A new Illumina MiSeq high-throughput sequencing-based method for evaluating the composition of the Bacteroides community in the intestine using the rpsD gene sequence.

Bacteroides is a bacterial genus that is known to closely interact with the host. The potential role of this genus is associated with its ecological status and distribution in the intestine. However, the current 16S V3-V4 region sequencing method can only detect the abundance of this genus, revealing a need for a novel sequencing method that can elucidate the composition of Bacteroides in the human gut microbiota. In this study, a core gene, rpsD, was selected as a template for the design of a Bacteroides-specific primer set. We used this primer set to develop a novel assay based on the Illumina MiSeq sequencing platform that enabled an accurate assessment of the Bacteroides compositions in complex samples. Known amounts of genomic DNA from 10 Bacteroides species were mixed with a complex sample and used to evaluate the performance and detection limit of our assay. The results were highly consistent with those of direct sequencing with a low Bacteroides DNA detection threshold (0.01 ng), supporting the reliability of our assay. In addition, the assay could detect all the known Bacteroides species within the faecal sample. In summary, we provide a sensitive and specific approach to determining the Bacteroides species in complex samples.

Introduction

Bacteroides is among the most abundant gram‐negative bacterial genera in the human gut, accounting for up to 25% of the total intestinal microbiota (Ochoa‐Repáraz et al., 2010). According to data from the National Center for Biotechnology Information (NCBI) and the European Molecular Biology Laboratory (EMBL), the genomic data of 42 Bacteroides species can be collected from NCBI and EMBL. It has been reported as a relatively complex genus and includes diverse species (Ley et al., 2008). As commensal and mutualist bacteria, Bacteroides species can establish stable relationships with hosts (Faith et al., 2013) and play potential probiotic roles, such as resolving diseases (Mazmanian et al., 2008; Ochoa‐Repáraz et al., 2010; Hsiao et al., 2013), aiding digestion (Xu and Gordon, 2003) and enhancing immunity (Mazmanian et al., 2005). Interestingly, some of these beneficial functions are associated with the ecological status and distribution of Bacteroides species in the intestine (Ley et al., 2005, 2006). For example, studies have shown that a low abundance of Bacteroides uniformis in the intestine of a formula‐fed infant is associated with a high risk of obesity (Owen et al., 2005; Sanchez et al., 2011). The abundance of Bacteroides acidifaciens may be associated with metabolic diseases, such as diabetes and obesity (Yang et al., 2017). Similarly, a significantly increased abundance of other Bacteroides species, such as Bacteroides vulgatus, was observed in subjects with type 2 diabetics (Remely et al., 2016). Given the importance of Bacteroides spp. in human health, ongoing research has focused on assessments of the diversity and composition of this genus in the human intestinal tract.

The development of molecular biological methods, particularly next‐generation sequencing (NGS) based on the 16S rRNA gene, has enabled the thorough examination of a variety of samples. Consequently, a variety of methods based on this approach have been proposed. One previous study described a two‐step multiplex PCR assay based on the 16S rRNA gene, the 16S‐23S rRNA ISR and a variable region of the 23S rDNA that could be used to identify 10 Bacteroides fragilis group species (Liu et al., 2003). However, based on the current classification of B. fragilis group species, this method may be inaccurate since two species (Bacteroides distasonis and Bacteroides merdae) reported in that study have been classified into Parabacteroides species by a following study (Sakamoto and Benno, 2006). Another study designed 19 oligonucleotide primers based on the 16S rRNA genes and detected 14 Bacteroides spp. at different hierarchical levels using a method called hierarchical oligonucleotide primer extension (HOPE) (Hong et al., 2008). However, the sensitivity of the HOPE method may be affected by the excessively long poly(A) tail (Wu and Liu, 2007) and the presence of nontarget DNA in a complex environmental sample (Hong et al., 2008). The number of Bacteroides species (14) that can be detected by HOPE is also limited. Besides, one study has reported a gyrB‐based real‐time PCR system that can detect B. fragilis as a human‐specific marker of faecal contamination (Lee and Lee, 2010). But this method lack efficacy and can only be used to identify one species of Bacteroides (B. fragilis). Consequently, an improved dual‐indexing amplification and sequencing approach based on the V3–V4 region of the 16S rRNA gene and the Illumina MiSeq platform was developed to assess the composition of microbial communities in clinical samples (Fadrosh et al., 2014). However, an analysis of the V3–V4 region can only identify microbial flora at the genus level, which has limited the progress of Bacteroides‐specific research. Therefore, there is a need to develop a simple and direct method to elucidate the relative abundances of different Bacteroides spp. present in complex intestinal microbial communities.

A previous study demonstrated that because each species harbours a unique DNA sequence, the choice of the correct probe sequence and use of sufficiently stringent assay conditions can enable a very selective DNA sequence‐based detection method (Kreader, 1995). Recently, Lactobacillus‐specific and Bifidobacterium‐specific primer pairs based on a hypervariable core gene have been developed for the precise taxonomic identification and detection of intestinal Lactobacillus (Xie et al., 2019) and Bifidobacterium (Hu et al., 2017) species, respectively. Thus, the selection of a unique DNA sequence suggests a potential approach to the rapid detection and measurement of the relative abundances of other bacteria at species levels.

In this study, we developed a Bacteroides‐specific primer pair using the core gene rpsD, which is present in the genomic sequences of all Bacteroides species, as a discriminative marker. We then assessed the precision, quantification limit, detection limit and detection efficiency of this primer pair and developed a novel method to quantify Bacteroides spp. in human and mouse faecal samples using high‐throughput sequencing.

Results

Selection and phylogenetic analysis of the core gene

Thirty‐one core genes were identified as alternative target genes by Roary. Only the rpsD gene was demonstrated to have a higher discriminating power for Bacteroides than the 16S rRNA V3‐V4 region gene. As shown in the neighbour‐joining tree based on the hypervariable sequence region within the rpsD gene in Figure S1, different Bacteroides species were placed into different clades, while the same species were placed into a single clade. In contrast, the phylogenetic tree based on the V3–V4 region gene did not allow distinctions between closely related Bacteroides species. As shown in the Figure S2, B. uniformis dnLKV2 01149 and Bacteroides fluxus YIT12057 were resolved into one clade, as were Bacteroides faecis MAJ27 and Bacteroides thetaiotaomicron 7330. Thus, the rpsD gene was selected as the template because of its high resolving power for species discrimination in the Bacteroides genus.

Comparative analysis of the rpsD and 16S rRNA gene resolving power

The comparative analysis of the rpsD of 42 Bacteroides species showed that the per cent identity of the rpsD gene ranged from 76.38% to 100%, and the average value was 87.22%. The minimum and maximum per cent identity of the 16S rRNA gene were 84.77% and 100%, respectively, with an average value of 91.96%. This finding demonstrates that the rpsD gene had higher levels of taxonomic and phylogenetic resolving power at the Bacteroides species level than the 16S rRNA gene. Therefore, the rpsD gene was used as the template for Bacteroides‐specific primer design.

Design of the species‐specific primer

The full‐length rpsD gene is approximately 606‐bp long, which is unsuitable for a short‐read sequencing platform. Therefore, a partial Bacteroides rpsD gene was used for sequencing and designing a novel primer set as per the primer design criteria. Based on the results of a multiple sequence alignment, the 24–508‐bp region of the rpsD gene was targeted for PCR amplification. All of these rpsD genes have been deposited in NCBI database (GenBank accession number: MT152002–MT152101). A potential Bacteroides group‐specific primer pair, Bif‐rpsD‐F (5′‐AWCDAGAATHGCMCGTAA‐′3)/Bif‐rpsD‐R (5′‐YRTCCCAYTCCAACCA‐′3), was selected and manually designed based on the hypervariable sequence region within the rpsD gene using MEGA 5 and Primer Premier 6.0. The total volume per reaction was 50 μl, and each reaction contained 1 µl of DNA template, 25 µl of 2×Taq PCR MasterMix (Sangon, Shanghai, China), 1 µl of each primer at 10 µM (Sangon, Shanghai, China) and 22 µl of double‐distilled water (ddH2O). The following PCR conditions were used to amplify the 485‐bp rpsD target sequence: initial denaturation at 95°C for 8 min; 30 cycles of 95°C for 40 s, 50°C for 40 s and 72°C for 40 s; and a final extension at 72°C for 8 min.

Database construction based on rpsD genes

To construct a DNA database for sequencing on the Illumina MiSeq platform, all rpsD genes from reported Bacteroides species should be considered. Finally, 520 rpsD genes from 42 different species of Bacteroides were collected from NCBI and EMBL. These genes were used for the construction of the database and the identification of amplified sequences. Notably, sequence similarity of 97% could be clustered into one operational taxonomic unit (OTU). OTUs based on the rpsD gene composition of Bacteroides were comparable with those in database.

Detection of the specificity, accuracy and sensitivity of the novel primer set

In silico PCR using PRIMER‐BLAST generated only a single amplicon in the Bacteroides pan‐genome. We also performed PCR using genomic DNA extracted from known bacterial species, including 14 Bacteroides strains and 8 non‐Bacteroides strains. As shown in Figure 1, a PCR product was obtained only when the genomic DNA of Bacteroides species was used as the template DNA. In addition, all reads generated for the artificial sample (with known amounts of genomic DNA extracted from 10 known Bacteroides species) were compatible with the 10 known species, and a strong correlation was observed between the normalized relative abundance predicted for Bacteroides species and the relative abundance observed by an rpsD‐profiling analysis (Fig. 2A). Furthermore, the lowest detectable amount of Bacteroides DNA (amplified using the designed primer pair) was 0.01 ng, corresponding to a detection limit of 103 CFU (Fig. 2B).

Fig. 1

Specificity of the PCR amplification of the selected partial rpsD gene region using the designed primer set. Note: M, marker; 1, ddH2O; 2, Bifidobacterium longum; 3, Lactobacillus brevis; 4, L. plantarum; 5, Enterococcus faecalis; 6, Escherichia coli; 7, Pediococcus acidilactici; 8, L. fermentum; 9, Akkermansia muciniphila; 10, Bacteroides caccae; 11, B. dorei; 12, Bacteroides eggerthii; 13, B. faecis; 14, Bacteroides salyersiae; 15, B. uniformis; 16, Bacteroides ovatus; 17, B. fragilis; 18, B. vulgatus; 19, Bacteroides stercoris; 20, Bacteroides xylanisolvens; 21, B. thetaiotaomicron; 22, B. kribbi; 23, B. koreensis.

Fig. 2

Detection accuracy and detection limit of the novel designed primer set.

A. Relationship between the normalized relative abundance predicted for Bacteroides species and the relative abundance determined by the rpsD‐profiling analysis.

B. The detection limit of the novel designed primer set based on the selected partial rpsD gene sequence. CFU: colony‐forming units.

Comparison of the robustness of the rpsD gene and the 16S rRNA V3‐V4 region

PCR using our designed primer set generated 832 342 and 669 429 copies of the high‐quality 16S rRNA gene and 528 618 and 951 729 copies of the rpsD gene from the 20 human and 20 mouse faecal samples, respectively (Table 1). In addition, approximately 18.91% and 1.56% of the reads generated from the human and mice faecal samples, respectively, could be assigned to the Bacteroides genus when using the primer set targeting the V3‐V4 region gene (Fig. 3A and B). In contrast, almost all the sequences could be assigned to the Bacteroides genus when using the novel primer pair Bif‐rpsD‐F/Bif‐rpsD‐R (Fig. 3C and D). Furthermore, the primer pair designed to target the partial rpsD gene could identify Bacteroides at the species level, whereas the universal primer set that targeted the V3‐V4 region of 16S rRNA could identify Bacteroides only at the genus level.

Table 1

Overview of sequencing results for each sample.

Sample ID	Sequence number a (16S)	OTU number b (16S)	Sequence number (rpsD)	OTU number (rpsD)	Sample ID	Sequence number a (16S)	OTU number b (16S)	Sequence number (rpsD)	OTU number (rpsD)
M‐1	36 855	4069	92 624	3999	Hu‐1	29 441	3464	11 891	827
M‐2	26 708	3049	57 356	4589	Hu‐2	62 019	5151	12 130	848
M‐3	29 059	3476	63 827	4791	Hu‐3	56 297	4766	25 825	668
M‐4	34 669	3727	70 629	4744	Hu‐4	51 110	5530	30 837	1591
M‐5	48 116	4774	80 354	5887	Hu‐5	20 584	2174	41 508	2379
M‐6	35 925	3616	49 640	3685	Hu‐6	30 888	2233	15 648	1002
M‐7	48 782	5805	26 034	2059	Hu‐7	25 166	2699	28 256	1286
M‐8	27 033	3123	29 453	2664	Hu‐8	31 504	2558	37 582	1345
M‐9	40 273	4827	41 948	3413	Hu‐9	23 045	2539	59 035	2935
M‐10	45 963	4850	21 980	2055	Hu‐10	51 727	3657	15 959	881
M‐11	23 407	2757	33 698	2921	Hu‐11	43 126	3344	29 815	1200
M‐12	27 307	3201	45 208	3911	Hu‐12	48 765	5278	16 268	1115
M‐13	26 085	3146	41 987	3876	Hu‐13	35 439	3516	19 954	1124
M‐14	33 210	3596	43 417	3510	Hu‐14	60 166	4545	21 169	1292
M‐15	26 735	3171	46 157	3755	Hu‐15	40 807	2814	37 444	1534
M‐16	25 529	2902	55 517	4307	Hu‐16	45 431	5279	42 246	1920
M‐17	38 377	4488	44 553	3697	Hu‐17	32 437	4289	15 167	852
M‐18	24 024	3024	37 192	3273	Hu‐18	49 466	4178	13 247	800
M‐19	28 566	3091	27 451	2550	Hu‐19	42 768	4020	37 892	1666
M‐20	42 806	5028	42 704	3424	Hu‐20	52 156	4446	16 745	1115

The sequence number refers to the count of assembled sequences after quality filtering.

The OTU (Operational Taxonomic Units) number is presented for all sequences without rarefaction. M, mice sample; Hu, human sample.

Fig. 3

16S rRNA gene‐based and rpsD gene‐based profiles of human and mice faecal samples using the 341F/806R and Bif‐rpsD‐F/Bif‐rpsD‐R primer pairs. Bar plots of the genus‐level microbial compositions of (A) 20 human samples and (B) 20 mouse samples. Bar plots of the species‐level microbial compositions of the (C) 20 human samples and (D) 20 mouse samples.

Discussion

The Illumina MiSeq platform provides a scalable, high‐throughput and streamlined sequencing platform for analysing the community compositions of complex samples (Fadrosh et al., 2014). Based on this sequencing technology, some approaches based on a high throughput, long sequence read length and high level of accuracy, including single‐ (Caporaso et al., 2012) and dual‐ (Kozich et al., 2013) indexing methods targeting the hypervariable region of the 16S rRNA gene, have been developed and are used widely (Fadrosh et al., 2014). These approaches allow the in‐depth analysis of complex samples (Tringe and Hugenholtz, 2008). However, the targeted region of the 16S rRNA gene has a limited resolving power at the Bacteroides species level. Therefore, a novel molecular marker with a high resolving power is needed to identify Bacteroides species. Notably, previous studies have reported that an appropriate target gene for species‐specific primers should meet the following criteria: (i) the target gene region should be prevalent in the genus at a high resolving power; (ii) the target gene region should encompass a hypervariable region and two constant regions at both ends; and (iii) the PCR amplification region in the target gene should not be longer than 500 bp (Dieffenbach et al., 1993; Hu et al., 2017; Xie et al., 2019).

In this study, the core gene rpsD, which encodes the 30S ribosomal protein S4 and exists in all Bacteroides species, was identified by Roary. The rpsD gene was reported to exist in Bacillus subtilis and is co‐transcribed with the genes for initiation factor 1 and ribosomal proteins B, S13, S11 and L17 (Boylan et al., 1989). A previous study also revealed that this gene is monocistronic (Grundy and Henkin, 1990). In our study, we observed that the partial sequences of both ends of the 485‐bp (< 500‐bp) rpsD gene (24–508 bp) were highly conserved, whereas the other sequences were more variable. These results suggest that the rpsD gene fulfils all prerequisites and should be considered as a reliable alternative phylogenetic marker for Bacteroides species.

Based on the partial rpsD gene, we designed a pair of Bacteroides‐specific primers, Bif‐rpsD‐F/Bif‐rpsD‐R, which produce a 485‐bp (< 500‐bp) amplicon and enable the rapid discrimination of all known Bacteroides species from non‐Bacteroides species. We then developed a novel method based on the Illumina MiSeq sequencing platform that allowed an accurate assessment of the Bacteroides composition in complex samples. A complex sample containing 10 genomic DNA samples was evaluated, and the results were highly consistent with those of direct sequencing, thus supporting the reliability of our assay. Moreover, the minimum Bacteroides DNA detection threshold was 0.01 ng, indicating that the designed primer set possessed a higher sensitivity than those previously reported for Bifidobacterium‐specific primers (0.05 ng) (Hu et al., 2017) and Lactobacillus‐specific primers (0.05 ng) (Xie et al., 2019).

Compared with previously reported methods that could only detect 10–14 B. fragilis group species (Liu et al., 2003; Hong et al., 2008), our newly developed rpsD‐based sequencing method was more ‘broad‐spectrum,’ as it could identify 29 Bacteroides species in human and mouse faecal samples. Besides, only one pair of Bacteroides‐specific primers were used in the present method, which improved the convenience of the assay. However, our method did not detect Bacteroides ihuae and Bacteroides timonensis in the 20 Chinese human faecal samples. These two Bacteroides species have been reported only in the sputum of healthy Frenchwomen living in Marseille (Fonkou et al., 2017) and in the faecal sample of a 21‐year‐old French Caucasian woman with severe anorexia nervosa (Ramasamy et al., 2014), respectively, indicating that they may not exist in the faecal samples of the Chinese studies in our study. Notably, B. acidifaciens and Bacteroides caecimuris were detected in our human faecal samples, although to date, these species have been detected only in mouse samples (Miyamoto and Itoh, 2000; Lagkouvardos et al., 2016). In addition, some Bacteroides species, such as Bacteroides barnesiae, Bacteroides gallinarum, B. salanitronis and Bacteroides paurosaccharolyticus, were not detected in our human and mouse samples, consistent with the findings of previous studies that identified the first three species in chicken caecal samples (Lan et al., 2006) and the latter in rice‐straw residue from a methanogenic reactor that treated waste from cattle farms (Ueki et al., 2011). However, our method has some limitations. One possible drawback of the rpsD gene is the high level of similarity between some Bacteroides species, as observed between Bacteroides dorei DSM 17855 and B. vulgatus ATCC 8482 (99.5%); Bacteroides cellulosilyticus DSM 14838 and B. timonensis AP1 (99.34%); and B. faecis MAJ27 and B. thetaiotaomicron VPI‐5482 (99.5%). Similar low levels of discriminatory power have also been reported for the groEL gene in Bifidobacterial species (Hu et al., 2017) and Lactobacillus species (Xie et al., 2019). Therefore, caution should be exercised when using the rpsD gene as a marker for the identification of certain Bacteroides species. In addition, the complete genomic information of some Bacteroides species, such as Bacteroides kribbi and Bacteroides koreensis, is not available in the current database. Consequently, these species would be labelled as unassigned Bacteroides when annotated in the database. Regular updates of the Bacteroides database are warranted to enable the identification of all Bacteroides species.

In this study, it is theoretically possible that apply the Bacteroides species composition from the novel primers to break down the overall relative abundance of the Bacteroides genus as assessed by the V3‐V4 data. We have tried to verify the feasibility of this method. The result showed that B. fragilis and B. vulgatus account for 1.33% and 5.05% of the total gut microbial population in the 20 human faecal samples, respectively. Interestingly, one previous study (Ruseler‐van Embden and Both‐Patoir, 1983) has reported that B. vulgatus accounts for 6% of the total gut microbiota of healthy humans. Besides, the abundance of B. fragilis accounts for only up to 1% of the total gut microbial population (Rocha and Smith, 2013). These results reinforced that our method is feasible. Notably, caution should be exercised when apply the primers. More well‐conducted experiments are needed to propose the efficiency and specificity of the novel primers to assess the species composition of the Bacteroides genus as assessed by the V3‐V4 data.

In summary, we developed a powerful Illumina MiSeq sequencing platform‐based method for the accurate, sensitive and rapid identification of different Bacteroides species. This method enabled the determination of the relative abundances of different Bacteroides species at a low detection limit of 103 CFU ml−1. It also yielded a high resolving power for discriminating between Bacteroides species from complex samples such as human and mouse faeces. This method can enable the elucidation of Bacteroides diversity in different ecological systems, as well as the potential roles of different Bacteroides species in host health.

Experimental procedures

Bacterial strains, culture media and DNA extraction

All bacterial strains were obtained from the Culture Collection of Food Microorganisms of Jiangnan University (Wuxi, China) and cultured at 37 °C in an anaerobic workstation (N2, 85%; H2, 10%; CO2, 5%) in different culture media (Table 2). Genomic DNA was extracted from these bacteria using the TIANamp Bacteria DNA Kit (TianGen, Beijing, China) as per the manufacturer’s instructions.

Table 2

Bacteria were used in this study.

Number	Species	Isolation source	Strain	Medium and reference
1	Bifidobacterium longum	Human faecal	FGDLZ58M1	MRS (with 1% L(+)‐Cysteine)
2	Lactobacillus breris	Human faecal	X3
3	Lactobacillus plantarum	Human faecal	Z2
4	Lactobacillus fermentum	Human faecal	B7
5	Pediococcus acidilactici	Human faecal	B18	TSB broth
	Enterococcus faecalis	Human faecal	CCFM596	BHI (with 1% L(+)‐Cysteine)
7	Escherichia coli	Human faecal	CCFM21	BHI (with 1% L(+)‐Cysteine)
8	Akkermansia muciniphila	Human faecal	ATCC BAA‐835	Anaerobic medium (gastric mucin as the sole carbon and nitrogen source) (Derrien et al., 2004)
9	B. caccae	Human faecal	FSDTA‐ELH‐2.5MIC‐3	Modified BHI (Tan et al., 2019)
10	B. dorei	Human faecal	FSDTA‐HCK‐B‐6
11	B. eggerthii	Human faecal	FSDTA‐HCK‐B‐9
12	B. faecis	Human faecal	FNMHLBE10K3
13	B. salyersiae	Human faecal	FSDTA‐ELI‐BHI‐9
14	B. uniformis	Human faecal	FSDTA‐HCK‐B1
15	B. ovatus	Human faecal	FSDTA‐HCK‐B4
16	B. fragilis	Human faecal	FSDTA‐HCK‐B8
17	B. vulgatus	Human faecal	FSDTA‐HCM‐XY‐14
18	B. stercoris	Human faecal	FFJLY21K3
19	B. xylanisolvens	Human faecal	FFJLY22K22
20	B. thetaiotaomicron	Human faecal	FGSZY48K9
21	B. kribbi	Human faecal	FQHXN7K2
22	B. koreensis	Human faecal	FYNLJ19K1

B., Bacteroides; MRS, de Man, Rogosa and Sharp broth; LB, Luria‐Bertani broth; BHI, Brian Heart Infusion broth; TSB: Tryptic Soy Broth.

Faecal sample collection and genomic DNA extraction

The faecal samples of 20 mice (male C57BL/6 mice, 6 weeks old) and humans (20 healthy people from China) were collected rapidly after defecation in faecal collection tubes under aseptic conditions and stored at −80°C until genomic DNA extraction. Genomic DNA was extracted from these samples using the method described in the Fast DNA SPIN Kit for Feces (MP Biomedicals; Carlsbad, CA, USA), with the following modifications: 0.1 g faeces sample was used to extract genomic DNA extraction and 60 μl DNA eluent were added in the clean catch tube to collect purified DNA (this step was performed twice to increase the DNA concentration of collected samples).

As shown in the Table S1, one hundred genomic sequences from 42 Bacteroides species were collected from NCBI and EMBL. The pan‐genome of the Bacteroides genus was identified using Roary software, and the selected core gene data were aligned using the CLUSTAL_X program (Thompson et al., 1997). A tree of the homologous genes was then constructed and used to select the hypervariable sequence region that would allow the precise taxonomic identification and detection of all Bacteroides spp.

Design of Bacteroides‐specific primers

The selected partial homologous gene sequences from the pan‐genome of Bacteroides species were amplified by PCR using the barcoded fusion primers (341F/806R) designed in this study. Meanwhile, the PCR conditions for the region covered by this primer pair and for the 16S rRNA V3‐V4 regions were set as described in a previous report (Jia et al., 2016). The major PCR products of the selected (i.e. primer‐covered) region and V3‐V4 region gene sequences were electrophoresed on a 2.0% agarose gel in TBE buffer, stained with SYBR SAFE (Invitrogen, Eugene, OR, USA), purified and quantified using the QIAquick Gel Extraction Kit (Qiagen, Hilden, Germany) as per the manufacturer’s instructions. A further quantification step was performed using the Quant‐iT PicoGreen dsDNA Assay Kit (Life Technologies, Carlsbad, CA, USA). The selected gene regions from all known Bacteroides species, based on NCBI and EMBL, were used to construct a DNA amplicon sequence library. This library was then sequenced on the Illumina MiSeq platform as described in a previous study (Hu et al., 2017).

Detection of primer specificity

Four tests were performed to determine the specificities of the primer set. (i) The tree generated from the V3‐V4 region of the 16S rRNA gene and the tree of selected genes constructed from the alignment of Bacteroides sequences were used to compare the efficacy of the novel primer set via mega 5.02 (Tamura et al., 2011). (ii) In silico PCR was performed using PRIMER‐BLAST, with the NCBI nonredundant database as the template (Ye et al., 2012). (iii) The genomic DNA of 14 Bacteroides strains and 8 non‐Bacteroides strains (Table 2) were extracted and PCR amplified using the novel primer set. (iv) The genomic DNA extracted from 20 healthy human and 20 mouse faecal samples were PCR amplified using the novel primer set and the V3‐V4 region gene primer set. These sequence data have been submitted to the GenBank databases under accession number SRR11212985 (Fig. 3A); SRR11213067 (Fig. 3B); SRR11213107 (Fig. 3C); SRR11213137 (Fig. 3D). Microbiota analyses were performed using 16S rRNA gene amplicon sequencing and the Quantitative Insights Into Microbial Ecology (QIIME) version 1 software.

Evaluation of primer sensitivity and detection limit

Known amounts (0.001–50 ng) of genomic DNA extracted from 10 known Bacteroides species were mixed to be an artificial sample and used to evaluate the detection sensitivity and detection limit of the designed primer set. The genomic DNA served as the template for PCR amplification with this primer set, and the obtained amplicons were then sequenced on the Illumina MiSeq sequencing platform. These sequence data have been submitted to the GenBank databases under accession number SRR11212976. Among these amplicons, the lowest detectable concentration of gene copies and the colony‐forming units (CFU) of the corresponding strain was identified as the PCR detection limit. The copy numbers of the selected gene and the CFU of the corresponding strain were estimated using a dsDNA copy number calculator (Calculator for Determining the Number of Copies of a Template, URI Genomics & Sequencing Center).

Statistical analysis

Data analysis was performed using graphpad prism 8. The data corresponding to each treatment were reported as the mean ± standard error of the mean (SEM). The statistical significance of the data was determined at the P < 0.05 level. Mean values were subjected to an analysis of variance, and statistically significant results were compared using Tukey’s test.

Author contributions

CW, JXZ, HZ, QXZ and WC conceptualized and designed the study; YX, SSF and MLP organized the database; CW performed the statistical analysis and wrote the first draft of the manuscript; SSF wrote sections of the manuscript. All authors contributed to the manuscript revision process and read and approved the submitted version. QXZ takes primary responsibility for communication with the journal and editorial office during the submission process, throughout the peer review and during publication. All authors read and approved the final manuscript.

Conflict of interest

None declared.

Ethical approval

The human participants and animal experiments in this study were approved by the Ethics Committee in Jiangnan University, China. All the faecal samples from healthy persons were for public health purposes and these were the only human materials used in present study. Written informed consent for the use of their faecal samples was obtained from the participants or their legal guardians. No human experiments were involved. The collection of faecal sample had no risk of predictable harm or discomfort to the participants.

Fig. S1. Phylogenetic tree derived from a neighbour‐joining analysis of the rpsD gene region sequences. Note: Bootstrap values (%) based on 1000 replications are presented on each node. The bar indicates 2% sequence divergence.

Click here for additional data file.

Fig. S2. Phylogenetic tree derived from a neighbour‐joining analysis of the 16S V3‐V4 gene region sequences. Note: Bootstrap values (%) based on 1000 replications are presented on each node. The bar indicates 2% sequence divergence.

Click here for additional data file.

Table S1. Basic genomic information on Bacteroides species used for comparative analysis in this study.

Click here for additional data file.