The human intestinal microbiota is composed of nine bacterial divisions, which is far less than the number observed in soils (at least 20 divisions; Ley et al., 2006). Most gut colonizers do not grow outside the gut and are transmitted via direct person-to-person contacts. Children inherit their microbiota from their mothers (Dominguez-Bello et al., 2010), and intestinal microbiota is highly similar between identical twins (Turnbaugh et al., 2010). The MetaHIT project¹, a vast metagenomics study aimed at establishing an exhaustive catalog of the genes present in the intestinal microbiota, found that from the feces of 124 individuals, 18 species were found to be common to all individuals, and 57 were present in ≥90% individuals (Qin et al., 2010). Among those 57 species, the most abundant were from the phyla Bacteroidetes and Firmicutes, yet up to 2000-fold variation in relative species abundance was observed among individuals. Fecal enterobacteria, which mostly comprise Escherichia coli, are subdominant bacteria, making up to 10⁸ to 10⁹ colony-forming units (CFU) per gram of feces.

The intestinal microbiota is mostly composed of prokaryotes, even if eukaryotes, such as fungi, are also present. However, the intestinal microbiota plays a role of an organ in humans and provides various benefits to its host. One benefit is the barrier effect or colonization resistance, which refers to the ability of intestinal microbiota to prevent sustainable colonization by exogenous bacteria (Ducluzeau et al., 1977; Vollaard and Clasener, 1994). Most ingested bacteria only transit through the digestive tube and do not colonize the patient for a significant period of time (Buck and Cooke, 1969; Cooke et al., 1972). The barrier effect mainly relies on the fact that the endogenous intestinal microbiota leaves very few niches (available nutrients and attachment sites) for use by exogenous bacteria (Vollaard and Clasener, 1994). The intestinal microbiota is also involved in the maintenance of intestinal epithelial homeostasis (Rakoff-Nahoum et al., 2004) and promotes angiogenesis (Stappenbeck et al., 2002). Moreover, a link between the intestinal microbiota and metabolic disorders has been proposed; in particular, low species diversity in the intestinal microbiota has been linked to medical conditions, such as obesity, inflammatory bowel diseases, atopy, and diabetes (Ley et al., 2005; Turnbaugh et al., 2006; Round and Mazmanian, 2009; Qin et al., 2012).

Next-generation sequencing has not yet changed the way antibiotic resistance is investigated, perhaps because the resistance genes that may be harbored by subdominant bacteria remain inaccessible to the current sequencing methods (Lagier et al., 2012b). However, the intestinal tract is a major reservoir for antibiotic-resistant bacteria, including naturally resistant bacteria and those with acquired resistance-conferring genes carried on mobile genetic elements such as plasmids, conjugative transposons or integrative and conjugative elements (ICEs; Sommer et al., 2009). The diversity of the resistance genes among intestinal bacteria (i.e., the intestinal resistome) cannot be effectively assessed by conventional methods based upon culture on antibiotic-supplemented agar media because most intestinal bacteria cannot be cultured using conventional methods. The use of various culture media, atmospheric conditions and mass-spectrometry identification (“culturomics”) allowed the establishment of an inventory of bacterial species, including many new intestinal microbiota species, in numbers even greater than those discovered using pyrosequencing (Lagier et al., 2012a). Thus far, this technique has not addressed the question of global antibiotic resistance of the microbiota.

Currently, the best description of the diversity of resistance genes present has been obtained using culture-independent methods (Sommer et al., 2009). Sommer et al. (2009) cloned the metagenome of feces samples into susceptible E. coli and plated it on agar media supplemented with various antibiotics. When applied to an overnight aerobic culture of feces, 95% of the identified genes had >90% nucleic identity with sequences in GenBank² found in pathogenic bacteria. Indeed, the genes identified in the aerobic fraction had been reported, repeatedly, to occur on mobile genetic elements [bla_CTX-M, bla_TEM, aac(3)-II, aac(6′)-Ib, bla_AmpC] found in pathogenic bacteria, such as enterobacteria. Conversely, when applied to the feces metagenome with no previous aerobic culture, the average shared identity dropped to 60.7%. Thus, the intestinal resistome can be divided into (i) a “resident” resistome, composed of the resistance genes naturally present in permanent members of the intestinal microbiota, such as the beta-lactamase gene from Bacteroides sp., and (ii) a variable resistome, composed of exogenous genes present in transient bacteria or acquired by lateral transfer (Wellington et al., 2013). Enrichment of the variable resistome comes from ingestion of resistant bacteria through food (Ruimy et al., 2010) or fecal peril (Tangden et al., 2010). Although exogenous bacteria may not colonize because of the intestinal microbiota barrier effect (also called “resistance to colonization”), their resistance genes can be transferred to resident bacteria through horizontal gene transfer during transit (Duval-Iflah et al., 1980).

Antibiotic effects on the intestinal microbiota depend on (i) the colonic concentrations of the antibiotic (luminal and mucosal) and/or its active metabolites and (ii) the activity of these concentrations on the bacterial species present. The growth of susceptible bacteria will either be impeded (bacteriostatic effect), or they will be killed (bactericidal effect; Dethlefsen et al., 2008; Antonopoulos et al., 2009). Thus, the extent and persistence of the impact of antibiotics on the intestinal microbiota is highly drug-dependent (Nord et al., 1984; Taur et al., 2012). Even antibiotics of the same family and spectrum of activity can have a very different impact depending on their rate of intestinal excretion (Brautigam et al., 1988; Michea-Hamzehpour et al., 1988). Using next-generation sequencing, Dethlefsen and Relman (2011) precisely observed the fecal diversity of three healthy subjects during 300 days, during which they received 2 × 5-day courses of ciprofloxacin, at days 60 and 250. Ciprofloxacin caused a loss of diversity and a shift in community composition occurring within 3–4 days of drug initiation. This effect was somewhat surprising because most of the microbiota is composed of anaerobes that are weakly susceptible to ciprofloxacin (Nord and Edlund, 1989). However, concentrations of ciprofloxacin that accumulate in the colon during treatments (Fantin et al., 2009) are so high that they most likely overcame their minimal inhibitory concentrations. The perturbation created by antibiotic use took weeks to be resolved; furthermore, the composition of the intestinal microbiota remained altered from its initial state.

In newborns treated by a combination of ampicillin and gentamicin, the Actinobacteria and Firmicutes phyla, comprising bacteria with potential benefit (Bifidobacterium and Lactobacillus) were replaced by Proteobacteria, including Enterobacteriaceae (Fouhy et al., 2012). This effect could be of importance considering that Proteobacteria are enriched with mobile genetic elements, including antibiotic resistance encoding genes (Baquero et al., 2013). The increase of Proteobacteria was persistent after 8 weeks. Indeed, in patients undergoing allogeneic hematopoietic stem cell transplantation, metronidazole (an antibiotic with broad-spectrum activity against anaerobes) strongly reduced the diversity of the intestinal microbiota (Taur et al., 2012). In mice receiving a combination of amoxicillin, metronidazole, and bismuth, the composition of the intestinal microbiota was altered, but the perturbation was resolved within 2 weeks; in contrast, resolution took approximately 6 weeks for mice treated with cefoperazone, a wide-spectrum cephalosporin (Antonopoulos et al., 2009). The resilience of the composition of the intestinal microbiota is thus likely to be different according to the type of antibiotic given.

As described above, the barrier effect is mainly exerted by anaerobes (Ducluzeau et al., 1977; Vollaard and Clasener, 1994). Thus, antibiotics with high activity against anaerobes, such as clindamycin, potently affect the capacity of the microbiota to prevent colonization by exogenous microorganisms (van der Waaij et al., 1971). In the study by Donskey et al. (2000), 13 vancomycin-resistant enterococci (VRE)-carriers received antibiotics active against anaerobes, while 10 VRE-carriers received antibiotics poorly active against anaerobes . Strikingly, the average density of VRE (expressed in log CFU/g of feces) significantly increased (by 2.2 logs) in patients who received antibiotics that were active against anaerobes, whereas the average density decreased by 0.6 log in those patients receiving antibiotics with minimal activity against anaerobes. In a subsequent study, the same group has reported that the density of resistant Gram-negative bacilli also increased under exposure to antibiotics that were active against anaerobes (Bhalla et al., 2003). In the latter case, the resistant bacteria occupied niches that appeared to be left by anaerobes (Figure 1).

The barrier effect can also be studied by considering one species, such as E. coli. In healthy volunteers, the E. coli population is composed of a variable number of clones of different abundances: dominant and subdominant clones (Lidin-Janson et al., 1978). When these E. coli have different susceptibilities, antibiotic exposure will change their respective proportions and promote the growth of the resistant strains over that of the susceptible ones (Figure 1). This phenomenon has recently been reported for fluoroquinolones, which cause a sharp decrease in total counts of intestinal enterobacteria during treatments. The available niches can then be occupied by resistant enterobacteria that were initially present in low fecal concentrations (Fantin et al., 2009) or from a new acquisition (de Lastours et al., 2012). If the number of total enterobacteria remains unchanged, then the relative abundance of resistant enterobacteria increases (Figure 1).

Several studies have shown that patients with infections caused by resistant bacteria were more likely to have taken antibiotics recently (Ben-Ami et al., 2009). Indeed, the link between antibiotic exposure and antibiotic-resistant infection could lie in the intestinal microbiota, as antibiotics would allow the overgrowth of resistant bacteria (i.e., increase the density of resistant bacteria). Interestingly, the association between antibiotic use and infections caused by resistant bacteria is also found for antibiotics with little effect on anaerobes, e.g., co-trimoxazole or quinolones, suggesting that it is not only the overgrowth of resistant bacteria within niches left empty by anaerobes that increases the risk of infection by resistant bacteria, but more likely the augmentation of the relative abundance of resistant bacteria (i.e., the augmentation of the proportion of resistant bacteria) within specific niches (Figure 1). MDRE are extensively antibiotic-resistant and not only to beta-lactams, but also to many other antibiotics including among others fluoroquinolones, aminoglycosides, or co-trimoxazole (Pitout, 2009). Thus, MDRE can overgrow and increase the risk of their involvement in further infection under almost any antibiotic exposure.

Urinary-tract infections (UTIs) are the most common bacterial infections (Hooton, 2012) and are most often caused by enterobacteria, and especially E. coli. There is some evidence that in most cases, the infecting clone originates from the intestinal microbiota (Yamamoto et al., 1997), even if it cannot always be retrieved in the stool at the onset of the symptoms (Moreno et al., 2008). Specific strains of E. coli appear to have the ability (through virulence factors) to colonize the urethra and bladder, causing cystitis; others are able to further colonize the ureter to cause pyelonephritis (Plos et al., 1995). According to this “virulence theory,” some strains of E. coli could cause UTIs, even if they are subdominant in the intestines (low relative abundance). Alternatively, the relative abundance of the various clones of E. coli present in the feces may also play a key role in the pathogenesis of UTIs, in that the dominant clone of E. coli would have the maximum likelihood to colonize the urinary tract (Moreno et al., 2008). In 42 women with cystitis caused by E. coli, Moreno et al. (2008) compared the UTI-causing strain to 30 randomly picked colonies from concomitant stool samples. In 90% of the women, the urine clones were found in the feces. In 71% of the women, the urine clone that was present in the feces was also the dominant fecal clone. Urine clones mostly belonged to B2 and D phylogroups and had an increased content of urovirulence factors. For one given fecal clone, predictors of infection were that the urine clone belonged to the B2 phylogroup and was dominant. Thus, the findings of this study reconciled the dominance and the virulence theories.

Translocation is defined as the passage of viable indigenous bacteria from the gastrointestinal tract to the mesenteric lymph nodes (Berg and Garlington, 1979). This passage is the first step for bacteria to settle in other locations, such as the bloodstream, or cause secondary forms of infections, such as in the cardiac valves. Bacterial translocation can be part of normal physiological processes in healthy subjects, but to a limited extent and without deleterious consequences. By contrast, sustained translocation is observed in subjects with specific deficiencies, such as neutropenia, starvation, or hemorrhagic shock and then leads to severe septic consequences (Tancrede and Andremont, 1985; Youssef et al., 1998). A key point determining bacterial translocation is the intestinal density: the translocating bacteria are mostly dominant within the intestinal microbiota (Youssef et al., 1998; Taur et al., 2012). Furthermore, translocation of enterobacteria has been reported in immunocompetent mice exposed to penicillin, clindamycin, and metronidazole (Berg, 1981). In the same report, the measured intestinal density of enterobacteria increased by 3–5 logs. Thus, antibiotics that increase the density of resistant bacteria would increase the risk of their translocation.

In the above-mentioned study from Donskey et al. (2000), the surrounding environment of 10 VRE-carrying patients was investigated for the presence of VRE. Environmental samples from 21 patients were analyzed and compared according to the intestinal density of VRE. Strikingly, when the density was <4 logs CFU/g of stool, VRE were found in the patient’s environment only in one out of nine sample sets (11%). Conversely, when density was ≥4 logs CFU/g of stool, VRE were found in 10 of 12 sets (83%). Other studies have shown that patients can acquire resistant bacteria from a former occupant of the room through the persistence of the bacteria on environmental surfaces (Datta et al., 2011). Although no study has shown that cross-transmission occurs less often when the density of resistant bacteria was low, the results from Donskey et al. (2000) support this notion. To date, no data are available for MDRE.

Selective digestive decontamination (SDD) aims to eliminate MDRE from the intestinal lumen and prevent further infection and dissemination. In SDD, the patient receives orally and/or parenterally administered wide-spectrum antibiotics for a short time, with the aim of eradicating most potentially pathogenic bacteria and/or the resistant bacteria from the intestinal microbiota, while sparing anaerobes as much as possible. SDD has mostly been used to eliminate all enterobacteria in patients at risk of infections, such as in onco-hematology and intensive care ones. In 1985, SDD was successfully employed in France for the control of a major outbreak involving MDRE (Brun-Buisson et al., 1989). The SDD regimen comprised oral neomycin, colistin, and nalidixic acid. The rate of MDRE infections decreased, yet the overall rate of nosocomial infections did not.

More recently, two decontamination regimens have been compared: selective oropharyngeal decontamination (SOD: an oropharyngeal topical administration of tobramycin, colistin, and amphotericin B, an antifungal agent) and SDD (SOD + intravenous cefotaxime), each given for 4 days (de Smet et al., 2011). SDD and SOD caused a slight reduction in mortality compared to the control group. Unexpectedly, SDD-receiving patients carried fewer enterobacteria resistant to ciprofloxacin, ceftazidime, or aminoglycosides than the standard care group. In a later study, the authors showed that when no Gram-negative bacilli were found in the intestinal microbiota, the rate of Gram-negative bacilli bacteremia decreased threefold (Oostdijk et al., 2011). However, if patients continued to be colonized, the rate of bacteremia was not different from that of patients receiving standard care (Oostdijk et al., 2011).

A recently published post hoc analysis focused on a subgroup of 507 patients with detectable Gram-negative bacilli in their intestinal microbiota prior to SDD (Oostdijk et al., 2012). As expected, the eradication rates for enterobacteria differed according to the resistance pattern. The difference was statistically significant for aminoglycoside resistance (62 vs. 81% in the standard care group; p < 0.05), but only a trend was evidenced for third-generation cephalosporins resistance (73 vs. 80% in the standard care group; p = 0.053). The study did not mention whether colistin-resistant bacteria were eradicated or not. Indeed, SDD raises the issue of using a last-resort antibiotic, colistin, with the risk of selecting MDRE in the microbiota which would also be resistant to the drug (Kumarasamy et al., 2010). So far, SDD has mainly been studied in countries with low MDRE prevalence (de Smet et al., 2011; Overdevest et al., 2011) and this risk could be much higher in countries where it is high.

A study in Israel, a country with a high prevalence of bacteria that produce the Klebsiella pneumoniae carbapenemase (KPC)-type carbapenemase, has tested a 7-day colistin + gentamicin SDD regimen in patients colonized with such bacteria (Saidel-Odes et al., 2012). The results showed that at day 7 (at the end of SDD), KPC-producing bacteria were no longer detectable in the feces of 61% of patients vs. 16% in placebo. Nevertheless, when SDD was discontinued, KPC-producing bacteria could be detected again in some formerly negative patients, suggesting that the concentrations of these bacteria were below the limit of culture detection while SDD was being applied. Nevertheless, keeping MDRE at low intestinal concentrations in the absence of total eradication may be sufficient to prevent further infections or cross-transmission. Interestingly, no colistin or gentamicin resistance emerged among the recovered KPC-producing bacteria.

Instead of killing resistant bacteria, another approach would be to prevent their overgrowth by inactivating the antibiotics in the intestine during treatments. Orally administered antibiotics are primarily absorbed in the proximal jejunum, yet a fraction reaches the colon, where the density of bacteria is maximal. Parenterally administered antibiotics are filtered by the liver, and a fraction is excreted through the gall bladder into the jejunum and then reaches the colon. Thus, the concept of designing drugs or beta-lactamases capable of inactivating antibiotics in the colon, but not earlier, has arisen.

The protective effect of a recombinant beta-lactamase, P1A, has been evaluated in 34 human volunteers taking ampicillin (Tarkkanen et al., 2009). In the ampicillin group without P1A, a decrease of Bifidobacterium, Streptococcus, Lactobacillus, and an increase of E. coli and yeasts, were noted in the intestinal microbiota. Conversely, in the ampicillin + P1A group, no significant changes in the composition of the intestinal microbiota could be observed. Furthermore, the relative abundance of ampicillin-resistant E. coli increased from 2.1% at baseline to >72.7% at day 5 under ampicillin only, while it remained under 10% in the ampicillin + P1A group. In addition to target-specific inactivation strategies, adsorbents, such as colonic delivery of activated charcoal, were shown to be efficient in trapping ciprofloxacin in a rat model (Khoder et al., 2010). Further clinical studies are expected to assess the efficacy of such a strategy.

Probiotics are defined as live microorganisms that may confer a health benefit to their host. The most used probiotics are lactic acid bacteria, Bifidobacteria, E. coli Nissle 1917 or Saccharomyces boulardii, a yeast. It is unknown if these probiotics could exert a barrier effect for resistant bacteria. A study from New Zealand competed E. coli Nissle with a fluoroquinolone-resistant E. coli in elderly residents, and the results showed that there was eventually no difference in terms of carriage of the fluoroquinolone-resistant E. coli between the Nissle group and the placebo group (Tannock et al., 2011). Another growing concern about probiotics is that there remains little evidence that the massive ingestion of one single species can restore all the intestinal microbiota at a significant extent. Another potential way to restore microbiota is a fecal microbiota transplantation (FMT), which refers to the process of instilling a liquid suspension of stool from a healthy donor into the gastrointestinal tract of a patient to restore the intestinal microbiota immediately after any perturbation, such as that caused by antibiotics.

When mice with intestinal microbiota affected by antibiotics, were caged with mice without previous antibiotic exposure, a faster restoration of the intestinal microbiota was observed. It was suggested that this was due to the transfer of a normal microbiota to the antibiotic treated mice, likely through coprophagy (Antonopoulos et al., 2009). In humans, FMT has demonstrated high efficacy in patients with recurrent Clostridium difficile infections (Gough et al., 2011; van Nood et al., 2013). The main limitation of FMT is its obvious repellence, which could be overcome by rectal instillations instead of oral routes (Bakken, 2009). Another caveat is that the fecal samples administered could include undetected pathogens. This caveat could be overcome by auto-banking (Tosh and McDonald, 2012) or by using preparation containing a cocktail of defined strains (Hamilton et al., 2012). This has recently been done to lower the density of VRE in mice (Ubeda et al., 2013) and of MDRE in chickens (Nuotio et al., 2013). However, these approaches have not been used in humans so far. If data support the efficacy of FMT in resolving C. difficile infections, studies assessing its efficacy in the context of outbreaks of MDRE to lower the risk of their transmissions and infections are warranted.

Another strategy for combating antibiotic-induced perturbations and to keep MDRE at low densities is to improve the use of antibiotics through antibiotic stewardship programs (ASPs; MacDougall and Polk, 2005). Indeed, these programs can be beneficial to the intestinal microbiota at three levels: (i) avoid prescriptions when antibiotics are not justified (Willemsen et al., 2010), (ii) scale down from the use of empirical wide-spectrum antibiotics to the narrowest spectrum possible, guided by antibiotic-susceptibility tests (Cosgrove et al., 2007), and (iii) choose the antibiotic with the lowest impact on the intestinal microbiota whenever possible (Lesprit and Brun-Buisson, 2008). In a Cochrane-based review, ASPs have been showed to decrease the overall antibiotic resistance, as well as C. difficile infection, suggesting their role in intestinal microbiota preservation (Davey et al., 2005). To reduce the use of wide-spectrum antibiotics further, new rapid diagnostic tests that identify resistance traits in strains in clinical samples or feces are being developed and have attracted interest from clinicians (Cuzon et al., 2012).