Approximately 10¹³ microorganisms, collectively known as microbiota, reside in the human gastrointestinal tract (Sender et al., 2016) and have increasingly received well-deserved attention regarding their deep impact on the physiology and well-being of the mammalian host. The microbiota composition is characterized by vast inter-individual variations and is furthermore influenced by numerous factors, including genetics (Org et al., 2015), mode of delivery (Biasucci et al., 2010), age (Palmer et al., 2007), diet (Ericsson and Franklin, 2015), hospitalization (Penders et al., 2006), exposure to pathogens and/or antibiotics (Buffie et al., 2012; Carding et al., 2015). It has been shown that bacteria belonging to the phyla Firmicutes, Bacteroidetes, Proteobacteria, Actinobacteria, and Fusobacteria comprise the vast majority of the mammalian intestinal microbiota (Eckburg et al., 2005; Palmer et al., 2007; Hill et al., 2012) and both bacterial density and diversity increase from the proximal to the distal gut (Sommer and Backhed, 2013). Firmicutes are Gram-positive bacteria, which include the large class of Clostridia and the lactic acid bacteria (Zhang et al., 2015), while Escherichia coli (Gram-negative, rod-shaped bacteria belonging to the family Enterobacteriaceae and phylum Proteobacteria) represent the predominant facultative anaerobe member of the mammalian gastrointestinal tract (Finegold et al., 1983; Tenaillon et al., 2010). The small intestinal microbiota is dominated by the families Lactobacillaceae and Enterobacteriaceae, whereas species from the families Bacteroidaceae, Prevotellaceae, Rikenellaceae, Lachnospiraceae, and Ruminococcaceae can be detected in the colon (Donaldson et al., 2016). While the microbiota is of paramount importance for numerous metabolic processes, including vitamin synthesis (LeBlanc et al., 2013) and digestion of dietary compounds (Backhed et al., 2005), compelling evidence points toward its impact on the maturation, development and function of the innate and adaptive immune system of the host (Macpherson and Harris, 2004; Sommer and Backhed, 2013).

Lactic acid bacteria exert important functions in the modulation of immune responses and have successfully been used as probiotics in inflammatory conditions of mice and men. For instance, Lactobacillus rhamnosus GG has been shown to exert preventive and therapeutic effects in atopic eczema and dermatitis (Isolauri et al., 2000; Kalliomaki et al., 2001, 2003; Viljanen et al., 2005). Moreover, treating IL-10 deficient mice with L. plantarum attenuated the severity of colonic inflammation by reducing mucosal IL-12p40 and IFN-γ levels (Schultz et al., 2002). In our own previous work, we could further demonstrate that a single commensal L. johnsonii strain was able to attenuate both intestinal mucosal and systemic pro-inflammatory immune responses upon murine infection with the enteropathogen Campylobacter jejuni (Bereswill et al., 2017). Kwon et al. (2010) have proposed that the observed beneficial anti-inflammatory properties may be elicited due to induction of regulatory DCs and regulatory T cells (Treg).

Gram-negative commensals, such as members of the family Enterobacteriaceae, tend to be neglected, but may, nevertheless, be potent immune-modulators (Zeuthen et al., 2006). This is best corroborated by the probiotic strain E. coli Nissle 1917, which has conclusively been shown efficient in the treatment of ulcerative colitis and is considered as an effective alternative to the standard maintenance therapy, i.e., mesalazine (Kruis et al., 2004; Henker et al., 2008; Schultz, 2008).

Importantly, Hessle et al. (2000, 2005) and Skovbjerg et al. (2010) have shown the differing impact of Gram-positive and Gram-negative bacteria on the modulation of cytokine production patterns, given that in vitro co-culturing of human PBMC with Gram-positive bacteria leads to elevated levels of IL-12, IL-1β, IFN-γ, and TNF, while Gram-negative bacteria rather induce IL-6, IL-8, and IL-10. In contrast, from monocytes derived human DC have been shown to produce comparable levels of IL-12 and TNF in response to commensal Gram-positive and Gram-negative bacteria (Karlsson et al., 2004; Zeuthen et al., 2006).

Murine studies have further underlined the impact of individual bacterial species on distinct immune cell populations. For instance, segmented filamentous bacteria (SFB) have been identified to induce the expansion of IL-17 producing Th17 cells (Ivanov et al., 2008), while Clostridium species of clusters IV and XIVa promoted accumulation of Treg in the colonic lamina propria (LP) of mice (Atarashi et al., 2011). The differentiation of CD4+ T cells into Treg locally in the LP as well as in the circulation is also supported by TLR-2 mediated sensing of the Gram-negative bacterium Bacteroides fragilis (Round et al., 2011). Moreover, treatment of IL-10^-/- mice with L. plantarum has been shown to attenuate the severity of colonic inflammation by reducing mucosal IL-12p40 and IFN-γ levels (Schultz et al., 2002). However, elucidating the differential impact of Gram-positive and Gram-negative commensals on the immune cell homeostasis in in vivo models, and identifying the species specific mechanisms of immunomodulation remains difficult and challenging, yet of utmost interest.

In our previous works, we have shown that normally developed mice rendered void of intestinal microbiota through a quintuple antibiotic therapy (i.e., secondary abiotic mice, ABx mice), display numerous changes of intestinal mucosal and systemic immune cell subsets, which can, however, almost completely be restored through reintroduction of intestinal antigens via fecal microbiota transplantation (FMT) (Ekmekciu et al., 2017a,b). Furthermore we were able to demonstrate that recolonization of ABx mice with VSL#3, a probiotic mixture consisting of eight bacterial species (namely Streptococcus thermophilus, Bifidobacterium breve, B. longum, B. infantis, Lactobacillus acidophilus, L. plantarum, L. paracasei, and L. delbrueckii subsp. bulgaricus), elicits IL-10 production by lymphocytes in the small and large intestinal LP, MLN and spleen, but does not influence the production of pro-inflammatory cytokines (Ekmekciu et al., 2017a,b). The impact of probiotics on cytokine production has been extensively investigated, and it has been proposed, that some strains might promote mainly IL-12 production, thus inducing Th1 type immune responses, which are known to protect from pathogens. Other strains, however, more effectively impacted the production of the anti-inflammatory IL-10, thereby limiting excessive immune responses as observed in autoimmune and allergic diseases, for instance (Fujiwara et al., 2004; Foligne et al., 2007; Shida et al., 2011).

Niess et al. (2008) have previously further addressed the role of commensal organisms in the recruitment of pro-inflammatory cytokine producing CD4+ cells in the colonic lamina propria and proposed that these cell types might contribute to the immunopathogenesis of colitis. However, a fine-tuned balance between pro- and anti-inflammatory immune responses at mucosal and systemic sites is of utmost importance for the vertebrate host. For instance, a reduction of the IL-17 producing Th17 cell compartment playing a pivotal role in protection against bacterial and fungal pathogens (Aujla et al., 2007) may explain the increased susceptibility of microbiota depleted mice to pathogens (Croswell et al., 2009; Stiemsma et al., 2014).

In the present study, we investigated the role of two important representatives of Gram-negative (namely E. coli) and Gram-positive (namely L. johnsonii) commensals, both derived from the gut microbiota of healthy mice, in restoring numbers of distinct immune cell subsets following antibiotic treatment as compared to FMT in C57Bl/6j ABx mice. To address this, we analyzed the immune responses exerted by lymphocytes within the small and large intestinal LP, MLN and spleen of microbiota-depleted mice upon reassociation with E. coli or L. johnsonii and upon FMT at day (d) 7 and d28 post recolonization. We furthermore surveyed the cytokine production patterns assessing TNF, IFN-γ, IL-17, IL-22, and IL-10 secretion by CD4+ lymphocytes in respective compartments.

In the present study, we aimed at comparing the short- and long-term effects of a single representative Gram-negative and Gram-positive intestinal commensal species, namely of E. coli and L. johnsonii, respectively, versus the complex gut microbiota in shaping host immunity following antibiotics-induced microbiota depletion. These species were chosen for the following reasons. Firstly, both commensals are common inhabitants of the intestinal ecosystem, and although lactobacilli are quantitatively more abundant than E. coli (Castillo et al., 2006), the latter represents the predominant facultative anaerobe Gram-negative strain within the mammalian gastrointestinal tract (Finegold et al., 1983; Tenaillon et al., 2010). Furthermore, in our previous works we could confirm differential immunomodulatory properties of either strain under defined immunopathological conditions. For instance, E. coli, but not L. johnsonii, aggravated Th1-driven pro-inflammatory immune responses in murine ileitis (Heimesaat et al., 2006, 2007). Furthermore, L. johnsonii was able to attenuate intestinal and systemic pro-inflammatory and to augment anti-inflammatory immune responses upon C. jejuni infection of secondary abiotic mice (Bereswill et al., 2017), and has further been shown to be effective against enteric including infectious morbidities (Lievin-Le Moal and Servin, 2014), hence resulting in its commercial probiotic application (e.g., Nestlé LC1).

Here, we unraveled immunomodulatory properties of respective commensals at two different time points following recolonization of antibiotics-treated mice, but in the absence of immunopathological conditions. Performing kinetic analysis in this context is important, given that intestinal responses are highly dynamic over time after the transfer of microbiota into germ-free animals (El Aidy et al., 2012; Tomas et al., 2015).

Cultural analysis of fecal samples revealed that either commensal strain was able to stably colonize the murine intestinal tract at high intestinal loads throughout the observation period. Our previous work revealed that upon cessation of antibiotic therapy neither regrowth of intestinal bacterial commensals nor changes in the immune cell populations could be observed, further supporting that the observed immune responses of the mucosal and systemic sites can be exclusively attributed to defined bacterial commensal recolonization (Ekmekciu et al., 2017a,b).

Overall, single-strain recolonization appeared to be, to some extent, less effective in restoring immune cell populations after broad-spectrum antibiotic treatment than the complex microbiota upon FMT, given that mono-colonized mice displayed a lack of full recovery of affected cell populations. Following mono-colonization of mice both common and strict strain-specific immune responses could be observed over time. For instance, E. coli and L. johnsonii had a minor impact on the relative abundances of intestinal CD4+ and CD8+ cells when compared to antibiotics-treated mice, while FMT resulted in cell percentages comparable to naive conventionally colonized counterparts. Furthermore, E. coli, but not L. johnsonii recolonization could increase colonic CD4+ abundances at d7, but not later on. Whereas small intestinal CD8+ cells could be reestablished only rather late upon mono-colonization, colonic CD8+ cells were virtually unaffected. In the systemic compartment, however, the observed increases in percentages of splenic CD8+ cells in microbiota-depleted mice could be reversed by both E. coli and L. johnsonii until d7, and remained lower in E. coli recolonized mice later on. These results underline the importance of examining immunological sequelae of gut microbiota-interventive strategies, not only at mucosal surfaces, but also on the systemic level of the immune system, given that the presence/absence and distinct composition of the gut microbiota might also affect splenic lymphocytes. This rationale is supported by the former findings that the Bacteroides fragilis driven differentiation of Treg was not limited to the LP, but could also be detected in the circulation (Round et al., 2011).

Interestingly, recolonization with L. johnsonii resulted in highest CD4+ lymphocyte numbers in the small intestinal LP, supporting former evidence from a clinical investigation, where the application of another Lactobacillus strain, namely L. reuteri, led to a significant increase in ileal CD4+ cells (Valeur et al., 2004). In our study, either commensal strain was able to promote CD25 expression on CD4+ lymphocytes in the small and large intestines as well as in the spleen, but only L. johnsonii was able to achieve the same effect in MLN. Of note, at d7 following recolonization, L. johnsonii was a stronger inducer of the colonic activated T cells (including Treg) than E. coli. The dependence of Treg on intestinal bacterial antigens has been demonstrated before. Atarashi et al. (2011) showed that treatment of mice with vancomycin, an antibiotic compound directed mainly against Gram-positive bacteria, resulted in reduced numbers of colonic Treg (Zarkan et al., 2016). A study conducted with germ-free rats revealed that colonic CD25+ cells were higher in animals co-colonized with E. coli and L. plantarum 299v, than in those harboring E. coli alone (Herías et al., 1999). This indicates that increased microbiota diversity and synergistic effects between species may elicit a stronger immunomodulatory effect on activated T cells and the Treg population, which is supported by our findings that FMT induced the most prominent expression of CD25 in all analyzed immunological compartments.

Unlike FMT, single-strain recolonization resulted in an only partial restoration of the intestinal activated DC (CD86+) cells, which remained, however, lower than in untreated SPF mice. Former evidence suggests, that the activation, maturation and cytokine production of DC upon bacterial stimulation is highly dose dependent (Evrard et al., 2011). Our data indicate that the activation of DC may furthermore be dependent on the diversity of presented microbial antigens.

In vitro studies applying PBMC revealed that Gram-positive and Gram-negative bacteria differentially induced cytokine expression patterns (Hessle et al., 2000, 2005; Skovbjerg et al., 2010). Therefore we analyzed the production of pro- and anti-inflammatory cytokines, including TNF, IFN-γ, IL-17, IL-22, and IL-10 in mice following recolonization with either E. coli or L. johnsonii. Intestinal E. coli colonization had a more prominent effect on TNF production by lymphocytes than L. johnsonii. At d7 following E. coli application, TNF levels measured in intestinal, peripheral and systemic immune compartments were comparable to those in naive mice. However, this effect could be preserved until d28 in MLN only. While excessive TNF production has conclusively been linked to chronic inflammation in numerous organs including the gastrointestinal tract (Kollias et al., 1999; Popa et al., 2007), it is important to recognize the protective roles of TNF against epithelial injury and disease susceptibility. In a murine ileitis model increased TNF levels resulted in improved epithelial barrier functions and prevention from disease onset (Pagnini et al., 2009). Moreover, TNF deficient mice were more prone to acute colitis, thus leading to the conclusion that TNF might have protective functions in normal gut homeostasis and intestinal epithelial integrity (Naito et al., 2003). Rakoff-Nahoum et al. (2004) suggested that the production of intestinal epithelial tissue-protective factors, including TNF, is regulated via TLR-mediated recognition of commensal bacteria, while we here provide further evidence that the adaptive immune system may also contribute to this effect.

Furthermore, both E. coli and L. johnsonii were able to promote the small intestinal and colonic production of IFN-γ by CD4+ cells as early as d7 post recolonization, whereas this cytokine remained highly expressed until d28 only in the small intestines of E. coli recolonized mice. The abundance of IFN-γ expressing CD4+ lymphocytes in the MLN and spleen remained unaffected by any of the interventions. These data seem to contradict in vitro findings, which suggest that Gram-positive bacteria tend to be stronger inducers of IFN-γ and TNF than Gram-negative species (Hessle et al., 2000, 2005; Skovbjerg et al., 2010). Apart from the general difficulty of directly applying in vitro results to vertebrates exhibiting a complex intestinal microbiota and distinct intra-luminal milieu interacting with the fine-tuned immune system in health and disease, it is also highly likely that human PBMC and murine CD4+ cells differently respond to bacterial stimuli, given former evidence regarding the differences in cytokine patterns between human monocytes and human DC upon bacterial stimulation (Karlsson et al., 2004).

Moreover, E. coli recolonization seemed to favor a Th17 immune cell differentiation, particularly in the small intestine, given that the production of the two key cytokines of this cell population, i.e., IL-17 and IL-22, was significantly higher in CD4+ cells isolated from with E. coli recolonized mice as compared to mice harboring L. johnsonii only. Th17 polarization is determined by the co-stimulation and cytokine provision from the DC (Perona-Wright et al., 2009), whereby IL-23 plays a particularly important role in the sustenance of Th17 cell responses in vivo (Mangan et al., 2006; McGeachy et al., 2009). It is tempting to speculate that commensal E. coli more strongly induce IL-23 production of DC, thus leading to a Th17 cell differentiation. Interestingly, a study conducted with human PBMC revealed that myeloid DC do indeed produce higher levels of IL-23 when stimulated with E. coli (Manuzak et al., 2012). Furthermore, E. coli heat-labile enterotoxin (synergizing with LPS) has been shown to induce IL-1β and IL-23 secretion by DC, which in turn promoted IL-17 and IFN-γ production by CD4+ T cells (Brereton et al., 2011).

Lactobacillus johnsonii recolonized mice, however, displayed higher abundances of CD4+IL-17+ cells in the colon and spleen and of CD4+IL22+ cells in the small intestinal LP and MLN as compared to their secondary abiotic counterparts at d7 post recolonization. Nonetheless, all the aforementioned immune cell populations declined back to the levels observed in secondary abiotic mice until 4 weeks after L. johnsonii recolonization. It has been known for a while now, that specific bacterial species such as SFB are required for the differentiation of small intestinal Th17 cells in germ-free mice (Ivanov et al., 2008; Gaboriau-Routhiau et al., 2009). A more recent investigation revealed that extracellular pathogens, such as Citrobacter rodentium and E. coli O157:H7 can also induce Th17 differentiation following adhesion to intestinal epithelial cells (Atarashi et al., 2015). Importantly, a common pilus adherence factor, which mediates adhesion to epithelial cells, has been reported for both pathogenic and commensal E. coli species (Rendón et al., 2007), indicating that the observed effects on the Th17 cells may be mediated through adherence of this commensal E. coli.

Furthermore, both E. coli and L. johnsonii could restore the antibiotics-induced reduction of IL-10 producing CD4+ cells in the small intestines, but only the latter could achieve the same effect in the colon, indicating commensal species-specific anti-inflammatory properties. Our results are well in line with a comprehensive analysis of numerous Lactobacillus species suggesting that species derived from healthy mice can be largely classified as potentially anti-inflammatory (Gorska et al., 2016). Notably, these data support the already proposed concept and evidence, that probiotic bacteria including L. johnsonii exert health-promoting effects through induction of regulatory/anti-inflammatory host immune responses (Di Giacinto et al., 2005; Kwon et al., 2010). However, it is highly likely that L. johnsonii and other probiotic species exert additional beneficial functions through IL-10 independent mechanisms, given their capability to attenuate the severity of colonic inflammation in IL-10 deficient mice.

The reintroduction of complex microbiota into the host via FMT is a well-known therapy dating back to the Chinese Dong-jin dynasty in the fourth century (Zhang et al., 2012) and has undergone a renaissance recently as a therapeutic option for the treatment of recurrent and refractory Clostridium difficile toxin induced acute necrotizing pseudo-membranous enterocolitis (van Nood et al., 2009; Rohlke et al., 2010; Brandt et al., 2012; Fischer et al., 2016; Scaldaferri et al., 2016).

In summary, the restoration of immune cell populations following broad-spectrum antibiotic treatment by commensal bacterial mono-colonization appears to be less effective than by complex intestinal microbiota association upon FMT, but depends both on the respective immune cell subtypes and analyzed compartments and does hence not follow a clear cut pattern. Nevertheless, mono-colonization of secondary abiotic mice with defined Gram-negative or Gram-positive species has a deep impact on shaping the host immune system and drives both pro- as well as anti-inflammatory immune responses on the mucosal, peripheral and systemic level in parallel. However, E. coli appears to favor pro-inflammatory immune responses, while L. johnsonii has a stronger impact on anti-inflammatory immune cell subsets.

Future studies should further unravel the fine-tuned interplay between the well-orchestrated concert within the complex intestinal microbiota intestinal on one side and the different members of the immune system on the other in health and disease subsequently providing potential novel treatment options for immunopathological intestinal and extra-intestinal morbidities.

IE: Performed experiments, analyzed data, and wrote paper. EvK: Performed experiments. CN: Suggested critical parameters in design of experiments, supplied antibodies. PB: Suggested critical parameters in design of experiments, supplied antibodies. AS: Provided advice in design and performance of experiments. SB: Provided advice in design and performance of experiments, co-edited paper. MH: Designed and performed experiments, analyzed data, and co-wrote paper.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.