A versatile genetic toolbox for Prevotella copri enables studying polysaccharide utilization systems.

Prevotella copri is a prevalent inhabitant of the human gut and has been associated with plant-rich diet consumption and diverse health states. The underlying genetic basis of these associations remains enigmatic due to the lack of genetic tools. Here, we developed a novel versatile genetic toolbox for rapid and efficient genetic insertion and allelic exchange applicable to P. copri strains from multiple clades. Enabled by the genetic platform, we systematically investigated the specificity of polysaccharide utilization loci (PULs) and identified four highly conserved PULs for utilizing arabinan, pectic galactan, arabinoxylan, and inulin, respectively. Further genetic and functional analysis of arabinan utilization systems illustrate that P. copri has evolved two distinct types of arabinan-processing PULs (PULAra ) and that the type-II PULAra is significantly enriched in individuals consuming a vegan diet compared to other diets. In summary, this genetic toolbox will enable functional genetic studies for P. copri in future.

Introduction

The complex microbial communities residing in the intestine affect the physiology of the host influencing the balance between health and disease (Bäckhed et al, 2005; Hooper, 2009). Yet, extensive interpersonal variability in the human gut microbiota composition and function complicates the establishment of links between the presence of specific bacterial gene content to human phenotypes. Functional genetic study of these bacteria is essential to dissect the genetic basis underlying the microbe‐driven host phenotypes. However, many commensal bacterial species have so far eluded efforts for genetic engineering.

For instance, no genetic tools have been established for Prevotella copri, a common human gut microbe, whose prevalence and relative abundance have been linked to various beneficial and detrimental effects on human health (Ley, 2016; Claus, 2019; Maeda & Takeda, 2019). Specifically, P. copri has been found to be enriched in individuals at risk for rheumatoid arthritis (Scher et al, 2013; Alpizar‐Rodriguez et al, 2019; Wells et al, 2020) and in patients with enhanced insulin resistance and glucose intolerance (Pedersen et al, 2016). Conversely, others found P. copri to be positively correlated with improved glucose and insulin tolerance during intake of fiber‐rich prebiotic diets (Kovatcheva‐Datchary et al, 2015; De Vadder et al, 2016). Besides the lack of tools for genetic engineering, the establishment of functional links between P. copri and disease outcomes has been additionally complicated by its fastidious nature in vitro, tremendous strain‐level diversity, resulting in the recent recognition of multiple genetically distinct clades and the lack of corresponding diverse isolates (Tett et al, 2019).

In contrast to P. copri, members of the genus Bacteroides, as the best example, have been extensively studied via a variety of genetic tools (Koropatkin et al, 2008; Goodman et al, 2011; Mimee et al, 2015; Lim et al, 2017; García‐Bayona & Comstock, 2019; Bencivenga‐Barry et al, 2020). These studies have, for instance, identified genes required for various bacterial physiological functions and provide approaches to investigate bacteria–host interactions. Of those, the genes for degrading plant‐ and animal‐derived polysaccharides that are resistant to human digestion have been highlighted due to their important role in affecting bacterial fitness in the microbiome (Kaoutari et al, 2013; Porter & Martens, 2017; Wexler & Goodman, 2017). These genes are typically organized in so called polysaccharide utilization loci (PULs) that differ in polysaccharide specificity (Koropatkin et al, 2012). PULs are defined by the presence of one or more genes homologous to Bacteroides thetaiotaomicron susD and susC encoding outer membrane proteins that bind and import starch oligosaccharides (Shipman et al, 2000; Martens et al, 2009). The SusC/D protein complex (Glenwright et al, 2017) cooperates with diverse carbohydrate‐degrading enzymes (CAZyme), e.g., glycosyl hydrolases (GHs) and polysaccharide lyases (PLs), which are typically encoded in close proximity to the susC/D homologs in the genome. Many PULs in B. thetaiotaomicron contain genes encoding sensor‐regulator systems, such as hybrid two‐component systems (HTCSs) (Xu et al, 2003; Sonnenburg et al, 2006). HTCS proteins are chimeric proteins harboring the functional domains of a periplasmic sensor, a histidine kinase, and a DNA‐binding response regulator enabling HTCSs to recognize distinct signal components degraded from complex carbohydrates and to initiate the upregulation of CAZyme‐encoding genes in a positive feedback loop (Sonnenburg et al, 2006; Martens et al, 2011).

Notably, higher prevalence of intestinal Prevotella spp. was found in populations consuming a plant‐rich diet, e.g., vegetarians in Western populations (De Filippo et al, 2010; Wu et al, 2011; Ruengsomwong et al, 2016; Fragiadakis et al, 2019), suggesting that they encode efficient machineries for degradation of plant‐derived polysaccharides. Yet, due to the lack of genetic tools, the characterization of carbohydrate utilization has been limited to bioinformatic and phenotypic studies (De Filippis et al, 2019; Fehlner‐Peach et al, 2019). Specifically, two recent studies described extensive variability among clades and also strains within clades in the ability to directly utilize diverse complex plant carbohydrates (Fehlner‐Peach et al, 2019; Tett et al, 2019). While combinations of comparative genomics and phenotypic assays can be used to predict the ability to utilize specific polysaccharides, such approaches rely on well‐characterized genetic elements as a reference, which makes it difficult to identify genes with unknown functions and thereby hinders the establishment of casual relationship between the genetic content and phenotypes. Moreover, substrate predictions based on gene annotations from genetically distinct bacteria might be incomplete or inaccurate. This observation is supported by the presented data below that some P. copri strains harboring PULs lacking susC/D genes can still grow on the substrates of those PULs, suggesting the functional redundancy between various PUL components.

Here, we described a newly established genetic toolbox for approaching gene insertion, deletion, and complementation in P. copri. Using the genetic tools as well as high‐throughput sequencing and bioinformatic analysis, we identified four highly conserved P. copri PULs responsible for utilization of specific plant polysaccharides via HTCS activation, and demonstrated that P. copri strains have evolved two types of arabinan processing PULs. These studies not only build up a universal genetic manipulation system for an abundant bacterial species in the microbiome, but also present its applications on future efforts of understanding P. copri biology, e.g., nutrient acquisition. Because the workflow of establishing the genetic manipulation system for P. copri can be potentially modified and applied to other underexplored gut bacteria, our studies shed light on the future microbiome research on intricate interactions between bacteria–bacteria and host–bacteria during human health and disease.

Results

Development of conjugation‐based gene insertion system for P. copri

As targeted gene inactivation approaches enable gene function studies, they are frequently carried out in Bacteroides spp. by transferring a suicide plasmid from a donor strain into the recipient followed by selection of bacterial clones, which underwent homologous recombination (Koropatkin et al, 2008; García‐Bayona & Comstock, 2019; Bencivenga‐Barry et al, 2020). To adapt the system for P. copri, we considered several key differences between Bacteroides spp. and P. copri including oxygen and antibiotic sensitivity as well as promoter sequences driving expression of the selectable marker gene.

Because oxygen exposure has been reported to promote mating between Escherichia coli (donor) and Bacteroides spp. (recipient) (Salyers et al, 1999), conjugation for Bacteroides spp. is routinely performed for at least 15 h under aerobic conditions followed by transferring the cultures to anaerobic conditions permitting growth (García‐Bayona & Comstock, 2019; Bencivenga‐Barry et al, 2020). We initially tested aerotolerance of three P. copri strains, the type strain (DSM18205) and two strains (HDD04 and HDB01) from our laboratory collection containing recent isolates from healthy and diseased individuals (Fig 1A and Dataset EV1). Exposure to air decreased viability three to four orders of magnitude for the P. copri strains within only 4 h, which is in sharp contrast to B. thetaiotaomicron that displayed only a 28.6% drop in viability (Fig EV1A). Hence, all genetic manipulations for P. copri were subsequently carried out in anaerobic conditions.

Figure 1

Development of a conjugation‐based gene insertion platform for Prevotella copri strains from multiple clades

Phylogenetic tree of P. copri species complex using reference strains (n = 17) from four P. copri clades (Tett et al, 2019) and novel isolates (n = 11) used in this study. Prevotella copri clades are indicated by different colors.

Schematic illustration for targeted gene insertion system in P. copri. Primer binding sites (P1–P6) are indicated.

Detection of plasmid integration in P. copri DSM 18205 and HDD04 by PCR.

Optimization of conjugation efficacy: Influence on yield of transconjugants of promoter sequences of selection marker (D), length of homology arm (E), donor Escherichia coli strain (F), conjugation ratio of donor to recipient strains (G), recipient P. copri strains that are erythromycin‐sensitive (H), or tetracycline‐sensitive (I). Comparisons in (D–G) were performed using P. copri HDD04. ND: not detectable.

Data information: Values and error bars represent the mean of at least three biological replicates and their standard deviations (SDs), respectively. Statistical significance between groups was calculated by Student's t test (*P < 0.05; **P < 0.01; ***P < 0.001; and ****P < 0.0001; NS, P > 0.05, not statically significant).

Figure EV1

Defining the impact of potential key factors on adapting conjugation system for Prevotella copri

Survival of Bacteroides thetaiotaomicron VPI‐5482, P. copri HDD04, HDB01, and DSM 18205 in BHI + S broth in aerobic conditions. CFUs in the cultures were determined at indicated time points after exposure to normal air.

Growth of Escherichia coli S17‐1 λpir and E. coli β2155 in BHI + S broth in the presence or absence of diaminopimelic acid (DAP).

Quantification of bacterial CFUs of E. coli S17‐1 λpir and E. coli β2155 grown on BHI blood agar plates with DAP for three different time points.

Transcription levels of seven genes in P. copri HDD04 grown in BHI + S broth. The promoter sequences of these genes were used for driving ermG expression in the plasmid in Fig 1D.

Data information: Values and error bars represent at least three biological replicates and their standard deviations (SDs), respectively. Asterisks indicate the significant differences (*P < 0.05; ***P < 0.001) calculated by Student's t test. In (A), asterisks represent the significant differences between B. thetaiotaomicron VPI‐5482 and either one of three P. copri strains.

Escherichia coli S17‐1 λpir is commonly used as a donor for Bacteroides spp. in aerobic conditions, but it may show impaired growth under anaerobic conditions. Hence, we compared anaerobic growth of E. coli S17‐1 λpir and another donor strain, E. coli β2155 (Dehio & Meyer, 1997; Demarre et al, 2005). Notably, while E. coli β2155 is auxotrophic for diaminopimelic acid (DAP), it displayed in the presence of DAP faster and more robust anaerobic growth compared to E. coli S17‐1 λpir both in liquid culture and on agar plates (Fig EV1B and C). Thus, we further tested the possibility of using E. coli β2155 as the donor for delivering vectors into recipient P. copri.

The suicide plasmid, pExchange‐tdk, is extensively used for gene deletion in Bacteroides spp. (Koropatkin et al, 2008). This plasmid possesses (i) a R6K origin limiting it to replicate only in host–bacteria carrying the pir gene, and (ii) an erythromycin‐resistant gene (ermG) for selecting Bacteroides transconjugants. Firstly, no pir homologs were identified in P. copri strains, suggesting the feasibility of using pExchange‐tdk‐based vectors for plasmid integration in P. copri. Secondly, the susceptibility of P. copri strains to erythromycin was tested. From one type strain and 11 distinct strains of our collection representing four P. copri clades, eight strains from two clades (Fig 1A) were sensitive to erythromycin driving us to initially utilize an erythromycin‐based selection system (Dataset EV2). To achieve a stable expression of ermG in P. copri, we inserted a strong promoter of a P. copri housekeeping gene, i.e., elongation factor Tu gene (tuf) (Fig EV1D), in front of the ermG coding sequence into pExchange‐tdk and removed the counterselection marker (thymidine kinase gene). These modifications resulted in a new shuttle vector referred to as pEx‐insertion‐ermG (Fig 1B and Appendix Fig S1A). To consider potential negative positional effects for P. copri growth due to plasmid insertion, we individually cloned three different 3‐kb regions (DSM18205_00642‐43, 00941‐42, and 02334‐35) containing the 3′‐end coding sequences of genes from P. copri DSM 18205, as homology arms for approaching plasmid integration without disrupting any functional genes. In addition, these DNA regions are relatively conserved in genomes of our P. copri isolates, allowing us to rapidly expand the testing into different strains.

The initial conjugation was performed between E. coli β2155 carrying the respective plasmids and P. copri DSM 18205. Of note, E. coli β2155 can grow weakly on BHI blood agar plates even without DAP based on our observations. Hence, after the co‐incubation of E. coli and P. copri on BHI blood agar with DAP, besides erythromycin, we additionally added gentamicin to inhibit the E. coli donor and positively select for P. copri transconjugants. This selection yielded eight erythromycin resistant (ErmR) colonies only when pEx‐insertion‐ermG‐DSM18205_02334‐35 was used, which indeed suggests the presence of positional effects (Fig 1B). Since DSM18205_02334 encodes a putative β‐glycosidase gene (bgl), we refer to the plasmid as pEx‐insertion‐ermG‐DSM‐bgl. Integration of the transferred plasmid was determined in three individual colonies and colony PCR amplified fragments expected for successful integration (Fig 1C and Dataset EV1). DNA sequences of PCR products were further confirmed by Sanger sequencing. We next attempted conjugation using pEx‐insertion‐ermG‐HDD04‐bgl and HDD04 as the conjugation recipient, which strikingly resulted in a more than 10,000 ErmR colonies suggesting a large strain variability in conjugation efficiency and prompting us to further optimize the approach.

Optimization of conjugation‐based gene insertion for P. copri

Building on these proof‐of‐concept data, the genetic elements of the plasmid and the conjugation procedures were systematically adjusted by varying factors that likely affect the conjugation efficiency. This included the promoter of ermG, the length of homology arm for plasmid integration, the donor E. coli strains and ratio of donor to recipient for conjugation, as well as the recipient P. copri strains.

First, an efficient expression of the selection marker is the prerequisite to obtain ErmR transconjugants. Besides the promoter of the tuf gene, we chose another six different promoters of housekeeping genes showing diverse gene expression in P. copri HDD04 in BHI liquid media supplemented with fetal bovine serum (BHI+S; Figs 1D and EV1D; Dataset EV5). Notably, the numbers of transconjugants varied approximately 300‐fold depending on the promoter, but none of the other promoters yielded higher numbers than tuf, suggesting that high ermG expression levels are required for transconjugant survival under erythromycin selection (Figs 1D and EV1D). Second, comparison of homology arms between 0.5 and 4 kb demonstrated the highest yield of transconjugants with homology arms of 3 and 4 kb (Fig 1E). Third, we assessed the above‐mentioned E. coli strains as conjugation donors. In line with the ability for anaerobic growth, conjugation with donor E. coli β2155 increased the yield of transconjugants for both DSM 18205 and HDD04 approximately 85‐fold compared to E. coli S17‐1 λpir, indicating the advantage of donor strain for advancing anaerobic conjugation (Fig 1F). More transconjugants were also obtained as ratio of donor to recipient increased until 100:1, after which it decreased again (Fig 1G). Last, we evaluated a larger panel of ErmS P. copri strains using strain‐specific homology arms as extensive sequence variations among strains are present. Except for one strain with undetectable production of ErmR colonies all other seven strains exhibited extensive diversity in the number of transconjugants varying by approximately 104‐fold between the strains with the lowest (RPA01, mean = 5 CFUs) and highest (HDA03, mean = 2.4 × 104 CFUs) yield (Fig 1H).

While these iterative improvements allowed the targeted insertion in seven strains from the clade A and “E” (a newly observed clade, unpublished observation), the other four strains from the clade A, C, and D could not be assessed due to their ErmR phenotype (Dataset EV2). Hence, we screened the antibiotic susceptibility of HDD04 and DSM 18205 identifying tetracycline, chloramphenicol, and spectinomycin as additional selective antibiotics (Dataset EV2). Next, P. copri HDD04 was utilized to test the feasibility of tetracycline, chloramphenicol, and spectinomycin resistance genes, i.e., tetQ, catA, aadA, for selection. Conjugation using pEx‐insertion carrying tetQ (pEx‐insertion‐tetQ‐bgl, Appendix Fig S1B), but not catA or aadA, successfully resulted in HDD04 transconjugants after selection with the respective antibiotics. We therefore performed the conjugation for four ErmR but tetracyclin‐sensitive (TetS) strains, i.e., RHA03, HDE04, HDE06, and HDD05, followed by tetracycline selection, resulting in tetracycline resistant (TetR) transconjugants (mean = 8.3 to 1.2 × 104 CFUs; Fig 1I).

In summary, the comprehensive and stepwise adaptation of a gene insertion system to a genetically inaccessible bacterium, i.e., P. copri, illustrates the significant influence of multiple variables for a successful production of genetic mutants, thereby providing a valuable template for initiating the construction of genetic tools for other commensals in future. These experiments together demonstrate the feasibility of gene insertion in P. copri strains from distinct clades enabling functional studies.

Genetic inactivation of PUL regulators to identify their polysaccharide substrates

The prevalence and relative abundance of P. copri has been linked to plant‐rich diets in humans and mouse models (De Filippo et al, 2010; Wu et al, 2011; Ruengsomwong et al, 2016; Fragiadakis et al, 2019; Kovatcheva‐Datchary et al, 2019; Gálvez et al, 2020), yet the underlying molecular mechanism is still poorly understood. To demonstrate the utility of our gene insertion system in identifying gene functions, we decided to investigate the genetic basis for utilization of distinct carbohydrates in P. copri. Since no available chemically defined cultivation system for P. copri exist thus far, we modified the minimal medium (MM) (Martens et al, 2008) originally used for cultivation of B. thetaiotaomicron by supplementing additional defined nutrients (see Materials and Methods), enabling the growth of all P. copri strains tested from our strain collection (n = 12) with glucose as a sole carbon source (Dataset EV3). Specifically, 10 out of 12 strains reached a maximal optical density (OD600 max) of 0.6–1.0 in MM + Glucose overnight, while two strains (HDA03 and RHA03) showed only moderate growth (approximately OD600 max 0.3). This minimal medium was then used to extensively characterize polysaccharide utilization in HDD04, as it showed robust growth in MM and high number of transconjugants. HDD04 grew on various plant cell wall glycans, such as arabinan, arabinogalactan, and arabinoxylan (Dataset EV3). Beyond the utilization of plant cell wall glycans, HDD04 also showed growth on plant and animal cell storage carbohydrates such as inulin (0.651 ± 0.016) and glycogen (0.818 ± 0.026), but grew poorly on levan (0.122 ± 0.009), and could not grow on starch. In parallel, PULs were identified in P. copri using a bioinformatic approach PULpy (preprint: Stewart et al, 2018) followed by manual curation. Specifically, the PUL repertoire of HDD04 was predicted based on susC/D‐like pairs resulting in 29 PULs in comparison with 19 PULs in the P. copri reference strain (DSM 18205; Dataset EV3), suggesting a much broader carbohydrate utilization capability of HDD04 compared to DSM 18205. CAZyme content in the PULs and their substrate were predicted using bioinformatic approaches (dbCAN2 tool and dbCAN‐PUL) (Zhang et al, 2018; Ausland et al, 2021) (Fig EV2).

Figure EV2

Selectivity of sacB‐based gene deletion system in Prevotella copri

Survival of Escherichia coli DH5α, P. copri HDD04 and DSM 18205 on agar plates supplemented with an increasing concentration of sucrose (A) or glucose (B). Serial dilutions of fresh E. coli β2155, P. copri HDD04 and DSM 18205 cultures were plated on YT or BHI + S agar plates with different concentrations of sucrose or glucose for counting CFUs. Growth rates were calculated by normalizing the CFU on YT + Sucrose/Glucose relative to BHI + S. ND: not detectable.

Survival of P. copri strains grown on YT agar plates supplemented with 5% sucrose. Growth rates were calculated as described in (A). The gray color represents the inability of five strains in growing on YT + 5% sucrose.

Viabilities of pEx‐deletion integrated P. copri strains on agar plates with indicated supplements.

Detection of gene deletion of the htcs Inu homologs in P. copri HDB01 and DSM 18205. The up‐ and down‐stream regions (dark gray) of the targeted gene and locations of primers (“OUT” and “IN”, arrowheads) for PCR are indicated.

Data information: In (A–D), data shown are the mean of at least three biological replicates ± SD. Statistical significance between groups was calculated by Student's t test (***P < 0.001; and ****P < 0.0001; NS, P > 0.05, not statically significant).

To directly link distinct PULs and growth phenotypes on polysaccharides, we focused on HTCS genes, the typical activator associated with PULs (Sonnenburg et al, 2006; Martens et al, 2011). Genome‐wide screening for HTCS genes in HDD04 by homology search using the known domains of HTCS (Terrapon et al, 2015) combined with a protein BLAST on National Center for Biotechnology Information (NCBI) identified 10 gene candidates as our targets. We associated nine out of 10 HTCS gene candidates with their closest predicted PULs (e.g. HDD04_00018 is named as htcs‐PUL3) with only HDD04_0019 being a solitary HTCS gene.

Ten HTCS insertion mutants were generated by integrating a modified pEx‐insertion‐ermG plasmid (Appendix Fig S1C) into the coding sequences of their periplasmic sensor domains followed by a screening for growth defects in MM plus polysaccharides to identify their respective substrates (Fig 2A). Of note, to block potential effects of transcriptional and translational readthrough for the HTCS genes after plasmid integration, pEx‐insertion‐ermG was modified to include T1–T2 terminators and TAA encoding stop codon in front of the cloned homology arm, respectively (pEx‐insertion‐ermG‐T1T2; Fig 2A and Appendix Fig S1C). A strain with plasmid integrated into an intergenic region (between HDD04_00165 and 00166) was utilized as a positive control. The polysaccharides (n = 14) and glucose that can support the growth of HDD04 to an OD600 max of > 0.2 within 120 h were investigated (Fig 2B). Compared to the control strain, 6/10 HTCS gene mutants displayed similar growth patterns, e.g., htcs‐PUL3, as shown in Fig 2B, demonstrating that they are not essential for growth on the tested polysaccharides. Strikingly, the other four HTCS mutants each showed dramatic growth defects (OD600 max < 0.1) on only one specific polysaccharide (Figs 2C and D). Specifically, gene disruptions of htcs‐PUL14, htcs‐PUL21, htcs‐PUL24, and htcs‐PUL26 abolished the capacities of HDD04 grown on arabinan, pectic galactan, arabinoxylan, and inulin, respectively (Fig 2C and D). To link the HTCS gene and the nearest PUL for easy identification, we have tentatively designated these genes as htcs D_Ara (htcs‐PUL14, HDD04_02372), htcs D_PecGal (htcs‐PUL21, HDD04_02939), htcs D_AraXyl (htcs‐PUL24, HDD04_03129), and htcs D_Inu (htcs‐PUL26, HDD04_03217).

Figure 2

Identification of HTCS and associated PULs essential for utilization of polysaccharides using targeted gene inactivation

Data information: Error bars represent the standard error of the means (SEMs) in (B) and SDs in (C) and (D) of the biological replicates from three carbohydrate arrays with each carbohydrate tested in duplicate, respectively.

Schematic illustration for gene inactivation strategy targeting HTCS gene candidates in Prevotella copri HDD04. A classical Reg_prop HTCS (Martens et al, 2011) harboring core domains is shown as an example. Partial coding sequence of the periplasmic sensor domain was cloned into the pEx‐insertion‐ermG‐T1T2 vector as the homolog arm for genetic insertion of the HTCS gene. This vector contains T1–T2 terminators (red ball‐and‐stick symbol) and TAA encoding stop codon (asterisk) for avoiding possible transcriptional and translational readthrough for the targeted gene after plasmid integration.

Growth of P. copri HDD04 strains with plasmid insertions at an intergenic region (control) and two representative putative HTCS genes, respectively, in minimal media (MM) supplemented with glucose or indicated polysaccharides.

Growth of P. copri HDD04 strains with respective htcs‐PUL14, htcs‐PUL24, and htcs‐PUL24 insertion compared to an intergenic insertion mutant on arabinan, pectic galactan, and arabinoxylan, respectively.

Maximum growth (OD600 max) of mutant strains (n = 11) in MM supplemented with glucose or indicated polysaccharides.

Together, these experiments not only demonstrate the utility of the gene inactivation strategies to perform functional studies in P. copri, but also uncovered the link between PUL‐associated regulatory genes and metabolic phenotypes on utilizing specific polysaccharides for P. copri.

Construction of an allelic exchange system for validating the function of PUL regulators on polysaccharide degradation

Although our genetic insertion system is efficient in generating mutants for rapid determinations of phenotypes, it some limitations: (i) The selective pressure provided by antibiotics is constantly required in the medium for plasmid integrants with corresponding antibiotic resistance markers; (ii) it would be challenging to target relatively smaller genes because the yield of conjugants is negatively correlated with the homology arm cloned in the conjugative plasmid as shown in Fig 1E; (iii) the integration of the plasmid may cause a polar effect on the expression of down‐stream genes, especially complicating the characterization of each gene's role in an operon. Therefore, we aimed to establish an allelic exchange system for unmarked gene deletion and complementation in P. copri. One of the most widely used system for allelic exchange in bacteria is based on the levansucrase gene (sacB), which catalyzes the hydrolysis of sucrose and synthesizes the toxic compound levan (Gay et al, 1985; Recorbet et al, 1993). Two key criteria have been identified to limit its application: (i) The inability of bacteria to grow properly on agar plates in the presence of relatively high concentrations of sucrose and (ii) whether the expression of sacB can effectively select gene deletion mutants in the presence of sucrose.

Hence, we first determined the growth of P. copri DSM18205 and HDD04 on agar plates with increasing concentrations of sucrose. As media base, we employed yeast extract and tryptone (YT) supplemented with horse blood instead of BHI to reduce salt concentrations, which have been demonstrated to decrease the sucrose sensitivity of E. coli (Blomfield et al, 1991) and have been observed by us to interfere with P. copri growth on high sucrose concentration. Both strains showed the normal colony numbers and morphology until a sucrose concentration of 6%, while E. coli tolerated up to 10% sucrose (Fig EV2A). The inability to grow under these conditions was likely caused by osmotic pressure, as similar results were obtained for P. copri strains with increasing the concentration of glucose in YT + blood agar (Fig EV2B). In order to ensure the selectivity of sucrose without affecting growth of P. copri, a working concentration of 5% sucrose was chosen. Notably, 5/10 strains were not able to grow well in the YT + blood media in the presence of 5% sucrose reflecting their distinct growth properties compared to other strains (Fig EV2C).

Next, a derivative vector of pEx‐insertion‐ermG named pEx‐deletion‐ermG was created by joining the promoter of gdhA gene to a promoterless copy of sacB and inserting it down‐stream of ermG (Fig 3A and Appendix Fig S1D). The homology arm for targeting the bgl gene (Fig 2A) was cloned into the pEx‐deletion‐ermG, resulting in pEx‐deletion‐ermG‐HDD04‐bgl. We individually integrated the pEx‐deletion‐ermG‐bgl and pEx‐insertion‐ermG‐bgl into HDD04 (Fig 3B). ErmR colonies were readily obtained for both plasmids and displayed normal colony morphology as wild type, indicating that the expression of sacB did not affect the growth of P. copri in the absence of sucrose (Fig 3B). Plating these ErmR colonies containing pEx‐deletion‐ermG‐bgl in the presence of sucrose significantly reduced CFU by 104‐fold (Fig 3B). In contrast, the same strain carrying pEx‐insertion‐ermG‐bgl exhibited equivalent viability in the presence and absence of sucrose. These results illustrated that expression of sacB effectively functioned as sucrose‐based selection.

Figure 3

Development of a conjugation‐based gene deletion and complementation platform for Prevotella copri strains

Data information: In (B, D, and E), the data represent the means of three biological replicates ± SDs.

Schematic illustration for allelic exchange using pEx‐deletion‐ermG.

Viability of P. copri HDD04 with integration of pEx‐insertion‐ermG‐bgl or pEx‐deletion‐ermG‐bgl into the chromosome on YT agar plates with indicated supplements. Error bars represent the mean of three biological replicates ± SDs (***P < 0.001).

Generation of P. copri HDD04 ∆htcs Ara and ∆htcs Inu and quantification of growth. Detection of deletion and complementation of two HTCS genes in P. copri HDD04 by PCR using “OUT” and “IN” primer pairs (C) and growth of indicated strains in arabinan or inulin (D), respectively.

Growth of the wild‐type and htcs Inu mutant strains of DSM 18205 and HDB01, respectively.

We subsequently assessed the false positive rate of this counterselection system by plating ErmR on YT+Erm+Suc plates and found that approximately 1 out of 6.7 × 105 cells in the bacterial population were SucR but still ErmR, i.e. carried the plasmid. Of note, 50 from 500 colonies were randomly picked and restreaked on YT + Erm and YT + Suc plates to evaluate whether the SucR phenotype of these “escapers” was due to genetic mutations. Unexpectedly, they all showed ErmR but SucS phenotypes. We further sequenced the sacB gene and its promoter sequences in 10 random‐selected clones, yet none of them had mutations. This suggested that the SucR phenotype of these escapers were attributable to phenotypic but not genetic causes. Similar level of selectivity in the sacB‐sucrose‐based system was recapitulated in other five P. copri strains (Appendix Fig S1D and E, and Fig EV2D). Taken together, these results demonstrate the utility of the sacB‐based counterselection system for targeted allelic exchange in P. copri.

Genetic deletion and complementation demonstrate essential function of HTCS genes on degrading plant polysaccharides

As a proof of concept, we chose two HTCS genes, htcs D_Ara and htcs D_Inu as our targets for genetic deletion and complementation. A schematic for describing the allelic exchange methodology during gene‐editing process including plasmid construction, allelic exchange, and mutant selection is presented in Fig 3A. In brief, we (i) cloned up‐ and down‐stream regions of the target gene into the pEx‐deletion‐ermG plasmid and transferred the plasmid into E. coli β2155; (ii) performed conjugation between P. copri and E. coli carrying the plasmids followed by selection of plasmid integrants (1st recombination); (iii) passaged the ErmR clones without any selection, permitting the spontaneous allelic exchange (2nd recombination); (iv) carried out the selection of bacteria that had lost the plasmid (revertant and deletion mutant); (v) validated the ErmS phenotype of selected clones and screened for the clones with gene deletion mutations. Of note, PCR screening and Sanger sequencing of ErmS clones obtained after the counterselection step revealed that 37.5–56.3% of clones are confirmed deletion mutants (Dataset EV4) with all remaining clones being revertants, showing the precise performance of our targeting system. Following these procedures, htcs D_Ara and htcs D_Inu were successfully deleted in HDD04, generating ∆htcs D_Ara, and ∆htcs D_Inu gene deletion strains accordingly (Fig 3C). We subsequently complemented the deletion mutants with the corresponding HTCS genes, respectively, through a similar genetic procedure except that we cloned the target gene and its flanking regions into pEx‐deletion‐ermG (Fig 3C). In line with our previous findings in Fig 2D, the HTCS‐deficient mutants failed to grow on their previously identified substrates, while complementation of HTCS genes in the mutant strains restored the growth to the levels of the wild‐type HDD04 strain (Fig 3D).

To demonstrate that the allelic exchange system can be also applied to P. copri strains with relatively lower yields of transconjugants, individual deletion of homologous htcs D_Inu genes were performed in the DSM 18205 (htcs DSM_Inu, DSM18205_02724) and HDB01 (htcs B_Inu, HDB01_02906) strains (Fig EV2E). Similarly, deletion of htcs D_Inu displayed dramatic grow defects in both DSM 18205 and HDB01, respectively (Fig 3E).

In conclusion, these results demonstrate the utility of our novel allelic exchange system in the type strain and other P. copri isolates for establishing a causal relationship between genotypes and phenotypes, as best exemplified by specific HTCS genes and the growth phenotypes on arabinan and inulin.

Plant‐derived polysaccharides induced the transcription of distinct PUL‐associated genes in vitro and in vivo

To determine whether polysaccharides induce the expression of specific PULs associated with the identified HTCS genes or rather broader changes in multiple PULs, we performed transcriptome profiling of P. copri HDD04 cultures grown in MM supplemented with either glucose or one of four plant polysaccharides as the sole carbohydrate (Dataset EV5). As we expected, the susC/D‐like elements in PUL14, 21, 24, and 26 exhibited the largest upregulation upon their respective polysaccharide substrates compared to MM + Glucose (Fig 4A), further confirming our genetic characterization (Figs 2 and 3). In line with the previous reports, these PULs also encode the CAZymes that likely primarily targeted the substrates, and were presumably activated via the cognate HTCS regulators in a feedback loop. We therefore defined these four PULs as PULD_Ara, PULD_PecGal, PULD_AraXyl, and PULD_Inu (Fig 4B). Using average fold change of susC/D‐like genes in each PUL as reference, PULD_Inu showed a relatively low (3.3‐fold) upregulation in response to inulin, whereas PULD_Ara, PULD_PecGal, and PULD_AraXyl displayed much higher induction levels (PULD_Ara: 266.1‐fold; PULD_PecGal: 983.7‐fold; PULD_AraXyl: 159.3‐fold; Fig 4A). It is worth noting that the disruption of HTCS gene in PUL24 resulted in the growth deficiency on a simple xylan, i.e., wheat arabinoxylan, where the xylan backbone is the main component with arabinose decorations, but did not affect the growth pattern on the corn xylan, which has extensive side residues forming heterogeneous and polydispersed glycans (Figs 2C and 4A). These data more precisely defined the specificity of PUL24 for arabinoxylan utilization compared to the previous prediction using bioinformatic analysis and growth assays (Fig EV3A) (Fehlner‐Peach et al, 2019). Interestingly, genome‐wide only susC/D‐like genes from PULD_Inu were significantly upregulated, indicating an extremely specific system for processing inulin by P. copri. Yet, multiple susC/D‐like elements were induced (> 10‐fold change) by the other three plant‐derived polysaccharides (Fig 4A). For instance, susC/D‐like pairs in PULD_PecGal (PUL24, 23.9‐fold) were also greatly expressed when cells were exposed to arabinan. The transcriptional response of susC homologs in four identified PULs to each tested polysaccharide were further validated by real‐time quantitative PCR (RT–qPCR; Fig 4C). The upregulation of multiple PULs during growth on pectic glycans, e.g., pectic galactan, have been also observed in Bacteroides spp. due to the complexity of these polysaccharides (Martens et al, 2011; Luis et al, 2018). For instance, PULD_PecGal was likely induced by the degraded components of pectin fragment attached to the main‐chain arabinofuranosyl residues of sugar beet arabinan (Fig EV3B). Surprisingly, PUL15 (HDD04_02377‐87) possesses genes encoding GH13 and GH97, which were previously demonstrated to degrade various glucans (Koropatkin et al, 2008; Cerqueira et al, 2020), was commonly upregulated by arabinan, pectic galactan, and arabinoxylan. Besides the regulation of PUL‐associated genes, there were also genes encoding polysaccharide catabolism enzymes that displayed >10‐fold upregulation by the polysaccharides, e.g. one putative extracellular exo‐alpha‐l‐arabinofuranosidase precursor gene in MM+Arabinan (HDD004_02362, Dataset EV5). Similarly, a putative gene operon was strongly upregulated in response to arabinan and arabinoxylan, which has a high similarity to the arabinose utilization system in B. thetaiotaomicron (Schwalm et al, 2016) (Fig EV3C).

Figure 4

Transcriptional adaptation of Prevotella copri HDD04 to distinct plant polysaccharides and in the human gut

Heatmap showing the induction of susC and susD homologs in each predicted PUL from P. copri HDD04 in MM supplemented with indicated polysaccharides (n = 3) or in a fecal sample (n = 1). Average gene expression of every susC/D element (two or four genes) was calculated and normalized to expression in glucose. The heatmap shows the average log2 fold change of susC/D pairs within the predicted PULs. PUL14 (PULD_Ara), PUL21 (PULD_PecGal), PUL24 (PULD_AraXyl), and PUL26 (PULD_Inu) are highlighted by black frames.

Genetic architectures of PULD_Ara, PULD_PecGal, PUL D_AraXyl, and PULD_Inu, in P. copri HDD04. Genes encoding proteins with predicted functions are displayed by arrows with different colors. Numbers above the gene represent the locus tags based on the sequenced genome of P. copri HDD04. GH families that genes encode are shown based on the bioformatic analysis as described in Dataset EV3 and Materials and Methods.

In vitro transcriptional response of targeted susC homologs in MM+indicated polysaccharide in comparison with MM+Glucose reference.

Figure EV3

Genetic and transcriptomic analysis of carbohydrate utilization by Prevotella copri strains

Gene organizations of homologous PULAraXyl in P. copri HDD04 and DSM 18205 in comparison with the predicted xylan utilization loci in two P. copri strains as described previously (Fehlner‐Peach et al, 2019). Genes in PULAraXyl are annotated as described in Fig 4B.

Graphic structures of arabinan, pectic galactan, arabinoxylan, and inulin used in this study. Different types of monosaccharides are represented according to “Symbol Nomenclature for Glycans” (https://www.ncbi.nlm.nih.gov/glycans/snfg.html). Detailed product information is provided on www.megazyme.com

Gene composition of the putative arabinose utilization operon from HDD04 and expression levels of genes in this operon in MM supplemented with glucose or arabinan. For each dot, the average of transcription levels from three replicates are shown. The gene functions were predicted by BLAST on NCBI.

To identify whether P. copri actively utilizes these PULs in vivo, a metatranscriptome analysis was performed from a stool sample collected from the donor, from which HDD04 was isolated. Strikingly, except PULD_Inu that displayed a slightly increased transcription compared to MM+Glucose (average of fold‐change susC/D‐like genes: 1.2‐fold), the susC/D homologs in the other three identified PULs (PULD_Ara: 4.2‐fold; PULD_PecGal: 215‐fold; PULD_AraXyl: 44.6‐fold) were actively expressed. Moreover, 16 other PULs displayed upregulation from 2.36‐fold (PUL1) to more than 8,000‐fold change (PUL25), which suggests additional substrates from the human diet can be targeted by various PULs in P. copri (Fig 4A). Collectively, these analyses illustrate that P. copri carries out an efficient and diverse polysaccharide processing by orchestrating its associated gene expression profile in vitro and in vivo. Further functional gene studies will be required to understand which polysaccharides can be utilized in vivo by P. copri.

PULD_Ara, PULD_PecGal, PULD_AraXyl, and PULD_Inu are conserved among genetically diverse P. copri strains

Recent studies reported that P. copri isolates exhibited extensive genomic and phenotypic variations (De Filippis et al, 2019; Fehlner‐Peach et al, 2019; Tett et al, 2019). To examine whether utilization of arabinan, pectic galactan, arabinoxylan, and inulin is a common capacity based on the genetic content of P. copri species, we performed a comparative genomic analysis to identify corresponding PULs in strains from our P. copri strain collection. Notably, PULs carrying homologous HTCS/SusC genes compared to HDD04 were found in each of the P. copri strains (Fig 5 and Dataset EV3; Fig EV4). The gene organization and content of these PULs varied from conserved, e.g., PULAraXyl, to variable, e.g., PULInu (Fig EV4). Hence, we next determined the growth in MM supplemented with arabinan, pectic galactan, arabinoxylan, or inulin. Most P. copri strains grew on the tested polysaccharides with the exception of HDA03 that could not grow on pectic galactan and arabinoxylan (Fig EV4). Strikingly, genetic evidence potentially explaining the inability to use specific polysaccharides could be easily identified (Fig EV4). Specifically, natural mutations in the HTCS genes of PULAraXyl, i.e., a truncation of HTCSAraXyl, and of PULPecGal, i.e., deletion, are likely responsible for the “no growth” phenotypes of HDA03 on these polysaccharides. Additionally, the cognate first susC gene of PULAraXyl as well as the SGBP gene of PULPecGal were truncated into two segments in HDA03, which could further contribute to the inability to utilize these polysaccharides.

Figure 5

Gene organizations of two types of arabinan processing PULs in multiple Prevotella copri and two Bacteroides type strains with the capacities of growing on arabinan

Figure EV4

Gene organizations of PULs for processing pectic galactan, arabinoxylan, and inulin among multiple Prevotella copri strains with the capacities of growing on these polysaccharides

Gene organizations of PULPecGal, PULAraXyl, and PULInu homologs among P. copri strains with distinct growth capacities on pectic galactan, arabinoxylan, and inulin, respectively. Maximum growth (OD600 max) of strains in MM supplemented with indicated polysaccharides is reported. Genes clustered in the PULs are displayed by arrows with different colors according to predicted functions of the gene products.

Notably, the PULPecGal in HDE04 and HDD12 and PULInu in HDD05 and HDD12 do not contain susC/D‐like elements, suggesting potential functional complementation between various PULs. These cases of “incomplete” PULs highlight the limitations of using susC/D homologs as genetic markers to predict the growth phenotypes on specific polysaccharides substrates.

Two types of arabinan processing PULs in P. copri display clade‐specific distribution and diet‐dependent expansion in the human gut microbiome

While functionally all tested strains were able to use arabinan, we found that the arabinan processing PULs genetically displayed two distinct structures among the 12 strains (Fig 5). Specifically, two strains from clade A and the single strain from clade E feature an almost identical gene organization with a SGBP‐like gene in front of two pairs of susC/D‐like genes, whereas the remaining strains from clade A, as well as the strains from clade C and D encode a single pair of susC/D‐like genes followed by a SGBP‐like gene. In the following, we refer to two types of PULs as type‐I (single copy) and type‐II (tandem repeat), and the two susC/D‐like pairs in type‐II PULAra as susC1, susD1, susC2, and susD2, respectively. Of note, similar PULAra types have been noticed in Phocaeicola vulgatus (formerly Bacteroides vulgatus) with type I and B. thetaiotaomicron with type II (Fig 5) (Martens et al, 2011; Lynch & Sonnenburg, 2012; Patnode et al, 2019).

Phylogenetic analysis of protein sequences encoded by HTCS genes from the 12 P. copri strains as well as Ph. vulgatus ATCC 8482 and B. thetaiotaomicron VPI‐5482 shows a clade‐driven evolutionary pattern, which closely resembled that of genome‐based tree of the P. copri complex (Figs 1A and 6A). The inability of the mutant strain DSM18205 htcs DSM_Ara to grow on arabinan validates that the HTCSs of two arabinan utilizing systems are functionally conserved in the two types of PULAra (Fig 6B). In contrast, the P. copri‐derived SusC and SusD homologs form three distinct evolutionary branches in the tree that differ from the proteins in B. thetaiotaomicron and Ph. vulgatus, but still show a relative similarity of these proteins corresponding to the PULAra types, respectively (Figs 6A and EV5A). Similar to the susC/D‐like genes, the SGBP‐like proteins are clustered by PULAra type rather than P. copri clade (Fig EV5B). We next performed functional studies complementing the phylogenetic analysis. Because the susC1 gene but not the susC2 gene is required in B. thetaiotaomicron for growth on arabinan as described previously (Luis et al, 2018), the genes encoding susC in type‐I, and susC1 and susC2 in type‐II system of P. copri were individually in‐frame deleted to explore their necessity in P. copri (Fig 6B). In agreement with previous observations in B. thetaiotaomicron, only susC1 but not susC2 is essential for type‐II PULAra carrier HDD04 (Fig 6B). Moreover, deletion of susC in type‐I PULAra abolished the growth capacity of P. copri DSM 18205 on arabinan (Fig 6B). These results indicated that P. copri strains encode highly similar sensory/regulatory systems for sensing arabinan‐derived ligands and transcriptional activation of PULAra, but that PULAra encodes distinct modules, i.e., susC‐susD‐SGBP in type I and SGBP‐susC1‐susD1‐susC2‐susD2 in type II, for carbohydrate binding and importing.

Figure 6

Genetic and phylogenetic analyses of PULAra types in members of the genera Prevotella, Phocaicola and Bacteroides

Phylogenetic trees of HTCS and SusC‐like proteins encoded by two types of arabinan processing PULs in diverse Prevotella copri strains, Bacteroides thetaiotaomicron VPI‐5482, and Phocaeicola vulgatus ATCC 8482. The isolates from distinct P. copri clades are indicated by dots in different colors. The proteins from type‐II PULAra are highlighted in bold.

Growth of P. copri DSM 18205 and HDD04 wild‐type strains and indicated mutants in MM+Arabinan.

Distribution of two types of arabinan processing PULs in the members of genera Prevotella, Bacteroides and Phocaeicola. Total number of genomes for each clade or species group that were analyzed is indicated above the bars.

The association between the two types of PULAra in P. copri and host dietary preference. The relative abundance of identified P. copri MAGs for each individual was grouped based on the presence or absence of two types of PULAra in each dietary habit. Asterisks indicate Wilcoxon U‐test significant differences (*P < 0.05; NS, P > 0.05, not statically significant).

Figure EV5

Phylogenetic analysis of SusD‐like and SGBP‐like proteins encoded by two types of arabinan processing PULs

Phylogenetic trees of SusD‐like (A) and SGBP‐like (B) proteins of two types of arabinan processing PULs from Prevotella copri strains, Bacteroides thetaiotaomicron VPI‐5482, and Phocaeicola vulgatus ATCC 8482. The strains from distinct P. copri clades are labeled by dots with different colors. The proteins from type‐II PULAra are shown in bold.

To gain a broader understanding of the prevalence of PULAra types in P. copri as well as related Prevotella spp., Bacteroides spp., and Phocaeicola spp., PULAra were predicted from 1,602 non‐redundant genomes retrieved from the NCBI genome database (n = 1,504) and from a recent comprehensive metagenomic P. copri survey (n = 98) (Tett et al, 2019). Together with our strains (n = 12), we identified that 499 out of 1,614 genomes encode either type‐I or type‐II PULAra, suggesting that the arabinan utilization potential is frequently found in these genera but not ubiquitous (Fig 6C and Dataset EV6). In agreement with the results from the analysis of our limited strain collection, type‐I PULAra is present in P. copri clades A (55.3%), C (72.7%) and D (75%), while the type‐II is encoded by P. copri strains from the clade A (36.8%) and the single strain from the clade E (HDD12; Fig 6C). Notably, none of the two types as well as the arabinose utilization operon we identified was found in any of the 53 screened genomes of clade B strains (Fig 6C), which is consistent with previous reports that this clade lacks the genes and capacity for arabinose and arabinan utilization (Tett et al, 2019). The two types of PULAra are also widespread in other members of the Prevotella spp., Bacteroides spp., and Phocaeicola spp. (Fig 6C). Notably, our extended analysis shows that while some species displayed either a type I‐ or type II‐dominated distribution, e.g., Ph. vulgatus (n = 90 genomes, type I: 90%, type II: 0%), other species such as Phocaeicola plebeius (n = 20 genomes, type I: 20%, type II: 50%) and B. thetaiotaomicron (n = 42 genomes, type I: 7.14%, type II: 88.1%) displayed both types of arabinan utilization systems similar to P. copri species (Fig 6C). As the evolutionary analysis of Bacteroides spp. and Phocaeicola spp. is limited, these dominations could also result from the clade‐specific distribution as P. copri. Interestingly, only 1 out of 55 Bacteroides ovatus genomes carry type‐I PULAra, which not only suggests the occurrence of horizontal gene transfer of type‐I PULAra between B. ovatus and other type‐I PULAra carriers (Fig 6C), but also provides an explanation for the apparent lack of arabinan utilization in most B. ovatus strains (Martens et al, 2011).

The increased relative abundance of P. copri in the gut microbiota has been associated with fiber‐rich diets (De Filippo et al, 2010; Wu et al, 2011; Ruengsomwong et al, 2016; Fragiadakis et al, 2019). As the in vitro results described above suggest a potential contribution of the PULAra for fitness advantage within the ecosystem, we investigate whether P. copri encoding different types of PULAra displays a diet‐modulated abundance in the human gut. Therefore, we performed a specialized analysis of a publicly available dataset from one recent study comparing the differences in the gut microbiota of individuals consuming omnivore, vegetarian, and vegan diet (De Filippis et al, 2019). In a previous study, we identified five distinct P. copri metagenome‐assembled genomes (MAGs) from four clades in individuals of that cohort (Fig 6D). MAG610 (clade C) encodes a type‐I PULAra, whereas two MAGs 609 and 611 (clade A and C) carry the type‐II counterpart. Two MAGs 612 and 613 (clade C and B) contain neither of the PULs (Fig 6D). The relative abundance of each MAG was then calculated per individual and grouped based on the presence of type‐I or type‐II PULAra in the individuals with three distinct dietary preferences (Fig 6D). While there was no significant difference based on PULAra presence and type in omnivore and vegetarian diets, remarkably, in vegans, PULAra type‐II positive MAGs showed higher abundance than those with type‐I PULAra (Fig 6D). A possible explanation for the expansion of P. copri strains encoding type‐II PULAra in vegans could be distinct capabilities of the two PULAra to degrade a so far unknown subtype of polysaccharide. Moreover, in the case of type‐II PULAra, it is possible that the two distinct SusC/D pairs not only form separate homodimer but also form heterodimers, which may enable type‐II PULAra to access more types of polysaccharides degradation intermediates. Further genetic and functional characterizations will be required to understand the differential precise nature of two PULAra systems. Yet, taken together our comprehensive analyses illustrate the importance of arabinan utilization systems for P. copri fitness in vitro and in vivo.

Discussion

Studying the biology of many human commensals is hindered by diverse obstacles, e.g., challenging cultivation, extensive strain‐level variation, low number of publicly available strains, and the lack of genetic tools (Ley, 2016; De Filippis et al, 2019; Fehlner‐Peach et al, 2019; Tett et al, 2019). In this study, we developed a genetic toolkit that allows versatile genetic manipulations for a wide range of P. copri strains, and applied our genetic platform coupled with genomic, transcriptomic, and phenotypic approaches to provide insights into the genetic basis of polysaccharide utilization of this prevalent gut bacterium. Particularly, the recognition that Prevotella spp. including P. copri are a dominant part of the “non‐westernized” microbiome as well as their unexplained antagonism with Bacteroides spp. has elevated the interest in the members of this genus (Arumugam et al, 2011; Ley, 2016; Johnson et al, 2017; Costea et al, 2018). While a series of genetic tools have been created for the genus Bacteroides (Koropatkin et al, 2008; Goodman et al, 2011; Mimee et al, 2015; Lim et al, 2017; García‐Bayona & Comstock, 2019; Bencivenga‐Barry et al, 2020), none have been reported for P. copri despite its diverse associations to human diseases. Moreover, few functional genetic studies reported on targeting genes in animal‐derived Prevotella, i.e., Prevotella ruminicola (Shoemaker et al, 1991; Gardner et al, 1996; Ogata et al, 1999) and Prevotella bryantii (Accetto et al, 2005; Accetto & Avguštin, 2007). Together, this suggests the existence of limitations preventing the simple transfer of established genetic tools to P. copri.

Here, by redesigning and optimizing the key genetic elements in the conjugative plasmids and experimental procedures, we developed an anaerobic conjugation‐based system, which overcomes the genetic intractability of diverse P. copri strains. For instance, the optimization of promoter strength for the selection marker resulted in a 306.6‐fold increase in conjugation outcomes. Another key factor was the evaluation of different donor and recipient strains improving conjugation outcome by 83.7‐fold and up to 14,160‐fold, respectively. Of note, naturally occurring antibiotic resistances can complicate gene targeting, but the combination of plasmids carrying either ermG or tetQ antibiotic markers allowed the generation of insertion mutants for 11 out of 12 P. copri strains in our strain collection representing multiple distinct clades. At least one of the strains may even represent a distinct species, i.e., the HDD12 strain was annotated as Prevotella hominis, further supporting the utility of our approach to diverse Prevotella strains.

In a proof of concept of our approach, we focused on one strain, i.e., HDD04, with high conjugation capacity and robust growth in chemically defined medium, targeting 10 HTCS regulators for controlling PUL expression, and identified four HTCS genes with essential functions in utilizing plant polysaccharides. Next, we adapted the sacB‐sucrose system to P. copri that enabled the selection of mutants after allelic exchange with a 3.1 × 10−7 to 4.9 × 10−6 false positive rate. This efficacy is similar to the that obtained for a recent system used for allelic exchange in Bacteroides spp. (Bencivenga‐Barry et al, 2020). Of note, compared to other counterselection systems, such as the genes encoding 30S ribosomal protein S12 (rpsL) for Proteobacteria and thymidine kinase (tdk) for Bacteroides spp. (Dean, 1981; Reyrat et al, 1998; Koropatkin et al, 2008), the sacB‐sucrose system does not require any genetic modification of the recipient strain in advance. The utility of the sacB‐sucrose system was demonstrated by deletion of HTCS genes in three strains, including ones with relatively lower conjugation capacity. Despite these advantages, the allelic exchange system for P. copri does have limitations; e.g., it requires multiple selection steps and the ability to tolerate the osmotic pressure of high sucrose concentrations. Other counterselection markers, such as inducible antibacterial effectors (Bte1 and Bfe1) utilized in Bacteroides species (García‐Bayona & Comstock, 2019; Bencivenga‐Barry et al, 2020), can be envisioned, yet it has been noticed that P. copri DSM 18205 was not affected by the Bfe1 effector (Chatzidaki‐Livanis et al, 2016). Another limitation relates to complementation of gene deletions, e.g., in B. thetaiotaomicron, genetic complementation is accomplished by integrating a plasmid carrying the complemented gene into the chromosome (Wang et al, 2000; Koropatkin et al, 2008). Specifically, the integration vector pNBU2 encodes a tyrosine integrase, which mediates sequence‐specific recombination between the attN site of pNBU2 and one of two attBT sites located in the 3′ ends of the two tRNASer genes on the B. thetaiotaomicron chromosome. Because no identical or similar attachment DNA sequences were found in P. copri, we so far carried out the complementation by the same allelic exchange approach as used for gene deletion. Thus, development of a similar integration vector for P. copri or a P. copri parent strain carrying attachment sites for pNBU2 has the potential to simplify the process of genetic complementation.

The high prevalence and increased abundance of P. copri in the intestinal microbiota is frequently associated with consumption of fiber‐rich diets, which has inspired the research for underlying genetic basis of polysaccharide utilization in P. copri. Combinations of comparative genomics and phenotypic assays have predicted the substrates for the PULs harboring well‐defined CAZymes‐coding genes (Fehlner‐Peach et al, 2019). However, the direct contribution of specific PULs for polysaccharide substrates has not been documented for P. copri. Our genetic studies confirmed the previous bioinformatic prediction of arabinan as the substrate for the PUL14 homologs (Fehlner‐Peach et al, 2019), Additionally, three new PUL/polysaccharide combinations were identified in P. copri. It is intriguing that PUL24, whose homologous gene cluster was predicted based on bioinformatic analysis as xylan‐processing PUL, was identified in this study to be specifically essential for wheat arabinoxylan utilization, but non‐essential for complex xylan from corn, suggesting the contribution of other CAZymes out of PUL24 to deconstruction of xylans with complicated structures. Two distinct PULs, PUL‐XylL and PUL‐XylS, in B. ovatus have been previously demonstrated to be induced by different types of xylans, but confer to growth on the simple and complex xylans, respectively. In P. copri HDD04, the CAZymes present in PUL25 (e.g., GH43_7, GH31, and GH3) as well as the ones that were not predicted as PUL components (e.g., HDD04_02869‐70 around PUL20) were also co‐upregulated with PUL24 by arabinoxylan to a lesser degree. These CAZymes are likely responsible for degrading other types of xylans besides wheat arabinoxylan. Collectively, genetic approaches coupled with phenotypic and transcriptome analyses present a framework for a more accurate characterization of PULs. Notably, the gene organization and content of PULs varied between P. copri strains from conserved to variable. For instance, the strictly conserved synteny of the arabinan and arabinoxylan processing PULs suggests that they have been under the positive selection pressure, as discussed below. In contrast, the gene content of the PULs for pectic galactan and inulin are relatively variable, suggesting the non‐essentiality of certain genes for supporting growth of P. copri on these two carbon sources, including even susC/D‐like genes. These observations support the model that susC/D‐like element in some but not all PULs are essential for the uptake of polysaccharides and are therefore suboptimal markers for carbohydrate utilization potential. Finally, P. copri strain carry either one of two types of PULAra, i.e., type‐I with a single susC/D pair or type‐II PULAra with tandem‐repeat susC/D, which appears to be shared phenomenon in the Bacteroides, Phocaeicola, and Prevotella genera. In P. copri, the distribution of the two distinct PUL types showed largely clade‐specific features, i.e., clade A encoded both types of PULAra, clades C and D only encoded type I, and none of identified PULAra was found in clade B (Fehlner‐Peach et al, 2019; Tett et al, 2019). While the HTCS regulators independent of the PULAra type shared high homology between members of the same clade, the susC/D‐like and SGBP‐like genes clustered by PUL type. Notably, in a limited explorative analysis PULAra type‐specific domination was observed in individuals consuming a vegan diet, suggesting the advantage of type‐II over type‐I in utilizing arabinan or potential other arabinose‐based polysaccharides from dietary fibers in the human gut.

In summary, we have demonstrated the versatile capacities of the genetic toolbox for, firstly, generating a series of individual gene insertion mutants for phenotypic screening in parallel; secondly, enabling targeted gene deletion and complementation to establish causal relationship between genotypes and phenotypes; thirdly, determining the impact of homologous genes in distinct P. copri strains on specific polysaccharide utilization. The toolbox will enable the dissection of more sophisticated biological interactions of P. copri with the human hosts during health and disease, such as investigate associations of P. copri to host metabolism in vivo (Kovatcheva‐Datchary et al, 2015; De Vadder et al, 2016; Pedersen et al, 2016). Importantly, the platform was designed using general principles highlighting key technical details that can be modified and applied to other Prevotella species and even prominent bacterial genera from humans and other habitats. Moreover, these principles can be utilized for further development of high‐throughput genetic screening, such as transposon mutagenesis (Goodman et al, 2011) and CRISPRi (Peters et al, 2016), thereby advancing studies into systematically understanding the ecological and metabolic processes of microbiota and their impacts on host health and disease.

Materials and Methods

Preparation of culture media for P. copri

BHI + S liquid medium

The fetal bovine serum was heated at 56°C for 30 min to inactivate complement. 9.25 g of brain heart infusion (BHI) powder was dissolved in 225 ml double‐distilled water (ddH2O) in 500‐ml glass bottle. The medium was supplemented with 10% fetal bovine serum (FBS), and placed on the hotplate stirrer with 250°C for 20 min. The heated medium was then cooled down to room temperature, supplemented with 1 µg/ml vitamin K3, and filter‐sterilized using a filter unit (0.22 mm pore diameter).

Minimal medium

Minimal medium (MM) was prepared on the basis of a minimal media supporting the growth of various Bacteroides spp. strains (Martens et al, 2008). It contained 100 mM KH2PO4 (pH 7.2), 15 mM NaCl, 8.5 mM (NH4)2SO4, 100 μM MgCl2, 1.4 μM FeSO4·7H2O, 50 μM CaCl2, 1.9 μM hematin, 1 mg/l vitamin K3, 5 μg/l vitamin B12, 0.5 g/l cysteine. We modified this media to contain additionally 10 ml/l amino acid mix solution (250 mg each of l‐alanine, l‐arginine, l‐asparagine, l‐aspartic acid, l‐cysteine, l‐glutamic acid, l‐glutamine, glycine, l‐histidine, l‐isoleucine, l‐leucine, l‐methionine, l‐phenylalanine, l‐proline, l‐serine, l‐threonine, l‐tryptophan, l‐tyrosine, l‐valine, and 312 mg of l‐lysine monohydrochloride into 1 l ddH2O), 10 ml/l purine/pyrimidine solution (200 mg each of adenine, guanine, thymine, cytosine, and uracil into 1 l ddH2O, pH 7.0), 10 ml/l ATCC Vitamin Mix, 10 ml/l ATCC Trace Mineral Mix, and 1 μl/l vitamin K1 solution were added. The commercial carbohydrates were prepared as described previously (Martens et al, 2011). Briefly, 10 g/l carbohydrate stock solutions (2× concentration) were sterilized by filtering (for glucose) or autoclaving at 121°C for 15 min (for complex polysaccharides, Dataset EV7). When needed, 10 g/l carbohydrate solutions were added into 2× MM at a volume ratio of 1: 1.

YT + S agar

Sucrose was dissolved in ddH2O at 0.5 g/ml (50%) as a stock solution. 5 g yeast extract and 10 g tryptone were dissolved in 900 ml ddH2O. The resulting medium (YT) was autoclaved, cooled down to 50°C, and supplemented with 5% sucrose (100 ml 50% sucrose stock solution) for counter‐selection for gene deletion and complementation of P. copri strains. When necessary, the different volumes of sucrose solution were added in media.

Bacterial culture conditions

All strains, plasmids, and primers used are listed in Dataset EV1. Escherichia coli strains were grown aerobically at 37°C on Luria‐Bertani (LB) media. Escherichia coli β2155 were specially cultured on LB supplemented with 0.3 mM 2,6‐Diaminopimelic acid (DAP) (Demarre et al, 2005). Prevotella copri was cultured in BHI+S liquid media, minimal media plus a carbon source, on BHI agar supplemented with 5% defibrinated horse blood or on YT agar supplemented with 5% defibrinated horse blood and 5% sucrose unless otherwise specified. Cultures were routinely grown and manipulated in an anaerobic chamber (Coy Laboratory Products) with an atmosphere of 20% CO2, 10% H2, and 70% N2 at 37°C. When necessary, antibiotics were added to the medium as follows: 7 µg/ml vancomycin, 100 µg/ml ampicillin, 200 µg/ml gentamicin, 20 µg/ml erythromycin for selecting P. copri DSM 18205 derived plasmid integrants; 5 µg/ml for other P. copri strains, and 20 µg/ml tetracycline for RHA03 and HDE06; 2.5 µg/ml for HDD04, HDE04 and HDD05.

Isolation of P. copri from humans

Prevotella copri was isolated from fecal samples of P. copri‐positive donors previously determined by 16S rRNA sequencing. Briefly, the fresh fecal samples were collected and further processed in an anaerobic chamber. A pea‐sized fecal pellet was resuspended in 5 ml BHI + S and filtered through 70 µm cell strainer. We performed a serial 10‐fold dilution of the flow‐through, and streaked out the diluted samples with the dilution factors of 10−3 to 10−6 on BHI blood agar plates supplemented with vancomycin. The plates were then incubated anaerobically at 37°C for 48–72 h. Individual colonies were picked into BHI+S broth and the resulting cultures were screened by PCR for P. copri‐positive cultures using P. copri‐specific primers (P_copri_69F/P_copri_853R). The pure P. copri isolates were obtained by steaking out the P. copri‐positive cultures above, and confirmed by Sanger sequencing using the 16S rRNA gene‐specific primers as described previously (16S_27F/16S_1492R) (Miller et al, 2013). The fresh culture of P. copri was mixed with an equal volume of 50% glycerol in BHI medium in sealed glass vials as bacterial glycerol stocks, and cryopreserved at −80°C immediately.

DNA extraction from human feces and P. copri cultures

The DNA extraction from fecal samples of P. copri‐positive donors or P. copri strains cultured in BHI+S broth (OD600 = 0.6) was performed using ZymoBIOMICS DNA Miniprep Kit based on the instruction manual. We measured the concentration of purified DNA samples by Qubit Fluorometer (Thermo Scientific), and analyzed by agarose gel electrophoresis, NanoDropTM 2000 (Thermo Scientific), and Bioanalyzer (Agilent Technologies).

Whole‐genome sequencing, assembly, and annotation

The DNA library for genome sequencing of P. copri strains was performed using NEBNext® Ultra™ II FS DNA Library Prep Kit (New England Biolabs) for Illumina with parameters as followed: 500 ng input DNA and 5 min at 37°C for fragmentation; > 550‐bp DNA fragments for size selection; primers from NEBNext Multiplex Oligos for Illumina Kit (New England Biolabs) for barcoding. The library was sequenced on the Illumina Miseq 2 × 250 bp The obtained reads were thus assembled with SPAdes version v3.10.0 using “careful” mode (Bankevich et al, 2012). Short contigs were then filtered by length and coverage (contigs > 500 bp and coverage > 5×). Gene prediction and annotation was performed using PROKKA version v1.13.3 (Seemann, 2014) with default parameters. The locus IDs of annotated genes in this study and the NCBI database are shown in Dataset EV8.

Phylogenetic analyses

Placement of P. copri complex. The phylogenomic analyses were conducted as previously described on the characterization of the P. copri Complex (Tett et al, 2019) using PhyloPhlAn3 (Asnicar et al, 2020) with reference set of P. copri strains (Tett et al, 2019) and the newly P. copri strains isolated in this study. The phylogenetic analysis in Fig 1A was built using the 400 universal marker genes of the PhyloPhlAn database using the parameters “‐‐diversity low”, and “‐‐accurate” option. The configuration file (config_file.cfg) was set with the following tools and parameters:

Diamond version v0.9.9.110 (Buchfink et al, 2015) with “Blastx” for the nucleotide‐based mapping, “Blastp” for the amino acid‐based mapping, and “‐‐more‐sensitive ‐‐id 50 ‐‐max‐hsps 35 ‐k 0” in both cases. MAFFT version v7.310 (Katoh & Standley, 2013), with “‐‐localpair ‐‐maxiterate 1,000 ‐‐anysymbol ‐‐auto” options. trimAl version 1.2rev59 (Capella‐Gutiérrez et al, 2009), with “‐gappyout” option. IQ‐TREE multicore version v1.6.9 (Nguyen et al, 2015), with “‐nt AUTO ‐m LG” options. RAxML version 8.1.15 (Stamatakis, 2014), with “‐p 1989 ‐m GTRCAT ‐t” options.

For Figs 6 and EV5, the amino acid sequences of HTCS, SusC‐like, SusD‐like, and SGBP‐like proteins encoded by arabinan processing PULs (PULAra) from B. thetaiotaomicron VPI‐5482, Ph. vulgatus ATCC 8482 and 12 P. copri strains were used for the phylogenetic trees via the MEGA‐X software, respectively. The evolutionary history was inferred using the Neighbor‐Joining method (Saitou & Nei, 1987) with 1,000 bootstrap replicates. The two types of PULAra extracted from all recovered assemblies from the genus Prevotella (n = 8), Bacteroides (n = 13) and Phocaeicola (n = 3) was calculated for the analysis of phylogenetic distributions, respectively.

Determination of P. copri sensitivity to oxygen

Bacteroides thetaiotaomicron VPI‐5482, P. copri HDD04, HDB01, and DSM 18205 were grown in BHI+S broth anaerobically. The fresh bacterial cultures were divided into 1 ml aliquots into 2 ml tubes with the caps being open, respectively. These aliquots were aerobically incubated at 37°C. At four time points (0, 1, 2, 4 h), three aliquots from respective cultures were placed back to the anaerobic chamber and performed serial dilutions for counting CFUs.

Determination of P. copri sensitivity to antibiotics

The wells of a non‐tissue culture flat bottom 96‐well were loaded with 198 µl BHI+S media in the presence of 2‐fold serial dilutions of erythromycin, tetracycline, chloramphenicol, spectinomycin, apramycin, and hygromycin ranging from 0.04 to 400 µg/ml. BHI+S media without antibiotics was loaded as a positive control. Prevotella copri strains were grown in BHI+S broth to an optical density (OD600) of 0.5–0.7. 2 µl bacterial culture was then inoculated into each well (inoculation ratio of 1:100). Absorbance at OD600 of each well was measured at an interval of 1 h for 5 days using the microplate reader (BioTek). Assays were performed in triplicate. To ensure that the concentrations of erythromycin and tetracycline used for selecting transconjugants of different P. copri strains were sufficient for killing all the wild‐type P. copri cells, 1 ml of fresh P. copri culture (108–109 bacterial cells) was plated on the BHI blood agar with respective concentration of antibiotics in triplicate.

Prediction of HTCS genes in P. copri

The genes of P. copri HDD04 that encodes proteins containing all the domains of HTCS (PF07494‐PF07495‐PF00512‐PF02518‐PF00072‐PF12833) according to the Pfam classification were identified as described previously (Terrapon et al, 2015). The hmmsearch was carried out using default parameters (Eddy, 1998).

Molecular cloning

The relevant primers and plasmids are described in Dataset EV1. PCR amplification for cloning was carried out using Q5 High Fidelity DNA Polymerase (New England Biolabs). The PCR products were purified, and followed by DNA assembling with PCR amplified plasmid using Gibson reaction (HiFi DNA Assembly Master Mix, New England Biolabs). The assembled products were transformed into E. coli β2155 by chemical transformation. The resulting colonies were randomly picked to detect the inserts and their sizes by colony PCR using OneTaq DNA Polymerase (New England Biolabs). Genetic modifications generated on plasmids were verified by sequencing at Microsynth Seqlab (Microsynth AG, Germany).

Specifically, the pEx‐insertion vector was constructed as follows: Firstly, the thymidine kinase gene (tdk) and its promoter was deleted from the multiple cloning site using DNA assembling with PCR amplified plasmid, resulting in pExchange. Secondly, 300 bp of the tuf (elongation factor Tu, DSM18205_02600) promoter was inserted into pExchange exactly before the coding sequence of ermG, generating the pEx‐insertion‐ermG vector. For first trial of genetic insertion in P. copri, 3‐kb DNA sequences from DSM18205_00642‐43, 00941‐42, and 02334‐35 were cloned into the multiple cloning site of pEx‐insertion‐ermG as the homology arm for plasmid integration. The pEx‐insertion‐ermG with the DNA region from DSM_02334‐35 (DSM_02334: putative β‐glycoside hydrolase, bgl) was designated as pEx‐insertion‐ermG‐DSM‐bgl.

Based on pEx‐insertion‐ermG as the vector backbone, similar cloning procedures were performed for constructing various plasmids carrying: (i) different promoter sequences for driving selective marker; (ii) different sizes of homology arms varying from 0.5 kb to 4 kb for integration in P. copri HDD04; (iii) 3‐kb cloned homologous regions from different P. copri strains in our collection; (iv) a pLGB30‐derived tetQ selective marker (García‐Bayona & Comstock, 2019) instead of ermG; (v) the T1‐T2 terminators copied from pSAM (Goodman et al, 2011) for blocking the transcriptional readthrough for the HTCS genes after plasmid integration.

The pEx‐deletion‐ermG and pEx‐deletion‐tetQ vectors were created by inserting the counter‐selection marker following ermG and tetQ, respectively. The counter‐selection marker is a DNA fragment generated by splicing 300‐bp of the gdhA (HDD04_01507) promoter and the sacB gene from the pEX18Ap plasmid (Hoang et al, 1998). The pEx‐deletion‐ermG‐bgl was similarly generated as described above. For in‐frame deletion of genes in P. copri, the approximately 2‐kb regions flanking the target gene were amplified, and assembled with PCR amplified pEx‐deletion‐ermG. For gene complementation, the target gene flanking with approximately 1‐kb up‐ and down‐stream regions were entirely amplified, and cloned into pEx‐deletion‐ermG.

Genetic manipulations of P. copri

Overnight culture of the E. coli donor strain was subcultured into LB medium containing ampicillin and DAP and P. copri subcultured into BHI+S medium. When they were grown to exponential phase (OD600 = 0.5–0.7), E. coli culture was transferred into the anaerobic chamber. The following procedures of genetic manipulations for P. copri including plasmid insertion, in‐frame deletion, and complementation were performed in the anaerobic chamber. For conjugation, 1 ml E. coli culture (˜ 5 × 109 CFUs) was centrifuged at 8,000 g for 3 min to pellet the bacterial cells, followed by resuspension in 100 µl fresh P. copri culture (˜ 5 × 107 CFUs) to get a ratio of donor:recipient of 100:1. Specially, if P. copri HDA04 or HDD12 culture was used as the recipient strain for conjugation, to obtain the same donor/recipient ratio above, 20 µl HDA04 culture plus 80 µl BHI + S medium or 10 µl HDD12 culture plus 90 µl BHI + S medium was used to resuspend the E. coli pellet, respectively. The resuspension was then plated on a BHI blood agar with DAP for 18 h at 37°C for bacterial conjugation unless otherwise stated. Bacterial cells were washed off from the plate using 1 ml BHI + S medium, mixed well, and plated serial dilutions or the whole bacterial pellet after centrifugation on BHI blood agar plates containing gentamicin in addition of erythromycin or tetracycline. Colonies generated from transconjugants were visible after incubation of plates for 2–4 days according to properties of the P. copri derivatives. If necessary, the CFUs were counted for quantification of transconjugant yields. Insertion of the plasmid was verified by amplifying two joints between the bacterial chromosome and vector via colony PCR using P3/P4 and P5/P6 primer pairs, with P1/P2 amplified DNA as a control.

For in‐frame deletion and complementation, the insertion mutants were grown in liquid BHI+S without selection and then subcultured every 12 h for allelic exchange. The final culture was plated onto YT agar plates supplemented with 5% sucrose to select the revertants (wild type) and gene deletion mutants with loss of the vector. After incubation of plates for 2–4 days, individual colonies were restreaked onto BHI blood plates in the presence and absence of erythromycin using the same inoculating loop, respectively, to further confirm erythromycin sensitivity of the clones. Erythromycin‐sensitive clones were subsequently screened for the genetic modifications (gene deletion or complementation) by PCR and verified by sequencing at Microsynth Seqlab (Germany).

Prediction of PULs in P. copri genomes

The prediction of PULs in P. copri genomes and MAGs was described previously (Gálvez et al, 2020). Briefly, PULs and susC/D‐like gene annotations were carried out using PULpy (preprint: Stewart et al, 2018) (commit 8955cdb, https://github.com/WatsonLab/PULpy). Annotation of carbohydrate‐active enzymes (CAZymes) surrounding the susC/susD‐like pairs was performed using dbCAN2 tool (Zhang et al, 2018) version v2.0.6 (CAZy‐DB=07312019, https://github.com/linnabrown/run_dbcan). Putative substrates of these CAZymes were predicted by dbCAN‐PUL as described previously (Ausland et al, 2021).

Measurement of P. copri growth on a carbohydrate array

The growth curves of P. copri strains cultured in minimal medium (MM) supplemented with a sole carbohydrate were measured as previously described with the following modifications (Martens et al, 2011). The wells of a non‐tissue culture flat bottom 96‐well were loaded with 100 µl sterilized carbohydrate stocks (2× concentration). Each carbohydrate was added into at least three wells. Prevotella copri was grown in MM + Glucose to an OD600 value of approximately 0.6. 400 µl culture was then centrifuged to pellet the bacterial cells. The pellet was washed by 1 ml 2× MM without any carbohydrates and resuspended in 10 ml 2× MM as a seed culture. Each well of the plate was loaded with 100 µl seed culture. Absorbance at OD600 of each well was measured for 5 days by the microplate reader (BioTek) at 1‐h intervals with 15‐s pre‐shaking.

In Figs 2D and EV4, and Dataset EV3, the maximal OD600 values subtracting the background reads (OD600 max) in the curves were identified for calculating means and standard deviations (SDs). Because we observed that the presence of erythromycin in MM significantly affect the duration of lag phase, but not the growth pattern of P. copri. The growth curves of HTCS gene insertion mutants and relevant intergenic insertion control were therefore shown starting from the OD600 values increased by 10% of the OD600 max to the OD600 max in Fig 2B and C.

RNA extraction from human feces and metatranscriptome sequencing

The fecal sample from the human donor carrying P. copri HDD04 was immediately collected into DNA/RNA Shield Fecal Collections Tubes and stored at 4°C for stabilizing RNA. An aliquot of 400 µl content from the tube was used for isolating RNA using ZymoBIOMICS RNA Miniprep Kit following the instruction manual.

RNA extraction and RNA‐seq library preparation

Prevotella copri HDD04 was grown in BHI+S broth to the exponential phase (OD600 = 0.6). 5 ml fresh cultures were treated by RNAprotect (New England Biolabs) based on the manufacturer's instructions, pelleted by centrifugation, and stored at −80°C until further processing. The bacterial RNA was isolated using ZymoBIOMICS RNA Miniprep Kit following the instruction manual. RNA quality was evaluated by agarose gel electrophoresis, NanoDropTM 2000 (Thermo Scientific), and Bioanalyzer (Agilent Technologies) according to RNA integrity score (RIN > 8.0). Bacterial ribosomal RNA (rRNA) was then depleted by Ribo‐Zero Gold rRNA Removal Kit (Epidemiology) as described in the commercial protocol. Libraries for Illumina sequencing were prepared using the NEBNext® Ultra™ Directional RNA Library Prep Kit for Illumina® (New England Biolabs) following manufacturer's protocol. For each sample, 100 ng of fragmented mRNA was used as an input for cDNA synthesis and Illumina sequencing adaptor ligation.

For other treatment in Fig 4, P. copri HDD04 was initially grown in minimal media plus glucose (MM + Glucose) to an OD600 value of 0.5. 2 ml culture was then centrifuged to pellet the bacterial cells and resuspended using an equal volume of minimal media without carbohydrates (MM), followed by another centrifugation and resuspending in the same volume of MM. 40 µl suspension was inoculated into 4 ml MM plus glucose, arabinan, arabinoxylan, pectic galactan, and inulin, respectively. Three replicates were performed for each carbon source. Once P. copri grown to OD600 = 0.5, 750 µl culture was taken and treated by RNAprotect (New England Biolabs). Bacterial ribosomal RNA (rRNA) was thus depleted using Pan‐prokaryote riboPOOL™ Kit (siTOOLs Biotech) as described in the manual. The cDNA library preparation and sequencing was carried out as described above.

RNA‐seq analysis

Reads were quality filtered using Trimmomatic (Bolger et al, 2014) version v0.33 with as follow parameters (LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15 MINLEN:35 HEADCROP:3). After quality control reads were aligned to each P. copri reference genome using STAR (Dobin et al, 2013) version v2.5.2a. Reads count was performed using HTSeq (Anders et al, 2015) version v0.11.2. With the aim to control for interspecies multi‐mapping, reads were split by mapping to multiple references using (BBsplit). References genomes were selected from the reconstructed MAGs and one representative strain for each of the P. copri clades in combination with each donor's isolate from the clade A.

For in vivo and in vitro differential gene expression, gene read counts were transformed using TPMs normalization and differential gene expression was quantified in R using the DEseq analysis with a single replicate (iDEG) package (Li et al, 2019).

For the transcriptome in vitro with supplemented polysaccharides, samples were proceeded as described above. Normalization and differential expression were quantified in R using the DEseq2 package (Love et al, 2014) version v1.26.0 using the samples grown in MM + glucose as a control.

Measurements of gene transcription by RT–qPCR

The preparation of P. copri cultures and extraction of total RNA were performed as described above. Reverse transcription was carried out with ProtoScript® II First Strand cDNA Synthesis Kit (New England Biolabs) using Random Primer Mix and 800 ng purified RNA as template for 20 µl reaction. The abundance of transcript for target susC‐like genes and reference tuf gene was quantified with KAPA SYBR® FAST qPCR mix (KAPA Biosystems) using 0.5 ng/µl template cDNA, 25 nM of each target gene‐specific primer. The reaction was performed in a 96‐well plate on the Roche Lightcycler 480. Using the ddCT method, raw values were normalized to values for the tuf gene and then the fold change was calculated by dividing MM+specific polysaccharide values by values obtained from MM+glucose.

Reconstruction of P. copri MAGs

The reconstruction of P. copri MAGs from a recent dataset (De Filippis et al, 2019) was performed as described previously (Gálvez et al, 2020). In brief, the sequencing data of the gut microbiome from 101 healthy Italian individuals with distinct diets (Omnivore, n = 25; Vegetarian, n = 39; Vegan, n = 37; NCBI SRA: SRP126540 and SRP083099) were analyzed as follows: (i) Sample‐wise assembly, annotation, and integrative genomic binning was carried out with ATLAS metagenomic workflow (Kieser et al, 2020) (commit a007857, https://github.com/metagenome‐atlas/atlas); (ii) genome abundance estimates were calculated for each sample by mapping the reads to the non‐redundant MAGs using BBmap and determining the median coverage across each of the MAGs.

Quantification and statistical analysis

Statistical analysis

Statistical analyses were carried out in R (R Core Team, 2019) and figures were produced using 690 the package ggplot2 (Wickham, 2016). Datasets were analyzed using the GraphPad Prism 8. Pairwise comparisons were performed using Student's t test with a paired, two‐tailed distribution. More statistical details are indicated in the associated figure legends when required.

Author contributions

JL and TS designed the experiments and wrote the paper. JL conducted the most of experiments and data analysis. LA and AI performed the strain isolation. JL, EJCG, and TRL performed the bioinformatic analysis. LA and EA assisted in growth assays. LA and AAB assisted in RNA preparation for RNA‐seq. EA and E‐MS assisted in molecular cloning and genetic manipulation.

Conflict of interest

The authors declare that they have no conflict of interest.

Appendix

Click here for additional data file.

Expanded View Figures PDF

Click here for additional data file.

Dataset EV1

Click here for additional data file.

Dataset EV2

Click here for additional data file.

Dataset EV3

Click here for additional data file.

Dataset EV4

Click here for additional data file.

Dataset EV5

Click here for additional data file.

Dataset EV6

Click here for additional data file.

Dataset EV7

Click here for additional data file.

Dataset EV8

Click here for additional data file.