification of all five APIs was achieved using a standard HPLC method. The concentration of the calibration solutions ranged from 0.002 to 1.00 mg/mL. The detection limit was 0.0002 mg/mL. The correlation coefficient (*r 2* ) ranged from 0.999 to 1.00.

A recovery study on the SBWE method was conducted using spiked samples (known amount of APIs) to validate the homemade SBWE system. The recoveries of the five APIs investigated in this work ranged from 95 to 102%, similar to that achieved in our previous study on SBWE of vanillin and coumarin [17]. This shows that the SBWE system is reliable. The subcritical extraction of *Salvia miltiorrhiza* was conducted at four different temperatures of 75, 100, 125, and 150 ◦C. Then, the SBWE extracts were characterized using GC/MS. Various analytes in the herbal extracts were identified by GC/MS by matching both GC retention times and mass spectra of standard samples. Among the identified analytes, five of them were anticancer agents: protocatechualdehyde, caffeic acid, ferulic acid, tanshinone

I, and tanshinone IIA. Figure 1 shows the elution of the five compounds with an internal standard on GC/MS.

**Figure 1.** Total ion GC/MS chromatogram of a *Salvia miltiorrhiza* herbal extract obtained by subcritical water extraction (SBWE) at 150 ◦C for 30 min. Peak identification: 1, protocatechualdehyde; 2, propyl paraben (internal standard); 3, caffeic acid; 4, ferulic acid; 5, tanshinone IIA; and 6, tanshinone I.

Figure 2 shows the HPLC separation of a standard solution (Figure 2a), methylene chloride phase after liquid–liquid extraction of an herbal extract obtained by SBWE at 125 ◦C (Figure 2b), and water phase (methanol was added) of an herbal extract obtained by SBWE at 125 ◦C (Figure 2c). As one can see, all five analytes and the internal standard were well separated.

**Figure 2.** *Cont*.

**Figure 2.** HPLC chromatograms of *Salvia miltiorrhiza* herbal extract obtained with 125 ◦C extraction temperature and evaluated in the Alltech Adsorbosil C18 column at ambient temperature. (**a**) Analyte standard solution; (**b**) methylene chloride phase; (**c**) water phase. Flow rate: 1.0 mL/min. UV detection: 254 nm. Mobile phase: A, 100 mM phosphoric acid in water; B, 100% methanol. Gradient: 0–4 min, 2% methanol; 4–8 min, 2–10% methanol; 8–23 min, 10–30% methanol; 23–32 min, 30–60% methanol; 32–43 min, 60% methanol; 43–49 min, 60–70% methanol; 49–61 min, 70–80% methanol; and 61–68 min, 80–2% methanol. Peak identification: 1, protocatechualdehyde; 2, caffeic acid; 3, ferulic acid; 4, propyl paraben; 5, tanshinone I; and 6, tanshinone IIA.

Table 1 shows the quantification results of the five analytes present in the SBWE herbal extracts obtained at four different temperatures of 75, 100, 125, and 150 ◦C. The quantification results indicate that the protocatechualdehyde quantity extracted increased by 2-fold with the increase of extraction temperature from 75 to 100 ◦C and by 24-fold with further increase of extraction temperature from 100 to 125 ◦C. Then, with further increase of temperature from 125 to 150 ◦C, the extracted protocatechualdehyde quantity was enhanced 2.5-fold. There was no significant temperature effect on extraction efficiency of caffeic acid in the temperature range of 75 to 125 ◦C. However, the caffeic acid quantity found in

the extract decreased at 150 ◦C due to possible degradation at such a high temperature. Ferulic acid was not detected at 75 and 100 ◦C, while its quantity extracted was improved by 32-fold when the temperature increased from 125 to 150 ◦C. The extraction efficiency, measured by analyte concentration in herbal extracts, of the two tanshinone compounds was clearly enhanced with increasing temperature, as shown in Table 1.

**Table 1.** Comparison of active pharmaceutical ingredient (API) concentrations found in *Salvia miltiorrhiza* obtained by traditional herbal decoction and subcritical water extraction.


<sup>a</sup> Triplicate measurements. <sup>b</sup> Not detected.

Table 1 also includes the quantities of the five analytes found in the THD extracts. One can easily see that tanshinone concentrations obtained by SBWE at all temperatures were much higher than those achieved by THD extractions. Specifically, tanshinone I concentration achieved by SBWE at 150 ◦C was 370-fold higher than that obtained by THD extraction, as demonstrated in Table 1.

We conducted the *t*-test on API concentrations in both THD and SBWE extracts. Our statistical analysis revealed that tanshinone I and tanshinone IIA concentrations obtained by THD at 100 ◦C and by SBWE at all four elevated temperatures were significantly different beyond the 99.9% confidence level. While there were no differences between caffeic acid concentrations achieved by THD at 100 ◦C and SBWE at 75–125 ◦C, the concentrations of caffeic acid obtained by THD and SBWE at 150 ◦C were significantly different at the 99.5% confidence level. Protocatechualdehyde concentrations achieved by THD at 100 ◦C and SBWE at all four temperatures were significantly different, mostly beyond the 99% confidence level.

#### *2.2. Reproduction Assay of Caenorhabditis Elegans*

First, we studied the impact of different concentrations (2, 10, and 50 times dilution with deionized water) of the herbal extract on *Caenorhabditis elegans* mortality. *Salvia miltiorrhiza* herbal extract was obtained by SBWE at 150 ◦C. After 30 h exposure to the three different diluted SBWE herbal extracts, the 10 times diluted herbal extract showed higher mortality rate than the other diluted herbal extracts. Therefore, the 10 times dilution factor was chosen for the remainder of the reproduction study. The API concentrations used for the reproduction assay are given in Table 2.

**Table 2.** Concentration of API in *Salvia miltiorrhiza* extracts used for reproduction study.


<sup>a</sup> Not detected.

In order to optimize the preparation conditions of efficacious herbal medication through subcritical water extraction, the reproduction inhibition of the SBWE herbal extracts obtained at four different temperatures (75, 100, 125, and 150 ◦C) was evaluated on *Caenorhabditis elegans*. All SBWE herbal extracts were diluted 10 times with deionized water. Table 3 shows the reproduction assay of *Caenorhabditis elegans* after 30 h exposure to the 10 times diluted SBWE water extracts obtained at 75 to 150 ◦C. The reproduction inhibition of the extracts increased with higher extraction temperature. The SBWE extraction temperature also influenced mortality, as shown in Table 3. In general, the worm survival rate decreased with the increase in extraction temperature except at 150 ◦C. The main reason for lower mortality of *Caenorhabditis elegans* with the herbal extract obtained at 150 ◦C may be attributed to the less intake of highly concentrated herbal extract through the skin of *Caenorhabditis elegans*. Another reason for the lower mortality of worms may be due to the degradation of compounds associated with the mortality of worms, such as caffeic acid, at 150 ◦C.

**Table 3.** Percentage reproduction inhibition and mortality of *Caenorhabditis elegans* after 30 h exposure to the 10 times diluted traditional herbal decoction and subcritical water extractions of *Salvia miltiorrhiza* at 75 to 150 ◦C.


<sup>a</sup> Total of eggs and larva average per worm over three days. <sup>b</sup> Five replicates.

The reproduction assay of *Caenorhabditis elegans* was also carried out using the THD extract of *Salvia miltiorrhiza* for comparison purposes. Both reproduction inhibition and mortality achieved by SBWE extracts at 100 ◦C and above were higher than those obtained by traditional herbal decoction, as shown in Table 3. The reproduction inhibition results indicate that SBWE is a much more efficient extraction technique than traditional herbal decoction, and it may be used to develop efficacious herbal medicine in the future.

We also carried out the *t*-test on reproduction assay results. Our statistical analysis showed that there was no difference in average reproduction and reproduction inhibition between the herbal extracts obtained by THD at 100 ◦C and SBWE at 75 ◦C, while there were significant differences between the extracts of THD 100 ◦C and SBWE at the other three elevated temperatures, mostly beyond the 99.9% confidence level. There were significant differences in mortality for the herbal extracts obtained by THD at 100 ◦C and SBWE at all four temperatures, mostly beyond the 99% confidence level.

#### **3. Materials and Methods**

#### *3.1. Reagents and Supplies*

Tanshinone I and tanshinone IIA were obtained from LKT Laboratories, Inc. (St. Paul, MN, USA). Protocatechualdehyde, caffeic acid, ferulic acid, sodium chloride, sodium hydroxide, agar, cholesterol, calcium chloride, calcium chloride dehydrate, and sodium phosphate dibasic heptahydrate were purchased from Sigma Aldrich (St. Louis, MO, USA). Sand, peptone, tryptone, magnesium sulfate, and magnesium sulfate heptahydrate were acquired from Fisher Scientific (Fair Lawn, NJ, USA). Potassium phosphate, dipotassium phosphate, yeast extract, and HPLC-grade methanol were purchased from Alfa Aesar (Ward Hill, MA, USA). Methylene chloride was obtained from Acros Organics (Fair Lawn, NJ, USA). Top Job bleaching solution was obtained from the local store. Deionized water (18 MΩ-cm) was prepared in our laboratory using a Purelab Ultra system from ELGA (Lowell, MA, USA). GD/X PVDF membrane filters (0.45 µm) were acquired from Whatman

(Florham Park, NJ, USA). Strata SPE silica-2 sample (3 mL) tubes were received from Phenomenex (Torrance, CA, USA). Petri dishes (6 cm) were obtained from BD Falcon (Franklin Lakes, NJ, USA). Alltech Adsorbosil C18 column (4.6 × 150 mm, 5 µm) was purchased from Alltech Associates, Inc. (Deerfield, IL, USA). An Empty stainless steel tube (5 × 1.00 cm I.D. with 1.27 cm O.D.) and end fittings were received from Chrom Tech, Inc. (Apple Valley, MN, USA). OP50 and *Caenorhabditis elegans* N2 Bristol wild-type worm were obtained from *Caenorhabditis* Genetics Center (University of Minnesota, Minneapolis, MN, USA).

#### *3.2. Preparation of Solutions*

Propyl paraben was used as an internal standard. This solution was prepared by adding 0.0500 g of propyl paraben to a 50 mL volumetric flask and diluted to the mark with methanol. A stock solution was prepared by adding 0.00020 g each of tanshinone I and tanshinone IIA to a 10 mL volumetric flask. Then, 4.00 mL of dichloromethane was added into the volumetric flask. The volumetric flask was vortexed to obtain a homogeneous mixture. Then, 0.0100 g each of protocatechualdehyde, caffeic acid, and ferulic acid were added to the same volumetric flask and diluted to the mark with methanol. Calibrated standard solutions were prepared using both stock and internal standard solutions. The concentrations of the calibration solutions ranged from 0.002 to 1.00 mg/mL.

#### *3.3. Subcritical Water Extraction of Salvia Miltiorrhiza*

The extraction of *Salvia miltiorrhiza* was carried out using a home-made subcritical water extraction system, as shown in Figure 3. Both end fittings of a stainless steel extraction vessel were wrapped with Teflon tape for proper sealing. One end of the vessel was sealed with an end fitting first. Approximately 2 g (the actual weight was recorded to four decimal place) of *Salvia miltiorrhiza* (finely cut to small pieces, a few millimeters in length) was added to the stainless steel vessel. The void space of extraction vessel was filled with precleaned sand. The other end of the stainless steel vessel was then sealed with another end fitting. The loaded vessel was placed in an oven (HP gas chromatograph 5890 Series 2, Hewlett Packard, Avondale, PA, USA), as shown in Figure 3.

**Figure 3.** Block diagram of subcritical water extraction (V1 and V2 are needle valves) followed by GC/MS identification and HPLC quantification of APIs.

An ISCO model 260 D syringe pump (Lincoln, NE, USA) was used to supply 18 MΩcm water by opening V<sup>1</sup> and closing V<sup>2</sup> to fill the loaded vessel. Leak check of the extraction vessel was performed in the constant-pressure mode. It should be pointed out that a delay between the actual temperature of the extraction vessel and oven temperature was determined. The delay was 10 min for 75 ◦C, 12 min for 100 ◦C, 14 min for 125 ◦C, and 16 min for 150 ◦C. Static extraction was performed for 30 min after the delay time was compensated. A pressure of 15 to 25 atm was applied to keep hot water in the liquid state for all experiments. After 30 min of heating, approximately 10 mL of herbal extract was collected at 1 mL/min into a 25 mL glass vial by opening V2. Triplicate SBWE experiments were conducted at all temperatures.

#### *3.4. Traditional Herbal Decoction of Salvia Miltiorrhiza*

Approximately 2 g (the actual weight was recorded to four decimal place) of *Salvia miltiorrhiza* (finely cut to small pieces, a few millimeters in length) was added to a 50 mL glass beaker. Then, 10.00 mL of deionized water was added to it. The beaker was covered with a watch glass and heated up to boiling on a hot plate. Then, the temperature was adjusted to ensure the water kept boiling for 30 min. Triplicate THD experiments were conducted.

#### *3.5. Sample Treatment*

For characterization of SBWE water–herbal extracts on GC/MS, solid-phase extraction (SPE) was carried out using a silica phase cartridge and methanol as the elution solvent. At first, the silica cartridge was cleaned with approximately 5 mL of methanol followed by 10 mL of water. Then, the herbal extract was run through the silica cartridge and eluted using 1.00 mL of methanol into a 2 mL glass vial. Then, 30 µL of propyl paraben internal standard solution was added.

For HPLC analysis of tanshinone I and tanshinone IIA, liquid–liquid extraction was conducted. First, 1.00 mL of methylene chloride was added to each glass vial containing SBWE water–herbal extract. These vials were then sealed with aluminum-lined caps. These vials were vortexed to effectively mix the two phases. After separation of the two phases, the methylene chloride phase was transferred into an empty 5 mL glass vial. The same liquid–liquid extraction procedure for the SBWE water–herbal extract was repeated with another 1.00 mL of fresh methylene chloride. Again, the methylene chloride layer was removed and combined with the first fraction of the methylene chloride extract in the 5 mL vial. Then, 30.00 µL of propyl paraben internal standard was added to the methylene chloride phase.

To the aqueous phase of the herbal extract sample, 1.00 mL of methanol was added. Then, 300 µL of propyl paraben internal standard was added and mixed well. This sample was then filtered through a Whatman GDX filter into a glass vial for chromatographic analysis of protocatechualdehyde, caffeic acid, and ferulic acid.

#### *3.6. HPLC Analysis*

Please note that *Salvia miltiorrhiza* contains tens if not hundreds of compounds, and the five APIs investigated in this work are mixed with all other compounds in the extract. Therefore, we needed a standard method such as HPLC to achieve separation and quantification of our analytes to ensure the quality of this research. Thus, Shimadzu Nexera UFLC was employed for separation and quantification of *Salvia miltiorrhiza* extracts on an Alltech Adsorbosil C18 column using a methanol–water mixture as the mobile phase with 1.0 mL/min at ambient temperature. The eluents were detected at 254 nm.

#### *3.7. GC/MS Analysis*

In order to separate and identify the five APIs studied in this work, Agilent Technologies 6890N Network GC System (Santa Clara, CA, USA) coupled with a JEOL Ltd. JMS-GCmate II MS System (Tokyo, Japan) was employed for the characterization of SBWE extracts of *Salvia miltiorrhiza*. The GC separations were carried out on an Agilent HP-5MS (5% phenyl)-methylpolysiloxane (30 m × 0.250 mm, 0.25 µm film thickness) capillary column with 1.0 mL/min flow of a helium carrier gas. The sample volume was 1 µL, and it was injected using split mode by keeping the injector temperature at 250 ◦C. The GC/MS interface and the MSD ion chamber were set at 250 ◦C. The MS solvent delay time was 3 min. The GC oven temperature programming was as follows. The initial temperature was held at 30 ◦C for 3 min. Then, it was increased at 7.4 ◦C/min to 250 ◦C and maintained at 250 ◦C for 16 min. TSSPro Version 3.0 (Shrader Analytical and Consulting Laboratories, Inc., Detroit, Michigan, USA) was used for data acquisition and analysis.

#### *3.8. Reproduction Studies on Caenorhabditis Elegans*

A hermaphrodite, *Caenorhabditis elegans* N2 Bristol wild-type worm, was used for the reproduction assay to determine the reproduction rate of the herbal extracts. Synchronized L1 worms were cultured on NGM with *Escherichia coli* bacteria (OP50) as food. The NGM was supplemented with SBWE herbal extracts with a certain fold of dilution, which were obtained at extraction temperatures of 75, 100, 125, and 150 ◦C. Each treatment contained 15 worms with five biological replicates. These plates were incubated at 20 ◦C. After 30 h, these worms were washed off from each plate using M9 buffer into an Eppendorf tube. The tubes were then centrifuged twice with M9 buffer to wash worms of the herbal extract. Then, from each tube, about four worms were transferred to each plate already seeded with OP50 food. These plates were continuously monitored for egg laying. When worms started laying eggs, time was noted, and the plates were labeled as day 1 plates. These plates were incubated for another 24 h. The number of laid eggs was recorded for each day for three continuous days. The following equations were used to calculate reproduction inhibition and mortality rate.

$$\text{Reproduction inhibition} = \frac{\text{Control average reproduction} - \text{Herbal extract average reproduction}}{\text{Control average reproduction}} \times 100\tag{1}$$

$$\% \text{Mortality} = \frac{\text{Number of words died}}{\text{Total number of words}} \times 100\tag{2}$$

#### **4. Conclusions**

The research described in this work is different from any other existing herbal medicine preparation techniques. Unlike traditional herbal decoction, subcritical water extraction is conducted in a closed system where no loss of volatile active pharmaceutical ingredients to open environment occurs except analyte degradation at higher temperatures. Because temperatures other than 100 ◦C (the condition used in traditional herbal decoction) can also be employed in SBWE, subcritical water extraction at the optimized temperature should be more efficient in extracting active pharmaceutical ingredients from medicinal herbs than traditional herbal decoction. The higher extraction efficiency of SBWE should allow the subcritical water extraction time to be shortened, thus reducing the chance for degradation of the active pharmaceutical ingredients. These three factors should assure that optimized SBWE conditions would produce herbal medicine containing higher API concentrations than traditional herbal decoction. Compared with other industrial processes of making herbal medicine, subcritical water extraction eliminates toxic organic solvents. Therefore, it is not only environment friendly but also produces safer herbal medicine for patients.

Our results showed that the API quantity obtained by subcritical water extraction of *Salvia miltiorrhiza* increased by up to 4-fold by increasing the extraction temperature from 75 to 100 ◦C. They were then further enhanced by up to 26-fold with the increase of temperature from 100 to 125 ◦C, except for caffeic acid. When the extraction temperature was raised from 125 to 150 ◦C, API concentrations in SBWE extracts were further increased up to 4-fold, except for caffeic acid and protocatechualdehyde. Both caffeic acid and protocatechualdehyde might be degraded at 125 ◦C or higher. When comparing the tanshinone concentrations achieved by SBWE of *Salvia miltiorrhiza* with that obtained by THD, the SBWE extracts contained much higher tanshinone concentrations than the THD extracts.

The extraction temperature also plays an important role in the reproduction inhibition rate of the SBWE herbal extracts collected at four different temperatures. Reproduction inhibition evaluation of *Caenorhabditis elegans* revealed that the three-day average reproduction of worms decreased with increasing extraction temperature, while the reproduction inhibition rate increased from 6 to 46% when the SBWE temperature was raised from 75 to 150 ◦C. Please note that the reproduction inhibition of SBWE herbal extracts obtained at all temperatures from 75 to 150 ◦C was higher than that of traditional herbal decoction extracts.

In closing, besides subcritical water extraction being a more efficient technique than traditional herbal decoction in extracting anticancer agents from *Salvia miltiorrhiza,* SBWE herbal extracts also have higher reproduction inhibition rate than THD extracts according to our reproduction inhibition study. These findings demonstrate the potential of employing subcritical water extraction technique to develop high API-containing herbal medicine from *Salvia miltiorrhiza*.

**Author Contributions:** Manuscript conception, Y.Y.; writing and original draft preparation, Y.Y. and B.Z.; subcritical water extractions, B.K. and N.D.; chromatography analysis, B.K. and N.D. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

**Sample Availability:** Samples of *Salvia miltiorrhiza* are available from the authors.

#### **Abbreviations**


#### **References**

