**1. Introduction**

Natural products have been used in the field of medicine and cosmetics for centuries. Their potential to treat various skin diseases and to improve skin condition is well-known. As ultraviolet (UV) radiation is a contributing cause for sunburns, wrinkles, premature aging, cancer and reduced immunity against infections, there is an increasing demand for products that provide protection against UV radiation [1]. Tyrosinase is a key enzyme that catalyzes the initial rate-limiting steps of melanin synthesis [2,3]. Abnormal and excessive melanin synthesis is the primary cause for skin disorders including melasma, senile lentigo, freckles and age spots [4]. As a result, finding novel and effective melanogenesis inhibitors has profound importance in controlling melanin production and

pigmentation-related disorders [5]. In an attempt to find potent and safe tyrosinase inhibitors, this experiment evaluated the in vitro tyrosinase inhibitory activity of *Angelica keiskei*.

*A. keiskei* Koidzumi (Japanese name: 'Ashitaba', Umbelliferae) is a hardy perennial herb with exuberant vitality and multi-bioactive components. It originated in Hachijojima, Japan, and now mainly grows along the Pacific coast of Japan. *A. keiskei* is traditionally used as a diuretic, laxative, analeptic and galactagogue [6–8]. It is a dark green leafy vegetable that has been recognized as a medicinally important herb and cultivated in many Asian countries because of its health benefits. The stems and leaves have been consumed commercially as health foods and the roots have also been used as a food additive and medicine to alleviate pain and diabetic symptoms [9]. Various chalcones such as xanthoangelol, 4-hydroxyderricin and coumarins like xanthotoxin and laserpitin have been isolated and characterized from this plant [6,8].

Chalcones and coumarins are the main bioactive compounds in *A. keiskei*. Chalcones have been widely studied and are known to contain antioxidants [10], anticancer agents [11,12] and α-glucosidase inhibitors [13]. Coumarins, on the other hand, were proved to have antioxidants [14], antidepressants [15] and anticancer agents [8,16,17].

An ultra-performance liquid chromatography (UPLC) coupled with the high resolution MS/MS method is the advanced method for the identification and quantification of phenol compounds from plants and food [18]. The full scan analysis reveals the molecular weight (MW) of the unknown compounds, while tandem MS reveals aspects of the chemical structure of the precursor ion via fragmentation. Ion trap mass spectrometers can carry out sequential fragmentations of the precursor ion to form product ions. By using this method, unknown compounds are identified from the exact mass and MS/MS fragmentation. Since compounds are scanned separately and identified, the UPLC-MS/MS approach has increased sensitivity and provides more structural information based on fragmentation pattern of the analyte [19].

The aim of this study was to establish a rapid method by using UPLC-MS/MS to identify the components in *A. keiskei* that regulate tyrosinase activity via statistical analysis.

#### **2. Results**

#### *2.1. The Tyrosinase Modulatory E*ff*ect of Angelica keiskei*

According to the current understanding, *A. keiskei* has anti-inflammatory, antimicrobial and antihypertensive characteristics [14], while its effect on tyrosinase activity remains largely unknown [20,21]. With the lack of research in this field, our study focused on the analysis of tyrosinase activity in the presence of different concentrations of *A. keiskei* extracts from leaf and root parts. As shown in Figure 1, the preliminary results from our study showed that at the concentration of 500 μg/mL the tyrosinase activity remained unchanged and similar to blank when treated with *A. keiskei* leaves. On the other hand, a dosage-dependent inhibition of tyrosinase activity was observed when treated with *A. keiskei* roots. This trend of the inhibitory effect started at 200 μg/mL and increased with the higher concentration of *A. keiskei* roots. At the 1000 μg/mL level, conditions even exceeded the inhibitory effect achieved by 10 μM kojic acid (positive control). Since the leaf extract showed no evident change at the 500 μg/mL concentration, the subsequent experiments were performed using the root extract of *A. keiskei* with the purpose of further examining its modulatory effect on tyrosinase activity.

**Figure 1.** The comparative ratio of tyrosinase activity of the *A. keiskei* Leaf (*AK*. Leaf) at 500 μg/mL and roots (*AK*. Root) at 200, 500, 750 and 1000 μg/mL concentrations using kojic acid (10 and 20 μM) as the positive control.

#### *2.2. Screening Tyrosinase Modulator by UPLC-MS*/*MS*

In order to investigate the compounds of *A. keiskei* roots that modulate the tyrosinase activity in a timely manner, we analyzed the samples using the simple LC-MS protocol optimized by our lab [22]. The base peak chromatograms (BPC) showed that under the ESI (+) MS mode in Figure 2, there were differences in signal intensities for the roots of *A. keiskei* treated with or without tyrosinase at the retention times of 7.93, 10.83, 10.84, 14.86, 14.88, 15.52, 18.56 and 19.04 min (Table 1). The base peak chromatograms (BPC) ESI (−) MS mode are shown in Figure 3; differences in signal intensities were observed at the retention times of 7.65, 10.08, 10.87, 13.20, 13.82, 14.32, 16.35 and 19.95 min (Table 2). Consequently, we suspect that components in the roots of *A. keiskei* might have a modulatory effect on tyrosinase activity.

**Figure 2.** The base peak chromatograms (BPC) of untreated (**A**) and tyrosinase-treated samples (**B**) of *A. keiskei* in positive ion mode.

**Figure 3.** The base peak chromatograms (BPC) of untreated (**A**) and tyrosinase-treated samples (**B**) of *A. keiskei* in negative ion mode.

#### *2.3. Multivariate Analysis to Di*ff*erentiate Processed A. keiskei*

The highly complex UPLC-MS spectra are difficult to visually link to various components; therefore, multivariate data analyses were performed to comprehensively characterize the distinct composition of various metabolites from the untreated samples and the tyrosinase-treated samples. For each condition, there were three biological replicates of *A. keiskei* metabolites (*n* = 3). After performing UPLC-orbitrap-MS-based profiling on individual samples, each dataset was processed using SIEVE software to align and extract meanings. All datasets obtained from the two treatment groups were then analyzed with PCA, OPLS-DA and S-plot functions within the SIMCA-P software to find the candidates of interest.

Due to the similarity in compositions, the differences between untreated (blank) and tyrosinase-treated (test) samples were hard to identify based on the BPC chromatograms (shown in Figures 2 and 3). Henceforth, a two-component PCA score plot of UPLC-MS data was utilized to depict general variations in components among the *A. keiskei* samples (Figure 4). According to the PCA scores plot in Figure 4, the distribution can be readily classified into two clusters, with the untreated (blank) and treated (test) samples clearly separated by principal component 1 (PC1) (Figure 4).

First, OPLS-DA was performed to compare untreated and tyrosinase-treated samples. An S-plot analysis was then used to select the critical variables that allowed differentiation. In the S-plot, each point represents *m*/*z*-RT pairs of molecules. To further reveal the tyrosinase inhibiting effect of *A. keiskei*, an OPLS-DA model was carried out between blank and test groups. The model fit well with R2Y value 0.996 and Q2 value 0.821 in positive mode, and R2Y value 1.0 and Q2 value 0.999 in negative mode. The score plot (Figure 4A,C) showed good separation, confirming the tyrosinase inhibiting effect. The corresponding S-plot was shown in Figure 4B, where coordinates in the lower-left quadrant were metabolites significantly increased in the blank group compared with the test group, while those in the upper-right quadrant represent the decreased ones. In Figure 4D, the components elevated in

untreated samples were shown in the upper-right quadrant of the S-plot, while the lower-left quadrant revealed the components elevated in tyrosinase-treated samples.

**Figure 4.** OPLS-DA score plot (**A**) and S-plot (**B**) of positive mode, OPLS-DA score plot (**C**) and S-plot (**D**) of negative mode, mass spectra obtained from untreated (blank) and tyrosinase-treated (test) groups.
