Treg–inducing capacity of genomic DNA of Bifidobacterium longum subsp. infantis

Abstract

Background:

Allergic and autoimmune diseases comprise a group of inflammatory disorders caused by aberrant immune responses in which CD25⁺ forkhead box P3–positive regulatory T cells (Treg) cells that normally suppress inflammatory events are often poorly functioning. This has stimulated an intensive investigative effort to find ways of increasing Tregs as a method of therapy for these conditions. Commensal microbiota known to have health benefits in humans include the lactic acid–producing, probiotic bacteria B. longum subsp. infantis and Lactobacillus rhamnosus. Mechanistically, several mechanisms have been proposed to explain how probiotics may favorably affect host immunity, including the induction of Tregs. Analysis of emerging data from several laboratories, including our own, suggest that DNA methylation may be an important determinant of immune reactivity responsible for Treg induction. Although methylated CpG moieties in normal mammalian DNA are both noninflammatory and lack immunogenicity, unmethylated CpGs, found largely in microbial DNA, are immunostimulatory and display proinflammatory properties.

Objective:

We hypothesize that microbiota with more DNA methylation may potentiate Treg induction to a greater degree than microbiota with a lower content of methylation. The purpose of the present study was to test this hypothesis by studying the methylation status of whole genomic DNA (gDNA) and the Treg-inducing capacity of purified gDNA in each of the probiotic bacteria B. longum subsp. infantis and L. rhamnosus, and a pathogenic Escherichia coli strain B.

Results:

We showed that gDNA from B. longum subsp. infantis is a potent Treg inducer that displays a dose-dependent response pattern at a dose threshold of 20 µg of gDNA. No similar Treg-inducing responses were observed with the gDNA from L. rhamnosus or E. coli. We identified a unique CpG methylated motif in the gDNA sequencing of B. longum subsp. infantis which was not found in L. rhamnosus or E. coli strain B.

Conclusion:

Although the literature indicates that both B. longum subsp. infantis and L. rhamnosus strains contribute to health, our data suggest that they do so by different mechanisms. Further, because of its small molecular size, low cost, ease of synthesis, and unique Treg-inducing feature, this methylated CpG oligodeoxynucleotide (ODN) from B. longum would offer many attractive features for an ideal novel therapeutic vaccine candidate for the treatment of immunologic diseases, such as the allergic and autoimmune disorders, in which Treg populations are diminished.

The gastrointestinal (GI) tract is the main portal of entry into the host for foreign agents from the external environment.¹ Because it is known that the GI tract houses nearly 70% of the entire immune system and is challenged daily by pathogens (e.g., bacteria, protozoa, fungi, viruses) and other toxic substances, the GI immune system is highly adapted to resist invading pathogens while peacefully coexisting with a robust and diverse population of friendly commensal bacteria.^2–4 In infants born by vaginal delivery, the initial seeding of maternal intestinal microflora plays a crucial role in the subsequent development of both the intestinal and peripheral immune systems that are responsible for the prevention of allergic disease.^5–7 Several studies have shown that a variety of commensal gut microbiota within the GI microbiome can favorably affect host health by directly regulating the induction of regulatory T cells (Treg), a finding that underscores the potential use of microbial therapies in treating allergic and autoimmune disorders.^8–10

Among the commensal microbiota known to have health benefits related to induction of Tregs are the lactic acid–producing bacteria B. longum subsp. infantis and Lactobacillus rhamnosus.¹¹ However, the identification of the “effector” components of these microbes that control Treg induction is a subject of ongoing debate. One of the major mechanism(s) by which these commensals stimulate Tregs to control inflammation are thought to be through extracellular microbial products that include short chain fatty acids^12–14 and polysaccharide moieties.¹⁵

Recently, genomic DNA (gDNA) and DNA oligonucleotides (in either their methylated or un-methylated forms) have been suggested as effector candidates responsible for the induction of Treg cells. To date, three epigenetic mechanisms have been associated with host immune reactivity, including DNA methylation (DNAme), histone modification, and microRNA activity, in which DNAme has been reported to be a key determinant that controls inflammation in patients with multiple sclerosis,¹⁶ rheumatoid arthritis (RA) and systemic lupus erythematosus (SLE) through the production of Tregs cells.¹⁷ It is important to stress that for each of these types of patient, DNA demethylation, and not DNA methylation, is the crucial event for Treg stability and function. Structurally, genomic bacterial DNA differs from mammalian genomic DNA and varies with species to produce highly complex effects on the immune system.¹⁸ Generally, methylated CpG moieties in normal mammalian DNA are both non-inflammatory and lacking in immunogenicity, while those associated with microbial DNA found in bacteria consists largely of unmethylated DNA which display proinflammatory properties.¹⁸

Methylated DNA, such as mammalian calf thymus (CT DNA), has long been recognized for its capacity for Treg induction. However, the epigenetic mechanism(s) responsible for Treg have rarely been explored in the context of microbial DNA methylation. A study by Notley et al. demonstrated that methylated DNA mediates Treg activation either directly or through dendritic cell interaction¹⁷. In addition to CT DNA, human placental DNA and synthetic phosphodiester or phosphorothioate oligonucleotides containing polyguanines are also reported to inhibit immune activation by bacterial DNA.^19,20 Using a plasmid as the DNA vehicle (pGT), Chen et al.²¹ reported that methylated CT DNA inhibits immune activation by E. coli DNA and that CpG methylation played a key role in this inhibitory effect by blocking the activation of nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) and Activator Protein-1 (AP-1). Consistent with these findings are the results of our previous study²², which showed that a synthetic methylated DNA 8-mer sequence (CpG type) could promote in vitro FOXP3 expression in human CD4⁺ T cell populations.

To more clearly define mechanisms by which probiotic bacterial genomic DNAs exert their anti-inflammatory influence, several suppressive (sup-) unmethylated DNA motifs, first found in mammalian telomeres, have been proposed.^23,24 A review by Bayik et al.²⁵ provides an excellent summary of these immunosuppressive oligonucleotides (ODNs), correlating their therapeutic properties with their structure and mechanisms of action. In several animal models, a suppressive ODN (TTAGGG) motif was shown to prevent the phosphorylation of Signal Transducer and Activator of Transcription1 (STAT1) and STAT4, and to inhibit B-cell proliferation.²³ Further, ODN BL07S derived from the B. longum BB536 decreased Th-2 cytokine production. Treg induction was not found in these sup-ODN-treated animals.^26,27 Even though unmethylated CpG motifs are highly conserved in the genus Bifidobactrium, no differences in enrichment of these unmethylated motifs were seen in the bacterial isolates from either clinically healthy or allergic infants.²⁸ Given the lack of Treg induction in these sup-unmethylated DNA studies, together with the absence of an inclusive correlation between allergic response and the frequency of genomic effectors in probiotics described above, anti-inflammatory effects of microbial DNA require an alternative explanation.

We hypothesize that microbiota with more DNA methylation may potentiate Treg induction to a greater degree than those with a lower degree of methylation. The purpose of the present study was to test this hypothesis by studying the Treg-inducing capacity of purified gDNA preparations from B. longum subsp. infantis and L. rhamnosus, and comparing with that seen with gDNA from pathogenic E. coli strain B.

MATERIALS AND METHODS

Cell Growth Conditions and Methods of DNA Extraction

Three bacterial strains obtained from ATCC (American Type Culture Collection, Gaithersburg, MD) were used for DNA preparation²⁹: Bifidobacterium longum subsp. infantis (ATCC 15697), Lactobacillus rhamnosus (ATCC 53103), and E. coli strain B (ATCC 11303). Cell growth conditions and methods of DNA extraction are provided in full detail in the Online Supplementary Material-A.

Isolation of CD4⁺ T Cells from Peripheral Blood Mononuclear Cells (PBMCs)

The Georgetown University Institutional Review Board IRB approved the human research protocol. After informed consent was obtained, 20 mL of EDTA-treated blood specimens were obtained from three healthy donors for CD4⁺ T cell isolation following the method provided in full detail in the Online Supplementary Material-A.

In Vitro Induction of Treg by Bacterial DNAs, FACS Analysis

Reagents used for fluorescence-activated cell sorting (FACS) are provided in Supplementary Material-A. Two sets of experiments were conducted for study of the Treg-inducing effect(s) of bacterial gDNA. The first set included a series of dose-response analyses performed at single or at dual time intervals using bacterial gDNA preparations from each of the three microorganisms at different concentrations. The second set analyzed the effects of gDNAs on Treg induction in the presence of IL-2 and TGF-β stimulating cytokines. For both experimental approaches, individual wells in 96-well plates were precoated with 50-µL aliquots that contained 10 µg/mL and 5 µg/mL of anti-CD3 and anti-CD28 incubated at 4°C overnight; 50 µL phosphate buffered saline solution (PBS) was parallel coated as methodologic controls. After washing three times with PBS, 150-µL aliquots of isolated CD4⁺ T cells (3 × 10⁵) were seeded into each well on day 1. The Treg-inducing effects of each gDNA were performed in 50-µL aliquots that contained either bacterial gDNA alone or bacterial gDNA supplemented with Treg-promoting cytokines IL-2 (60 IU)/TGF-β (50 IU).³⁰ CD4⁺ T cells were cultured in X-vivo 15 medium and incubated at 37°C, 5% CO₂. On day 4, half volumes of the stimulated CD4⁺ T-cell preparations in each culture well were transferred to a second 96-well plate that was precoated with the same anti-CD3 and anti-CD28 treatment regimen described above. After the addition of sufficient culture medium to achieve a final volume of 200 µL, both culture plates were incubated at 37°C in a 5% CO₂ incubator. FACS analyses were performed in duplicate on blood samples obtained from each of the three donors.

On day 7, the stimulated cells were harvested and stained with CD4 and CD25, and FOXP3 biomarkers as previously described.³¹ Determination of FOXP3 populations was performed using a FACSSymphony (BD) and analyzed with FlowJo software. The gating strategy for measurement of FOXP3 and CD25 is based on the same gating strategy for IgG control as provided in the Online Supplementary Fig. S1. To obviate the possible artifactual input effects of cytokines and gDNA introduced singly, in each experiment the gDNAs and cytokines were always pre-mixed and were introduced prior to gDNA stimulation.

Statistical Analysis

A two-way analysis of variance methodology with a Bonferroni posttest was used to determine statistical significance of the percentages and the numbers of Tregs cells between each gDNA or gDNA cytokine-enhanced regimen and each control without gDNA stimulation. All data are expressed as arithmetic mean ± standard error. The level of statistically significance was set at p < 0.05 (*) or p < 0.01 (**). All statistical evaluations were performed using the GraphPad Prism 5 (San Diego, CA) statistical software package.

Whole Genome Sequencing, de novo Genome Assembly, and Methylation Analysis

Whole genome sequencing was performed using single molecule, real-time (SMRT) sequencing on PacBio RS II and Sequel instruments (Pacific Biosciences, Menlo Park, CA), shown previously to identify the roles of epigenetics in eukaryotic and prokaryotic species.³²

The references for L. rhamnosus and B. longum subsp. infantis are Genbank identifier FM179322 and Genbank CP001095.1. As the entire genomic sequence of this E.coli (ATCC 11303) has not been previously reported, ancestral strain NCTC 86 with GenBank identifier LT601384.1 was used as the reference sequence.^33,34

Detection of DNA modification and motif analysis were performed by methods provided in the Online Supplementary Material-A using the PacBIO SMRT Link v7.1.0 package with the MotifMaker minimum Qmod score (–minScore) set to 100 and all other parameters were set on default.³⁵

RESULTS

Growth Characteristics and DNA Fragmentation of B. longum subsp. infantis, L. rhamnosus, and E. coli

Optimal growth conditions for each bacterial gDNA preparation were observed by using tryptic soy and MRS media for E. coli and L. rhamnosus grown under aerobic, 200 rpm shaking conditions. The optimal growth condition for B. longum subsp. infantis was assessed in MRS-cysteine cultures carried out under anaerobic stationary conditions, as shown in Fig. S2A & S2B. Genomic DNA (gDNA) samples from each bacterium demonstrated high quality (>10 kb) on agarose gel (Fig. S2C) and electropherogram analysis (Fig. S2D). Results of Growth Characteristics and DNA Fragmentation of B. longum subsp. infantis, L. rhamnosus and E. coli are provided in full detail in the Online Supplementary Material-B.

Stimulatory Effects of gDNA from B. longum subsp. infantis on Treg Induction

The quantity of Treg induction was assessed by flow cytometric analysis by using CD4⁺CD25^high/FOXP3⁺ as a signature biomarker. The effects of bacterial gDNA on Treg expansion on day 7 were measured in a series of experiments where the concentrations of anti-CD3 (α-CD3), anti-CD28 (α-CD28) antibodies, and IL-2/TGF-β were varied. The baseline values in CD4⁺ cells in the absence of gDNA and cytokines served as the first methodologic control (CD4 co.) and were found to be less than 1% as shown in Fig. 1A.1. The second control of Treg cell induction consisted of CD4⁺cells stimulated by α-CD3 and α-CD28 (AB co.) (Fig. 1A.2). The third control consisted of Treg induction by IL-2 (60 IU)/ TGF-β (50 IU) measured in the presence of α-CD3/ α-CD28 (AB-CK co.) (Fig. 1A.3). A significant increase of CD4⁺CD25^high/FOXP3⁺ expression was seen in the IL-2/TGF-β treated cells (Fig. 1A.3) in contrast to that seen in the AB co. (Fig. 1A.2).

Figure 1.

Flow cytometric analysis of the effects of gDNA on T regulatory cell induction. Representative flow cytometric assessments expressed as percentages of CD4⁺CD25^high/FOXP3⁺ (T regulatory) cells induced by different concentrations of gDNAs. (A) Three controls for gDNA-Treg response: un-activated CD4⁺ cells (A.1); the CD4⁺ cells activated by αCD3/αCD28 (A.2) as control for CD4⁺ cells treated with gDNAs alone; and αCD3/αCD28 + IL-2/TGF-β control (A.3) for gDNA stimuli with IL-2/TGFβ. (B) Sampled CD4⁺CD25^high/FOXP3⁺ percentages with high dose (50 µg) and low dose (0.08 µg) of each bacterial gDNA stimulus without two cytokines (top panel) and with two cytokines (bottom panel). (C and D) The overall dose-response results of Treg induction by bacterial gDNA stimulation. The bacterial gDNA-Treg response (percentage) at gDNA concentration ranging from 0.0625 µg ∼ 50 µg without (C) and with IL-2/TGF-β (D) were averaged from three individual CD4⁺ samples derived from three healthy subjects (n = 3). B. longum represents B. longum subsp. infantis in all figures. The dash lines in C & D represent baseline values of anti-CD3 and anti-CD28 antibody (AB) and AB+ cytokine (AB-CK) treatments, respectively. Statistical analyses versus AB co. (C) and AB-CK co. (D) were performed by two-ANOVA analysis. Data are expressed as mean ± SEM; “*”: p < 0.05; “**”: p < 0.01 and “***”: p < 0.001.

A set of dose-response experiments was next performed to examine the effects of gDNA on Treg induction. The gDNA-Treg responses of each bacterium at concentrations of 50 µg and 0.08 µg from one CD4⁺ T cells donor are shown in Fig. 1B and the gDNA-Treg responses at doses of concentrations of 50 µg, 20 µg, 10 µg, 2 µg, 0.4 µg, and 0.08 µg in CD4⁺ T from three donors are displayed in Figs. 1C. When B. longum subsp. infantis gDNA was used at concentrations of 50 µg and 20 µg, Treg percentages increased to 23.3% ± 11.6% and 18.3% ± 10.1% (mean values ± SEM), respectively, values that were significantly higher than those seen with the gDNA-absent (AB co.) (p < 0.001 and p < 0.05), as shown in Fig. 1C. The gDNA-Treg–induced effects at concentrations of 50 µg and 20 µg of B. longum subsp. infantis gDNA were greater than that obtained with the IL-2/TGF-β treatment (AB-CK co.), although the differences were not statistically significant. However, at gDNA concentrations of <10 µg, these DNA-Treg effects for B. longum subsp. infantis were rapidly diminished (Fig. 1C).

The dosage effects on Treg induction for L. rhamnosus and E. coli revealed a nonlinear correlation between DNA concentrations and Treg effects as well (Figs. 1B and 1C). Paradoxically, unlike the Treg-augmenting responses seen at high doses of B. longum subsp. infantis gDNA, the DNA-Treg effects at 50 µg gDNAs from L. rhamnosus and E. coli were reduced. (Fig. 1B). The maximum Treg effects of L. rhamnosus and of E. coli were 9.0% ± 2.5% and 10.7% ± 4.0%, respectively, at a gDNA dose of 20 µg which were not statistically significant from those observed with AB co. (Fig. 1C). Although none of the three bacterial gDNAs displayed a linear Treg stimulatory response, a dose of 20 µg gDNA for all three and ≥20 µg for B. longum subsp. infantis seemed to provide optimal effective concentrations for Treg activity (Fig. 1C). These results clearly demonstrated that the intact gDNA from B. longum subsp. infantis was a potent Treg inducer, uniquely different from the gDNA of L. rhamnosus and E. coli, and only operative at a dose threshold ≥ 20 µg gDNA.

Different Treg-Inducing Effects of Bacterial gDNAs in the Presence of IL-2/TGF-β

The combination of two cytokines (IL-2 and TFG-β) has been widely recognized for its potent in vitro Treg-inducing effects.^36,37 In the absence of gDNA, treatment with two cytokines (AB-CK co.) promoted an 11.2% Treg-induction rate (Fig. 1C) which was not able to be sustained when low doses of each gDNA was introduced (Fig. 1D). In contrast to a consistent linear increase of Tregs cells induced by B. longum subsp. infantis at doses ranging from 20 µg to 50 µg, a slight increase of IL-2/TGF-β–Treg induction was also seen at 20 µg of E. coli gDNA (Fig. 1D). Of interest, the inherent capacities of L. rhamnosus and E. coli gDNAs for Treg induction were improved in the presence of IL-2/TGF-β, particularly with E. coli gDNA. With this species, gDNA doses of <10 µg, together with IL-2/TGFβ resulted in 2 ∼ 4-fold higher levels of Treg induction.

The Inhibitory Effect of L. rhamnosus and E. coli gDNAs, but not B. longum subsp. infantis gDNA, on CD4⁺ T cell Proliferation

During flow analysis, we observed that the quantity of total numbers of CD4⁺ T cells collected for final analysis was significantly decreased when 50 µg L. rhamnosus or E. coli gDNA was used (Fig. 1B). When IL-2/TGF-β was available, although the percentage of Treg cells at 50 µg E. coli gDNA was comparable to that seen at higher doses of B. longum gDNA, the absolute number of Treg cells was greatly diminished (bottom panel, Fig. 1B). For better evaluation of the gDNATreg effects, the absolute number of CD4⁺CD25^high/FOXP3⁺ cells in each experimental condition was also determined. As shown in Fig. 2A, the absolute number of Treg cells were significantly higher in B. longum gDNA at concentrations ranging from 2 to 50 µg than those seen in the AB co. (p < 0.001 ∼ p < 0.05), findings also seen with L. rhamnosus gDNA concentrations of < 2 µg. In the presence of IL-2/TGF-β, the cytokine-induced Treg effects remained only at doses of 50 µg of B. longum subsp. infantis gDNA and < 2 µg of E. coli gDNA. Nevertheless, the total numbers of CD4⁺CD25^high/FOXP3⁺ were lower than AB-CK co. (p < 0.05 in Fig. 2A).

Figure 2.

Inhibitory effects of gDNAs on Treg cell differentiation (A) and CD4⁺ cell expansion (B). The total numbers of induced CD25^highFOXP3⁺ Treg cells and CD4⁺ cells are shown which resulted from bacterial gDNA (s) at different concentrations (n = 3). The gDNA concentrations and cytokines treatments used were the same as those shown in Fig. 1. The controls used for statistical analyses included gDNA alone and gDNA+ cytokines. Data were performed by two-ANOVA analysis and are expressed as mean ± SEM. “*”: p < 0.05; “**”: p < 0.01 and “***”: p < 0.001. C. Representative photomicrographs showing CD4⁺ cell expansion in response to high (50 µg), intermediate (20 µg) and low (0.08 µg) doses of each bacterial gDNA on day 7 cultures.

To test whether the inhibitory effect on Treg induction with high doses of gDNA from L. rhamnosus or E. coli was a consequence of overall inhibition of CD4⁺ T cell proliferation, the total number of CD4⁺ T cells and morphologic characteristics of polarized CD4⁺ T cells under various treatment regimens were next examined. A quantitative decrease was observed in the total number of CD4⁺ T cells in cultures treated with a high dose of gDNA from L. rhamnosus (>20 µg) or E. coli (>50 µg) (Fig. 2B). In contrast, the suppression of CD4⁺ proliferation was less robust, with a high concentration of B. longum susp. infantis gDNA. The CD4⁺ T cell proliferative effects were further improved in the presence of IL-2/TGF-β cytokines, with low doses of L. rhamnosus or E. coli gDNA treatment but not with B. longum gDNA stimulation (Fig. 2B). On microscopic examination, the untreated CD4⁺ T cell (i.e., those without treatment with α-CD3/α-CD28, gDNA, and IL-2/TGF-β cytokines) remained rounded (Fig. 2C) and showed no proliferative effects, with ∼80% viability at day 7. This viability rate was similar to that seen when CD4⁺ T cell were treated with low or intermediate concentrations of gDNAs. In the presence of cytokines, however, the viability of CD4⁺ T cell was generally improved (>90%) even in the presence of high doses of gDNA(s) from L. rhamnosus or E. coli. Morphologically, the CD4⁺ T cell expansion at low concentrations (i.e., <10 µg) of any of the gDNA moieties led to the formation of larger clusters than those seen at higher concentrations of gDNA (Fig. 2C). These morphologic events with the formation of large clusters of CD4⁺ T cells were accompanied by a quantitative increase in the total number of CD4⁺ T cells, as seen in flow analysis (Fig. 1B). In contrast to the variable sizes of CD4⁺ T cells with poorly formed clusters produced by inputs of 20 to ∼50 µg E. coli gDNA (Fig. 2C), CD4⁺ T cells to gDNA(s) from Lactobacillus and B. longum were more uniform and more easily sustained the formation of large clusters. Thus, our results revealed a less inhibitory effect of B. longum subsp. infantis gDNA in high doses on CD4⁺ T cells proliferation than with L. rhamnosus or E. coli, which was inhibited by high doses of gDNAs alone (Fig. 2B). Therefore, we suggest that the Treg induction produced by low doses of L. rhamnosus or E. coli gDNA is likely the direct consequence of CD4⁺ T cell expansion. Cytokine treatment partially reversed the inhibitory effects of gDNA from L. rhamnosus or E. coli on CD4⁺ T cell proliferation when used at concentrations ≤20 µg gDNAs (Fig. 2B).

These observations supported the participation of some specific component of B. longum subsp. infantis gDNA that directs or targets the Treg-induction pathway. This unidentified component of B. longum gDNA would have to be in sufficient quantity to protect the inhibitory effects of gDNA on Treg induction as seen in L. rhamnosus or E. coli. The opposite cytokine-induced Treg inductive effects of E. coli gDNA at concentrations of ≤2 µg (Figs. 1B and 2A), noted above, however, infer the participation of other indirect effects of E. coli gDNA on Treg induction.

Highly Induced Tregs by B. longum subsp. infantis gDNA Sustained with Dual gDNA Treatment

Although the percentages and the total number of Treg cells were effectively induced with 50 µg of gDNA from B. longum susp. infantis gDNA alone or gDNA plus IL-2/TGF-β treatment, the total number of CD4⁺ T cells stimulated with this high concentration of B. longum subsp. infantis gDNA was significantly less than CD4⁺ T cell responses treated with IL-2/TGF-β alone (AB-CK co.) (p < 0.05) (Fig. 2B). Therefore, the characteristics of high doses of B. longum subsp. infantis DNA with or without two cytokines on Treg induction were further evaluated by using a dual stimulatory regimen that involved temporally repeating gDNA stimulation on day 1 and day 4.

Notably, the second gDNA input on day 4 was introduced in a proinflammatory environment that presumably resulted from stimulation of CD4⁺ T cells by gDNA on day 1. In these experiments, an intermediate concentration of B. longum subsp. infantis gDNA (20 µg) was also included to avoid possible inhibition of CD4⁺ T cell expansion that was observed with 50 µg gDNA(s) from L. rhamnosus or E. coli. As shown in Figs. 3, A and B, in the absence of IL-2/TGF-β treatment, the highly inducible Treg activity by B. longum subsp. infantis gDNA was sustained only when the second dose was 50 µg gDNA (p < 0.001, versus to AB co.). Unexpectedly, the total CD4⁺ T cells were not suppressed but actually were increased 4-fold and even, under microscopic clusters, (Fig. 3C). However, at a dose of 20 µg gDNA, the Treg-inducible effects of B. longum subsp. infantis gDNA were diminished after the second gDNA stimulus, with or without the presence of cytokines. After a second dose of either 20 µg or 50 µg of B. longum subsp. infantis gDNA, the suppressive effects on Treg activity were more apparent than when the gDNA was given simultaneously with cytokines, offering a dual counterbalancing pro-inflammatory–anti-inflammatory effect of B. longum subsp. infantis gDNA mediated by cytokine modulation.³⁸

Figure 3.

The percentages of stimulated CD25^highFOXP3⁺ cells following single or double stimulation with B. longum subsp. infantis gDNA with or without IL-2/TGF-β. (A) Representative flow cytometric analyses of 20 or 50 µg of B. longum gDNA at single (on day 1) or double stimulation (on day 1 and day 4) were compared. (B) The double gDNA treatment at 20 µg of B. longum gDNA (2^nd_DNA) induced less Treg than single DNA (1^st_DNA), resulted in no significance versus AB control in terms of Treg percentages (n = 3). With IL-2/TGF-β, 50 µg of B. longum gDNA (2^nd_DNA_CK) shown the same inhibitory effects as 20 µg of B. longum gDNA observed above. Data are represented as mean ± SEM; statistical analyses were performed by one-tail t-test. *p < 0.05 and **p < 0.01. C. Representative photomicrographs showing morphological characteristics of gDNA-treated CD4⁺ cells following one or two B. longum gDNA stimulations with or without IL-2/TGF-β treatment.

Distribution of CTCpGAG Motif Was Found in gDNA of B. longum subsp. infantis

The detection of DNA methylation was performed through SMRT sequencing of all three strains, resulted in the generation of high-quality complete genomes and resultant de novo assemblies, which were fully co-linear with their reference genomes. The sequencing data from each bacterial species and generating contigs are shown in Table 1. In general, the GC content of B. longum subsp. infantis (59.86%) is higher than that of E. coli (50.78%) or L. rhamnosus (46.76%). Because we previously reported that a methylated CpG motif was responsible for Treg induction, we focused on the m5C CpG methylated motif.

Table 1.

Summary of the number of 5-methylcytosine (m5C) methylation sites in three bacterial species

Strains	Reference Sequence	Genomic Size	G + C (%)*	No. m5C		Fraction of m5C/bp¶
				De novo#	Ref.§
Bifidobacterium longum, ATCC15697	CP001095.1	2,862,748	59.86	11066	10992	0.00384
Lactobacillus rhamnosus, ATCC53103	NC013198.1	3,010,111	46.76	10432	10647	0.00347
Escherichia coli strain B, ATCC11303	LT601384.1	5,144,392	50.78	22852	23461	0.00444

^*Derived from genomic DNA assembly found in this study.

^#m5C events detected in genomic DNA assembly of each strain found in this study.

^§Events detected in genomic DNA sequence of each strain from the reference sequence in GenBank.

^¶Fraction of m5C modification per base in each genomic DNA set.

The presence of Treg effector ODN1 (m5C motif), demonstrated in our previous study,²² was investigated among three bacterial methylomes via CC/CTCGAG and CTCGAG/GG tags. There are three methylation types that result from base modifications in bacterial genomes derived from a series of methyltransferases that recognize unmethylated motifs: (1) 6-methyladenine (m6A), (2) 4-methylcytosin (m4C), and (3) 5-methylcytosine (m5C) (Fig. 4A) We found two positive hits in B. longum subsp. infantis, with CpG placed precisely at the middle of the motif context (21st character), as shown in Table 2. In addition, two more suspected hits were also found in the B. longum subsp. infantis genome, which were designated as m4C types by the SMRT analysis in silico model. This methylated motif is similar but not identical to DNA sequencing of BioAORF1145p binding (categorized as ytCpGar).³⁹ In that report, two m5C motifs, ggCpGcc and srtCpGays, were identified from B. longum subsp. infantis, which were identified by the DNA binding specifically to type II MTases (methyltransferases), BioAIP, and BioAORF1145p, respectively. By using the same approach, a search for this m5C motif found no hits in L. rhamnosus or E. coli.

Figure 4.

Results of the genomic DNA sequencing studies of B. longum subsp. Infantis, L. rhamnosus and E. coli genomes. (A) Schematic representation of the molecular structures of common base modifications in bacterial DNAs. (B) Scatter plot results of DNA modifications in L. rhamnosus (B1 & B2) and E. coli (B3 & B4) with per-strand coverage (left panel) and with distribution of modified DNA bases in variable modQV (modification quality value) (right panel). At each genomic position, modQV was computed as the −10 log (P-value) for each modified base position, based on the distributions of interpulse durations (IPD ratios) versus to in silico kinetic reference values from all reads covering this position. The color specified the nucleotide bases that have been detected positive for the modification. Adenine is colored in red, guanine is colored in blue, cytosine is colored in green, and thymine is colored in purple. (C) Scatter plot results of DNA modifications (C1 & C2), m5C motif analysis (C3) and distribution of identified m5C motifs in B. longum (C4). The descending greyish marker on the left of the Fig. indicates the possible range where m5C and m4C contexts are located. (C4) The distribution of three types of m5C motifs (RGCGGCGCC, YGCGGCGCC and CCCTCGAG) are displayed by colored dots.

Table 2.

The list of CTCpGAG sequence contexts in the B. longum genome

	Context	Type	Score (threshold >20)	Coverage	Interpulse Duration Ratio	Identification Qv
Bifidobacterium longum subsp. infantis	Context = GCCCTGGAGAAGGAAACCCTCGAGACCCAGGACGAGGCCCT	Modified base	32	191	1.57	—
	Context = GCCCGAGGAGCCCTTGCCCTCGAGGACGCGTCGGGCGCATG	m5C	98	180	2.80	56
	Context = GATGACGCTGTTCAGGCCCTCGAGGACCGCGTTGGTGCGCC	m4C	126	143	2.79	26
	Context = CTTGACGAGAACATGTCCCTCGAGGACGCGGCAAGGAACGC	m4C	116	143	3.32	21

m5C = 5-methylcytosine; m4C = 4-methylcytosine.

The B. longum subsp. infantis Genome Contains a Unique Methylated CpG Motif

All kinetic data were deposited in Sequence Read Archive (SRA) under accession number PRJNA609353. Having shown no quantitative advantage in m5C content in B. longum subsp. infantis genome, we next directed our research strategy to asking if a qualitative difference of motif could provide a more discriminative biomarker for the putative moiety responsible for Treg induction. Analysis of the methylated motifs of the three bacteria by SMRT v7.0.0 revealed that each bacterium had its own unique methylation signature and that there were no overlapping motifs (Fig. 4). For example, B. longum subsp. infantis carries m4C, m5C, and m6A; L. rhamnosus has m5C and m6A; and E. coli strain B displays only m6A (Fig. 4B3). Interestingly, although the gDNA of E. coli quantitatively has a higher GC content; no m5C motifs could be identified (Table 3). Inversely, a total of 39,180 m6A motifs were identified in the E. coli genome, which were found to be 50- and 10-fold higher than those found in L. rhamnosus and B. longum subsp. infantis genomes.

Table 3.

Base methylated motifs in three bacteria

	Motif Name	Motif String*	Type	CentPos#	nDetect§	nGenome¶	Fraction||	Mean Score	Mean Ipd Ratio
B. longum subsp. infantis	Bl-T1	GAGGAC	m6A	5	1486	1490	1.0	241.5	5.32
	Bl-T2	RGCGGCGCC	m5C	3	42	160	0.38	144.6	2.54
	Bl-T3	SNCNNTGGCGCC	m4C	3	40	203	0.20	133.9	3.04
	Total	m6A: 1486; m5C: 42; m4C: 40
L. rhamnosus	Lr-T1	YACAGC	m6A	4	1734	1741	1.0	633.0	6.3
	Lr-T2	GTRAAT	m6A	5	1797	3227	0.56	141.2	2.2
	Lr-T3	GNNCNNGNGANCNCGT	m5C	14	113	276	0.41	166.9	1.9
	Lr-T4	GCGTNNNNSNGCGNG	m5C	12	18	120	0.15	162.6	1.6
	Lr-T5	CCNNGGNCTNNTANNC	m5C	16	5	5	1.0	191.2	1.6
	Lr-T6	HNCTNCNCNNGCNNNNGC	m5C	12	19	119	0.16	169.1	2.0
	Lr-T7	CGNGNNNGNNCANGCT	m5C	15	19	123	0.15	158.0	1.5
	Total	m6A: 3531; m5C/CpG motifs: 131; m5C/CpN motifs: 43
E. coli strain B	Ec-T1	GATC	m6A	2	37806	37896	1.0	342.9	4.9
	Ec-T2	AGCANNNNNNNNTCA	m6A	4	686	692	1.0	303.2	4.7
	Ec-T3	TGANNNNNNNNTGCT	m6A	4	688	692	1.0	308.9	4.5
	Total	m6A: 39180

m6A= 6-Methyladenine; m5C = 5-methylcytosine; m4C = 4-methylcytosine; CentPos = center position of methylated base; nDetect = number of methylated motif detected in methylome; nGenome = number of motifs found in genome; Fraction = methylated motifs/total motifs; Mean Score = average score of nDetect; MeanIpdRatio = mean of Interpulse duration ratio; Bl = B longum subsp. infantis; Lr = L rhamnosus; Ec = E coli.

^*Underscored base denotes methylated base in each motif.

^#Central position of modified base in each motif.

^§No. motifs with designated methylated type in genome.

^¶No. motif sites in genome.

^||Fraction of positive methylated type.

For a m5C-type methylation, differences were found in the amounts and sequence motifs between B. longum subsp. infantis and L. rhamnosus. Of 174 m5C contexts that were detected in L. rhamnosus, with an average IPD ratio that ranged from 1.5 to ∼2.0, 131 motifs contained CpG modifications (Lr-T3 and Lr-T4) (Table 3). Unlike the diversity of m5C motifs seen in L. rhamnosus, 42 contexts that contained the m5C type of CpG modifications were identified from the B. longum subsp. infantis genome, with a mean IPD ratio >2.0. All of these contexts (Bl-T2) in B. longum subsp. infantis shared the same m5C motifs, GCpGGCGCC; 33 of which were R (A/G) GCpGGCGCC and 9 were Y(C/T)GCpGGCGCC, as shown in Table 3. Although this CpG subtype of m5C contexts in B. longum subsp. infantis was present in insufficient quantity that did not allow detection of clustering in the default motif analysis platform (Fig. 4C3) its presence was found to be evenly distributed in the genome Fig. 4C4. The low frequency of the m4C type (0.005%) of base modification (SNCNNTGGCGCC), with a similar mean score of m5C, evaded detection by its inability to form clusters (Fig. 4C3). The overall m5C motifs accounted for 0.025% or 0.005% of all cytosines in L. rhamnosus and B. longum genomes, respectively.

DISCUSSION

Several studies have shown that a variety of commensal gut microbiota can favorably affect host health by directly regulating the induction of Tregs.^12–15 In the present study, we evaluated the Treg-inducing capacity of the gDNA from B. longum subsp. infantis, L. rhamnosus, and E. coli strain B, and correlated this with the degree of methylation at the DNA level. We demonstrated that gDNA from B. longum subsp. infantis is a potent Treg inducer in the absence of IL-2 and TGF-β co-stimulation, operating in a dose-dependent manner with a dose threshold of 20 μg gDNA. By contrast, the gDNA of E. coli at low dose stimulates Treg induction in the presence of IL-2 and TGF-β. Unlike the other two bacteria, L. rhamnosus gDNA demonstrated only a weak Treg-inducing effect, both in the presence and absence of IL-2 and TGF-β stimulation.

The immunomodulatory effects of probiotics have long been investigated in clinical studies and animal models.^40,41 The study by Jeon, et al,⁴¹ in which BALB/c mice were orally treated with either B. breve or L. casei, revealed that only the B. breve but not the L. casei-treated mice induced intestinal IL-10-producing Tr1 regulatory cells. One of the major mechanism(s) by which these commensals stimulate Treg cells to control inflammation are thought to be through extracellular microbial products that include short chain fatty acids (SCFAs)^12–14 and polysaccharide (PSA) moieties¹⁵. Although the short chain fatty acids and polysaccharides have been shown to induce FOXP3 expression in CD4⁺ T cells and to promote the function of FOXP3⁺ Tregs in intestinal lamina propria,⁴² a recent study from our group demonstrated the relevance of methylated DNA as another molecular determinant of Treg induction²². These previously published results are consistent with the findings described here, which demonstrate the selectively greater Treg induction by gDNA from Bifidobacterium longum subsp. infantis, than from Lactobacillus rhamnosus or Escherichia coli strain B.

The results of the present study offer an alternative mechanism for Treg-inducing effects in commensals related to the methylation status of microbial gDNA itself. To counteract inflammation-inducing effects of hypomethylated DNA-induced allergic and autoimmune conditions, the immunosuppressive effects of Tregs induced by methylated gDNA moieties of Bifidobacterium longum subsp. infantis now have a clinical rationale and are supported by the results of the present study. Here, we not only showed a greater capacity for Treg induction of gDNA from B. longum subsp. infantis, but we correlated this activity with its methylation status. Our methylome sequencing studies demonstrated a unique bacterial motif (B1-T2, RGCmeGGCGCC) in the genome of B. longum subsp. infantis, which was not seen in the genome of L. rhamnosus or E. coli strain B (Table 3). The absence of methylated CpG motifs in the L. rhamnosus and E. coli genomes, which were shown to lack potent Treg induction, strongly supports the possibility of a methylated CpG-Treg mechanism by bacterial gDNAs. Further in vivo investigations will be required to unambiguously validate the correlation of the relative enrichment of unique methylated CpG motifs of B. longum subsp. infantis gDNA with Treg induction. Treg induction was associated with enrichment of four unmethylated TTAGGG suppressive motifs in a study by Bode et al.⁴³ and other investigators,^23,26,27 although epigenetic mechanisms were not considered in those studies. Once we factored in the effects of the inflammatory CpG motifs found in E. coli genomic data, we were obliged to suggest that the interaction between microbial DNA effectors and the immune system are more complicated than we first suspected.

In the present study, the microbial gDNA-Treg–inducing mechanisms were affected by various ancillary cytokine constituents at different concentrations. We found some degree of Treg induction at low concentrations of E. coli gDNA in the presence of IL-2/TGF-β. However, high doses of gDNA from E. coli or L. rhamnoses exerted a “toxic effect” on CD4⁺ proliferation. In contrast to the immunoinflammatory function of unmethylated CpG from E. coli, there also is compelling evidence to suggest an anti-inflammatory capacity for suppressive motifs identified from Lactobacillus DNA.⁴⁴ The sup-motifs also seem to explain the suppressive effects of high concentrations of L. rhamnosus gDNA on CD4⁺ T cell proliferation observed in this study although this is presumably carried out by a different set of mechanism(s). Analysis of our data also explained the beneficial effects of Bifidobacterium-containing probiotics in the treatment of allergic rhinitis report both human and mouse models by restoring Th2/Treg imbalance and eliminating gut dysbiosis.¹¹

The discovery of these critical genetic or epigenetic mechanisms also is now beginning to find clinical application. For example, ODNs that contain unmethylated CpG motifs have been used as adjuvants for immunoregulation and immune-enhancing responses due to their inflammation-promoting potential. By using this approach in studies of immunotherapy, Creticos et al.⁴⁵ showed that conjugated CpG ODN and Amb a1 in adults with allergic rhinitis alleviated acute allergic responses, increased Amb a1-specific Ig G and reduced levels of Amb-specific Ig E. Collectively, analysis of these data supports the idea that a balance between stimulatory and regulatory DNA motifs contributes to the induction of controlled immune responses seen in many immunologically mediated clinical conditions such as allergic and autoimmune disorders, and underscores the potential use of microbial therapies in treating these conditions.

CONCLUSION

With regard to clinical applications involving therapies that increase Treg induction, our findings add some additional depth to our understanding of the complex host-microbiota interactions involved and offer an alternative approach in which the proinflammatory capacity of bacterial DNA can be harnessed by its endogenous content of suppressive methylated motifs. A report by Martino et al.⁴⁷, describes the role of epigenetic modifications, including DNA methylation, in mechanisms of allergic disease and food allergy and suggests they may provide ideal biomarkers of modifiable disease pathways important in the formulation of new therapeutic modalities. If the in vitro Treg inductive properties of a methylated CpG ODN can be replicated in in vivo animal models and, ultimately, in humans, this could provide a basis for novel therapeutic development of methylated CpG ODN vaccines.⁴⁶ Because of its small molecular size, low cost, ease of synthesis, lack of immunogenicity, and wide applicability for use through a variety of systemic and mucosal delivery routes (e.g., oral, nasal, respiratory, and dermal), a methylated CpG ODN offers many attractive features for an ideal vaccine candidate for the following: allergic rhinitis, asthma, and atopic dermatitis and for other autoimmune and neurodegenerative disorders (Table 4).

Table 4.

Potential clinical applications of a therapeutic vaccine that contains a unique methylated Bifidobacterium longum CpG motif

Disease Example	Route of Administration	Mechanism of Action (related references)
Allergic disease		Treg induction (48–51)
Allergic rhinitis	Nasal
Asthma	Inhalation
Food allergy	Oral
Atopic dermatitis	Oral or topical
Autoimmune disorders		Treg induction (52–54)
Systemic lupus erythematosus	Intramuscular
Rheumatoid arthritis
Hashimoto thyroiditis
Neurodegenerative		Treg induction (55, 56)
Alzheimer disease	Intramuscular
Multiple sclerosis	Nasal
Parkinson disease

Treg = Regulatory T cell.