To culture or not to culture: a snapshot of culture-dependent and culture-independent bacterial diversity from peanut rhizosphere

Sample collection

A total of five samples of groundnut/peanut (Arachis hypogea) rhizosphere were collected from two nearby farms each, near Porbandar, Gujarat, INDIA (21.618389 N, 69.866106 E). The samples were collected during the nodulating phase of the crop and named as R3 (Rhizosphere3; from Farm1) and R4 (Rhizosphere4; from Farm2). The samples used in this study are part of another unpublished study on peanut rhizosphere and sample names are used as such. The soil surrounding the plants was carefully removed before uprooting the plants. Plants were then shaken to remove any loosely attached soil. Soil attached on the roots (Rhizosphere) was then collected by washing the roots in 40 ml sterile Normal Saline buffer in sterile 50 ml tubes. Five different plants were uprooted, and samples were collected from them during both sampling. Samples for culture-dependent (CD) approach were transported to lab in cool conditions (at 4–10 °C) and proceeded further next day. Samples for culture-independent (CI) approach were transferred to lab at −20 °C and stored at −80 °C till further processing.

Sample processing

For CD approach, all five samples from each farm were pooled and diluted serially upto 10⁻⁶ using sterile water. The last two serial dilutions, 10⁻⁵ and 10⁻⁶, were used for spreading and plated on eight different media supplemented with cycloheximide (60 µg/ml of media) as listed in Table 1. The media were selected based on their selective enrichment of common soil organisms as well as their routine use in isolating soil-borne organisms. The plates were incubated at 27 ± 2 °C and 37 ± 2 °C. The spreading of each dilution was done in triplicates for given media and temperature, respectively. After incubation of 14 days, entire plate surface (this should cover all the colonies same or different) was scrapped and collected in Qiagen RNAprotect Bacteria Reagent (Qiagen, Germany). Colonies were pooled in a tube for each media and each farm, i.e., colonies from 12 plates (2 temperatures $\times$ 2 dilutions $\times$ 3 replicates = 12 plates) pooled in a single tube. Total 16 tubes were obtained (2 farms $\times$ 8 media). All the colonies were mixed properly, and DNA was extracted from the pool of colonies using QIAGEN QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) following manufacturer’s protocol of bacterial genomic DNA extraction. Extracted DNA was checked on agarose gel for quality and quantified using Qubit 3.0 (Invitrogen, Waltham, CA, USA).

For CI approach, all samples were processed separately. Samples were thawed and centrifuged briefly to collect soil at bottom. The buffer was decanted, and remaining soil was mixed to homogenize. DNA was extracted from 0.2 gm homogenized soil using QIAGEN PowerSoil DNA Extraction Kit (Qiagen, Hilden, Germany) following manufacturer’s protocol. Extracted DNA was checked on agarose gel for its quality and quantified using Qubit 3.0 (Invitrogen, Waltham, CA, USA).

Library preparation and sequencing

DNA from CI approach samples (n = 10) and CD approach samples (n = 16) were diluted to 5 ng/µl. PCR was performed with diluted DNA using Illumina adapter fused primers 341F and 785R targeting V3–V4 region of 16S rRNA gene as illustrated in Illumina 16S Library Preparation Guide (Illumina, San Diageo, CA, USA) (Klindworth et al., 2013). Further steps were followed as given in the protocol for Illumina 16S Library preparation guide. Prepared libraries were checked on Agilent 2100 Bioanalyser (Agilent, Santa Clara, CA, USA) for their size and quantified on Qubit 3.0 (Invitrogen,Waltham, CA, USA). The CD and CI samples’ libraries were sequenced in separate runs on Illumina MiSeq system using 250 $\times$ 2 v2 paired-end chemistry.

Data analysis

Generated data was imported in R v3.6.2 environment wherein all the analysis was carried out. Sequences were analyzed using DADA2 package v1.14.0 following analysis pipeline as given in tutorial (https://benjjneb.github.io/dada2/tutorial.html) (Callahan et al., 2016). Since samples of CD and CI were sequenced in separate runs, runs were separately processed till sequence table construction and then merged for further steps as instructed in “big data” tutorial (https://benjjneb.github.io/dada2/bigdata.html). Further, taxonomy was assigned using SILVA v132. The ASV table, assigned taxonomy and sample metadata information were combined as a phyloseq object (phyloseq package version 1.30.0) (McMurdie & Holmes, 2013). This object was then further used for alpha diversity indices estimation, between-sample diversity estimation by Bray–Curtis distance based PCoA, taxonomy agglomeration as well as for presence-absence based comparison. Other R packages used include microbiome v1.8.0 (Lahti & Shetty, 2017), vegan v2.5.6 (Oksanen et al., 2019) and ggpubr v0.2.5 (Kassambara, 2020) with all the plots were made using ggplot2 v3.2.1 (Wickham, 2016).

The complete script/steps for analysis along with the data can be obtained from https://doi.org/10.5281/zenodo.4896308 for reproduction of results. The raw data is also submitted in SRA under the BioProject PRJNA665712. The accession numbers for CD samples are SRR12728805–SRR12728820 and for CI samples are SRR12732125–SRR12732134.

Alpha diversity

As expected, the CI approach showed a greater number of ASVs (Wilcoxon test, BH (Benjamini–Hochberg) adjusted p-value = 0.00003) as well as higher Shannon Index (Wilcoxon test, BH adjusted p-value = 0.0000004) than CD samples (Fig. 1). Further, no significant differences (Wilcoxon test, BH adjusted p-value < 0.05) was observed between two farms (R3 and R4; see methods section) for both CD and CI approach. A rarefaction plot also showed asymptotes for all the samples (Fig. S1).

Taxonomic profile

All ASVs were classified as Bacteria. However, 83 ASVs remained unassigned at phylum level. From 29 detected phyla, ASVs from only six phyla namely, Proteobacteria (377), Firmicutes (72), Actinobacteria (62), Bacteroidetes (43), Deinococcus-Thermus (1) and Verrucomicrobia (1) were detected in CD samples (Fig. 2A). Further, the greatest number of ASVs belonged to Proteobacteria phyla (1,487) followed by Planctomycetes (810) and Acidobacteria (692). Proteobacteria was the most dominating phyla in both approaches with 31.5% of average proportion in CI approach while, 46% to 98% in CD approach (Fig. 2B). Majority of the proportion in CD approach samples was of Proteobacteria phylum alone. Other phyla like Acidobacteria, Actinobacteria, Bacteroidetes, Chloroflexi, Firmicutes, Gammatimonadetes, Planctomycetes and Verrucomicrobia were detected with abundance more than 1% in CI approach.

Presence-absence of ASVs and Genera

A total of 3,112, 3,902, 262 and 390 ASVs were detected in R3, R4, Cultured-R3 and Cultured-R4 samples, respectively (Fig. 3A, Fig. S2). Out of those ASVs, 1,129, 1,891, 97 and 205 ASVs were exclusively observed in R3, R4, Cultured-R3 and Cultured-R4 samples, respectively. Overall, 322 ASVs (5.98%) were detected exclusively in CD approach compared to 4825 ASVs (89.67%) detected exclusively in CI approach, while only 234 ASVs (4.35%) were shared among both approaches. PERMANOVA test on the Bray–Curtis distance of presence-absence matrix showed that CD and CI group of samples differ significantly (R² = 0.37505, P = 0.001) while lesser significant differences were observed between collections (R² = 0.08435, P = 0.049). The same result when plotted through PCoA also highlighted these differences (Fig. S3). Taxonomically, CD-exclusive ASVs belonged to Proteobacteria (226), Firmicutes (42), Actinobacteria (38), Bacteroidetes (15) and Verrucomicrobia (1) (Fig. S4). At genus level, Bacillus (9%) was the most dominant followed by Pseudomonas (3%), Enterobacter (2%), Nocardioides (2%), Streptomyces (2%), Paenibacillus (2%), Allorhizobium-Neorhizobium-Pararhizobium-Rhizobium (2%). Further, many of these ASVs were also assigned at the species level as well, for example, Bacillus endophyticus, Bacillus subtilis, Bacillus infantis, Pseudomonas indica, Pseudomonas stutzeri, Pseudomonas taiwanensis and many others. All these genera were also observed in notable amount in CI approach. However, many of the species were not observed. Therefore, to provide more leniency in thresholds, we also compared samples at genus level to find if any genera are exclusively detected in CD samples only.

Around 433, 460, 94 and 82 genera were detected in R3, R4, Cultured-R3 and Cultured-R4 samples, respectively from 527 unique genera (Fig. 3B). Forty-nine genera were common among all four groups while, 22 and 408 genera were exclusively present in CD and CI groups, respectively. These 22 genera were Aeromonas, Agrococcus, Amycolatopsis, Arthrobacter, Cellulomonas, Comamonas, Cronobacter, Curtobacterium, Georgenia, Kocuria, Luteibacter, Methylobacterium, Methylophilus, Micrococcus, Nocardiopsis, Prauserella, Prevotella_1, Promicromonospora, Rheinheimera and unknown genera from families Intrasporangiaceae, p-251-o5 and Planococcaceae. A total of Nine of these genera were detected on more than one culture media. Also, majority of these genera belonged to Actinobacteria phylum (12), followed by Proteobacteria (7: 6 Gammaproteobacteria and 1 Alphaproteobacteria), Bacteroidetes (2) and Firmicutes (1).