Seasonal and Organ-Specific Variations of Alkaloids in Buxus obtusifolia (Mildbr.) Hutch: A Multivariate LC/MS Study

1. Introduction

Buxus obtusifolia (Mildbr.) Hutch is an evergreen shrub endemic to Kenya and Tanzania, and holds a significant ethnomedicinal value among some East African communities who utilize its root decoctions to treat chest ailments [1]. Our recent phytochemical investigation of the species’ leaves led to identification of 24 aminosteroid alkaloids. Notably, some of these compounds exhibited promising in vitro antiprotozoal activity against Trypanosoma brucei rhodesiense (Tbr), the causative agent of East African Human Trypanosomiasis and Plasmodium falciparum (Pf), the primary parasite responsible for tropical Malaria [2]. It is widely established that abiotic factors such as fluctuating temperatures, light intensity, nutrient availability, and water stress influence growth as well as biosynthesis and accumulation of active principles in medicinal plants [3,4]. Consequently, these environmental variations can significantly alter the overall therapeutic efficacy of the resulting herbal materials. Numerous studies have demonstrated that the total alkaloid content in both higher plants [5,6,7] and their associated endophytic fungi [8,9] is greatly influenced by environmental factors. Using a multivariate data analysis (MVDA) approach, our group has recently evaluated seasonal, organ-specific and location variability in the alkaloid profiles of Buxus sempervirens var. arborescens and suffruticosa [10], and Pachysandra terminalis [11]. These investigations identified various aminosteroid alkaloids as the key contributors to the observed chemical differences among the plant organs (leaves, twigs and flowers) across seasons.

Our recent study on B. obtusifolia represented the first phytochemical and pharmacological investigation on this species [2]. Accordingly, to the best of our knowledge, no previous studies have examined seasonal and organ-specific variability in its phytochemical profile. The aim of the present study was therefore to investigate the variations in the alkaloid content of B. obtusifolia, considering both the seasonal dynamics over a full annual vegetation cycle as well as differences between the primary aerial organs. Samples of the leaves and twigs harvested monthly for one year were prepared and analyzed using Ultra High Performance Liquid Chromatography/positive Mode-Electrospray Ionization Quadrupole Time-of-Flight-Tandem Mass Spectrometry (UHPLC/+ESI-QqTOF-MS/MS, henceforth abbreviated LC/MS). In order to extract meaningful information from the resulting complex LC/MS data matrix, principal component analysis (PCA) and volcano plot were employed to identify the characteristic aminosteroid alkaloids responsible for seasonal variation and differences between leaves and twigs.

2. Results

2.1. Sample Preparation, LC/MS Characterization, and Data Preprocessing

To establish a seasonal profile, the leaves and twigs of B. obtusifolia were collected monthly over a period of one year (February 2024–January 2025). The air-dried plant materials were separately extracted with dichloromethane to obtain 24 extracts which were subsequently analyzed using LC/MS. The resulting LC/MS raw data were processed using MetaboScape 3.0, where molecular features were extracted to generate a bucket table of 2195 <t_R:m/z> variables from 84 MS chromatograms (24×3 measurements for leaf and twig samples and 12 measurements for a quality control sample). In order to identify seasonal trends, this complex dataset was imported to ProfileAnalysis 2.1 and evaluated using two multivariate statistical methods, principal component analysis (PCA) and volcano plot.

2.2. Quantification of Annual Variability of Previously Isolated B. obtusifolia Aminosteroid Alkaloids

Our recent investigation reported 24 aminosteroids (1–24, Supplementary Materials, Figure S1) along with their antiprotozoal activity from the leaves of B. obtusifolia [2]. By comparing their exact retention times and mass-to-charge ratios <t_R:m/z>, variables corresponding to each of the previously isolated compounds were traced within the bucket table, and their relative abundance across the 24 samples was assessed. In order to track the relative quantities of these constituents throughout an annual vegetative cycle, bar charts were constructed in Microsoft Excel to visualize the MS signal intensities across the sampling period. For each feature, the mean signal intensity of the two most reproducible replicate measurements per month was calculated, with standard deviation represented as error bars. Samples were arranged chronologically from February 2024 to January 2025, allowing for a comparative analysis of each compound across the year. For instance, Figure 1 shows bar chart plots of 29-trimethoxybenzoyloxy cycloprotobuxoline-C (5, A) and obtusiepoxamine-A (21, B) with promising bioactivity [2], representing the 9β,10β-cyclo-5α-pregnane and 9 (10→19) abeo-5α-pregnane derivatives, respectively.

On average, both compounds are more abundant in the leaves (red bars) than in the twigs (green bars), showing the leaves as the preferred starting material for their isolation. Furthermore, 29-trimethoxybenzoyloxy cycloprotobuxoline-C (5), which exhibited the strongest antiplasmodial activity (IC₅₀ = 0.5 μmol/L vs. 5.9 μmol/L for cytotoxicity; selectivity index, SI = 12) in our previous work [2], showed the highest concentration in January. In contrast, obtusiepoxamine-A (21), which showed moderate antitrypanosomal activity (IC₅₀ = 2.2 μmol/L) and the highest selectivity index (SI = 16) among the 9 (10→19) abeo-5α-pregnane derivatives [2], reached its peak concentration in March. Additionally, both compounds 5 and 21 exhibited similar but minimally elevated concentrations, occurring in February–March for 5 and January–February for 21, with both compounds also showing high levels in December. Overall, it can be deduced from these plots that the optimal time to harvest the leaves for isolating these two compounds would be from December to March.

Using the same criteria, the optimal harvest period for all B. obtusifolia aminosteroid alkaloids reported in our recent study [2] was determined (Supplementary Materials, Figure S2). The results (Table 1) indicated that, except compounds 11, 14 and 16, all the other alkaloids exhibit their peak concentrations during the first quarter of the year. Additionally, except for compounds 8a + 8b, 11, and 24, the rest of the compounds were mainly found in higher quantities in the leaves compared to the twigs.

To correlate the optimum harvesting period with environmental conditions during the sampling period, we analyzed two abiotic factors, that is, temperature and precipitation. Abiotic stress, particulary drought and elevated tepmereatures have been widely demonstrated to induce the accumulation of nitrogen-containing plant secondary metabolites, including alkaloids [12,13]. This is consistent with our observations, which showed accumulation of most compounds between January and March, coinciding with high temperatures (>30 °C) and low precipitation (Figure 2). Physiologically, drought and heat stress can trigger stress-response pathways, such as abscisic acid (ABA) signaling and oxidative stress mechanisms, which in turn upregulate the genes responsible for secondary metabolite biosynthesis [14,15]. Consequently, water-stressed plants often exhibit higher levels of defensive alkaloids that may contribute to stress tolerance or herbivore deterrence [16]. Furthermore, it is worth noting that reduced plant growth under water-limited conditions may lead to higher relative metabolite concentrations when expressed on a dry-weight basis.

2.3. Principal Component Analysis (PCA)

Principal component analysis (PCA) is a widely-known unsupervised multivariate statistical method used to reduce the dimensionality of complex datasets while retaining the maximum possible variance [17]. PCA simplifies data interpretation with minimal loss of information by transforming a large set of correlated variables into a smaller number of orthogonal, uncorrelated variables known as principal components (PCs). The first principal component (PC1) captures the highest degree of variability, while subsequent components (PC2, PC3, amongst others) explain the remaining variance in descending order [18]. Following the methodology established in previous studies by our group [10,11], PCA models were calculated and cross-validated using three different scaling approaches (Level, Unit Variance, and Pareto scaling) as well as analysis of the unscaled data. When compared, all four approaches produced models that effectively discriminated between leaves and twigs samples of B. obtusifolia (Supplementary Materials, Figure S3). The Pareto-scaled model was selected for further analysis, as it provides the most balanced representation of variance between high and low abundance metabolites, as previously described by both Szabó and Schmidt and Schäfer et al. of our group [10,11].

Using ProfileAnalysis 2.1, samples are visualized in a score plot, which shows their relative positions in the PC space and enables the identification of natural groupings or trends. The corresponding loadings plot is used to determine which original variables contribute most significantly to each principal component. The results of the PCA model obtained with the Pareto-scaled data are presented in Figure 3.

The scores plot (A), displaying the second PC (PC2) plotted against the first PC (PC1), clearly demonstrates a separation between the two organs of B. obtusifolia. PC1 differentiates between the leaf and twig samples and accounts for 43.2% of the total variance, while PC2 explains an additional 10.0% of the variance and reflects the temporal variability of the samples. Together, the PC2 vs. PC1 scores plot captures 53.2% of the total variance in the bucket table. The quality control (QC_mix), consisting of an equal mixture of all samples, is located in the middle of the plot as expected. The associated loadings plot (B) shows the distribution of bucket variables <t_R:m/z> within the PC coordinate system. Variables characteristic of the leaves are located on the right (high positive values on PC1), whereas those characteristic of twigs are on the left (strongly negative values on PC1). Consequently, these variables represent the primary contributors to the observed differences between the two plant organs. Interestingly, PC2 appears to reflect a relatively large difference between December and January (negative scores on PC2) and February and March (positive scores on PC2).

2.4. Identification of Aminosteroids Contributing to the Chemical Differences Between the Organs and Temporal Variation

The loadings plot (Figure 4) was examined to extract the molecular masses of the most influential bucket variables, to identify the aminosteroids responsible for differences between leaf and twig samples. The molecular formulae of these variables were determined by analyzing the corresponding +ESI-QqTOF MS/MS spectra, and compared with literature data as well as previously isolated compounds from B. obtusifolia [2]. Furthermore, the fragmentation patterns of these variables were evaluated and compared with diagnostic fragments characteristic of the steroidal skeletons found in Sarcococca [19], Pachysandra [20], and Buxus [21], all members of the Buxaceae family. These diagnostic fragments have been previously compiled and described in earlier studies of our group [10,11].

Among the leaf-specific compounds, 29-trimethoxybenzoyloxy cycloprotobuxoline-C (5) (5.85 min:m/z 314.2287) and deoxycyclovirobuxeine-B (12) (5.58 min:m/z 200.1962) were identified as characteristic constituents of the leaves of B. obtusifolia. Notably, these two compounds demonstrated the highest antiplasmodial (IC₅₀ = 0.5 μmol/L for compound 5) and antitrypanosomal activity (IC₅₀ = 0.8 μmol/L for compound 12) in our recent study on the leaves of this species [2]. Correlation with the scores plot (Figure 3A) further indicates that deoxycyclovirobuxeine-B (12) is most strongly associated with leaf samples collected in January and December. Similarly, O-benzoyl-cycloprotobuxoline-D (6.03 min:m/z 254.2065) and N-benzoyl-cycloprotobuxolin-C (6.14 min:m/z 261.2143) were also identified as characteristic compounds of the leaves. The former was previously reported from Buxus sempervirens L. by Szabó et al. [24], while the latter was described from the same species by Kupchan et al. [22]. Based on the positions of the corresponding buckets in the loadings plot (Figure 4), the two compounds are characteristic of samples collected in February and March (Figure 3A). Cyclobuxophylline O (6.89 min:m/z 356.2941) and buxtauine M (7.63 min:m/z 372.3270) were identified as characteristic compounds of the twigs of B. obtusifolia. Finally, all the 24 compounds described in our previous study from this species could be located in the PC2 vs. PC1 loadings plot (Supplementary Materials, Figure S4). Among these, compounds 1, 2, 4, 5, 10, 12, 16, 21, 22 were found to significantly influence the organ-specific PCA model.

2.5. Seasonal Differences in the Annual Alkaloid Profile of Buxus obtusifolia

To further investigate the seasonal variability of the alkaloid profile of B. obtusifolia, the third principal component (PC3) was plotted against the first (PC1) (Figure 5). Since PC3 accounts for an additional 10.0% of the variance, PC3 vs. PC1 scores plot captures a total of 60.2% of the overall variance.

Based on monthly precipitation data provided by the Kenya Meteorological Department (Figure 2), samples were classified into rainy and dry seasons. Months with an average precipitation > 70 mm (April, May, June, November and December) were designated as the “rainy season,” while the remaining months were considered as the “dry season.” Given that sampling was conducted in a tropical region, temperature variation across the study period was minimal, except January–March, when average monthly temperatures exceeded 30 °C (Figure 2). This relatively stable temperature pattern likely explains the absence of a clear and consistent seasonal separation among the samples of the two organs. However, the leaf samples collected during the warmest months (January-March) cluster in the lower right region of the scores plot (Figure 5), indicating that the corresponding bucket variables are characteristic of the dry and warmest season. In contrast, the twig samples collected in July appear distinctly separated in the upper left region of the scores plot.

Accordingly, cycloprotobuxoline-C (1) (5.06 min:m/z 209.2021), 29-trimethoxybenzoyloxy cycloprotobuxoline-C (5) (5.85 min:m/z 314.2287), and obtusiepoxamine-A (21) (7.53 min:m/z 289.1908) (Figure 6) previously isolated in our study [2], were identified as characteristic of leaf samples from the warmest season. In addition, a previously undescribed aminosteroid with the molecular formula C₂₃H₃₉NO₂ (7.68 min:m/z 362.3075) was found to be characteristic of twig samples collected during the dry and coolest month of July.

2.6. Volcano Plot-Based Comparison of Leaves and Twigs Alkloid Profile

The two PCA models (PC2 vs. PC1 and PC3 vs. PC1) revealed a clear separation between the leaf and twig samples of B. obtusifolia and highlighted the major compounds contributing to this separation. However, compounds present in low concentrations are normally overlooked due to their limited influence on PCA models and are therefore difficult to identify through this approach alone. To complement the PCA results, a volcano plot based on direct pairwise t-test comparison between the two organs was performed. Each bucket was individually evaluated by plotting the statistical significance of the difference between the leaves and twigs (−log10(p-value)) against the magnitude of change (log2(fold change). This allowed visualization of both the magnitude and significance of concentration differences without less abundant compounds being dominated by the more abundant ones. From Figure 7, bucket variables corresponding to leaf compounds occur to the far right on the x-axis, while those occurring mainly in the twigs are on the left side of the x-axis. Out of the 2,195 variables, 1,190 showed statistically significant differences between leaves and twigs (p-value < 0.05, |log2 fold change| ≥ 1), with 510 features concentrated in the twigs (green bubbles) and 680 concentrated in the leaves (red bubbles). Notably, all variables with the highest PCA loadings (PC2 vs. PC1, Figure 4) were also statistically significant in the volcano plot, although they were not necessarily among the most differentially expressed variables. This finding is consistent with observations reported for B. sempervirens in the earlier study by our group [10].

Among the significant compounds, buxtauine M (7.63 min:m/z 372.3270) and the two new isomers (6.63 min:m/z 354.2792 and 6.92 min:m/z 354.2789), all of which had been identified as characteristic compounds of the twigs of B. obtusifolia (Figure 4), were additionally detected to be significantly more concentrated in the twigs than in the leaves (Figure 7). The volcano plot also highlighted another minor variable belonging to a previously unknown compound (9.77 min:m/z 275.2029) with a high log2(fold-change) of -4.97 as a characteristic compound of the twigs. Conversely, buxaustroine A (6.82 min:m/z 388.3209) was shown to be in significantly higher concentration in the leaves compared to the twigs. Two additional variables (10.86 min:m/z 653.4121 and 11.12 min:m/z 649.4170) exhibited particularly substantial log2(fold-changes) of 5.32 and 4.37, respectively. Based on their molecular mass, these variables were attributed to minor, previously unreported aminosteroids that are concentrated in the leaves.

3. Materials and Methods

3.1. Plant Material Processing and Extraction

The aerial parts of Buxus obtusifolia were harvested monthly from the same population site described in our previous study [2], over a period of one year (February 2024–January 2025). Identification of the plant material was done by Mr. Patrick Mutiso, a taxonomist from the Faculty of Science and Technology, University of Nairobi. The voucher specimens were deposited at both the University of Nairobi Herbarium (UoN_JM 2022_002) and at the Institute of Pharmaceutical Biology and Phytochemistry, University of Münster (IPBP 916 - TS_JM_2022_002). To complement the seasonal alkaloid variation, organ-specific variability was also evaluated by separately processing the leaves and twigs. The plant materials were air-dried under shade at room temperature to a constant weight and pulverized into fine powder using a mill. For each sample, 5 g of the powdered material was exhaustively extracted with dichloromethane (2 × 30 mL) under continuous agitation on a magnetic stirrer for 30 min. The obtained extracts were combined and thoroughly evaporated under reduced pressure at 40 °C.

3.2. UHPLC/+ESI-QqTOF-MS/MS-Analysis

The resulting extracts were analyzed using the previously described LC/MS method by our research group [11], with minor changes. The extracts were dissolved in methanol at a concentration of 10 mg/mL and analysed using a Bruker Daltonics micrOTOFQII time-of-flight mass spectrometer (Bruker Daltonics GmbH, Bremen, Germany). Chromatographic separation was achieved using a binary gradient of H₂O (+0.1% formic acid; A) and Acetonitrile (+0.1% formic acid; B) at a flow rate of 0.4 mL/min and column temperature of 40 °C. The elution profile described in our recent article [2] was used: 0–1.88 min: linear from 15% B to 30% B; 1.88–7.88 min: linear from 30% B to 33% B; 7.88–9.9 min: linear from 33% B to 50% B; 9.9–9.93 min: linear from 50% B to 100% B; 9.93–15.88: isocratic 100% B; 15.88–15.98 min: linear from 100% B to 15% B; 15.98–20.0 min: isocratic 15% B. In order to monitor injection precision and instrument sensitivity, 10 µL of papaverine (0.25 mg/mL) was added to each sample as an internal standard. For quality control, 20 μL of each sample was pooled into a collective LC/MS vial (QCmix) and measured after every six samples to assess sensitivity shifts and peak alignment consistency. Additionally, a methanol blank was measured after every 20 samples to monitor any carry-over of analytes. To obtain more information about the fragmentation patterns, some representative samples were initially subjected to auto MS/MS analysis at a collision energy of 20 eV. These initial runs also served to saturate the column matrix, ensuring reproducible retention times throughout the sequence. Following the matrix saturation measurements, the complete sample set was analyzed in triplicate using full-scan mode without intermediate recording of fragment spectra to obtain maximum number of data points per peak. An injection volume of 2 µL was employed for all analyses. The robustness of the dataset was ensured by measuring the samples in three distinct sequences: once sequentially from January 2025 to December 2024, followed by a randomized sequence and concluding with a sequential repetition.

3.3. Preprocessing of LC/MS Data

The raw LC/MS data from 84 sample measurements (24×3 measurements for leaf and twig samples and 12 measurements for a quality control sample) were processed using MetaboScape 3.0 software (Bruker Daltonics, Bremen, Germany). Molecular features were extracted to generate a comprehensive bucket table of <t_R:m/z> variables. The sample groups were defined by organ (leaves and twigs) and harvest month (January–December). The filter parameters were defined as follows: minimum features for extraction, 20/84 (accounting for molecular features detected in at least 20/84 samples during extraction); presence of features, 20/84 (representing features only present in at least 20/84); filter features by occurrence in groups, month 10% (accounting for features occurring in at least 10% of the samples within each group). The peak detection parameters were set as: intensity threshold, 1000 counts; minimum peak length, 8 spectra; minimum peak length (recursive), 7 spectra; retention time range, 1–15 min; mass range, m/z 50–1500. Ion deconvolution parameters were set as: EIC correlation, 0.7; primary ions, [M + H]⁺; seed ions, [M + 2H]²⁺, [M + Na]⁺, [M + K]⁺; common ions, [M + H−H₂O]⁺. The resulting bucket table containing 2195 variables x 84 analyses, was exported to Bruker ProfileAnalysis 2.1 software (Bruker Daltonics, Bremen, Germany) for multivariate statistical evaluation.

3.4. Principal Component Analysis (PCA) Modelling

Prior to PCA modelling, the 2195 bucket variables’ intensity values of each analysis were normalized to those of the internal standard papaverine (t_R 5.84 min:m/z 340.1580) in the respective run. To optimize the statistical output, three different scaling approaches were evaluated, such as level, unit variance, Pareto, as well as analysis of the unscaled data (Supplementary Materials, Figure S3). All PCA models were cross-validated using leave-one-out cross-validation to ensure robustness of the resulting data. While all four approaches demonstrated good sample separation, the Pareto-scaled model was selected for further analysis due to its superior representation of the variance between high and low abundance metabolites.

3.5. Volcano Plot

Using the bucket table exported from Mataboscape, a t-test was performed in Microsoft Excel to compare mean signal intensities between twig and leaf samples of B. obtusifolia. Log2 fold change and −log10(p-value) were calculated for each bucket and visualized as a bubble volcano plot, with bubble size proportional to the maximum signal intensity observed across all samples. Significance thresholds were set at |log2 fold change| ≥ 1 and p-value < 0.05.

4. Conclusions

The present study investigated variations in the alkaloid content of the leaf and twigs samples of B. obtusifolia over a full annual vegetation cycle using a multivariate data analysis approach. Principal component analysis and volcano plot of the LC/MS dataset revealed clear organ-specific differences in alkaloid profiles, along with some minimal trends in seasonal metabolite variation. In total, 15 compounds were identified as key contributors to the observed patterns. Additionally, among the 24 aminosteroids described in our previous study from this species, compounds 1, 2, 4, 5, 10, 12, 16, 21, 22 exhibited the most profound organ-specific differences as reflected in the PCA model. These findings provide valuable insights into the phytochemical dynamics of B. obtusifolia, and offer a basis for the targeted isolation of bioactive aminosteroids in future pharmacological studies.