Comparison of the diversity of cultured and total bacterial communities in marine sediment using culture-dependent and sequencing methods

Marine sediment collection

Three marine sediment samples (C1, I1, and X4) were collected from the South China Sea in June 2014 (Table S1). These samples were packed into sterile centrifuge tubes, transported to the laboratory and stored at 4 °C for later use.

DNA extraction and 16S rRNA gene fragment sequencing

DNA was extracted with an E.Z.N.A.^® Soil DNA Kit according to the supplier’s instructions (Omega Bio-Tek, USA). The DNA concentration was measured with a NanoDrop-200 microspectrophotometer (Allsheng, China). The 16S rRNA gene was amplified using the barcoded primers 515F (5′-GWA TTA CCG CGG CKG CTG-3′) and 806R (5′-CCG TCA ATT CMT TTR AGT TT-3′) (Bergmann et al., 2011). A negative control (water instead of DNA) was included in the PCR, and no bands were observed in the negative control. PCR products were sent to BGI Tech Co. Ltd. (Wuhan, China) for further purification and HTS using the HiSeq PE250 system.

Isolation and characterization of culturable bacterial strains

Culturable bacteria were isolated by six different media: Emerson agar (EM), Zobell 2216E (2216E), R2A agar (R2A), modified yeast mannitol agar medium (MCCM), modified chemically defined medium (CDM), and mineral basal medium (MBM) (Table 1). A 5.0-g sediment sample was dissolved in 20 ml of sterilized artificial seawater and placed in a conical flask in a shaker (30 min, 180 rpm). Then, the mixture was incubated until the two phases separated. The supernatants were serially diluted with sterilized artificial seawater, and 100 µl diluents were spread onto plates. The plates were incubated at 16 ° C until stable colonies formed. Each gradient was performed in triplicate. Negative controls included sterilized artificial seawater, and no colonies were noted in the negative controls during the culture process. To count the number of strains, plates with 25–250 colonies were selected (Tomasiewicz et al., 1980). Colonies with different morphologies were subcultured into pure cultures by inoculating them into freshly prepared agar plates (Selvin et al., 2019). The number of similar colonies was recorded for subsequent identification and statistics. The total DNA of bacteria was extracted using a Biospin Bacteria Genomic DNA Extraction Kit (Bioer, China). The 16S rRNA gene was amplified using the primers 27F (5′-AGA GTT TGA TCC TGG CTC AG-3′) and 1492R (5′-CGG TTA CCT TGT TAC GAC TT-3′) (Weisburg et al., 1991). PCR products were detected on a 1.5% agarose gel and sequenced using the Sanger sequencing method (BGI Inc., Guangzhou, China).

Physicochemical analysis of media

The total organic carbon (TOC), total phosphorus (TP), organic nitrogen (ON) and inorganic nitrogen (IN) contents of the media were measured. All the liquid media were sterilized before physicochemical analysis. A TOC analyzer (TOC-5000A, Shimadzu, Japan) was used to determine TOC, and the media were diluted 1000 times with ultrapure water before testing. Total nitrogen (TN) was measured using an alkaline potassium persulfate oxidation-ultraviolet spectrometer (UV-2450, Shimadzu, Japan). Ammoniacal nitrogen (NH₄⁺-N) was measured by spectrophotometry with salicylic acid. Nitrate nitrogen (NO₃⁻-N) was measured using UV spectrophotometry without digestion (Bao, 2011). Nitrite nitrogen (NO₂⁻-N) was measured using a spectrophotometer by a coupling reaction with naphthylenediamine hydrochloride. The concentration of IN was determined by the following equation: c(IN)=c(NH₄⁺-N)+c(NO₃⁻-N)+c(NO₂⁻-N). The ON concentration was determined by c(ON)=c(TN)-c(IN) (Li et al., 2006; Ren et al., 2006). TP was determined by the molybdoantimonyl phosphoric acid photometric method (UV-2450, Shimadzu, Japan) using persulfate as the reducing agent (121 °C, 30 min) (Guo, 2009). The media listed above were diluted 20 times before testing.

Bioinformatics and statistical analysis

For HTS data, sequencing reads were assigned to each subsample according to unique barcodes. Then, paired-end reads from the original DNA fragments were merged using Fast Length Adjustment of SHort reads (FLASH, v1.2.11) after removing the barcode and primer sequences (Magoč & Salzberg, 2011). QIIME and USEARCH (v7.0.1090) were used to filter the reads and to select the operational taxonomic units (OTUs) with 97% similarity, respectively (Edgar, 2013). Finally, the resulting OTUs_(97%) were aligned against the Silva (Silva_119_release_aligned) database. To evaluate microbial richness and diversity, the abundance-based coverage estimator (ACE), Chao1, Shannon and Simpson indices were calculated using the mothur program. A Venn diagram was constructed in R (v3.1.1) with the ”VennDiagram” package. To characterize similarities in microbial communities among different sediment samples, hierarchical cluster analysis at the genus level was performed by Past3 software (version 3.22) (Aguilar et al., 2016). To compare the recoveries of bacteria between culture-dependent and HTS methods, the sequences obtained from cultured strains and the HTS dataset were combined and clustered with 97% identity by the UPARSE algorithm (Lee et al., 2016). The 16S rRNA gene sequences were deposited in the NCBI Sequence Read Archive under accession number PRJNA577514.

For the 16S rRNA gene sequences, bacterial taxonomy was assigned at the OTUs_(98%) level using a 98% confidence threshold (Stackebrandt & Ebers, 2006). The EzBioCloud platform (https://www.ezbiocloud.net/identify) was employed to identify the most closely related taxon according to the 16S rRNA gene sequence (Kim et al., 2012). The linear discriminant analysis (LDA) effect size (LEfSe) approach was used to determine bacterial community differences and search for biomarkers in the media. The LEfSe approach was implemented by online tools with default settings (http://huttenhower.sph.harvard.edu/galaxy/) (Segata et al., 2011). All tests for significance were two sided, and p-values <0.05 were considered statistically significant. To determine the associations between culturable bacterial communities and medium components, redundancy analysis (RDA) was performed using CANOCO software (version 5.0). The relationships between the microbes and individual physicochemical factors were determined by correlation tests using Pearson correlation coefficient analysis software (version 3.6.1). To characterize similarities in the culturable bacterial communities between sediment samples, hierarchical cluster analysis was performed by Past3 software (version 3.22) (Aguilar et al., 2016). The 16S rRNA gene sequences of representative taxa for each sediment sample were deposited in the NCBI GenBank database with accession numbers MK841046 –MK841311.

Sediment microbial communities detected by the HTS method

A total of 110,628 high-quality partial 16S rRNA sequences were obtained from the three sediment samples by the HTS method. Gammaproteobacteria (37%), Bacilli (18%), Flavobacteriia (17%), Epsilonproteobacteria (11%) and Clostridia (8%) were the dominant classes in these sediment samples (Fig. 1A). Despite their low relative abundances levels(<1%), Actinobacteria, Alphaproteobacteria, and Betaproteobacteria were present in every sample. Planomicrobium (17%), Psychrobacter (15%) and Arcobacter (11%) were the dominant genera in the sediment samples (Fig. 1B). Cluster analysis results showed that the microbial composition in I1 was more similar to that in X4 than to that in C1 (Fig. 1D). In sediment samples, the range of the ACE was 296-314, while the range of the Chao1 index was 300-314 (Table S2), which are richness indices. C1 produced the largest number of OTUs_(97%); correspondingly, the ACE and Chao1 index levels were also highest, followed by those in I1 and X4. C1 contained the most OTUs_(97%), followed by I1 and X4, as well as the most unique OTUs_(97%) (Fig. 1C).

The diversity of culturable bacteria detected with different media

A total of 594 pure cultures were obtained in this study, including 201 pure cultures from C1, 196 pure cultures from I1, and 188 pure cultures from X4. The culturable bacterial strains were classified into 53 OTUs_(98%), 35 genera and 5 classes. At the class level, Actinobacteria (10%–84%), Gammaproteobacteria (11%–50%) and Bacilli (0%–60%) were the main groups determined by the culture method (Fig. 2B). Cytophagia (0%–1%) was a rare class, which was isolated only with R2A and MCCM (Fig. 2B). At the genus level, Glutamicibacter, Nocardioides, Planococcus, Planomicrobium, Paeniglutamicibacter, and Halomonas were the main genera. We obtained the most genera from C1 with MCCM. However, we obtained the most genera from I1 and X4 with R2A. Fewer culturable genera were found with CDM and MBM, especially MBM, with which only 16 genera were isolated. Interestingly, we obtained many Pseudomonas strains with CDM and MBM, while this genus was not found or was found at low abundance with other media (Fig. 2C). At the OTUs_(98%) level, we obtained the most OTUs_(98%) with MCCM (40), followed by R2A (37) and 2216E (37). MBM (20) produced the fewest culturable OTUs_(98%), followed by CDM (27) and EM (31) (Fig. 2A). No medium resulted in the detection of all 53 culturable OTUs_(98%). The total number of culturable OTUs_(98%) obtained with these six media was significantly larger than the numbers obtained with single medium (p <  0.05) (Fig. S1). In addition, four potentially novel bacterial strains belonging to Oceaniovalibus, Psychrobacter and Salegentibacter were isolated (16S rDNA similarity <  98%) (Kim et al., 2014).

Among these six media, the shared OTUs_(98%) belonged to 8 genera, namely, Glutamicibacter, Planococcus, Psychrobacter, Pseudoalteromonas, Halomonas, Dietzia, Nesterenkonia, and Arthrobacter (Fig. 3A). MCCM produced the most unique OTUs_(98%), such as members of Pantoea, Janibacter, Primorskyibacter, Devosia and Marinomonas. Interestingly, no unique OTUs_(98%) were observed in other media, indicating that MCCM is a special medium for unique bacterial isolation. In addition, the relative abundance levels of Flavobacteriales, Flavobacteria, Flavobacteriaceae and Bacteroidetes in R2A and Pseudomonadaceae and Pseudomonas in MBM were substantially higher than those in other media (Fig. 3B).

Comparisons of the in situ and cultured microbial diversity

The culture-dependent and HTS approaches showed different taxonomic distributions at the phylum level, and the number of phyla obtained by the culture-dependent approach was significantly smaller than that obtained by the HTS method (p <  0.01) (Fig. S2). Of the 33 phyla in the HTS dataset, only 4 were recovered in the culture collection (Figs. 4A, 4B). A large number of OTUs_(97%) in the HTS dataset were assigned to Proteobacteria (175), Firmicutes (64) and Bacteroidetes (56), but the corresponding numbers in the culture collection were much lower (26, 4, and 5, respectively). Among a total of 478 OTUs_(97%) from the HTS dataset, 27 were represented by culturable strains, which belonged to 24 different genera, such as Psychrobacter, Pseudomonas and Arthrobacter, accounting for 45% of the total reads in the HTS dataset (Figs. 4C, 4D). Surprisingly, we obtained culturable strains belonging to Aeromicrobium, Janibacter, Maribacter, Nesterenkonia, and 3 other genera that were not detected by HTS. In the HTS dataset, 58 OTUs_(97%) (12% of the total OTUs_(97%)) had a relative abundance greater than 0.1%, accounting for 95% of the total reads. Among these OTUs(97%), 14 (24%) were shared with the culturable strains (Table S3). These OTUs_(97%) were affiliated with the genera Planomicrobium, Psychrobacter, Pseudoalteromonas, Gramella, Pseudomonas, Salegentibacter, Flavobacterium, Arthrobacter, Cobetia and Halomonas, which accounted for 45% of the total reads in the HTS dataset (Table S3).

Physiochemical parameters of the media

The inorganic carbon (IC) concentrations of these six media were between 15.52 mmol/L and 22.33 mmol/L, and no obvious differences were observed. In contrast, a difference in the organic carbon content was identified (Table 2). EM possessed the highest concentration of organic carbon, and the concentration of organic carbon in MBM was the lowest. EM and 2216E were rich in ON, while the amounts of IN in CDM and MBM were greater than the amounts in other media. The nitrogen content in MCCM was surprisingly low. On the basis of these results, EM is a nutrient-rich medium with abundant carbon and nitrogen sources. MCCM is a high-carbon and low-nitrogen medium. CDM and MBM contained a single carbon/nitrogen source.

Correlations between the culturable bacterial communities and physicochemical factors

The total amount of variation in culturable microbial community composition explained by the first two axes (RDA1 and RDA2) reached 87%. RDA1 was mainly associated with TOC, ON and IN. The IC and TP contents had the largest associations with RDA2 (Fig. 3C). Salinibacterium, Planococcus, Salegentibacter, Psychrobacter, and Planomicrobium were projected in the negative direction of RDA1; thus, these microbial genera were positively correlated with TOC and ON but negatively correlated with IN and TP. RDA also indicated that only IN contributed significantly to the medium-physicochemical factor relationship (p <  0.05), providing 50% of the RDA explanatory power. Moreover, the two media with the greatest IN influence were MBM and CDM, and the medium with the smallest effect was MCCM (Fig. 3D). Although not significant (p = 0.052), ON provided 23% of the explanatory power, and 2216E, EM, and CDM were the three most affected media. In addition, the relative abundance levels of Paeniglutamicibacter, Pseudomonas, and Halomonas were positively related to IN (Fig. S3). The relative abundance of Planomicrobium was significantly positively correlated with TOC. Furthermore, the relative abundance of Arthrobacter increased significantly with increasing TP (Fig. S3).

Effects of different culture methods on bacterial composition among sediment samples

Compared with the culturable bacterial community in C1, that in I1 was more similar to that in X4 (Fig. 5). This result was consistent with that obtained with the HTS method (Fig. 1D). Furthermore, we obtained the largest numbers of culturable OTUs_(98%) and unique culturable OTUs_(98%) from C1, followed by I1 and X4 (Fig. S4). Similar results were obtained with the HTS method.