Optimization of cacao beans fermentation by native species and electromagnetic fields

Abstract

Acid and bitter notes of the cocoa clone Cacao Castro Naranjal 51 (CCN 51) negatively affect the final quality of the chocolate. Thence, the fermentative process of cocoa beans using native species and electromagnetic fields (EMF) was carried out to evaluate the effect on the yield and quality of CCN 51 cocoa beans. The variables magnetic field density (D), exposure time (T), and inoculum concentration (IC) were optimized through response surface methodology to obtain two statistically validated second-order models, explaining 88.39% and 92.51% of the variability in the yield and quality of the beans, respectively. In the coordinate: 5 mT(D), 22.5 min (T), and 1.6% (CI), yield and bean quality improved to 110% and 120% above the control (without magnetic field). The metagenomic analysis showed that the changes in the microbial communities favored the aroma profile at low and intermediate field densities (5–42 mT) with high yields and floral, fruity, and nutty notes. Conversely, field densities (80 mT) were evaluated with low yields and undesirable notes of acidity and bitterness. The findings revealed that EMF effectively improves the yield and quality of CCN 51 cocoa beans with future applications in the development and quality of chocolate products.

1. Introduction

The fact that various microbial species in different conditions and environments are susceptible to being influenced by electromagnetic fields (EMF) is a prominent issue in the scientific community today [[1], [2], [3]]. However, thermal and non-thermal effects can mainly cause these changes. The first is related to the production of heat by cells, and the second is cell growth [4,5].

It has been seen that the non-thermal effects of the electromagnetic field (EMF) can affect not only microbial growth but also metabolism [6,7]; however, it has also been shown on several occasions that the metabolic activity induced by EMF is linked to the activation of some areas of DNA [8]. Furthermore, some reports describe that EMF can change many processes in a microorganism, such as increasing antibiotic sensitivity and membrane transport [2,9], changing morphology [10], biofilm formation [11], and reproduction [12]. These electromagnetic field effects have been found in distinct species of bacteria, both Gram-positive and Gram-negative [9,13], and in yeast and filamentous fungi [5].

Although these effects cannot be explained totally, cells' response to EMF varies depending on the density, frequency, and exposure time [14]. In addition, it has also been seen that the first concentration of microorganisms and the characteristics of the culture medium are crucial factors that influence the cellular response to EMF [5].

Fermentation of cocoa beans goes through several stages where native microbial species are selected naturally, giving the chocolate its typical organoleptic characteristics [15]. As a result, each variety of cacao plants has distinctive characteristics from the others. The quality of cocoa depends on the first instance of the microbial species' interaction with beans during the fermentative process in that ecosystem [16]. Different microbial groups such as yeasts, lactic acid bacteria, and acetic bacteria that intervene in the fermentation of cocoa beans have been described, among which stand out genera Saccharomyces, Lactobacillus, and Acetobacter [[15], [16], [17]]. However, fermentative processes of the cocoa beans are developed under uncontrolled conditions that produce contamination and decrease the quality of the final product. These damages are irreversible and depend on the type and the growth of the MO naturally selected [[18], [19], [20]]. Thus, an effective controlled fermentative process and optimization of parameters like time, temperature, and inoculum concentration are essential.

The Cocoa Castro Naranjal 51 (CCN 51) is considered a coa with an intense chocolate flavor, widely productive, resistant, and adaptable to various environmental conditions, however, its acid and; however have a negative impact on the aroma fine market [21]. fine aromaudies have proposed several alternatives to improve its sensory profile and mitigate the negative impacts of this importaeffectse on the cocoa market [22]. These alternatives include the control of the fermentative process [23], the introduction of grain pre-drying steps [24], and the use of starter microorganisms during grain fermentation [25], among the rest [26]. However, research in this field is still in its infancy, and novel alternatives are required to redefine the flavor profile of CCN51. To study the changes induced by OMF in the fermentation process of CCN 51 cocoa beans and optimize this process, a 2^3 factorial design extended to a central composite design (CCD) was carried out to understand the fermentation performance and the quality of the bean of the variety of cocoa beans CCN 51, exposed to OMF of 5–80 mT, inoculum concentration of 1–2%, and exposure time of 15–30 min, in contrast to the non-exposed ones. The procedure was resolved by measuring the grains at the end of fermentation (6–7 days), to estimate the degree of fermentation resulting in each experimental run.

2. Materials and methods

2.1. Electromagnetic fields (MEF) system

Helmholtz coils designed on a laboratory scale were used to generate the electromagnetic field as described by Hu [26]. This system consisted of two circular coils of 0.205 m in radius and connected in series, running with a variable voltage source at 60 Hz, generating a variable density magnetic field (of 1–120 mT). The samples were placed between both coils with a separation of 0.205 m during the magnetic field treatment. To control EMF, a sensor Hall-effect Tesla meter (SS49E) was connected in the middle of the coils. A specialized electronic board (Arduino nano) was used to convert this sensor-emitted signal to a digital signal in the computer [27]. The magnetic field was read in the interface of the Laboratory Virtual Instrument Engineering Workbench (LabVIEW) software version 9.0 (Fig. 1).

Fig. 1.

MEF system. (1) Coil 1 (2); Coil 2; (3) Magnetic sensor (Hall-effect Tesla meter); (4) Variable voltage source; (5) Voltage indicator; (6) Current Indicator; (7) Ammeter; (8) Arduino nano; (9) Magnetic field flux lines; (10) Sensor module; (11) PC.

2.2. The cocoa beans processing

2.2.1. Fermentation procedure

The cocoa pods were selected without any defects or damage caused by pests and with a uniform coloration (an indicator of the degree of maturity) [28]. The pods obtained were duly packed in polyethylene bags, transported, and stored (7–12 °C) in a cold room until the test (no more than 32 h). Each pod was adequately treated with sodium hypochlorite at 200 ppm in aseptic procedures for 5 min. Then 3 kg of cocoa beans were obtained for each treatment. The experimental units obtained consisted of 3 kg of cocoa beans of the CCN 51 variety ready to ferment. The beans were placed in the fermentation chamber for 168 h, where the temperature (37 °C ± 0.5) and the humidity (85% ± 0.8) of the system were controlled.

2.2.2. Drying and roasting the beans

After fermentation, the obtained beans were placed in different metal meshes, to avoid contact between replicates, inside rotary drying equipment at 60 °C for 4 h until reaching a humidity of 10%. Later, in the same equipment, the temperature was raised to 120 °C for 30 min for the roasting process. The beans obtained from each experimental unit were sensory evaluated.

2.2.3. Preparation of the inoculum

The natural inoculum was obtained from cocoa pulp CCN 51 previously fermented at 37 °C for 18 h. For this, the fermented grains were decanted, obtaining a homogeneous suspension of the pulp as inoculum [[28], [29], [30]]. This suspension was inoculated by spraying in the fermentation systems at 0.5–2.5% v/v [30]. The same inoculation procedure was performed for each experimental unit.

2.3. Evaluation procedure

2.3.1. Cutting test

The cut test allowed the identification and quantification of the grains. Initially, the fermentation index of each group (described below) was determined according to Refs. [[31], [32], [33]], and later the different fermentation and contamination rates were calculated. One hundred fermented cocoa beans were cut longitudinally to expose the inner surface of the cotyledon [28,29,33]. The beans were classified according to their degree of fermentation into well-fermented (brown or reddish-brown cotyledons, accompanied by well-open streaks), moderately fermented (partially striated cotyledons and purple stripes on the edges), poorly fermented (cotyledons of intense violet color), and contaminated (cotyledon black or gray) unfermented bean highly damaged by insects and fungi [[33], [34], [35]].

2.3.2. Sensorial profile

The pre-dried and roasted cocoa beans were analyzed by a sensory panel composed of seven judges from the National Institute for Agricultural Research (INIAP) of Ecuador trained in the sensory profile of cocoa. The appropriate protocols for protecting the rights and privacy of all judges participants were utilized during the execution of the sensory analysis. Sensory attributes (cocoa, floral, fruity, nutty, sweet, bitter, astringent, green, color, odor, texture) were evaluated using a predetermined visual scale for sensory evaluation of cocoa [34,36]. This scale presents ten measurable points for each attribute from a minimum sensation of the attribute (1) to a maximum (10). Scores were compared to a reference cocoa bean derived from Ecuadorian cocoa beans [21]. In each sample, three measurements of each attribute were made; according to the criteria evaluated by the panel, the instrument's reliability was measured using Cronbach's alpha coefficient (α = 0.879) [37].

2.3.3. Metagenomic analysis

Samples were taken for metagenomic analysis in each experimental run to distinguish the predominant microbial genera during the first 18 h of fermentation.

2.3.3.1. Cocoa beans sample preparation and DNA extraction

The cocoa beans collected at 18 h were resuspended in 10 ml of saline solution and vortexed with an average speed of 30 s. Subsequently, the almonds were removed, and the obtained solution was centrifuged at 6000 rpm. Next, the obtained pellet was resuspended in 2 ml of 1.5 M sorbitol buffer, and this procedure was performed twice. In the last step, the obtained pellet was frozen for 1 h to later proceed with the DNA extraction protocol [38].

The DNA extraction process began with the cell lysis of the microbial communities. Then, it was carried out using the same extraction protocol used by Florac De Bruyn 2015 [39] for coffee seeds, with some modifications in the proportions of chloroform: phenol: alcohol. Isoamyl (49.5: 49.5: 1) is used to remove proteins. During DNA extraction, the QIAamp genomic DNA Kits were also produced with some modifications and using a DNA affinity column with two ultracentrifugation steps. Finally, for each sample, the concentration and quality of the DNA extraction were measured by spectrophotometry (NanoDrop 2000, brand Thermo Fischer Scientific) [40].

2.3.3.2. Amplification and sequencing of microbial communities

The DNA obtained from the microbial communities was amplified by PCR with specific primers of the 16s rDNA region for the V4 sections in F515 (5′-GTGCCAGCMGCCGCGGTAA-3′) and R806 (5′-GGACTACVSGGGTATCTAAT-3′) bacteria. 39). The primers BITS (5'-ACCTGCGGARGGATCA-3′) and B58S3(5′-GAGATCCRTTGYTRAAAGTT-3′) were also used for the ITS region of fungi and yeasts. PCR products were run on agarose gel under standard conditions. PCR products retained on the gel were cleaned using the Kit Wizard SV gel and PCR cleanup System (Promega) [41]. Sizing of sequences (genomic libraries) was performed using Agencourt AMPure XP PCR for magnetic bead purification (Beckman Coulter). Once purified PCR products >100 bp, DNA concentration, and quality were measured by fluorescence (Qubit 2.0; Thermo Fisher Scientific). The amplified DNA sequences of the microbial communities were sequenced by synthesis using the Illumina MiSeq platform (Illumina, San Diego, California, USA) through the VUB-ULB (Brussels, Belgium) inter-university agreement. Results were obtained in FASTQ format for analysis with bioinformatics tools [40,42].

2.3.3.3. Sequence analysis and taxonomic assignment

The sequences obtained from the NGS were subjected to quality cleaning with the DADA2 package version 1.6 [32,33] and performing order and alignment of the sequences for analysis with R language, version 3.6.3. The sequences obtained by Ilumina for bacteria were compared with the SILVA 16s rRNA databases (http://www.Arb-silva.de; version 128). For the sequences amplified with the BITS primer (due to their length), less demanding quality filtering was performed, and these unique sequences were taxonomically classified with the UNITE database (http://www.unite.ut.ee; version 8, sh 99) [43,44]. The relative abundance of metagenomic sequences of the different genera obtained was used to calculate the rates in the groups (LDR, IDR, and HDR) and to determine the Shannon diversity index (H).

2.3.3.4. Biodiversity analysis

A phyloseq object was used to analyze biodiversity, comprising the following files: taxonomic assignment, sequence counter, and test characteristics. The sequences were then grouped by operational taxonomic units (OTU) based on a genus. Next, it was filtered according to a 1E-05 mean read count threshold. Subsequently, the sequencing depth was determined; from this point, found the alpha diversity [45].

2.4. Variable selections

2.4.1. Fermentations yields

Three variables of the fermentation process (Y1, Y2, and Y3) were determined to relate the rates between those exposed and the controls: The total fermentation yield (Eq. (1)), the best fermentation yield (Eq. (2)), and the contamination rate (Eq. (3)). Barrel's criteria (rate greater than or equal to 60%) was used to decide the best rate of fermentation quality [46].

Y1 = ((v3 − v4) /vt)/((w3 − w4)/wt) × 100 (1)

Y2 = (v1/vt)/(w1/wt) × 100 (2)

Y3 = (v4/vt)/(w4/wt) × 100 (3)

where:

• v1 – mean values of well-fermented beans exposed

• v2 – mean values of moderate fermented beans exposed

• v3 – mean values of fermented beans exposed

• v3 – mean values of contaminated beans exposed

• vt – total beans exposed

• w1 – mean values of well-fermented beans controls

• w2 – mean values of moderate fermented beans controls

• w3 – mean values of fermented beans controls

• w3 – mean values of contaminated beans controls

• wt – total beans controls.

2.4.2. Roasted bean quality

The variable quality of roasted beans (Equation (4)) was obtained from the mean of the mean values of each attribute obtained in the sensory test, according to the following formula:

Y4 = ((m1 + m2 + m3)/3)/((n1 + n2 + n3)/3) × 100 (4)

where:

• m1 – mean values of color exposed

• m2 – mean values of texture exposed

• m3 – mean values of odor exposed

• n1 – mean values of color controls

• n2 – mean values of texture controls

• n3 – mean values of odor controls

2.5. Statistical procedures for the optimization process

Optimization procedures were performed using experimental design (DOE) in the Design Expert v 13 software (Stat-Ease). A Face-Centered Cube (α = 1) Central Composite Design (CCCFCD) was carried out (Table 1). Consisted of six center points (0, 0) (for lack-of-fit (LoF) test), 18 axial (–1; +1) points, and 27 factorial (−1; +1) points, making a total of 47 experimental runs performed in random order (Table 1). In the experimental design, three groups are distinguished depending on the density of the magnetic field: LMD (Low Magnetic Density), IMD (Intermediate Magnetic Density Runs), HMD (High Magnetic Density Runs), and three groups are distinguished depending on the starter culture concentration LIC (Low Inoculum Concentration), IIC (Intermediate Inoculum Concentration) and HIC (High Inoculum Concentration Runs).

Table 1.

Matrix of actual and coded factors of the CCCFCD. Responses values of the 2^3 factorial design extended to face-centered cube central composite design: fermentative yield (Y2) and bean quality (Y4); three repetitions, 48 experimental runs.

			Actual and coded factors
			A:MEF density	B:Exposure time	C:Inoculum concentration	Responses values
		runs	(mT)	(min)	% (v/v)	Y21	Y21	Y21	Y41	Y41	Y41
ffactorialpoints	LDR	F1	5 (−1)	15 (−1)	0.5 (−1)	87.94	80.43	79.58	111.27	101.07	113.23
		F2	5 (−1)	30 (+1)	0.5 (−1)	92.46	91.84	91.13	101.62	110.38	106.25
		F3	5 (−1)	15 (−1)	2.5 (+1)	95.66	93.02	88.66	102.25	97.62	92.29
		F4	5 (−1)	30 (+1)	2.5 (+1)	100.21	112.01	107.25	67.86	62.45	65.25
	HDR	F5	80 (+1)	15 (−1)	0,5 (−1)	77.49	82.06	89.65	60.61	67.71	60.10
		F6	80 (+1)	30 (+1)	0.5 (−1)	80.93	76.45	80.78	69.26	59.26	60.92
		F7	80 (+1)	15 (−1)	2.5 (+1)	80.92	76.92	82.96	58.20	48.40	50.05
		F8	80 (+1)	30 (+1)	2.5 (+1)	78.12	72.44	70.56	60.86	57.70	51.31
central point	F9*	42.5* (0)	22.5* (0)	1.5* (0)	112.47	109.40	113.43	88.47	83.34	89.49
axial points	IDR	A10	5 (−1)	22.5 (0)	1.5 (0)	120.82	120.70	115.01	105.75	10.62	109.13
		A11	80 (+1)	22.5 (0)	1.5 (0)	104.05	10.26	102.15	54.27	67.96	69.53
		A12	42.5 (0)	22.5 (0)	2.5 (+1)	125.46	120.59	117.04	67.86	62.45	65.25
		A13	42.5 (0)	22.5 (0)	0.5 (−1)	117.12	114.48	115.59	79.86	68.80	77.07
		A14	42.5 (0)	30 (+1)	1.5 (0)	122.18	116.16	124.53	59.10	75.14	62.80
		A15	42.5 (0)	15 (−1)	1.5 (0)	119.16	119.78	116.41	57.36	65.44	79.35
		A16	42.5* (0)	22.5* (0)	1.5* (0)	118.69	124.37	122.05	79.62	78.15	75.71

Replicas (3) runs (48) blocks (2) center points* (6) factorial points (24) and axial points (18). F: factorial points. A: axial points.

2.5.1. Experimental limits

Factors and corresponding levels were selected to find which of them and their interactions can significantly affect the fermentation process of cocoa beans CCN51 and how they do it. The MEF density (5–80 mT), time (15–60 min), and inoculum concentration (1–1.5% v/v) were the limit selected for the design. The established levels of the factors density and time of the MEF were as far apart as possible to generate a significant difference in the response (Δy); for that, it was proved as a strategy to analyze the entire experimental region delimited by the maximum and minimum range of the electromagnetic device where stable MFE emissions were obtained over time [47]. In the case of inoculum concentration, the percentage ranges (v/v), which do not cause excessive humidity (by exudations) in the fermentation system, were selected [20].

2.5.2. Model fit

Since the responses are proportions that express quantities, second-order polynomial functions were used to fit them. In this sense, the data were also normalized using square root transformations to stabilize the variance and achieve a reasonable and satisfactory fit [47]. In each case, the adjustment of each Model was verified by the statistical analysis of variance, and it was reduced, neglecting the terms that were not significant (p < 0.05). Likewise, each Model tried to find an adequate approximation of the functional relationship between the response variables and factors [48].

2.5.3. Optimization approach

A multi-response optimization approach was carried out to maximize the process variables, using the desirability function provided by the Design Expert optimizer, which is an integral mathematical protocol [47,48]. In these conditions, maximum desirability of 70% was obtained in the following factorial coordinate: A, B, C = 5, 22.5, 1.6., The models obtained were evaluated employing experimental replicates in this coordinate to see the agreement between each response's calculated and experimental values (Fig. S1) [48]. The theoretical and practical results obtained (Table S2) show an agreement in the observed data mean and calculated predicted mean values, for both responses (Y2 and Y4), with a 95% confidence, which ratifies this Model can be used to navigate in the design space (Table S2).

2.5.4. Design output

The F-values of the Y1 and Y3 model responses were insignificant compared to the noise, and significant terms were not found in the model for these responses. However, the quadratic regression model for fermentative yield (Y2) and bean quality (Y4) were significant (p < 0.05) with F-values of 44.79; 97.10, respectively. Consequently, with a probability of more than 99.99% confidence, it is clear that the studied factors influence the yield and quality of beans [48] (Table S2).

For Y2-response, A, B, C, AB, AC, A², B², and C² are significant model terms. For Y4 responses: A, C, A², B², and C² are significant model terms (p < 0.05). The lack of fit F-value in both models is not significant (p > 0.05) compared to the pure error, which suggested that in the study region, both models adequately explain the response variations [44,48] (Table S2).

The Predicted R² of 0.84 for Y2 is in reasonable agreement with the adjusted R² of 0.88. The same occurs for Y4: Predicted R² of 0.90 is in reasonable agreement with the Adjusted R² of 0.92 [48]. The coefficients of variation obtained for each response written down (Y2: CV = 3.51 and Y4: CV = 6.82) proved the reliability of the experiments performed. The residual diagnosis reveals no statistical problems [47,48] (Fig. S1).

2.6. Principal component analysis (PCA)

To analyze the sensory attributes most affected by the magnetic field, a PCA was performed for each DOE treatment (individuals). The results of the data matrix were processed using the R (version 4.1.3) programming language environment with the help of the Bioconductor packages (version 3.14). In the PCA, the multidimensional data of the quality attributes of the previously standardized cocoa beans (x = 0; SD = 1) were projected to a reduced dimensional space of n variables [22,23]. The PC chosen were those with a maximum variance, and the values of each biplot (score) were obtained from a linear combination of the reduced variables [23]. Correlations were also made between other variables to interpret obtained data better.

3. Results

3.1. SRM results

Response-surface and contour graphs (Fig. 2a, b, c, d) show two curves that obey the quadratic terms of 2 s-order polynomial Eqs. (1), (2). Polynomial equations reveal that whether the inoculum concentration or exposure time is high or low, the yield (Y2) and quality (Y4) are consistently lower with increasing field density (A). This implies a strong negative effect of OMF on both factors, time and inoculum concentration (C and B).

Fig. 2.

Response-surface (RS) and contour graphs (CG). (a, d) RS: vertical axes: fermentative yield (a) and bean quality(d), horizontal axes: OMF density and Inoculum concentration; (b, d) CG: vertical axes: Inoculum concentrations, horizontal axes: OMF density, contour line: fermentative yield (b) and bean quality (d); Exposure time supported at best levels.

Consequently, plots show that an increase in OMF density causes a substantial decrease in fermentation yield (Fig. 2a, b) and grain quality (Fig. 2a, c). Conversely, a reduction in OMF density to values close to 5 mT causes an increase in both responses and the ratio between exposed and controls (closer contour lines) (Fig. 2b, d).

The effect of culture concentration and exposure time depends on OMF density values, and better grain quality can be obtained with lower OMF density values (Fig. 2a, b, c, d). In the coordinate: 5 mT(A), 22.5 min (B), and 1.6% (C), yield and bean quality improved to 110% and 120% above the control. However, the graph indicates that at densities below 5 mT, optimum grain quality values could be reached.

Sqrt(Y2) = 10.8528 − 0.370432A + 0.0733862 B + 0.102468C − 0.221457AB − 0.21719AC − 0.641664 A^2 − 0.270573B^2 − 0.331708C^2 (5)

Sqrt(Y4) = 9.04851 − 1.23934 A + 0.0241599 B − 0.256541C + 0.46818 A^2 − 0.581385 B^2 − 0.3519C^2 (6)

3.2. PCA results

Two principal components (PC1 and PC2) explained 77% and 10% of the total variance of the data set, respectively, where two groups separated by the central axis were distinguished (Fig. 3 a). The notes with the highest floral, fruity, and nutty scores perceived by the judges were found in group A, associated with the LDR and IDR. On the contrary, the highest scores in astringency, acidity, green, and bitterness were found in group B, corresponding to HDR (opposite vectors in the biplot graph showing a negative correlation) (Fig. 3a).

Fig. 3.

Biplot PCA: (a) 2D view (group A: LDR and IDR, group B HDR); Spider plot: (b) LDR and IDR scores, (c) HDR scores.

In the radial graph (Fig. 3c, d), two sensory profiles are also distinguished and associated with PCA results. LDR and IDR (Fig. 3b) corresponded with group A (Fig. 3a), and HDR (Fig. 3c) corresponded with group B. In summary, the judges perceived a greater intensity of aroma, flavor, and better texture of the bean in LDR and IDR, having a more favorable effect on the quality of the roasted bean (fine flavors) than the HDR scores (unpleasant flavors). This result corresponds with the polynomial equations obtained in DOE, where an increase in field density decreases yield and grain quality while a decrease has the opposite effect.

3.3. Correlations results

In the correlation graph between microbial genera and sensory quality attributes (Fig. 4), the following positive correlations stand out: Weisella, Frauteria, and Pichia with flavors cocoa and fruity, Luteibacter with flavors cocoa and floral, Gluconobacter and Luteibacter with cocoa and floral (blue ovals). On the other hand, the negative correlations of the genera Acetobacter, Bacillus, Klebsiella, Isatechia, and Mucor are also remarkable, with the sensory attributes: of florality, fruitiness, nuttiness, and cocoa (red ovals) (Fig. 4).

Fig. 4.

Correlation graph between microbial genera and quality attributes: blue ovals: positive correlation; red ovals: negative correlation; darker ovals: higher correlation. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

3.4. Metagenomic results

After 24 h of fermentation, five yeast genera and 13 bacterial genera were identified in the different groups (Fig. 5a, b, c). The highest RAR of these metagenomic sequences corresponded to LDR (39.8%), followed by IDR (32.09%), and finally, HDR (28.03%) (Fig. 5d). Microbial genus greater diversity could also be noted in LDR and IDR.

Fig. 5.

Heatmap of microbial genera by groups. (a) Metagenomic sequences of the HDL group: light yellow color represents low RA. a dark color represents high RA; (b) IDL group of metagenomic sequences: light blue color represents low RA. a dark color represents high AR; (c) LDL group of metagenomic sequences: light green color represents low RA. a dark color represents high RA; (d) Alluvial plot: yellow color represents RAR of HDL group; blue color represents RAR of IDL group and. Green color represents RAR of LDL group. Shannon diversity index (H). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

The relative abundance (RA) of enterobacteria genus in order of declining was: Tatumella (72,83), Pseudomonas (22,05), and Klebsiella (18,11). Tatumella was found in the three groups, but Pseudomonas and Klebsiella were present exclusively in HDR (Fig. 5a). The highest RAR of enterobacteria metagenomic sequences corresponded to IDR (36.05%), followed by LDR (32.44%), and finally, HDR (31.50%) (Fig. 5d).

Three yeast genera were identified in order of declining abundance: Hanseniaspora (93.0), Issatchenkia (68.6), and Pichia (0.77) (Fig. 5a, b, c). Hanseniaspora was present in the three groups together with the genus Pichia. However, the genus Issatchenkia only occurred in IDR and HDR (Fig. 5a, b). Metagenomic sequences of filamentous fungi of the Mucor (2.37) genus were notable in HDR (Fig. 5a), but these sequences did not appear in LDR and IDR (Fig. 5b, c). The yeasts' RAR in order of declining abundance corresponded to IDR (36.8%), LDR (34.2%), and HDR (31.9%) (Fig. 5d). Other sequences above the genus, not identified, were also found. In decreasing order of RAR, they correspond to IDR (36.8%), LDR (33.8%), and HDR (31.9%) (Fig. 5d).

The relative abundance of metagenomic sequences of two LAB genera was highlighted: the Lactobacillus (36.49) genus, which was only present in IDR, and Weissella (3.95), present in all three groups (Fig. 5a, b, c). The highest RAR of these metagenomic sequences corresponded to IDR (49.14%), followed by LDR (44.38%) and HDR (6.48%) (Fig. 5d).

Two genera of AAB were found: Gluconobacter (6.42), present in all three groups, with a higher RA, and Acetobacter (0.14), which only occurred in IDR with a lower RA (Fig. 5b). In addition, an increased relative abundance of the Frauteria genus was also found a genus with metabolic characteristics like AAB (Fig. 5b). In these cases, the RAR was as follows: LDR (86.24%), IDR (10.64%), and HDR (10.33%) (Fig. 5d).

Another important finding was the presence of the genus Bacillus (20.47), which was also presented exclusively in LDR (Fig. 5c). Other three genera of bacteria soil were also found in decreasing order of abundance: Stenotrophomonas (85.71), Luteibacter (51.02), and Variovorax (28.57), which are not frequent among CFCs but appeared with high RA in LDR (Fig. 5c). Metagenomic sequences of the genus Cellulosimicrobium (6.58) in LDR and IDR were also highlighted (Fig. 5b, c). Although, the CFS metagenomic analysis did not previously detect this genus.

4. Discussion

The results suggest that the OMF density had the highest contribution to fermentation quality and performance. In this case, it was successfully implemented to obtain the best conditions that influenced the efficiency of the fermentation process. OMF treatment in CCN 51 cocoa beans fermentation process effectively increased both the fermentative yield and bean quality, implying a satisfactory alternative for the optimization fermentation process of this essential product. A quadratic behavior in the Model of both responses was observed when the OMF changed. This behavior in both models was consistent with the findings of Anaya et al. (2021), who reported that OMF 60 Hz/220 V and 5 mT influences the quadratic behavior of the colony's diameter of various species of fungi [5].

4.1. Changes in sensory profile

Cocoa beans' sensory quality and performance improvement results induced by MEF during cocoa fermentation agree with different processes modulated by electromagnetic fields. Konopacka et al. (2019) [7], and Andrade et al. (2021) [49], found that a rotating magnetic field of 16–18.5 mT by continuous exposure up to 72 h and a magnetic flux density of 10 mT by 24 h, respectively, increases the bioethanol production process by S. cerevisiae. Boeira et al. point out that a magnetic field of 35 mT exposure between 24 and 40 h reduces 56.5% nivalenol in alcoholic fermentation by S. cerevisiae [50]. Also, Lin et al. demonstrated that the inactivation efficacy of PMF on E. coli O157:H7 was proportional to the pulse number and intensity of PMF [51]. In contrast, Nakasono et al. report that a 300 mT exposure for 5–24 h in yeast cultures has no effect [52].

4.2. Microbial ecosystem changes

The higher microbial diversity in LDR and minor diversity in HDR at 24 h could be explained based on EMF effects on microorganisms. Several authors confirm that EMFs can condition both the stimulation and the inhibition of microbial species, affecting the dynamics of microbial populations in each ecosystem [[6], [7], [8], [9], [10], [11], [12], [13], [14]]. For example, Strašák et al. demonstrated that OMF of density 2.7–10 mT, frequency 50 Hz, and time 0–12 min reduces the growth of Escherichia coli [53]. On the other hand, Novák et al. report that OMF density of 10 mT, frequency 50 Hz, and time of 24 min decreased the growth of S. cerevisiae [54]. This fact was also corroborated by Bayraktar et al., who also found that OMF density of 5 mT, frequency of 160 Hz, and time of 30 min decrease the growth of S. cerevisiae [55]. For their part, Gao et al. also found that OMF of density 0.2–1 mT, frequency 50 Hz, and time 4–8 h increased the growth of Aspergillus niger [56].

4.2.1. Effects in enterobacterial communities

In SCFs, the enterobacterial colonizes the early hours and is involved in the pectinolytic degradation of cocoa pulp and the assimilation of citric acid [57]. Recent investigations confirm that Tatumella correlated positively with gluconate in the cocoa pulp and with ethyl isovalerate and benzaldehyde in the cocoa beans, which conditions good flavors in the chocolate [58]. However, Pseudomonas is a primary colonizer that prolonge the lactic acid phase and limits enzyme activities [59]. In this sense, the high notes of bitterness and acidity in HDR may be related to the development of this genera.

4.2.2. Effects on yeast communities

The genera Hanseniaspora, Issatchenkia, and Pichia have also been reported as predominant in the first hours of FCs due to their active participation in the enzymatic degradation of cocoa pulp. These yeasts have high fermentative activity and condition important flavor precursors in the first hours of the SFC by transforming sugars into aldehydes, esters, and alcohols, compounds related to pleasant notes in the cocoa bean [57,58]. Recent studies indicate that Hanseniaspora and Pichia can produce monoterpenes (nerolidol, linalool, geraniol, and citronellol) related to the floral flavor of cocoa beans [59,60]. The absence of the Saccharomyces genus occasionally isolated from SFCs is also a notable result in this study since most of the SFCs highlight the prevalence of this yeast genus. The Saccharomyces genus is particularly sensitive to MEF and can be stimulated or inhibited at different densities [12,52,54]. The relative abundance of these genera in HDR and HDR could also condition the differences in the notes found as a function of the field density.

4.2.3. Effects in LABs communities

LABs are generally considered an important group developed in the anaerobic phase of SFCs and associated with producing cocoa bean flavor and color precursor molecules [[60], [61], [62], [63], [64], [65], [66]] by transforming citric acid and pulp sugars into lactic acid, acetic acid, and mannitol [60,61]. However, no other LAB genera were found except Weissella and Lactobacillus. Although several authors highlight the role of the LAB in CFS, others question its relevance in CFS [67]. Another notable result was the low AR of the Lactobacillus genus, which was only found in LDR despite being the most abundant genus in SFCs, which is evidence of the inhibitory effect of magnetic fields on some bacterial genera [68].

4.2.4. Effects in AABs communities

In the aerobic phase, AABs are also the most common in SFCs. However, they can also be present from the beginning of fermentation, oxidizing ethanol and lactic acid (produced by yeasts and LABs) to acetic acid [[69], [70], [71], [72]]. This process promotes the death of the embryo and the development of different flavor precursors in SFC [31,61,63]. Although the Acetobacter genus is the most abundant in SFCs, in its place, Frateuria stood out in LDR, IDR, and HDR. This genus produces acetic acid from glucose and ethanol and can oxidize lactate. Consequently, the relative abundance of these genera in LDR and, to a lesser extent, in HDR could also determine differences in flavors and odors found as a function of magnetic field density.

4.2.5. Effects in other microbial genera

Unlike other genera that commonly develop more abundantly in the anaerobic phase, the high abundance of four soil bacterial genera, such as Stenotrophomonas, Luteibacter, Variovorax, and Bacillus, was very notable in LDR, which shows more remarkable development of species at low field density. Most of these genera are related to the pectinolytic decomposition of the pulp, facilitating the aeration of the medium and the start of the aerobic phase [[50], [51], [52], [53], [54],63]. Recently, this activity in the Bacillus genus has been related to flavor precursors and the formation of pyrazines [71,72].

4.2.6. EMF conclusive evidence

There is evidence that EMF can condition changes in microbial metabolism (inducing the synthesis of different primary and secondary metabolites), such as those reported by Liao et al. [73], who demonstrated that the low-frequency magnetic field improved the performance of pigments of Monascus purpureus significantly reducing citrinin production and Alvarez et al. [74], who affirm that the densities of the sinusoidal magnetic field of 5–20 mT in 4–12 h managed to increase the production of nisin by Lactococcus lactis. In this sense, it could be guessed that both the interaction and single effects of the OMF and the concentration of the inoculum could have a direct and indirect impact on the different autochthonous microbial species of the cocoa bean [60,75], which could also occur, giving rise to favorable conditions in the fermentation process of the cocoa bean. These findings could be the slight notes of flavor fine found in CCN 51.

On the other hand, although a change in magnetic energy is insignificant compared to covalent bond energies, it has been seen that the magnetic field can or not change the energy levels in a biological system and its constituent molecules by changing the speed of biochemical reactions (especially those involving pairs of radicals) [58,59,76]. Thus, changes in the genetic expression of various microbial species induced by EMFs have also been shown, such as that reported by Wang et al., 2017 who found that ambient magnetic fields of 5 mT increase the expressions of genes related to signal transduction, cell motility and the activity of ammonium-oxidizing aerobic bacteria [77].

Since the metagenomic study was limited to the first 24 h of fermentation and the optimal growth of the main genera in SFCs is established between 24 and 96 h [15], the microbial population analysis found could have been underestimated. However, the first microorganisms to colonize can determine the course and progress of fermentation [47,55,60]. In this regard, the change observed in the cocoa bean fermentation could be an indirect effect of modifications in both the gene expression of the microbial species present and their metabolic activity.

Thus, in the coordinates obtained in the study, favorable conditions were generated for good fermentative performance and grain quality. Consequently, the appearance of undesirable microbial genera (Pseudomonas, Mucor) in HDR could lead to poor grain quality. In contrast, the microbial genera found in LDR and IDR could develop more favorable characteristics in the grain, conditioning better fermentative performance and sensory quality. Another hand, the genus Cellulosimicrobium found in LDR and IDR could be a contaminant obtained during the DNA extraction process as described in other research, as this genus is usually found in the soil [78].

On the other hand, although the process was optimized, the study coordinates were limited to the OMF emission range of the electromagnetic device, where stable emissions were obtained over time. However, results indicate that optimal values of cocoa bean quality could still be reached at densities below 5 mT. Therefore, before carrying out a process on a larger scale, evaluation of the lower OMF density ranges that fell outside the limits of this study should be considered.

5. Conclusions

This study proposes a new concept to improve fermentation performance and cocoa bean quality using electromagnetic fields and a new direction to explore the mechanism of magnetic field influence on the cocoa bean fermentation process. In addition, the observed phenomena should also be explained in terms of new process parameters (e.g., abiotic factors, microbial populations, volatile components, and kinetic variables). Nevertheless, despite the production costs required to control the large-scale cocoa bean fermentation process, the quality assurance of cocoa providing a stable market is promising. Thus, technological synergies will bring advances and future contributions to the science and industry of chocolate.

Author contributions

The authors contributed in the following manner to this research project. Conceptualization, Methodology, Validation, Formal analysis: T.M.G.A: Resources, writing – original draft; J.R.N: Resources, Investigation, supervision, writing – original draft & review; L.R.G: Investigation, Resources, Writing – review; LSG: writing – review, visualization & editing. All authors have read and agreed to the published version of the manuscript.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article: