Comparative analysis of three experimental methods for revealing human fecal microbial diversity

Abstract

Due to the heterogeneity of the human gut environment, the gut microbiota is complex and diverse, and has been insufficiently explored. In this study, one fresh fecal sample was cultured using 12 commercial or modified media and incubation of culture plates anaerobically and aerobically, the conventional experienced colony picking (ECP) was first used to isolate the colonies and obtain pure culture strains. On this basis, all the colonies grown on the culture plates were collected for culture-enriched metagenomic sequencing (CEMS), and the original sample was also subjected to direct culture-independent metagenomic sequencing (CIMS), the study compared the effects of three methods for analyzing the microbiota contained in the sample. It was found that compared with CEMS, conventional ECP failed to detect a large proportion of strains grown in culture media, resulting in missed detection of culturable microorganisms in the gut. Microbes identified by CEMS and CIMS showed a low degree of overlap (18% of species), whereas species identified by CEMS and CIMS alone accounted for 36.5% and 45.5%, respectively. It suggests that both culture-dependent and culture-independent approaches are essential in revealing gut microbial diversity. Moreover, based on the CEMS results, growth rate index (GRiD) values for various strains on different media were calculated to predict the optimal medium for bacterial growth; this method can be used to design new media for intestinal microbial isolation, promote the recovery of specific microbiota, and obtain new insights into the human microbiome diversity. This is among the first studies on CEMS of the human gut microbiota.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12866-025-03985-7.

Background

With the widespread application of next-generation sequencing (NGS) technology, metagenomic sequencing has improved our understanding of the complex human gut microbiome. However, this method has the following drawbacks. First, sequencing technologies yield rich microbiome datasets; but current databases have limitations, many sequences remain unassigned, corresponding to the microbial dark matter [1–3]. Second, data obtained from molecular approaches alone remain limited, and microorganism culture and isolation is required to define the roles of specific bacteria in health and disease states. Third, most molecular sequencing methods are based on DNA extracted from microbial cells and cannot determine whether the resulting microorganisms are alive or dead [4].

Microbial culture makes up for the above problems existing in metagenomic sequencing to some extent [5]. In 1876, Koch established pure culture technology, which is based on the selection of characteristic colonies using the naked eye to obtain isolated strains [6]. Although pure cultures of microbial strains can be obtained, this method also has limitations such as a heavy workload for colony selection, and high cost in terms of human and financial resources. Recent studies have attempted to characterise the human gut microbiome by combining culturomics and NGS techniques, defined as cultureomics [7]. In one study, faecal samples cultured under 212 distinct conditions yielded 340 bacterial species through culturomics, whereas metagenomic sequencing identified 698 species; only 51 species were common between methods, shown that a combination of the two methods can capture more gut microbes [4]. Another study showed that culturomics enable a 20% higher richness when compared to molecular approaches [8]. Lau et al. identified culturable gut microbiota by comparing the operational taxonomic units (OTUs) recovered from 16S sequencing of culture plates with OTUs derived from culture-independent sequencing of fecal samples. Their analysis demonstrated that characterizing plate-grown communities through this approach revealed greater bacterial diversity than culture-independent sequencing alone [9]. However, because this amplicon sequence variant approach is accurate only to the genus level, and it cannot be used to elucidate the functional potential of bacteria identified through culture [10]. The above studies show that microbial culture remain valuable for exploring the gut microbiome, but the current related research on culture still has certain limitations.

In view of the above background, here, this study proposes a new idea, we used 24 cultivation conditions and high-throughput metagenomic sequencing to determine the phylogenetic diversity of bacteria grown on different cultivation media, referred to as culture-enriched metagenomic sequencing (CEMS). We also conducted “experienced colony picking” (ECP) and culture-independent metagenomic sequencing (CIMS) to identify as many microbial species in faecal samples as possible. The degree of overlap among microbiota obtained by manual ECP and CEMS was evaluated to determine the effectiveness of ECP. The bacterial species obtained by CEMS and CIMS were also compared and analysed, to evaluate the effectiveness of both methods in identifying sample microorganisms. In addition, we used metagenomic sequencing technology to determine the function and growth rate index (GRiD) of bacteria grown on different media.

Methods

The experimental design of this study is illustrated in Fig. 1A, and the data analysis workflow is illustrated in Fig. 1B.

Fig. 1.

(A) Experimental design, Stool sample was plated onto 12 media and incubated either aerobically or anaerobically. Routine empirical picking of colonies grown on each media type (Experienced colony picking). Metagenomic sequencing was conducted total colonies grown on each media type (Culture-enriched metagenomic sequencing) as well as on the stool sample (Culture-independent metagenomic sequencing). (B) Data analysis workflow

Sample collection

Sample was obtained from a healthy buryat mongolian child from Hushuo Gacha (a traditional animal husbandry village), Xisumu Bayan, Hulunbuir Ewenki Banner, Inner Mongolia, the minority population is small, and no studies have reported their gut microbial diversity. The informed consent was obtained from the subject’s guardian prior to collecting stool samples from children. The subject had no gastrointestinal symptoms and had not used antibiotics or probiotics within 6 months of the study. The subject was required to self-collect a faecal sample into airtight sterile stool specimen collection tubes provided in advance, and return them to members of the research team. The sample was then frozen immediately in liquid nitrogen and transported to the laboratory on dry ice within 12 h. The sample was stored at − 80 °C until further analysis.

Medium design

In this study, we used 12 media (Table S1) based on those commonly used for intestinal microbial isolation, including commercial media and modifications thereof. Commercial media were purchased from Qingdao Haibo Biotechnology Co., Ltd. (Qingdao, China). We initially selected eight medium types; among these, LGAM, PYG, GLB and MGAM are nutrient—rich media (Type I) for culturing intestinal bacteria. PYA and PYD are media (Type II) for enriching the culture of probiotics. PGAM (Type III), DGAM (Type IV),and MAR (Type V) are selective media with high acid, bile salt and salt content respectively. 1/10GAM is an oligotrophic medium (Type VI). There are also two other types, MRS-L (Type VII) and RG (Type VIII), which are media for selectively culturing Bifidobacterium and Lactobacillus respectively. All media were prepared following the manufacturer’s instructions, or as described previously [4, 11, 12], unless otherwise stated.

Culturing of faecal samples

All manipulations were performed in a Type B Vinyl Anaerobic Chamber (Coy Lab Products) filled with an atmosphere of 95% nitrogen and 5% hydrogen. The sample was thawed with cold water. Faecal samples (0.5 g) were shaken well with 4.5 g distilled water and tenfold dilutions were prepared in 0.85% NaCl solution. To cultivate as many microorganisms as possible, five gradients of dilutions 10^–3, 10^–4, 10^–5, 10^–6, and 10^–7 were used. We plated 200 μL of the prepared dilutions on agar plates containing the various types of media. One set of media was incubated at 37 °C for 5 days in an anaerobic chamber, and another was incubated at 37 °C in an aerobic constant-temperature incubator for 5 days. Among these, 1/10 GAM contained the fewest nutrients; therefore, we extended the culture time to 7 days. After incubation, we selected one or two single colonies of the same type from the plate, evaluated them in terms of size, shape, colour, and protrusion degree, and recorded their shapes. The selected colonies were streaked and purified on solid medium to obtain single-strain pure culture, and then subcultured to obtain pure culture isolates; some colonies were mixed with 10% skim milk and stored at − 80 °C, and others were used to extract bacterial DNA. For each medium type, colonies on the same culture media, including those remaining after single-colony selection, were collected from each plate by adding 1 mL of 0.85% NaCl solution and scraping the plate surface with a cell scraper. Anaerobic and aerobic of each medium (10^–3 to 10^–7) were combined and 5 mL of the harvested colonies, i.e. bacterial culture (after centrifugation to remove the supernatant) was frozen in 10% skim milk at − 80 °C as stock, and the other 5 mL used for DNA extraction.

DNA extraction, shotgun metagenomic sequencing, and quality control

DNA of single-bacterium isolated strains was extracted using the TIANamp Bacteria DNA Kit (Qiagen GmbH, Hilden, Germany) according to the manufacturer’s instructions. The quality of the DNA extracts was determined using 1% agarose gel electrophoresis and a Nanodrop spectrophotometer (Thermo Scientific, Waltham, MA, USA). Polymerase chain reaction (PCR) amplification of the 16S rRNA gene was performed on the qualified DNA, and the amplified products were sent to Paisenuo Biological Co. (Shanghai, China) for Sanger sequencing.

Metagenomic DNA was extracted from a total of 12 bacterial cultures (5 mL each) and approximately 100 mg of stool sample using the QIAamp Fast DNA Stool Mini Kit (Qiagen GmbH), according to the manufacturer’s instructions. The quality of the DNA extracts was examined using 0.6% agarose gel electrophoresis and a Nanodrop spectrophotometer, and indexed by the absorbance ratio at 260 nm/280 nm. Shotgun metagenomic sequencing was performed using the Illumina HiSeq 2500 instrument (Illumina, San Diego, CA, USA). Libraries were constructed using DNA fragments approximately 300 bp in length, and paired-end reads were generated using 100 bp in the forward and reverse directions. Low-quality reads, adaptor sequences, and contaminating host reads were removed.

Identification of microbiota

We used the SeqMan contig assembly module (DNASTAR, Madison, WI, USA) to assemble the front and back end sequences of single-bacterium isolates. After obtaining complete error-free sequences, first the sequences were identified preliminarily using the BLAST sequence alignment tool from the National Center for Biotechnology Information (NCBI). Phylogenetic analysis was then performed using MEGA 7.0 software [13] by constructing a 16S rRNA gene-based tree with closely related reference strains to further determine its taxonomic position at the species level.

For metagenomic sequencing, an average of 6.73 Gb of high-quality paired-end reads were obtained for each sample, 13 samples, totaling 87.44 Gb of high-quality data. We used HUMANN2 (ver. 0.11.2) [14] for metagenomic analysis. It relied on MetaPhlAn2 for microbial composition profiling. For function, DIAMOND was used to align reads to UniRef, and gene families were mapped to ChocoPhlAn. Metabolic pathways were inferred via KEGG.

Illumina metagenomic assembly and binning

Illumina metagenomic sequence data for all samples were assembled using metaSPAdes (ver. 3.13.0) [15], with the following parameters: -k 33,55,77,99,111 -meta. QUAST (ver. 5.0.0) [16] was used to evaluate the metagenomic assembly results, and assembled scaffolds > 2,000 bp in length were used to generate metagenome-assembled genomes (MAGs). MetaBAT 2 (ver. 2.12.1) [17] and MaxBin2 (ver. 2.2.1) [18] were used to bin the assemblies. The coverage depth required for binning was inferred by mapping the raw reads to their assemblies using Bowtie2, and the corresponding read depths of individual scaffolds were calculated using SAMtools (ver. 1.9) [19] and the jgi_summarize_bam_contig_depths function in MetaBAT 2 (ver. 2.12.1). The completeness and contamination of each MAG were estimated using CheckM (ver. 1.0.18) [20] using lineage_wf workflow. High-quality MAGs (completeness > 80%, contamination < 5%) were selected for downstream analysis.

Species-level genome bin refinement and GRiD calculation

Next, high-quality MAGs were clustered into species-level genome bins (SGBs) using dRep (ver. 2.2.4) [21] with the following parameters: -pa 0.95 -sa 0.95 -nc 0.30 -cm larger. For each SGB, open reading frames (ORFs) were predicted using Prodigal (v.2.6.3; meta option) [22] with the default parameters. SGBs replication rates were calculated by GRiD (ver. 1. 3) [23].

Carbohydrate-active enzyme annotation

Carbohydrate-active enzymes (CAZymes) carried by SGBs were annotated using the dbCAN2 database with the HMMER tool [24].

Statistical analyses

All statistical analyses were performed in R v4.0.2. The vegan package (ver. 2.5–2) (https://CRAN.R-project.org/package=vegan) was applied for principal coordinates analysis (PCoA) using the Jaccard distance. The ggplot2 package (ver. 3.1.0) was used for data visualisation. A heatmap was constructed using the pheatmap package (ver. 1.0.12).

Results

Analysis of microbiota composition by the ECP method

We isolated 105 pure strains by the ECP method, including under anaerobic and aerobic conditions. Four strains were discarded due to double sequencing peaks, for a final total of 101 single strains; 11 of the strains were suspected to be new strains according to 16S rRNA gene identification. Under anaerobic conditions, 57 strains were isolated and 16 species were identified, whereas under aerobic conditions 44 strains were isolated and 16 species were identified. The detailed results are provided in Table S2. Consistent with common knowledge, more bacteria were isolated from faeces under anaerobic than aerobic conditions. Lactococcus garvieae can be isolated using 9 types of culture media. Bifidobacterium longum and Clostridium perfringens can be obtained with 8 types of culture media. Bacillus sp., Enterococcus durans, Enterococcus faecalis, L. garvieae and Vagococcus fluvialis can be isolated under both aerobic and anaerobic conditions (Fig. 2).

Fig. 2.

Number of medium and culture conditions for bacterial strains isolated from gut microbiome samples using the “experienced colony picking” (ECP) method

Analysis of microbiota composition by CEMS and CIMS method

The phyla with the highest average relative content among the tested samples according to CEMS were Firmicutes (73.59%), Proteobacteria (24.65%), and Actinobacteria (1.72%), whereas those identified by CIMS were Bacteroidetes (64.41%), Firmicutes (30.47%), Actinobacteria (3.43%), and Proteobacteria (1.44%), indicating that most unculturable bacteria belonged to Bacteroidetes (Fig. S1). The genera showing the highest content according to CEMS were Lactococcus (30.70%), Clostridium (23.58%), and Escherichia (21.77%); the Lactobacillus content was higher in GLB and RG media, while Staphylococcus and Acinetobacter contents were higher in 1/10 GAM and GLB media, respectively (Fig. 3A). The genera showing the highest average relative content according to CIMS were Prevotella (56.50%) and Faecalibacterium (12.62%); Lactococcus and Clostridium had the lowest contents (< 0.5%), but the average relative contents of those genera were higher with culture (Fig. 3B). The dominant species identified by CEMS were C. perfringens (23.32%), Escherichia coli (21.77%), Lactococcus petauri (18.04%), and L. garvieae (12.66%) (Fig. 3C), while the contents for Prevotella sp AM42 24 (22.88%), Prevotella stercorea (17.64%), Prevotella copri (15.44%) and Faecalibacterium prausnitzii (12.62%), were higher with CIMS (Fig. 3D). In summary, the dominant bacterial types identified by CEMS were distinct from those identified by CIMS at all taxonomic levels. In addition, some bacterial genera with high relative abundance according to CEMS analysis had < 1% average relative content according to CIMS, indicating the advantage of culture for detecting low-abundance bacteria in the gut microbiota.

Fig. 3.

Stacked bar charts showing the relative abundances of the microbiota identified by culture-enriched metagenomic sequencing (CEMS) and culture-independent metagenomic sequencing (CIMS) at the (A,B) genus and (C,D) species levels

Comparison of the number of species identified by ECP and CEMS

Because the microbiota identified by ECP and CEMS were derived from the same media, except that a few strains picked from the LGAM, GLB, PGAM, and RG culture media are not included in the results identified by CEMS (At the genus level, all the species identified by the ECP method are included in the results identified by CEMS, considering due to sequencing or other methodology-related errors), the bacterial species identified by ECP in the remaining culture media are all included in the results identified by CEMS. The largest numbers of species were identified in GLB, PYD, and MGAM culture media, among which GLB and MGAM are nutrient-rich (Fig. S2). Only 10 species were obtained from MRS-L medium, confirming its high selectivity, and the main bacterial genera isolated were Clostridium and Bifidobacterium (Fig. 3A; Fig. S2).

Comparison of the number of species identified by CEMS and CIMS

CEMs and CIMS analysis identified fewer common than unique species, indicating that the species identified by culture differed greatly from those identified by sequencing. Thus, both culture and sequencing are important for comprehensive analysis of microbial species, and the combination of two methods greatly enhances gut microbe identification (Fig. 4A). In addition, the study focused on the average relative abundance of bacteria identified by CIMS of sample was greater than 0% (that is, all bacteria), 0% ~ 0.01%, 0.01% ~ 0.1%, 0.1% ~ 1%, and more than 1%, we calculated the proportion of these bacteria identified in CEMS. It was found that using average relative abundance thresholds of 0.01 ~ 0.1% and 0.1% ~ 1% (in CIMS), 32.61% and 42.75% of the strains were cultured, respectively. The proportion of the content greater than 1% was only 27.27%, and the total proportion was 28.35% (Fig. 4B).

Fig. 4.

(A) The number of bacterial species identified as common and unique by CIMS and CEMS. (B) Comparison of species identified through CEMS of faecal samples with those identified through CIMS. Color gradients represent the average relative abundance of bacterial species according to CIMS data. Proportional values indicate the percentage of bacterial species detected by CEMS relative to those detected by CIMS, stratified by ranges of average relative abundance

ECP, CEMS, and CIMS for comprehensive microorganism identification

A total of 203 bacterial species were identified in the samples using all three methods, among which 32 (15.76%) were obtained by ECP, 109 (53.69%) by CEMS, and 127 (62.56%) by direct CIMS. Bacterial species identified by ECP accounted for only 26.61% of those identified by CEMS, indicating that the traditional empirical approach of manually selecting colonies for isolation and identification neglects a large proportion of bacterial species. Only 36 species were identified by both CEMS and CIMS; 73 and 91 species were uniquely identified by CEMS and CIMS, respectively (Fig. 5).

Fig. 5.

Venn diagrams showing common and unique bacterial species identified by ECP (blue), CEMS (pink), and CIMS (green)

Microbiota composition and functions identified by CEMS and CIMS

We performed species-level principal coordinate analysis (PCoA) based on the Jaccard distance to examine the microbiota composition and functions identified by CEMS and CIMS. We found clear differences between the microbiota isolated from most culture media and those identified by CIMS (Fig. 6A, C). Combined with the cluster analysis results, our findings showed that similar microbiota were isolated from GLB and PYD media, and from LGAM, DGAM, and MGAM media, and from PYA, PGAM and PYG media. PCoA also showed that microbiota identified through the CEMS method were within a narrow functional range similar to those identified by direct CIMS, except for those cultured on PYG, PYA, PGAM, RG, and MRS-L (which were scattered throughout the coordinate system). Thus, the isolated microbiota differed somewhat from those identified through direct sequencing, but had similar functions (Fig. 6B, D).

Fig. 6.

Principal coordinates analysis (PCoA) results were plotted according to the (A) composition and (B) functions of microbiota identified through CEMS and CIMS. Clusters based on the Jaccard distance indicate the (C) composition and (D) functions of microbiota identified by CEMS and CIMS. Metabolic pathways (E) and carbohydrate-active enzymes (CAZymes) (F) enriched by microbiota identified by CEMS and CIMS. Note: Different colors represent different major categories of culture media, and CIMS is represented by a single color

Metabolic pathway (avergage relative content > 1%) and CAZymes (the number) analysis of bacteria was performed to determine the functions enriched in bacteria identified through CEMS and CIMS. Significant differences in enriched metabolic pathways among different media were seen in the Fig. 6E, including adenine and adenosine salvage III (PWY-6609) on PYG, PYA, PGAM and RG (high), pyruvate fermentation yielding acetate and lactate II (PWY − 5100) on RG (high), dTDP-L-rhamnose biosynthesis I (DTDPRHAMSYN-PWY) on RG (high), GDP-mannose biosynthesis (PWY-5659) on MRS-L and 1/10 GAM (low), guanosine ribonucleotides de novo biosynthesis (PWY-7221) on PYG and PYA (low), and L − threonine biosynthesis superpathway (THRESYN-PWY) on PYG (low). CAZyme enrichment was high in the glycoside hydrolases (GHs) category. Polysaccharide lyases (PLs) enrichment was higher in CIMS. Regarding the CEMS results, CAZyme enrichment was lowest in bacteria obtained from MRS-L medium. The direct CIMS results were similar to the culture results for MGAM medium. Generally, CAZyme enrichment differed little between bacteria identified by CEMS and CIMS, but was slightly higher content for bacteria identified by CIMS (Fig. 6F).

GRiD values of strains identified by CEMS on different culture media

The GRiD values of SGBs were analysed based on the genome replication ratio to identify strains showing active growth on various media, where GRiD = 1 indicates no bacterial genome replication, 1 < GRiD < 2 indicates faster replication, and GRiD > 2 suggests multiple-fork replication. The results are shown in Fig. 7. We screened strains with coverage > 0.5, heterogeneity < 0.3 and GRiD > 1, and obtained 20 SGBs; most of these were opportunistic pathogens (Acinetobacter lwoffii, Aerococcus viridans, C. perfringens, Enterococcus sp., Myroides odoratimimus, Psychrobacter immobilis, Shigella sonnei, and Staphylococcus sp.). C. perfringens showed a GRiD > 1 in all 12 media, indicating that a variety of media support its growth. Lactobacillus grew well on GLB, RG, and PGAM media, while Bifidobacterium grew better on MRS-L, PYA, and PYG. Moreover, in the ECP method, a relatively large number of Lactobacillus and Bifidobacterium were also isolated from these culture media (Table S2).

Fig. 7.

Growth rate index (GRiD) values of strains identified by CEMS on various media. Note: Media with the same colour are of the same type

Discussion

The human gut microbiome is a current research hotspot due to its key roles in health and disease. In the past, culture-based methods are generally thought to capture only a small fraction of the microorganisms in the gut; therefore, non-culture sequencing technologies should be applied to better reflect the gut microbiota [25]. The recent proliferation of culture technologies, including the encapsulation of bacteria into microdroplets [26], diffusion chambers that simulate the natural environment of the samples [27], microfabricated cultivation chips [28] and design of more effective culture media [29] with high-throughput identification of the cultured bacteria [30], has re-emphasised the importance of culture for gut microbiome research. Previous studies reported that only 15% of species were detected by both non-culture sequencing and culturomics approaches [12], and that methodological differences between these two approaches presents a challenge to the interpretation of culturable microbiota and overall microbiome diversity in the gut [30]. In this study, to compare the diversity of human gut microorganisms capable of growth under various cultivation conditions with those identified using culture-independent methods, we used metagenomic sequencing to determine the phylogenetic diversity of bacteria among 12 different cultivation media. Most cells recovered from culture were in samples diluted 1,000–1,000,000-fold, further reducing the likelihood of DNA in non-growing cells surviving incubation. In the present study, all media were incubated for 5 days, except for the 1/10 GAM oligonutrient medium (which was incubated for 7 days). A previous study found that on most of the tested media, colony numbers were stable by day 3, increased significantly from day 5 to 7 in a low-nutrition medium, and generally increased with culture time [31]. Therefore, to obtain as many microorganisms as possible, we extended the culture time by 2 days in the present study.

The sample from this study derived from a traditional husbandry village in Mongolia, which is an understudied population. This study found that the content of Bacteroidetes was higher than Firmicutes in the gut of child in this area, and the dominant genus was Prevotella. Previous studies have shown that children with a diet rich in starch, fiber and plant protein and low in fat and animal protein show significant enrichment of Bacteroidetes and reduction of Firmicutes, and the diet with a high content of Prevotella is also characterized by a vegetarian diet [32, 33]. The local population in this study preferes dairy and meat products, and their diet tend to be animal-based. However, by asking the parents of the subject, the child usually ate less meat and preferred vegetarian foods such as potatoes, so it is speculated that the characteristics of her gut microbiota were related to personal eating habits.

Compared with conventional methods of cultureomics, the ECP method used in this study relies on artificial naked eye to select characteristic colonies, resulting in a small number of selected colonies and poor experimental effect, which requires special attention in later research. The gut microbiota is mainly composed of anaerobic bacteria, facultative anaerobic bacteria, and aerobic bacteria; anaerobic bacteria account for the highest proportion [34]. Some research has reported that a number of Clostridia/Clostridium species were isolated only from culture. Clostridia species are endospore formers that can be challenging to lyse in spore form, making it difficult to obtain their DNA; some Clostridia strains were not identified through sequencing in this study [35]. This also shows the advantage of culture methods for detecting low-abundance bacteria in gut microbiota. ECP is a common method for bacterial isolation and identification, which allows for rapid isolation and the acquisition of pure cultures of sampled strains. However, this method is relatively poor in terms of its coverage of cultivated live bacteria. CEMS identified 109 microbial species, only 32 of which were identified by ECP. Although our results were influenced by sample size and sequencing accuracy, it also reminds us to select as many colonies with similar morphology as possible in future studies using a manual (naked-eye) colony selection approach. It is also likely that morphological similarities among colonies of different species complicated the selection process, such that some target species were overlooked.

In this study, the proportion of microbes identified using both culture and sequencing methods was not high, with only 28.35% of strains identified by CEMS also being identified by CIMS; in contrast, sequencing identified many strains not obtained from culture. Some bacteria may have been omitted from the culture process because their DNA were from dead cells; other bacteria may have been outcompeted because their growth requires a larger nutrient supply than less selective bacterial species. The gut is a complex nutritional environment that provides a rich and diverse diet for bacteria, such that the simple media compositions used in this study may have lacked the nutritional factors necessary for some strains (thereby affecting the growth of specific strains or their symbiotic bacteria). However, the research also found that many strains were obtained only via culture in our study. Microbial metagenomic sequencing is associated with accuracy and depth issues; the sequencing depth used in most microbiome studies may lead to bacteria that are present in faeces samples in amounts < 10⁶ cells/g being overlooked; culturing can help identify these low-abundance species, especially in selective media that filter highly abundant non-target species [4]. In addition, during the extraction of microbial metagenomic DNA, some bacterial species will inevitably go undetected during sample processing. In this study, the microbiota identified in each medium were compared with the stool sample sequencing results, and none of the media well-represented the faecal sample composition. The microbial communities cultured on different media vary, reflecting the inter—individual heterogeneity of the microbiota. Therefore, it is difficult to predict the minimum set of media that can capture most of the bacterial species in all samples [36]. In conclusion, a combination of culture and direct sequencing methods is an effective approach for studying the composition of the gut microbiome. However, further studies with larger sample sizes are required to confirm this finding.

The Jaccard distance, which considers only the presence or absence of species and functional potential without considering their abundance, was used herein to study microbiota composition and functions according to the CEMS and CIMS results. The composition of the microbiota differed, although bacterial functions were similar; in some cases, the composition was similar but the functions differed, perhaps due to bacterial species heterogeneity [34]. Our analysis of enriched metabolic pathways showed that the oligonutrient medium 1/10 GAM provided results closest to the original sequencing sample; nutrient dilution somewhat limited the growth of the dominant bacteria, restoring the functions of the microbiota [37, 38], and this idea can serve as a reference for future bacterial isolation and culture studies. MRS-L medium contains the antioxidant L-cysteine hydrochloride, and is therefore highly selective for the culture of anaerobic Bifidobacterium bacteria; thus, few microorganisms were isolated and functional enrichment was minimal. Psychrobacter exhibits a degree of salt tolerance; its content was higher in DGAM and MAR media, which have higher salt content [39]. The functions of bacteria isolated from PGAM and RG media were similar due to the higher acid tolerance of the cultured strains. Thus, due to the specificity and complexity of the media, the distribution of functions was similar among species types. No medium can fully represent the functions of all microorganisms within a faecal sample, and classification based on the initial medium categories (e.g. nutrient-rich, oligonutrient, high bile salt, high salt, high acid, and probiotic-selective) was difficult. The study also found that the bacteria identified by CIMS were enriched in PLs, which further indicated that the bacteria enriched in the sample had a strong utilization ability for starch, fiber and other plant substances.

Conclusion

This study showed that the CEMS approach helped culture-independent sequencing to capture more gut microbial diversity and restored some low-abundance bacterial species in the gut; the CEMS approach provides a basis for the design of culture media to incubate new microorganisms, and for the targeted screening of bacterial strains. CEMS will also allow for functional characterisation of bacterial populations, which will contribute to elucidation of their biological activity during host–bacteria and bacterial interactions in of the context of healthy and diseased states. We also found that CEMS approach could effectively identify many species missed by the ECP in the process of identifying species in the samples.

Abbreviations

ECP

Experienced colony picking

CEMS

Culture-enriched metagenomic sequencing

CIMS

Culture-independent metagenomic sequencing

GRiD

Growth rate index

NGS

Next-generation sequencing

NCBI

National center for biotechnology information

MAGs

Metagenome-assembled genomes

SGBs

Species-level genome bins;

CAZymes

Carbohydrate-active enzymes

Authors’ contributions

C.Q.Y., and J.G.W. designed the experiments. C.Q.Y., Y.Z., L.J.Y. and J.J.E. analyzed the data. C.Q.Y., and J.G.W. drafted the manuscript. All authors revised the manuscript.

Funding

This study was supported by the Major Program of Natural Science Foundation of Inner Mongolia (2018ZD14), the Fundamental Research Program of Shanxi Province (2024L154), and the Scientific and Technological Innovation Programs of Higher Education Institutions in Shanxi (202403021212101).

Data availability

The raw sequencing dataset in this study have been deposited to the NCBI SRA database (BioProject: PRJNA856821).

Declarations

Ethics approval and consent to participate

All humans and/or human data, or material used in this study adhered to the Declaration of Helsinki. Specifically, the study protocol was reviewed and approved by the Ethical Committee of the Affiliated Hospital of Inner Mongolia Medical University in Hohhot, China (approval no. KY2020014). The informed consent was obtained from the subject’s guardian prior to collecting stool samples from children (under the age of 16).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.