The Distribution of Glucosinolates in Different Phenotypes of Lepidium peruvianum and Their Role as Acetyl- and Butyrylcholinesterase Inhibitors—In Silico and In Vitro Studies

The aim of the study was to present the fingerprint of different Lepidium peruvianum tuber extracts showing glucosinolates-containing substances possibly playing an important role in preventinting dementia and other memory disorders. Different phenotypes of Lepidium peruvianum (Brassicaceae) tubers were analysed for their glucosinolate profile using a liquid chromatograph coupled with mass spectrometer (HPLC-ESI-QTOF-MS/MS platform). Qualitative analysis in 50% ethanolic extracts confirmed the presence of ten compounds: aliphatic, indolyl, and aromatic glucosinolates, with glucotropaeolin being the leading one, detected at levels between 0–1.57% depending on phenotype, size, processing, and collection site. The PCA analysis showed important variations in glucosinolate content between the samples and different ratios of the detected compounds. Applied in vitro activity tests confirmed inhibitory properties of extracts and single glucosinolates against acetylcholinesterase (AChE) (15.3–28.9% for the extracts and 55.95–57.60% for individual compounds) and butyrylcholinesterase (BuChE) (71.3–77.2% for the extracts and 36.2–39.9% for individual compounds). The molecular basis for the activity of glucosinolates was explained through molecular docking studies showing that the tested metabolites interacted with tryptophan and histidine residues of the enzymes, most likely blocking their active catalytic side. Based on the obtained results and described mechanism of action, it could be concluded that glucosinolates exhibit inhibitory properties against two cholinesterases present in the synaptic cleft, which indicates that selected phenotypes of L. peruvianum tubers cultivated under well-defined environmental and ecological conditions may present a valuable plant material to be considered for the development of therapeutic products with memory-stimulating properties.


Introduction
According to available statistics, average life expectancy and longevity increased from about 45 years in 1900 to over 75 years today. This can be attributed to advances in medical, Glucosinolates are well-defined biochemical plant components belonging to a group of metabolites characterized by the presence of nitrogen and sulphur atoms in their structure. They are spread across the representatives of the following botanical families: Brassicaceae, Tovariaceae, Capparidaceae, Stegnospermaceae, Moringaceae, Resedaceae, and other unrelated plant families, such as Limnanthaceae, Caricaceae, and Tropaeolaceae. When the structure of plant cells is physically disrupted or damaged, glucosinolates are hydrolyzed by the enzyme called myrosinase, which usually occurs during the physical processing of tubers. As a consequence, a sugar molecule is released, whereas the aglycon degrades, forming oxazolidine-2-thiones, nitriles, isothiocyanates, epithionitriles, thiocyanates, and other derivatives of interest from a pharmacological perspective [20,21]. Recent studies confirm the broad medicinal properties of glucosinolates and isothiocyanates that include anticancer and antimicrobial activity [22] and provide protection to the cardiovascular and central nervous systems, which may be important in the prevention of diabetic nephropathy and neuropathy [23].
Only a few up-to-date publications dealt with the anticholinesterase properties of this group of secondary metabolites. Previously, Burcul and colleagues determined the inhibitory properties of 11 compounds from the isothiocyanate group. Of these, 2-methoxyphenyl isothiocyanate showed the best inhibitory properties against AChE with IC 50 of 0.57 mM, while in the BuChE inhibition test, 3-methoxyphenyl isocyanate showed the best result of 49.2% at 1.14 mM [24]. Additionally, Blazevic et al. [25] investigated volatile isolates from Bunias erucago, with glucosinalbin and glucoraphanin as the leading components inhibiting both AChE and BuChE. Two components of extracts from maca tubers, namely N-benzylhexadecanamide and N-acetylbenzylamine, were already reported to exhibit neuroprotective effects, presumably partly due to inhibition of the AChE and BuChE enzymes [26], but the scientific literature still lacks information on the potential of glucosinolates from maca.
This fact encouraged the authors to embark on further study to assess inhibitory properties of glucosinolates present specifically in Peruvian maca (Lepidium peruvianum syn. L. meyenii) and to explain their effect towards the AChE and BuChE enzymes. Another reason for choosing this plant for the study is the fact that Peruvian maca has been demonstrated to contain about 100 times more glucosinolates than other well-known cruciferous plants, such as cabbage or broccoli [14].
These goals are complemented by a desire to compare the content of glucosinolates and to study the fingerprints of the rich collection of five prime Peruvian maca phenotypes labelled as "Yellow", "Black", "Red", "Purple", and "Gray" that were grown in two locations of the Peruvian Andean Highlands. In this research, the HPLC-ESI-QTOF-MS/MS technique and the chemometric approach were adopted to analyse the obtained results, which were extended for measuring the in vitro inhibitory potential of individual glucosinolate fractions towards AChE and BuChE. Additionally, the explanation of the grounds for the memory-enhancing activity of glucosinolates using an in silico approach linked to modeling studies was undertaken and was specifically adopted in an attempt to explain the dynamics behind utilizing the inhibitory potential of these metabolites towards both the AChE and BuChE enzymes.

Qualitative Analysis of Extracts from Lepidium peruvianum by Liquid Chromatography Coupled with Mass Spectrometry (HPLC-ESI-QTOF-MS/MS)
Glucosinolates ( Figure 1) are composed of a glucose structure which is combined with an O-sulfated thiohydroximate moiety and are variable in length-side chain is variable in derivatives of amino acids. The aliphatic chain is biosynthesized most often from methionine, aromatic moiety from phenylalanine or tyrosine, and the indole components from tryptophan [27,28]. Isothiocyanates are formed during the hydrolysis of glucosinolates. Using the example of glucotropaeolin (GTP), we will follow the stages of hydrolysis and the end products. During the hydrolysis of GTP with the participation of myrosinase, the glucose molecule and the hydroxysulfate ion are detached from the structure. As a result, a biologically active compound called benzyl isothiocyanate is formed, which is the major intermediate to benzylamine formation [29]. An understanding of the mechanisms of glucosinolates transformation is essential to determine their presence in plant extracts. methionine, aromatic moiety from phenylalanine or tyrosine, and the indole comp from tryptophan [27,28]. Isothiocyanates are formed during the hydrolysis of glu olates. Using the example of glucotropaeolin (GTP), we will follow the stages of hy sis and the end products. During the hydrolysis of GTP with the participation of m nase, the glucose molecule and the hydroxysulfate ion are detached from the stru As a result, a biologically active compound called benzyl isothiocyanate is formed, is the major intermediate to benzylamine formation [29]. An understanding of the anisms of glucosinolates transformation is essential to determine their presence in extracts. The applied chromatographic conditions allowed for the determination of t metabolites in the extracts of L. peruvianum that were visible in the negative ioni mode (also see Figures S1 and S2 in the Supplementary File), as presented in Table  identification was based on high-resolution molecular weight measurements, anal their MS/MS spectra, retention times, open MS databases, and scientific literature in line with the previously published data of other researchers [14,26,30]. The to chromatogram (see Figure S3 in Supplementary File) of the selected extract, the ext ion chromatograms of all glucosinolates whose presence was confirmed in the ne mode [26], and the MS/MS spectra of the tentatively identified structures are presen Table S1 in the Supplementary File.

Data Preparation
Glucosinolates were found to constitute the leading role in the analyzed L. peruvianum extracts. Based on the qualitative determination of the extracts' constituents presented in the Table 1, the further aim of the study was to compare the content of glucosinolates and to analyse their relative concentration in the extracts.
For that reason, the chemometric analyses of the presented dataset of L. peruviaum extracts were prepared in order to compare absolute (ABS) and relative (REL) glucosinolate compositions of the studied samples. For this purpose, the peak areas of nine compounds (G1-G9, see Table 1) were collected in triplicate from the recorded chromatograms of every extract injected into the HPLC-MS instrument. For further information on data preparation, please consult Materials and Methods (Section 3.4).

The "Absolute" Approach (ABS)
Principal component analysis (PCA) performed for the ABS dataset (Table S2 in (Table S4 in the Supplementary File).
PCA and the correlation analysis (Figures 2 and S4) have shown that the absolute extraction ratios of aromatic glucosinolates (G1, G2, G3) and aliphatic glucoalyssin (G8), constituting Dim1, were strongly correlated. This means that the content of G1, G2, and G3 is related to the content of G8. The higher content of the aromatic glucosinolates in the extracts of maca, the higher the amount of G8.  It is worth noting that the first dimension (Dim1), represented by a bimodal distribution, was responsible for the division of the dataset into two subsets of samples, which was also clearly demonstrated by the hierarchical clustering (HC, Figure 3). The obtained results classify the tested samples into two different groups-one of which collects juice samples obtained from fresh tubers, technologically altered gelatinized samples, maca tubers collected in the Ancash area, and yellow and purple tubers that seem to have different profiles from the remaining phenotypes-whereas the other collects large tubers of differently coloured phenotypes, mostly collected from Junin location.   Figure 3). Interestingly, gelatinized samples in general contained low amounts of aromatic glucosinolates and G8 (Dim1) in comparison with non-gelatinized tubers, except for BL_gel-AN (located at the middle of the Dim1 axis). However, they all exhibited similar, medium ratios of G4, lower than unprocessed samples.
As a raw material containing a large quantity of starch, in order to improve their nutritional properties, maca tubers are subjected to a pretreatment process-gelatinization. The results presented above show that this process, which takes place at an elevated temperature, induces thermal decomposition of glucosinolates because their content in gelatinized samples is lower in relation to the untreated samples.
In the study, the composition and content of glucosinolates was found to be colourand size-dependent. As previously mentioned, the above presented dendrogram ( Figure  3) classifies the analysed extracts into two major groups. Interestingly, a majority of the extracts from black, grey, and red phenotypes tend to show a similar composition, whereas the extracts from yellow and purple phenotypes are more distant from other samples in terms of glucosinolate content. Specifically, small, grey tubers (GR_D-J, GR_Z-J) contained higher amounts of indolyl glucosinolates, i.e., G4 and G9 (Dim2) than the rest of the tubers. Large red, grey, and black tubers (A, B, X) were poor in G4 and G9, in contrast to the small ones (for which the trend was not applicable to purple and yellow species). Red tubers in general contained low amounts of indolyl glucosinolates (Dim2), while all the red species (excluding RE_Bp-J) contained smaller quantities of glucosinolates and G8, constituting the Dim1. Red and purple tubers produced lower amounts of G6 and aliphatic glucosinolate G7 (Dim3). On the contrary, black tubers excelled in ratios of G6 and G7 produced, with the exclusion of juiced samples (BL_JU) and extra small phenotypes (BL_Z-J). Generally, large tubers seem to differ in composition in comparison to small tubers, which can be seen in the above dendrogram. The glucosinolates' profiles of yellow and purple phenotypes are alike. These results are in accordance with the study of Clement et al. [41], who calculated total glucosinolate content in their work on four phenotypes of maca. In their study, the grey phenotype was found to contain the highest amount of total glucosinolates (54.78 μmol/g d.w.), followed by the red (44.10 μmol/g d.w.), yellow (37.23 μmol/g d.w.), and purple (33.22 μmol/g d.w.) phenotypes. For that reason, the maca phenotypes tested herein are divided into two groups-one containing black, grey, and red tubers and another with yellow and purple tubers-seems to be in accordance with former observations on the total glucosinolates content. Linear maps of samples ( Figure 2), combined by the cluster dendrogram resulting from HC ( Figure 3), have revealed that juices contained very similar amounts of glucosinolates, with BL_JU (No 7) slightly differentiating from the rest ( Figure 3). These observations are in line with previous findings on the formation of glucosinolates in the plant tissues that underline that the process of oxidation is necessary to trigger the conformational changes in their structures [27]. Juices that were squeezed out from fresh tubers and immediately analysed did not contain high quantity of glucosinolates yet.
Moreover, the following samples collected in mid-October 2017: MIX_gel-AN, BL_gel-AN, BL_Bp-AN, RE_Bp-J, RE_gel-AN, RE_SM, RE_LA, and RE_LAgel contained similar amounts of glucosinolates ( Figure 3). Interestingly, gelatinized samples in general contained low amounts of aromatic glucosinolates and G8 (Dim1) in comparison with non-gelatinized tubers, except for BL_gel-AN (located at the middle of the Dim1 axis). However, they all exhibited similar, medium ratios of G4, lower than unprocessed samples.
As a raw material containing a large quantity of starch, in order to improve their nutritional properties, maca tubers are subjected to a pretreatment process-gelatinization. The results presented above show that this process, which takes place at an elevated temperature, induces thermal decomposition of glucosinolates because their content in gelatinized samples is lower in relation to the untreated samples.
In the study, the composition and content of glucosinolates was found to be colourand size-dependent. As previously mentioned, the above presented dendrogram (Figure 3) classifies the analysed extracts into two major groups. Interestingly, a majority of the extracts from black, grey, and red phenotypes tend to show a similar composition, whereas the extracts from yellow and purple phenotypes are more distant from other samples in terms of glucosinolate content. Specifically, small, grey tubers (GR_D-J, GR_Z-J) contained higher amounts of indolyl glucosinolates, i.e., G4 and G9 (Dim2) than the rest of the tubers. Large red, grey, and black tubers (A, B, X) were poor in G4 and G9, in contrast to the small ones (for which the trend was not applicable to purple and yellow species). Red tubers in general contained low amounts of indolyl glucosinolates (Dim2), while all the red species (excluding RE_Bp-J) contained smaller quantities of glucosinolates and G8, constituting the Dim1. Red and purple tubers produced lower amounts of G6 and aliphatic glucosinolate G7 (Dim3). On the contrary, black tubers excelled in ratios of G6 and G7 produced, with the exclusion of juiced samples (BL_JU) and extra small phenotypes (BL_Z-J). Generally, large tubers seem to differ in composition in comparison to small tubers, which can be seen in the above dendrogram. The glucosinolates' profiles of yellow and purple phenotypes are alike. These results are in accordance with the study of Clement et al. [41], who calculated total glucosinolate content in their work on four phenotypes of maca. In their study, the grey phenotype was found to contain the highest amount of total glucosinolates (54.78 µmol/g d.w.), followed by the red (44.10 µmol/g d.w.), yellow (37.23 µmol/g d.w.), and purple (33.22 µmol/g d.w.) phenotypes. For that reason, the maca phenotypes tested herein are divided into two groups-one containing black, grey, and red tubers and another with yellow and purple tubers-seems to be in accordance with former observations on the total glucosinolates content.
Finally, the chicha morada (CH) species was also poor in G1, G2, G3, and G8 (Dim1), probably due to the fact that they were not dried sufficiently at the place of collection.

The "Relative" Approach (REL)
According to the authors' knowledge, the relative approach to the content of glucosinolates in maca tubers was applied for the first time. This method of data analysis allows for a study on the relations between these metabolites in the maca extracts. Principal component analysis (PCA) performed for the REL dataset (Table S5 in the Supplementary File) also resulted in selection of three principal components (PCs) which explain 90.5% of the total variance (Table S6 in the Supplementary File). Correlation analysis, as well as the VARIMAX rotation, resulted in a linear map of variables in the space of three varivectors, further referred to as "dimensions" (Figures 4 and S5 in the Supplementary File). This revealed that the relative amounts of indolyl glucosinolates G5 and G6 were strongly yet reversely correlated with aromatic glucosinolates G2 and G3 and indolyl glucosinolate G8, while all five compounds constituted the first dimension of the REL dataset, Dim1. These observations show that the synthesis of the individual groups of glucosinolates shall engage different biosynthetic pathways which are dependent on one another.
Additionally, aliphatic glucosinolate G7 was reversely correlated with G2, G3, and G8, yet PCA revealed its stronger connections with indolyl glucosinolate G9 and aromatic glucosinolate G1. Therefore, G1, G7, and G9 defined the second dimension of the studied dataset, Dim2. The relative ratios of indolyl glucosinolate G4 solely constituted the third dimension of the REL matrix, Dim3, which means that the biosynthesis of this compound has no relations to other glucosinolates analysed in this work. This dataset is well explained by a three-dimensional set of factors since no original variables escaped into higher dimensions (Tables S6 and S7 in the Supplementary).
Further PCA and HC studies (Figures 4 and 5) have shown that the size of the tubers was correlated with the relative composition of the extracts, which additionally confirms the outcomes of the described above ABS approach.
The extracts that are located close to one another are characterized by similar relative proportions of glucosinolates.
Juices were observed to form a very separated subgroup within the REL dataset, with BL_JU being a "black sheep" of the juicy family, as it contains different proportions of the analysed glucosinolates as other samples. This differentiation was observed on the cluster dendrogram and the linear map of samples and mainly resulted from the first dimension (G2, G3, G5, G6, and G8). Gelatinization did not impact the relative composition of the glucosinolates within the black phenotype family, which explains that the eventual loss in the content of glucosinolates is the same for all compounds. However, the gelatinized MIX_gel-AN sample was a clear outlier in the REL dataset, substantially differentiating its relative composition from the rest of the extracts mainly due to its high Dim1 score (yet lower than juiced) and highest-by-far Dim2 score (G1, G7, and G9, Figure 4). A high Dim1 score indicates a marked relative concentration of G5 and G6, but low G2, G3, and G8, whereas the high Dim2 score of the sample is induced by increased relative contents of G1, G7, and G9 in comparison to other glucosinolates. Therefore, MIX_gel-AN was also the last sample to join the cluster dendrogram with a Euclidean distance above~13 ( Figure 5), which indicates that this sample exhibited extremely different relative composition in comparison to the others.    The REL dataset analysis shows that there are important correlations between certain groups of glucosinolates that may come from the similar, yet competing pathways of their biosynthesis in the plant.

Quantitative Determination of Glucotropaeolin Content in Lepidium peruvianum Extracts Using HPLC-ESI-QTOF-MS/MS
A comparative quantitative analysis of the glucotropaeolin (G2) content-the leading glucosinolate of the tested samples-was performed (Table 2) using the HPLC-MS methodology to visualize the existing differences between the analysed extracts. The analysis performed based on triplicate injections determined the content of this major glucosinolate in the 50% (v/v) alcoholic extract as ranging from 0 to 1.572%. Fresh juice samples did not contain G2, whereas its presence was confirmed in all remaining dried tuber samples.
Next, the relationship between the content of G2 and the color of tubers was investigated. For all five tested varieties of maca-black, purple, grey, yellow, and red-marked differences between the injections were noted. When analyzing the size B of all varieties, grey, red and black maca tubers were found to contain the highest quantity of this compound. The extract from black maca (BL_B-J) contained 0.965% of G2; from red maca (RE_B-J), 0.923%; whereas the extract from grey maca (GR_B-J) contained 1.010% (see Table 2). The smallest content of G2 was calculated for the purple variety: 0.239%. Interestingly, the level of this glucosinolate in the Chica morada (CH) variety was equal to 0.089%.
A drop in G2 concentration was noted for the gelatinized samples. Black gelatinized tubers (BL_gel-AN) contained 0.624% G2 (a drop from 0.965%), small gelatinized red tubers contained (RE _gel-AN) 0.103% G2 (without processing: 0.242%), and large red gelatinized maca tubers (RE_LAgel) 0.061% G2 (without processing: 0.344%). The REL dataset analysis shows that there are important correlations between certain groups of glucosinolates that may come from the similar, yet competing pathways of their biosynthesis in the plant.

Quantitative Determination of Glucotropaeolin Content in Lepidium peruvianum Extracts Using HPLC-ESI-QTOF-MS/MS
A comparative quantitative analysis of the glucotropaeolin (G2) content-the leading glucosinolate of the tested samples-was performed (Table 2) using the HPLC-MS methodology to visualize the existing differences between the analysed extracts. The analysis performed based on triplicate injections determined the content of this major glucosinolate in the 50% (v/v) alcoholic extract as ranging from 0 to 1.572%. Fresh juice samples did not contain G2, whereas its presence was confirmed in all remaining dried tuber samples.
Next, the relationship between the content of G2 and the color of tubers was investigated. For all five tested varieties of maca-black, purple, grey, yellow, and red-marked differences between the injections were noted. When analyzing the size B of all varieties, grey, red and black maca tubers were found to contain the highest quantity of this compound. The extract from black maca (BL_B-J) contained 0.965% of G2; from red maca (RE_B-J), 0.923%; whereas the extract from grey maca (GR_B-J) contained 1.010% (see Table 2). The smallest content of G2 was calculated for the purple variety: 0.239%. Interestingly, the level of this glucosinolate in the Chica morada (CH) variety was equal to 0.089%.
A drop in G2 concentration was noted for the gelatinized samples. Black gelatinized tubers (BL_gel-AN) contained 0.624% G2 (a drop from 0.965%), small gelatinized red tubers contained (RE _gel-AN) 0.103% G2 (without processing: 0.242%), and large red gelatinized maca tubers (RE_LAgel) 0.061% G2 (without processing: 0.344%). Previous studies discuss the content of glucosinolates in maca samples of different origins. It has been observed that batches from different producers and differently coloured phenotypes significantly differed in quantity of glucosinolates. According to Clement et al. [41], the content of glucotropaeolin was the highest in lead-coloured (grey) hypocotyls compared to red, yellow, and purple phenotypes, which is in accordance with the herein presented studies. Another study by the authors [42] confirms an important role of cultivation and cultivation conditions in the biosynthesis of secondary metabolites. The extracts from the same phenotypes of maca (lead-colored, red, yellow, and purple) were examined, but they differed with collection site. The content of G2 in the lead-coloured phenotype was calculated as 44.23-45.35 µmol/g of dry matter from the Patala area and only 2.37-2.89 µmol/g of dry matter from Alpacayan area. Additionally, our studies confirmed marked differences between the samples collected in the Junín and Ancash areas, with a higher content of these secondary metabolites from the former group.

Inhibitory Activity of Individual Glucosinolates and the Total Extract towards AChE and BuChE
The obtained results confirmed the inhibitory activity of the tested total extracts and the standards.
Concerning the results of anti-enzymatic activity exhibited by the reference compounds, G4 was found to be the strongest component. Glucobrassicin derivative showed the best inhibitory properties (enzyme inhibition value: 57.6%) towards AChE among the tested compounds in relation to BuChE, for which the calculated inhibition percentage of G4 was equal to 39.9% (see , Table 3). Then, in descending order, for AChE G1 > G3 > G2, with 38.9, 37.05, and 36.2 for BuChE and 56.9, 56.5, and 55.95 for AChE, respectively. It should be underlined that G4 was the leading enzyme inhibitor, whereas the following compounds were characterized by a similar inhibitory strength. The results obtained in the tests on reference compounds were set together with the outcomes of the anti-enzymatic assays performed for some selected total extracts. All samples were tested at the concentration of 10 mg/mL, so they can be directly compared with one another. It is important to note that, based on the results of reference compounds, the share of glucosinolates in the total enzyme inhibitory activity can be marked, especially for AChE enzymes. The AChE inhibitory activity of the total extracts ranges from approximately 15-25%, whereas the activities of single glucosinolates are within the range of 55-57%. Based on these observations, they can noticeably influence the total activity of the extract. In the case of BuChE, where the activity of the reference compounds is between 37-40%, the total inhibition rate of the extracts is higher (71-77%), so it must be influenced also by other components present in the extracts.
However, the quantitative data obtained in this study seem to correspond with the results of the enzymatic assay presented herein. Both AChE and BuChE inhibition decrease together with a decrease in size of tubers and with a decrease in the content of glucosinolates, as can be noted in the case of black maca (see Table 3).
Additionally, color-dependent variations were observed between the samples. Black, grey, and red tubers (size B) show a similar anti-enzymatic potential, whereas the yellow variety is both a weaker enzyme inhibitor and a weaker source of glucosinolates.
Rubio et al. reported that aqueous and hydroalcoholic extracts of Lepidium meyenii reduced activity of AChE in the brain by 45% after memory impairment caused by scopolamine [43]. Three different maca varieties (black, yellow, and red) were tested to evaluate the effects on memory and depression in ovariectomized mice. Of the three cultivars, the black maca variety showed the best activity [44].
L. peruvianum is rich in compounds called glucosinolates. It has been proven that glucosinolates have the ability to inhibit AChE in a dose-dependent manner. Blazevic et al. demonstrated that these secondary metabolites from Bunias erucago extracts, like gluconapin, glucoerucin, and glucoraphanin, inhibited the AChE enzyme by about 53% [45], whereas glucosinalbin did so by about 40.9% [25]. Recent studies also confirm that the products of glucosinolate hydrolysis, isothiocyanates, have AChE-inhibiting properties. Phenyl derivatives of isothiocyanates show promising AChE-inhibitory properties [24].
In the presented work, we used a standard method evaluating the AChE activity of newly characterized glucosinolates present in maca. Due to the use of unified methodology, we are able to make direct comparisons with previously characterized compounds. We must emphasize that the inhibitory activity of glucosinolates was within one order of magnitude (at a similar level) of a number of efficient cholinesterase inhibitors-terpenes and phenylpropanoids-previously studied by us in the very same manner [46]. Moreover, glucosinolates exerted much higher inhibitory activity than physodic acid isolated from Hypogymnia physodes, characterized in one of our recent works [47]. Although the cholinesterase-inhibiting activity of glucosinolates is lower than the activity of very efficient inhibitors-flavonoids (e.g., quercetin, rutin, or kaempferol) recently identified in selected fruits [48])-this activity can be judged applicable.
The results presented herein underline that next to the colour of Peruvian maca phenotype, the tuber size is an important factor in terms of glucosinolate content. This observation is crucial, as so far there is no specific method to classify tubers according to their size, and the cultivated tubers from a given plantation are collected together. The obtained results draw attention to the fact that in the context of glucosinolate content and possible preventive use of the tubers in memory impairment conditions, the important issue remains proper crop collection.

Modelling
The ligand-protein-binding energies found during docking simulations are given in Table 4 and graphically illustrated in Figure 6. Apart from compound 1-isothiocyanato-9-methanesulfinylnonane, which displays significantly distinct structure, the rest of compounds exhibit relatively similar magnitudes of binding energies. Independently of the considered protein, the strongest binding is always exhibited by compound G9, whereas the weakest is exhibited by 1-isothiocyanato-9-methanesulfinylnonane. All negative energy values clearly speak for favorable binding in all considered cases.
A series of interesting correlations is revealed when looking closer at calculated binding energies ( Figure 6). Firstly, the binding energies are fairly well but negatively correlated with the molecular sizes of docked ligands (expressed here as molecular volume). This suggests that the increased size of the -R substituent (see Figure 6) positively influences the binding strength and, therefore, inhibition potency. Secondly, the binding energies obtained for a series of X-ray crystal structures of human BuChE are strongly correlated with analogous values for equine BuChE structure obtained from homology modeling. This result is expected due to high sequence identity between these two proteins. Nevertheless, it also speaks for the correctness of the homology model. Finally, a good correlation is observed between binding energies obtained for AChE and BuChE. This also is expected due to the close nature of both enzymes and structural similarities in their binding sites. It is worth noting that in the last two cases, except for a good correlation, there exist only minor deviations in absolute binding energies determined for the same ligand and a given pair of proteins (on average, 0.35 and 0.57 kcal/mol, respectively). ing. This result is expected due to high sequence identity between these two proteins. Nevertheless, it also speaks for the correctness of the homology model. Finally, a good correlation is observed between binding energies obtained for AChE and BuChE. This also is expected due to the close nature of both enzymes and structural similarities in their binding sites. It is worth noting that in the last two cases, except for a good correlation, there exist only minor deviations in absolute binding energies determined for the same ligand and a given pair of proteins (on average, 0.35 and 0.57 kcal/mol, respectively). The results of the docking studies have been analysed with respect to the mechanistic interaction pattern that may be significant in the context of interpretation of the obtained binding energies. The summary given below relies on analyzing the ligand-protein contacts that take place if the distance between any corresponding atom pair is smaller than the arbitrarily accepted value of 0.4 nm. Figure 7 shows the superposition of all most favorable ligand poses with the most characteristic ligand-protein interactions. Let us remark that compound 1-isothiocyanato-9-methanesulfinylnonane, sharing little structural similarity with remaining compounds, was excluded from the discussion below. All ligands prefer roughly the same binding position in the enzyme cavity, which The selected amino acids creating the most essential ligand-protein contacts are given separately for each of ligand moieties. (E,F) The most favorable location of compound glucotropaeolin molecule bound to either the AChE (E) or BuChE (F) structure. The ligand molecule is shown as thick sticks, whereas all the closest amino acid residues (of distances no larger than 0.4 nm) are represented by thin sticks. The description of the interaction types is given in the text.
All ligands prefer roughly the same binding position in the enzyme cavity, which enables them to block catalytic sites (proximity to the tryptophan and histidine residues). The relative molecular orientation is (in majority of the cases) the same among compounds interacting with the same protein.
In the case of AChE, the -R substituent (usually of either aromatic or aliphatic character) interacts with the cluster of aromatic amino acid sidechains (Phe330, Phe331, and Tyr334). The corresponding interactions are of either p-p, CH-p stacking or simply hydrophobic nature. The opposite site of the ligand molecule (glucopyranose moiety) interacts with hydrogen-bonding-acceptor-containing amino acid residues (Glu199, Ser200, Tyr130, and His440) mainly by hydrogen bonding. Additionally, the CH-p stacking involving Trp84 and hydrophobic patch in the glucose ring is observed (This type of interaction is characteristic of non-charged carbohydrate-protein binding [49]. As indicated by the dispersion of the poses (including secondary and tertiary ones), the glucopyranosidic moiety is much more conformationally flexible in comparison to the -R substituent. Finally, the charged sulfate group, located in the central part of the ligand molecule, creates some attractive, hydrogen-bonding-mediated interactions with Ser122 and Asn85 (backbone fragment). As the sulfate moiety is located close to the entrance to the binding cavity, it is also possible that multiple interactions with polar water molecules are possible (as no explicit solvent was considered during the docking procedure, this has not been definitively confirmed).
In the case of BuChE, an important reorientation of the ligand in the binding cavity may be observed in comparison to AChE. Namely, the -R substituent is shifted toward Trp110 (structural analogue of Trp84 in Ache), interacting with it through p-p stacking. P-p interactions are possible also in the case of Phe357 but are probably significantly less intense due to increased R-Phe357 distance. At the same time, the glucose moiety exhibits hydrogen bonding with backbone fragments of Ser315, Gly145, and the side chain of His466 (structural analogue of His440 in AChE). Thus, the whole ligand molecule is slightly shifted toward the more polar edge of the binding cavity. The sulfate group is again located closest to the binding cavity entry and, thus, probably interacts with water molecules. Additionally, its hydrogen binding with Thr148 and Asp98 (backbone) were identified.
The observed differences between most favorable orientations of the docked ligands interacting either with AChE or BuChE can be ascribed to the diverse composition of the corresponding binding cavities. The cavity of BuChE lacks the cluster of aromatic residues Tyr-Phe-Phe which was responsible for favorable interactions with -R in the case of AChE. Therefore, reoriented -R moiety seeks attractive interactions with Trp110, whereas the glucose group can still interact with Ser226 and His466.
Finally, let us note that in the case of some compounds and-especially in secondary, tertiary, or quaternary poses-systematic reorientations of the ligand in the binding cavity were observed. Namely, the position of the -R substituent and glucopyranosidic group are swapped. In other words, AChE-characteristic poses become close to the BuChE ones and vice versa. Such an alternative interaction pattern is reasonable to accept, as the carbohydrate moiety can exhibit attractive interactions with either Trp or Phe/Tyr moieties. The same can be said about the aromatic -R substituent. Although the present docking study classifies such binding modes as only secondary ones, the accompanying binding energies are close to the most favorable values (difference by at most 1 kcal/mol). Thus, it cannot be excluded that in real systems, when the solvent and temperature effects are present, the ligand position can be a superposition of those two alternative orientations.

Plant Material and Extraction Conditions
Plant material studied in this paper represents dried tubers of Maca (Lepidium peruvianum syn. L. meyenii) collected in the highlands of the Peruvian Andes in the Junín plateau (2017 and 2019 harvesting seasons) and Ancash cultivation area (2017 harvest) as summarised in Table 5.
Maca cultivated in the Junín plateau was harvested on two occasions. Sampling J-17 was conducted on 1 October 2017 from plantation in Tarma Province, Tapo District, some 30 km west of Tarma city at 4174 m above sea level (a.s.l. ±20 m). Tubers were sampled from an area of approximately 250 m radius surrounding the geolocation coordinates 11 •    All L. peruvianum extracts used in the study were obtained with 50% aqueous ethanol [37] (v/v) by ultrasound-assisted extraction technique. Three extraction cycles of 30 min each were performed on 2 g of the same plant material using fresh portions of extracting solvent at room temperature. The selected solvent to solid ratio was 1:1 (v/v). After extraction, the extracts were joined and evaporated to dryness using a vacuum evaporator at 50 °C. The obtained residue was weighed and stored in a refrigerator at 5 °C until the analysis. Juice samples were produced by cutting fresh bulbs of L. peruvianum and placing them into a slow juicer. The collected samples were later centrifuged, filtered through a syringe filter (nominal pore size of 0.2 µm), and used for chromatographic analysis.     All L. peruvianum extracts used in the study were obtained with 50% aqueous ethanol [37] (v/v) by ultrasound-assisted extraction technique. Three extraction cycles of 30 min each were performed on 2 g of the same plant material using fresh portions of extracting solvent at room temperature. The selected solvent to solid ratio was 1:1 (v/v). After extraction, the extracts were joined and evaporated to dryness using a vacuum evaporator at 50 • C. The obtained residue was weighed and stored in a refrigerator at 5 • C until the analysis. Juice samples were produced by cutting fresh bulbs of L. peruvianum and placing them into a slow juicer. The collected samples were later centrifuged, filtered through a syringe filter (nominal pore size of 0.2 µm), and used for chromatographic analysis.

Qualitative and Quantitative Analysis of the Extracts by High Performance Liquid Chromatography Coupled with Mass Spectrometry (HPLC-ESI-QTOF-MS/MS)
The HPLC-ESI-QTOF-MS/MS platform from Agilent Technologies (Santa Clara, California, USA) was used to analyse the composition of L. peruvianum extracts. For the separation process, an HPLC chromatograph (1200 series) equipped with a Zorbax Eclipse Plus RP-18 chromatography column (150 mm × 2.1 mm; dp = 3.5 µm), a degasser (G1322A), a binary pump (G1312C), an autosampler (G1329B), a photodiode array detector (G1315D), and a mass spectrometer (G6530B) were used. Agilent MassHunter Workstation Software (version B.10.00) was used to acquire the MS spectra and process the data.
The temperature of the HPLC thermostat was set at 25 • C and the UV detection wavelengths at 254, 280, 320, and 365 nm. The operating wavelength range of UV-Vis/DAD detector was 190-600 nm. Chromatographic separation was performed using a 10 µL injection volume with a flow rate of 0.2 mL/min in a 45 min gradient elution program. Mobile phases consisted of eluent A (0.1% formic acid in water, v/v) and eluent B (acetonitrile solution with 0.1% formic acid added). The gradient elution was as follows: 0 min-1% of B, 4 min-20% of B, 20 min-60% B, 35-38 min-95% B, 39-45 min-1% of B.
The mass spectrometer measurements were carried out under the following conditions: gas temperature and shield gas temperature were 350 and 325 • C, respectively; gas flows were set at 12 L/min; capillary voltage at 3500 V; fragmentor voltage at 120 V; collision energies at 10 and 20 V; skimmer voltage at 65 V; nebulizer pressure at 35 psig. The collected spectra were scanned in the m/z 100-1700 Da range in the negative ionization mode that visualizes glucosinolates, as well as in the positive ionization mode. Two of the most intense signals seen in the TIC spectrum were automatically fragmented to obtain MS/MS spectra. After collecting two spectra for a given m/z value, the selected signals were excluded for 0.2 min from further fragmentation.

Absolute (ABS) Approach vs. Relative (REL) Approach
In the ABS approach, nine variables (nine compounds) representing the absolute detector responses, i.e., peak areas for a given compound, were autoscaled in order to enable standard procedures of exploratory analysis, i.e., principal component analysis (PCA) and hierarchical clustering (HC) ( Table S2 in the Supplementary File). In the REL approach, all peak areas for nine traced compounds of every given sample were summed to 100%, hence the amount of every compound was represented as a percentage share of a total detector response of a given sample (Table S5 in the Supplementary File). After such a percentage rescaling, all the variables were also autoscaled in a similar fashion to the ABS approach.

Principal Component Analysis (PCA)
After the standard, formal decomposition of the covariance matrices were calculated for the ABS (absolute) and REL (relative) autoscaled datasets, three principal components (PCs) were first considered relevant in both cases. After the selection of the relevant PCs, their vectors were rotated in space in order to maximize the values of correlation coefficients between the original variables and the three orthogonal factors using the VARIMAX algorithm. In every case, samples scores in the space of the resulting varivectors (dimensions) were calculated by multiplying the matrix of the original variables' loadings in the space of varivectors by the matrix of the autoscaled ABS/REL dataset.

Correlation Analysis
Matrices of the coefficients of linear correlation (R) were prepared using Pearson's definition of linear correlation.

Hierarchical Clustering (HC)
Hierarchical clustering (HC) for the ABS and REL approaches was performed using the Ward method based on a standard matrix of Euclidean distances between the samples.

Quantitative Analysis of Glucotropaeolin in the Analysed Extracts by HPLC-ESI-QTOF-MS/MS
Glucotropaeolin, as the most abundant glucosinolate in the tested extracts, was selected for the quantitative analysis to present the differences in its content between the tested extracts. The quantitative results were prepared for the standard compound (purity >98%, Extrasynthese, Genay, France) on the basis of its calibration curve, plotted for 7 different concentrations within the range of 0.0039-0.25 mg/mL that were obtained by diluting the stock solution with a concentration of 1 mg/mL. The same volumes (10 µL) of every solution were injected on the column and developed in the same above-described chromatographic method as for the qualitative assessment. The linearity range was observed for the concentration range of glucotropaeolin from 0.0039-0.0625 mg/mL. For these concentrations, the calculated calibration curve equation was y = 1,970,017,678x + 2,232,216 and the coefficient of determination (R 2 ) was equal to 0.99084.
The dried extracts from different varieties of L. peruvianum were redissolved in 50% ethanol (v/v) to the final concentration of 5 mg/mL, filtered through nylon syringe filters (nominal pore size 0.22 um), and subjected to chromatographic analysis in triplicate.

AChE-and BuChE Inhibition Assay
The selected total extracts (BL_A-J, BL_B-J, BL_C-J, BL_D-J, BL_Z-J, BL_B-J, RE_B-J, GR_B-J, YE_B-J) dissolved in DMSO at a concentration of 10 mg/mL and reference compounds (GL, GTP, GB, GS) at a concentration of 1 mg/mL. These samples were tested for their BuChE-and AChE-inhibitory activity according to the previously published protocol described below.
The assay was essentially performed as described by Szwajgier et al. (2021) [56]. AChE and BuChE were at a concentration of 0.28 U/mL, ATChI and BuTChI were at a concentration of 1.5 mmol/L, and DTNB (0.3 mmol/L) contained 11.89 mg DTNB, 58.44 mg NaCl, and 40.66 mg MgCl 2 ·6H 2 O, all per 100 mL of Tris buffer. The measurement of the absorbance at 405 nm was performed using a GloMax ® Explorer GM3500 plate reader (Promega, Madison, WI, USA) using the Software version 3.2.3. All samples were tested in triplicate. Three blank tests were also performed together with one false positive control for every samples. In the first blank test, the sample was replaced with 10 µL DMSO, the second contained 10 µL Tris instead of the sample; in the third, the enzyme and the sample were replaced with Tris buffer and "false-positive" blank sample according to Rhee et al. [57].

Molecular Docking of Glucosinolates to AChE and BuChE
The ligand molecules were drawn by using Avogadro 1.1.1 [58] and initially optimized within the UFF force field [59] (5000 steps, steepest descent algorithm). Flexible and optimized ligand molecules were docked into the binding pocket of the protein structures either found in the PDB database or prepared by using a homology modeling procedure. In the case of AChE, the 1EEA PDB entry (AChE from electric eel, X-ray resolution: 0.45 nm) was used. In the case of BuChE, no equine BuChE structure was deposited in PDB. Therefore, seven different human BuChE structures (PDB: 6EQP, 4BDS, 1XLW, 4XII, 6ESJ, 6ESY, and 6QAC of resolutions equal to 0.235, 0.210, 0.210, 0.270, 0.298, 0.280, and 0.277 nm, respectively) and, additionally, a homology model prepared on the basis of the highest resolution 4BDS structure, were used. Note that the equine BuChE has~90% amino acid similarity with the human sequence [60] and nearly all differences are located outside binding cavity. The homology model was prepared by using the SWISS-MODEL server [61,62] and all inherent, default procedures.
Docking simulations were carried out in the AutoDock Vina software [63]. The procedure was performed within the cubic region of dimensions of 18 × 18 × 18 Å 3 , which covers all the originally cocrystallized ligands present in the considered PDB structures as well as the closest amino acid residues which exhibit contact with those ligands. All the default procedures and algorithms implemented in AutoDock Vina were applied during the docking procedure. The torsional angles in ligand molecules as well as the selected amino acid sidechains within the binding cavities of both proteins were allowed to rotate. In the case of human BuChE, the predicted binding energies were averaged over all seven protein structures. The visual inspections of each pose of the docked ligand (after aligning with respect to the given protein structure) were carried out in order to assure that the partial energies contributing the final, averaged, binding energy value, correspond to the structurally analogous orientations. Only the lowest energy values, corresponding to the most energetically favorable complexes, were considered during subsequent analysis. The procedure was initially validated by docking the ligands cocrystalized with proteins and present in the 6EQP and 1E3Q (AChE from Torpedo californica). The graphical illustration of the validation results is given in Figure 7.

Conclusions
The results presented above provide evidence of the existence of substantial antienzymatic activity of several well-identified maca compounds against AChE and BuChE, thus confirming that L. peruvianum is a rich source of active glucosinolates, thus showing promise for possible application of selected maca phenotypes in developing therapeutic products for treating medical conditions related to memory impairment. It was proved that the potential of the plant is related both to the phenotype (color) and to the size of Peruvian maca tubers.