Synbiotic Matchmaking in Lactobacillus plantarum: Substrate Screening and Gene-Trait Matching To Characterize Strain-Specific Carbohydrate Utilization

Synbiotics combine prebiotics and probiotics to synergistically enhance the health benefits associated with these ingredients. Lactobacillus plantarum is encountered as a natural inhabitant of the gastrointestinal tract, and specific strains are marketed as probiotics based on their strain-specific health-promoting activities. Strain-specific stimulation of growth through prebiotic substrates could enhance the persistence and/or activity of L. plantarum in situ. Our study establishes a high-throughput screening model for prebiotic substrate utilization by individual strains of bacteria, which can be readily employed for synbiotic matchmaking approaches that aim to enhance the intestinal delivery of probiotics through strain-specific, selective growth stimulation.

ABSTRACT

Synbiotics are food supplements that combine probiotics and prebiotics to synergistically elicit a health effect in humans. Lactobacillus plantarum exhibits remarkable genetic and phenotypic diversity, in particular in strain-specific carbohydrate utilization capacities, and several strains are marketed as probiotics. We have screened 77 L. plantarum strains for their abilities to utilize specific prebiotic fibers, revealing variable and strain-specific growth efficiencies on isomalto- and galactooligosaccharides. We identified a single strain within the screening panel that was able to effectively utilize inulin and fructooligosaccharides (FOS), which did not support efficient growth of the rest of the strains. In the panel we tested, we did not find strains that could utilize arabinoxylooligosaccharides or sulfated fucoidan. The strain-specific growth phenotype on isomaltooligosaccharides was further analyzed using high-performance anion-exchange chromatography, which revealed distinct substrate utilization phenotypes within the strain panel. The strain-specific phenotypes could be linked to the strains’ genotypes by identifying gene clusters coding for carbohydrate membrane transport systems that are predicted to be involved in the utilization of isomaltose and other (unidentified) oligosaccharides in the isomaltooligosaccharide substrate.

IMPORTANCE Synbiotics combine prebiotics and probiotics to synergistically enhance the health benefits associated with these ingredients. Lactobacillus plantarum is encountered as a natural inhabitant of the gastrointestinal tract, and specific strains are marketed as probiotics based on their strain-specific health-promoting activities. Strain-specific stimulation of growth through prebiotic substrates could enhance the persistence and/or activity of L. plantarum in situ. Our study establishes a high-throughput screening model for prebiotic substrate utilization by individual strains of bacteria, which can be readily employed for synbiotic matchmaking approaches that aim to enhance the intestinal delivery of probiotics through strain-specific, selective growth stimulation.

INTRODUCTION

Probiotics were defined by the FAO and WHO in 2001 as “live microorganisms which when administered in adequate amounts confer a health benefit to the host” (1). This represents the most widely accepted definition worldwide. The correct usage of the term and the scope of probiotics have been reviewed with close scrutiny, and more-nuanced guidelines have been proposed, although the original definition offered by the FAO and WHO is still regarded as sufficiently broad to enable a wide range of products while maintaining narrow core requirements (2). The term “prebiotic” was originally defined in 1995 as “a non-digestible food ingredient that beneficially affects the host by selectively stimulating the growth and/or activity of one or a limited number of bacteria in the colon, and thus improves health” (3), but consensus definitions have changed over the years (4, 5). The definition that is commonly used at present defines a prebiotic as “a substrate that is selectively utilized by host microorganisms conferring a health benefit,” thereby allowing the possibility of inclusion of noncarbohydrate components, not limited to food ingredients, and application to body parts other than the gastrointestinal (GI) tract, while still requiring selective microbiota-stimulating activities (6). Synbiotics are food ingredients or dietary supplements that combine probiotics and prebiotics and thus synergistically elicit a health effect in the consumer (7); the synergy between pro- and prebiotics has been emphasized by the recommendation of the FAO (8). This synergy can be interpreted in different ways, where the modulatory effect of the prebiotic compound can be designed to directly favor the survival and persistence of the probiotic administered, or the prebiotic functions by modulating the endogenous microbiota, providing a potential health benefit in parallel with the host response to the probiotic administered. These scenarios are not mutually exclusive, suggesting that both mechanisms of action can simultaneously contribute to the health benefits of synbiotics (7, 9).

Lactobacillus plantarum is a species of lactic acid bacteria (LAB) that is found in a wide variety of ecological habitats, ranging from decaying plant materials and fermented food products to the GI tracts of humans and animals (10, 11). L. plantarum strains have been marketed as probiotics, and their strain-specific beneficial effects on human or animal health have been reported in a number of cases (12–18). Notably, the widely studied probiotic strain L. plantarum 299v has been shown to elicit several health-promoting effects, including reduction of cortisol levels in induced stress in young adults, enhancement of iron absorption in women of reproductive age, improvement of irritable bowel syndrome (IBS) symptoms, and reduction of Clostridium difficile colonization in critically ill patients treated with antibiotics (12–18). A recent study has reported that another probiotic L. plantarum strain (ATCC 202195), when administered as a synbiotic together with fructooligosaccharides (FOS), led to significant reductions in neonatal sepsis and mortality in rural India, illustrating the important health potential of L. plantarum-containing synbiotic products (14). However, the exact relationship between the pro- and prebiotic in that study remains unclear, because there is no information on separate pre- or probiotic interventions within the same population, and no information about the ability of L. plantarum strain ATCC 202195 to utilize FOS. Therefore, the underlying and potentially synergistic activities of the synbiotic ingredients that were used in this landmark study remain to be determined.

Various L. plantarum strains have been shown to have high survival capacity during transit in the human GI tract and have been reported in several studies to be detectable in fecal samples 3 to 11 days postadministration (19–22). Prebiotic compounds in a synbiotic combination could potentially enhance colonization efficiency by providing L. plantarum with a suitable carbon source while it resides in the intestinal niche, thus improving the survival or persistence of the bacterium in situ (5, 7). Oligo- and polysaccharide fermentation is of great importance for the performance of L. plantarum in food fermentations as well as during intestinal transit (23, 24), but the current understanding of oligosaccharide metabolism in many Lactobacillus species, including L. plantarum, remains unclear, and comprehensive characterization of metabolic pathways involved in oligosaccharide utilization is limited (25).

L. plantarum displays considerable phenotypic diversity, in particular in strain-specific carbohydrate utilization capacity repertoires, which is reflected by an array of genomic “lifestyle” islands encoding functional carbohydrate utilization cassettes that appear to be located close to the origin of replication (10, 26). Comparative genomics analysis of 54 L. plantarum strains revealed a pangenome encompassing 7,107 orthologous groups (OGs), of which 1,957 were found to form the L. plantarum core genome, leaving 5,150 OGs to be differentially present and absent in strains (i.e., the variome). However, the pangenome did not approach saturation with the 54 available genomes, demonstrating that the genomic diversity of the species is substantially larger than the current pangenome estimate (11). The variability in carbohydrate utilization capacity in this species offers good potential for synbiotic matchmaking with the aim of selectively stimulating the growth of specific strains of the species with specific prebiotic substrates. Furthermore, the availability of genome sequences of many strains of the species could be exploited to identify genetic elements involved in carbohydrate utilization (i.e., prebiotic oligo- and polysaccharides) by gene-trait matching (10, 11, 27).

In this study, 77 L. plantarum strains were screened for their abilities to grow on a selection of generally established prebiotic substrates, including galactooligosaccharides (GOS), inulin, and FOS, as well as putative prebiotic substrates, including isomaltooligosaccharides (IMO), arabinoxylooligosaccharides (AXOS), and fucoidan. Within a subset of the strain panel (23 strains) that displayed variable growth on an isomaltooligosaccharide substrate (IMO2), the specific substrate utilization pattern was determined for individual strains using high-performance anion-exchange chromatography (HPAEC), revealing distinct IMO2 utilization phenotypes These phenotypes could be linked to specific genetic determinants encoded by the corresponding strains by use of gene-trait-matching approaches, revealing the L. plantarum gene clusters that are predicted to be involved in the utilization of isomaltose and longer-chain oligosaccharide compounds that are present in the IMO2 prebiotic substrate.

RESULTS

Growth of L. plantarum strains on prebiotics.

The ability to grow on a selection of prebiotic substrates was analyzed for 77 L. plantarum strains isolated from diverse fermented foods and human host-associated niches from different geographic locations (Table 1). This collection includes the reference strain L. plantarum WCFS1, as well as 38 other strains whose genome sequences are known and which were previously reported to encompass remarkable phenotypic and genomic diversity, particularly in carbohydrate metabolism (10, 26). Besides these strains, a panel of strains was obtained from the Probi A/B culture collection (Table 1).

TABLE 1.

L. plantarum strains used in this study

Strain	NIZO identifier	Source	Geographical origin	GenBank accession no.	Reference
WCFS1	NIZO1836	Human saliva	England	AL935263	32
CIP104440	NIZO1838	Human stool	France	LUWA00000000	10
SF2A35B	NIZO1839	Sour cassava	South America	LUWB00000000	10
NCIMB12120	NIZO1840	Fermented cereal (ogi)	Nigeria	LUWC00000000	10
MLC43	NIZO2029	Raw cheese with rennet	Italy	LUWD00000000	62
CIP104441	NIZO2256	Human stool	France	LUWE00000000	10
CIP104450	NIZO2257	Human stool	France	LUWF00000000	10
CIP104451	NIZO2258	Human urine	France	LUWG00000000	10
CIP104452	NIZO2259	Human tooth abscess	France	LUWH00000000	10
LM3	NIZO2262	Silage	NAa	LUWJ00000000	10
LP80	NIZO2263	Silage	NA	LUWK00000000	10
CHEO3	NIZO2457	Pickled sour pork sausage	Vietnam	LUWM00000000	10
NCTH19-1	NIZO2484	Pickled sour pork sausage	Vietnam	LUWO00000000	10
NCTH19-2	NIZO2485	Pickled sour pork sausage	Vietnam	LUWP00000000	10
NCTH27	NIZO2494	Pickled sour pork sausage	Vietnam	LUWQ00000000	10
LD2	NIZO2535	Fermented orange	Vietnam	LUWR00000000	10
ATCC 8014	NIZO2726	Maize silage	NA	LUWS00000000	10
NOS140	NIZO2741	Cabbage kimchi	Japan	LUWT00000000	63
Q1	NIZO2753	Fermented sourdough	Italy	LUWU00000000	62
H4	NIZO2757	Fermented sourdough	Italy	LUWV00000000	10
H15	NIZO2766	Fermented sourdough	Italy	LUWW00000000	10
ST1CECT4645	NIZO2776	Cheese	NA	LUWX00000000	10
KOG18	NIZO2801	Pickled turnip	Japan	LUWY00000000	63
KOG24	NIZO2802	Cheese	Japan	LUWZ00000000	10
LMG9208	NIZO2806	Sauerkraut	United Kingdom	LUXA00000000	64
Lp 95	NIZO2814	Wine, red grapes	Italy	LUXB00000000	10
BLL(EI31)	NIZO2830	NA	NA	LUXC00000000	10
CECT221(24Ab04)	NIZO2831	Grass silage	United States	LUXD00000000	10
N58	NIZO2855	Pickled sour pork sausage	Vietnam	LUXE00000000	10
X17	NIZO2877	Hot dog	Vietnam	LKHZ00000000	65
LAC7	NIZO2889	Fermented banana	Vietnam	LUXF00000000	10
LD3	NIZO2891	Pickled radish	Vietnam	LUXG00000000	10
LMG18021	NIZO3400	Milk	Senegal	LUXH00000000	10
NA	NIZO3892	Human spinal fluid	France	LUXI00000000	11
NA	NIZO3893	Human stool	France	LUXJ00000000	11
NA	NIZO3894	Vegetables	NA	LUXK00000000	11
Lp I 177	NA	Human children, GI tract	Nordic country	NA	This study
Lp I 154	NA	Human children, GI tract	Nordic country	NA	This study
Lp I 123	NA	Human children, GI tract	Nordic country	NA	This study
Lp E 12	NA	Human children, GI tract	Nordic country	NA	This study
Lp S 71	NA	Human children, GI tract	Nordic country	NA	This study
Lp S 114	NA	Human children, GI tract	Nordic country	NA	This study
Lp S 126	NA	Human children, GI tract	Nordic country	NA	This study
Lp S 144	NA	Human children, GI tract	Nordic country	NA	This study
Lp S 177	NA	Human children, GI tract	Nordic country	NA	This study
Lp S 204	NA	Human children, GI tract	Nordic country	NA	This study
Lp S 265	NA	Human children, GI tract	Nordic country	NA	This study
Lp S 268	NA	Human children, GI tract	Nordic country	NA	This study
Lp 42A	NA	Human rectum	Sweden	NA	This study
Lp 62C	NA	Human tongue	Sweden	NA	This study
Lp 35C	NA	Human tongue	Sweden	NA	This study
Lp 67B	NA	Human rectum	Sweden	NA	This study
Lp 88C	NA	Human tongue	Sweden	NA	This study
Lp 36D	NA	Human tongue	Sweden	NA	This study
Lp 81D	NA	Human tongue	Sweden	NA	This study
Lp 6A	NA	Human rectum	Sweden	NA	This study
Lp 12D	NA	Human tongue	Sweden	NA	This study
Lp 962	NA	Unknown	Unknown	NA	This study
Lp 12C	NA	Human tongue	Sweden	NA	This study
Lp 32A	NA	Human tongue	Sweden	NA	This study
Lp 86C	NA	Human tongue	Sweden	NA	This study
Lp 33C	NA	Unknown	Unknown	NA	This study
Lp 37C	NA	Human tongue	Sweden	NA	This study
Lp 78B	NA	Human tongue	Sweden	NA	This study
Lp 8014	NA	Unknown	Unknown	NA	This study
Lp 39	NA	Ogi (red sorghum)	Nigeria	NA	This study
Lp 49	NA	Ogi (red sorghum)	Nigeria	NA	This study
Lp 900	NA	Ogi (red sorghum)	Nigeria	NA	This study
Lp 59	NA	Ogi (red sorghum)	Nigeria	NA	This study
Lp 74	NA	Ogi (red sorghum)	Nigeria	NA	This study
Lp 56	NA	Ogi (red sorghum)	Nigeria	NA	This study
299v	NIZO2260	Human intestine	Sweden	LEAV00000000	21
299	NIZO1837	Human colon	Sweden	LTAU00000000	21
Heal 99	NA	Human GI tract	Sweden	NA	This study
SD5870	NA	Human GI tract	Sweden	NA	This study
Heal 9	NA	Human GI tract	Sweden	NA	66
Heal 19	NA	Human GI tract	Sweden	NA	67

^aNA, not available.

We established a 96-well high-throughput screening approach to evaluate the growth of the L. plantarum strains on a selection of prebiotic substrates with variable degrees of polymerization (DP), monosaccharide compositions, and linkages. The substrates employed here included galactooligosaccharides (Vivinal GOS), fructooligosaccharides (FOS and inulin), arabinoxylooligosaccharides (AXOS), isomaltooligosaccharides (IMO1 and IMO2), and the seaweed-derived sulfated polysaccharide fucoidan (Table 2). The optical densities (ODs) reached after overnight growth on a prebiotic- or glucose-supplemented medium were corrected by subtraction of the OD reached on the medium without a supplement. These corrected ODs revealed different degrees of strain-specific growth capacities within the strain panel (Fig. 1A).

TABLE 2.

Characteristics and origins of prebiotics used in this study based on information received from their respective suppliers

Compound	Supplier	Composition (purity [%])	Monosaccharide composition	DP range	Mono- and disaccharide content (%)
Orafti P95	BENEO-Orafti	Oligofructose (93.2–97.5)	Glucose, fructose	2–8a	∼2–7
Orafti GR	BENEO-Orafti	Inulin (90)	Glucose, fructose	2–60a	∼10
Vivinal GOS	FrieslandCampina Domo	Galactooligosaccharides (70)	Glucose, galactose	2–8a	∼30
AXOS	Cargill	Arabinoxylan oligosaccharides (71)	Xylose, arabinose	2–9a	∼16
IMO1	Cargill	Isomaltooligosaccharides (DP, 3–4) (41.2)	Glucose	2–7b	∼30
IMO2	Cargill	Isomaltooligosaccharides (DP, 3–4) (39.4)	Glucose	2–7b	∼40
Fucoidan	South Product Co., Ltd.	Sulfated fucopolysaccharides (86.7)	Fucose, glucuronic acid	NAc	NA

^aInformation provided by the supplier.

^bInformation provided by in-house analyses.

^cNA, no detailed information available.

FIG 1.

L. plantarum strain-specific relative growth on prebiotic substrates. (A and B) The relative growth (expressed as OD₆₀₀) of L. plantarum strains on prebiotic substrates in comparison to their growth on glucose was hierarchically clustered and displayed in a heat map (A) and in distribution plots where horizontal lines indicate averages and standard errors of the means (B). (C) Relative OD₆₀₀ (black bars) and pHs (gray bars) of a subset of selected L. plantarum strains grown on IMO2 compared to glucose in controlled-growth experiments, and their congruency with the relative OD₆₀₀ (white bars) obtained in the screening.

Only a single strain (Lp 900) displayed growth efficiencies on a FOS- or inulin-containing medium that were comparable to its growth on the preferred substrate for growth, glucose. All other strains in the panel displayed only minor amounts of growth on FOS or inulin, which is likely explained by the utilization of the mono-, di-, and possibly trisaccharide (28) compounds present in these prebiotic substrates (Table 2). Analogously, very limited growth was observed for all strains on media containing AXOS or fucoidan, which never appeared to exceed the expected background levels of growth based on the ability of L. plantarum strains to utilize common mono- and disaccharides in these substrates (i.e., glucose, arabinose, and cellobiose). Conversely, all strains were able to grow on Vivinal GOS, although the extent of growth relative to growth on glucose ranged from 58% to 100% (Fig. 1B). Finally, substantial strain-specific variation of growth capacity was detected for both the IMO substrates, where the relative growth (compared to growth on a glucose-containing medium) ranged from 37% to 97% for IMO1 and from 4% to 96% for IMO2 (Fig. 1B). The difference in growth capacity in the strain panel between the two IMO substrates can be attributed to their distinct differences in chemical composition, most notably the larger amounts of isomaltose and other α-1,6 linked glucooligosaccharides in IMO2 (see Table S1 in the supplemental material).

Since the growth capacity on IMO2 displayed the largest degree of variation within the strain panel, this substrate was selected for further investigation. A subset of strains (23 strains) with known genome sequences was selected for verification of the substrate-screening results in controlled-growth experiments (growth in culture tubes rather than 96-well plates). This subset of strains was selected in such a way that their growth capacities on IMO2 represented the range of growth capacities observed in the original screening. The strain-specific relative growth of the 23 selected strains under controlled-growth conditions in IMO2-containing medium (expressed as the OD at 600 nm [OD₆₀₀]) was strongly correlated (Spearman correlation [ρ] = 0.7940; P < 0.0001) with that obtained in the screening procedure. Moreover, strain-specific growth on a particular substrate was also strongly correlated with the final pH reached after overnight growth (Spearman correlation [ρ] = –0.9380; P < 0.0001) (Fig. 1C). These observations establish the reliability of the outcome generated in the high-throughput screening procedure and indicate that both the final OD₆₀₀ and the pH can adequately reflect the strain-specific ability to utilize the substrate provided in such a screening procedure (Fig. S1).

Refining L. plantarum strain-specific IMO2 substrate utilization.

The IMO2 substrate is composed of a diverse range of specific carbohydrates, including isomalto-, malto-, and other glucooligosaccharides with various degrees of polymerization and glyosidic linkages (Table S1). Analogously, IMO2 complexity was confirmed by HPAEC analysis, revealing 38 distinct peaks, including glucose, maltose, isomaltose, isomaltotriose, maltotriose, panose, and maltotetraose, which were identified through standards and/or HPAEC comparative analyses (Fig. 2A). The variation in the abilities of the L. plantarum strains to grow on IMO2 could depend on strain-specific utilization of a subfraction(s) of the IMO2 substrate. Therefore, the spent culture supernatants of the strains were analyzed by HPAEC and compared to the original IMO2 in order to estimate strain-specific peak utilization by calculating the ratio of the area under the curve of each peak in the spent medium and the original IMO2 substrate (Fig. 2B).

FIG 2.

Refinement of IMO2 utilization by an L. plantarum subpanel. (A) HPAEC-PAD elution pattern of the IMO2 substrate, in which the compounds identified include isomaltose (peak 3), isomaltotriose (peak 8), maltose (peak 12), panose (peak 18), maltotriose (peak 23), and maltotetraose (peak 29). (B) Heat map and hierarchical clustering of relative peak surface area decreases in the HPAEC-PAD responses of spent culture supernatants of strains grown on IMO2 relative to peak surface areas for the original IMO2 substrate. (C) Exemplary HPAEC-PAD elution patterns of five distinct L. plantarum phenotypes of IMO2 utilization, with the original IMO2 substrate (black) (elution pattern 1), a “nonuser” (Nizo2535) (red) (elution pattern 2), utilization of peak 16 (Nizo2766) (orange) (elution pattern 3), partial utilization of HDP compounds (Nizo2808) (dark green) (elution pattern 4), utilization of HDP compounds (Nizo2029) (light green) (elution pattern 5), and utilization of isomaltose and HDP compounds (Nizo3892) (blue) (elution pattern 6). Arrows indicate decreases in peaks corresponding to the utilization of specific compounds in IMO2 (shown in panel A). A 2% shift in the x axis is set for each successive chromatogram following that for elution pattern 1.

These analyses revealed that only four L. plantarum strains (Nizo3892, Nizo2830, Nizo2726, and Nizo2256) appeared to utilize isomaltose (peak 4), together with an unknown compound (peak 5). The high isomaltose prevalence in IMO2 could thus explain why these strains also displayed the highest relative growth on IMO2 (Fig. 3). The isomaltose-utilizing ability of these L. plantarum strains was verified by their growth in a medium supplemented with pure isomaltose, where these strains reached a final OD₆₀₀ comparable to that in a medium supplemented with the corresponding amount of glucose (Fig. S2). Notably, a subset of strains that were found not to utilize isomaltose (Nizo2535, Nizo2891, WCFS1, and Nizo2029) only reached final OD₆₀₀ values in isomaltose-supplemented medium that were comparable to those in the same medium without carbon source supplementation (Fig. S2). Remarkably, the isomaltose utilization phenotype always coincided with the utilization of IMO2 components with the highest retention time in HPAEC (peaks 25, 27, and 30 to 38), whose chemical nature is unknown but which represent oligosaccharides with higher DP (tetra-, penta-, and hexasaccharides [data not shown]), to which we will refer as high-DP (HDP) compounds. These high-DP compounds could also be utilized by 13 other strains of L. plantarum (Nizo2258, Nizo3894, Nizo2753, Nizo2485, Nizo2029, Nizo2263, Nizo2259, Nizo2814, Nizo2484, Nizo2257, Nizo1840, Nizo2889, and WCFS1), whereas 1 strain (Nizo2802) appeared to utilize only a smaller subset of these high-DP compounds (peaks 25, 27, 30, and 33 to 37) (Fig. 2B). Notably, this panel of strains could not utilize isomaltose, indicating that the isomaltose and late-eluting-compound utilization phenotypes are independent, or at least not necessarily linked. Finally, a small set of strains (Nizo2535, Nizo2891, Nizo2766, Nizo1838, Nizo2256) did not utilize either the high-DP compounds or isomaltose (Fig. 2B). However, three of these five strains (Nizo2766, Nizo1838, and Nizo2256) could exclusively utilize a low-abundance compound of an unknown chemical nature (peak 16) (Fig. 2B). Exemplary chromatograms of isomaltose, peak 16, and high-DP peak utilizations are shown in Fig. 2C. Comparative analysis of growth data (optical densities acquired in screening and controlled-growth experiments) and strain-specific peak utilization revealed that on average, high-DP peak utilization results in optical densities higher than those of the nonutilizers and peak 16-utilizing strains that still achieve some level of growth due to the presence of maltose and glucose (which are readily utilized by all L. plantarum strains) (Fig. 3). Therefore, high-DP compounds are utilized for growth.

FIG 3.

Correspondence between growth efficiency and IMO2 utilization phenotypes. The relative growth (expressed as OD₆₀₀) on the IMO2 substrate obtained in the screening and controlled-growth experiments was related to the differential IMO2 utilization phenotypes (i.e., HPAEC peak utilization phenotypes) observed for the strains in the panel. Strain-specific IMO2 peak utilization phenotypes are indicated as follows: blue, isomaltose and HDP utilization; light green, HDP utilization; dark green, partial HDP utilization; orange, utilization of peak 16; red, nonutilizers. See the text for details.

Candidate L. plantarum genes involved in the utilization of specific IMO2 compounds.

Comparative genomics, specifically predicting orthology relations between translated protein sequences of the publicly available L. plantarum genomes using the orthAgogue tool (29), was performed to assemble an OG matrix for the 23 L. plantarum strains selected from the screening panel. The genome sizes of the selected strains ranged from 3 to 3.5 Mb, encompassing a core genome of 2,189 OGs (OGs found in all strains) and a “variome” of 3,353 OGs; collectively, these were assembled into a pangenome of 5,542 OGs (total amount of OGs). The numbers of OGs found in the core genome, pangenome, and variome of the 23 strains are in (relative) agreement with those reported previously for the comparison of 54 strains of L. plantarum (11).

The OG matrix for the 23 strains was used to correlate the IMO2 compound-specific utilization patterns with the corresponding strain genotypes, with the aim of identifying OGs that are putatively involved in the observed peak-specific IMO2 utilization phenotypes. This procedure is often referred to as “gene-trait matching” (GTM) (27, 30, 31). GTM revealed a perfect correlation between the ability to utilize isomaltose and the presence of a specific gene cluster (cluster A), which was uniquely present in isomaltose-utilizing strains (Nizo3892, Nizo2830, Nizo2726, Nizo2831) (Fig. 4). The cluster encompasses nine genes encoding the canonical EIIA, EIIB, and EIIC components (OG3438, OG3687, and OG3276, respectively) of a phosphotransferase system (PTS) that is annotated to transport glucose or maltose, a transcriptional regulator (OG3685), a maltose-6′-phosphate glucosidase belonging to glycoside hydrolase family 4 (OG3277), a β-phosphoglucomutase (OG3701), and a hypothetical protein (OG3684) (Fig. 4B). Notably, the genes encoding PTS subunit IIC (OG3276) and maltose-6′-phosphate glucosidase (OG3277) appear to be duplicated (100% identical duplicate copies) in the operon, while the PTS IIA subunit (OG3438) was also found in strains Nizo2484 and Nizo2485, which could not utilize isomaltose. However, in the latter two strains, this PTS IIA subunit (OG3438) was flanked by genes encoding components of a PTS predicted to transport cellobiose and/or other β-glucosides (data not shown), suggesting that its clustering in OG3438 depends on general conservation of PTS IIA genes rather than the substrate specificity of the corresponding PTS IIC component. Importantly, the maltose-6′-phosphate glucosidase (OG3277) lacks a predicted signal peptide, indicating that this enzyme most likely functions intracellularly and is not involved in extracellular degradation of isomaltose prior to import. This implies that the PTS system identified is likely the importer for isomaltose, leading to intracellular isomaltose phosphate, which is subsequently converted intracellularly, possibly involving the maltose-6′-phosphate glucosidase and/or the β-phosphoglucomutase. Notably, gene-trait matching also revealed an ABC transporter permease protein (OG3690) that was uniquely present in the four isomaltose-utilizing strains (Nizo3892, Nizo2830, Nizo2726, Nizo2831). However, closer inspection of this gene revealed that it was a truncated version (i.e., pseudogene) of the ABC transporter permease protein that is present in all the other strains (OG2361), indicating that it is not involved in isomaltose import (Data Set S1). One strain tested in the screening panel (Lp 56) (Fig. 1A) for which no genome sequence is available exhibited a comparable isomaltose-utilizing phenotype (Fig. S3A). Therefore, we predicted that this strain also harbors cluster A. The presence of the PTS IIC subunit (OG3276) in Lp 56 was verified by PCR (Fig. S3B) and thereby supports the genotype-phenotype causality of cluster A and isomaltose utilization.

FIG 4.

Gene cluster A, identified in L. plantarum strains that utilize isomaltose. (A) Heat map displaying the in silico gene-trait-matching results based on the association between the presence (green) or absence (red) of specific OGs and the utilization (strain names in boldface) or lack of utilization (strain names in lightface) of isomaltose. The asterisk indicates an OG not found exclusively in isomaltose-utilizing strains but still considered to play a role in isomaltose utilization (see the text for details). (B) Predicted functions of genes in gene cluster A. (C) Gene map of cluster A in strain Nizo2830, with OGs identified by gene-trait matching (dark blue arrows); OG3438, which is expected to be a part of cluster A but is also found in two strains that do not utilize isomaltose (labeled with an asterisk); and genes flanking the cluster (gray arrows).

The four isomaltose-utilizing strains (Nizo3892, Nizo2830, Nizo2726, and Nizo2831) encode two additional gene clusters (designated clusters B and C [Fig. S4]) directly downstream of the isomaltose utilization-associated cluster that encompass genes encoding various carbohydrate utilization functions, which are also found in the genomes of 13 of the 17 strains that are able to utilize the high-DP compounds in the IMO2 substrate (Nizo3892, Nizo2726, Nizo2830, Nizo2831, Nizo2029, Nizo2263, Nizo2259, Nizo2814, Nizo3894, Nizo2484, Nizo2485, Nizo2889, and WCFS1). Notably, gene clusters B and C are consistently absent from the five strains that do not utilize any high-DP compounds (Nizo2535, Nizo2891, Nizo2766, Nizo1838, and Nizo2256), suggesting that these gene clusters could play a role in the utilization of the high-DP fraction in IMO2. Intriguingly, strain Nizo2802, which could utilize only a restricted set of the high-DP compounds, harbored only gene cluster C, supporting a role of these clusters in the high-DP utilization phenotype. However, four strains that do utilize the high-DP compounds do not carry gene clusters B and C (Nizo2258, Nizo2257, Nizo1840, Nizo2753), suggesting that these strains may carry different genes involved in the high-DP utilization phenotype, although we could not identify such alternative genes or gene clusters through GTM analysis. Gene cluster B encodes several glycoside hydrolases, including several putative (exo-)α-mannosidases (OG2786, OG2705, OG2781) (Fig. S4). Gene cluster C contains a gene encoding a putative intracellular glucan α-1,6 glucosidase that is predicted to belong to family 13 of glycoside hydrolases (OG2683) and that was previously annotated as a dextranase, DexB, in L. plantarum strain WCFS1 (32). Cluster C also encodes an ABC transport system consisting of a solute-binding protein (OG2748), two transmembrane proteins (OG2742 and OG2755 [genes annotated as msmG and msmF, respectively]), and two ATP-binding domains (OG2744 and OG2743), which is predicted to be involved in the import of multiple sugars (Fig. S4). As in gene cluster A, all putative glycoside hydrolases in clusters B and C lack a predicted signal peptide and are predicted to function intracellularly. Consequently, due to the apparent absence of an extracellular glycan hydrolase, utilization of the high-DP compounds is most likely dependent on their import through the ABC transporter system encoded within cluster C, which is remarkable in view of the tetra-, penta-, and even hexasaccharide nature of these high-DP compounds.

In analogy to the perfectly matching association of gene cluster A with isomaltose utilization, the strains that could utilize the chemically uncharacterized peak 16 (Nizo2766, Nizo1838, and Nizo2256) in the IMO2 HPAEC chromatogram uniquely carried a gene cluster (designated cluster D) that is absent in all other strains. Gene cluster D encodes an ABC transport system, including two transmembrane proteins (OG3638 and OG3817 [genes annotated as malG and malF, respectively]), a substrate-binding protein (OG3820 [gene annotated as malE]), and a sugar phosphorylase (OG3823), which are predicted to be involved in the import of maltose and/or maltodextrin. Gene cluster D is consistently flanked by OGs that are part of the core genome of this strain panel, one of which encodes an ATP-binding protein (OG36) that we propose to be associated with the ABC transport system encoded downstream in the strains harboring cluster D (Fig. 5).

FIG 5.

Gene cluster D, identified in L. plantarum strains that utilize IMO2 peak 16. (A) Heat map displaying the in silico gene-trait-matching results based on the association between the presence (green) or absence (red) of specific OGs and the utilization (strain names in boldface) or lack of utilization (strain names in lightface) of peak 16 in the IMO2 HPAEC elution pattern. (B) Predicted functions of the genes in gene cluster D. FMN, flavin mononucleotide; FAD, flavin adenine dinucleotide. (C) Gene map of cluster D in strain Nizo2766, with OGs identified by gene-trait matching (dark blue arrows); OG36, which is expected to be a part of cluster D but is also part of the core genome (light blue arrow); and genes flanking the cluster (gray arrows).

DISCUSSION

In this study, we set up a high-throughput growth-screening model to identify (potential prebiotic) oligo- or polysaccharides that can selectively stimulate the growth of one or several strains of L. plantarum within a panel of 77 strains. Moreover, using the IMO2 substrate as an example, we illustrate how the screening results can be used to identify genes that are predicted to be involved in the utilization of specific carbohydrate substrates using a gene-trait-matching approach.

The growth-screening results revealed no strains that could grow effectively on either AXOS or fucoidan, which is in agreement with the lack of literature-reported growth of L. plantarum on these substrates. Nevertheless, AXOS- or fucoidan-utilizing L. plantarum strains could possibly be isolated from niches enriched in these carbohydrate polymers, such as maize silage for AXOS (33) and brown seaweeds for fucoidan (34). Only a single strain was identified that grew to high cell densities on inulin and FOS, a phenotype that has been reported before for specific strains of this species (35), while the majority of strains are likely to utilize only the short-chain fructooligosaccharides (up to DP3), as has been shown for L. plantarum WCFS1 (28). The galacto- and isomaltooligosaccharides (Vivinal GOS, IMO1, and IMO2) supported the growth of virtually all strains but to variable extents. These are the most complex substrates used here, containing a variety of compounds that differ in glycosidic linkages and DP (36, 37), as a result of their production by enzymatic polymerization using disaccharides as a substrate (37). In contrast, AXOS, inulin, FOS, and fucoidan are naturally occurring plant polysaccharides or hydrolyzation products thereof, with more-homogeneous glycosidic linkages.

The largest variation in growth capacity was found for one of the IMO preparations that contains relatively high levels of isomaltose (IMO2). In a subpanel of 23 L. plantarum strains, comparative HPAEC allowed the identification of five distinct IMO2 utilization phenotypes, which were characterized by the specific utilization of a single compound, or a subset of the HPAEC-separated compounds, present in the IMO2 substrate. The availability of (draft) genome sequences for these 23 strains enabled the prediction of genes involved in these distinctive IMO2 utilization phenotypes by gene-trait matching.

Isomaltose (α-1,6 linked glucose disaccharide) can be digested by sucrase–isomaltase (EC 3.2.1.48) in the brush border of the small intestine in most mammals (38, 39), disqualifying it as a prebiotic or a suitable ingredient in a synbiotic combination. Nevertheless, it did selectively stimulate the growth of four L. plantarum strains, which was associated with the presence of highly conserved genes encoding a putative isomaltose-importing PTS (OG3276; cluster A; up to 100% identity among the four strains) that also resembles other PTS subunit IIC components (>70% identity) found in lactobacilli, including Lactobacillus pentosus, Lactobacillus rhamnosus, and Lactobacillus paracasei. The high identity of the cluster A genes (up to 100% identity) in the isomaltose-utilizing strains (Nizo2830, Nizo2831, Nizo2726, and Nizo3892) coincides with their previously reported phylogenetic relatedness (11) in a cluster of strains that also contains the L. plantarum JDM1 and ER strains, in which cluster A appears to be conserved. These strains were isolated from various niches (11), suggesting that these genes do not reflect a clear niche adaptation, in agreement with the proposed nomadic lifestyle of L. plantarum as well as several other Lactobacillus species (11, 40). This gene-trait-matching result could be further confirmed by the PCR-based detection of OG3276 in a fifth strain that could utilize isomaltose but for which the genome sequence remains to be determined.

The commonly annotated substrate specificity of the cluster A-encoded PTS is α-glucoside or maltose, which is in agreement with its genetic linkage with a predicted maltose-6-phosphate glucosidase (OG3277). However, our results indicate that this PTS imports isomaltose rather than maltose, corroborating that substrate specificity annotations of carbohydrate transporters and hydrolases are often unreliable and may cause misleading carbohydrate utilization predictions (41, 42). However, our results cannot exclude a dual role of the PTS identified in the import of both maltose and isomaltose. Moreover, the genetically linked maltose-6′-phosphate glucosidase (OG3277) belongs to glycoside hydrolase family 4, which has been proposed to hydrolyze various 6-phospho-α-glucosides (43), thereby not allowing a more-precise substrate specificity prediction. Nevertheless, our findings support the involvement of these genes in isomaltose import and isomaltose-6-phosphate hydrolysis, and it remains to be established whether these genes could also play a role in maltose utilization. Interestingly, downstream of the proposed isomaltose PTS operon, a core genome-associated oligo-1,6-glucosidase (OG5; malL) is encoded, which was previously shown to hydrolyze isomaltose and isomaltulose (α-1,6 linked glucose and fructose disaccharide), but not isomaltotriose (α-1,6 linked glucose trisaccharide) or panose (α-1,6- and α-1,4-linked glucose trisaccharide) in L. plantarum LL441 (44). However, several additional α-1,6 glucosidase-encoding genes have been annotated in the L. plantarum WCFS1 genome (32), suggesting that this is a redundant function in this species, although the substrate specificities and/or affinities of these enzymes may be (subtly) different. Taking these findings together, it appears that many strains of L. plantarum encode a repertoire of α-1,6-glucosidases, some of which could potentially hydrolyze isomaltose, but that growth on isomaltose as a substrate appears to be constrained by import, and only strains that encode the PTS identified and its cognate isomaltose-6-phosphate hydrolase can grow effectively on this substrate.

In view of the above considerations, it is relevant that maltose import in lactobacilli is usually considered to be facilitated by ABC transporters or maltose-H⁺ symporters and has not been reported to involve a PTS function (45–47). For example, in L. plantarum WCFS1, the mdxE, mdxF, mdxG, and msmX genes are annotated to encode the maltose ABC transporter (32) and correspond to OG2436, OG2451, OG2439, and OG36 in the OG matrix used in this study. Notably, the ATP-binding protein (OG36) is part of the core genome of our strain panel, while the cognate OG2436, OG2451, and OG2451 were found to be conserved among all strains except those that uniquely utilize the unknown compound “peak 16” (Nizo1838, Nizo2256, and Nizo2766) (Fig. 2). Intriguingly, the latter strains uniquely harbor an alternative maltose ABC transport system, encompassing the (cluster D-encoded) canonical substrate-binding protein (MalE; OG3820) and two transmembrane transport proteins (MalF; OG3817 and 3638) as well as the corresponding ATP-binding protein (MalG), which is a member of OG36. Since all strains utilized maltose, these findings suggest that both ABC transport systems (MdxEFG and MalEFG) facilitate maltose import but that only the malEFG-encoded ABC importer also facilitates the import of an additional carbohydrate, peak 16. The chemical nature of the compound detected as peak 16 remains unknown, but based on its retention time in HPAEC, it is likely a di- or trisaccharide. The selective utilization of this compound is of interest, and its chemical characterization and purification could be the start of investigating its potential as a prebiotic, which could be implemented in “selective” synbiotic combinations with L. plantarum probiotic strains that encode the MalEFG import system.

A relatively larger subset of the 23 strains (78%) utilized several tetra-, penta-, and hexaoligosaccharides present in the IMO2 substrate (designated HDP compounds). Interestingly, these compounds are not α-1,4-linked maltodextrins and are unlikely to be isomaltooligosaccharides, which are consistently linked through α-1,6 bonds, since HPAEC analyses demonstrated that none of the strains could utilize the α-1,6-linked trisaccharide isomaltotriose. Furthermore, L. plantarum has not been reported to utilize extracellular α-1,6-linked isomaltooligosaccharides and has even been used to remove residual sugars from IMO preparations for a purer oligosaccharide end product (48). Therefore, these oligosaccharides are unlikely to represent typical α-1,4- and/or α-1,6-linked isomaltooligosaccharides; they probably represent by-products of industrial IMO production (37). Although GTM analysis did not reveal an unambiguous gene association with this prevalent phenotype, two gene clusters (clusters B and C) are likely to be involved. Analogously, previous GTM analyses reported on the often-ambiguous associations between genotypes and phenotypes, without disallowing the identification of genes that are credibly involved in specific phenotypes (27, 30, 49). Intriguingly, gene clusters B and C do not appear to encode glycan hydrolases that are predicted to be secreted and could thereby break down these HDP oligosaccharides to liberate their monosaccharide constituents for import and metabolization by L. plantarum, suggesting that these HDP compounds are imported and hydrolyzed intracellularly. Although the import and hydrolysis systems involved in disaccharide metabolism in lactobacilli often are also able to import and utilize chemically related trisaccharides and, in some cases, tetrasaccharides as well (25, 28), it is uncommon that oligosaccharides of a higher DP are imported. Nevertheless, GTM suggests that the HDP compounds utilized in IMO2 are imported prior to their intracellular hydrolysis and metabolization by the cluster C-encoded ABC transporter. Notably, import of long-chain saccharide compounds by ABC importers has been described previously, including alginate import in Sphingomonas sp. strain A1 (50) and import of a long-chain FOS in Streptococcus pneumoniae (51). As stated above for peak 16, the structural elucidation of these HDP compounds, combined with unravelling of the mechanism by which they are utilized by L. plantarum, may pave the way to more-selective synbiotic combinations.

In conclusion, we established a high-throughput growth-screening model for in vitro screening of prebiotic substrate utilization, which we employed for 77 L. plantarum strains, but this model could equally be employed to screen other lactobacilli or other genera. Notably, in our study, we employed the model using semiaerobic conditions, suitable for the aerotolerant species L. plantarum, but it could readily be adapted for the screening of anaerobic microbial species by use of an anaerobic cabinet. More-precise characterization of specific compound utilization phenotypes by HPAEC demonstrates the potential for identification of individual compounds that display the most selective growth stimulation. Chemical characterization of such compounds and verification that they are undigested by the host-associated enzyme repertoire offers approaches toward their inclusion in synbiotics to selectively stimulate the growth of the accompanying probiotic that can utilize these compounds. Moreover, we were able to identify genotype information associated with specific compound utilization phenotypes, which enables dedicated genotype-screening efforts to predict the abilities of strains to utilize specific prebiotic compounds based on their genetic repertoires. Taking our results together, our work illustrates an effective approach for synbiotic matchmaking, which could improve the efficacy of in situ delivery and growth of probiotics to enhance their health effects.

MATERIALS AND METHODS

Bacterial strains and prebiotic compounds.

The 77 L. plantarum strains used in this study were selected from the NIZO and Probi A/B strain collections (Table 1). They were originally isolated from fermented foods of European and non-European origins or from different niches associated with the human body (mostly from samples obtained from oral and intestinal niches [Table 1]).

The prebiotic substrates used in this study are listed in Table 2. The compositions, degrees of polymerization (DP), and levels of purity of the compounds are listed according to the manufacturers’ data sheets. Vivinal GOS was provided by FrieslandCampina Domo (Amersfoort, the Netherlands); FOS and inulin were provided by BENEO-Orafti (Oreye, Belgium); AXOS, IMO1, and IMO2 were provided by Cargill (Minneapolis, MN, USA), and fucoidan was provided by South Product Co., Ltd. (Uruma, Japan). Differences in the composition of IMO preparations are described in Table S1 in the supplemental material.

Screening for bacterial growth on prebiotic substrates.

L. plantarum strains were grown overnight at 37°C in De Man, Rogosa, and Sharpe (MRS) broth (Merck, Kenilworth, NJ, USA). Cells were harvested by centrifugation (4,000 × g, 10 min, 4°C), washed with phosphate-buffered saline (PBS) solution, resuspended in an equal volume of PBS containing 15% (vol/vol) glycerol, and stored in 300 μl in 96-well microtiter plates (Corning, Inc., Corning, NY, USA) at –80°C to create “master plates” for subsequent inoculation in different carbohydrate-supplemented media. Culture-containing wells were interspaced with empty wells so as to avoid false-positive results due to cross-contamination during inoculation. The medium used for screening of strain-specific growth on prebiotic substrates, 2-fold-diluted MRS without the addition of any carbon source (½ MRS-C), was prepared and supplemented with 0.5% (wt/vol) prebiotic substrate (GOS, FOS, inulin, IMO1, IMO2, AXOS, or fucoidan) or glucose. Due to the different levels of mono- and disaccharide content in the prebiotics screened, the screening model employed limiting amounts of total carbon sources by using 2-fold-diluted MRS-C and reduced the prebiotic or glucose content to 0.5% (wt/vol) instead of the regular 2.0% (wt/vol) in standard MRS so as to enable distinctions in growth between strains. The strains were also grown on ½ MRS-C without carbon source supplementation in order to account for background growth on carbohydrates that are endogenously present in the constituents of this medium, such as yeast and meat extract.

Strains were inoculated in triplicate from the glycerol master plates using a microplate replicator (Boekel Scientific, Feasterville-Trevose, PA, USA) in 96-well plates containing 200 μl of the specific medium and were grown for 24 h at 37°C. Noninoculated wells interspaced between culture-containing wells were included as negative controls. After 24 h, bacterial cultures were resuspended, and the final cell density was determined by measurement of the optical density at 600 nm (OD₆₀₀) with a microplate spectrophotometer (SpectraMax M5; Molecular Devices, San Jose, CA, USA). Optical densities determined for the strains grown on ½ MRS-C were subtracted from those obtained from growth on a supplemented medium. The strain-specific relative growth capacity on each prebiotic substrate was calculated by dividing the OD₆₀₀ reached in prebiotic-supplemented medium by the OD₆₀₀ reached in glucose-supplemented medium. Since glucose is anticipated to support the most effective growth of these L. plantarum strains, the maximum relative growth on prebiotic-supplemented media was set to 1.0, and the minimum relative growth was set to 0.0, even though subtraction of growth on MRS-C resulted in values below 0.0 in some cases. To visualize the screening results, the relative growth values for each of the prebiotic substrates were represented in a heat map, made in R, version 3.4.0 (52), using the pheatmap 1.0.7 package (53).

Controlled-growth experiments.

To consolidate the 96-well screening results for the IMO2 substrate, a subset of 23 strains was selected that spanned the range of relative growth values observed for the IMO2 substrate in the screening. The 23 selected strains were inoculated in 15-ml Falcon tubes (Corning, Inc., Corning, NY, USA) containing ½ MRS-C supplemented with either 0.5% IMO2 or 0.5% (wt/vol) glucose and were grown for 24 h at 37°C. In the resulting cultures, the final OD₆₀₀ (SpectraMax M5 microplate spectrophotometer) and pH were determined. Cells were removed from these cultures by centrifugation (4,000 × g, 10 min, 4°C), and spent culture supernatants were stored at –20°C for further analysis (see below). Within this subset of strains, the growth of four strains on isomaltose was confirmed by preculturing these strains (and a selected panel of negative-control strains) on ½ MRS-C supplemented with 0.5% (wt/vol) glucose and passaging them (1:500 inoculum) in ½ MRS-C supplemented with 0.5% (wt/vol) isomaltose or glucose as the sole carbon source.

HPAEC peak profile characterization.

Spent culture supernatants were diluted 1:20 in Milli-Q, and eventual debris was removed by centrifugation (16,000 × g, 15 min, room temperature) and subsequently analyzed by high-performance anion-exchange chromatography with pulsed amperometric detection (HPAEC-PAD), using an ICS5000 HPAEC (Dionex, Sunnyvale, CA, USA) with a CarboPac PA1 (2 by 250 mm) column in combination with a CarboPac PA1 (2 by 50 mm) guard column. Eluents A (0.1 M sodium hydroxide) and B (1 M sodium acetate in 0.1 M sodium hydroxide) were used at a flow rate of 0.3 ml/min in a linear gradient from 0 to 40% B (25 min) and 40 to 100% B (10 min), followed by column washing with 100% B (5 min) and column reequilibration with 100% A (20 min). The elution was monitored by a pulsed amperometric detector (Dionex ICS‐5000 ED); the column oven was kept at 20°C; and the sample tray was set at 10°C. Peak profile analysis was performed with Chromeleon software (version 7.2; Thermo Fisher Scientific, Waltham, MA, USA). Relative peak utilization per strain was estimated by determination of the area under the curve in the spent culture supernatant relative to that of the corresponding peak in the original IMO2 substrate and was visualized in heat maps using pheatmap package 1.0.7 (53) in R, version 3.4.0.

Gene-trait matching.

The genomes of the 23 L. plantarum strains, for which IMO2 utilization was analyzed with HPAEC, were obtained from the National Center for Biotechnology Information (NCBI) database (54). Protein sequences were aligned with an all-versus-all bidirectional BLASTP (cutoff E value, <1e–05) (55), and the resulting BLAST output was processed by OrthAgogue (29) for clustering of orthologue genes into orthologous groups (OGs). The orthologue gene families obtained were appointed as either part of the core genome or part of the variome based on their presence in all of the strains analyzed or in just a subset of the panel. The resulting OG assignments were used to establish an OG matrix where each orthologous group was linked to the corresponding genome locus tag for each of the strains, which was subsequently converted to a binary gene presence (1)-or-absence (0) matrix (see Data Set S1 in the supplemental material). In silico genotype-phenotype gene-trait matching (GTM) was performed based on the binary matrix of orthologous genes generated, correlating the presence and absence of genes with the observed peak-specific substrate utilization phenotype visualized via HPAEC. Lists of candidate OGs involved in a particular phenotype were selected by sorting the OG matrix for those OGs that displayed the highest degree of presence in strains that could utilize a specific peak, while the same OGs had the lowest presence (or complete absence) in the strains that could not. The genomic regions containing the selected candidate OGs were visualized with Artemis, version 16.0.0 (56), while the presence or absence of candidate OGs and their corresponding operons in different strains was visualized in heat maps using pheatmap 1.0.7 (53) in R (version 3.4.0) and in gene maps using Easyfig 2.2.2 (57). Candidate OGs were screened for those with predicted functions associated with carbon source utilization (e.g., glycoside hydrolases and membrane transport proteins), while the annotation of candidate OGs with “hypothetical protein” annotations was further investigated. This investigation of the function annotation of the candidate OGs used additional annotation or database platforms such as InterProScan 5 (58) and CAZy (59). Finally, for the candidate OGs, the predicted cellular localization of the proteins encoded was investigated using SignalP, version 5.0 (60), and the TMHMM server, version 2.0 (61). A DNA fragment of 302 bp in OG3276 was amplified from strains Lp 56 and Nizo2831 by colony PCR with Taq DNA polymerase (Promega, Madison, WI, USA) using primers S1 (5′-TGCTGAATTGGTGTCAGTTCTT-3′) and S2 (5′-CCTTACATTGAACGTGCACAAG-3′) and was loaded onto a 1.5% agarose gel. A 100-bp DNA ladder (Promega) was used as a reference for amplicon size.

Statistical analysis.

Shapiro-Wilk normality tests were performed with a confidence interval of 95%. Two-tailed Spearman correlations were performed with a confidence interval of 95%. Both analyses were performed using GraphPad Prism, version 5.00 for Windows (GraphPad Software, San Diego, CA, USA).