**2.5 Antimicrobial spectrum of the selected isolates**

The antimicrobial spectrum of the selected isolates was determined against Gram-positive and Gram-negative bacteria, as previously described by Garzon et al. [33]. The indicator bacteria were: *Salmonella enterica subsp. enterica* (Kauffmann and Edwards) Le Minor and Popoff ATCC 51741, *Salmonella enterica subsp. enterica* serovar Abaetetuba ATCC35640, *Escherichia coli* ATCC25922, *E. coli* ATCC10536, *Shigella sonnei* ATCC25931, *Streptococcus thermophilus* ATCC19258, *Enterobacter aerogenes* UTNEag1 (laboratory collection), S*almonella* UTNSm2 (laboratory collection), *Shigella* UTNShg1 (laboratory collection), and *E. coli* UTNEc1 (laboratory collection). LAB was grown in MRS broth at 34°C for 24–27 hours and the supernatants were collected by centrifugation at 13,000 x g for 20 minutes at 4°C. The crude extract (CE) was recovered and filtered with a 0.22 μm porosity syringe filter. The indicator strains (100 μl) were grown in broth medium (7 log CFU/ml) and mixed with 3.5 ml of soft MRS agar (0.75%). It was then overlaid on nutrient agar plates and incubated at 37°C for 2 hours. The CE of each strain (100 μl) was transferred onto the reaction wells (6 mm) on overlaid agar, incubated at 37°C and subsequently examined for the presence of an inhibition zone at 48 hours. To rule out the possible inhibitory activity of organic acids, the CE was heated at 80°C for 10 minutes, the pH adjusted at 6.0 and the activity was determined. *Lactobacillus plantarum* ATCC8014 and *L. fermentum* CNCM 1-2998 (Lacc) were used as reference strains. MRS broth was used as a negative control. The experiments were run in triplicate and the mean value of the inhibition zone was determined. A numeric scale from zero to ten was included in the statistical analysis and the results were also qualitatively defined as narrow (inhibit less than 5 indicator strains) or broad activity (inhibit more than 5 indicator strains).

### **2.6 Statistical and phylogenetic analyses**

The interpretation of the antibiogram results was assisted by the package AMR [34], which provided corresponding frequencies on the qualitative responses. The distances (Bray-Curtis) among samples were then projected in canonical space through a non-metric multidimensional scaling. Either putative genera of bacteria, assigned through the RDP Bayesian classifier algorithm [28], or the host plants were included as the grouping variable.

The metabolic profiling resulted in a matrix that could be interpreted in binary form, and from which it was possible to determine a set of distances (binary Bray-Curtis) for classifying samples through a cluster analysis (unweighted pair group method with arithmetic averages, UPGMA). Ordination methods were carried out through vegan [35] and ggdendro [36] in R [30].

**21**

*Microbiota of Wild Fruits from the Amazon Region of Ecuador: Linking Diversity…*

**3.1 Wild fruits: A microenvironment of diverse lactic acid bacteria**

(31 isolates), followed by *Lactococcus* (3 isolates), *Weissella* (3 isolates), and

Out of 41 isolates, the most frequently observed genus was *Lactiplantibacillus*

*Enterococcus* (1 isolate). Three isolates showed large divergence and were not identified by the taxonomic assignment algorithm (i.e. UTN39, UTN41, and UTN88) (**Figure 2a**). The former three isolates may represent unreported lineages or species. The presence of *Lactococcus* in plants or fruit is rare: thus, we only found few isolates

Isolates showed a remarkable distance to the outgroup reference samples, and most were included within a clade formed by *Lactiplantibacillus,* where they showed relatively small distances (i.e. small branch lengths). However, within this inclusive group of *Lactiplantibacillus* there was an ingroup with strong support (Bayesian posterior probabilities = 1) and formed by *Weissella, Enterococcus,* and *Lactococcus*. This paraphyletic ingroup, within a more inclusive group formed mostly by *Lactiplantibacillus* should be eventually resolved by the inclusion of additional molecular markers (**Figure 2a**). The paraphyletic ingroup contains samples that belong to the plant species *Costus* sp., which occurs exclusively to this

**3.2 The metabolic profile reveals the divergent properties of selected LAB**

metabolic profile, hints to four main groups of isolates (**Figure 4**).

LAB strains may present specific metabolic traits as a result of their microenvironmental origin (i.e. different species of fruits) and possess a unique portfolio of enzymes that allow them to metabolize various compounds found in the host plant or fruit matrices. We present a metabolic profile together with other properties that were analyzed in the obtained isolates (**Figure 3**). Within *Lactiplantibacillus*, the isolates showed a variable capacity to ferment sugars and hydrolyze esculin. These features were strain-dependent. Among the available isolates, UTN39 and UTN76 were the only two samples that metabolized ARG, while UTN37 and UTN39 hydrolyzed urea. The latter is a relevant characteristic for selecting probiotic strains [2]. Of particular relevance were the locations of UTN76 and UTN39 on the previously presented phylogenetic hypotheses (**Figure 2**); these two isolates showed large genetic distances relative to the other samples, or unique positions in clades that diagnose them as different or remarkable lineages. The observed patterns in the metabolic profile are rather complex. UTN37, UTN39, and UTN76 show noteworthy properties; they, however, differ in their response to other substrates (**Figure 3**); thus, the corresponding dendrogram, which results from clustering the observed

Sequences were aligned trough Clustal Omega, as implemented in Geneious Prime 2020.2.3 [27]. A proper substitution model was obtained through jModeltest v. 2.1.10 [37] and selected by a Bayesian information criterion. A phylogenic hypothesis was inferred by Bayesian inference with Mr. Bayes 3.2.6 [38]. The selected HKY85 model included a proportion of invariable sites and varying rates across sites with a discretized gamma distribution (HKY85 + I + G). The Bayesian analysis included 1.1 million generations and four chains per run. Hypotheses were sampled every 200 generations and the first 10% of these samples were discarded. The remaining 90% of the trees and parameters were respectively summarized in a

*DOI: http://dx.doi.org/10.5772/intechopen.94179*

50% majority-rule consensus tree.

**3. Results and discussion**

belonging to such genus.

clade (**Figure 2b**).

*Microbiota of Wild Fruits from the Amazon Region of Ecuador: Linking Diversity… DOI: http://dx.doi.org/10.5772/intechopen.94179*

Sequences were aligned trough Clustal Omega, as implemented in Geneious Prime 2020.2.3 [27]. A proper substitution model was obtained through jModeltest v. 2.1.10 [37] and selected by a Bayesian information criterion. A phylogenic hypothesis was inferred by Bayesian inference with Mr. Bayes 3.2.6 [38]. The selected HKY85 model included a proportion of invariable sites and varying rates across sites with a discretized gamma distribution (HKY85 + I + G). The Bayesian analysis included 1.1 million generations and four chains per run. Hypotheses were sampled every 200 generations and the first 10% of these samples were discarded. The remaining 90% of the trees and parameters were respectively summarized in a 50% majority-rule consensus tree.
