LC-MS Based Phytochemical Profiling towards the Identification of Antioxidant Markers in Some Endemic Aloe Species from Mascarene Islands

Aloe plant species have been used for centuries in traditional medicine and are reported to be an important source of natural products. However, despite the large number of species within the Aloe genus, only a few have been investigated chemotaxonomically. A Molecular Network approach was used to highlight the different chemical classes characterizing the leaves of five Aloe species: Aloe macra, Aloe vera, Aloe tormentorii, Aloe ferox, and Aloe purpurea. Aloe macra, A. tormentorii, and A. purpurea are endemic from the Mascarene Islands comprising Reunion, Mauritius, and Rodrigues. UHPLC-MS/MS analysis followed by a dereplication process allowed the characterization of 93 metabolites. The newly developed MolNotator algorithm was usedfor molecular networking and allowed a better exploration of the Aloe metabolome chemodiversity. The five species appeared rich in polyphenols (anthracene derivatives, flavonoids, phenolic acids). Therefore, the total phenolic content and antioxidant activity of the five species were evaluated, and a DPPH-On-Line-HPLC assay was used to determine the metabolites responsible for the radical scavenging activity. The use of computational tools allowed a better description of the comparative phytochemical profiling of five Aloe species, which showed differences in their metabolite composition, both qualitative and quantitative. Moreover, the molecular network approach combined with the On-Line-HPLC assay allowed the identification of 9 metabolites responsible for the antioxidant activity. Two of them, aloeresin A and coumaroylaloesin, could be the principal metabolites responsible for the activity. From 374 metabolites calculated by MolNator, 93 could be characterized. Therefore, the Aloe species can be a rich source of new chemical structures that need to be discovered.


Introduction
The Aloe genus counts over 500 species, mainly distributed in arid areas, predominantly in Africa, but also in India [1]. Aloe species are members of the Asphodelaceae family, and the studied species belong to the same subfamily Alooidae. Aloe vera (L.) Burm. f. (synonym A. barbadensis Mill.) and Aloe ferox Mill. (synonym Cape Aloe), are two of the most well-known species. Within the Alooidae subfamily, the Aloe section Lomatophyllum (Willd.) G.D. Rowley includes species such as Aloe tormentorii (Marais) L.E.

UHPLC-MS/MS Analysis
The LC-MS/MS analysis was performed on a Thermo Scientific Dionex 3000 Ultra High-Performance Liquid Chromatography system (UHPLC) coupled to a Bruker Impact II Q-TOF high-resolution mass spectrometer equipped with an electrospray ionization source (ESI). The chromatographic separation was carried on an Agilent Zorbax Eclipse Plus C18 column (2.1 × 100 mm, 1.8 µm) at 43 • C. Ultrapure water (A) (LC-MS grade, Carlo Erba, Italy) and acetonitrile (B) (LC-MS grade, Carlo Erba, Italy), both acidified with 0.1% formic acid (LC-MS grade, Carlo Erba, Italy), were used as mobile phases. The injection volume was 1 µL for all samples, and the flow rate of the mobile phase was 0.8 mL/min. The following gradient was applied: isocratic hold at 5% B for 2 min, 5-50% B over 2-17 min, 50-100% B over 17-27 min, then isocratic hold at 100% B for 2 min (27-29 min), followed by a decrease to 5% B in 1 min (29-30 min), held at 5% B over 30-33 min for the column equilibration for the next experiment.
Mass spectrometry data were acquired from m/z 50 to 1200, using both positive (+) and negative (-) modes. The following parameters were used for the Q-TOF in both ionization modes: end plate offset at 500 V; nebulizer gas (N 2 ) pressure at 3.5 bar; dry gas flow (N 2 ) at 12 L/min; drying temperature at 200 • C; acquisition rate at 4 Hz. The capillary voltage was set at 3500 V for positive mode and 3000 V for negative mode.
A data-dependent acquisition (DDA) protocol was used; therefore, MS/MS fragmentation spectra were obtained automatically for the three most abundant precursor ions using mixed collision energy 20-40 eV (in stepping mode).
A solution of Sodium Formate Acetate was used as a calibration to obtain high mass accuracy (2-5 ppm) and was automatically injected at the beginning of each run.
However, the issue regarding a non-calibrated precursor ion value in .mzXML, highlighted by Zdouc et al. in 2021, persists in Bruker DataAnalysis software [10], which means that the export of raw data sets results in calibrated MS/MS data, with non-calibrated precursor ion information (<precursorMz></precursorMz>) for each MS/MS scan. Zdouc et al. suggested using a script in DataAnalysis software to export to .mgf files, which contain calibrated precursor ion information, and a Perl5 script to correct precursor information in the .mzXML data. The Perl5 script could not be used in the present work because of the difference between Bruker Compass versions used for the acquisition of raw data. Therefore, the Perl5 script was adapted to be used with Bruker Compass version 5.0 SR1 64-bit. It was also adjusted to correct the data exported with GNPS Vendor in 32 and 64 bits. The script is freely available on https://github.com/elnurgar/dataanalysis.git (accessed on 13 November 2022).

Data Processing
Exported .mzXML data were pre-processed using MZmine software, version 3.0. The processing workflow includes raw data file import, mass detection, chromatogram building, chromatogram deconvolution, feature list deisotoping, alignment between analytical replicates, filtering, and plant species [8,11,12]. The features present in blank methanol runs were removed from the features list. Parameters of each step used for processing can be seen in Table S1 of Supplementary Materials.
Processed data were exported (mgf and CSV files) and applied to MolNotator software. First, the AdNotator module allows the identification of the adducts of a compound, and FragNotator identifies the in-source fragments. Then, the MolNet module allows the construction of a molecular network [13]. The Adnotator precursor and fragments ion mass tolerance were set to 0.002 and the retention time tolerance to 6 s. Detailed parameters used for MolNotator software are presented in the Supplementary Materials. Cytoscape 3.9.1 software was used to visualize the resulting network.
The table with 374 metabolites and their peak areas in the studied Aloe species, generated by MolNotator, was submitted on Metaboanalyst 5.0 platform for hierarchical cluster analysis in order to produce a dendrogram illustrating molecular similarity between five Aloe species [14].

Total Phenolics Content (TPC)
The Folin-Ciocalteu method was adapted from El Hosry et al. [15]. The previously dried hydro-ethanolic extracts were prepared at 3 mg/mL in EtOH 50% (v/v). A volume of 5 mL of the prepared solutions was mixed with 1 mL of Folin-Ciocalteu reagent (Sigma Aldrich, lot BCBP2077V) and 4 mL of Na 2 CO 3 7.5% (m/v) (Fluka Biochemika, 347579/1 596 lot.71345) in a 100 mL volumetric flask. The volume was completed with distilled water. Samples were incubated in an oven for 2h30 at 30 • C in the dark. The absorbance of the solutions was measured at 760 nm using a UV/Vis spectrophotometer (Thermo Scientific Genesys 10S UV-Vis). TPC was expressed as g of gallic acid equivalent (GAE) per 100 g of extracts.
Gallic acid (Extrasynthese, 6079 lot.04900102), used as a positive control, was solubilized in MeOH and diluted to obtain concentrations in the range of 0.5-5 µg/mL.
A fresh DPPH methanolic solution at a concentration of 10 −4 M was prepared every day by dissolving 10 mg of DDPH (Sigma Aldrich, D9132-5G lot.STBD2362V) in 250 mL MeOH and kept at room temperature in the dark for 3 h before use.
Gallic acid (Extrasynthese, 6079 lot.04900102), used as a positive control, was solubilized in MeOH and diluted to obtain concentrations in the range of 0.5-5 µg/mL. A fresh DPPH methanolic solution at a concentration of 10 −4 M was prepared every day by dissolving 10 mg of DDPH (Sigma Aldrich, D9132-5G lot.STBD2362V) in 250 mL MeOH and kept at room temperature in the dark for 3 h before use.
DPPH assay was carried out in 96-well plates (Sterilin Ltd., Newport UK) with one blank row, one negative control row, three columns with sample solutions at different concentrations and in triplicates and one column of sample blank solution. The 96-well plate layout can be seen in Figure   The 96-well plate was placed in the spectrophotometer (BioTek EON, USA) and was incubated for 1 h at 25 °C. Absorbance was then read at 515 nm. The scavenging activity (%) was expressed as EC50 (concentration corresponding to 50% inhibition).
Statistical analysis was performed by ordinary one-way ANOVA test followed by Dunnett's multiple comparisons tests.

On-Line RP-HPLC-DPPH
A rapid On-Line method allowing targeting compounds with radical scavenging activity in complex mixtures was realized according to Koleva et al. [17]. A scheme of the online system is given in Figure 2. The HPLC Agilent 1260 system coupled with Agilent 1200 consisted of the following: The 96-well plate was placed in the spectrophotometer (BioTek EON, Providence, RI, USA) and was incubated for 1 h at 25 • C. Absorbance was then read at 515 nm. The scavenging activity (%) was expressed as EC 50 (concentration corresponding to 50% inhibition).
Statistical analysis was performed by ordinary one-way ANOVA test followed by Dunnett's multiple comparisons tests.

On-Line RP-HPLC-DPPH
A rapid On-Line method allowing targeting compounds with radical scavenging activity in complex mixtures was realized according to Koleva et al. [17]. A scheme of the online system is given in Figure 2. The HPLC Agilent 1260 system coupled with Agilent 1200 consisted of the following:
• Agilent 1200: a second HPLC pump (quaternary pump G1311A) for the delivery of the DPPH solution; a DAD UV-Vis detector (DAD G1315B).
Chromatographic separation was carried out on an Agilent Zorbax Eclipse Plus C18 column (2.1 × 100 mm, 1.8 µm). The reaction coil was a 10 m × 0.25 mm i.d. stainless steel tube. The UV detection wavelength for the tested samples was set at 325 nm. Detection of DPPH solution bleaching was carried out at 515 nm. Dried hydro-ethanolic extracts were prepared at a concentration of 10 mg/mL in EtOH 70% (v/v), then filtered using 0.22 µm PTFE filters into glass vials.
A fresh DPPH methanolic solution at a concentration of 2.10 −4 M was prepared every day by dissolving 40 mg of DDPH in 500 mL MeOH and kept at room temperature in the dark for a minimum of 3 h before use. This allowed the stabilization of the solution absorbance.
All solvents used were of HPLC grade. Ultrapure water (A) and acetonitrile (B) (Carlo Erba, Italy), both acidified with 0.1% formic acid (Carlo Erba, Italy), were used as the mobile phase. The injection volume of all samples was 2 µL, and separation was carried at 43 °C, with a mobile phase flow of 0.2 mL/min. The following gradient was applied: isocratic hold at 5% B for 2 min, 5-50% B over 2-17 min, 50-100% B in 1 min (17-18 min), then isocratic hold at 100% B for 2 min (18-20 min), followed by a decrease to 5% B in 0.1 min (20-20.1 min), held at 5% B over 20.1-30 min for the column's equilibration for the next experiment.
The second HPLC pump delivering the DPPH at 2.10 −4 M in methanol solution into the reaction coil was set at a flow of 0.2 mL/min. The reaction coil temperature was set at 60 °C. The used flow rate and the reaction coil dimensions allowed a reaction time of 1 min 14 s between samples and the DDPH solution.
LC-MS/MS analysis was carried out using the same analytical method as On-Line DPPH for the annotation of peaks presenting radical scavenging activity. The mass spectrometer parameters used were the same as in 2.3. UHPLC-MS/MS analysis. Chromatographic separation was carried out on an Agilent Zorbax Eclipse Plus C18 column (2.1 × 100 mm, 1.8 µm). The reaction coil was a 10 m × 0.25 mm i.d. stainless steel tube. The UV detection wavelength for the tested samples was set at 325 nm. Detection of DPPH solution bleaching was carried out at 515 nm.

Molecular Network and Chemotaxonomic Study
Dried hydro-ethanolic extracts were prepared at a concentration of 10 mg/mL in EtOH 70% (v/v), then filtered using 0.22 µm PTFE filters into glass vials.
A fresh DPPH methanolic solution at a concentration of 2 × 10 −4 M was prepared every day by dissolving 40 mg of DDPH in 500 mL MeOH and kept at room temperature in the dark for a minimum of 3 h before use. This allowed the stabilization of the solution absorbance.
All solvents used were of HPLC grade. Ultrapure water (A) and acetonitrile (B) (Carlo Erba, Italy), both acidified with 0.1% formic acid (Carlo Erba, Italy), were used as the mobile phase. The injection volume of all samples was 2 µL, and separation was carried at 43 • C, with a mobile phase flow of 0.2 mL/min. The following gradient was applied: isocratic hold at 5% B for 2 min, 5-50% B over 2-17 min, 50-100% B in 1 min (17-18 min), then isocratic hold at 100% B for 2 min (18-20 min), followed by a decrease to 5% B in 0.1 min (20-20.1 min), held at 5% B over 20.1-30 min for the column's equilibration for the next experiment.
The second HPLC pump delivering the DPPH at 2 × 10 −4 M in methanol solution into the reaction coil was set at a flow of 0.2 mL/min. The reaction coil temperature was set at 60 • C. The used flow rate and the reaction coil dimensions allowed a reaction time of 1 min 14 s between samples and the DDPH solution.
LC-MS/MS analysis was carried out using the same analytical method as On-Line DPPH for the annotation of peaks presenting radical scavenging activity. The mass spectrometer parameters used were the same as in Section 2.3.

Molecular Network and Chemotaxonomic Study
The use of a Molecular Network approach was chosen to explore the phytochemical composition of the five Aloe species. The leaves of five Aloe species collected at Reunion Island have been extracted with 70% ethanol. Metabolomics spectral data from the hydroethanolic leaf extracts of the Aloe species were acquired in both ionization modes (ESI +/−) using an LC-MS quadrupole time-of-flight (Q-TOF).
Raw data sets were converted into open .mzXML format and processed using MZmine. The final features lists contained 1370 features in positive mode and 1596 in negative mode. The use of MolNotator allowed us to identify the different adducts generated during the Antioxidants 2023, 12, 50 7 of 25 ionization process using a triangulation method and predict the neutral metabolites [13]. The algorithm also allowed the identification of in-source fragments with high efficiency, therefore decreasing the number of false positives. The MolNotator algorithm proposed 374 metabolites out of the 2966 chemical features.
Based on the list of the 374 metabolites, a hierarchical clustering analysis was generated using the Metaboanalyst platform and can be found in Figure 3A. The MolNotator molecular network was generated based on cosine similarity between different MS/MS spectra of predicted metabolites ( Figure 3B).
composition of the five Aloe species. The leaves of five Aloe species collected at Reunion Island have been extracted with 70% ethanol. Metabolomics spectral data from the hydroethanolic leaf extracts of the Aloe species were acquired in both ionization modes (ESI +/−) using an LC-MS quadrupole time-of-flight (Q-TOF).
Raw data sets were converted into open. mzXML format and processed using MZmine. The final features lists contained 1370 features in positive mode and 1596 in negative mode. The use of MolNotator allowed us to identify the different adducts generated during the ionization process using a triangulation method and predict the neutral metabolites [13]. The algorithm also allowed the identification of in-source fragments with high efficiency, therefore decreasing the number of false positives. The MolNotator algorithm proposed 374 metabolites out of the 2966 chemical features.
Based on the list of the 374 metabolites, a hierarchical clustering analysis was generated using the Metaboanalyst platform and can be found in Figure 3A. The MolNotator molecular network was generated based on cosine similarity between different MS/MS spectra of predicted metabolites ( Figure 3B).  The visualization of the metabolome through a molecular network approach revealed structurally related chemical families in the five Aloe species. Visual exploration of the network showed that MS/MS spectra were grouped according to their chemical class.
The automatized dereplication process combined with manual annotation allowed for the putative and full identification of 94 metabolites, including metabolites so far undescribed in the Aloe genus. In addition, chemotaxonomic characteristics reported for the Aloe genus helped to improve the structural annotation process [6,18].
Identified compounds were classified following the levels of confidence proposed by Schymanski et al.: level 1 (L1): structure confirmed by the reference standard with MS, MS/MS spectra, and retention time matching; level 2a (L2a): probable structure using library spectrum match or literature match; level 2b (L2b): diagnostic of structure using MS/MS fragments or ionization behavior, with no literature confirmation; level 3 (L3): tentative candidates with uncertainties (for example positional isomers) [19].
All annotated metabolites can be found in Table 1. Spectral data acquired in both positive and negative ionization modes were used to annotate the Aloe species metabolomes.

Major Chemical Classes of the Aloe Species Metabolome
Three main families of metabolites could be described in the five Aloe species. Firstly, a range of phenylpropanoids (C6-C3), more precisely cinnamic acid derivatives, such as chlorogenic acids, were observed in the extracts and can easily be recognized by the presence of a characteristic caffeoyl substituent (fragment ion at m/z 163.0390 in positive mode). Likewise, coumaroylquinic acid and feruloylquinic acids and their derivatives presented a fragment ion at m/z 147.0440 and m/z 177.0550 in positive mode. They are all strongly linked with cosine scores above 0.95, suggesting their similar fragmentation patterns.
Coumaroylquinic acid and its derivatives are ubiquitous, as they were found in all five Aloe species. The three naturally occurring isomers of caffeoylquinic acid were observed in many Aloe species [26,38]. Chlorogenic acid (5-O-caffeoylquinic acid) (8) was identified by comparing its retention time and MS/MS spectrum with a commercial standard. The two other mono-caffeoylquinic acid isomers were also detected (5; 13) and annotated through their characteristic MS/MS fragmentation patterns [23]. The distribution of described isomers was found to be species-dependent: chlorogenic acid (8) was found only in A. vera, while the two other isomers, neochlorogenic acid (3-O-caffeoylquinic acid) (5) and cryptochlorogenic acid (4-O-caffeoylquinic acid) (12) were mostly found in A. ferox, A. macra, and A. purpurea. As opposed to the other investigated Aloe species, A. tormentorii contained few cinnamic acid derivatives: only a glycosylated derivative of coumaric acid was annotated (9).
Secondly, the five studied species were rich in flavonoids, which are biosynthesized through the phenylpropanoid pathway. Various subclasses can be found in plants due to the action of reductases, isomerases, dioxygenases, and hydrolases. Thus, flavonoids can be found with structurally diverse aglycone backbones, namely chalcones, flavones, isoflavones, flavanols, flavonols, flavanones, and anthocyanidins. These backbones exist in various modified forms through hydroxylation, methylation, and glycosylation by transferases.
The loss of a sugar attachment represents the main MS fragmentation path for flavonoids: a neutral loss of 162.0528 amu represents a loss of a hexose, while the neutral loss of 132.119 characterizes a pentose.
In the studied Aloe species, flavones were the most frequent subclass of flavonoids: apigenin and luteolin were found with diverse sugar attachments and in many isomeric forms. C-glycosyl flavones, such as isoorientin and isovitexin, showed to be especially present. The second subclass that could be found was flavonols, such as kaempferol. No free flavone or flavonol aglycones were detected in the studied Aloe species.
Lastly, several lipids were identified: phospholipids, more precisely lysophosphatidylcholine (LPC) derivatives, were found in all five species. LPC, also called lysolecithins, is a class of lipids resulting from the cleavage of phosphatidylcholine (PC) via the action of phospholipase A 2 (PLA 2 ) and/or the transfer of fatty acids to free cholesterol via lecithin-cholesterol acyltransferase (LCAT).
Two fatty acids, linoleic acid (92) and α-linolenic acid (85) were also annotated in all studied species and are widely described in plants as they contribute to the integrity of the cellular membrane [39].

Chemotaxonomic Exploration of the Chemical Composition of the Five Aloe Species
From 374 metabolites found, 241 metabolites (64.4%) were common to two or more species, and 133 metabolites (35.6%) were specific to a species.
Among the metabolites present in all studied Aloe species, major compounds previously described in the Aloe genus could be highlighted: aloin A (61) and B (58), which are members of the anthracene chemical group [41]. Aloesin (10) and aloeresin A (68), members of the chromone family, were also detected. These four metabolites are largely described in the literature as specific to the Aloe genus [6,22,42].
Aloe macra presented the richest chemical diversity, as it contains 219 metabolites in total, 26 being specific. Aloe purpurea presented the second chemical diversity with 209 metabolites in total. However, only 19 of them were specific to A. purpurea. Aloe vera contains 190 metabolites, but 55 were uniquely found in the species. Amongst the five studies species, A. vera showed the most specific metabolites. Aloe ferox contains 163 metabolites, 27 specific to the species. Aloe tormentorii appeared to have the lowest metabolic diversity, with 157 metabolites and only six specific ones.
The highest specificity of metabolites observed for A. vera and A. ferox can be explained by the fact that these species do not belong to the Lomatophyllum section. These results show that Mascarene Aloe species, A. purpurea, A. tormentorii, and A. macra possess a different metabolome compared to other species of the genus Aloe.
Hierarchical clustering analysis, generated by the Metaboanalyst platform ( Figure 3A), demonstrated the differences between the metabolic fingerprints of A. vera and A. ferox compared to those characterizing the Aloe species uniquely found in the Mascarene islands. Within the Mascarene Aloe species, the metabolomic fingerprints were closer between A. tormentorii and A. purpurea compared to A. macra, which is endemic to Reunion Island [43]. These results are also in agreement with the study from Ranghoo-Sanmukhiya et al. [2], where genetic similarities were determined between the same Mascarene Aloe species: A. purpurea, A. tormentorii, A. macra compared to A. vera. The authors showed that A. purpurea and A. tormentorii share more genetic similarities with A. macra than A. vera. Herein, we corroborated such classification from a metabolomic point of view.
A detailed exploration of the molecular network highlighted that the major and most abundant metabolites of the Aloe genus could be found at the center of the main cluster. Aloin A (61) and B (58), aloesin (10), and aloeresin (coumaroylaloesin) (68) form the main cluster with their derivatives.
A number of malonylnataloin derivatives were also detected and can be seen at the top of the network. Only two of them seemed to be specific to A. purpurea.
One of the known major metabolites, aloesin (10), was detected in all five species. Interestingly, some metabolites in the aloesin derivatives cluster were mostly detected in A. macra.
Aloe vera appeared richer in coumaroylaloesin derivatives, which are almost entirely specific to this species (except for one metabolite that could also be found in A. ferox). Two other chromones, isoaloeresin D (59) and isorabaichromone (55), were only present in A. vera.
It appears that the main cluster links a range of different chemical classes, but all belong to the phenolic class. Phenolic acids, and more precisely, cinnamic acid derivatives, can be found in the middle of the cluster, between aloesin and aloin derivatives. They form the denser part of the cluster, with a great number of nodes, and are strongly linked to Aloin and its derivatives.
Flavonols and flavones can easily be identified on the network, as they form two subclusters. Visualization of the chemical space of the five Aloe species through a molecular network pointed out that A. macra contain a larger number of metabolites, while A. vera contains a lower diversity of metabolites, but more specific ones: for example, flavonols such as kaempferol-3-glucoside, also named astragalin (51), were found strictly in A. vera, and kaempferol-3-O-rutinoside (48) was found in A. vera, with traces in A. ferox.
Flavones were distributed across all five studied species. However, apigenin-Chexoside-O-hexoside (28) was found only in A. vera, and narcissin or narcissoside (52) only in A. ferox. This can indicate the presence of a specific flavonoid biosynthesis pathway for A. vera.
Aloe tormentorii showed no specific subcluster and appeared to contain less diverse metabolites than the four other species.
Lipids identified through annotation were not found as part of a cluster on the network and are distributed among the single nodes.

Total Phenolic Content (TPC)
From the 93 annotated metabolites, more than 2/3 belong to the polyphenols class. Therefore, the total phenolic content was evaluated for hydro-ethanolic extracts of the Aloe species.
The total phenolic contents of the five species, expressed in gallic acid equivalent, were determined according to the Folin-Ciocalteu method [15]. Phenolic contents ranged from 1.1 to 2.8 g GAE/100 g extract (Table 2), with the highest TPC for A. purpurea. Higher TPC is often correlated with higher radical scavenging activity, which was evaluated in this study using the DPPH assay.

DPPH Assay in 96-Well Plates
This method was developed by Blois to determine the antioxidant activity of compounds using a stable free radical α,α-diphenyl-β-picrylhydrazyl (DPPH) [16]. The assay measures the scavenging capacity of antioxidant compounds towards DPPH. The single electron of the DPPH's nitrogen atom is reduced by receiving a hydrogen atom from the scavenging compound, forming the corresponding hydrazine [44].
The 96-well plate assay was realized as a preliminary step to determine whether or not extracts possessed a DPPH scavenging activity. EC 50 is defined here as the concentration of substrate that causes a 50% reduction in the DPPH absorbance at 515 nm. The lower the EC 50 , the higher the scavenging activity. In our study, the extracts with the EC 50 below 200 µg/mL were considered active. Three extracts showed antioxidant activity: A. macra (EC 50 = 172 µg/mL), A. ferox (EC 50 = 151 µg/mL), with the best result obtained from A. purpurea hydro-ethanolic extract with an EC 50 of 88 µg/mL. The extracts of A. vera and A. tormentorii with an EC 50 superior to 200 µg/mL were considered inactive. The statistical analysis shows that results are significant for all Aloe extracts with a p-value ≤ 0.05, excluding the results for Aloe purpurea extract.
It is to be noted that results between TPC and DPPH assays were not directly correlated: this can be explained by the diversity in the polyphenolic compounds composition of the extracts, the presence of groups on phenolics that can interfere with the colorimetric reaction of TPC assay, or the presence of compounds in extracts that may also act as false positives. However, two species, A. macra and A. purpurea, with the highest metabolite diversity, were richest in TPC, with relatively high antioxidant DPPH activity. The same tendency regarding the activity of Mascarene Aloe species is observed in the previous study conducted by Govinden-Soulange et al., where A. macra (from Reunion Island) and A. purpurea were more active than A. vera and A. tormentorii [45]. In order to determine which metabolites contribute the most to the radical scavenging activity, the 96-well plates DPPH assay was followed by the On-Line RP HPLC DPPH assay.

On-Line RP HPLC DPPH Assay
The method described by Koleva et al. can be applied to complex mixtures such as plant extracts and/or fractions for rapid detection of radical scavenging components [17].
Such a method was applied to hydro-ethanolic extracts of all five studied species in triplicates. Combined UV and DPPH bleaching Visible (Vis) chromatograms of active extracts can be seen in Figure 4. The chromatograms of inactive extracts are presented in Supplementary Materials. Aloe vera and A. tormentorii, both inactive extracts, showed no decrease in absorbance on the Vis chromatogram, which means that none of the separated compounds within the extracts induced bleaching of the DPPH solution. This could mean that (1) compounds present in A. vera and A. tormentorii possess low to no antioxidant properties, (2) their concentration in the extracts is too low to be effective, and (3) compounds act antagonistically as antioxidants.

On-Line RP HPLC DPPH Assay
The method described by Koleva et al. can be applied to complex mixtures such as plant extracts and/or fractions for rapid detection of radical scavenging components [17].
Such a method was applied to hydro-ethanolic extracts of all five studied species in triplicates. Combined UV and DPPH bleaching Visible (Vis) chromatograms of active extracts can be seen in Figure 4. The chromatograms of inactive extracts are presented in Supplementary Materials. Aloe vera and A. tormentorii, both inactive extracts, showed no decrease in absorbance on the Vis chromatogram, which means that none of the separated compounds within the extracts induced bleaching of the DPPH solution. This could mean that (1) compounds present in A. vera and A. tormentorii possess low to no antioxidant properties, (2) their concentration in the extracts is too low to be effective, and (3) compounds act antagonistically as antioxidants.
On the other hand, the extracts that had antioxidant activity in 96-well plates also showed radical scavenging properties with the On-Line approach. A. macra, A. ferox, and A. purpurea showed 5 to 6 negative peaks on the DPPH Vis 515 nm chromatogram. On the other hand, the extracts that had antioxidant activity in 96-well plates also showed radical scavenging properties with the On-Line approach. A. macra, A. ferox, and A. purpurea showed 5 to 6 negative peaks on the DPPH Vis 515 nm chromatogram.
The aim of the On-Line RP HPLC DPPH assay was to identify compounds responsible for the antioxidant properties of the extracts.
The extracts were also analyzed by LC-MS/MS using the same analytical method as On-Line DPPH. The acquired MS data led to the annotation of active metabolites by comparing their MS/MS spectra to those from previously annotated metabolites (Section 2.1).
That way, aloesin (10), aloeresin A (2-O-p-coumaroylaloesin) (68), 2 -O-trans -p-coumaroylaloenin (59), two caffeoylquinic acid isomers (5; 12), luteolin-C-glucoside-Opentoside (30), isoorientin (31) and one new compound, yet to be annotated, and codified, were identified as metabolites responsible for the radical scavenging properties. This new metabolite was found to be uniquely detected in A. macra and A. purpurea, underlining the potential of these species for the discovery of new bioactive compounds. However, most of the metabolites identified as responsible for the activity were not specific to the active species, namely A. purpurea, A. ferox, and A. macra, as they were present in all five species.
The quantitative factor may play an important role in the antioxidant properties of an extract. Our hypothesis is that metabolites identified as responsible for the radical scavenging properties can be found in different amounts and proportions across the five studied species. In order to confirm that the proportion of antioxidant compounds in active extracts is higher than in the other extracts, their peak heights were compared. The results, presented in Table 3, show a good correlation between the peak height of an antioxidant compound detected at 325 nm and the activity of the extracts. Even though some compounds responsible for the radical scavenging activity are present in inactive extracts, their content is too low to induce a decrease in the DPPH chromatogram baseline at 515 nm. Table 3. Comparative quantitative analysis of antioxidant compounds using peak height in UV at 325 nm.