Spontaneous fermentations were performed at the Fazenda Apucarana located in the Cerrado Mineiro region (18°55’59.4” S, 46°50’41.5” W) at Minas Gerais, Brazil. Freshly harvested coffee (Coffea arabica var. Catuaí) cherries were depulped using a BDSV-04 depulper (Pinhalense, São Paulo, Brazil) obtaining beans with a surrounding layer of mucilage (15). Fermentations were conducted for 24 h in cement tanks with a nominal volume of 4.5 m³, containing 20 kg of depulped beans and approx. 500 L of fresh water, in accordance with the local wet processing method. At the end of the process, fermented beans were sun-dried for 20 days to 11-12% moisture, as measured by a moisture meter (model AL-102 ECO; Agrologic, São Leopoldo, Brazil). Environmental temperature during the experimental procedure was 24-32 °C (day) and 12-15 °C (night). Samples (fermenting coffee bean pulp mass) were collected at random at 0, 12 and 24 h for HTS and target metabolic analysis.

For extraction of total DNA from the samples, 1 mL of coffee bean pulp mass was centrifuged at 12 000×g for 1 min (centrifuge model 5430; Eppendorf, Hamburg, Germany). Cell pellet was resuspended in 500 µL of Tris-EDTA, homogenized with 10 µL of lysozyme solution (20 mg/mL; Sigma-Aldrich, Arklow, Ireland) and incubated at 30 °C for 60 min. Then, 50 µL of sodium dodecyl sulphate (SDS; 10%, by mass per volume) and 10 μL of proteinase K solution at 20 mg/mL (Sigma-Aldrich) were added to the lysis solution, followed by homogenization and incubation at 60 °C for 60 min. A volume of 150 µL of phenol/chloroform (25:24; Sigma-Aldrich) was added, homogenized by inversion and centrifuged at 12 000×g (model 5430R; Eppendorf) for 5 min. Supernatant was removed and the DNA was precipitated with 3× (by volume) absolute ethanol (Sigma-Aldrich). Pellets was washed with 80% ethanol, dried and resuspended in Mili-Q® ultrapure water (Merck, Kenilworth, NJ, USA). Total DNA was quantified with the Nanodrop 2000 instrument (Thermo Fisher Scientific, Inc., Waltham, MA, USA).

A fragment of the 16S rRNA gene was amplified from the total DNA extracted using primers for the V4 region (bases 515 to 806), containing complementary adaptors for Illumina platform (16) using KlenTAQ polymerase (Sigma-Aldrich). Amplification was performed using the degenerated primers 515F (5’-GTGCCAGCMGCCGCGGTAA-3’) and 806R (5’-GGACTACHVGGGTWTCTAAT-3’), where M is A/C, H is A/C/T, V is A/C/G and W is A/T (17). Bar-coded amplicons were generated by PCR under the following conditions: 95 °C for 3 min, followed by 18 cycles at 95 °C for 30 s, annealing at 50 °C for 30 s, extension at 68 °C for 60 s, and final extension at 68 °C for 10 min. Samples were sequenced in the MiSeq (Illumina Inc) platform using 500 V2 kit (Illumina Inc), following standard Illumina protocols.

Data generated by sequencing went through a rigorous quality system that involved: (i) identification and removal of sequences containing more than one ambiguous base (N), and (ii) evaluation of the presence and complementarity of primer and barcode sequences. Chimeric sequence detection, removal of noises from pre-cluster and taxonomic attribution were also performed using standard parameters of QIIME (Quantitative Insights Into Microbial Ecology) software package, v. 1.9.0 (17). Applying the UCLUST method (18), sequences presenting identity above 97% were considered the same operational taxonomic units (OTUs) according to the SILVA database (19).

The concentration of reducing sugars (glucose and fructose), organic acids (acetic, succinic, lactic and propionic acids) and ethanol was determined during coffee bean fermentation by high-performance liquid chromatography (HPLC). Samples were centrifuged at 6000×g (centrifuge model CT-6000; Cientec, Porto Alegre, Brazil) and filtered through 
0.22-µm pore size filter (Sartorius Stedim, Goettingen, Germany) in order to remove debris. Analysis parameters were determined according to de Carvalho Neto et al. (20). Filtered samples were injected into HPLC system equipped with an Aminex HPX 87 H column (300 mm×7.8 mm; Bio-Rad, Richmond, CA, USA) and a refractive index (RI) detector (model HPG1362A; Hewlett-Packard Company, São Paulo, Brazil). The column was eluted in isocratic mode with a mobile phase of 5 mM H₂SO₄ at 60 °C and a flow rate of 0.6 mL·min.

Table 1 shows the evolution of sugar consumption, metabolite formation and pH decrease during fermentation of coffee bean pulp. The observed increase in the concentration of reducing sugars (glucose and fructose) at 12 h of fermentation can be attributed to the hydrolysis of sucrose by the action of yeast invertase (21). These sugars were partially consumed after 24 h of fermentation, with a final residual content of 3.2 and 4.5 g/L of glucose and fructose, respectively. Lactic acid (0.32 g/L) was the most important organic compound formed during fermentation, followed by succinic and acetic acids (0.08 and 0.05 g/L, respectively). Lactic acid is an important organic compound for coffee bean fermentation that assists in the coffee pulp acidification without interfering with the final product quality (22). The accentuated production of lactic acid is in agreement with the strong dominance of lactic acid bacteria found in the present study (Fig. 1), resulting in pH decrease from 5.3 to 4.0 at the end of fermentation (Table 1). The reduction of pH below 4.5 is a widely used method by coffee producers to determine the end of fermentation of coffee bean during wet processing (23).

A total of 440 524 high-quality sequences of the hypervariable V3 region of the 16S rRNA gene region were obtained after trimming on the Illumina MiSeq sequencing, with an average length of 250 bp. A great coverage was obtained in all samples as demonstrated by the rarefaction curves (Fig. 2).

Studies evaluating the microbiology of coffee fermentation have been performed over the last 100 years in several coffee-producing regions, evidencing the dominant species during the post-harvest processing (6, 7, 24-29). On average, nine bacterial genera had been reported in previous studies using culture-dependent methods (7, 15, 30-33). Our work demonstrates that these findings are an underestimate, since over eighty genera of bacteria have been identified by HTS. High frequency and abundance of readings corresponding to Proteobacteria (e.g. Erwinia, Pseudomonas and Methylobacterium) and Firmicutes (e.g. Bacillus, Fructobacillus, Leuconostoc and Lactococcus) were observed. The possible habitat origins of these microbial groups are: human contact, e.g. Pseudomonas sp., Enterobacter, Erwinia and Actinobacteria (34), soil or aerial parts of coffee plants, e.g. Mesorhizobium, Methylobacterium, Stentrophomonas, Sphingobium and Sphingomonas (35-37), the water source used for wet processing, e.g. Planctomyces, Luteimonas, Devosia and Brevundimonas (38), and the air surrounding the fermentation tank, e.g. Janthinobacterium, Pedobacter, Burkholderia and Kaistobacter (39). These findings indicate the need for a program of research to understand the microbial ecology origin of coffee cherries and processing sites.

The rich and complex bacterial diversity revealed in this study demonstrates the potential of coffee terroir as a source of microorganism species with biotechnological application. An example is the first report of the presence of Fructobacillus in coffee fermentation. This LAB group has a unique biochemical metabolism when compared to other LAB, having preference for fructose consumption and the necessity of an elector acceptor when in presence of glucose (40). Fructobacillus microorganisms were found in gastrointestinal tracts of insects feeding on fructose-rich diet and presented symbiotic interactions with its hosts (41, 42). A survey of previous studies demonstrates significant amount of residual pulp fructose at the end of coffee fermentations conducted under field conditions (20), even by using selected starter cultures (4, 15, 43). With these findings, the isolation and further implementation of Fructobacillus may assist in the fructose metabolism, contributing to drying of coffee beans.

Bacterial composition and dynamics shown in Fig. 1 reveal that, despite the presence of a high bacterial diversity associated with coffee fermentation environment, several microorganisms are suppressed by the growth and dominance of LAB group. Reads assigned to LAB genera, including Lactobacillus, Pediococcus, Enterococcus, Leuconostoc, Lactococcus and Fructobacillus, corresponded to 26.32% at the start of the process and reached a total of 97.59% of the total operational taxonomic units (OTU) at 24 h. The high availability of fermentable sugars coupled with the low presence of dissolved oxygen creates a propitious environment for the rapid growth and colonization of these species, which promote an efficient conversion of sugars into mainly lactic acid (44).

Within the LAB group, Leuconostoc and Lactococcus shared dominance. Species of Leuconostoc, such as L. mesenteroides, L. pseudomesenteroides and L. citreum, have already been reported as dominant LAB in coffee fermentations performed in Mexico, Colombia, India and Taiwan (6, 45, 46), while Lactococcus species dominates coffee fermentations performed in Taiwan and Brazil (45, 47). Co-dominance of LAB enables the production of a wide range of organic compounds (e.g. acetate, acetaldehyde, ethanol, short-chain fatty acids) by heterofermentation (e.g. Leuconostoc sp.) and a high production of lactic acid through homofermentation (e.g. Lactococcus sp.), which promotes yeast growth and reduces the prevalence of spoilage microorganisms.