Glycophenotyping of mutants of Lacticaseibacillus paracasei by lectin microarray

ABSTRACT

We previously identified a gene cluster of Lacticaseibacillus paracasei strain Shirota (YIT 9029) for cell surface long-chain polysaccharides (LCPS-1) biosynthesis, which modulates YIT 9029 activity to induce cytokine production in immune cells, and showed that a lectin microarray can be useful for distinguishing the profile of bacterial cell-surface polysaccharide (PS) structures. Therefore, we isolated disruptive mutant strains of 51 genes predicted to be involved in cell wall PS biosynthesis in YIT 9029. Their binding profiles to lectins in conjunction with their binding abilities to YIT 9029-specific monoclonal antibody (MAb) were compared. The mutants defective in binding to the MAb all had defects within the cps1 gene cluster. Some mutants were partially bound to MAb, indicating that these genes may influence the synthesis and maturation of LCPS-1. An advanced lectin microarray analyzed the cell surface glycosylation properties of YIT 9029 and its mutants. YIT 9029 bound to a rhamnose (Rha)-specific lectin CSA, and three additional lectins, including an O-glycan binder (rDiscoidin II) and two mannose (Man)-binders (rOrysata and rBanana). Lectin binding specificity was confirmed by a gene complementation assay for the cps1C gene and a carbohydrate inhibition assay. When the binding profiles of individual cps1A through cps1J knockout mutants were compared, typical and specific binding profile patterns were observed, in which some similarities in the functions of each gene could be predicted. In conclusion, the combined use of lectin microarray and a YIT 9029 mutant strain library is a powerful tool for identifying unknown bacterial gene functions related to the cell surface glycome.

IMPORTANCE

Previously, only a limited number of methods have been available for studying mutations in bacterial cell surface polysaccharide structures in relation to gene function. In this study, we focused on the lectin-binding properties of Lacticaseibacillus paracasei YIT 9029 (wild type; WT) and investigated the lectin-binding capabilities of 51 cell wall biosynthesis gene disruption strains using lectin microarrays. The results indicated that lectin-binding properties in gene-disrupted strains varied significantly with the presence or absence of long-chain polysaccharides (LCPS-1), ranging from similar to WT to distinctly different. The use of lectin microarrays in conjunction with the YIT 9029 mutant library has been shown to be a highly effective method for identifying the functions of unknown bacterial genes related to cell-surface glycomes. This innovative approach to glycophenotyping allows for the determination of cell wall glycomes associated with bacterial gene functions using lectin microarrays.

INTRODUCTION

Bacterial cell surface components, such as polysaccharide (PS) capsules, glycoproteins, or glycolipids, comprise a thick peptidoglycan (PG) layer that surrounds the cytoplasmic membrane (1) and are important signaling factors that trigger various host responses, including pathogenesis, host-microbe interaction, immune modulation, and symbiosis (2). Bacterial cell walls have unique structures related to these phenomena (3, 4). Carbohydrates in the form of capsular polysaccharides (CPS) in Gram-positive bacteria and/or lipopolysaccharides in Gram-negative bacteria are the major components on the surface of bacteria. The cell wall of Gram-positive bacteria is a complex assembly of glycopolymers, teichoic acids (TA), and proteins. These molecules possess a thick PG layer surrounding the cytoplasmic membrane (1). Bacterial cell-surface glycans differ substantially between species and strains (4). Moreover, the cell wall PS of lactic acid bacteria (LAB) contains rhamnose (Rha), and the PS structures are highly diverse, depending on the bacterial strain (5, 6).

LAB are industrially important microorganisms for fermented food production. Many LAB strains have been reported to exert beneficial effects through immune modulation of host cells. Lacticaseibacillus paracasei (formerly Lactobacillus casei) strain Shirota (YIT 9029) exerts immunomodulatory activities in vitro (7, 8) and in humans (9–16). In addition, YIT 9029 enhances NK cell activity in healthy volunteers after regular oral feeding with intact cells (17).

YIT 9029 possesses two types of CPS: long-chain polysaccharides (LCPS-1 [18] and LCPS-2 [19], formerly called PS-1 and PS-2, respectively). In previous studies, we first focused on the role of CPS in the immune modulation activities of this bacterium and revealed that certain gene-knockout strains defective in producing LCPS-1 had altered immune modulating activities toward cultured mouse macrophage-like cells (RAW264.7 and J774.1 cells), suggesting that the PS moieties of cell surface structures, including LCPS-1 (20) and LCPS-2 (19), play an important role in YIT 9029 immune modulation activities. To clarify and compare the surface structural characteristics of YIT 9029 and its mutants with different immune modulation activities, we introduced a novel approach for determining the binding profiles of these cells to lectins using liquid-phase lectin microarray technology. From this analysis, we successfully detected structural alterations on the YIT 9029 cell surface in mutants with defects in certain possible glycosylation enzymes (20).

In this study, we aimed to identify the genes of YIT 9029 that are possibly involved in the biosynthesis of genes or influence the structure of the cell surface PS moieties of the cell wall of YIT 9029 by collecting as many genes as possible and by determining the lectin-binding profiles of the gene-disrupted mutant cells in the lectin microarray system. Although databases of human and bacterial glycosyltransferase genes are being developed each year, it is still difficult to determine the role of a particular gene involved in cell wall biosynthesis. Therefore, it would be worthwhile to systematically correlate the changes in the binding profiles of these mutants with lectin probes. We identified all possible genes for the biosynthesis and modification of the cell-wall PS of YIT 9029 based on the similarity of the gene sequences with known proteins, constructed gene knockout mutants, and employed these mutants in the liquid-phase lectin microarray system.

Through this approach using lectin microarray technology, we hoped to gain new insights into the cell surface structures to evaluate other reactivity with YIT 9029-specific monoclonal antibody (MAb) (21) and the characteristics of the bacterial strain YIT 9029.

RESULTS

Identification of a cluster of genes associated with the cell wall biosynthesis of YIT 9029

In a previous study, we identified the biosynthesis genes of LCPS-1; the cps1 gene cluster consisted of 10 genes designated cps1A to cps1J (GenBank AB470649) (20). To further analyze the genes responsible for the glycosylation of cell surface molecules of L. paracasei, we attempted to identify the genes of YIT 9029 that may participate in the biosynthesis and modification of cell wall PS, TA, and PG, including enzymes for sugar conversion, glycosylation, capsular PS polymerization, and PS repeat unit transporters from the whole genome sequence of YIT 9029. Based on amino acid sequence similarity to known genes of other bacterial strains, we selected 51 candidate genes, including the 10 cps1 cluster genes described above (Table 1). RmlA, rmlC, rmlB, and rmlD are the biosynthesis genes of the Rha substance to Glc for cell-wall PSs (22), and two sets of rml cluster genes were found on the chromosome of YIT 9029 (Table 1). Unidentified genes located within possible operons of the PS synthetase genes were also included. For glycophenotype analysis, gene-deficient mutants were constructed toward these 51 genes having either deletion (Δ1932 and Δcps1C) or insertion (marked “Ω”, described in Materials and Methods) within the individual genes (Table S1). Target genes are shown as scattered images in the genome of YIT 9029 (Fig. 1).

TABLE 1.

Gene library of L. paracasei Shirota (YIT 9029) annotated from other lactic acid bacteria and L. casei BL23 (23) via amino acid sequence similarities of the cell wall biosynthesis-related genes

L. paracasei strain Shirota (YIT 9029)		Annotation of gene before genome information of L. casei BL23 was released	Annotation of genome information of L. casei BL23 (23)
Target gene			Protein ID
1	CDS0209	Lbul epsM	CAQ65363.1
2	CDS0211	Llac rgpB/rhamnosyltransferase	CAQ65364.1
3	CDS0212	Lbul epsH/glucosyltransferase	CAQ65365.1
4	CDS0213	Lrha epsB/polysaccharide biosynthesis protein	CAQ65366.1
5	CDS0214	Latilactobacillus sakei putative autolytic 1,4 beta-MurNAc ase/L. sakei putative autolytic 1,4 beta-N-acetylmuramidase	CAQ65367.1
6	CDS0215	Sthe epsI/glycosyltransferase/similar to Streptococcus thermophilus EpsI protein/repeat unit transporter	CAQ65368.1
7	CDS0216	Spne cps19bQ/probable glycosyl/rhamnosyl transferase	CAQ65369.1
8	CDS0228	Transcription repressor/transcription repressor	CAQ65380.1
9	CDS0229	Lbul epsM/glycosyltransferase/spore coat polysaccharide biosynthesis homolog yveR—Bacillus subtilis	CAQ65381.1
10	CDS0230	Hypothetical protein	CAQ65382.1
11	CDS0231	No hit	CAQ65383.1
12	CDS0661	Putative glycosyltransferase	CAQ65850.1
13	CDS0704	Dolichol phosphate mannose synthase/glycosyl transferase	CAQ65898.1
14	CDS0705	Hypothetical protein ykcB of B. sub	CAQ65899.1
15	CDS0822	Glycosyltransferase	CAQ66021.1
16	CDS0823	Galactosyltransferase homolog	CAQ66022.1
17	CDS0824	Conserved membrane protein	CAQ66023.1
18	CDS0838	Polysaccharide biosynthesis protein/polysaccharide transporter	CAQ66037.1
19	CDS0884	Lpla tagE2/glycosyltransferase/poly(glycerol-phosphate) alpha-glucosyltransferase	CAQ66090.1
20	CDS0885	Lpla tagE3/poly(glycerol-phosphate) alpha-glucosyltransferase	CAQ66091.1
21	CDS1062	Lactobacillus delbrueckii ArbX protein/glycosyltransferase	CAQ66283.1
22	CDS1063	D-galactan O antigen synthesis gene/glycosyl transferase	CAQ66284.1
23	CDS1064	Phospho-beta-glycosidase protein	CAQ66285.1
24	CDS1065	Alpha-galactosidase	CAQ66286.1
25	CDS1111	Lbul epsJ/epsM, Llac ycbH/raffinose-raffinose alpha-galactotransferase	CAQ66336.1
26	CDS1128	Llac ycbB/glycosyltransferase/stress response protein	CAQ66335.1
27	CDS1889	Glucosyl transferase	CAQ67242.1
28	CDS1892	Glycosyltransferase	CAQ67245.1
29	CDS1893	rmlD1 (dTDP-dehydrorhamnose reductase)	CAQ67246.1
30	CDS1894	rmlB1 (dTDP-glucose-4,6-dehydratase)	CAQ67247.1
31	CDS1895	rmlC1 (dTDP-dehydrorhamnose 3,5-epimerase)	CAQ67248.1
32	CDS1896	rmlA1 (glucose 1-phosphate thymidyltransferase)	CAQ67249.1
33	CDS1898	Glycosyltransferase	CAQ67251.1
34	CDS1899	Llac ycbD/UDP-glucose 4-epimerase	CAQ67252.1
35	CDS1926	Lrha epsB, Lbul epsD/capsular polysaccharide biosynthesis	CAQ67280.1
36	CDS1927	Lrha epsA/transcription regulator/membrane-bound protein	CAQ67281.1
37	CDS1932	rmlD2 (dTDP-dehydrorhamnose reductase)	CAQ67287.1
38	CDS1933	rmlB2 (dTDP-glucose-4,6-dehydratase)	CAQ67288.1
39	CDS1934	rmlC2 (dTDP-dehydrorhamnose 3,5-epimerase)	CAQ67289.1
40	CDS1935	rmlA2 (glucose 1-phosphate thymidyltransferase)	CAQ67290.1
41	CDS2708	Glycosyl transferase	CAQ68112.1
42	cps1A	Lbul epsB/capsular polysaccharide synthesis enzyme/chain length determination	CAQ67302.1
43	cps1B	Lbul epsC/capsular polysaccharide biosynthesis	CAQ67301.1
44	cps1C	Llac rgpA, Sthe cpsI/rhamnosyltransferase	CAQ67300.1
45	cps1D	Sthe cpsG/hexose transferase/glycosyltransferase	CAQ67299.1
46	cps1E	Llac ycbH, Sthe cpsI/epsG/spore coat polysaccharide biosynthesis protein/galactosyltransferase	CAQ67298.1
47	cps1F	Weakly ABC transporter protein	CAQ67297.1
48	cps1G	Galactoside acetyltransferase (lacA)	CAQ67296.1
49	cps1H	Sthe epsI/polysaccharide biosynthesis protein/repeat unit transporter	CAQ67295.1
50	cps1I	Spne cps19bQ/rhamnosyl transferase/glycosyltransferase	CAQ67294.1
51	cps1J	Lbul epsE, Sher epsE/undecaprenyl-phosphate glycosyl-1-phosphate transferase (Lactobacillus rhamnosus)/sugar transferase	CAQ67291.1

Fig 1.

Distribution map of cell wall biosynthesis-related genes in the chromosome of YIT 9029. Fifty-six genes are required for cell wall biosynthesis. Twenty-four, 10, 16, and six genes are required for the biosynthesis of cell wall polysaccharide, sugar nucleotide substrates, peptidoglycan, and lipoteichoic acid, respectively. We succeeded in making 51 gene knockout mutants of YIT 9029. The remaining five were very difficult to isolate.

Ability of the mutants to bind to YIT 9029-specific MAb

The YIT 9029-specific MAb recognizes LCPS-1 on the cell surface of YIT 9029. The abilities of the MAb to the mutants were determined by the enzyme-linked immunosorbent assay (ELISA) method (21). Based on the color intensity expressed as actual intensities of absorbance (Table S3), the mutants were largely categorized into three groups: positive, slightly positive, and negative (Table 2). Only seven of the 51 mutants were negative for binding to MAb (21), namely cps1A, cps1B, cps1C, cps1D, cps1E, cps1G, and cps1J, constituting the cps1 cluster (20). Eight mutants, including the cps1F-deficient mutant, were classified into the slightly positive group and showed decreased antibody reactivity. Thirty-six of 51 mutants were classified as positive, indicating that there was no apparent change in the cell surface PS structure due to these mutations. These mutants are thought to conserve the cell surface structure recognized by the YIT 9029-specific MAb (21). Moreover, the reactivity of YIT 9029-specific MAb (21) in our laboratory collection strains with different antibody reactivity was as follows: YIT 9021 (24), YIT 9022, YIT 9036, and YIT 9037 were slightly positive, positive, negative, and negative, respectively (Table 2).

TABLE 2.

Reactivity of L. paracasei Shirota (YIT 9029) and mutants to L. paracasei YIT 9029-specific monoclonal antibody (21)^a

Antibody reactivity	Strain
Positive	YIT 9029 (wild type), Ω0209, Ω0211, Ω0214, Ω0215, Ω0216, Ω0228, Ω0229, Ω0230, Ω0231, Ω0661, Ω0704, Ω0705, Ω0822, Ω0884, Ω0885, Ω1062, Ω1063, Ω1064, Ω1065, Ω1111, Ω1128, Ω1889, Ω1892, Ω1893 (rmlD1), Ω1894 (rmlB1), Ω1895 (rmlC1), Ω1896 (rmlA1), Ω1898, Ω1899, Ω1926, Δ1932 (rmlD2), Ω1934 (rmlC2), Ω1935 (rmlA2), Ω1937 (cps1I), Ω1938 (cps1H), Ω2708, Δcps1C/cps1C, YIT 9022
Slightly positive	Ω0212, Ω0213, Ω0823, Ω0824, Ω0838, Ω1927, Ω1933 (rmlB2), Ω1940 (cps1F), YIT 9021
Negative	Ω1945 (cps1A), Ω1944 (cps1B), Δ1943 (cps1C), Ω1942 (cps1D), Ω1941 (cps1E), Ω1939 (cps1G), Ω1936 (cps1J), YIT 9036, YIT 9037, YIT 0180

^a The reactivity of the mutants to YIT 9029-specific monoclonal antibody was determined using a sandwich enzyme-linked immunosorbent assay, as described previously (20). The resultant fluorescence intensities of wild type and mutants were classified into three types: “positive” with full to half of the color intensity as that of wild-type YIT 9029, “negative” with very weak or no color, and “slightly positive” with weak or slight color intensity. An absorbance of ≥1.0 was considered positive and <0.25 was considered negative. The absorbance of ≥0.25 but <1.0 was considered slightly positive.

Therefore, the seven genes, cps1A, cps1B, cps1C, cps1D, cps1E, cps1G, and cps1J, among the 51 genes of YIT 9029, are essential for biosynthesizing LCPS-1.

Analysis of lectin binding properties of YIT 9029 and influence of cps1C gene on lectin binding affinity

We used a newly improved lectin microarray comprising 96 lectins, including 51 additional lectins (25). The lectin microarray format is shown in Fig. 2A, and the glycan-binding specificities of the lectins used in this study are listed in Table S4. YIT 9029 (WT) showed the affinity to an O-glycan binder (rDiscoidin II), two Man binders (rOrysata and rBanana), and a Rha-binder CSA (formerly CSL) (Fig. 2B, left) (26–32).

Fig 2.

Result of lectin microarray analysis to evaluate the binding properties of YIT 9029 and the influence of the cps1C gene on lectin binding affinity. (A) Spot pattern of the lectin microarray with 96 lectins immobilized to a glass slide in triplicate (25). (B) The lectin binding profiles of YIT 9029 (WT), Δcps1C (cps1C gene is knocked out from YIT 9029), and Δcps1C/cps1C (complemented by WT-cps1C to Δcps1C) (20). YIT 9029 bound to an O-glycan binder (rDiscoidin II) and two Man binders (rOrysata, rBanana) besides a Rha-binder CSA (formerly CSL) (left). White frames indicate four types of lectins bound to WT-YIT 9029. Among the four lectins bound to YIT 9029, Δcps1C did not bind to rDiscoidin II, Orysata, and rBanana, whereas it bound to CSA. On the other hand, Δcps1C strongly bound to GlcNAc-binders WGA, LEL, and STL, but did not bind to WT-YIT 9029. Yellow frames indicate three lectins specifically bound by Δcps1C (middle). The lectin binding profile of Δcps1C/cps1C was completely recovered to that of YIT 9029 by complementing the WT-cps1C gene into Δcps1C (right). (C) Summary of the results. Each mark indicates binding affinity to lectins; ++ indicates strong, + indicates weak, and – indicates not binding. YIT 9029, Lacticaseibacillus paracasei strain Shirota; WT, wild type; Man, mannose; Rha, rhamnose.

Next, the lectin binding profile of Δcps1C deficient in LCPS-1 (20) was different from that of WT; the binding signal to rDiscoidin II and rOrysata disappeared, and that to rBanana was reduced, whereas the binding to CSA was kept. Moreover, Δcps1C strongly bound to WGA, LEL, and STL, to all of which YIT 9029 did not bind (Fig. 2B, middle) (33–35). WGA, LEL, and STL are all GlcNAc-binders and often show specific preferences for polySia and polylactosamine structures. Other lectins to which Δcps1C weakly bound will be described later.

On the other hand, the lectin-binding profile of Δcps1C/cps1C, a derivative of Δcps1C complemented in trans by the WT-cps1C gene (20), was the same as that of the WT (Fig. 2B, right). The results are summarized in Fig. 2C. These data show that the lectin-binding profile of YIT 9029 is strictly controlled by cps1C, a key gene in the biosynthesis of LCPS-1 (20).

YIT 9029 binds to three additional lectins, rDiscoidin II, rOrysata, and rBanana, in addition to CSA (26–32). Moreover, these data indicate that the binding patterns of lectins to bacterial cells are closely associated with the gene functions involved and reflect the cell surface glycosylation profile.

Carbohydrate inhibition assay

The binding of lectins to YIT 9029 was predicted to be mediated by the affinity of lectins for the carbohydrate moiety of cell surface structures. To speculate and clarify the affinity points for rDiscoidin II, rOrysata, rBanana, and CSA on YIT 9029 (Fig. 2B, left), we attempted to detect possible interference of lectin binding by simple saccharides. We chose Gal, Glu, Lac, Man, Rha, Suc, and Fuc as inhibitors because each lectin bound to YIT 9029 is known to have an affinity for some glycoproteins containing some of these saccharides (36).

The binding affinity of YIT 9029 to rOrysata was inhibited by the addition of 1 mM Gal and Man, and 50 mM Glc, Man, and Suc (Fig. 3, left). Similarly, rBanana was inhibited by the addition of 1 mM Man (Fig. 3, middle). The binding affinity of YIT 9029 to CSA was inhibited by the addition of 50 mM Rha (Fig. 3, right). The binding affinity of YIT 9029 to rDiscoidin II was not inhibited by any saccharides used in this study (data not shown). Man competitively inhibited the binding to rOrysata and rBanana, whereas Rha competitively inhibited the binding to CSA (Fig. 3). These results were consistent with the lectin binding specificities of rOrysata, rBanana, and CSA (26–32).

Fig 3.

Carbohydrate inhibition assay. D-galactopyranose (Gal), D-glucopyranose (Glc), D-galactosylpyranosyl-(β1→4)-D-Glc (Lac), D-mannopyranose (Man), L-rhamnopyranose (Rha), D-fructofuranosyl-(2↔1)-D-glucopyranoside (Suc), or (3S,4R,5S,6S)-6-methyltetrahydro-2H-pyran-2,3,4,5-tetraol (Fuc) were added in each well at a concentration of 1 mM (above) or 50 mM (below) with YIT 9029 cells (2 × 10⁹ cells/well) labeled with SYTOX orange. None means a control assay in which no carbohydrate is added to the reaction between lectins and YIT 9029. The efficiency of carbohydrate inhibition is shown as the ratio of the fluorescence intensity with carbohydrate against that of the control with no added carbohydrate. The binding of YIT 9029 to rOrysata was inhibited by Gal, Glc, Man, and Suc (left), and that to rBanana and CSA was specifically inhibited by Man and Rha, respectively (middle and right).

Effect of cps1 cluster genes for LCPS-1 biosynthesis on lectin binding affinity

To analyze the effects of the genes responsible for biosynthesis of the cell wall polysaccharides of YIT 9029, we first determined the lectin-binding profiles of the gene knockout mutants within the cps1 gene cluster, which are essential for LCPS-1 biosynthesis using microarray technology (Fig. 2A). The lectin-binding profiles of the 10 knockout mutants from cps1A (CDS1945) to cps1J (CDS1936) were unique to one another (Fig. 4). All these mutants, except for Ω1938 (cps1H) and Ω1937 (cps1I), gained the abilities to bind to LEL and STL (34). The binding profiles of Ω1938 (cps1H) and Ω1937 (cps1I) were similar to that of WT (YIT 9029) (Fig. 2B, left). Considering the fact that YIT 9029-specific MAb (21) can bind to Ω1938 (cps1H) and Ω1937 (cps1I) (Table 2) and the previous report (20), it was concluded that the LCPS-1 structure (18) did not change in these two mutants.

Fig 4.

Effect of cps1 cluster genes essential for LCPS-1 biosynthesis on lectin binding affinity. White frames indicate four lectins, rDiscoidin II, rOrysata, rBanana, and CSA, to which L. paracasei strain Sirota (YIT 9029) has specific affinities. Yellow frames indicate three lectins, WGA, LEL, and STL, to which Δcps1C has specific affinities. The lectin binding profiles of Ωcps1A, Ωcps1B, and Δcps1C; Ωcps1D and Ωcps1E; Ωcps1F and Ωcps1J; Ωcps1H and Ωcps1I, respectively, were well matched. The lectin binding profile of Ω1938 (cps1H) and Ω1937 (cps1I) was completely different from the other eight mutants of disrupted cps1 cluster genes and similar to that of WT (YIT 9029) (Fig. 2B, left). LCPS-1, long-chain polysaccharide; WT, wild type. The result of Δcps1C is the same as that in the middle of Fig. 2B. Here, the result of Δcps1C is duplicated for comparison of all 10 cps1 genes involved in LCPS-1 biosynthesis.

The binding profiles of the mutants Ω1945 (cps1A), Ω1944 (cps1B), and Δ1943 (cps1C) were similar to one another because they lost the ability to bind to rDiscoidin II and rOrysata and almost lost the ability to bind to rBanana (26–30). Moreover, the binding profiles of the mutants Ω1942 (cps1D) and Ω1941 (cps1E) were similar to each other due to the loss of binding to LEL and rBanana and gain of binding to rMalectin (29, 30, 35). In addition, these mutants showed affinity for some common additional lectins such as rF17AG, rAOL, and TJAII. The mutants Ω1940 (cps1F) and Ω1936 (cps1J) had similar binding profiles by gaining binding ability to LEL, STL, and rMalectin, while maintaining the ability to bind to four lectins to which YIT 9029 binds. However, it is obvious that the structural change in Ω1940 (cps1F) is different from that in Ω1936, because Ω1940 is partially bound to the YIT 9029-specific MAb (21), while Ω1936 is not. The mutant Ω1940 had a unique lectin binding profile among the cps1 cluster genes knockout mutants (20), showing multiple lectin binding capacities.

In contrast, the lectin-binding affinities of Ωcps1A, Ωcps1B, Δcps1C, Ωcps1D, Ωcps1E, Ωcps1F, and Ωcps1J were partially common. These eight mutants were negative or slightly positive for the YIT 9029-specific MAb (21) (Table 2). On the other hand, for three lectins, WGA, LEL, and STL to which Δcps1C was bound (Fig. 2B, middle), these eight mutants strongly bound to both LEL and STL; however, the binding to WGA was not clear in Ωcps1A, Ωcps1B, and Ωcps1G. The other four lectins, rDiscoidin II, rOrysata, rBanana, and CSA, to which YIT 9029 (WT) bound (Fig. 2B, left), had different lectin-binding properties (26–32). Although binding to CSA was observed in these eight mutants, binding affinities to rDiscoidin II, rOrysata, and rBanana were different for each strain. Ωcps1F, Ωcps1G, and Ωcps1J strongly bound to these three lectins. Ωcps1A, Ωcps1B, and Δcps1C weakly bound to those lectins. It was distinctive that Ωcps1D and Ωcps1E only bound to rBanana (29, 30).

Moreover, the lectin binding profile of Ωcps1G was significantly different from other cps1 gene-disrupted strains. Ωcps1G strongly bound to not only extensive lectins such as rDiscoidin II, rOrysata, rBanana, CSA, LEL, and STL (26–36), but also to other lectins such as LFA, RCA120, rGal9N, BPL, rF17AG, rGRFT, CCA, Heltuba, rHeltuba, rCalsepa, rAOL, rGC2, ACA, and rMalectin. The sugar-binding specificities of lectins are listed in Table S4.

Among the genes essential for LCPS-1 biosynthesis in YIT 9029 (20), lectin-binding properties differed greatly depending on the gene. These results strongly suggest that the effect on the cell surface structure changes specifically and drastically for each gene. Based on these results, it was possible to analyze the similarity of gene functions by comparing the lectin-binding properties of genetically disrupted strains of YIT 9029.

Profiling 51 mutants of YIT 9029

Based on the above results, a statistical analysis was performed to compare the lectin-binding profiles of 51 mutants of YIT 9029 (WT), two complementary strains, and the WT with 96 lectin probes (Fig. 2A, Table S4) using advanced lectin microarray technology (37). Our laboratory collection strains with different antibody reactivity of YIT 9021 (24), YIT 9022, YIT 9036, and YIT 9037, and L. casei ATCC 334 were also tested. The lectin binding properties of all tested strains were classified into three major clusters: from the top, the clusters were named 1, 2, and 3 (Fig. 5). Clustering by lectin binding was consistent with the antibody reactivity (Table 2). Clusters 1 and 2 were composed of all positive strains and slightly positive strains to YIT 9029-specific Mab (21), except for Ωcps1F. Cluster 1 consisted of Ω0213 alone, and it did not bind to the Man-binders (rBanana and rOrysata) (27–30). Cluster 2 included the largest number of members: 43 mutants and WT (YIT 9029). These mutant strains bound to an O-glycan binder (rDiscoidin II), two Man binders (rOrysata and rBanana), and a Rha binder (CSA) (26–32). YIT 9021 (24), YIT 9022, Δcps1A/cps1A, and Δcps1C/cps1C were also classified into cluster 2.

Fig 5.

Relative binding of 51 mutants of L. paracasei strain Shirota (YIT 9029) with respect to lectin binding. Lectin-binding profiles of 51 mutants of YIT 9029 (WT) and WT with 96 lectin probes were compared using advanced lectin microarray technology (37). YIT 9021 (24), YIT 9022, YIT 9036, YIT 9037, and L. casei ATCC 334 were tested. The lectin-binding signals for each strain were normalized to the highest signal. The levels of lectin-binding signals are indicated by color change from blue (low binding levels) to black (high binding levels). The lectin binding properties of all tested strains were classified into three major clusters, from above, clusters one, two, and three. Cluster one classified Ω0213 alone, and that did not bind to the mannose binder (rBanana) and weakly bound to rOrysata. Cluster two classified 43 mutants of YIT 9029 with YIT 9029 bound to an O-glycan binder (rDiscoidin II), two mannose binders (rOrysata, rBanana), and a Rha binder (CSA). Clusters one and two had mutants of all positive strains and slightly positive strains except for Ωcps1F to YIT 9029-specific MAb reactivity. YIT 9021 (24), YIT 9022, Δcps1A/cps1A, and Δcps1C/cps1C were also classified in cluster two. Cluster three included Ωcps1A, Ωcps1B, Δcps1C, Ωcps1D, Ωcps1E, Ωcps1F, Ωcps1G, Ωcps1J, YIT 9036, and YIT 9037. All these mutants had negative and only one had slightly positive strain (Ωcps1F) to YIT 9029-specific MAb reactivity. Moreover, Δcps1A/cps1A was a positive strain to YIT 9029-specific MAb reactivity (20).

Cluster 3 included Ωcps1A, Ωcps1B, Δcps1C, Ωcps1D, Ωcps1E, Ωcps1F, Ωcps1G, Ωcps1J, YIT 9036, and YIT 9037, all of which showed negative binding to YIT 9029-specific Mab (21), except for Ωcps1F, which was slightly positive to the antibody binding. Moreover, based on the lectin-binding affinities of our four laboratory collections with different antibodies, YIT 9021 (24) and YIT 9022, and YIT 9036 and YIT 9037 were categorized in cluster 2 and cluster 3, respectively.

It was confirmed that the lectin-binding properties of the gene-disrupted strains closely matched the reactivity of these mutants with YIT 9029-specific Mab (21) with some exceptions. For the present analysis, we included some gene knockouts within the same predicted operon because genes in the same operon (transcriptional unit) often produce one final product. For this, the following four possible operon genes were chosen: Ω0209 to Ω0216, Ω0228 to Ω0231, Ω0822 to Ω0824, and Ω1062 to Ω1065. The lectin-binding profiles of the mutants within the same operon showed certain similarities. Cluster 2, composed of 43 mutants, showed profiles similar to that of YIT 9029. Though all mutants of this cluster bound to Man-binders (rOrysata and rBanana), only Ω0213 cells did not bind to rBanana (18). For this reason, only Ω0213 cells were classified as a different branch from other strains, which were classified under cluster 1. The reasons why only Ω0213 was clustered in cluster 1, and all mutants of YIT 9029 bound to the Rha binder CSA (34) are discussed in the Discussion.

DISCUSSION

It is well documented that LAB used in food are beneficial for human health; however, the interaction between bacterial cell surface components and host immune cells is limited (38). We investigated Lacticaseibacillus paracasei strain Shirota (YIT 9029) for its immunological action against host cells, and it plays a critical role in the modulation of immune cells (20). YIT 9029 is a cured strain of the bacteriophage FSW of Lactobacillus casei YIT 9018 (39), and both strains react with YIT 9029-specific Mab (21). It has unique PSs called LCPS-1 (18) and LCPS-2 (19) on its cell surface, and LCPS-1 greatly affects the physiological effects of this bacterium (20).

Based on the previous reports (20, 37), we investigated how each of the genes involved in cell surface PS synthesis of YIT 9029 affects the actual structure of this bacterium by analyzing which genes are linked to which lectin-binding properties. To answer these questions, we predicted 56 genes involved in cell wall polysaccharide biosynthesis of YIT 9029 from reported bacterial glycosyltransferases, etc., and 51 genes were disrupted (Fig. 1, Table 1). As a technical challenge, lectin microarray analysis was initially developed to analyze mammalian cells. However, it was successfully applied to analyze YIT 9029 mutants using an advanced version of the 96 lectin microarray (25) (Fig. 2A, Table S4). Hence, a series of gene disruption mutants of glycosylation genes and modification enzymes of YIT 9029 were constructed, and the structure and functional relationships of the cell surface molecules of YIT 9029 were elucidated in more detail. Furthermore, we compared their binding profiles to lectins in parallel with their binding abilities to YIT 9029-specific Mab, which is our original tool (21).

We showed that the Rha-binding lectin CSA (formerly CSL) binds to YIT 9029 (37). Three new lectins were identified that bind to YIT 9029; those were an O-glycan binder (rDiscoidin II) and two Man-binders (rOrysata/rBanana) (Fig. 2B, left). The binding affinities between YIT 9029 and Man-binding lectins, rOrysata/rBanana, were inhibited by 1 mM Man (Fig. 3). It is suggested that these two lectins bind to Man of the cell surface molecules in YIT 9029, but no Man was found in the cell wall PSs of YIT 9029 (18, 19). Although these facts may seem contradictory at first glance, based on the research of Sharon et al. (40), lectins typically contain two or more carbohydrate-combining sites per molecule. Generally, galactose-specific lectins do not react with glucose, and glucose-specific lectins do not react with galactose (40). We consider that the OH group at the fourth position of the Man is particularly important in the binding between rOrysata/rBanana and YIT 9029. Therefore, even if the presence of Man is not detected in YIT 9029, it is not inconsistent for the Man-binding lectin to bind to it.

In this study, several remarkable points were observed through the lectin microarray analyses. (i) The loss of LCPS-1 did not influence the number of lectins capable of binding to mutant cells. We considered LCPS-1 as a suppressive molecule that induces cytokine production by macrophages owing to the physical masking effect of LCPS-1 on the cell surface (20). (ii) The Ω1939 (cps1G) and Ω1936 (cps1J) mutants still have the affinity to rDiscoidin II, rOrysata, and rBanana, which are possible markers for LCPS-1 biosynthesis, although these mutants do not react with YIT 9029-specific MAb reactivity, nor produce LCPS-1 (20). (iii) The similarity of the lectin-binding profiles of Ω1942 (cps1D) and Ω1941 (cps1E) mutants probably indicates that these gene products have a similar role in the synthesis and maturation of LCPS-1. (iv) The Ω1939 (cps1G) mutant dramatically changes the binding pattern of lectins, indicating big changes in the surface structure of the mutant cells. All these indications and assumptions will be addressed in future research.

The mutation that reduced antibody reactivity in YIT 9029 was confirmed to be a deletion of LCPS-1 (20). However, it is unclear whether the effects of LCPS-1 deletion on the cell surface PS structures of YIT 9029 cells are uniform or diverse. Here, we demonstrated that the mutations in the cell surface PS that occur after the deletion of LCPS-1 vary greatly depending on the gene.

Another interesting result is that while Ω0213 develops LCPS-1 on the cell wall, the lectin binding profile is different from that of YIT 9029 (Fig. 5). In this mutant, the CDS0213 gene encoding a putative PS biosynthesis protein (CAQ65366.1) (23) was knocked out, and the sequence of CDS0213 was similar to eps7I of L. casei strain BL23 (23) (Table 1). This gene is widely conserved in bacterial genomes and is classified as the Glycosyltransferase Family 32 (http://www.cazy.org/GT32.html) with known activity as α-1,6-mannosyltransferase (EC 2.4.1.232). This result was a good match between the predicted gene function and lectin-binding in the bacteria.

On the other hand, among the 51 genes chosen for disruption, several groups of genes may form operons, including Ω0209 to Ω0216, Ω0228 to Ω0231, Ω0822 to Ω0824, and Ω1062 to Ω1065. As shown in Fig. 5, we confirmed that the lectin-binding profiles of the mutants within the same operon have certain similarities.

These observations suggest that the combined use of lectin microarray technology and a mutant strain library of YIT 9029 is a powerful tool for identifying unknown bacterial gene functions in terms of cell surface glycome, which is a novel approach to glycophenotyping. It is important to clarify the relationship between the phenotype and genotype of bacterial strains. The use of lectins in combination is an effective solution when antibodies cannot be produced by the bacterial strain.

We summarized the relationship among the lectin binding profile, the predicted cell surface structures, and the lectins bound to the cell surface PS molecules to focus on wild-type YIT 9029 (cluster 2), its mutant strains Ω0213 (cluster 1), and Δcps1C (cluster 3) (Fig. 6). From left to right, each panel shows lectin-binding affinities, schematic illustration of cell-surface PS structures, and lectins bound to cell-surface PS molecules. Here, the lectin-binding profiles of YIT 9029 and Δcps1C are duplicated for comparison with Ω0213.

Fig 6.

Possible lectin-binding profiles of wild-type YIT 9029 (A) and its mutant strains Ω0213 (B) and Δcps1C (C). From left to right, each panel shows lectin-binding affinities, a schematic illustration of cell-surface PS structures, and lectins bound to cell-surface PS molecules. (A) YIT 9029 has high and low molecular mass PSs named LCPS-1 (18) and LCPS-2 (19). (B) Ω0213 is predicted to have fewer LCPS-1 molecules per cell than YIT 9029, or it has modified LCPS-1 molecules. (C) Δcps1C is completely missing LCPS-1 (20). Here, the lectin-binding profiles of YIT 9029 and Δcps1C are duplicated for comparison with Ω0213.

Panel A: YIT 9029, classified into cluster 2, has high and low molecular mass PSs named LCPS-1 (18) and LCPS-2 (19). Considering the structure of LCPS-1 (18), CSA and rDiscoidin II are predicted to bind to Rha, the acetyl group, and lactose of LCPS-1. Though LCPS-1 has Glc at the non-reducing end, none of the lectins used in this study (Fig. 2A, Table S4) recognize glucose located at the non-reducing terminal. Therefore, Man-binding lectins (rOrisata/rBanana) are predicted to bind to two Glc molecules of LCPS-1 (26–32).

Panel B: Ω0213, classified into cluster 1, is predicted to have fewer LCPS-1 molecules per cell than those of YIT 9029, or it has modified LCPS-1 molecules from the results of the ability of the mutants to bind to YIT 9029-specific MAb (Table 2; Table S3). We predict rDiscoidin II and CSA bind to mutated LCPS-1. To analyze the structure of LCPS-1 in this strain is a challenge for the future.

Panel C: Δcps1C, classified into cluster 3, is completely missing LCPS-1 (20), and we predict WGA, LEL, STL, and CSA bind to LCPS-2 (19). Considering the structure of LCPS-2 (19), CSA is predicted to recognize Rha of LCPS-2 (37), and WGA, LEL, and STL are predicted to recognize the acetyl group in glucosamine and galactosamine of LCPS-2 (33–36). The existence of LTA has been suggested in YIT 9029 (41), but its structure has not been elucidated. Regarding the lectins that bind to LTA, this is a point to focus on in the future. In conclusion, the combined use of lectin microarray technology and a mutant strain library of YIT 9029 makes a powerful tool for identifying unknown bacterial gene functions in terms of cell surface glycome. This is a novel approach to glycophenotyping, and lectin microarrays enable the identification of the cell wall glycome in relation to bacterial gene function.

MATERIALS AND METHODS

Criteria for the selection of genes from YIT 9029 and annotation of the genes

The genes predicted to be involved in the biosynthesis of cell wall-associated PSs of YIT 9029 (AB470649, LCS853105–LCS853145, and GRN 429) were selected from the in-house genome laboratories’ data (T. Sato, et al., unpublished data; Table 1, left). These were based on the similarity of at least 20% of the amino acid sequences with other bacterial proteins known or predicted to be involved in extracellular or cell wall PSs biosynthesis in the GenBank database. In addition, genes were selected to contain motifs for glycosyl transferases and glycosylation enzymes in the genome of L. casei BL23 (23). The protein ID is listed in Table 1 (rightmost).

Bacterial strains and plasmids used in this study

The bacterial strains and plasmids used in this study are listed in Table S1. Fifty-one gene knockout mutants of YIT 9029 were employed for the analysis; some were newly constructed as described below, while others were from a previous study (20). L. casei ATCC 334 (YIT 0180), a neotype strain of L. casei (42), was purchased from the American Type Culture Collection (Manassas, VA). We tested four strains from our laboratory collection with different antibody reactivities: YIT 9021 (24), YIT 9022, YIT 9036, and YIT 9037. Escherichia coli JM109 was purchased from Toyobo Co., Ltd. (Osaka, Japan) as competent cells for DNA transformation.

Gene manipulation of YIT 9029 based on the homologous recombination principle

Forty-nine mutants of YIT 9029 designated “Ω” were produced by insertion of a plasmid with the respective truncated gene fragment deleting both N- and C-terminal coding regions as described previously (Table S1). The synthetic primers used to amplify the truncated gene fragments are listed in Table S2. All mutants produced by one-step homologous recombination were cultured in De Man–Rogosa–Sharpe (MRS) medium (Becton, Dickinson and Company, New Jersey, USA) containing 10 µg/mL erythromycin (39, 43) under static culture conditions. The mutant Δ1932 (rmlD2) of YIT 9029 was constructed by two-step homologous recombination and described as Δcps1C (20). The synthetic primer set for amplification of N- and C-terminal fragments is shown in Table S2. The plasmid pRD8 to isolate the mutant Δ1932 (rmlD2 deficient) of YIT 9029 is described in Fig. S1.

Reagents and chemicals for recombinant DNA technology

The reagents and DNA technology have been described in a previous report (20). Briefly, DNA was amplified by polymerase chain reaction (PCR) using KOD PLUS DNA polymerase (TOYOBO Co., Ltd., Osaka, Japan) or TaKaRa Ex Taq (Takara Bio Inc., Otsu, Japan). Restriction endonucleases, calf intestinal alkaline phosphatase, and a DNA Ligation Kit were purchased from Takara Bio Inc. or TOYOBO Co., Ltd. Plasmid purification was performed using the Wizard Plus SV Minipreps DNA Purification System (Promega K.K., Tokyo, Japan), and DNA fragments amplified by PCR were purified using the Qiaquick Gel Extraction Kit (QIAGEN K.K., Tokyo, Japan). Custom-made synthetic DNAs were purchased from Sigma-Aldrich Japan K.K. (Tokyo, Japan).

Recombinant plasmid construction and insertion, and deletion mutagenesis

The plasmids used in this study are listed in Table S1. The basic procedures for constructing recombinant plasmids for insertional mutagenesis, deletion mutagenesis, and isolation of bacterial clones harboring these chromosomal mutations have been described previously (20). Plasmid pRD8 was constructed as follows: DNA fragments containing N- and C-terminal coding regions of the CDS1932 (rmlD2) gene were amplified by PCR using the primers shown in Table S2, and the purified DNA fragments thus obtained were digested with the respective restriction enzymes. These two fragments and pBE31 (43), digested with Kpn I and Xba I, were mixed and ligated to obtain an in-frame deletion fragment of the rmlD2 gene, which was cloned on pBE31 (43). The resulting plasmid was named pRD8 (Fig. S1). Construction of deletion mutants of YIT 9029 at cps1A (Δcps1A) and cps1C (Δcps1C), and each revertant harboring the respective wild-type (WT) gene in trans on the chromosome (Δcps1A, Δcps1A/cps1A, and Δcps1C/cps1C) was described previously (20). All primers for deletion mutagenesis were designed to enable in-frame rejoining of the N- and C-terminal peptide fragments of the gene, thereby avoiding translational interruptions within an operon. YIT 9029 was transformed with these plasmids, and erythromycin-resistant clones (39, 42) were selected. These clones contained recombinant plasmids integrated into either side of the respective gene fragments via homologous recombination. After several cycles of subculturing (one thousandth inoculation into fresh medium, followed by full growth), erythromycin-sensitive clones were screened and checked for reversion or deletion.

Culture of bacterial strains and storage

Bacterial cells were cultured in 4 mL of MRS medium with or without erythromycin (10 µg/mL) for 22–24 h at 37°C under static culture conditions. After culturing (total cells: 1 × 10⁹–2 × 10¹⁰), the turbidity of the cultures was measured using a Klett-Summerson spectrophotometer (Klett MFG, New York, USA). Because the Klett value increased linearly with culture time up to 23 h (Fig. S2), the culture time for each mutant strain was fixed to be 22 h in this study. Then the cells were labeled with 10 µM SYTOX Orange Nucleic Acid Stain (44) (Molecular Probes Co., Ltd.) in phosphate-buffered saline (PBS) containing 1% bovine serum albumin (BSA) for the lectin microarray analysis. The fluorescence intensity of 2 × 10⁸ labeled cells was measured using an ARVO X3 apparatus (PerkinElmer) after 1 h of labeling. The fluorescence of cells labeled with SYTOX Orange (44) was stabilized by freezing at −20°C for 2 weeks. Hence, we stored labeled cells at −20°C and used them within 2 weeks. Fluorescence intensities for all the tested strains were adjusted to within two times (Fig. S3). Before testing for microarray, the cells were suspended in 360 µL of PBS containing 1% BSA (PBS/BSA) (37).

Determination of the reactivity of mutants to YIT 9029-specific monoclonal antibody

The reactivity of the mutants to YIT 9029-specific MAb (21) was determined using a sandwich ELISA, as described previously (20). The resultant fluorescence intensities of WT (YIT 9029) and mutants were classified into three types: “positive” with full to half of the color intensity that WT showed, “slightly positive” with weak or slight color intensity, and “negative” with very weak or no color. An absorbance of ≥1.0 and <0.25 was considered positive and negative. The absorbance of ≥0.25 but <1.0 was considered slightly positive.

Lectin microarray hybridization

The lectin microarray was prepared as described previously (25, 37, 45). Briefly, 96 lectins were dissolved at a concentration of 0.5 mg/mL in a spotting solution (Matsunami Glass) and spotted onto epoxysilane-coated glass slides (Schott) in triplicate using a non-contact microarray-printing robot (MicroSys4000; Genomic Solutions, Ann Arbor, MI). The origins and binding specificities of the tested lectins are listed in Table S4. YIT 9029, mutants of YIT 9029, and other tested strains were labeled with SYTOX Orange (44) as described above, then were added to each well of a glass slide containing immobilized lectins (1–2 × 10⁹ cells/100 µL/well) followed by incubation at 4°C for 1 h. In this study, washing buffer at room temperature (RT) was used because no difference was observed in the binding of lectin to bacteria for the temperature of the washing buffer between RT and 4°C. Unbound cells were mildly removed by immersing the inverted lectin microarray glass slides in more than 1 L of PBS at RT for 30 min. Cells bound with lectins immobilized on a glass slide were detected using an evanescent-field fluorescence scanner. Data are shown as the ratio of the fluorescence intensities of the 96 lectins (31) relative to the maximal fluorescence intensity on the lectin microarray. Levels of lectin-binding signals are indicated by a color change from blue (low binding levels) to black (high binding levels).

Effects of simple saccharides addition on the lectin microarray assay

To determine the effects of mono- and di-saccharides addition on the binding affinity of YIT 9029 to four lectins, rDiscoidin II, rBanan, Orysata, and CSA, the following saccharides were added to the assay system at a concentration of 1 or 50 mM: D-galactopyranose (Gal), D-glucopyranose (Glc), D-galactosylpyranosyl-(β1→4)-D-Glc (Lac), D-mannopyranose (Man), L-rhamnopyranose (Rha), D-fructofuranosyl-(2↔1)-D-glucopyranoside (Suc), or (3S,4R,5S,6S)-6-methyltetrahydro-2H-pyran-2,3,4,5-tetraol (Fuc). The efficiency of the inhibition of YIT 9029 binding to these lectins is shown as the ratio of the fluorescence intensity with each saccharide to that of the control with no added saccharides (36).

Profiling 51 mutants of YIT 9029

We compared the lectin-binding profiles of 51 YIT 9029 mutants (Table 1), YIT 9029 (WT), YIT 9022, YIT 9036, YIT 9037, and L. casei ATCC 334 (YIT 0180) with 96 lectin probes (Fig. 2A, Table S4) using advanced lectin microarray technology (36). To evaluate the similarity of gene functions, cluster analysis was performed to relate the cell wall PS biosynthesis genes in YIT 9029 and the lectin binding profiles. Considering the error between the arrays, YIT 9029 used two arrays in the comparison between the arrays. Unsupervised clustering was performed by employing the average linkage method using open-source Cluster 3.0 software developed by Michael Eisen of the Berkeley Lab. A heat map with clustering was generated using Java Treeview (Fig. 5).

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/aem.01707-24.

Figure S1. aem.01707-24-s0001.tif.

Preparation of plasmid pRD8.

Figure S2. aem.01707-24-s0002.tif.

Klett units.

Figure S3. aem.01707-24-s0003.tif.

Fluorescence intensity of all tested strains.

Supplemental legends. aem.01707-24-s0004.docx.

Legends for Fig. S1 to S3.

Table S1. aem.01707-24-s0005.docx.

Bacterial strains used in this study.

Table S2. aem.01707-24-s0006.docx.

Synthetic primers for truncated amplification.

Table S3. aem.01707-24-s0007.docx.

Reactivity of L. paracasei Shirota (YIT 9029) and mutants to L. paracasei Shirota-specific monoclonal antibody (MAb).

Table S4. aem.01707-24-s0008.docx.

Sugar-binding specificities of the lectins in this study.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.