Intestinal Bacteriophage Therapy: Looking for Optimal Efficacy

SUMMARY

Several human intestinal microbiota studies suggest that bacteriophages, viruses infecting bacteria, play a role in gut homeostasis. Currently, bacteriophages are considered a tool to precisely engineer the intestinal microbiota, but they have also attracted considerable attention as a possible solution to fight against bacterial pathogens resistant to antibiotics. These two applications necessitate bacteriophages to reach and kill their bacterial target within the gut environment. Unfortunately, exploitable clinical data in this field are scarce. Here, we review the administration of bacteriophages to target intestinal bacteria in mammalian experimental models. While bacteriophage amplification in the gut was often confirmed, we found that in most studies, it had no significant impact on the load of the targeted bacteria. In particular, we observed that the outcome of bacteriophage treatments is linked to the behavior of the target bacteria toward each animal model. Treatment efficacy ranges from poor in asymptomatic intestinal carriage to high in intestinal disease. This broad range of efficacy underlines the difficulties to reach a consensus on the impact of bacteriophages in the gut and calls for deeper investigations of key parameters that influence the success of such interventions before launching clinical trials.

INTRODUCTION

Advances in sequencing technologies have revealed the diversity of microbes in human-associated microbiota, in particular in the gut (1–4). Among these microbes, bacteria and their viruses, bacteriophages (phages), are the most abundant. Recently, several studies have shown an association between variations of intestinal phage communities and several human diseases or disorders such as inflammatory bowel disease (IBD), colorectal cancer, or child growth impairment (5–12). Although causality between intestinal phages and human diseases has not been established, the most convincing data supporting an active role for phages come from the field of fecal microbiota transplantation, used to treat Clostridioides difficile recurrent infections. Indeed, sterile fecal filtrate was found to be as effective as nonfiltrated fecal transplantation, and phages from the donors could colonize the recipient’s microbiota for at least 12 months (13, 14). More unexpected was the recent evidence that a prophage sequence from the intestinal bacterium Enterococcus hirae triggers a specific immune response, enhancing the cyclophosphamide treatment of cancer patients, showing that the role of phages in the gut is broader than anticipated (15).

Interestingly enough, the link between gut and phages roots from the original 1917 report of F. d’Herelle, who isolated novel ultramicrobes from feces of patients recovering from diarrhea caused by Shigella and named them bacteriophages (16). Following this report, phage therapy, the use of phages to treat bacterial infections, increased for a few decades before antibiotics became the most successful antibacterial treatment. Unfortunately, the growing expansion of bacteria becoming resistant to antibiotics is now recognized as a major public health threat for which solutions must be found. During the last 2 decades, this situation has given more weight to phage therapy, which is now becoming more accessible in Europe and the United States to patients in lack of other solutions (17–19). For more than 80 years, phage therapy treatments have also been used in Georgia and few other countries in Eastern Europe that can provide access to such treatment. However, no recent and well-documented clinical data are available impeding the translation of this clinical experience worldwide.

Given the initial observation by d’Herelle, one could expect that phage therapy of intestinal bacterial infections would be the most documented treatment. However, this is not the case. Worse, a unique recent (<30 years old) clinical trial that attempted to treat Escherichia coli diarrhea in Bangladeshi children failed (20). This failure could not be attributed to the lack of efficacy of phages targeting E. coli. Indeed, the authors uncovered an unexpected correlation between the abundance of strains of Streptococcus and diarrhea symptoms, questioning the direct causality of E. coli strains in this disease (21, 22). This observation considerably reduced the number of children in which an effect on the phage treatment could be found (20). In addition, the clinical status of children who had E. coli as the main pathogen was not dramatically improved upon treatment. Although this clinical trial and associated safety studies have firmly established that oral application of phages is safe (23), we must recognize a gap of knowledge in translating the in vitro efficacy of phages into in vivo optimal conditions, in particular when targeting intestinal bacterial pathogens.

Here, we focused this review on the literature reporting intestinal phage therapy experiments using exclusively mammalian gut models, since such models are generally considered as proxies to study the physiopathology of intestinal diseases in humans. Only a limited number of studies showed unambiguously that a phage treatment led to a significant reduction of the targeted bacteria coupled with amplification of phages and reduction of symptoms, which are the criteria fulfilling the definition of an active phage therapy treatment (24, 25). Given the anatomy of the digestive tract and the spatial distribution of microbes (75), it seems unlikely that passive phage therapy, which relies on the administration of single high dose of phages to reduce the bacterial load within a single infection cycle, would indeed be sufficient to provide an effective treatment in this organ. The different modalities that have been tested were also reviewed to delineate recommendations for future studies. Overall, we found that the success of phage treatments was more frequent in infected animals developing a disease than in animals merely or strongly colonized by an intestinal pathogen and not developing clinical signs of disease. We then aligned this analysis with the growing literature that identified some of the parameters affecting the efficacy of phages in the gut. We concluded that using phages as a treatment to cure intestinal infections or as tools for precisely engineering the intestinal microbiota remains a challenging concept that requires deeper mechanistic investigations.

SYNTHESIS OF REPORTS ON THE EFFICACY OF PHAGE TREATMENTS TARGETING INTESTINAL PATHOGENS IN MAMMALS

The state of the art of molecular mechanisms of phage infection relevant to phage therapy has been abundantly reviewed over the past few years (17, 27–29). However, these mechanisms were rarely studied within the context of the mammalian gut ecosystem. We reviewed the literature of experimental phage therapy studies on bacterial pathogens causing intestinal infections in mammals in order to highlight conditions of treatments and point to mechanisms involved in phage therapy efficacy.

Clostridioides difficile

C. difficile is an intestinal pathogen for which phage therapy would be particularly useful given its growing resistance to antibiotics (30). Since no virulent (strictly lytic) phage of C. difficile has been isolated thus far, studies were performed using temperate phages (Table 1), which are usually disregarded for therapy because of their ability to transfer undesirable bacterial genes such as those coding toxins or antibiotic resistance systems (31).

TABLE 1.

Experimental phage treatments of C. difficile-infected animals

Animals	Bacterial challenge and phage therapya	C. difficile (CFU g–1) and outcomes	Reference
Hamsters	Bacterial challenge: oral 103 CFU and a second after 2 wks	Control: 4 × 105; all dead ≤ 72 h	32
	Group a: 1 phage, oral 108 PFU at 0 h	a: 7 × 104; no death
	Group b: 1 phage, oral 108 PFU at 0 h and every 8 h up to 48 h	b: 7 × 105; no death
	Group c: 1 phage, oral 108 PFU at 0 h and every 8 h up to 72 h	c: 4 × 105; 1 dead ≤ 72 h
		All animals died after subsequent bacterial challenge
Hamsters	Bacterial challenge: oral 2 × 103 CFU	Control: 106–107; death ≤ 55 h	33
	Group a: 2 to 4 phages, oral 108 PFU ml−1 at 0 h and every 8 h up to 36 h	a: 104–105; death ≤ 88 h
Mice	Bacterial challenge: oral 105 spores	Control: 108	34
	Group a: oral 1 wild-type phage at 4 h and twice daily during 4 days	a: 108
	Group b: oral 1 recombinant phage at 4 h and twice daily during 4 days	b: 108 (transient 1-log reduction)
	Group c: oral 1 mutant wtPhage at 4 h and twice daily during 4 days	c: 107
	Group d: oral 1 mutant crPhage at 4 h and twice daily during 4 days	d: 107

^aThe bacterial challenge was initiated at t = 0.

Two studies were performed with hamsters and reported a longer survival rate upon phage treatment. Ramesh et al. showed that 17 of the 18 phage-treated animals survived the bacterial challenge for 2 weeks, while all untreated animals died within 72 h. The first oral administration of phages (10⁸ PFU) was performed immediately after the bacterial challenge, but in the absence of a longitudinal record of phages in feces along the experiment it remains unclear whether phage amplified. In addition, at the endpoint of the experiment, the C. difficile levels in ceca were similar in all groups (about 10⁵ CFU per ml [CFU ml⁻¹]), and phages were detected in only few animals, including untreated controls. Unsurprisingly, phage-treated animals were not protected by a second bacterial challenge administered 2 weeks after phage treatment (32). Using a four-phage cocktail (8 × 10⁷ PFU), Nale et al. reported a decrease of C. difficile counts in the lumens of the ceca and colons of phage-treated animals compared to control hamsters 36 h after the first phage dose (an ∼2-log CFU ml⁻¹ reduction). In agreement with these data, phage-treated animals survived up to 120 h compared to 60 h for the untreated controls (33). In a third and more recent study, a temperate phage was genetically modified to behave as a virulent phage, and its oral application in mice 2 days postchallenge led to a significant reduction of ∼2 logs CFU g⁻¹ of C. difficile levels in feces compared to controls and the wild-type phage-treated group (34).

The two studies performed with hamsters used similar protocols, but the authors observed opposite outcomes. While phage treatments were roughly comparable in dose and frequency, the nature of the C. difficile inoculum was different: Nale et al. used spores, while Ramesh et al. used cells (32, 33). When phages were administered shortly after the challenge bacteria, a rapid decrease of the infectious dose was expected since bacteria would not have the time to adapt to the gut environment. However, spores, which are resistant to phages, may challenge this expectation, and this may explain why the study by Nale et al. resulted only in a time shift to death. Finally, a recent study performed in mice with genetically modified temperate phages demonstrated that a strong reduction of their ability to form lysogens resulted in a higher efficacy at reducing the levels of C. difficile in the gut compared to the wild type (34).

Enterococcus faecalis

Duerkop et al. observed a modest but significant reduction, ∼0.7-log CFU g⁻¹, of the fecal levels of E. faecalis in mice 24 h after administration of a single phage once (10¹⁰ PFU) 6 h after bacterial gavage of germfree mouse drinking water supplemented with 10⁵ PFU ml⁻¹ of the same phage (35) (Table 2). However, from 48 h to 9 days posttreatment, E. faecalis fecal levels were no longer different between phage-treated and control groups. These authors also noticed that some E. faecalis clones became phage resistant as soon as 48 h posttreatment and that high levels of phages persisted in the feces (>10⁶ PFU g⁻¹).

TABLE 2.

Experimental phage treatments of E. faecalis-infected animals

Animals	Bacterial challenge and phage therapya	E. faecalis (CFU g−1)	Reference
Mice	Group a: 1 phage, oral 1 × 1010 PFU by oral gavage after a 6 h colonization period and 5 × 108 PFU ml−1 in drinking water	Control: 8 × 109 a: 7 × 109 (transient 3-fold reduction)	35

^aThe bacterial challenge was initiated at t = 0.

Escherichia coli

Within the broad genetic diversity of the E. coli species that includes both pathogenic and commensal strains (36), several studies using different animal models focused on the highly pathogenic strains of the serotype O157:H7, which were analyzed together, while strains from other serotypes included mostly murine models and were also analyzed together (Table 3).

TABLE 3.

Experimental phage treatments of E. coli O157:H7-infected animals

Animals	Bacterial challenge and phage therapya	E. coli O157:H7 (CFU g–1)	Reference
Sheep	Bacterial challenge: oral 108 CFU	Control: undetectable	40
	Group a: 1 phage, oral 1013 PFU at day 2	a: undetectable
Mice	Bacterial challenge: oral 109 CFU	Control: 102–103	37
	Group a: 3 phages, oral 108 PFU at day 2	a: 102–103 (transient 1-log reduction)
	Group b: 3 phages, oral 1010 PFU at day 2	b: 102–103 (transient 1-log reduction)
	Group c: 3 phages, oral 1010 PFU at day 2 and then daily	c: 101–102
Group a: sheep	Bacterial challenge a and c: oral 3.5 × 1010 CFU; b: oral 108 CFU; c: rectal 1010 CFU a: 1 phage, oral 1011 PFU at days 1, 9, 10, and 11	Control for group a: undetectable a: undetectable	39
Group b: mice	b: 1 phage, oral 1010 PFU at days 1, 2, and 3	Control for groups b and c: <101 b: undetectable (transient 3-log reduction)
Group c: mice	c: 2 phages, oral 1010 PFU at days 1, 2, and 3	c: undetectable (transient 2-log reduction)
Group d: steers	d: 2 phages, rectal 1011 PFU at days 8, 9, 10, and 15+ in drinking water 106 PFU ml−1 from day 8	Control d: 102 d: 102 (transient 1-log reduction)
Sheep	Bacterial challenge: oral 1010 CFU	Control: 106	41
	Group a: 1 phage, oral 1011 PFU at day 3	a: 104
Sheep	Bacterial challenge: oral 1010 CFU		43
	Group a: 8 phages, oral 109 PFU at 48 h and 72 h	Control for group a: 105 a: 104
	Group b: 8 phages, oral 1:1 PFU/CFU at 48 h and 72 h	Control for groups b, c, and d: 103 b: 101
	Group c: 8 phages, oral 10:1 PFU/CFU at 48 h and 72 h	c: 102
	Group d: 8 phages, oral 100:1 PFU/CFU at 48 h and 72 h	d: 103
Steers	Bacterial challenge: oral 5 × 1010 CFU	Control: <101	44
	Group a: 4 phages, oral 3.3 × 1011 PFU at days −2, 0, 2, 6, and 9	a: <101
	Group b: 4 phages, rectal 1.5 × 1011 PFU at days −2, 0, 2, 6, and 9	b: <101
	Group c: 4 phages, oral and rectal 4.8 × 1011 PFU at days −2, 0, 2, 6, and 9	c: <101
Steers	Bacterial challenge: oral 1011 CFU	Control: <101	46
	Group a: 4 encapsulated phages, oral 1010 PFU at days −1, 1, 3, 6, and 8	a: <101
	Group b: 4 encapsulated phages, in feed 1010 PFU at days −1, 1, 3, 6, and 8	b: <101
Steers	Bacterial challenge: oral 1010 CFU	Control: 101	45
	Group a: 2 phages, oral 1011 PFU at days 1, 2, and 3	Group a: 101
Sheep	Bacterial challenge: oral 1010 CFU	Control: 106	42
	Group a: 2 phages, oral 1011 PFU, at day 3	a: 5 × 103
	Group b: 1 phage naturally present in the gut	b: 2 × 103
Mice	Bacterial challenge: oral 108 CFU	Control: undetectable	38
	Group a: 16 phages, oral 109 PFU at −2.5 h, 2.5 h and twice daily during 3 days	a: undetectable (transient 0.2-log reduction)
	Group b: cocktail of 16 phages, 108 PFU at −2.5 h, 2.5 h and twice daily during 3 days	b: undetectable (transient 0.2-log reduction)
	Group c: cocktail of 4 phages, 109 PFU at −2.5 h, 2.5 h, and twice daily during 3 days	c: undetectable (transient 0.2-log reduction)

^aBacterial challenge was initiated at t = 0.

Escherichia coli O157:H7.

(i) Mice (three studies).

Tanji et al. (37) observed in phage-treated groups (single or repeated administrations started 2 days postchallenge) a decrease in E. coli fecal concentration of ∼1 log CFU g⁻¹ 5 days after bacterial challenge that was not observed 6 days later. Dissanayake et al. (38) obtained similar results within 1 day with an administration of phages twice a day before and after the bacterial challenge. In both studies, the impact of phage treatment on fecal level of E. coli was less pronounced at later time points (11 and 3 days, respectively). On the other hand, Sheng et al. (39) reported a prolonged impact on fecal E. coli levels of a single or two phages treatment started 24 h after the bacterial challenge and administered once a day during few days. Indeed, E. coli clones were detected in phage-treated groups up to 10 days only if an enrichment step was performed, while the untreated group excreted over 10² CFU g⁻¹.

(ii) Sheep (five studies).

Three studies showed no significant difference between phage-treated and control groups (39–41). Sheng et al. administered a single phage four times, starting 1 day after bacterial challenge and followed the animals for 21 days, after which phages were still detectable in the feces (39). Bach et al. reported the administration of a single dose of a single phage 2 days after the bacterial challenge, and monitored the animals for 28 days after phage treatment (40). The concentration of phages rapidly decreased and reached the detection threshold within 2 days. Raya et al. also used a single dose of a single phage, but they administered this dose 3 days after the bacterial challenge with a 2-day follow-up after phage treatment (41). In another study, Raya et al. reported a significant reduction in E. coli levels of 3 log CFU g⁻¹ throughout the gut 2 days after phage administration by a single dose on animals challenged 5 days before. Concomitantly, 10³ to 10⁶ PFU ml⁻¹ of phages were detected in the feces (42). Callaway et al. administered a single dose of an eight-phage cocktail at three different doses at 48 h and at 72 h postchallenge and found that the E. coli levels were significantly reduced at 72 h postchallenge (i.e., 24 h after the first phage administration) but not at later time points. However, at 96 h postchallenge, the bacterial loads in cecal and rectal contents were ∼2 log g⁻¹ lower in the phage-treated group given the lowest phage dose but not in the group given the 100-fold-higher dose (43).

(iii) Steers (four studies).

Three studies reported no difference between oral phage-treated and control groups over periods from 7 to 30 days posttreatment (44–46). The fourth study reported an initial decrease in fecal E. coli levels between 1 and 10 days after oral and rectal phage treatment, which did not last during the next 3 weeks (39). In all reports, phages isolated from feces were still able to infect the O157:H7 E. coli strain, but their impact on the E. coli intestinal load progressively decreased over time, independently of the administration route and frequency.

Other Escherichia coli serotypes.

(i) Mice/rats (nine studies).

Only three studies reported a significant reduction from 1 to 2 log g⁻¹ in the E. coli loads in feces from phage-treated animals by a single oral administration compared to controls (47–49). These data were independent of the administration time after bacterial challenge, being 1, 3, 8, or 10 days (Table 4). Two studies were performed with a cocktail of three phages (47, 48), while the third study used a single phage (49). In all studies, both phages and the targeted E. coli were still detected in feces, up to 2 weeks after phage treatment for the Galtier study (48).

TABLE 4.

Experimental phage treatments of E. coli-infected animals

Animals	Bacterial challenge and phage therapya	E. coli (CFU g−1)	Reference
Group a: calves	Bacterial challenge for a: oral 3 × 109 CFU; b and c: oral 3 × 108 CFU a: 2 phages, oral 1011 PFU at 1 h, at 8 h, or at onset of diarrhea	Control for group a: 1010; mortality 95% a: 4 × 104; mortality 35%	58
Group b: piglets	b: 2 phages, oral 1010 PFU at the onset of diarrhea	Control for group b: 5 × 107; mortality 60% b: 105; mortality 0%
Group c: lambs	c: one phage, oral 109 to 1010 PFU at 8 h	Control for group c: 4 × 109; mortality 30% c: 6 × 107; mortality 0%
Mice	Bacterial challenge for a, b, c, and d: oral 5 × 107 CFU; e and f: oral 104 CFU Group a: 4 phages, in drinking water 106 PFU ml−1	Control for group a: undetectable a: undetectable	53
	Group b: 4 phages, in drinking water 106 PFU ml−1 + ampicillin 20 μg ml−1 during the first week	Control for group b: 103 to 105 b: undetectable
	Group c: 4 phages, in drinking water 106 PFU ml−1 + ampicillin starting at day −3	Control for group c: undetectable to 107 c: undetectable to 101
	Group d: 1 phage, in drinking water 105 PFU ml−1, at day 7	d: 108 to 104 after treatment (no control group)
	Group e: 1 phage, in drinking water 105 PFU ml−1, at day 0	e: 109 and 106 to 104 after treatment (no control group)
	Group f: 1 phage, in drinking water 105 PFU ml−1, at day −7	f: 109 (no control group)
Mice	Bacterial challenge: oral 1010 CFU	Control: 105 to 108	50
	Group a: cocktail of 3 phages, 109 PFU ml−1 in drinking water	a: 105 to 108
Mice	Bacterial challenge: oral 5 × 107 CFU Group a: 2 phages, oral 5 × 104 PFU after colonization	Control for group a: 106 a: 106	51
	Group b: 2 phages, oral 107 PFU after colonization	b: 3 × 109 (data not shown for control group)
	Group c: 1 phage, oral 106 PFU after colonization	c: same as control (data not shown)
	Group d: 1 phage, oral 108 PFU after colonization	Control for group d: 107 d: 107
	Group e: 1 phage, oral 106 PFU at day −7	e: 109 (no control group)
	Group f: 1 phage, oral 106 PFU at day 3	f: high titers (data not shown)
Pigs	Bacterial challenge: oral 1010 CFU	Percentage of E. coli excretion	57
	Group a: 1 phage, oral 1010 PFU at 15 min	Control for group a: 83% a: 3 to 37%
	Group b: 3 phages, oral 109 PFU at 15 min	Control for group b: 63% b: 38%
	Group c: 2 phages, oral 108 PFU at 24 h, 30 h, and 36 h	Control for group c: 21% (108 CFU g−1) c: 80% (107 CFU g−1)
Mice	Bacterial challenge: oral 106 CFU	Control: 108	52
	Group a: 3 phages, in drinking water 3 × 108 PFU ml−1 at day 3 during 24 h	a: 108
Mice	Bacterial challenge: oral 106 CFU	Control: 109	54
	Group a: 3 phages, in drinking water 3 × 108 PFU ml−1 at day 3 during 24 h	a: 109
	Group b: 3 phages, in drinking water 3 × 1010 PFU ml−1 at day 3 during 24 h	b: 109 (transient 3-fold reduction)
Rats	Bacterial challenge: in drinking water 108 CFU ml−1	Control: 3 × 105	55
	Group a: 140 phages, in drinking water 107 PFU ml−1 during 20 days	a: 3 × 105 (transient 2-log reduction)
	Group b: 140 phages, oral 4 × 107 PFU, three times per day during 20 days	b: 3 × 105 (transient 2-log reduction)
	Group c: 140 phages, oral capsules 5 × 106 PFU, three times per day during 20 days	c: 5 × 105 (transient 4-log reduction)
Mice	Bacterial challenge: oral 107 CFU	Control for groups a and b: 109	47
	Group a: 1 phage, oral 2 × 107 PFU at day 4	a: 109
	Group b: 3 phages, oral 2 × 107 PFU at day 4	b: 108
	Group c: 3 phages, oral 6 × 105 PFU at day 7	Control for groups c and d: 108 c: 105
	Group d: 3 phages, oral 6 × 107 PFU at day 7	d: 105
Mice	Bacterial challenge for a: oral 109 CFU; b and c: oral 108 CFU	Control: 108	48
	Group a: 3 phages, oral 3 × 107 PFU at day 1 (2 doses)	a: 106
	Group b: 3 phages, oral 3 × 107 PFU at day 8	b: 106
	Group c: 3 phages, oral 3 × 107 PFU at day 10	c: 105
Rabbits	Bacterial challenge: oral 1010 CFU	eaeA gene concn (10−6 μg/μl) Control: 16.0	56
	Group a: 1 phage, oral 1010 PFU, at day 3	a: 1.1
Mice	Bacterial challenge: oral 4 × 106 CFU	Control: 8 × 108	49
	Group a: 1 phage, oral 4 × 108 PFU, at 24 h	a: 5 × 107

^aThe bacterial challenge was initiated at t = 0.

The six other studies showed no (50–53) or only transient (54, 55) reductions in the fecal levels of E. coli. This reduction was dose dependent in the study of Maura et al. since a 100-fold reduction in the oral phage dose had no significant impact 24 h after phage administration compared to controls (54). Nevertheless, fecal levels of phages remained roughly stable over 2 weeks at ∼10⁷ PFU g⁻¹ (52). By administering a cocktail of 140 phages continuously in drinking water, or three times a day for 20 days after bacterial challenge, Abdulamir et al. observed a transient reduction of E. coli fecal titers of ∼2 log CFU ml⁻¹ between days 5 and 8, followed by a progressive increase until day 20. Parallel to this bacterial increase, the fecal titer of phages decreased from day 8 and was no longer detectable on day 20 (55). In all other studies (50, 51, 53), whether the phage was administered alone or in a cocktail or before or after the bacterial challenge, no significant reduction in E. coli fecal levels was observed or only in some parts of the gut, such as the stomach or the small intestine.

(ii) Other species (rabbit, pig, calf, and lamb—five studies).

Zhao et al. (56) noticed a 15-fold decrease in the cecal levels of challenged bacteria 3 days after administration of a single oral dose of one phage to rabbits or a single oral dose of 20 mg of ciprofloxacin. Using pigs, which developed diarrhea upon administration of an E. coli strain, Jamalludeen et al. (57) reported a significant decrease in the E. coli fecal level 6 days postchallenge compared to the control when a single dose of one phage was orally administered, but not with a three-phage cocktail. Surprisingly, the fecal levels of phages following the single phage treatment were below the threshold of detection, whereas they were 1 to 2 log g⁻¹ above the threshold of detection after the treatment with the cocktail. Nevertheless, the authors reported a reduction in the diarrhea score with both treatments. With three animal species and a different E. coli serotype for each, Smith et al. (58) observed a reduction in animal mortality congruent with a lower bacterial shedding when they administered one or two phages once the animals showed signs of diarrhea.

Synthesis of experiments with E. coli (all serotypes).

In mice, seven studies reported a decrease in E. coli levels in phage-treated groups, whereas four studies showed no effect, and one study did not include enough animals to be conclusive (53). In most cases, the levels of the E. coli strain introduced in the guts of mice remained low and sometimes tended to decrease within a few days. Under these conditions, the impact of phage application on E. coli levels was often significant (37–39, 47–49). In contrast, studies performed with models of high and stable levels of E. coli gut colonization showed weak or no phage impact (51–54). Therefore, the data showed that the density of the target bacteria critically affects the efficacy of phages in the murine gut.

Experiments performed with sheep revealed the same shortcoming of unstable E. coli gut colonization seen in mice and lasting no more than 1 week. Interestingly, despite no significant impact on E. coli fecal titers, oral phage administration reduced the E. coli levels in intestinal organs (41–43). With steers, E. coli colonization was more stable over several days, with a decline observed only after several weeks. Nevertheless, weak or no efficacy of phages was also observed (44–46). In studies with pigs and calves, E. coli gut colonization was stable, and phages administered orally led to a significant reduction in E. coli titers in the feces (57, 58). In the latter two animals, E. coli intestinal colonization was associated with a disease; this was not the case for any of the sheep, steer, or mouse models. This observation suggests that the pronounced efficacy of phage treatments under such conditions could be linked to the additive action of the immune defense against the pathogen. An alternative explanation would be that the physiology of the pathogen causing the disease may favor phage replication.

Listeria monocytogenes

A single study with mice reported that a moderate dose (10⁵ PFU) of a six-phage cocktail administered daily before and after the bacterial challenge led to a decrease in the fecal level of L. monocytogenes 3 days after bacterial challenge (Table 5). This decrease was as strong as the one observed for antibiotic-treated animals. However, phage titers in feces and cecal contents were below the threshold of detection. This is consistent with a weak amplification of phages caused by the low abundance of L. monocytogenes in intestinal organs, i.e., ∼10² CFU g⁻¹ (59).

TABLE 5.

Experimental phage treatments of L. monocytogenes-infected animals

Animals	Bacterial challenge and phage therapya	L. monocytogenes (CFU g−1)	Reference
Mice	Bacterial challenge: oral 105 CFU	Control: 90	59
	Group a: 6 phages, oral 1 × 105 PFU daily from days −3 to 3	a: <10

^aThe bacterial challenge was initiated at t = 0.

Pseudomonas aeruginosa

In a murine model of intestinal sepsis, a single phage was administered orally once (Table 6). The observations at 10 days postchallenge revealed that a pretreatment 1 day before or after the challenge did not affect the survival rate of animals compared to untreated controls. In contrast, a treatment administered 6 days postchallenge increased the survival up to 66%. The fecal concentrations of P. aeruginosa were only significantly reduced by <1 log CFU g⁻¹ for both postchallenge treatments (60).

TABLE 6.

Experimental phage treatments of P. aeruginosa-infected animals

Animals	Bacterial challenge and phage therapya	P. aeruginosa (CFU g−1)	Reference
Mice	Bacterial challenge: in drinking water 108 CFU ml−1 during 3 days	Control: 3 × 105	60
	Group a: 1 phage, oral 1010 PFU at day −1	a: 2 × 105
	Group b: 1 phage, oral 1010 PFU at day 1	b: 2 × 104
	Group c: 1 phage, oral 1010 PFU at day 6	c: 5 × 104

^aThe bacterial challenge was initiated at t = 0.

Salmonella enterica Serovar Typhimurium

All of the studies included (n = 7) were performed with pigs with the goal to propose phages for lowering food contamination (Table 7). Three studies reported a significant reduction of S. Typhimurium levels in gut sections (ileum and cecum) but not in the feces of phage-treated animals compared to control groups at both 6 and 48 h posttreatment (61–63). In these studies, phages were administered by either oral, intramuscular, or intraperitoneal routes a few hours after the bacterial challenge. Saez et al. used microencapsulated phages orally administered three times every 2 h postchallenge and observed a significant reduction in Salmonella counts of 2 logs in the ileum at 6 h postchallenge compared to untreated animals. When phages were administered daily for 5 days before the bacterial challenge, their levels 6 h postchallenge reached higher values in ileal and cecal contents compared to animals that received phages every 2 h postchallenge (>10⁶ PFU ml⁻¹ versus >10³ PFU ml⁻¹) (63). The study performed by Callaway et al. reported a trend toward a reduction in Salmonella counts at 96 h postchallenge, which was, however, not significant. These researchers used a cocktail of two phages administered orally 24 and 48 h after the bacterial challenge, and the phage levels reached 10⁴ PFU ml⁻¹ in the intestinal contents at 96 h (64). Using a six-phage cocktail administered at different concentrations 2 days postchallenge, Albino et al. observed no impact of Salmonella counts compared to controls 18 h later (65). When phages were mixed with food (2 × 10⁹ PFU kg⁻¹), pigs displayed lower fecal shedding scores than did the control groups on days 7 and 14 postchallenge (66). At 6 h after the administration of 15 microencapsulated phages, Wall et al. reported that Salmonella counts in ileal and cecal samples were reduced, but all feces remained positive for Salmonella in phage-treated and control groups (67).

TABLE 7.

Experimental phage treatments of S. Typhimurium-infected animals

Animals	Bacterial challenge and phage therapya	S. Typhimurium (CFU g−1)	Reference
Pigs	Bacterial challenge: intranasal 108 CFU	Control for groups a, b, and c: 104	61
	Group a: 26 phages, intraperitoneal 1.2 × 109 PFU at 18 h	a: 5 × 104
	Group b: 26 phages, intramuscular 1.2 × 109 PFU at 18 h	b: 7 × 103
	Group c: 26 phages, oral 1.2 × 109 PFU at 18 h	c: 5 × 1043
	Group d: 1 phage, oral and intramuscular 2 × 1010 at 3 h	Control for group d: 2 × 10 d: <102
Pigs	Bacterial challenge: intranasal 5 × 108 CFU	Control: 2 × 103	62
	Group a: 1 phage, oral and intramuscular 2 × 1010 PFU at 3 h	a: <102
Pigs	Bacterial challenge for a: oral 5 × 108 CFU; b: oral 5 × 109 CFU Group a: 15 microencapsulated phages, oral 1010 PFU at 0 h, 2 h, and 4 h	Control for group a: 4 × 103 a: 3	67
	Group b: 15 microencapsulated phages, oral 1010 PFU at 48 h, 50 h, and 52 h	Control for group b: 8 × 102 b: 3 × 101
Pigs	Bacterial challenge: oral 109 CFU Group a: 1 phage, in feed 2 × 109 PFU kg−1	Shedding score (0 to 3) Control: 1.4 a: 0.2	66
Pigs	Bacterial challenge: oral 5 × 109 CFU	Control: 5 × 103	63
	Group a: 14 microencapsulated phages, in feed 5 × 1011 PFU per day from day −5 to day 0	a: 5 × 102
	Group b: 14 microencapsulated phages, oral 5 × 1011 PFU at 0, 2, 4, and 6 h	b: 2 × 103
Pigs	Bacterial challenge: oral 2 × 1010 CFU	Control: 2 × 101	64
	Group a: 2 phages, oral 3 × 109 PFU at 24 h and 48 h	a: 2
Pigs	Bacterial challenge: oral 5 × 105 CFU (2 doses)	Control: 3 × 103	65
	Group a: 6 phages, orally (a) 103 PFU ml−1 at day 3	a: 3 × 103
	Group b: 6 phages, orally (a) 105 PFU ml−1 at day 3	b: 3 × 103
	Group c: 6 phages, orally (a) 107 PFU ml−1 at day 3	c: 3 × 103
	Group d: 6 phages, orally (a) 109 PFU ml−1 at day 3	d: 3 × 103

^aThe bacterial challenge was initiated at t = 0.

Most of the experiments reported were performed to address a highly specific challenge: the suitability of a phage application prior to meat processing (61, 62, 65, 67). Therefore, experimental settings included phage administration only 6 h before sacrifice but did not always include the intestinal readout of such treatment. The results showed that Salmonella levels in the gut tend to decrease, but never significantly, compared to the controls. Encouragingly, the only long-term study (2 weeks), during which the authors analyzed the impact of phages in pigs carrying Salmonella showed a significant reduction of the Salmonella titers in fecal samples (66).

Shigella sonnei

The oral administration to mice of a commercial product including five phages (ShigActive) 1 h before or after the bacterial challenge, or both, led to reduced fecal levels of S. sonnei 48 h postchallenge (Table 8). These regimens were more effective than ampicillin at both 24 and 48 h. It should also be mentioned that mice naturally eliminated Shigella within 72 h (68).

TABLE 8.

Experimental phage treatments of S. sonnei-infected animals

Animals	Bacterial challenge and phage therapya	P. aeruginosa (CFU pellet−1)	Reference
Mice	Bacterial challenge: oral 108 CFU	Control: 1114	68
	Group a: 5 phages, oral 109 PFU at −1 h	a: lower than control (data not shown)
	Group b: 5 phages, oral 109 PFU at 1 h	b: 110
	Group c: 5 phages, oral 109 PFU at 3 h	c: lower than control (data not shown)
	Group d: 5 phages, oral 109 PFU at −1 h and 1 h	d: 26 (most effective group)

^aThe bacterial challenge was initiated at t = 0.

Vibrio cholerae

Four studies with animal models of V. cholerae infection were analyzed (Table 9). In 1963, Dutta et al. noted that rabbits treated with a single phage 1 or 8 h postchallenge survived longer than untreated controls. These results were obtained with five different phages (69). In a more recent study, rabbits infected by the oral administration of V. cholerae did not develop symptoms when a single phage treatment was given either before or after the challenge. Both treatment regimens led to a reduction in V. cholerae counts in the cecal fluids by 24 h postchallenge. High levels of phage replication, up to 10⁷ PFU g⁻¹, were detected in the gut when the treatment was performed postchallenge. Slightly lower numbers, ∼10⁵ PFU g⁻¹, were recovered with the prophylactic treatment (70). Using murine and rabbit models, Yen et al. tested a preventive phage treatment, which reduced the levels of V. cholera in the gut 24 h postchallenge. These researchers also showed that a three-phage cocktail performed better than did individual phages (71). In another study, a five-phage cocktail significantly decreased the levels of viable V. cholerae in the intestinal tissues at days 1 and 4 postchallenge of phage-treated mice compared to untreated controls (72).

TABLE 9.

Experimental phage treatments of V. cholerae-infected animals

Animals	Bacterial challenge and phage therapya	V. cholerae (CFU g–1) and outcomes	Reference
Rabbits	Bacterial challenge: intraintestinal 104 vibrios per 100 g of body wt Five different phages tested individually Group a: at −1 h Group b: at 8 h Group c: at 16 h Group d: after onset of diarrhea	Mortality rate (n/n) Control: 4/4 Phage R: a: 0/6; b: 2/8; c and d: 6/6 Phage 138: a: 0/4; b: 2/4; c and d: 4/4 Phage 145: a: 1/4; b: 4/4; c and d: 4/4 Phage 149: a: 1/4; b: 2/4; c and d: 4/4 Phage 163: a: 0/4; b: 0/4; c and d: 4/4	69
Mice	Bacterial challenge: oral 5 × 107 CFU	Control: 1010	72
	Group a: 5 phages, oral 1 × 107 PFU at days 1, 2, and 3	a: 9 × 103
Group a: mice	Bacterial challenge for a: oral 5 × 105 CFU	Control for group a: 2 × 107 to 9 × 107	71
	a1: 1 phage, oral 106–107 at −3 h	a1: 105
	a2: 1 phage, oral 106–107 at −3 h	a2: 102
	a3: 1 phage, oral 106–107 at −3 h	a3: 101
	a4: 3 phages, oral 106–107 at −3 h	a4: 101
	a5: 3 phages, oral 5 × 105 to 9 × 105 at −6 h	a5: 101
	a6: 3 phages, oral 5 × 105 to 9 × 105 at −12 h	a6: 4 × 104
	a7: 3 phages, oral 5 × 105 to 9 × 105 at −24 h	a7: 2 × 106
	a8: 3 phages, oral 108 PFU at −6 h	a8: 105
	a9: 3 phages, oral 108 PFU at −12 h	a9: 4 × 105
	a10: 3 phages, oral 108 PFU at −24 h	a10: 8 × 105
Group b: rabbits	Bacterial challenge for b: oral 5 × 108 CFU	Control for group b: 1010
	b1: 3 phages, oral 4 × 109 to 8 × 109 PFU at −3 h	b1: 102
	b2: 3 phages, oral 4 × 109 to 8 × 109 PFU at −24 h	b2: 3 × 107
Rabbits	Bacterial challenge: oral 5 × 108 CFU	Control: 2 × 107	70
	Group a: 1 phage, oral 109 PFU at −6 h	a: 3 × 103
	Group b: 1 phage, oral 109 PFU at 6 h	b: 3 × 103

^aThe bacterial challenge was initiated at t = 0.

Cholera is one of the earliest human infections for which phage therapy has been tested (73). Experimental results confirmed the efficacy of this treatment in both mice and hamsters (70–72). Impressively, the results are consistent among studies performed 65 years apart. Here, as with E. coli in calf and pig models, the efficacy of phage treatments was observed when administered to animals developing a disease.

Yersinia enterocolitica

In a murine model, oral administration of a single dose of one phage 6 h after bacterial challenge resulted in a significant reduction of the bacterial load within 18 h (4-log CFU g⁻¹ reduction) in the cecum and colon (Table 10). This reduction persisted until 144 h postchallenge, but to a lesser extent (2 log CFU g⁻¹). A limited increase in the phage titer was observed in the cecum and colon at 12 h postadministration, and a decrease was subsequently observed until 48 h, after which phages were no longer detected. No significant histopathologic lesions in the cecum and a lower level of proinflammatory cytokines were observed in phage-treated mice as opposed to the control untreated group (74).

TABLE 10.

Experimental phage treatments of Y. enterocolitica-infected animals

Animals	Bacterial challenge and phage therapya	Y. enterocolitica (CFU g−1)	Reference
Mice	Bacterial challenge: oral 2 × 108 CFU	Control: 104	74
	Group a: 1 phage, oral 109 PFU ml−1 at 6 h	a: 102

^aThe bacterial challenge was initiated at t = 0.

OVERVIEW

Altogether, two-thirds of the studies discussed above showed that phage treatments had either a transient or no efficacy in reducing the levels of the target bacteria, showing clearly that intestinal phage therapy is not as easy as one could expect. Therefore, in-depth analysis of studies must be undertaken to highlight the conditions driving successful phage therapy treatments in the gut. We organized this analysis under the form of documented answers to three key questions.

Is the Amplification of Phages in the Gut Supported by Experimental Data?

The answer to this question, which is expected to be positive given the self-amplification of phages on their target bacteria, is not so definitive. In some studies, the data are lacking, while in others, the phage levels remained relatively low, mirroring the levels of the target bacteria. Nevertheless, the transit of phages throughout the gut was not found to be a major hurdle since phages were recovered from the fecal contents of control animals. In addition, a buffer to neutralize the gastric acidity was often used (23, 51, 75). Therefore, in most of the studies phage amplification was reported. Sometimes, an increased number of phages compared to the dose administered was observed within 24 h, testifying unambiguously to phage replication (35, 47, 48, 52, 57, 70). More striking was the continuous presence of phages in feces during weeks from animals that received a single phage administration. However, the overall consequence of in vivo phage replication did not match with a parallel reduction in target bacteria except when animals developed intestinal diseases. In fact, our analysis of both phage and bacteria populations distinguished three situations depending on the behavior of a given bacterial strain in animals (Fig. 1): (i) when the phage-targeted bacterium mimics the gut transit of an opportunistic pathogen with no stable colonization and without developing a disease, phage application accelerates the pace at which the bacterium is washed out (transit model); and (ii) when the bacterium stably colonizes the gut without affecting the animal’s health, behaving like a commensal strain, long-term coexistence of phage and bacteria populations is observed, without major impacts on bacterial colonization levels (coexistence model). Recently, a study by Hsu et al., in which multiple bacteria were targeted altogether, reported the coexistence of phage T4 and its E. coli target after an initial drop of the fecal level of E. coli, suggesting that such coexistence is probably more frequent than anticipated (76). (iii) Finally, when intestinal bacterial infections occur, the administration of phages significantly decreases the level of pathogenic bacteria (infection model). This model would fulfill the expectations of efficient intestinal phage therapy. While in such a situation successive cycles of phage amplification may be expected, it remains possible that a single strong dose could be sufficient to reduce the burden of the targeted intestinal pathogen (passive phage therapy) and allow other antibacterials, mainly antibiotics, and the immune system to eliminate the remaining infectious bacteria.

FIG 1.

Three models recapitulate the variation of the bacterial populations in the gut of mammals receiving a phage treatment (red) compared to untreated controls (blue). Following phage introduction in animals colonized by their target bacteria, three models were defined: transit, coexistence, and infection. These models correspond to the outcomes directly linked to the behavior of bacteria toward animals. Either bacteria could not stably colonize an animal’s gut (transit model), or they colonized stably at high levels (coexistence model), or they induced an intestinal infection (infection model).

Are Transit, Coexistence, and Infection Models Relevant in Humans?

First, the transit model could be hard to assess during a clinical trial, since intestinal colonization may not last long enough to directly test the efficacy of phage treatments. Indeed, human intestinal viromes have revealed that phages are abundant and diverse (77). These resident phages may reduce the duration of a transient intestinal colonization or even prevent the colonization of incoming opportunistic pathogens. Nevertheless, frequent samples of feces could be collected, and interactions between intestinal phages and bacteria could be experimentally tested. A recent study of E. coli phages residing in the gut of children revealed that in a few cases some phages were dominating the E. coli phage population (78). This observation suggests that prompt amplifications of resident phages could reflect their protective role against incoming bacteria.

Second, the coexistence model is in agreement with the longitudinal study of viromes from healthy humans, revealing their long-term stability and strong link to the bacterial microbiome (79). This stability expresses the resilience of the microbiota as a community able to maintain its equilibrium state. This is a characteristic that is unique to this environment compared to others, such as skin or lungs, where the efficacy of phage therapy appears to be more promising (80, 81). Therefore, the lack of evidence from experimental models, supporting that oral phage application can efficiently decrease intestinal bacteria, may be in part related to the absence in these models of a disease context that could provide a more favorable environment for phage activity.

Third, the infection model relates to classical phage therapy treatments. Here, the goal of the phage administration is to reduce the load of the pathogen in order to restore a healthy environment. In immunocompetent animals, this would most likely rely on the synergistic action of phages and immune cells, defined as immunophage synergy, which has been experimentally demonstrated during pulmonary phage therapy but not yet during the treatment of intestinal infections (18). It is also worth mentioning that the immune response itself can provoke host damages that could be more deleterious than bacterial multiplication (82). We can also hypothesize that other antibacterial defenses, such as antimicrobial peptides or commensal members of the gut microbiota (colonization resistance), would likely participate in the overall success of the treatment, as well as antibiotics, which are very often administered with phages during compassionate treatments (83).

Overall, the three proposed models sound plausible in the human context, but they await clinical data to be more firmly supported. In particular, extrapolation of data from rodent models ignores the specific behaviors of these animals in terms of food regimens and nocturnal activities, among other characteristics (84). Perhaps, more acutely, the dose of the pathogen used in experimental models is often several orders of magnitude higher than the bacterial load to which a human may be usually exposed. For instance, infecting a 20-g mouse with 10⁸ CFU of Salmonella is equivalent to a human of 70 kg ingesting a piece of food contaminated by 3.5 × 10¹¹ CFU, whereas the infectious dose during outbreaks in human was estimated to be in the 10³- to 10⁵-CFU range (85). A large inoculum in experimental models leads to an increase in the abundance of the prey population for phages but also speeds the infection of the mammalian host, which may result in lethal cellular damage despite clearance of the pathogen.

What Are the Limiting Factors Affecting Phage Efficacy in the Gut?

The optimistic approach to rely on the amplification of phages to decrease the density of intestinal bacteria, based on their rapid in vitro efficacy, is not unambiguously supported by the data, suggesting that the in vivo efficacy of phages is jeopardized. Three main barriers to phage infection in the gut are anticipated. First, phages need to reach the gut. Second, phages must find a susceptible target. Third, phages have to achieve a successful infection. How phages will mechanistically overcome these barriers will depend not only on their nature (physical, chemical, or biological challenges) but also on the environment phages will face at three levels that are the host, the organ, and the bacterial cell (Fig. 2).

FIG 2.

Several factors in the host, organ, and cell influence phage activity in the gut. The health state of the host (left) imposes a global physiological environment with prolonged consequences on the intestinal microbiota including phage-bacteria interactions. In the organ (middle), the cellular environment (immune cells, other microbes) affects bacterial physiology (pH, oxygen) with direct consequences on phage dynamics (prophage induction and bacterial susceptibility to phages). Finally, in the cell (right), different bacterial defense mechanisms will impact the outcome of phage-bacterium interactions. Epithelial cells, pink; mucus layer, gray; DNA molecules correspond to bacterial (blue) and phage (red) genetic material; CRISPR/Cas, clustered regularly interspaced short palindromic repeats/CRISPR associated.

In addition to an acidic pH, which was mentioned earlier, hydrolytic enzymes can also reduce the number of phage particles transiting in the gut. In studies with a cocktail of T4-like phages fed to mice not colonized with bacteria supporting phage replication, Denou et al. estimated that a substantial decrease in phages occurred during gut transit. These authors also observed that a 3-log-lower dose of phage administered to human volunteers still led to the detection of phages in all fecal samples, while the gastric acidity is lower in humans than in mice (50). These observations were reported multiple times with mouse models (23, 51, 75). This illustrates the necessity of performing biodistribution studies to narrow the time and numbers of phage particles reaching each gut section to get a better understanding of the phage/bacterium ratio needed for phage amplification. An example of a study that includes a large set of treatment modalities was published by Smith et al. in 1987, with phage doses ranging from 50 to 10¹⁰ PFU and administration times ranging from pre- to postinfection, as well as from the continuous exposure via the environment (calf litter) to manage E. coli diarrhea (86). It should, however, be noted that in this study a nonenteropathogenic E. coli isolate and an uncharacterized Lactobacillus strain were always administered with the E. coli pathogenic strain targeted by the phages, without any information provided on whether or not phages were able to infect the nonenteropathogenic strain, which raises questions regarding the conclusions drawn by the authors. A solution to the degradation of phage particles during the intestinal transit may reside in the use of phages encapsulated in liposomes, polymers, or other formulations. Promising observations have been reported but need to be confirmed, and the kinetics of phage release in the gut must be thoroughly investigated (87, 88). Finally, recent data from a mammalian cell model showed that the gut epithelium internalizes virions, a phenomenon that could further reduce the number of available phages in the gut (89).

Nonetheless, once the phages have reached the gut section, where the target bacteria reside, these phages need to find a susceptible host. Indeed, the gut environment, such as the availability of nutrients or the oxygen concentration, among other factors, affects the bacterial physiology, which in turn could affect the susceptibility of bacteria to phages (90–92). Moreover, a pathogen will likely perturb the host physiology, which in turn could affect bacterial susceptibility to phages (Fig. 2) (93, 94). In addition, the spatial distribution of phages and bacteria in gut sections may be beneficial for one or the other population, promoting their coexistence, as has been recently shown in vivo (75). The property of some phages to bind to mucins would alter their diffusion but may increase their ability to reach bacteria embedded in this extracellular structure lying at the surfaces of epithelial cells (95, 96). Therefore, solutions to overcome the aforementioned challenges reside in the selection of phages with improved abilities to reach their target, which requires a preexisting knowledge of what is or are the specific behaviors of these targets in the gut.

Once they reach the surfaces of bacterial cells, phages have to defeat bacterial defense systems that drive the growth of phage-resistant bacteria. These defenses include mutations, surface modifications, restriction-modification, abortive infection, and CRISPR (clustered regularly interspaced short palindromic repeats) systems (Fig. 2). A variety of novel systems has rapidly developed during the past few years but, thankfully for phages, not every bacterium possesses all of them (28). Indeed, developing phage resistance can affect bacterial fitness, as shown by the reduced virulence of such phage-resistant bacteria in animal models of infection, which overall could be beneficial for patients treated with phages (97–100). In addition, most of the recently uncovered phage resistance systems do not provide a full phage resistance phenotype but instead decrease susceptibility by several logs, leaving opportunities for phages to amplify. Indeed, through ages of coevolution with bacteria, phages defeated these defense systems by evolving genetic variants, modifying their nucleotidic bases, or even inventing new functions, such as anti-CRISPR proteins (Fig. 2). Reciprocally for bacteria, not all phages possess the ability to develop a large variety of counter-resistance measures. It has also been shown that phages can jump from susceptible to resistant bacteria in the mouse gut by a single mutation in their tail fibers, which demonstrates that phages can rather quickly adapt to fluctuating bacterial populations in order to persist in this organ (101). To conclude, every phage-bacterium combination possesses its own potential for coevolution; this highly complicates the choice of phages and their combinations (phage cocktails) for therapeutic applications.

Finally, another layer of complexity is provided by temperate phages hidden within bacterial genomes. Their abundance in the gut has been shown to be higher in IBD patients compared to healthy adults (12). This abundance is necessarily linked to phage activity, i.e., their excision from bacteria. Some factors, such as diet and inflammation, induce prophage excision in the gut, but many remain to be identified, as well as the molecular cascade involved (102, 103). Lysed bacteria upon prophage induction will increase bacterial debris to which virulent phages could bind, and this may lower the dose of available phages for therapeutic interventions. Likewise, when both temperate and virulent phages recognize the same receptor, they could compete with each other. It has also been reported in a study of model phages (lambda and T4) that lysogens, bacteria carrying prophages, prevent the infection of bacteria by expressing defense systems against virulent phages (abortive infection) (104). In addition, homoimmunity prevents lysogens from being infected by closely related temperate phages (105). Nevertheless, temperate phages may also be used as a solution when virulent phages are not readily isolated. Their genetic modifications, while inserted in the bacterial chromosome, are accessible in order to make them virulent and therefore more suitable for applications, as shown in animal models of C. difficile infections or, recently, during the compassionate treatment of a Mycobacterium abscessus infection in a cystic fibrosis patient (34, 106). Overall, both virulent and temperate phages will face the same challenges in the gut, with perhaps an advantage to the temperate phages, since during their stay within intestinal bacteria they may acquire functions useful to survive in this environment.

CONCLUSIONS

During billions of years of evolution, phages have infected bacteria in different environments, and the guts of mammals are certainly not an exception. Therefore, the lack of success in using phages in therapeutic interventions in the gut reflects more our limited knowledge of this ecosystem than the intrinsic capacity of phages to infect intestinal bacteria. In particular, there is a very poor mechanistic appreciation of the factors that govern phage activity in the gut. The multiple experimental models examined here clearly show a breadth of possibilities rather than merely highlighting a path to success. In order to improve our knowledge obtained from animal models, we recommend that future studies incorporate the following: (i) timely quantification of viable phage particles supporting phage amplification; (ii) monitoring of phage resistance over time and treatments; (iii) avoidance of models for which the bacterial transit time is lower than 72 h, unless it reflects intestinal acute infections; and (iv) administration of phages at least 4 h after the bacterial challenge to allow bacteria to adapt to the intestinal environment, unless pretreatment (at least 4 h ahead of the bacterial challenge) is being evaluated.

Future research in this area will require bridging data from metagenomic studies with simplified experimental models. One possible way to progress in this direction would be to build increasingly complex synthetic systems, such as those based on gnotobiotic/isobiotic murine models (107). Although still imperfect in several ways, these models offer a very high reproducibility backbone that would help in addressing specific questions regarding molecular mechanisms. Nevertheless, the most convincing intestinal phage therapy data examined here arose from models in which an intestinal disease was ongoing, which is directly in line with the growing list of successful compassionate treatments currently provided to patients facing therapeutic inefficacy of antibiotic treatments. This argues for the necessity to consider the host, in addition to bacteria and phages, in identifying the conditions required for optimal efficacy of phage applications in the gut.

Biographies

François Javaudin, M.D., M.S., is Assistant Professor of Emergency Medicine at Nantes University Hospital, Nantes, France. He is currently a Ph.D. student at the MiHAR Lab (Microbiotas, Hosts, Antibiotics, and bacterial Resistances, University of Nantes), where the prevention of antibiotic resistance is a key research topic, as well as in the emergency department of the Nantes University Hospital. Experimental research in the MiHAR Lab includes the use of bacteriophages to reduce the intestinal carriage of multiresistant bacteria.

Chloé Latour, M.D., is a physician at Pontivy Hospital, Pontivy, France. She received her medical degree at Nantes University Hospital, where she worked for several years in the emergency unit.

Laurent Debarbieux, Ph.D., is leading the Bacteriophage, Bacterium, Host Laboratory of the Institut Pasteur. Following an initial training in the molecular biology of the bacterial cell at the University of Lille, France, and at Harvard Medical School, Boston, MA, he turned his attention to bacteriophages in 2006. Since then, he has used mainly experimental animal models to decipher the mechanisms governing the activity of bacteriophages targeting bacterial pathogens.

Quentin Lamy-Besnier is a Ph.D. student at Institut Pasteur. He graduated from the biology department of the Ecole Normale Supérieure with a M.S.c in ecology and evolution. His research focuses on understanding bacteriophage-bacterium interactions in the mammalian gut across scales from animal studies to molecular mechanisms. He also has bioinformatics skills that he deploys to analyze viromes.