Integrated LC–HRMS and HPLC Profiling of Fourteen Anatolian Hypericum Extracts Reveals Distinct Chemometric and Bioactivity Patterns

TL;DR

This study uses advanced analytical methods to compare the chemical and biological properties of 14 Hypericum species from Anatolia, revealing distinct patterns and potential uses.

Contribution

The integration of LC–HRMS and HPLC with chemometric analysis provides new insights into Hypericum chemotypes and their bioactivity correlations.

Findings

Hyperoside was identified as a consistently abundant marker in Hypericum extracts.

H. triquetrifolium showed the strongest acetylcholinesterase inhibition and notable antioxidant capacity.

Selected Hypericum species exhibited high tyrosinase or α-glucosidase inhibition and limited antimicrobial activity.

Abstract

Background/Objectives: Anatolia hosts a rich diversity of Hypericum taxa; however, the chemical and biological properties of most species remain insufficiently characterized. Methods: This study combined liquid chromatography–high-resolution mass spectrometry (LC–HRMS) with high-performance liquid chromatography coupled to diode-array detection (HPLC–DAD) to profile 14 extracts obtained from 12 Anatolian species together with H. perforatum, and to examine whether metabolic variation aligns with bioactivity trends. Results: Chemometric analyses (Principal Component Analysis—PCA—and Hierarchical Cluster Analysis—HCA) revealed distinct chemotypes primarily driven by phenolic acids and flavonol glycosides, with hyperoside emerging as a consistently abundant marker. Phenolic-rich extracts displayed enhanced functional properties in multiple assays. Among them, H. triquetrifolium showed the strongest acetylcholinesterase inhibition and notable antioxidant capacity, whereas the selected taxa exhibited comparatively high tyrosinase or α-glucosidase inhibition. Antimicrobial testing indicated narrow-spectrum activity limited to Gram-positive bacteria and Candida spp. Conclusions: The integration of metabolite profiling with chemometric and functional data highlights clear chemotype–bioactivity associations across Anatolian Hypericum taxa and underscores the analytical value of LC–HRMS-guided chemotyping for pharmacognostic assessments. These results broaden the phytochemical and bioactivity landscape of this genus and provide a comparative framework for prioritizing species for future targeted investigations.

Figures

Full text

1. Introduction

The genus Hypericum (Hypericaceae) comprises a chemically diverse range of temperate taxa that accumulate naphthodianthrones (e.g., hypericin and pseudohypericin) and acylphloroglucinols (e.g., hyperforin), together with abundant flavonol glycosides and phenolic acids [1,2]. This phytochemical complexity underpins the long-standing use of Hypericum preparations; however, the overwhelming majority of pharmacognostic and pharmacological studies have focused on H. perforatum, leaving many regional taxa comparatively underprofiled.

Türkiye harbors nearly 120 Hypericum taxa (98 species, 49 endemic) distributed across steep geo-ecological gradients, which are likely to foster distinct chemotypes [3]. Several endemics accumulate unique or enriched metabolites with pharmacological potential [4,5], but ethnobotanical documentation remains uneven relative to H. perforatum [6]. In local trade, multiple species are frequently subsumed under the vernacular umbrella “kantaron,” increasing the risk of substitution and admixture and underscoring the need for species authentication and marker-based quality control.

From an analytical perspective, most existing data are derived from targeted various chromatographic assays or single-species surveys that provide only partial coverage of the genus [1,6,7]. Moreover, environmental and phenological gradients, including altitude and developmental stage, are known to markedly influence naphthodianthrone, acylphloroglucinol, and flavonoid titers [8], whereas misidentification among closely related taxa can confound apparent chemical patterns [5]. Therefore, in the present study, strict taxonomic authentication was combined with liquid chromatography–high-resolution mass spectrometry (LC–HRMS) based chemoprofiling under uniform experimental conditions. Building on this framework, the present study integrates complementary analytical and chemometric approaches to systematically characterize regional Hypericum chemotypes. Rather than aiming at the discovery of new chemical entities, this work focuses on chemotype-level variation and the relationship between chemical profiles and supportive bioactivity readouts.

Specifically, we integrated LC–HRMS profiling with high-performance liquid chromatography coupled to diode-array detection (HPLC–DAD) to map chemotypes across 12 Anatolian Hypericum taxa, benchmarked against H. perforatum. Feature lists were annotated using accurate mass, UV signatures, and HRMS fragmentation patterns, and autoscaled data matrices were explored using Principal Component Analysis and Hierarchical Cluster Analysis (PCA/HCA) to resolve species-level structure and phenological effects. To contextualize chemical differentiation, chemotype–bioactivity associations were examined using supportive in vitro readouts (acetylcholinesterase, butyrylcholinesterase, tyrosinase, $[eqn]$ -glucosidase; antimicrobial) with Spearman’s correlations and Benjamini–Hochberg FDR control, while antioxidant indices (DPPH/ABTS/CUPRAC) and bulk phenolics (TPC/TFC) are provided in the Supplementary Materials as chemical context.

Accordingly, this study aimed to: (i) establish LC–HRMS-based chemotaxonomic maps for Anatolian Hypericum under a single, harmonized workflow; (ii) annotate metabolites using transparent confidence levels and quantify diagnostic markers by HPLC–DAD; (iii) resolve species- and phenology-driven chemotypes; and (iv) evaluate whether major metabolite classes coherently track supportive bioassay trends without implying direct clinical efficacy.

2. Results

2.1. Phytochemical Profiling of Hypericum Extracts

2.1.1. Targeted LC–HRMS Profiling of Reference Phytochemicals

The targeted LC–HRMS workflow enabled the screening of 63 reference phytochemicals, of which 35 were unequivocally detected in at least 1 extract (Table 1). A salient outcome was the ubiquitous presence of eight metabolites—ascorbic acid, vanillic acid, hyperoside, rutin, myricetin, quercetin, salicylic acid, and 3,4-dihydroxybenzaldehyde—in every extract, delineating a conserved biochemical fingerprint across the Anatolian Hypericum taxa. Among these, hyperoside was the most prominent compound. Its abundance consistently saturated the detector response, indicating that it was the dominant flavonoid among these taxa. Because detector saturation precludes accurate back-calculation within a validated calibration model, hyperoside is qualitatively reported as a major constituent (+++) in Table 1. This observation is consistent with the prevalence of flavonol-rich Hypericum chemotypes [9].

In contrast, a defined subset of reference standards (e.g., fumaric acid, sinapic acid, luteolin-7-O-glucoside, apigenin-7-O-glucoside, luteolin, nepetin, 3′-O-methylquercetin, hispidulin glycosides, gypsogenic acid, shatavarin, pyrrolizidine alkaloids, cynarin, carnosic acid, and cirsilineol) was not detected (ND) in any sample. Their collective absence further supports a flavonol/phenolic-acid–centered chemical space in these Anatolian taxa [10,11], distinguishing this regional chemical profile from that of Hypericum populations reported in other biogeographic contexts [12,13].

Two primary trends emerged in the profiling dataset. Flavonol glycosides (e.g., rutin and hyperoside) co-occurred with diverse phenolic acids (e.g., vanillic and p-coumaric), delineating a phenolic–flavonol-centered chemical space across the Anatolian taxa with substantial between-taxon variation (Table 1). Variation within this domain showed an associative correspondence with the bioassay readouts.

2.1.2. HPLC–DAD Quantification of Hyperforin and Naphthodianthrones

For the HPLC analysis, three standard compounds, hyperforin, hypericin, and pseudohypericin, were used. For accurate quantification, an individual calibration curve was constructed for each standard using a series of dilutions. The calibration started at 10.00 $[eqn]$ for hypericin (correlation coefficient $[eqn]$ ) and 10.00 $[eqn]$ for pseudohypericin ( $[eqn]$ ); for hyperforin, the initial concentration was 50.00 $[eqn]$ ( $[eqn]$ ). Each standard was serially diluted six times to yield seven data points per calibration curve. To ensure consistency and reliability, all calibrations were performed under the same analytical conditions as those used for the samples.

Quantification across extracts revealed the highest concentrations in h14: hyperforin $[eqn]$   $[eqn]$ , pseudohypericin $[eqn]$   $[eqn]$ , and hypericin $[eqn]$   $[eqn]$ . Samples h3 and h9 contained only trace amounts or were not detected, indicating substantial variability in phytochemical content across species and developmental stages. The HPLC results are presented in Table 2. Reported concentrations therefore reflect the actual content of the extracts after back-calculation, rather than the injected standard solutions.

The observed variability, particularly the lower or undetectable levels in h3 and h9, may reflect phenological effects as well as genetic and site-related factors [12,14]. Given that pharmacological responses in H. perforatum are attributed to the concerted action of multiple metabolites, between-sample divergence in hypericins/hyperforin may be associated with differences in efficacy and, potentially, safety variability (e.g., phototoxicity linked to hypericins), if not controlled [15].

2.2. Antioxidant Capacity of Hypericum Extracts

2.2.1. Total Phenolic Content

The total phenolic content was determined as pyrocatechol equivalents (PEs) using the calibration equation $[eqn]$ ( $[eqn]$ ). As shown in Table 3, h14 exhibited the highest phenolic content at $[eqn]$ PEs/mg extract, followed by h2 with $[eqn]$ PEs/mg extract. The lowest value was observed in h4  $[eqn]$ (see also Figure 1a).

2.2.2. Total Flavonoid Content

Total flavonoids were determined as quercetin equivalents (QEs) using $[eqn]$ $[eqn]$ ( $[eqn]$ ). As in Table 3, h2 and h14 showed the highest flavonoid contents ( $[eqn]$ and $[eqn]$ , respectively), whereas h4 was lower ( $[eqn]$ ) (see Figure 1b).

2.2.3. DPPH Radical Scavenging Capacity

The DPPH radical-scavenging capacities (IC_50_) of the extracts ranged from $[eqn]$ to $[eqn]$ . h14 was the most active (lowest IC_50_), whereas h4, h5, h10, and h11 were in the higher (weaker) range, just above $[eqn]$ . h1 showed moderate activity $[eqn]$ ; other samples varied within the reported range. Standards BHA, $[eqn]$ -tocopherol ( $[eqn]$ -TOC), and BHT gave IC_50_ values of $[eqn]$ , $[eqn]$ , and $[eqn]$ , respectively. The results are presented in Table 3 and Figure 1c.

2.2.4. ABTS Cation Scavenging Capacity

ABTS radical-scavenging IC_50_ values are summarised in Table 3. h14 was the most active $[eqn]$ , followed by h4 and h5  $[eqn]$ and $[eqn]$ . h1, h6, h8, and h9 showed moderate activity (approximately 7– $[eqn]$ ). The remaining extracts had IC_50_ values from $[eqn]$ to $[eqn]$ . Controls: BHA $[eqn]$ , $[eqn]$ -TOC $[eqn]$ , BHT $[eqn]$ (see also Figure 1d).

2.2.5. Cupric Reducing Antioxidant Capacity (CUPRAC)

CUPRAC A_0.5_ values ranged from $[eqn]$ to $[eqn]$ (Table 3). h14 showed the greatest reducing power (lowest IC_50_), and h5 was also notable. h4, h10, and h11 were in the 30– $[eqn]$ range. h1, h6, h7, h8, and h9 showed intermediate reducing powers (20– $[eqn]$ ). Controls: BHA $[eqn]$ , $[eqn]$ -TOC $[eqn]$ , BHT $[eqn]$ . See Figure 1e.

Concordant rankings across DPPH, ABTS, and CUPRAC place h14, and to a lesser extent h2, among the most active extracts, mirroring their higher TPC/TFC. The assays probe complementary mechanisms (hydrogen-atom transfer vs electron transfer), explaining the consistent cross-assay performance.

2.3. Enzyme Inhibitory Activity

2.3.1. Cholinesterase Inhibition

Cholinesterases, i.e., acetylcholinesterase (AChE) and butyrylcholinesterase (BChE), rapidly hydrolyze acetylcholine (ACh) in the synaptic cleft, ensuring tight control of cholinergic signalling. Pharmacological blockade of these enzymes is a cornerstone symptomatic approach in Alzheimer’s disease (AD); by prolonging ACh residence time, approved inhibitors such as donepezil, galantamine, and rivastigmine help stabilise cognition and daily living [16].

Beyond their established use in AD, a growing body of behavioral and translational work suggests that low doses of reversible cholinesterase inhibitors can elicit antidepressant-like effects in rodent models of stress and despair, pointing to a broader role of cholinergic tone in mood regulation. Notably, donepezil and several structurally diverse inhibitors reduced immobility in the forced-swim test at sub-cognitive doses, displaying a characteristic U-shaped dose–response [17]. This evidence suggests that cholinesterase modulation is a promising avenue for treating mood disorders that are comorbid with or independent of neurodegeneration.

Against this backdrop, we evaluated Hypericum extracts for AChE and BChE inhibition. Previous reports on H. repens, have highlighted meaningful anticholinesterase activity along with antioxidant and antimicrobial properties, underscoring the therapeutic potential of the genus [18]. The inhibition profiles obtained here therefore extend current phytochemical research, offering dual relevance: (i) they reinforce the plant-derived search for safer symptomatic agents in AD, and (ii) they provide a pharmacological context for exploring Hypericum extracts as modulators of affective disorders where cholinergic imbalance is implicated. The ensuing discussion compares our IC_50_ values and enzyme-selectivity indices with the existing literature, examines putative structure–activity relationships, and considers how the multi-target pharmacology of the extracts might translate into neuroprotective and mood-ameliorating benefits.

For AChE inhibition, h1, h4, h6, h7, h8, h10, h11, h12, h13, and h14 displayed IC_50_ values ranging from $[eqn]$ to $[eqn]$ . Notably, h14 showed the lowest IC_50_, whereas h2, h3, h5, and h9 exceeded the tested range, indicating weaker or negligible inhibition at these levels.

For BChE inhibition, h9 was most active (IC_50_ $[eqn]$ ), followed by h7 and h12 (IC_50_ $[eqn]$ and $[eqn]$ , respectively). Therefore, these samples have emerged as candidates for further studies on conditions associated with cholinesterase dysregulation.

Galantamine, used as the reference standard, potently inhibited AChE and BChE with IC_50_ values of $[eqn]$ and $[eqn]$ , respectively. This benchmark highlights the relative efficacy of the extracts, and frames the potential of Hypericum as a source of cholinesterase inhibitors. The results are summarized in Table 4 (see also Figure 2a,b).

The cholinesterase inhibitory effects of Hypericum species are generally attributed to phloroglucinol derivatives [19] and phenolic compounds [20,21]. The relatively stronger AChE inhibition observed for h14 coincided with its overall phenolic richness and with higher levels of hyperforin/hypericins in the HPLC assay (Table 2); however, these co-variations are associative and do not establish causality. Likewise, the selective BChE activity of h9 co-occurred with a distinct phenolic profile in the targeted panel.

Spearman’s correlation analysis ( $[eqn]$ ), followed by Benjamini–Hochberg correction ( $[eqn]$ ), yielded 12 variables significantly associated with AChE inhibition (Table S1). Five compounds showed negative correlations; hence, variables negatively associated with inhibition included apigenin 7-O-acylglucoside ( $[eqn]$ ), vanillic acid ( $[eqn]$ ), total flavonoid content ( $[eqn]$ ), pseudohypericin ( $[eqn]$ ), and chrysin ( $[eqn]$ ). Conversely, chlorogenic acid and pyrogallol were positively correlated ( $[eqn]$ ), indicating that higher levels coincided with a weaker inhibition.

These trends were in accordance with the literature; flavone glycosides, including luteolin glycosides, can inhibit AChE at low micromolar levels, whereas apigenin-7-O-glucoside is typically much weaker [22,23]. Vanillic acid shows modest AChE effects [24], and reports on chlorogenic acids are mixed, with both inhibitory binding and antagonistic interactions described in crude matrices [25,26]. Chrysin is a moderate AChE inhibitor [27]. For pseudohypericin, docking suggests potential interactions with cholinesterases; however, direct enzyme data are limited [28]. Collectively, this supports flavonoid glycosides as major correlates of the activity of the most potent extracts (h14, h7, and h6), while phenolic acids and phloroglucinols may contribute additively; bio-guided fractionation will be required to test these possibilities.

For BChE, FDR-corrected analysis (Table S2) identified seven variables significantly associated with activity. Verbascoside and lithospermic acid showed the strongest negative associations with BChE IC_50_ ( $[eqn]$ and $[eqn]$ , respectively), nominating these metabolites as candidates for follow-up testing. Conversely, (+)-trans-taxifolin, rosmarinic acid, (−)-epigallocatechin, and emodin showed positive associations with BChE IC_50_, indicating that higher levels tended to coincide with weaker inhibition in this dataset. These findings are associative and require confirmation with isolated compounds and mixture studies.

The dataset comprised of 14 samples against 12 quantified secondary metabolites, yielding an $[eqn]$ ratio close to unity. Although Spearman–FDR provides robust rank-level insights, mechanistic claims await confirmation with isolated compounds and isobolographic synergy tests.

2.3.2. Tyrosinase Inhibition

Next, we examined the tyrosinase inhibitory potential. As tyrosinase catalyzes the rate-limiting steps in melanogenesis, its inhibitors are valuable for treating hyperpigmentation disorders and in cosmetic skin-lightening formulations [29]. Beyond cutaneous melanogenesis, tyrosinase participates in neuromelanin formation and has been linked to oxidative-stress pathways implicated in Parkinson’s disease [30]. Fungal tyrosinases also contribute to spore wall maturation and virulence, making them emerging antifungal targets [31].

Among the 14 samples, three extracts stood out. h10 and h12 were most potent (IC_50_  $[eqn]$ and $[eqn]$ , respectively); h7 followed at $[eqn]$ . By comparison, h2 and h13 were least active (IC_50_  $[eqn]$ and $[eqn]$ ). The remaining extracts displayed moderate inhibition (IC_50_  $[eqn]$ to $[eqn]$ ). Kojic acid, used as a positive control, yielded IC_50_  $[eqn]$ , confirming that none of the plant extracts approached the reference potency (Table 4, Figure 2c).

A growing body of evidence indicates that Hypericum extracts can inhibit tyrosinase through a constellation of polyphenols rather than a single “silver-bullet” constituent. For example, methanolic berries of H. androsaemum inhibit the enzyme at IC_50_ = 229 $[eqn]$ g/mL while simultaneously inhibiting MAO-A at an order-of-magnitude lower concentration (9.3 $[eqn]$ g/mL), a dual profile ascribed to extraordinarily high shikimic and chlorogenic acid levels [32]. However, in an anti-aging screen of the three Anatolian taxa, the most chlorogenic acid rich species, H. calycinum, did not outperform its congeners in tyrosinase assays, despite excelling collagenase, elastase, and hyaluronidase [33]. Likewise, Peruvian H. laricifolium leaves yielded the best tyrosinase read-out among 50 medicinal plants (IC_50_ = 120.9 $[eqn]$ g/mL), yet bio-guided fractionation pinpointed quercetin (IC_50_ = 14.3 $[eqn]$ M) as the principal inhibitor rather than chlorogenic acid [34].

These seemingly contradictory trends align with our dataset: chlorogenic-acid-rich h8 was a mediocre inhibitor, whereas flavonol-enriched h10 and h12 were the most potent (IC_50_ ≈ 200 $[eqn]$ g/mL). Recent docking and kinetics studies reinforce this observation-showing that quercetin, isoquercitrin, myricetin and their glycosides adopt bidentate orientations that chelate the binuclear Cu^2+^ active site with higher affinity than mono-caffeoyl esters such as chlorogenic acid [35]. Moreover, mixed-type inhibition and cooperative binding have been demonstrated for H. pruinatum and H. scabrum extracts, whose activities correlate with total flavonol equivalents rather than with any single phenolic marker [36].

Taken together, the emerging picture is consistent with a multicomponent interpretation, in which the co-occurrence of different phenolic classes may underlie the observed inhibition patterns. Caffeoylquinic acids may differentially interact with the enzyme environment, whereas flavonols are reported to exhibit higher copper-chelating affinity required for effective turnover inhibition. Consequently, chemotypes that co-accumulate in both classes, such as h10 and h12, represent attractive leads for dermatocosmetic or agro-food anti-browning applications. Targeted fractionation of these extracts, followed by combination assays, is essential for quantifying interaction indices and determining whether dual-action profiles (e.g., MAO-A + tyrosinase) can be harnessed for multi functional formulations.

Spearman’s correlation analysis ( $[eqn]$ ) followed by Benjamini–Hochberg FDR correction ( $[eqn]$ ) provided insights into the potential drivers of tyrosinase inhibition (Table S3). Notably, (−)-epigallocatechin displayed a strong positive correlation ( $[eqn]$ ) with the IC_50_ values, suggesting that higher levels of this catechin coincided with weaker anti-tyrosinase activity (or higher IC_50_). This implies that the potent inhibition observed in H. malatyanum and H. scabrum is likely not driven by monomeric catechins, but rather by their specific profile of flavonols (e.g., quercetin glycosides) and phenolic acids. Since no single ion feature passed the strict FDR threshold as a discrete “magic bullet” inhibitor after removing tentative annotations, the data suggest that the anti-tyrosinase potency is likely a multicomponent phenomenon involving the synergistic chelation of the copper active site of the enzyme by the complex phenolic matrix.

Literature indicates that Licorice-derived triterpenoids inhibit mushroom tyrosinase and reduce UV-induced pigmentation $[eqn]$ [37]. In contrast, catechins behave heterogeneously: galloylated analogs such as EGCG inhibit tyrosinase $[eqn]$ , whereas non-galloylated (−)-epigallocatechin is much weaker and can act as an alternative oxidative substrate [38,39]. These trends were consistent with the positive $[eqn]$ values observed for (−)-epigallocatechin.

The present dataset covered a defined phenolic panel quantified by HPLC–DAD and annotated via LC-HRMS; consequently, the full metabolome was not exhaustively captured. Regarding the correlation analysis, the positive association observed for (−)-epigallocatechin may reflect its potential as an alternative substrate for tyrosinase (masking inhibition), rather than true enzyme activation. These statistical associations are hypothesis-generating and will be verified in future studies using mechanistic assays employing isolated constituents and combinatorial testing.

2.3.3. α-Glucosidase Inhibition

We assessed the ability of the extracts to inhibit $[eqn]$ -glucosidase, a key therapeutic target for post-prandial hyperglycemia management [40]. h3 and h8 were most active (IC_50_  $[eqn]$ and $[eqn]$ , respectively), followed by h4 (IC_50_  $[eqn]$ ) and h7 (IC_50_  $[eqn]$ ). In contrast, h2 and h5 were only weakly active (IC_50_  $[eqn]$ ), and the remaining eight extracts showed no measurable inhibition within the tested range (Table 4, Figure 2d).

Acarbose, included as a positive control, gave an IC_50_ of $[eqn]$ under our assay conditions, indicating that the best Hypericum samples were ca. five-fold more potent than the standard compound.

Literature comparisons revealed pronounced geo-chemotypic variation. Mandrone et al. reported IC_50_  $[eqn]$ for an Italian H. perforatum extract [41], whereas the Turkish H. perforatum sample (h1) tested here was inactive. A similar discrepancy applies to H. scabrum; a Persian population inhibited the enzyme at $[eqn]$ [42], yet our Anatolian material (h12) showed no effect. Such divergence likely reflects differences in phytochemical profiles driven by climate, soil and harvest stage, underscoring the importance of region-specific chemotyping when prospecting for $[eqn]$ -glucosidase inhibitors [43].

Spearman’s correlation analysis ( $[eqn]$ ) followed by Benjamini–Hochberg FDR correction ( $[eqn]$ ) identified 11 variables that were significantly associated with $[eqn]$ -glucosidase inhibition (Table S4). Strong negative correlations ( $[eqn]$ ) were observed for (−)-epigallocatechin, rosmarinic acid, and emodin ( $[eqn]$ ), as well as for rutin and apigenin ( $[eqn]$ ). Moderate correlations ( $[eqn]$ to $[eqn]$ ) included chrysoeriol, vanillic acid, 3,4-dihydroxybenzaldehyde, (−)-epicatechin, and p-coumaric acid.

Literature comparisons further validated these statistical findings. Rutin [44] and rosmarinic acid [45] have repeatedly exhibited micromolar in vitro activity, often outperforming acarbose, which supports their high correlation rankings in our study. Similarly, apigenin and emodin are established low-micromolar inhibitors in purified-enzyme assays [46]. Although kaempferol did not reach statistical significance in our dataset ( $[eqn]$ ), its presence likely contributes to the overall efficacy. Vanillic acid [47], a weak inhibitor, is hypothesized to act synergistically when co-present with flavonoids [48].

It should be emphasized that the correlation does not equate to causation. The present analysis was hypothesis -driven and limited to a focused phenolic panel quantified by HPLC–DAD and LC-HRMS; untargeted metabolites were not captured. Nonetheless, the convergence between statistical and literature evidence justifies the prioritization of (−)-epigallocatechin, emodin, apigenin and rutin for future bio-guided fractionation and mechanistic studies.

2.4. Antimicrobial Activity

Hypericum extracts (h1–h14) were evaluated against a panel of Gram–negative bacteria (Pseudomonas aeruginosa ATCC 27853, Escherichia coli ATCC 25922, Klebsiella pneumoniae ATCC 4352, Proteus mirabilis ATCC 14153), Gram–positive bacteria (Staphylococcus aureus ATCC 29213, Staphylococcus epidermidis ATCC 12228, Enterococcus faecalis ATCC 29212), and yeasts (Candida albicans ATCC 10231, C. tropicalis ATCC 750, and C. parapsilosis ATCC 22019). Minimum inhibitory concentrations (MICs) are summarized in Table 5. Following the benchmark proposed by Ríos and Recio, crude extracts were considered to exhibit notable antimicrobial activity at MIC values ≤ 100 $[eqn]$ for bacteria and ≤200 $[eqn]$ for fungi [49,50].

Against Gram–positive strains, the extracts displayed selective but appreciable activity, with MIC values ranging from 78.12 to 1250 $[eqn]$ . Among the panel, h12 exhibited the strongest inhibition against both S. aureus and E. faecalis (78.12 $[eqn]$ ), while h11 also showed marked activity against S. aureus. In contrast, none of the extracts inhibited the tested Gram–negative bacteria within the examined concentration range (>1250 $[eqn]$ ).

Fungal assays revealed moderate but consistent inhibition of Candida tropicalis, with h14 showing the lowest MIC (78.12 $[eqn]$ ). Activity against C. albicans and C. parapsilosis was negligible across the panel. Reference antimicrobials yielded MIC values within expected CLSI/EUCAST quality-control ranges for the corresponding ATCC strains, confirming the validity of the assay conditions.

Overall, the antimicrobial profile observed here was selective rather than broad-spectrum, with activity largely confined to Gram–positive bacteria and Candida tropicalis. This pattern is consistent with previous reports indicating that Hypericum extracts preferentially inhibit Gram–positive organisms, whereas Gram–negative bacteria are comparatively less susceptible due to the permeability barrier imposed by their outer membrane [50,51].

Antimicrobial potency within the genus is further shaped by chemotype and extraction polarity. Lipophilic fractions enriched in acylphloroglucinols, such as hyperforin and related derivatives, frequently display enhanced activity against Staphylococcus spp. and Candida spp., whereas polar hydroalcoholic or aqueous extracts often require substantially higher concentrations to achieve comparable effects [52,53]. In this context, the absence of Gram–negative inhibition in the present study is consistent with the use of hydroalcoholic extracts dominated by flavonol glycosides and phenolic acids rather than strongly lipophilic phloroglucinols.

Importantly, several studies have demonstrated that crude Hypericum extracts can equal or exceed the antimicrobial potency of isolated constituents, supporting the contribution of multicomponent interactions to activity [54,55]. Accordingly, the comparatively strong effects observed for h12 and h14 against Gram–positive bacteria and C. tropicalis are best interpreted within a chemotype-dependent and synergistic framework, rather than being attributed to a single dominant metabolite.

2.5. Chemometric and Statistical Analyses

Fourteen Hypericum extracts (13 species) were analyzed for their chemical composition using HPLC and LC–HRMS. Principal component analysis (PCA) and hierarchical cluster analysis (HCA) were performed on 40 compounds. According to the PCA, PC1 accounted for 26.5% of the total variance and PC2 accounted for 19.1%. The score and loading plots are shown in Figure 3a,b, respectively. Based on LC–HRMS and HPLC data, h1 was clearly separated from the remaining species. The closest sample to h1 was h14, whereas h4, h5, h10, h12, h9, h11 and h7 clustered together by virtue of a shared compound set. In parallel, h3 differed from the others owing to the higher levels of apigenin, ascorbic acid, salicylic acid, p-coumaric acid, myricetin, ellagic acid and cirsimaritin. The loading plot visualizes the drivers of separation.

HCA (Ward linkage, and Euclidean distance) were used to summarize inter-sample relatedness (Figure 3c). Two major clusters were evident: h1 formed one branch and the remaining species formed the other. Within the latter, h6, h7, h10 and h11 were the closest, whereas h12 and h14 constituted the distinct subgroups.

3. Discussion

The integrated LC–HRMS and HPLC datasets indicate that the Anatolian Hypericum taxa examined in this study share a conserved chemical core dominated by flavonol glycosides and phenolic acids. Hyperoside was ubiquitous and exceptionally abundant across all extracts, supporting its role as a dominant flavonoid marker within this regional dataset. In contrast, several reference compounds were consistently absent, collectively delineating a flavonol/phenolic-acid–centered chemical space relative to profiles reported from other biogeographic regions.

Chemometric modeling (PCA and HCA) further resolved chemotype-level structure across the taxa. Variability in flavonol glycoside and phenolic acid loadings accounted for the primary axes of separation, while phenological stage (e.g., flowering versus fruiting) contributed additional dispersion within multivariate space. Importantly, taxa enriched along these loadings tended to exhibit stronger responses in several bioassays, consistent with an association between chemical composition and screening-level bioactivity rather than a causal or mechanistic relationship.

With respect to antioxidant assessment, the three assays employed (DPPH, ABTS, and CUPRAC) capture partially distinct response patterns reflecting different chemical mechanisms. In the present dataset, all three antioxidant indices showed a predominant alignment with total phenolic content (TPC), as evidenced by H. triquetrifolium and H. pamphylicum (flowering stage), which possessed the highest phenolic densities and superior scavenging and reducing activities. The DPPH assay appeared particularly responsive to the overall polyphenol profile, while ABTS and CUPRAC responses highlighted the broad electron-donating capacity of the extracts. The potent activity of H. triquetrifolium across all assays, despite not having the highest TFC, suggests that specific phenolic acids or synergistic interactions between flavonoids and other phenolics, rather than total flavonoid content alone, drive the antioxidant potential in these Anatolian taxa. This multi-method approach confirms that the combined use of DPPH, ABTS, and CUPRAC provides a comprehensive and complementary validation of the antioxidant landscape.

Across the biological assays, activity patterns were selective rather than uniform. H. triquetrifolium emerged as a consistently high-performing chemotype, particularly in antioxidant capacity and acetylcholinesterase inhibition, whereas fruiting-stage H. pamphylicum displayed pronounced selectivity toward butyrylcholinesterase inhibition, distinguishing it from its flowering-stage counterpart. Tyrosinase inhibition was comparatively strongest in extracts of H. malatyanum and H. scabrum, while $[eqn]$ -glucosidase inhibition was confined to a limited subset of taxa. Antimicrobial activity followed a similarly selective profile, with measurable effects largely restricted to Gram-positive bacteria and Candida tropicalis, whereas Gram-negative strains remained resistant within the tested concentration range.

Importantly, the observed bioactivities should be regarded as moderate and primarily comparative in nature, reflecting screening-level effects of crude extracts rather than direct pharmaceutical efficacy. As such, these findings do not imply immediate therapeutic relevance but instead provide a rational basis for prioritizing specific chemotypes for further investigation.

Correlation screening using Spearman’s analysis with Benjamini–Hochberg false discovery rate correction further supported these interpretations by identifying statistically significant associations between selected metabolite features and bioactivity endpoints. In particular, negative correlations between tyrosinase IC_50_ values and several flavonoid-related ion features suggest that the anti-tyrosinase potency observed for H. malatyanum and H. scabrum may be associated with the co-accumulation of copper-chelating polyphenols rather than the presence of a single dominant compound. Associations between flavonols and antioxidant or cholinesterase inhibition were likewise consistent with expectations from prior phytochemical and pharmacological studies. Nevertheless, given the complexity of crude extracts and the limited sample size, these relationships should be interpreted as associative and hypothesis-generating.

Several limitations of the present study should be acknowledged. Enzyme inhibition assays were conducted using non-human isoforms of tyrosinase and $[eqn]$ -glucosidase, and crude extracts are inherently susceptible to matrix effects that may obscure the contributions of individual metabolites. In addition, a proportion of LC–HRMS features remained annotated at MSI confidence levels 2–3. Future investigations should therefore focus on bio-guided fractionation, isolation and structural confirmation of key metabolites, validation of bioactivities using human-relevant enzyme isoforms, conversion of potencies to molar units, and kinetic characterization of inhibition modes. Accordingly, all chemometric and correlation-based inferences presented here should be considered associative rather than mechanistic.

Finally, the inclusion of a commercial drug sample (Hyperici herba, h1) was intended to serve as a benchmark reference rather than to imply strict chemotaxonomic equivalence with wild-collected taxa. Comparisons involving this sample should therefore be interpreted in a benchmark-based and exploratory context, aimed at assessing relative chemical composition and bioactivity performance rather than definitive equivalence.

4. Materials and Methods

4.1. Chemicals and Reagents

All reagents and solvents were of analytical or LC–MS grade as appropriate. Unless otherwise stated, chemicals (including reference standards where available) were obtained from Sigma-Aldrich (Merck, Darmstadt, Germany). LC–MS grade methanol, acetonitrile, and formic acid were used for chromatographic analyses; ultrapure water (18.2 M $[eqn]$ cm) was used throughout.

4.2. Plant Material and Extraction Procedure

The Hypericum taxa listed in Table 6 were collected from Türkiye during the flowering stage (except h9, which was fruiting stage). Specimens were identified by Ş. Kültür (Istanbul University) and deposited in the Herbarium of the Faculty of Pharmacy, Istanbul University (ISTE) under the voucher codes. H. perforatum (h1) was sourced from a commercial supplier of botanical raw material.

The plant material was air $[eqn]$ dried in the shade at ambient temperature and milled (Waring, Torrington, CT, USA), and each powdered sample was macerated with ethanol (96%) at a 1:10 (w/v) ratio at room temperature in the dark with daily solvent renewal until apparent exhaustion. The combined filtrates were concentrated under reduced pressure at 40 °C (Büchi R $[eqn]$ 210), frozen at $[eqn]$ °C and lyophilized (Labconco, Kansas City, MO, USA). The dried extracts were stored at $[eqn]$ °C until further analysis. For chromatographic analysis, stock solutions (e.g., 10 mg mL^−1^) were prepared in methanol (LC–MS grade) and filtered through 0.22 $[eqn]$ m membranes; for bioassays, dimethyl sulfoxide (DMSO) was used as a cosolvent where required (final DMSO ≤ 1% v/v).

4.3. Phytochemical Investigations

4.3.1. LC–HRMS Analysis

Phenolic constituents were profiled following a previously described procedure [56], with minor modifications. Dried extracts (100 mg) were dissolved in methanol (1.8 mL) and spiked with an internal standard solution (0.2 mL; dihydrocapsaicin; ≥85% purity; Sigma-Aldrich) to achieve a final extract concentration of 50 mg mL^−1^. Suspensions were clarified by centrifugation (if needed) and filtered through 0.45 $[eqn]$ m membrane filters prior to injection.

Analyses were performed using a Thermo Scientific Q Exactive Orbitrap system (Bremen, Germany). Separation was performed using a Fortis UniverSil C18 column (150 mm × 3.0 mm, 3 $[eqn]$ m; UK) maintained at 40 °C. The mobile phases were (A) water with 1% formic acid and (B) methanol with 1% formic acid. The gradient was: 50% B (0–1 min), 50 → 100% B (1–3 min), 100% B (3–6 min), 100 → 50% B (6–7 min), and 50% B (7–10 min). Flow rate was 0.35 mL min^−1^ and the injection volume was 2 $[eqn]$ L.

Putative identification was based on accurate mass data and retention time matching to authenticated reference standards (purity 95–99%) and the in-house spectral library of the Bezmialem Vakıf University Drug Application and Research Center (ILMER). To monitor repeatability and mitigate ionization-related variation, dihydrocapsaicin (95% purity) was used as an internal standard during LC–HRMS measurements. When authentic standards were unavailable, assignments were reported as tentative and comparatively used.

4.3.2. HPLC–DAD Analysis

Targeted quantitation of hypericin, pseudohypericin, and hyperforin followed the European Pharmacopoeia procedure with minor adjustments [57]. Experiments were performed using a Shimadzu 20A series HPLC system (LC–20AD pump, SPD–M20A diode-array detector, SIL–20AD autosampler; Shimadzu, Kyoto, Japan).

Hypericin and Pseudohypericin (Isocratic)

Chromatographic separation was achieved on a Thermo Fisher C18 column (250 × 4.6 mm, 5 $[eqn]$ m; USA) maintained at 40 °C using an isocratic mobile phase composed of ethyl acetate/15.6 g L^−1^ sodium dihydrogen phosphate (adjusted to pH 2 with phosphoric acid)/methanol (39:41:160, v/v/v). The flow rate was set at 1.0 mL min^−1^, the injection volume was 20 $[eqn]$ L, and detection was carried out at 590 nm.

Hyperforin (Gradient)

For hyperforin, the mobile phases consisted of 0.3% formic acid in water (A) and 0.3% formic acid in acetonitrile (B). The gradient program was as follows: 18% B (0–8 min), 18 → 53% B (8–18 min), 53 → 97% B (18–18.1 min), 97% B (18.1–29 min), and 97 → 18% B (29–30 min). The column temperature was maintained at 40 °C, the flow rate was 1.0 mL min^−1^, the injection volume was 10 $[eqn]$ L, and hyperforin was monitored at 275 nm.

All solvents were filtered through 0.45 $[eqn]$ m membranes and degassed by ultrasonication prior to use. System control and data processing were performed using Shimadzu LCsolution software (version 1.25). Quantitation was performed using external calibration with authenticated reference standards, following established pharmacopoeial practice for Hypericum constituents. All analytes quantified in the main dataset were evaluated within validated linear ranges under identical chromatographic conditions. Where authentic standards were unavailable, the results were reported as class-based estimates and are clearly identified as such in the Supplementary Information.

Calibration curves were constructed using a six-point external calibration prepared by serial dilution of stock standard solutions (pseudohypericin and hypericin: starting at 10 ppm; hyperforin: starting at 50 ppm), covering the validated linear concentration ranges. Extract samples were diluted as necessary prior to injection to ensure that detector responses fell within the calibration range, and concentrations were calculated from the calibration equation and back-calculated to the original extract.

4.4. Antioxidant and Bulk Compositional Assays

Total phenolic content (TPC), total flavonoid content (TFC), and antioxidant capacities (DPPH, ABTS, and CUPRAC) were assessed using established spectrophotometric protocols to provide supportive chemical context for the LC–HRMS-based profiles. TPC was determined using the Folin–Ciocalteu method with pyrocatechol as the reference standard [58], while TFC was measured by the aluminum nitrate assay using quercetin as standard [59]. Free radical scavenging and reducing capacities were evaluated using DPPH [60], ABTS [61], and CUPRAC [62] assays under standard conditions. Detailed experimental parameters are provided in the Supplementary Materials.

4.5. Enzyme Inhibitory Activity Assays

4.5.1. Cholinesterase Inhibitory Assays

Acetylcholinesterase (AChE) and butyrylcholinesterase (BChE) inhibition were measured using the microplate adaptation of Ellman’s method [63]. Each well contained 130 $[eqn]$ L sodium phosphate buffer (pH 8.0), 10 $[eqn]$ L of 4 mM sample solution, and 20 $[eqn]$ L AChE or BChE solution, and was incubated for 15 min at 25 °C. The reactions were initiated with 10 $[eqn]$ L DTNB and 10 $[eqn]$ L acetylthiocholine iodide or butyrylthiocholine iodide. The final test concentration of the extracts was 200 $[eqn]$ g/mL. Hydrolysis was monitored at 412 nm (BioTek Power Wave XS microplate reader (BioTek Instruments, Winooski, VT, USA)) via formation of 5-thio-2-nitrobenzoate. Galantamine was used as the positive control.

4.5.2. Tyrosinase Inhibitory Assay

Tyrosinase inhibition was assessed according to the method described by Hearing and Jimenez [64]. Diphenolase activity was assessed using L-DOPA a substrate. Mushroom tyrosinase (EC 1.14.18.1; 30 U, 28 mM) was dissolved in a sodium phosphate buffer (pH 6.8, 50 nM). The compounds were preincubated with the enzyme for 10 min at room temperature, and the reaction was initiated by the addition of L-DOPA (0.5 mM). The change in absorbance was monitored at 475 nm at 37 °C, and kojic acid was used as a positive control.

4.5.3. α-Glucosidase Inhibitory Assay

The $[eqn]$ -glucosidase assay the method described by followed Schmidt et al. [65] with minor modifications. Briefly, 10 $[eqn]$ L of extracts (in DMSO) were combined with 90 $[eqn]$ L phosphate buffer (pH 7.5; Na_2_HPO_4_, NaH_2_PO_4_, Milli-Q water, and 0.02% NaN_3_). Then 80 $[eqn]$ L $[eqn]$ -glucosidase Type I (0.05 U/mL) was added, the mixtures were incubated for 10 min at 28 °C. Substrate (20 $[eqn]$ L; p-nitrophenol $[eqn]$ -D-glucopyranoside, 1.0 mM) was then added. Blanks contained buffer (10% DMSO) instead of the sample. Absorbance at 405 nm was recorded every 40 s for 35 min, and the initial slopes were used to minimize interference from colored samples. Acarbose was used as the positive control.

Calculation. For all enzyme assays, percentage inhibition was computed as follows:

$$[eqn]$$

where A denotes absorbance.

4.6. Antimicrobial Activity Assays

Minimum inhibitory concentrations (MICs) were determined by broth microdilution, according to the Clinical and Laboratory Standards Institute (CLSI) [66,67]. Overnight bacterial and yeast cultures were adjusted to 0.5 × McFarland (≈1– $[eqn]$ CFU mL^−1^ for bacteria and 1– $[eqn]$ CFU mL^−1^ for yeasts) and diluted 1:100 in test media immediately before use. Crude extracts were dissolved in DMSO (final DMSO $[eqn]$  v/v) to 10 mg mL^−1^ stocks, and two-fold serial dilutions from 1250 to 0.06 $[eqn]$ g mL^−1^ were prepared in Mueller–Hinton broth (bacteria) or MOPS-buffered RPMI-1640 (yeast).

The activity was tested against Gram-negative Pseudomonas aeruginosa ATCC 27853, Escherichia coli ATCC 25922, Klebsiella pneumoniae ATCC 4352, Proteus mirabilis ATCC 14153; Gram-positive Staphylococcus aureus ATCC 29213, Staphylococcus epidermidis ATCC 12228, Enterococcus faecalis ATCC 29212; and yeasts Candida albicans ATCC 10231, Candida parapsilosis ATCC 22019, Candida tropicalis ATCC 750. Microtiter plates were incubated at 37 °C for bacteria (18–24 h) and 35 °C for yeasts (24–48 h). The MIC was defined as the lowest extract concentration that showed no visible growth relative to that of the untreated control. Ceftazidime, cefuroxime-Na, and amikacin were used as antibacterial controls, and clotrimazole and amphotericin B were used as antifungal controls.

4.7. Statistical Analysis

Measurements of each enzyme inhibitory assay were performed in triplicate ( $[eqn]$ ). Statistical comparisons were performed using two-tailed Student’s t-tests ( $[eqn]$ ), and the results are presented as mean ± SD. When multiple comparisons were involved, raw p-values were adjusted by the Benjamini–Hochberg false discovery rate (FDR) procedure with significance at $[eqn]$ (families defined per enzyme/endpoint). Associations between metabolite abundance and bioactivity readouts were evaluated using Spearman’s rank correlation ( $[eqn]$ ; two-sided, 95% CIs). Chemometric analyses (PCA and, HCA using Euclidean distance and Ward’s linkage) were performed using MINITAB 16.0. The hypothesis tests and FDR-adjusted correlations were computed using GraphPad Prism 10.0.

5. Conclusions

This study provides the first harmonized LC–HRMS/HPLC chemoprofiling of 14 Anatolian Hypericum taxa, demonstrating a chemical space dominated by flavonol glycosides and phenolic acids, with hyperoside as a dominant chemotaxonomic marker. Variations in phloroglucinols and naphthodianthrones further differentiated the species and were associated with assay readouts in an exploratory manner. Several taxa, particularly H. triquetrifolium, fruiting-stage H. pamphylicum, H. malatyanum, and H. scabrum, emerged as priority chemotypes for follow-up pharmacognostic work. Accordingly, all chemometric and correlation-based inferences should be interpreted as associative and hypothesis-generating rather than mechanistic.