Soil samples were collected from the municipalities of Girón (Sotavento farm - 6° 58´ 20.4024” N 73° 8´ 13.47” W) and Páramo (Vista Hermosa farm – 6° 27´ 42.0552” N 73° 9´ 4.3308” W) in the department of Santander, Colombia. Permission for the sampling and processing of Colombian biodiversity was granted by the National Ministry of Environment through contract ARG-167-2017, addenda N°1 and N°2. Seventeen soil samples were collected in Girón from three different locations (plots 1-3), dedicated to the organic production of Citrus latifolia and C. sinensis. As reference, one additional sample from a non-cultivated plot in Girón (plot 0) was collected. Also, samples from the two organic fertilizers used in the farm, prepared from chicken manure (B20) and worm castings (B21), were retrieved for microbiological analysis. In Páramo, the non-organic production crops of C. latifolia and C. limon within the farm are organized as a single plot (plot 4). From there, six soil samples were retrieved. To collect the top layer of the mineral soil (A horizon), 3 cm of topsoil was removed, and samples were taken at 30 cm from the plant stem and from a depth of 20-30 cm. Each sample was processed by resuspending 2 g of soil or fertilizing material in 18 mL PBS for microbiological analysis, and 5 g were collected in sterile centrifuge tubes for molecular analyses. Subsequently, 500g of soil were placed in sterile vacuum-sealed bags for physicochemical analysis (for methods see Supplementary Material).

Enrichment cultures in liquid modified M9 (Reasoner and Geldreich, 1985; Puentes-Cala et al., 2023), and R2A (Chaudhary et al., 2019) were prepared from each sampled plot. According to the type of microorganism to be enriched the medium and carbon source varied. For cellulose-degrading bacteria the modified M9 medium was supplemented with 2 g/L carboxymethylcellulose (CMC), while for plastic degrading microorganisms 2 g/L 500 µm low-density polyethylene granules (LDPE) were added. When preparing solid M9 medium, agar-agar (20 g/L) was washed three times with distilled water to eliminate traces of contaminants. The M9 enrichments were incubated for 10 days at room temperature to allow for microbial growth. This was followed by a transfer to fresh media and allowed to grow for 10 additional days. Subsequently, 100µL of 10^-6 and 10^-8 dilutions were transferred to solid M9. Cellulolytic isolates were identified by the formation of a clear halo stained with a 0.1% w/v Congo red solution (Teather and Wood, 1982; Dantur et al., 2015). Potential low-density polyethylene (LDPE) degraders were identified by their ability to grow on M9 agar plates with LDPE as the only carbon source. Concurrently, R2A enrichments supplemented with heavy metals were employed to select for metal-tolerant bacteria. The modified R2A medium contained (g/L): yeast extract 0.5, peptone 0.5, sodium casein 0.5, glucose 0.5, soluble starch 0.5, sodium pyruvate 0.3, K₂HPO₄ 0.3, and MgSO₄ 0.024. pH was adjusted to 7 before autoclaving. To select for tolerant microbes, metal concentrations were gradually increased. First, samples were used to inoculate R2A broth supplemented with 1 mM Pb(NO₃)₂ or 5mM ZnSO₄. After 14 days at room temperature, enriched organisms were subjected to two cycles of inoculation in fresh R2A. In the first round, twice the initial concentrations of Pb²⁺ or Zn²⁺ (i.e., 2mM or 10mM, respectively) were used, followed by a second transfer into R2A with three times the initial metal concentration (3 mM Pb(NO₃)₂ or 15mM ZnSO₄.7H₂O). Finally, tolerant bacteria were isolated on solid R2A medium and preserved at -80°C in 20% glycerol.

Metagenomic DNA was extracted from the soil samples with the DNeasy PowerSoil kit (Qiagen, Hilden, Germany) according to the manufacturer’s protocol. DNA concentration and quality were determined using a NanoPhotometer^® NP80 (OD260/OD280). Subsequently, DNA samples were sent to Novogene Corporation Inc. (California, USA) for partially sequencing the 16S rRNA gene using primers 341F CCTAYGGGRBGCASCAG and 806R GGACTACNNGGGTATCTAAT on Illumina Novaseq PE250 platform (Illumina, USA). Genomic DNA of the isolates was extracted using the protocol described by (Martín-Platero et al., 2007). DNA quality was assessed by agarose gel electrophoresis and measuring the concentration using the Qubit™ Fluorometer (ThermoFisher-Scientific). Sequencing of the 16S rRNA gene was carried out with a MinION Mk1C (Oxford Nanopore Tech., UK), using the 16S Barcoding Kit 1-24 (SQK-16S024) and FLO-MIN106 (R9.4.1) flow cells. Whole genome of the cellulolytic strain CP102 was sequenced using Illumina NovaSeq 6000 (PE150) and Nanopore MinION platforms (Ligation Sequencing Kit (SQK-LSK109)).

We determined the physicochemical properties of citrus production soils from four plots in Girón and one plot in Páramo, located in the eastern Andes region of Colombia (Supplementary Table 1). Although geographically close, Girón and Páramo exhibit slightly different climatic conditions. On average, Girón is situated 777 meters above sea level, has a mean annual temperature of 24.5°C, and receives annual precipitation of 933 mm. In contrast, Páramo is on average at 1,564 meters above sea level, with mean annual temperatures of 21°C and annual precipitation of 1,600 mm. Since Colombia lies close to the equator, no strong seasonal variations take place in any of the municipalities.

Differences in soil textures were observed between the two sampling locations. While soil samples from Giron exhibited a predominantly sandy texture, those from Páramo were classified as sandy loam. This distinction underscores the higher moisture retention capacity of soils in Páramo. Humidity, organic carbon, potassium, sodium, and iron content were higher in Páramo than in Girón (p ≤ 0.005) (Supplementary Table 1). Interestingly, the uncultivated and unfertilized plot in Girón (plot 0) exhibited the lowest levels of electrical conductivity, humidity, and nitrogen content. In all cases, the pH values were in the slightly acidic range, with plots 1-3 the closest to neutrality. The humidity content in Girón ranged from 4.96-9.8%, while in Páramo had a remarkable 25.03% despite being collected 2 weeks apart, suggesting substantial differences in soil water content, which is crucial for microbial activity and nutrient uptake. Organic carbon, a key component for soil fertility and structure, also varied, with Plot 4 showing the highest percentage (2.36%) and plot 1 the lowest (0.437%). This variation in organic carbon may impact the microbial diversity and activity in these soils.

Phosphorus (P) and nitrogen (N) levels, essential for plant growth varied within plots with the highest nitrogen content observed in Plot 4 (0.182%) and the highest phosphorus content in Plot 3 (0.3%). These physicochemical characteristics provide a foundation for understanding the diverse environmental conditions faced by citrus crops in this region and their potential impact on the soil-associated bacteriome.

To characterize the microbiome in soil plots (A horizon) from two citrus plantations, 18 soil samples were collected from an organic citrus plantation and five 5 from a farm utilizing mineral fertilizers. These samples underwent metataxonomics analysis and bacterial isolation in minimal media. For the former, the amplicon sequencing and subsequent data processing generated 3,588,159 16S rRNA reads, which were classified into 8,650 amplicon sequence variants (ASV). The observed ASV counts displayed considerable variability, ranging from 1,103 to 1,925 across samples. The alpha diversity, which describes the richness (number of taxa) within a community, was contextualized using the Chao1 estimator, which suggested potential species richness values spanning from 1,220 to 2,480, indicating that the actual biodiversity might exceed the observed counts. Moreover, we used Shannon diversity index, which considers both species abundance and evenness, ranged from 4.64 to 6.73 across the samples. These values reflect the amount of species and their relative abundances within the soil communities among the different plots sampled in this study. Next, we assessed the richness differences between the Páramo and Girón localities, pairwise Wilcoxon rank-sum exact tests were applied to the observed ASVs, Chao1 richness estimates, and Shannon diversity indices. We did not detect significant differences in species richness between the two locations (p-values: 0.19 for ASVs, 0.22 for Chao1, and 0.49 for the Shannon index). To compare the variability in community composition (beta diversity) between Páramo and Girón, as well as the physicochemical soil properties, we used a constrained ordination analysis (Supplementary Figure S1). This analysis showed a clear separation between the plots from Páramo and Girón, driven by humidity, organic carbon, organic matter, potassium and iron, although none were significant. The location factor accounted for only a small proportion of the variance (r2 = 0.015).

At phylum level, the data revealed a significant presence of Actinobacteria and Proteobacteria (Figure 1), two key bacterial phyla in soil ecological processes (Huang et al., 2014; Deng et al., 2018). Actinobacteria are essential decomposers of organic matter, releasing nutrients crucial for plant growth and producing antibiotics and hormones (Kopecky et al., 2011; Lewin et al., 2016; Chen et al., 2023). Meanwhile, Proteobacteria contribute to nutrient cycling, soil fertility, and plant health (Wei et al., 2018; Mhete et al., 2020; Zhou et al., 2023). Other well-represented phyla included Acidobacteria and Bacteroidota, also relevant for their roles in decomposing complex organic substrates and contributing to carbon and nitrogen cycling, sulfur biogeochemistry and enzyme secretion (Banerjee et al., 2018; Kalam et al., 2020; Larsbrink and McKee, 2020; Kruczyńska et al., 2023). The soil microbiomes of Girón and Páramo portrayed a diverse array of bacterial orders, featuring the presence of Gaiellales, which among others like Burkholderiales, contribute to intricate processes such as promoting soil fertility, plant health, and ecosystem resilience through their involvement in organic matter decomposition and the nitrogen cycle (Tong et al., 2015; Zhang et al., 2019). The variation in Gaiellales abundance across samples suggests potential associations with specific soil characteristics or environmental conditions that favor their proliferation. These findings underscore not only the complexity of the microbial ecosystem, but also the specialized roles that distinct bacterial orders may play in soil health and functionality. At lower phylogenetic levels, it is noteworthy that the populations within the samples are predominantly dominated by taxa represented by less than 1% of the sequences. This observation emphasizes the prevalence of highly diverse and less abundant microbial taxa, further highlighting the nuanced composition and dynamics within the soil microbiomes under study.

The microbial communities in the fertilizer based on chicken manure were mostly dominated by Gram positive bacteria of the phyla Firmicutes (class Bacilli) and Actinobacteria (class Actinobacteria) (Figure 1). Even though fresh chicken manure is rich in diverse gut microbes, specially Proteobacteria, these usually do not reach the soil, since microbiomes of composted chicken manure are usually dominated by Gram-positive bacteria (Maeda et al., 2011; Zhang et al., 2018; Haq et al., 2021). Such dominance is explained by the ability of Gram-positives to withstand the relatively high temperatures reached during the composting process (Partanen et al., 2010; Aguilar-Paredes et al., 2023). In contrast, the worm castings were mostly populated by Gram negatives of the phyla Proteobacteria (Beta-, Delta- and Gammaproteobacteria) and Gemmatimonadetes (C_Gemm-1). Vermicomposting occurs at room temperature and under aerobic conditions, eliminating the temperature-based selection of microbes (Huang et al., 2014; Cai et al., 2018).

Culture-based methods were focused on isolating k-strategist bacteria capable of cellulase production, heavy metal resistance and LDPE degradation. Twenty-five isolates from 8 bacterial genera were identified through full 16S rRNA gene sequencing (Table 1). The strains showcased a range of morphological colony traits, and functional activities such as the degradation of natural (CMC) and artificial (LDPE) polymers, and tolerance to heavy metals. These isolates, with their promising traits, are prime candidates for further research and potential applications in bioremediation, bioprospecting, and sustainable agricultural practices (Kyaw et al., 2012; Ahmed et al., 2018; Alvarado-Campo et al., 2023; Nademo et al., 2023).

Isolate CP102 emerged as a particularly interesting candidate due to its pronounced capability for degrading carboxymethyl cellulose (CMC) on agar plates. Genomic sequencing techniques were employed for the production of a closed bacterial genome and the identification of an accompanying plasmid (Figure 2). Comparative genomic analysis showed that the closest relative of CP102 is Enterobacter soli ATCC BAA-2102 as determined by 16S rRNA gene comparison (identity: 99.54%), genome average nucleotide identity (95.73%, Figure 3) and dDDH (d6: 82.2%). These results suggest that isolate CP102 is most likely affiliated to the species Enterobacter soli. CP102’s genome, as detailed in Supplementary Table 2, consists of a single circular contig with a total length of 4,781,449 bp. The genome harbors 4,389 coding DNA sequences (CDS) and has a GC content of 53.9%. The chromosomal DNA includes 22 rRNA genes, 84 tRNA genes, and a single tmRNA, indicating a robust capacity for protein synthesis.

We conducted a search for genes associated with cellulolytic activity within the genome of CP102 and its close relatives, using BLASTp. En. Soli CP102 encoded four proteins that are potentially involved in the cellulolytic activity: Candidate protein KOMHPLAP_04250, is similar to a terephthalic acid (TPA) cellulase from Enterobacter soli (Acc. Number HDX4050896.1). The candidate TPA cellulase KOMHPLAP_042 may catalyze the hydrolysis of cellulose into its constituent glucose units, positing a significant role in the primary degradation of cellulose, which could be a cornerstone in the CP102 strain’s cellulolytic arsenal. Similarly, KOMHPLAP_04258, is a ‘TPA: endoglucanase’, linked to Enterobacter soli (HDX4050904.1). This enzyme appears to be akin to endoglucanases that act on the internal β-1,4-glucosidic bonds within cellulose. This catalytic activity is critical, as it disrupts the orderly structure of cellulose, rendering it more accessible for subsequent enzymatic action, and suggests a synergistic participation of KOMHPLAP_04258 in cellulose breakdown. The third protein, KOMHPLAP_01362, bears the annotation ‘TPA: glycoside hydrolase family 1 protein’ and is aligned with a protein from Enterobacter asburiae (HDR2889362.1). As a member of the glycoside hydrolase family, this protein is anticipated to possess a broad range of activity on glycosidic bonds, which could include cellulolytic functions, depending on the specific substrates the enzyme interacts with. This implies a potential versatility in the carbohydrate-active enzymes of CP102, warranting further biochemical validation. Lastly, KOMHPLAP_01787 is close homologous of a ‘6-phospho-beta-glucosidase’, derived from Enterobacter species (WP_014069886.1.). This enzyme has a higher specificity compared to the wider-range cellulolytic enzymes, which could provide a specialized niche, potentially linked to a unique phosphorolytic cleavage mechanism.

Strain CP102 harbors a plasmid, pCP102, with a length of 109,870 bp and 141 CDS (Figure 2B). This is the first report of a plasmid in En. soli isolates. Interestingly, in a BLAST search, 81% of the genes encoded by pCP102 aligned to plasmid p1 (109 kbp) from Cronobacter muytjensii JZ38 showing 89.84% nucleotide identity over the aligned sequence. Cr. muytjensii JZ38 is described as an endophytic bacteria that promotes plant growth in stressing environments (Eida et al., 2020). En. soli CP102 and Cr. muytjensii JZ38 showed 96.49% 16S rRNA sequence identity and a mere 77% ANI value. Although originally all members of the genus Cronobacter were collapsed into a single species within Enterobacter sakazakii, a reclassification generated the formulation of a new genus with seven described species thus far (Eida et al., 2020). The roles of p1 and pCP102 in the metabolism or survival of their hosts are still to be determined.