This study used either Gluconacetobacter xylinus (ATCC 23767) (now named Komagataeibacter xylinus) as a single isolate strain or a Kombucha starter culture (KSC—www.happykombucha.co.uk) to propagate bacterial cellulose over three generations. K. xylinus was stored at −80°C in a cryopreservative solution [3.6% KH₂PO₄, 12.6% K₂HPO₄, 0.9% Na₃C₆H₅O₇, and 1.8% (NH₄)₂SO₄]. When required, it was grown on H&S agar overnight at 30°C and was then used to inoculate a liquid medium for experiments detailed as follows. KSC, as provided by suppliers, comprises a solid surface (hereafter termed “pellicle”) and liquid phase. Storage of KSC was at room temperature in the absence of light in a sealed container, following manufacturer’s instructions.

Sterile “universal” tubes (25 ml) containing 10 ml of sterile H&S broth were inoculated with either 1 ml of K. xylinus or 1 g of KSC pellicles. The tubes were covered with loose-fitting lids and left at 30°C to incubate for 10 days—producing generation zero (G0). Following the 10-day incubation, pellicles had formed at the liquid–air interface in both starter cultures. Each pellicle was removed aseptically and divided using a sterile scalpel. Half of each pellicle was frozen at −80°C for DNA extraction and analysis (see below). The other half of the pellicle was further divided into two and used as the inoculum for generation 1 (in duplicate) 10 ml sterile H&S broth (G1P; Figure 1). Further duplicate G1 cultures were established using 10 ml of H&S broth that had been inoculated with 1 ml of the G0 liquid phase (G1L; Figure 1). This resulted in eight new starter cultures, inoculated from either the pellicle or liquid phase for both the community started by G. xylinus or the KSC. These cultures were incubated at 30°C for 30 days. After 30 days, the pellicle and broth were removed and sampled as mentioned earlier. The process was repeated to create generation 2 (G2) and generation 3 (G3), where each of the previous generation bacterial communities resulted in two new inoculations (one from the pellicle and one from the liquid phase) (Figure 1).

At the end of each generation, before the pellicle was removed from the pot, the thickness was measured, using a ruler on the outside of the pot, and recorded. Pellicles were photographed both before and after removal from the pot, and visual appearance observations (color and opacity/transparency) were noted.

Samples of all pellicles were prepared for scanning electron microscopy (SEM) analysis by immersing each sample in 0.1 M glutaraldehyde overnight (to fix and preserve the cell structure), removing, and then dehydrating by passing through sequential ethanol baths (10 min per bath) of increasing concentrations (50%/70%/80%/90%/99.9%). The samples were kept in a desiccator until ready to be viewed on the SEM (Carl Zeiss Ltd., model Supra 40VP).

Nucleic acid extractions were carried out using the DNeasy PowerSoil kit (QIAGEN Ltd., Manchester), cleaned using the ZR-96 DNA Sequencing Clean-Up kit (Zymo Research, United States), and quantified using the Qubit dsDNA high sensitivity assay kit (Invitrogen, Paisley).

Amplicon sequencing of bacterial ribosomal rRNA genes (16S rRNA) was undertaken, as previously described (Thompson et al., 2017). Briefly, PCR reactions for initial amplifications consisted of 5 µl NEB Q5 reaction buffer, 0.25 µl NEB Q5 High-Fidelity DNA Polymerase, 0.5 µl NEB X nM dNTPs, 15 µl DNA template (c. X ngµl⁻¹), and 0.5 µl X pM primers 515F (5′-GTGYCAGCMGCCGCGGTAA-3′) and 806_R (5′-GGACTACNVGGGTWTCTAAT-3′) or EUK1391_F (5′-GTA​CAC​ACC​GCC​CGT​C-3′) and 1510_R (5′-CCTTCYGCAGGTTCACCTAC -3′) for 16S and 18S rRNA genes, respectively, and made up to a final reaction volume of 25 µl with nuclease-free water. Cycling conditions comprised an initial denaturation at 94°C for 3 min, followed by 35 cycles of denaturation at 94°C for 45 s, annealing at 50°C for 60 s, and extension at 72°C for 90 s, with a final extension at 72°C for 10 min. Amplification was confirmed visually by 1.5% (w/v) Tris-acetate-EDTA (TAE)-agarose gel electrophoresis. A second-stage PCR was carried out to attach barcodes for Illumina sequencing. The constituents of the second-stage PCR reaction are as follows: 10 µl NEB Q5 reaction buffer, 0.5 µl NEB Q5 High-Fidelity DNA Polymerase, 1 µl NEB (X nM) dNTPs, 0.5 µl (X pM) forward primer, 0.5 µl (X pM) reverse primer, and 20 µl cleaned amplicon template, and they were made up to a final reaction volume of 50 µl with nuclease-free water. Cycling parameters comprised an initial denaturation at 98°C for 30 s, followed by 10 cycles of denaturation at 98°C for 10 s, annealing at 62°C for 20 s, and extension at 72°C for 30 s, with a final extension at 72°C for 2 min. Amplification and attachment were confirmed by 1.5% (w/v) TAE-agarose gel electrophoresis.

Following second-stage PCR, all amplified DNA strands were normalized using the SequalPrep Normalization kit (Fischer Scientific, Loughborough) and pooled into libraries for sequencing on a MiSeq Illumina platform with a MiSeq Reagent Kit v2 (300-cycle) flow cell.

Amplicon sequence variants (ASVs) were extracted from the raw sequence data using the dada2 pipeline (Callahan et al., 2016) using the default parameters. Taxa were assigned using the SILVA database and sequencing information files deposited in the NCBI BioProject database (BioProject PRJNA787576; accession numbers SAMN23827663–SAMN23827813). Statistical analysis is described in the results section, with α set at 0.05. All assumptions for parametric statistics were assessed visually prior to analysis. Multivariate ordination plots were calculated using the Bray–Curtis dissimilarity measure and non-metric multidimensional scaling (NMDS). Analysis of similarity (ANOSIM) was used to compare compositional differences in communities between two groups, with permutational ANOVAs (PERMANOVAs) used to analyze differences between communities as a factor of multiple factors using Bray–Curtis dissimilarity measures in the vegan package (v.3.5-0) (Oksanen et al., 2016).

Comparisons of community shifts were achieved using multivariate ANOVAs (MANOVAs). Univariate analyses were performed using repeated measures ANOVA (linear mixed-effects modeling fit by REML) in the lme4 (Bates et al., 2015) package, with significance assigned from the lmerTest (Kuznetsova et al., 2017) package, where generation was both a random effect and a fixed effect along with the original seed culture and phase measured as fixed effects.

Across all generations, the Kombucha (KSC) pellicles had a consistent pale brown gel-like appearance, as noted from previous studies (Jarrell et al., 2000; Domskiene et al., 2019). However, previous work has not highlighted the visual variations our study found in the K. xylinus pellicles. In contrast to the consistent KSC pellicles, white, opaque spots were observed within the K. xylinus pellicle structure, as indicated in Figure 2.

In contrast to the KSC pellicles, K. xylinus pellicles, from both inoculum types, were initially (generation 1) translucent with a gelatinous appearance and were difficult to remove from the vessel due to their gel-like adherent consistency. In both second and third generations, however, the pellicles developed regions of opaqueness and were more robust.

Pellicle thickness increased as the generations progressed (Table 1), from 1 mm in all samples observed in generation 1 to between 2.4 mm K. xylinus pellicles (KXP) and 3.9 mm Kombucha pellicles (KP) in generation 3, despite all generations having the same incubation time. This trend was observed regardless of the starter culture. However, the inoculum type had an effect on the pellicle yield in generation 3; the highest yields by the thickness of pellicle were achieved when a liquid phase inoculum had been used at some stage of the process, with a lower G3 yield when only pellicle phase inocula had been used throughout the preceding generations (Table 1). Dima et al. (2017) and Blanco Parte et al. (2020) noted that the bacterial communities responsible for BC production are more mobile in the liquid phase and have greater access to nutrients, thereby improving their BC production potential and suggesting a reason for the observed differences in pellicle thickness. This leads to the hypothesis that liquid inocula achieve higher yields of pellicles, with Kombucha-derived liquid inocula producing higher yields than pure strain K. xylinus.

Additionally, previous studies of the fermentation of the Kombucha tea state that the yeast species present in the starter culture are instrumental in making carbon sources freely available (by breaking down complex sugars into more simple molecules) and in increasing the acidity of the broth as a by-product of the metabolic process (Teoh et al., 2004; Marsh et al., 2014). Both are favored conditions for BC-producing bacteria. Thus, the presence of yeast in the Kombucha created an advantage over the single isolate in the development of BC pellicles, particularly in broths such as H&S used in our study where there is no adjustment of pH. Furthermore, Jefferson, (2004) suggested there may be an ability to “store” carbon sources as extracellular polymeric substances (EPS) within the bacterial cellulose fibrillar matrix, thus creating a supply of nutrients which can be utilized when the source in the growth broth is depleted. This could explain the more rapid and consistent development of pellicles from Kombucha starter cultures. Here, a sequestered carbon supply is present when Kombucha is cultured compared to a single isolate where the nutrient source declines.

All pellicles were examined by scanning electron microscopy. Figure 3 shows the comparisons of generation 3 pellicles as examples, illustrating similarities in the microfibrillar structure. However, SEM revealed pellicles developed from K. xylinus tended to develop more consistent and smoother microfibrillar mats. The pellicles developed from the Kombucha culture, whilst exhibiting a microfibrillar structure, also displayed a degree of contamination, which were suggested to be extracellular polymeric substances and a documented part of the Kombucha microbial consortium (Jefferson, 2004). Additionally, all pellicles showed similar size and distribution of nanofibrils, ranging from a (all broth inoculant) mean diameter of 81 (+/− 7.8) nm in K. xylinus to (mixed inoculant) a mean diameter of 92.5 (+/− 6.6) nm in Kombucha. The measured nanofibril size is in accordance with the findings of previous studies (Castro et al., 2011; Kumbhar et al., 2015; Chen et al., 2017). Furthermore, there were no other observed differences between the pellicle structures of those formed from liquid, solid, or mixed inocula.

The analysis of the 16S rRNA gene sequence data indicated that Komagataeibacter was the dominant genus in all samples, agreeing with findings of the previous work (Jarrell et al., 2000; Marsh et al., 2014; Gaggìa et al., 2019). The results indicate that pellicles inoculated with Kombucha (98.31% range: 44.09–99.99%) demonstrated a consistent (Wilcoxon rank sum test: W = 2,986 and p = 0.614) relative abundance of Komagataeibacter compared to those from a K. xylinus starter (97.70% range: 49.12–100%) (Tables 2, 3).

Additionally, the proportional abundance of Komagataeibacter in the pellicles was analyzed using the Berger–Parker index to establish the numerical importance of this as the most abundant species. Figure 4 illustrates the relative abundance of Komagataeibacter across generations of pellicles inoculated with either K. xylinus or Kombucha. Whilst Komagataeibacter is abundant in early generations of pellicles inoculated with K. xylinus, the trend of abundance slightly decreases over subsequent generations. Conversely, in pellicles inoculated with Kombucha, the trend of abundance of K. xylinus increases in later generations.

When considering the non-dominant composition data, there were significant differences between communities due to the inoculation type (ANOSIM: R = 0.28 and p < 0.001), during generations 0–2, and significant differences between cultures seeded from Kombucha or K. xylinus (ANOSIM: R = 0.14 and p < 0.001) across all generations. Figure 5 provides a visual illustration of these differences. The points represent the mean ordinations of the communities grown using liquid (circular points) or pellicle (triangular points) inocula, which were found to be significantly different (p < 0.001). Communities derived from a KSC inoculum (filled points) were found to be significantly different (p < 0.001) from those derived from a K. xylinus (KX) starter inoculum.

Using the Bray–Curtis similarity index to further explore and quantify the difference in the communities, a more marked drift away from the original pellicle community is noted in pellicles developed from a Kombucha inoculant (Figure 6) than those inoculated with K. xylinus. However, the rate of pellicle community composition change when compared to the previous generation is similar in trend for both Kombucha and K. xylinus (Figure 7).

Previous work studying the microbial communities in Kombucha tea liquid observed microbial community stabilization over time. In studies sampling tea broth, it was suggested that stabilization occurs after approximately 21 days (Chakravorty et al., 2016). This implies that stabilization of the community is an important factor when studying the BC pellicle-forming effects of both Kombucha and single isolates. Previous studies investigating the properties of the liquid culture consortium have suggested that there are critical times in fermentation at which the microbial community is most stable (Marsh et al., 2014; Chakravorty et al., 2016). However, it is important to note that some previous studies have used non-sterile sweetened black tea liquid as it is suggested the autoclave process can cause a build-up of toxins which inhibit the fermentation of Kombucha pellicles (Marsh et al., 2014). Further work will be required to tailor the methodology to ensure reproducibility of all aspects of BC production, for example, thickness of BC, and any specific properties required for commercialization. Therefore, it could be argued that the microbial diversity found in these studies was not adequately controlled. Nevertheless, the results of our study lead to the hypothesis that a community stabilization period’ (in the region of 30 days) is required to observe consistent, reproducible BC pellicle production.