How plant species origin, identity, and cereal type shape the microbial composition and functional properties of Munkoyo, a traditional fermented beverage

Abstract

Traditional fermented cereal products are important for food security and nutrition. In Zambia and neighbouring countries, Munkoyo is a traditional beverage produced by spontaneous fermentation of cooked grains inoculated with roots from specific wild plant species (“Munkoyo” plants). These plants host distinct root endosphere microbial communities that also vary across geographic locations. However, how plant species identity, geographic origin, and cereal type shape the microbial composition and functional properties of Munkoyo remains unclear. We conducted fermentation experiments to assess how plant species, field location, and cereal type affect the bacterial communities and functional properties of Munkoyo. Fermentation with different co-occurring plant species showed distinct acidification dynamics, with Rhynchosia insignis yielding lowest final pH and Eminia holubii showing faster acidification. These differences were accompanied by distinct Munkoyo bacterial communities, primary metabolites, and aroma profiles. Using R. insignis from three districts showed that geographic origin significantly influenced titratable acidity, aroma composition, and microbial community structure and metabolic potential, indicating terroir effects. In contrast, cereal type (maize vs. sorghum) had a limited impact on final pH but altered microbial community composition and metabolic profiles. Overall, plant species identity and geographic origin are key determinants of microbial and functional variation in Munkoyo beverage.

Sustainability Statement.

Munkoyo, a traditional fermented beverage produced by women at the household level and consumed by people of all ages, has strong potential to contribute to SDG2 (Zero Hunger). We show that the geographic origin of wild Munkoyo plants and their species identity are instrumental in shaping the microbial community composition and functional properties of Munkoyo beverage. This understanding contributes to improved nutrition from easy accessible fermented cereals and urges sustainable use of food ingredients sourced from the wild. Our work additionally contributes to SDG3 (Good Health and Well-being), SDG12 (Responsible Consumption and Production), and SDG15 (Life on Land) by showing that preserving the biological and cultural values of the Munkoyo beverage requires conserving the wild Munkoyo plant species and their associated wild root microbiomes. It urges against overharvesting these Munkoyo plants from natural woodlands and calls for assessing the resilience of the plants and their associated microbes to climate change. This needs to be taken into account in future studies aimed at upscaling of traditional Munkoyo processing to SME industrial levels.

Introduction

Traditional fermented cereal foods and beverages play essential roles in ensuring food security and enhancing dietary diversity globally (Schoustra et al. 2013, Phiri et al. 2019). In Africa, these products are primarily made from locally sourced staple cereals such as maize, wheat, sorghum, and millet, resulting in a diverse range of products that are fundamental parts of daily diets (Blandino et al. 2003). Apart from their appealing taste and aroma, fermented products offer nutritional benefits, extended shelf life, enhanced food safety, and are of profound cultural value (Blandino et al. 2003, Schoustra et al. 2013). However, despite their importance, our understanding of the raw materials and microbial communities driving traditional African fermentation processes remains limited, impeding broader initiatives for large-scale production and commercialization (Schoustra et al. 2013, Motlhanka et al. 2018).

Recent years have witnessed an increase of studies focusing on African fermented cereal-based products, aimed at elucidating the mechanisms underlying the fermentation process (Motlhanka et al. 2018). These studies dealt with a variety of products, including Togwa in Tanzania (Mugula et al. 2003), Mahewu in Zimbabwe and South Africa (Daji et al. 2022, Kudita et al. 2024), Mawè and Akpan in Benin (Carole Sanya et al. 2023, Addy et al. 2026), and Munkoyo in Zambia and the Democratic Republic of Congo (DRC; Phiri 2019, Obafemi et al. 2022). Across these products, lactic acid bacteria (LAB) and yeasts constitute the dominant microbial groups and interact throughout fermentation to determine product quality. LAB acidify the substrate through the production of organic acids, thereby improving microbial safety, extending shelf life, and contributing to flavour development. Yeasts complement bacterial activity by producing ethanol and volatile aroma compounds while contributing to flavour complexity. Characterization of these microbial communities has enabled the development of defined starter cultures for several traditional fermented foods, improving fermentation consistency while creating opportunities for product standardization and commercialization (Blandino et al. 2003, Motlhanka et al. 2018). Depending on the fermentation system, the microorganisms driving these fermentations are either present (i) in the environment, (ii) on processing equipment, (iii) within the food substrate, or (iv) intentionally added as starter cultures (Motlhanka et al. 2018).

Munkoyo is a traditional cereal-based fermented beverage widely consumed by all age classes in Zambia and the Democratic Republic of Congo (Foma et al. 2013, Phiri 2019). The beverage primarily consists of maize as the main cereal ingredient, although alternative ingredients such as sorghum and millet are also used in its preparation (Phiri 2019). Nutritional characterization indicates that Munkoyo is primarily an energy-rich beverage, containing high levels of carbohydrates together with variable amounts of protein, lipids, and essential minerals such as calcium, iron, and zinc, depending on the cereal substrate and processing method (Chileshe et al. 2020). Fermentation further enhances its nutritional and functional value by reducing antinutritional compounds such as phytates, improving starch digestibility and mineral bioavailability, and promoting the growth of LAB with reported probiotic and antioxidant potential (Chileshe et al. 2020, Phiri et al. 2021).

Unlike many cereal fermentations that rely primarily on spontaneous microbial inoculation, Munkoyo fermentation is initiated using roots from specific wild leguminous plants, principally Rhynchosia insignis, Rhynchosia heterophylla, and Eminia holubii, which serve as natural starter cultures (Foma et al. 2013, Kalumbilo et al. 2025). These plant species are traditionally preferred because their roots possess strong amylolytic activity and naturally harbour microbial communities that contribute to fermentation initiation and subsequent metabolic transformations. The debarked roots, either fresh or sun-dried, or their extracts, are added to warm cereal porridge and left to ferment at ambient temperature for 1–3 days, during which microbial fermentation processes transform the porridge into the characteristic Munkoyo beverage. The added Munkoyo roots are the source of LAB and amylolytic enzymes that drive the fermentation process (Zulu et al. 1997, Phiri 2019). The amylolytic enzymes degrade the starch in the cereal porridge into simple, fermentable sugars, mostly maltose, maltotriose, and glucose. The sugars are subsequently converted by the LAB into organic acids and aroma compounds that underlie the characteristic flavour, acidity, and preservation of the beverage (Foma et al. 2013, Phiri 2019).

Our prior investigation into the root-associated bacterial communities of three prominent Munkoyo plant species commonly used as starters, R. insignis, R. heterophylla, and E. holubii, showed microbial variations across different plant species and field sampling locations (Kalumbilo et al. 2026). Specifically, distinct root endosphere bacterial communities were identified among the different plant species, and within the same plant species, notable variations in root endosphere bacterial composition were evident across geographic locations of the sampling sites. These findings suggest that starter roots may introduce distinct microbial inocula into the fermentation process. However, it remains unknown what the impact of the use of specific plant species and their corresponding growth locations is on the characteristics and bacterial communities of the beverage. Understanding how starter plant identity, geographic origin, and cereal substrate influence fermentation outcomes is essential for developing evidence-based strategies to improve fermentation consistency while preserving the traditional characteristics of Munkoyo. Such knowledge provides a scientific basis for selecting starter materials, improving quality control, and facilitating future standardization and commercialization of indigenous cereal fermentations.

To our knowledge, this is the first study to experimentally disentangle the effects of starter plant species, geographic origin of starter roots, and cereal substrate on Munkoyo fermentation under controlled conditions. We simultaneously evaluated in the Munkoyo beverage the bacterial community assembly, predicted metabolic pathways, physicochemical characteristics, and metabolite production to determine how each of these factors contributes to fermentation outcomes. To do so, we conducted three fermentation experiments in which we aimed to investigate the impact of (i) plant species identity, (ii) geographic origin, and (iii) cereal type on the bacterial community composition and the functional properties of Munkoyo beverage. Given the documented influence of plant species and field sampling location on root endosphere microbiomes, we hypothesized that: (i) different plant species used as fermentation starters result in distinct bacterial communities and functional properties of Munkoyo; (ii) different geographic origins of the same plant species (terroir effect) lead to differences in both microbial community composition and functional properties of Munkoyo; (iii) different cereal substrates shape microbial community structure and functional properties of Munkoyo.

Materials and methods

Materials

Cereal ingredients used as the base for Munkoyo preparation were sourced from local markets. Maize roller meal (National Milling Corporation Ltd, Chilanga, Zambia) from Zea mays and red sorghum meal (Shais Foods, Zambia) from Sorghum bicolor were selected for this study.

Fresh Munkoyo roots of R. insignis were collected from three districts in Zambia: Kasempa in North-Western Province (13°19.089′ S, 26°00.585′ E), Mumbwa in Central Province (14°58.817′ S, 26°35.783′ E), and Rufunsa in Lusaka Province (15°12.053′ S, 29°12.836′ E). The roots of R. heterophylla and E. holubii were collected only in Kasempa, where all three starter plant species naturally co-occur. The roots were debarked, beaten, shredded, and then sun-dried until completely dry. The dried Munkoyo roots were stored at room temperature until further use. These three species were selected because they are among the principal wild plant species traditionally used as starter cultures for Munkoyo fermentation in Zambia and were previously confirmed through DNA barcoding and ethnobotanical surveys as authentic Munkoyo plants (Foma et al. 2013, Kalumbilo et al. 2025). The roots were debarked, beaten, shredded, and sun-dried until completely dry before being stored at room temperature until use in the fermentation experiments.

Munkoyo preparation

Munkoyo beverage was prepared using the Nyimba method (Phiri et al. 2020) with slight modifications. First, the Munkoyo root extract was prepared by soaking 5 g of Munkoyo root sample in 300 ml of cooled boiled tap water (40°C–45°C) for 2 h. Next, cereal porridge was prepared by heating 20 l of tap water to ∼60°C and gradually adding 2700 g of cereal powder while stirring. The mixture was allowed to cook for 2 h with periodic stirring and then cooled to 40°C–45°C. Subsequently, 200 ml of the porridge was aliquoted into sterilized Pyrex glass bottles, to which 300 ml of the Munkoyo root extract, together with the roots, was added and thoroughly mixed. The mixture was allowed to stand for 10–15 min, after which time the 0-h time (t0) samples were collected. The mixture was then incubated at 28°C for 72 h, after which Munkoyo samples were collected to determine the bacterial community composition, physicochemical properties (pH and titratable acidity), primary metabolites, and volatile organic compounds (VOCs).

Experimental design

Experiment one: influence of plant species on Munkoyo beverage characteristics

To determine the influence of plant species used as fermentation starters on the characteristics of Munkoyo beverage, we used maize meal as a base for the porridge along with three root species collected from the same geographic location within Kasempa district. The roots comprised three distinct Munkoyo plant species: R. insignis, R. heterophylla, and E. holubii. For each plant species, roots were collected from three individual plants, constituting the biological replicates. For the experiments, each biological replicate was used to make three technical replicates. As a negative control, we included three replicates of the maize porridge without the addition of Munkoyo roots to assess spontaneous fermentation in the absence of the traditional starter culture and to distinguish fermentation changes attributable to the addition of Munkoyo roots from those arising from the indigenous microbiota naturally present in the porridge and processing environment. The experimental setup is shown in Supplementary Fig. S1.

Experiment two: influence of geographic origin on Munkoyo beverage characteristics

To determine the influence of geographic origin of the Munkoyo plants on the characteristics of Munkoyo beverage, we used maize meal as a base for the porridge along with R. insignis roots collected from three districts in Zambia: Kasempa in North-Western Province, Mumbwa in Central Province, and Rufunsa in Lusaka Province. For each location, roots were collected from three individual plants, constituting the biological replicates. For each biological replicate, three technical replicates were prepared. As a negative control, we included three replicates of the maize porridge without the addition of roots. The experimental setup is shown in Supplementary Fig. S2.

Experiment three: influence of cereal types on Munkoyo beverage characteristics

To determine the influence of cereal types on the characteristics of Munkoyo beverage, we used R. insignis roots collected from Kasempa district. Roots were collected from three individual plants, constituting the biological replicates. For each biological replicate, three technical replicates were prepared. Maize meal and sorghum meal Munkoyo were made for comparison. As negative controls, we included three replicates of the maize and sorghum porridge without the addition of roots. The experimental setup is shown in Supplementary Fig. S3.

Functional properties of Munkoyo beverage

As proxies for the functional characterization of the Munkoyo beverage we quantified the pH, titratable acidity, the abundances of LAB, the metabolites (organic acids, sugars, ethanol, and acetoin) and the volatile organic compounds from the Munkoyo beverage. Below we explain in more detail how these different properties were quantified. Negative control fermentations (cereal porridge prepared without Munkoyo roots) were analysed alongside inoculated fermentations to characterize spontaneous fermentation by the indigenous microbiota naturally present in the cereal substrate and processing environment, thereby providing a baseline against which the effects of Munkoyo root inoculation could be evaluated.

Physicochemical properties of Munkoyo beverage

The pH and titratable acidity (TA) were recorded every 12 h for 72 h from the time the roots were added to the porridge. The pH was measured using a portable pH meter (HI 151 HANNA Instruments), and TA was determined by titrating 10 cm³ of the sample against sodium hydroxide using phenolphthalein as an indicator (Phiri et al. 2020).

Analysis of metabolites in Munkoyo beverage

Organic acids (lactate, acetate and formate), sugars (maltose and glucose), ethanol, and acetoin were analysed by High-Performance Liquid Chromatography (HPLC) using the Dionex™ UltiMate™ 3000 HPLC system (Thermo Fisher Scientific, Inc., Waltham, USA) equipped with the Aminex® HPX-87H column, 300 × 7.8 mm, and guard column (Bio-Rad Laboratories, Inc., Hercules, USA). RefractoMax 521 Refractive Index Detector (Thermo Fisher Scientific, Inc., Waltham, USA) in combination with UV at 220, 250, and 280 nm was used to detect the eluted molecules. Five millimolar H₂SO₄ was used as mobile phase at 0.6 ml/min and 40°C. Munkoyo samples were clarified before HPLC by first adding 0.25 ml of cold Carrez Clarification Reagent II (4.22% ZnSO₄ · 7 H₂O, w/v) to 0.5 ml of sample, followed by vortexing, and subsequent addition of 0.25 ml cold Carrez Clarification Reagent I (5.77% K4[Fe(CN)6] · 3 H₂O, w/v) and vortexing again. The samples were centrifuged at 18′850 × g and 4°C for 10 min. A supernatant of 0.2 ml was transferred into an HPLC vial and subjected to HPLC. Data were analysed using the chromatography data system Chromeleon® 7.2 (Thermo Fisher Scientific, Inc., Waltham, USA). Quantification was done using calibration curves for the individual compounds based on the peak area.

Analysis of VOCs from Munkoyo beverage

VOCs were analysed using Headspace Solid Phase Microextraction Gas Chromatography combined with Mass Spectrometry (HS-SPME GC-MS) performed on the Trace™ 1300 Gas Chromatograph equipped with the TriPlus™ RSH autosampler and the ISQ™ QD Mass Spectrometer (Thermo Fisher Scientific, Inc., Waltham, USA). A volume of 0.5 ml of each Munkoyo sample was transferred into a 5 ml GC glass vial and stored at –20°C until analysis. Frozen samples were incubated for 20 min at 60°C before the VOCs were extracted for 20 min at 60°C using the Supelco® SPME fibre with Divinylbenzene/Carboxen/Polydimethylsiloxane coating (Merck KGaA, Darmstadt, Germany). The VOCs were desorbed from the fibre for 2 min onto a Restek™ Stabilwax™-DA Crossbond™ Carbowax™ Polyethylene Glycol column with 30 m length, 0.25 mm ID, 0.5 μm df (Thermo Fisher Scientific, Inc., Waltham, USA). The programmable temperature vaporizer injector was heated to 250°C and operated in split-less mode. The gas chromatography oven temperature was kept at 35°C for 2 min, raised to 240°C with a slope of 10°C/min, and kept at 240°C for 5 min. Helium was used as a mobile phase with a 9a constant flow rate of 1.2 ml/min. Mass spectral data were measured over a range of 33–250 m/zin full-scan mode with 3.0030 scans/s. Data were analysed using the chromatography data system Chromeleon® 7.2 (Thermo Fisher Scientific, Inc., Waltham, USA). The ICIS algorithm was used for peak integration, and the VOCs’ MS profiles were matched with the MS profiles of known compounds in the NIST Standard Reference Database (Main EI MS Library). Peak areas were calculated based on the highest m/z peak per compound (MS quantification peak). To denoise the data, peak areas for compounds that had a value below 1′000 counts × min were classified as ‘not detected’, and all peak areas for non-detected compounds were set to 0 counts × min to enable data normalization.

Enumeration of (facultative) anaerobic LAB in the Munkoyo beverage

For the enumeration of the viable (facultative) anaerobic LAB, Munkoyo samples were serially diluted in Ringer’s solution. The diluted samples were plated on MRS agar plates, which were then incubated at 28°C for 48 h in anaerobic jars. After the incubation period, all plates were counted manually.

Bacterial community characterization of Munkoyo beverage

Total Genomic DNA was extracted from 2 ml of Munkoyo beverage using the DNeasy PowerSoil Pro kit (Qiagen, Germany), following the manufacturer’s instructions. For each sample, four replicates were concentrated on the same MB Spin Column and eluted with a 50 μl volume of solution C6 elution buffer. This process aimed to increase the DNA concentration sufficiently. DNA was quantified using a NanoDrop (ND-2000) Spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA), and the extracted DNA was stored at −20°C until further analysis.

The V3–V4 region of the bacterial 16S rRNA gene was targeted for amplification using the universal primers 341F (5′-CCTAYGGGRBGCASCAG-3′) and 806R (5′-GGACTACNNGGGTATCTAAT-3′) (Yu et al. 2005, Klindworth et al. 2013). The primers were designed to incorporate connecting barcodes, providing unique identifiers for individual samples during subsequent sequencing processes. Polymerase Chain Reaction (PCR) amplification, incorporation of sample-specific barcodes, library preparation, and sequencing were performed by Novogene (Cambridge, United Kingdom) using their standard amplicon sequencing workflow. A detailed description of library preparation, quality control, and sequencing procedures is provided in Supplementary Methods.

Bioinformatics

Raw paired-end reads were demultiplexed, quality filtered, and denoised using the DADA2 pipeline implemented in QIIME2 (Callahan et al. 2016, Bolyen et al. 2019). Taxonomy was assigned against the SILVA 138.1 database using the classify-sklearn classifier (Quast et al. 2013, Bokulich et al. 2018). Chloroplast, mitochondrial and non-bacterial sequences were removed before downstream analyses. ASVs representing less than 0.25% relative abundance were excluded (Reitmeier et al. 2021). Rarefaction curves confirmed adequate sequencing depth (Supplementary Fig. S4). Additional details of the bioinformatics pipeline are provided in Supplementary Methods.

Prediction of microbial functional pathways

Functional profiles of the bacterial communities were predicted using PICRUSt2 (Douglas et al. 2020). The ASV abundance table and representative sequences generated following DADA2 processing were used as input for functional prediction. Predicted gene family abundances were annotated against the Kyoto Encyclopaedia of Genes and Genomes (KEGG) database. Downstream analyses and visualization of predicted pathway abundances were performed in R using the ggpicrust2 package (v 2.5.17) (Yang et al. 2023) and R software (v 4.5.0). Differential pathway abundance among treatments was evaluated using ALDEx2 implemented within ggpicrust2, and significant pathways were visualized using heatmaps. Functional interpretation was focused on pathways associated with carbohydrate metabolism, amino acid metabolism, energy metabolism, environmental information processing, and genetic information processing because of their relevance to cereal fermentation (Addy et al. 2026).

Statistical analyses

The statistical analyses were performed with R software version 4.3.0 (R Core Team 2023).

The metabolite and VOC profiles of the Munkoyo beverages were analysed using principal component analysis (PCA) on log-transformed and column-normalized data using the pcaMethods package (Stacklies et al. 2007). PCA was used to visualize differences in metabolite and VOC composition between treatments, and compound contributions to principal components were assessed using loading values.

Bacterial community analyses were conducted using the phyloseq (McMurdie et al. 2013) and vegan (Oksanen et al. 2009) packages. Beta diversity of the bacterial communities in Munkoyo beverage was assessed using Bray–Curtis dissimilarity matrices calculated from rarefied ASV counts. Patterns in community composition were visualized using Principal Coordinate Analysis (PCoA) implemented in the phyloseq package. Differences in community composition among treatments were statistically tested using permutational multivariate analysis of variance (PERMANOVA) with 999 permutations, as implemented in the adonis2 function in the vegan package (v2.6–4) (Oksanen et al. 2009). Pairwise PERMANOVA comparisons were conducted to assess differences between individual treatment groups when overall treatment effects were significant.

Taxonomic composition was summarized at genus level using relative abundance data. Relative abundance plots were generated to visualize differences in dominant bacterial taxa across treatments while retaining biological replicate-level variation.

Differential abundance analysis of the metabolic pathways of the microbial communities in the Munkoyo beverages, in relation to plant species, plant geographic origin, or substrate, was conducted using ALDEx2 implemented in the ggpicrust2 package (v 2.5.4), and visualized through heatmaps.

Results

Influence of plant species on pH, titratable acidity, and viable cell count of LAB in Munkoyo beverage

The acid production during fermentation was evaluated through pH measurements and titratable acidity (TA) determination at various time points. The initial pH of maize porridge without added roots was 6.50 (SD < 0.01), while for all other samples, the initial pH ranged between 6.19 and 6.60 (Fig. 1a). Munkoyo samples prepared with E. holubii roots showed rapid acidification, reaching a pH of 4.09 (SD = 0.06) after 24 h. Comparisons of pH levels at 72 h showed significant differences among Munkoyo samples prepared with different plant species (Analysis of variance ANOVA; F_3,29 = 12.85, P < 0.001), with R. insignis Munkoyo displaying the lowest pH at 3.79 (SD = 0.21).

Figure 1.

Development of pH (a), titratable acidity (%) (b), and viable cell counts of (facultative) anaerobic LAB (c) during the fermentation of Munkoyo prepared using the roots of three plant species as starters (EH: Eminia holubii, RH: Rhynchosia heterophylla, RI: Rhynchosia insignis). Cooked maize porridge to which no roots were added (NR) served as a negative control. All samples were incubated at 28°C for 72 h. Results are shown as mean values from nine independent experiments, with each species having three biological replicates and three technical replicates. Error bars represent standard error. For LAB cell counts, the missing data points indicate values below the detection limit of 2.5 log₁₀(CFU/ml).

For non-inoculated maize porridge, the initial TA, representing the % lactic acid produced by the microbial community, was 0.02 (SD < 0.01), while for all other samples, initial TA ranged between 0.03 and 0.05 (Fig. 1b). After 72 h, Munkoyo prepared with R. insignis exhibited the highest TA at 0.45 (SD = 0.08), while non-inoculated samples showed the lowest TA at 0.19 (SD = 0.02).

The viability of LAB was assessed by plating samples on agar plates at 0 and 72 h. Initial LAB plate counts were highest for Munkoyo prepared with E. holubii (6.4 log₁₀CFU/mL, SD = 0.2) (Fig. 1c). Counts of LAB for Munkoyo prepared with R. heterophylla and R. insignis were lower, ranging between 5.8 and 5.4 log₁₀CFU/ml, respectively. Initial LAB counts for non-inoculated maize meal porridge were below the detection limit (< 2.5 log₁₀CFU/ml). After 72 h, LAB counts increased and were comparable for all samples, including the non-inoculated samples, ranging between 7.3 and 8.1 log₁₀CFU/ml, with no significant differences observed (ANOVA; F_3,29 = 0.45, P = 0.719).

Influence of plant species on bacterial community composition and predicted functional pathways in Munkoyo beverage

PCoA was used to assess the β-diversity (dissimilarity) of bacterial communities of Munkoyo beverage samples fermented for 72 h. Bacterial community composition differed substantially between Munkoyo samples, with the first PCoA axis explaining 40.8% and the second axis 13.7% of the variation (Fig. 2).

Figure 2.

PCoA derived from Bray-Curtis distances of bacterial communities in Munkoyo beverages prepared using roots of three plant species as starters: Eminia holubii (EH), Rhynchosia heterophylla (RH), and Rhynchosia insignis (RI). Negative control samples (NR) consisted of cooked maize porridge without added roots. Letters A, B, and C after the plant species name indicate the different biological replicates and numbers 1, 2, 3 identify the technical replicate. All samples were incubated at 28°C for 72 h. Root samples were collected from a single site within the Kasempa district. The % of variation explained by each axis is shown in square brackets (Fig. 2).

While Munkoyo bacterial communities generally clustered by plant species, some exhibited overlap, suggesting similarities in bacterial communities across some Munkoyo samples made with different plant species. Nevertheless, PERMANOVA analysis corroborated the findings of the plant species imprint, demonstrating that plant species significantly influenced the dissimilarity of bacterial communities in the Munkoyo samples (R² = 0.195, P = 0.019). Pairwise PERMANOVA comparisons further elucidated distinct patterns, with Munkoyo prepared with R. insignis roots significantly differing from Munkoyo made with E. holubii (R² = 0.199, P = 0.011) and non-inoculated fermented porridge samples (R² = 0.157, P = 0.040); other pairwise comparisons did not yield significant differences.

Genus-level taxonomic profiles revealed clear differences in bacterial community composition among Munkoyo beverages prepared with different starter plant species (Fig. 3). Munkoyo inoculated with R. insignis and R. heterophylla displayed highly similar bacterial communities, characterized by the predominance of Enterococcus, together with lower relative abundances of Clostridium sensu stricto 1, Pediococcus, Lactococcus, and Enterobacter. A number of sequences in these samples remained assigned as unidentified genera within the fifteen most abundant taxa. In contrast, Munkoyo prepared with E. holubii roots was predominantly dominated by Bacillus, although one biological replicate exhibited increased abundances of unidentified genera and Enterococcus, producing a community profile more similar to those observed in the two Rhynchosia species. Technical replicates clustered closely within each biological replicate, indicating high reproducibility of the fermentation process. The non-inoculated control samples displayed highly consistent bacterial communities across replicates and were primarily composed of Enterococcus, Clostridium sensu stricto 1, and Enterobacter, with substantially lower proportions of unidentified genera than inoculated fermentations.

Figure 3.

Genus-level relative abundance plot depicting bacterial community composition depicting the top 15 bacterial genera, with remaining taxa grouped as ‘Other’, found in Munkoyo, produced using three distinct Munkoyo plant species: R. insignis (RI), R. heterophylla (RH), and E. holubii (EH). The ‘No roots’ (NR) category represents samples where no roots were added. On the X-axis, the letters A, B, and C after the plant species name indicate the biological replicates and numbers 1, 2, 3 identify the technical replicate. All root samples were collected from the same site within the Kasempa district.

The PICRUSt2 analysis revealed that between plant species, the predicted functional pathways of the bacterial communities in the Munkoyo beverage were those involved in metabolism (75.98%), genetic information processing (12.27%), and environmental information processing (11.75%) (Table S1). Particularly, functional pathways involved in phosphotransferase systems (PTS), ribosome, and nucleotide excision repair highlighted genetic information processing (Supplementary Fig. S5). Also, starch and sucrose metabolism, amino-sugar metabolism, and riboflavin metabolism were highly abundant, signifying increased energy, amino acid, and vitamin metabolism (Supplementary Fig. S5). However, the differential abundance analysis revealed nonsignificant differences in predicted functional pathways (Supplementary_File1), suggesting that microbial communities across Munkoyo produced from different plant species harboured a similar functional repertoire of predicted genomic pathways.

Influence of plant species on primary metabolites and volatile organic compounds in Munkoyo beverage

Principal component analysis (PCA) was used to assess the similarities and dissimilarities in primary metabolites and VOC profiles of Munkoyo samples fermented with different plant species as inoculum (Fig. 4). Primary metabolites, including lactate, formate, maltose, acetate, glucose, acetoin, and ethanol, were compared. PCA analysis revealed that the variation in the metabolite data could be adequately explained by the first two principal components, accounting for 59.36% of the total variation (Fig. 4a). PC1 explained 40.28% of the variation, while PC2 accounted for 19.08%. The Munkoyo samples clustered primarily according to the plant species used as fermentation inoculum, with PC1 segregating E. holubii from R. insignis and R. heterophylla, and PC2 further differentiating R. insignis from R. heterophylla and E. holubii. To further interpret the separation observed in the PCA of primary metabolites, the contribution of individual compounds to the principal components was examined (Supplementary Table S2A) and visualized in a PCA biplot (Supplementary Fig. S6). Variation along the first PCA axis was primarily driven by opposing contributions of acetoin and lactate (negative loadings) versus ethanol and acetate (positive loadings), indicating a shift between different fermentation end-products. In contrast, the second PCA axis was largely influenced by glucose (strong negative loading) and formate (positive loading), with additional contributions from maltose, suggesting differences in residual sugars and fermentation intermediates. Together, these patterns indicate that plant species-dependent differences in Munkoyo are associated with distinct metabolic pathways, reflecting variation in carbohydrate utilization and organic acid and alcohol production.

Figure 4.

PCA of (a) primary metabolites and (b) volatile organic compounds in 72-h Munkoyo samples using roots from three plant species: Eminia holubii (EH), Rhynchosia heterophylla (RH), and Rhynchosia insignis (RI). Negative controls (NR) lacked added roots. Root samples were from Kasempa district. Original values are ln(x + 1)-transformed, unit-variance scaled, and SVD with imputation calculated principal components. Axes: PC1 and PC2 with a contribution to total variance in %. Prediction ellipses (P = 0.95) show group observation bounds.

Similarly, VOC profiles varied among Munkoyo samples prepared with the different plant species (Fig. 4b). PC1 explained 39.04% of the variation, while PC2 explained 29.32%, collectively accounting for 68.36% of the total variation. As observed in the primary metabolites analysis, PC1 of the VOC profiles predominantly distinguished E. holubii from R. insignis and R. heterophylla Munkoyo, while PC2 further separated R. insignis from R. heterophylla and E. holubii Munkoyo. PERMANOVA analysis corroborated these findings, indicating that the plant species used to prepare Munkoyo significantly influenced both primary metabolites (R² = 0.753, P < 0.001) and VOC blends (R² = 0.651, P < 0.001) in Munkoyo beverages. To further interpret the separation observed in the VOC PCA, the contributions of individual compounds to the principal components were examined (Supplementary Table S2B) and visualized in a PCA biplot (Supplementary Fig. S7). Compounds with the highest loadings on PC1 included several esters and alcohols such as butanoic acid esters and 1-butanol, as well as hydrocarbons (e.g. 2,4-dimethylhexane and 4-methylheptane), indicating that variation along PC1 was primarily driven by differences in ester- and hydrocarbon-related aroma compounds. In contrast, PC2 was largely influenced by compounds such as acetic acid, ethanol, phenolic compounds, and furan derivatives, suggesting differentiation associated with fermentation-derived metabolites. Together, these results indicate that the observed separation between plant species in VOC profiles is driven by distinct combinations of fermentation products and aroma-active compounds.

Influence of plant geographic origin on pH, titratable acidity, and viable cell count of LAB in Munkoyo beverage

The initial pH of maize porridge without added roots was 6.70 (SD < 0.01), while for all other samples, the initial pH ranged between 6.57 and 6.70 (Fig. 5a). Munkoyo samples initiated with roots from Kasempa and Mumbwa showed fastest acidification, reaching a pH below 5.0 after 24 h. After 72 h, there were no significant differences in the pH levels for all samples (ANOVA; F_3,29 = 1.36, P = 0.276). However, the pH levels in samples inoculated with roots, ranging between 4.30 and 4.36, were lower than in non-inoculated samples, 4.68 (SD = 0.21). The initial TA for non-inoculated maize porridge was 0.01 (SD = 0.00), significantly lower than the initial TA for all other samples, around 0.04 (Fig. 5b). After 72 h, significant differences in the TA of all samples were observed (ANOVA; F_3,29 = 9.87, p < 0.001). Notably, the TA for Kasempa root-inoculated samples was the highest at 0.25 (SD = 0.07), whereas those inoculated with roots from Mumbwa and Rufunsa were similar in TA, at 0.19 (SD = 0.01) and 0.18 (SD = 0.02), respectively. The lowest TA was recorded in the non-inoculated samples, 0.11 (SD = 0.02).

Figure 5.

Development of pH (a), titratable acidity (%) (b), and viable cell counts of (facultative) anaerobic LAB (c) during the fermentation of Munkoyo prepared using R. insignis roots collected from three geographic origins, KA: Kasempa, MU: Mumbwa and RU: Rufunsa. Negative control samples (NR) consisted of cooked maize porridge without added roots. All samples were incubated at 28°C for 72 h. Error bars represent standard error. Results are shown as mean values from nine independent experiments, from three biological replicates and three technical replicates. For LAB cell counts, the missing data points indicate values below the detection limit of 2.5 log₁₀(CFU/ml).

The initial LAB plate counts were highest for Munkoyo prepared using roots from Kasempa (3.5 log₁₀CFU/ml, SD = 0.2) and Mumbwa (3.5 log₁₀CFU/ml, SD = 0.3) (Fig. 5c). The initial LAB counts for Rufunsa-inoculated samples were 2.8 log₁₀CFU/ml, SD = 0.4. Initial LAB counts for non-inoculated maize meal porridge were below the detection limit (< 2.5 log10CFU/ml). After 72 h, LAB counts for the root-inoculated samples were comparable, ranging between 7.8 and 8.2 log₁₀CFU/ml, whereas the counts for non-inoculated were significantly lower at 7.5 log₁₀CFU/ml (ANOVA; F_3,29 = 5.16, P < 0.01).

Influence of plant geographic origin on bacterial community composition and predicted functional pathways in Munkoyo beverage

The PCoA analysis of the Munkoyo beverage bacterial communities at 72 h, inoculated with roots of the same plant species but originating from different field locations, revealed notable variations in bacterial community composition. The first PCoA axis explained 32.8% of the variation, with the second axis explaining an additional 19.7%, resulting in a total explained variation of 52.5% (Fig. 6). Samples inoculated with R. insignis roots from Mumbwa and Rufunsa exhibited a close clustering pattern, albeit with a few exceptions. In contrast, Kasempa samples displayed a closer overlap with Mumbwa samples than with Rufunsa samples. Interestingly, the bacterial communities in non-inoculated samples formed a distinct cluster, separate from the bacterial communities in the root-inoculated samples.

Figure 6.

Bray–Curtis distance-based principal coordinate analysis of bacterial communities found in Munkoyo, produced using R. insignis roots from three geographic origins, KA: Kasempa, MU: Mumbwa and RU: Rufunsa. Negative control samples (NR) consisted of cooked maize porridge without added roots. Letters A, B, and C after the field site name indicate the biological replicates and numbers 1, 2, 3 identify the technical replicate. All samples were incubated at 28°C for 72 h.

These findings were further supported by PERMANOVA analysis, which demonstrated that geographic location of the field sampling sites significantly influenced the variation in Munkoyo bacterial communities (R² = 0.158, P = 0.009). Pairwise PERMANOVA comparisons revealed distinct patterns in the Munkoyo bacterial communities due to R. insignis root origin, with Munkoyo inoculated with roots from Mumbwa significantly differing from those of Rufunsa (R² = 0.158, P = 0.009) and non-inoculated samples (R² = 0.519, P = 0.005). Similarly, bacterial communities in Munkoyo prepared with roots from Kasempa exhibited significant differences compared to Munkoyo prepared with roots from Rufunsa (R² = 0.189, P = 0.003) and non-inoculated samples (R² = 0.391, P = 0.008). Additionally, bacterial communities in Munkoyo prepared with roots from Rufunsa differed significantly from non-inoculated samples (R² = 0.589, P = 0.004), while other pairwise comparisons did not yield significant differences.

Genus-level bacterial profiles differed among geographic origins (Fig. 7). Munkoyo prepared with roots collected from Mumbwa and Rufunsa exhibited broadly similar bacterial communities dominated by Enterococcus together with Clostridium sensu stricto 1 and Lactococcus, whereas Kasempa samples displayed greater variability among biological replicates and included prevalence of Streptococcus. Non-inoculated controls formed a distinct community dominated primarily by Enterococcus and Bacillus. Technical replicates showed high reproducibility within biological replicates.

Figure 7.

Genus-level relative abundance plot depicting bacterial communities found in Munkoyo, produced using R. insignis roots from three geographic origins, MU: Mumbwa, RU: Rufunsa, and KA: Kasempa, and Negative control samples (NR). Letters A, B, and C after the field site name indicate the biological replicates and numbers 1, 2, 3 identify the technical replicate.

Predicted functional profiles of Munkoyo microbial communities also varied according to the geographic origin of R. insignis roots. Overall, pathways associated with metabolism (69.61%), genetic information processing (21.26%), and environmental information processing (9.12%) predominated (Table S1). Several pathways differed in relative abundance among field locations, including two-component systems, fatty acid metabolism, α-linolenic acid metabolism, lysine and arginine metabolism, and pathways involved in cofactor and vitamin biosynthesis, including biotin, nicotinate and nicotinamide metabolism (Fig. 8; Supplementary File 2). These findings suggest that geographic origin influences the predicted metabolic potential of microbial communities associated with Munkoyo fermentation.

Figure 8.

Heatmap of the PICRUSt2-predicted functional potential of bacterial communities in Munkoyo prepared with Rhynchosia insignis roots collected from three different geographic locations (sites); MU: Mumbwa, RU: Rufunsa, and KA: Kasempa, and Negative control samples (NR).

Impact of plant geographic origin on primary metabolites and volatile organic compounds in Munkoyo beverage

PCA analysis of primary metabolites and VOC profiles of fermented Munkoyo samples, inoculated with R. insignis roots from different geographic origins, revealed significant impacts of geographic origin of the roots used as inoculum (Fig. 9). For primary metabolites in Munkoyo beverage, the first two principal components explained 49.66% of the total variation, with PC1 and PC2 accounting for 34.41% and 15.25% of the variation, respectively (Fig. 9a). Samples primarily clustered based on geographic location of the field site, with PC1 segregating Munkoyo prepared with roots from Rufunsa from Munkoyo prepared with roots from Mumbwa and Kasempa, and PC2 further differentiating Munkoyo prepared with roots from Kasempa from Munkoyo prepared with roots from Mumbwa and Rufunsa.

Figure 9.

PCA of (a) primary metabolites and (b) volatile organic compounds in 72-h Munkoyo samples R. insignis roots from three geographic origins, MU: Mumbwa, RU: Rufunsa, and KA: Kasempa. Negative control samples (NR) consisted of cooked maize porridge without added roots. Root samples were from Kasempa district. Original values are ln(x + 1)-transformed, unit variance scaled, and SVD with imputation calculated principal components. Axes: PC1 and PC2 with a contribution to total variance in %. Prediction ellipses (P = 0.95) show group observation bounds.

Similarly, VOC profiles exhibited variations among Munkoyo samples prepared with R. insignis roots from different geographic origin. PC1 explained 43.77% of the variation, while PC2 explained 27.44%, collectively accounting for 71.21% of the total variation (Fig. 9b). PC1 predominantly distinguished non-inoculated samples from inoculated ones, and further differentiated Munkoyo prepared with roots from Rufunsa from Munkoyo prepared with roots from Mumbwa and Kasempa. PC2 further differentiated Munkoyo prepared with roots from Kasempa from Munkoyo prepared with roots from Mumbwa and Rufunsa, with minimal overlap.

PERMANOVA analysis supported these findings, indicating that field sampling site geographic location significantly influenced both primary metabolites (R² = 0.627, P < 0.001) and VOC blends (R² = 0.337, P < 0.001) in the Munkoyo beverage.

To further interpret the observed separation among the geographic origin of the roots used to prepare the Munkoyo beverage, the contribution of individual compounds to the principal components was examined (Supplementary Table S3A, B) and visualized using PCA biplots for primary metabolites and VOCs (Supplementary Figs S8 and S9). For primary metabolites, variation along PC1 was mainly driven by differences in fermentation end-products, with organic acids and alcohols contributing strongly to sample separation, while PC2 reflected variation in residual sugars and intermediate metabolites (Table S3 A). These patterns suggest geographic differences in the fermentation dynamics and substrate utilization by the R. insignis root endosphere microbes.

For VOC profiles, separation among geographic origin was driven by a combination of esters, alcohols, acids, and phenolic compounds (Fig. 9b, Table S3 B). Compounds with high loadings on PC1 and PC2 included key aroma-active molecules associated with fermentation and substrate degradation, indicating that geographic origin influences the balance of volatile metabolites produced during fermentation. Together, these results demonstrate that field sampling site geographic location (terroir) shapes both metabolic output and aroma composition of Munkoyo.

Influence of cereal types on pH, titratable acidity, and viable cell count of LAB in Munkoyo beverage

The initial pH of all samples ranged between 6.35 and 6.50 (Fig. 10a), with no significant differences observed. Root-inoculated samples for both cereal types exhibited a faster acidification rate compared to the non-inoculated samples. After 72 h, significant differences in pH levels were observed for all samples (ANOVA; F_3,29 = 7.51, P < 0.01). Non-inoculated sorghum samples had a significantly higher final pH of 5.27, while non-inoculated maize samples had a final pH of 5.03. There was no significant difference in the final pH between sorghum- and maize-based Munkoyo, which were 4.33 and 4.48, respectively.

Figure 10.

Development of pH (a), titratable acidity (%) (b), and viable cell counts of (facultative) anaerobic LAB (c) during the fermentation of sorghum-based (SM) and maize-based (MM) Munkoyo prepared using R. insignis roots collected from the same field site in Kasempa. Negative control samples (MNR for maize and SNR for sorghum) consisted of cooked maize or sorghum porridge without added roots. All samples were incubated at 28°C for 72 h. Error bars represent standard error. Results are shown as mean values from nine replicates for each cereal type and three replicates for each of the non-inoculated samples. For LAB cell counts, the missing data points indicate values below the detection limit of 2.5 log₁₀(CFU/ml).

TA showed similar trends to pH, with inoculated samples exhibiting higher TA, 0.07 (SD = 0.01) for both sorghum and maize-based samples, compared to non-inoculated samples, which showed 0.03 (SD = 0.01) for both cereals (Fig. 10b). After 72 h, significant differences in TA were observed for all samples (ANOVA; F_3,29 = 7.68, P < 0.01). Non-inoculated sorghum samples had a significantly lower final TA of 0.10 (SD = 0.01), while non-inoculated maize samples had a final TA of 0.13 (SD = 0.03). There was no significant difference in the final TA between sorghum and maize-based Munkoyo, which were 0.29 (SD = 0.08) and 0.30 (SD = 0.10), respectively.

The initial LAB plate counts were higher for both sorghum-based and maize-based samples where roots were added compared to porridge without root addition (4.0 log₁₀CFU/ml, SD = 0.7 for sorghum and SD = 0.9 for maize) (Fig. 10c). The initial LAB counts for non-inoculated sorghum and maize meal porridge were below the detection limit (< 2.5 log₁₀CFU/ml). After 72 h, LAB counts for root-inoculated sorghum and maize samples were both 8.0 log₁₀CFU/ml, whereas the counts for non-inoculated were significantly lower at 7.5 log₁₀CFU/ml, SD = 0.06 for non-inoculated sorghum and 7.5 log₁₀CFU/ml, SD = 0.03 for maize.

Influence of cereal type on bacterial community composition and predicted functional pathways in Munkoyo beverage

The PCoA analysis conducted on bacterial communities of sorghum-based and maize-based Munkoyo, along with their respective non-inoculated samples after 72 h of fermentation, revealed significant variations in community composition. The first PCoA axis explained 30.1% of the variation, with the second axis contributing an additional 21.0%, resulting in a cumulative explained variation of 51.1% (Fig. 11). Samples predominantly clustered according to cereal types, although there were substantial overlaps observed between sample types. Furthermore, PERMANOVA analysis demonstrated a significant influence of cereal type on the variation in Munkoyo bacterial communities (R² = 0.267, P = 0.003). Pairwise PERMANOVA comparisons provided additional insights, revealing distinct patterns. Specifically, sorghum-based and maize-based Munkoyo samples exhibited significant differences in bacterial community composition (R² = 0.153, P = 0.005), which also differed from non-inoculated sorghum samples (R² = 0.244, P = 0.007) and non-inoculated maize-based samples (R² = 0.193, P = 0.05). Other pairwise comparisons did not yield significant differences.

Figure 11.

PCoA derived from Bray-Curtis distances of bacterial communities in Munkoyo made using different cereal types, including sorghum meal (SM, So) and maize meal (MM, Ma), along with their respective non-inoculated samples (SNR, So_NR for sorghum-based and MNR, Ma_NR for maize-based). Letters A, B, and C after the cereal type indicate the biological replicates and numbers 1, 2, 3 identify the technical replicate. All samples were incubated at 28°C for 72 h, with R. insignis root samples collected from a single site within the Kasempa district. The percentage of variation explained by each axis is indicated in square brackets, with colours representing different cereal types.

Analysis of the genus-level relative abundance of bacterial communities in Munkoyo made from different cereal types revealed notable variations (Fig. 12). Sorghum-based Munkoyo samples were characterized by higher relative abundances of Enterococcus, Bacillus, and Lactococcus, whereas maize fermentations contained lower proportions of these genera. Technical replicates were highly consistent within biological replicates. In contrast, uninoculated maize and sorghum porridges exhibited highly similar bacterial communities despite differences between the inoculated fermentations.

Figure 12.

Genus-level relative abundance plot depicting bacterial communities found in Munkoyo, made using different cereal types, including sorghum meal (So) and maize meal (Ma), along with their respective non-inoculated samples (So_NR for sorghum-based and Ma_NR for maize-based). Letters A, B, and C after the cereal type indicate the biological replicates and numbers 1, 2, 3 identify the technical replicate. All samples were incubated at 28°C for 72 h, with R. insignis root samples collected from a single site within the Kasempa district.

Predicted functional profiles of the Munkoyo microbial communities associated with the two cereal substrates were similarly dominated by pathways involved in metabolism (76.18%), followed by genetic information processing (12.40%) and environmental information processing (11.41%) (Table S1). The most abundant predicted pathways included two-component systems, fatty acid metabolism, α-linolenic acid metabolism, lysine and arginine metabolism, and vitamin metabolism, including vitamin B6 and folate biosynthesis (Supplementary Fig. S10). However, differential abundance analysis revealed no significant differences in predicted functional pathways between maize- and sorghum-based fermentations (Supplementary File 3), indicating that despite taxonomic differences, both cereal substrates supported microbial communities with similar predicted functional capacities.

Impact of cereal type on primary metabolites and volatile organic compounds in Munkoyo beverage

PCA analysis was used to assess the similarities or dissimilarities in primary metabolites and VOC profiles between sorghum-based and maize-based Munkoyo, along with their respective non-inoculated samples, revealing noteworthy variations (Fig. 13). For primary metabolites, the data variation was effectively explained by the first two principal components, collectively accounting for 56.43% of the total variation (Fig. 13a). PC1 accounted for 30.37% of the variation, with PC2 contributing 26.06%. Samples primarily clustered according to cereal types with the non-inoculated samples clustering closely together and root-inoculated samples also clustering together.

Figure 13.

PCA of (a) primary metabolites and (b) volatile organic compounds in 72-h Munkoyo made using different cereal types, including sorghum meal (SM) and maize meal (MM), along with their respective non-inoculated samples (SNR for sorghum-based and MNR for maize-based). Original values are ln(x + 1)-transformed, unit variance scaled, and SVD with imputation calculated principal components. Axes: PC1 and PC2 with a contribution to total variance in %. Prediction ellipses (p = 0.95) show group observation bounds.

Regarding VOC profiles, PC1 explained 46.98% of the variation, while PC2 explained 20.81%, collectively representing 67.79% of the total variation (Fig. 13b). PC1 primarily differentiated the VOC profiles of root-inoculated samples from non-inoculated ones. For the root-inoculated samples, sorghum-based Munkoyo samples clustered closely together, while maize-based Munkoyo samples showed some overlap with the sorghum-based ones. Furthermore, PERMANOVA analysis showed that cereal type significantly influenced both primary metabolites (R² = 0.604, P < 0.001) and VOC blends (R² = 0.357, P < 0.001) in Munkoyo beverage.

To further interpret the observed separation between sorghum- and maize-based fermentations, the contribution of individual compounds to the principal components was examined (Supplementary Table S4A, B) and visualized using PCA biplots for primary metabolites and VOCs (Supplementary Figs S11 and S12). For primary metabolites in Munkoyo, PC1 was strongly influenced by core carbohydrate and fermentation-related compounds, with maltose, glucose, and ethanol showing high negative loadings, while acetoin contributed positively (Table S4 A). This suggests that PC1 primarily reflects differences in carbohydrate availability and early fermentation activity between cereal types. PC2 was largely driven by organic acids, particularly lactate (strong positive loading) and formate (strong negative loading), indicating that this axis represents variation in fermentation end-product accumulation and acidification patterns (Table S4 A). Together, these patterns suggest cereal-dependent differences in substrate composition and downstream microbial metabolism, with potential implications for fermentation efficiency and acid production dynamics in Munkoyo.

For VOC profiles, Munkoyo sample separation was driven by a broad suite of fermentation- and substrate-derived volatiles, including esters, alcohols, acids, aldehydes, and phenolic compounds (Supplementary Table S4B). Compounds such as octanoic acid, ethyl esters (e.g. butanoic acid ethyl ester and acetic acid ethyl ester), and phenolic derivatives (e.g. 4-ethylphenol and 4-ethyl-2-methoxyphenol) showed strong contributions to PC1 and PC2, highlighting their role in distinguishing cereal-based fermentations. In contrast, several aliphatic hydrocarbons and ketones (e.g. 4-methylheptane, 2,4-dimethylhexane, and 2-octanone) loaded negatively, suggesting an opposing chemical signature associated with specific cereal substrates. Overall, these results indicate that cereal type not only influences the concentration of key fermentation metabolites but also shapes the balance between desirable aroma-active esters and phenolic compounds versus more neutral hydrocarbon and ketone fractions.

Collectively, these findings demonstrate that cereal type is an important determinant of both primary metabolic fluxes and volatile aroma composition in Munkoyo, with sorghum- and maize-based fermentation exhibiting distinct biochemical signatures that likely reflect differences in substrate composition, microbial activity, and fermentation pathways.

Discussion

Munkoyo is an important source of energy and nutrients for many people in Zambia and surrounding countries. In this study, we conducted an in-depth investigation into the influence of plant species used as fermentation starters, their geographic origin, and the cereal types used on the characteristics of the beverage. The primary goal was to determine how these factors impact the microbial communities, pH and titratable acidity, and metabolic and VOC profiles of Munkoyo beverage. Our findings highlight the significant role of plant species in shaping Munkoyo characteristics, with notable variations in the beverage’s bacterial communities, acidity, viable LAB cell counts, and metabolite profiles across the different Munkoyo plant species. Additionally, geographic origin of the Munkoyo plants and cereal types contributed to differences in Munkoyo characteristics, albeit to a lesser extent than the Munkoyo plant species.

Plant-driven variations in Munkoyo bacterial communities and beverage characteristics

Previous studies showed that for the production of Munkoyo beverage, Munkoyo roots are the primary source of fermenting microbes (Phiri 2019). Furthermore, it was recently shown that the root endosphere bacterial communities differ significantly between Munkoyo plant species (Kalumbilo et al. 2026). Based on these findings, we hypothesized that these microbial differences would manifest in the beverage’s bacterial community composition, acidity, and metabolite profiles. Our results confirmed this hypothesis, showing that bacterial communities and other Munkoyo properties indeed varied across the plant species used as fermentation starters. Specifically, R. insignis and R. heterophylla Munkoyo were dominated by Enterococcus, whereas E. holubii Munkoyo exhibited a prevalence of Bacillus. The presence of these taxa that were consistently abundant across all Munkoyo samples suggests their fundamental role in the fermentation process. However, the presence of unique microbial taxa in each plant species points to the potential for plant-specific fermentation dynamics. Previous studies on microbial communities in Munkoyo beverage have reported that these bacterial genera are among the most abundant LAB in Munkoyo (Schoustra et al. 2013, Phiri et al. 2019). Additionally, several members of these bacterial taxa, such as Enterococcus species within Enterococcaceae, are known for their acidification capabilities and have been identified as important fermentative agents in other fermented products like sourdough (Oshiro et al. 2021). Similarly, Bacillus species within Bacillaceae, which were prevalent in E. holubii Munkoyo, are known for their robust enzymatic activities that contribute to starch breakdown and flavour development in various traditional fermented foods (Li et al. 2023).

The microbial metabolism of carbohydrates, lipids, and amino acids is known to be fundamental to the development of the physicochemical properties of fermented products, including the production of volatile organic compounds, which are major determinants of aroma profiles (Smid and Kleerebezem 2014, Liu et al. 2023). Our analysis of the functional potential of the microbial communities in the Munkoyo samples indeed showed prevalence of these metabolic pathways. Similarly, we found that functional pathways associated with the phosphotransferase system were relatively enriched, signifying ecological adaptation and carbohydrate uptake within Munkoyo derived from varied plant species (Jamal et al. 2013). However, we did not find significant differences in metabolic potential of the microbial communities in Munkoyo prepared with the different plant species. Nevertheless, the Munkoyo beverage did show distinct profiles in VOCs and primary metabolites between Munkoyo derived from Rhynchosia and Eminia.

Several studies have reported that LAB typically dominate the fermentation of cereals, where they metabolize carbohydrates to form lactic acid, a primary fermentation product in cereal fermentations (Blandino et al. 2003, Mashau et al. 2021). We found that different plant species led to variations in the acidity of the beverage, with significant differences observed in Munkoyo pH and TA after 72 h of fermentation, with pH ranging between 3.5 and 4.5 and TA ranging between 0.2 and 0.5. Similar results on the acidity of the beverage have been reported in previous studies (Schoustra et al. 2013, Phiri et al. 2019). These differences can be attributed to the distinct bacterial communities associated with each plant species, which could possess different capacities for acid production (Kalumbilo et al. 2026). Previous studies on LAB genera that are also known to be important in Munkoyo fermentation (Schoustra et al. 2013) revealed that members of the genus Weissella, which consist of obligately heterofermentative LAB, produce less titratable acids from the same amount of glucose compared to the homofermentative Lactococcus species (Björkroth et al. 2015, Teuber 2015). Lactococcus species are generally known to produce lactic acid rapidly (Gutiérrez-Méndez et al. 2010). Furthermore, we observed that the initial viable LAB cell counts were higher when roots were added, further indicating that the roots act as the primary source of fermenting microbes. Among the species tested, the initial viable LAB cell counts were particularly high for E. holubii, leading to more rapid acidification (Fig. 1). This suggests that differences in Munkoyo characteristics are not only due to variations in microbial composition but also due to differences in the abundance of LAB in the roots.

Interestingly, for the non-inoculated samples, the initial LAB cell counts were below the detection limit; however, after 72 h, LAB cell counts were not significantly different from those in the inoculated samples. This suggests that spontaneous fermentation occurred, likely due to LAB present in the environment. This finding suggests the potential contribution of environmental LAB to the fermentation process, even in the absence of deliberately added microbial inoculants. However, the environmental LAB may differ in their fermentation abilities from the LAB directly derived from the Munkoyo roots. A study by Ruiz-Rodríguez et al. (2019) on LAB isolated from various wild fruits and flowers demonstrated that LAB colony counts obtained by direct isolation were variable and dependent on the sample source. Similarly, isolating LAB from the roots of different Munkoyo plant species could provide a comprehensive understanding of LAB diversity and abundance, which is crucial for studying fermentation dynamics and identifying the organisms that contribute to the distinct characteristics of Munkoyo.

Microbial communities involved in fermentation produce various volatile organic compounds in addition to primary metabolites such as different sugars (Longo and Sanromán 2005). The observed variations in bacterial communities between Munkoyo produced with different plant species were reflected in the metabolic profiles of the beverage. Our results indicate that the primary metabolites and VOC profiles varied between Munkoyo species, as indicated by PCA showing clear separations between plant species. Variations in metabolite profiles in Munkoyo beverages sampled from different regions have been reported; however, the origin of the variation was not revealed (Phiri 2019). As the same cereal was used in that comparative study, the variations in metabolic profiles of the Munkoyo beverage from different geographic regions may have been due to using different plant species to initiate fermentation.

Moreover, the differences in the bacterial communities of non-inoculated samples resulted in different metabolic profiles compared to root-inoculated samples, especially for primary metabolites. One possible reason for this could be the spontaneous fermentation by environmental LAB differed from the LAB populations introduced via the Munkoyo roots. This environmental LAB, which may have originated from cereal dust or other sources, contributed to the fermentation process, leading to distinct metabolic profiles.

Additionally, Munkoyo roots are known to be sources of enzymes and flavonoids, which play crucial roles in the fermentation process and the development of characteristic flavours (Zulu et al. 1994, 1997). The absence of these specific root-derived enzymes and flavonoids in non-inoculated samples likely resulted in a fermentation process that differs from the traditional Munkoyo, leading to a beverage with different sensory attributes. This suggests that while spontaneous fermentation can occur, the resulting product may not possess the typical characteristics of Munkoyo, emphasizing the importance of the roots in achieving the desired beverage profile.

Plant geographic origin influences bacterial communities and metabolic profiles in Munkoyo beverage

Our previous study on Munkoyo root microbiomes (Kalumbilo et al. 2026) revealed a significant influence of the geographic origin of the plant samples on the root endosphere microbiomes. Against this background, we hypothesized that using the same plant species but sourced from field locations in different geographic districts as a starter would result in differences in the Munkoyo bacterial communities, physicochemical characteristics (pH and titratable acidity), primary metabolites, and VOCs. Our results validate the hypothesis, revealing a significant influence of field sites on the bacterial communities and metabolic profiles of the beverage. However, no significant differences in the pH were found, though TA varied across the geographic origins of the roots used in the fermentation.

PCoA analysis and PERMANOVA results showed that the geographic location of the R. insignis field sampling sites significantly influenced the variation in Munkoyo bacterial communities, with Munkoyo prepared with roots from Rufunsa being significantly different from Munkoyo prepared with roots from Kasempa and Mumbwa. One possible reason for this could be that Rufunsa is geographically distant from Kasempa and Mumbwa, potentially leading to differences in the bulk soil microbiome from which Munkoyo plants assemble their root microbiomes (Kalumbilo et al. 2026). A study by Zhou et al. (2022) reported that variation in microbiome composition in the legume Melilotus officinalis was associated with the geographic distance between the locations where the plant species was collected. The bacterial community composition, as indicated by genus-level relative abundance analysis, showed variations notably between the root-inoculated and non-inoculated fermented porridge samples. Munkoyo beverage prepared with roots were predominantly dominated by Enterococcus together with Clostridium sensu stricto 1, Lactococcus, and Streptococcus, while non-inoculated controls were dominated by Enterococcus and Bacillus.

Whereas no significant differences in the pH of the final product were observed, variations in titratable acidity (TA) were noted, with Munkoyo prepared with roots from Kasempa having relatively higher TA. One possible reason for this could be differences in bacterial communities, as evidenced by the presence of members of the LAB genus Streptococcus in Munkoyo prepared with roots from Kasempa that could have different acid production capabilities (Björkroth et al. 2015, Teuber 2015). Isolating and characterizing LAB species from Munkoyo roots of the same species from different geographic origins could be crucial for studying fermentation dynamics and understanding which organisms contribute to the differences in Munkoyo characteristics. Additionally, measuring the physicochemical properties of soil in the different regions could decipher potential links between soil properties and specific plant-derived LAB. Besides bacteria, yeasts are also known to play an important role in flavour formation in Munkoyo (Zulu et al. 1997). It could be useful to characterize the yeasts associated with Munkoyo plant species.

The differences in bacterial communities in Munkoyo prepared with roots from different geographic origins were reflected in the metabolic profiles of the Munkoyo beverage. Moreover, several microbial metabolic pathways differed in relative abundance between Munkoyo prepared with roots from different geographic origin, notably fatty acid and α-linolenic acid metabolism, lysine and arginine metabolism, and pathways involved in biosynthesis of vitamins, including biotin, nicotinate and nicotinamide. It is known that fermentation of cereals can result in vitamin enrichment in the fermented product, such as in Munkoyo, and that these levels can differ depending on the substrate (Chileshe et al. 2020, Ignat et al. 2020). However, as in this second experiment, we used the same substrate our results indicate that there may be source material geographic effects (terroir) on the potential and actual metabolic profile of Munkoyo, in line with studies on wine (Liu et al. 2019). Furthermore, PERMANOVA analysis showed that the R. insignis sample geographic origin significantly influenced both primary metabolites and VOC blends. PCA results showed that the metabolic profiles of Munkoyo prepared with roots from Mumbwa and Rufunsa were closer than Munkoyo prepared with roots from Kasempa. In addition to bacterial community differences, such as the presence of LAB from the genus Streptococcus in Kasempa samples, the environmental conditions might have an influence. Kasempa is located in agroecological zone III (with annual rainfall above 1000 mm), whereas Mumbwa and Rufunsa are in agroecological zones II and I, respectively and receive annual rainfall below 1000 mm. Regional climatic variation has been reported to influence plant root microbiomes (Sharma et al.  2022).

Cereal types influence bacterial community composition and metabolite profiles in Munkoyo beverage

We hypothesized that using different cereal types, even with the same Munkoyo root species from the same field origin, would lead to variations in the bacterial communities, and physicochemical (pH and titratable acidity), primary metabolites, and VOCs of Munkoyo. The influence of cereal types on Munkoyo bacterial community composition and metabolite profiles was evident in our study. Sorghum-based Munkoyo samples exhibited distinct bacterial community compositions compared to maize-based Munkoyo, as supported by PCoA and PERMANOVA analyses. Specifically, sorghum-based samples were characterized by a higher prevalence of Enterococcus, Bacillus, and Lactococcus, compared to maize-based fermentations. This divergence likely stemmed from the inherent compositional variations between sorghum and maize, which could have influenced the growth dynamics of LAB and other microorganisms, consequently shaping the resultant metabolite profiles (Dabija et al. 2021). Notably, these differences were particularly pronounced in the primary metabolites, with significant variations observed between sorghum-based and maize-based Munkoyo samples. Additionally, VOC blends also exhibited differences between cereal types, further emphasizing the impact of cereal choice on the overall metabolic profile of Munkoyo beverage.

At the level of microbial metabolic potential, we found the presence of alpha-linoleic acid metabolism across both cereal types, suggesting a role of LAB in lipid metabolism, concordant with previous reports which revealed that Lactiplantibacillus spp. were capable of converting linoleic acid to functional and health-beneficial fatty acid metabolites (Aziz et al. 2021). Functional pathways associated with vitamin B6 and folate were also relatively abundant across both cereal types, highlighting B-complex vitamin enrichment potential by fermenting microbial communities. Likewise, cereal-based fermented products have been shown to harbour B-complex vitamins (Chileshe et al. 2020). Also, two-component systems and amino acid metabolism, including lysine and arginine, were predicted to be relatively enriched across both maize- and sorghum-based Munkoyo. Both the two-component systems and amino acid metabolism can be enhanced during LAB fermentation as a mechanism to mediate acid stress (Zhang et al. 2012, Hu et al. 2018). This can ultimately influence aroma profiles in the fermented environment (Liu et al. 2023).

These findings highlight the importance of considering cereal diversity in Munkoyo production, as even though the metabolic potential of the Munkoyo microbiomes was similar, the microbial community composition and the resultant metabolite profiles of maize- and sorghum-based Munkoyo were distinct, ultimately contributing to the diverse sensory characteristics of the beverage. Therefore, next to the origin of the Munkoyo roots, the cereal substrate type will be important to consider in efforts of upscaling of traditional Munkoyo processing.

Conclusion

The results of this study support the hypothesis that plant species, geographic origin, and cereal types all influence the microbial communities, physicochemical properties, primary metabolites, and VOCs of the Munkoyo beverage. Significant differences were observed in the bacterial community composition and metabolic profiles of Munkoyo based on the plant species used as fermentation starters, their geographic origin, and the types of cereals involved. These findings enhance our understanding of the factors shaping microbial communities in traditional Munkoyo fermentation, emphasizing the importance of considering both plant-specific and environmental factors. Moreover, the choice of cereals plays a crucial role in determining the beverage’s microbial and metabolic profiles, highlighting opportunities for optimizing Munkoyo production to enhance food security and economic development. These findings can inform further studies aimed at optimizing and upscaling of traditional Munkoyo processing.

Author contributions

Mubonda Kalumbilo (Conceptualization [equal], Formal Analysis [lead], Investigation [equal], Methodology [equal], Writing – original draft [lead], Writing – review & editing [equal]), Mukuka Mwaba (Formal Analysis [equal], Investigation [equal], Writing – original draft [equal], Writing – review & editing [equal]), David Chuba (Conceptualization [equal], Supervision [equal], Writing – original draft [supporting], Writing – review & editing [equal]), Agripina Banda (Conceptualization [equal], Methodology [equal], Supervision [equal], Writing – original draft [supporting], Writing – review & editing [equal]), Anna Alekseeva (Formal Analysis [equal], Methodology [equal], Writing – original draft [supporting], Writing – review & editing [equal]), Shepherd Nehanda (Data curation [equal], Formal Analysis [equal]), Oscar van Mastrigt (Formal Analysis [supporting], Methodology [equal], Writing – original draft [supporting], Writing – review & editing [equal]), Sijmen E. Schoustra (Conceptualization [equal], Methodology [equal], Resources [equal], Supervision [equal], Writing – original draft [supporting], Writing – review & editing [equal]), Eddy J. Smid (Conceptualization [equal], Methodology [equal], Resources [equal], Supervision [equal], Writing – original draft [supporting], Writing – review & editing [equal]), Gerlinde B. De Deyn (Conceptualization [equal], Data curation [equal], Funding acquisition [lead], Project administration [lead], Resources [equal], Supervision [equal], Writing – original draft [supporting], Writing – review & editing [equal])

Conflicts of interest

The authors declare no conflict of interest.

Data availability

All data supporting the findings of this study, including 16S rRNA gene sequencing data, metabolite profiles (HPLC and GC–MS), and physicochemical measurements, are available in the 4TU—ResearchData repository at https://doi.org/10.4121/1d03823f-22fa-47b2-b060-e057d878992a. The dataset is currently under embargo and will be made fully accessible upon publication of the associated article.