The coffee-machine bacteriome: biodiversity and colonisation of the wasted coffee tray leach

Microbial communities are ubiquitous in both natural and artificial environments. However, microbial diversity is usually reduced under strong selection pressures, such as those present in habitats rich in recalcitrant or toxic compounds displaying antimicrobial properties. Caffeine is a natural alkaloid present in coffee, tea and soft drinks with well-known antibacterial properties. Here we present the first systematic analysis of coffee machine-associated bacteria. We sampled the coffee waste reservoir of ten different Nespresso machines and conducted a dynamic monitoring of the colonization process in a new machine. Our results reveal the existence of a varied bacterial community in all the machines sampled, and a rapid colonisation process of the coffee leach. The community developed from a pioneering pool of enterobacteria and other opportunistic taxa to a mature but still highly variable microbiome rich in coffee-adapted bacteria. The bacterial communities described here, for the first time, are potential drivers of biotechnologically relevant processes including decaffeination and bioremediation.

We have sampled the inner drip tray below the capsule container, in which coffee lyxiviate accumulates. We have analysed ten domestic and semi-domestic machines and studied the dynamic colonization process in a brand new Inissia Krups machine operated in our laboratory. This is the first systematic analysis of the microbial diversity associated to coffee machines. Our results may shed light on the microbial arsenal of caffeine degraders with important implications for both medicine and biotechnology.

Results and Discussion
The waste coffee in the capsule container of nine different Nespresso machines operated for at least one year was sampled (Fig. 1A). In one case (CityZ model), the cup tray was also sampled independently as it does not connect with the capsule container. The high throughput sequencing and analysis of the 16S rRNA gene amplicons from all the machines revealed a significant bacterial diversity, with the total number of identified genera ranging from 35 to 67. Although relatively similar microbial profiles were detected, there was an important variation in the frequency of particular taxa. Enterococcus sp. and Pseudomonas sp. proved to be the main taxa as they were moderately to highly abundant in nine out of the ten samples analysed. Other frequent genera were Stenotrophomonas, Sphingobacterium, Acinetobacter and, to a lesser extent, Coprococcus, Paenibacillus or Agrobacterium. Dysgomonas was very frequent in the Inissia machine, accounting for 15% of the sequences (Fig. 1B). No differences were detected between machine models (Table 1) or use (domestic vs. communal).
One of the two most frequent genera found in the coffee machines was Pseudomonas, which is also one of the few reported examples of a caffeine-degrading bacterium. Indeed, Pseudomonas sp. has been known to catabolise caffeine since the seventies 13 , and is reported to degrade up to 15 g/L of caffeine through an N-demethylation reaction, which along with C-8 oxydation represent the two potential catabolic pathways 14 . Species reported to display caffeine degradation abilities are P. alcaligenes 15 and P. putida (strains C1, CBB1 or CBB5). In fact, P. putida N-demethylation genes have been used to genetically   16 , and caffeine removal from sewage by bioremediation with P. putida has also been proposed 17 . The abundance of Enterococcus spp. in caffeine-rich leach might not necessarily involve unreported caffeine degradation abilities in Enterococcus, but it might simply be a consequence of tolerance to certain caffeine levels. The same applies to other frequent taxa. Interestingly, this genus has previously been associated with coffee 18 , along with several others detected in this work. For example, Acinetobacter sp. has been isolated during coffee fermentation 19 , while Stenotrophomonas sp., Curtobacterium sp., and Pseudomonas sp. are abundant in the coffee seed 20 .
The colonisation process of the wasted coffee leach was studied in an experiment using a brand new Krups Inissia machine (located in a separate room within our laboratory). The experiment lasted two months, during which leach samples were taken and bacterial diversity analysed, with a significant variation in the taxonomic profiles detected. The initially high species richness was substituted by a relatively simpler, but still highly variable, species composition (species richness significantly dropped 14 days after the beginning of the experiment; t-test p-value = 0.039). During the first 11-13 days, Pantoea sp., Cloacomonas sp. and, to a lesser extent, Brevundimonas sp. were relatively abundant but amounts decreased to undetectable levels by the end of the experiment. All these taxa were largely substituted by Pseudomonas sp., Acinetobacter sp., and Sphingobium species, which reached a peak and then fluctuated (Sphingobium sp., Bacillus sp.) or reached the highest levels at the end of the experiment (Pseudomonas sp., Acinetobacter sp.) as shown in Fig. 2A. The first 30 days exhibited greater instability in the bacterial communities, as deduced by the consecutive peaks of very abundant taxa, which were substituted by a more balanced bacterial composition after one month. As in other studies on different environments 21,22 , these results strongly suggest a long ecological succession during the first month, in which generalist bacterial taxa, including enterobacteriaceae genera such as Pantoea, are the first colonizers but are then displaced by successive waves of other taxa. The main keyplayers observed during this succession were, in order ( Fig. 2B): enterobacteria (genus Pantoea; peaking 4-11 days after the beginning of the experiment), Firmicutes (three genera of the bacillaceae family: Bacillus, Terribacillus, Paenibacillus; peaking after 14-21 days); and, finally, the sphingomonadales genus Sphingobium (proteobacteria), the actinomycetales genus Curtobacterium (actinobacteria), and the pseudomonadales genus Acinetobacter (proteobacteria), peaking after 28, 31 and 49 days, respectively. These taxa gave way to a different bacterial profile dominated by Pseudomonas sp. and Enterococcus sp. after two months of the experiment. This profile was very similar to that found in the nine other coffee machines sampled (Fig. 1B) which had been operated for a longer time, suggesting that the particular physico-chemical conditions (cycles of high temperature, constant caffeine accumulation, etc.) of coffee leach, rather than the influence of the user or the number of uses, are the main force shaping the composition of the microbial community. A mathematical modelling performed on the dynamic series of 16S rRNA gene data revealed statistically significant correlations among the detected taxa, indicating that the distribution of bacterial genera in time is not random (Fig. 3).
Most of the taxa we identified during the colonisation process of the coffee machine operated in our laboratory have previously been found in natural coffee-related environments. Species belonging to the genera Acinetobacter and Bacillus, and also some enterobacteria, have been detected during the natural fermentation of coffee beans 23,19 , whereas Paenibacillus and other Bacteroidetes and Firmicutes species have proved abundant in the composting process of coffee hulls 24,25 . Despite some reports describing the ability of different Sphingobium species to degrade toxic molecules, such as bisphenols 26 and hydroquinones 27 , this is the first report where Sphingobium sp. has been associated to a caffeine-rich environment.
In addition to the 16S rRNA gene monitoring, we followed up changes in the coffee leach microbial diversity through scanning electron microscopy (SEM). Figure 4 shows a dynamic series of samples taken at different time points (4,8,14, and 21 days after the first day of operation) during the first month. Microbial biomass increased throughout the analysis, and variations in the composition and viscosity of the coffee leach were also evident. For example, a filamentous matrix was observed at days 8 and 21 (Fig. 4). At day 14, the sample was dominated by a single shape of bacterial cells, which interestingly coincided with an overwhelming relative abundance of Bacillus spp. in that sample ( Fig. 2A). Further experiments are needed to determine whether microbial community changes are the cause or the effect of the variations in the composition of coffee leach as shown by SEM.
Our results show, for the first time, that coffee leach from standard capsule machines is a rich substrate for bacterial growth; that caffeine content does not prevent a rich bacterial biodiversity from rapidly colonising coffee leach; and that microbial succession from an initial pool of generalist bacteria gives way to an apparently coffee-adapted but still highly variable bacteriome. This bacteriome is rich in species previously reported to be associated with the coffee plant and/or the coffee fermentation processes. Colonising bacteria might be of environmental origin (no cultivable microorganisms nor bacterial DNA was detected in coffee capsules, data not shown), whereas heterogeneity of bacterial composition may relate to factors such as cleaning habits and, specially, the frequency of machine use (with higher frequencies presumably correlating with increased volume and temperature of the coffee leach). Further studies comprising more coffee machines, deep genome sequencing of the microbial communities therein, and The presence of bacterial genera with pathogenic properties and the fast recovery of the communities after rinsing the capsule container, strongly suggest the need for frequent maintenance of the capsule container of these machines. Maintenance should employ bacteriostatic compounds, and avoid contact of the coffee leach with other parts of the machine to avoid unintended contamination of the beverage. On the other hand, the resistant microbial communities we describe here (microbial consortia, individual caffeine degrading/tolerant species or as a source of metabolic pathways and genes) may represent a promising tool for biological coffee decaffeination processes and for environmental caffeine decontamination.

Methods
Sampling. Nine Nespresso machines, which had been operated for at least one year, used either at home (domestic) or in academic departments, institutes or biotechnology companies (communal) in the Valencia (Spain) area were aseptically sampled ( Table 1). The coffee lixiviate from wasted capsules present in the drip tray placed below the capsule container was sampled using a sterile Pasteur pipette. In one case (CityZ), the small space under the drip grid and cup support facilities was sampled and treated separately ( Fig. 1A; Table 1). In all cases, the average temperature of the room where the coffee machines were operated was close to 25 °C.
Additionally to these machines, a Krups Inissia machine was purchased for the present work and normally operated in our laboratory for five months with a daily use of around 1-5 capsules per day. Sampling was performed at increasing time lapses and consisted of the removal of most (except 5 mL, approximately) of the lixiviated liquid. When the volume of the lixiviate was smaller than 5 mL, no samples were taken. All the detachable pieces (drip grid and cup support, drip tray and capsule container were thoroughly rinsed once (day 28) with tap water.
In all cases, 2 mL aliquots of the samples were immediately used for DNA isolation, whereas the remaining volume was stored at − 80 °C.
Scanning electron microscopy. Aliquots from each sample were filtered through a 0.2 micrometre filter (Corning Inc.), fixed with a 2% paraformalin -2.5% glutaraldehyde solution, and lightly washed with filtered-sterile phosphate buffer (PBS). Then, small pieces of the filters were placed inside microporous capsules (30 μ m pore size, available from Ted Pella Inc. product number 4619) and subjected to successive dehydration steps in growing ethanol solutions up to 100% ethanol, following critical point drying in an Autosamdri 814 (Tousimis). These fragments were then placed on SEM stubs with silver conducting paint TAAB S269 and examined under a scanning electron microscope Hitachi S-4100. DNA isolation and PCR amplification. Two mL aliquots of each sample were centrifuged at 11,000 g for 3 min and the resulting pellets were washed twice with sterile PBS buffer (NaCl 8 g/L, KCl 0.2 g/L, Na 2 HPO 4 1.44 g/L, KH 2 PO 4 0.24 g/L, pH adjusted to 7.4). A widely-used DNA isolation protocol based on lysozyme treatment and alkaline lysis, suitable for the treatment of both Gram negative and Gram positive bacteria, was used to obtain metagenomic DNA. Briefly, each sample was treated with 2 mg/mL lysozyme (30 min, 37 °C) to ensure the lysis of Gram positive bacteria, and DNA was isolated with an alkaline solution (Tris 300 mM, SDS 1.25%, sucrose 5%, EDTA 10 mM; adjusted to pH 8.0) followed by precipitation with 3M potassium acetate (pH 5.0) and isopropanol. The quality of the DNA was checked on a 0.8% (w/v) agarose gel and quantified with Nanodrop-1000 Spectrophotometer (Thermo Scientific, Wilmington, DE). Universal primer sequences 28F (5′ -GAG TTT GAT CNT GGC TCA G-3′ ) and 519R (5′ -GTN TTA CNG CGG CKG CTG-3′ ), were chosen for the PCR-amplification of the 16S ribosomal RNA gene since they targeted the V1-V3 hypervariable region (commonly analysed in metagenomic studies) and produced an amplicon of a suitable length (500 bp) for the subsequent sequencing library construction. A short (9-11 nucleotides) barcode sequence followed by a four-nucleotide spacer (CGAT) was included at the 5′ end of the oligonucleotides used as forward primers to enable sequence assignment to samples after high-throughput sequencing. All the amplifications were performed under the following conditions: initial denaturing at 95 °C for 5 min, followed by 35 cycles of denaturing at 95 °C for 30 s, annealing at 54 °C for 30 s, and extension at 72 °C for 1 min; finalized by a 10-min elongation at 72 °C. The resulting amplicons were checked on a 0.8% (w/v) agarose gel and purified by precipitation with 3M potassium acetate (pH: 5) and isopropanol. Pure amplicons were quantified with the Qubit ® 2.0 Fluorometer (Invitrogen, Carlsbad, CA, USA) and an equimolar pool of amplicons was prepared from all the samples.
16S rRNA gene sequencing and bioinformatic data analysis. A sequencing library was constructed with 100 ng of the equimolar pool by the amplicon fusion method (Ion Plus Fragment Library Kit, MAN0006846, Life Technologies). The library was quantified with the Agilent 2100 Bioanalizer (Agilent Technologies Inc, Palo Alto, California) prior to clonal amplification, and emulsion PCRs were carried out applying the Ion PGM Template OT2 400 kit as described in the user-guide (MAN0007218, Revision 3.0 Lifetechnologies) provided by the manufacturer. The library was sequenced in an Ion 318 Chip v2 on a Personal Genome Machine (PGM IonTorrent, Lifetechnologies) at Lifesequencing S.L (Lifesequencing, Valencia, Spain), using the Ion PGM Sequencing 400 kit following the manufacturer's protocol (Publication Number MAN0007242, Revision 2.0, Lifetechnologies). Raw sequences obtained from the sequencing centre were processed with the MOTHUR software 28 . A summary of sequencing statistics is available in Supplementary Table 1. Short (< 100 bp) and low quality (< Q10, < 90% accuracy) reads were removed in a first step, and sequences were then assigned to samples based on barcode matches (allowing a maximum of 3 mismatches in primer search and 1 mismatch barcode search). The resulting sequences were trimmed by removing primer, barcode, and spacer sequences. All sequences were aligned to the ribosomal 16S Greengenes database using the kmer method (8-mers) for finding template sequences and the Needleman method for sequence alignment. Penalties for mismatch, gap opening, and gap extension were set as default (− 1, − 2, and − 1, respectively). Finally, sequences were classified using BLAST searches against the same database. The similarity percentage cut off was set at 70%.

Consortia identification and visualization.
In order to identify microbial interactions, we used taxonomic data of samples from the ten machines. A method used a recently described multidimensional scaling and a biological significance filtering of the interactions 29 . Basically, the procedure was as follows: fluctuation scaling of replicas was validated by a Poisson distribution selection. Then, linear correlation coefficients were converted into distances and displayed by multidimensional scaling. Finally, a discrete Lotka-Volterra model with relative abundances 30 was used to filter biologically significant interactions from the correlations identified in the last step.