Tunable control of B. infantis abundance and gut metabolites by co-administration of human milk oligosaccharides

ABSTRACT

Precision engineering of the gut microbiome holds promise as an effective therapeutic approach for diseases associated with a disruption in this microbial community. Engrafting a live biotherapeutic product (LBP) in a predictable, controllable manner is key to the consistent success of this approach and has remained a challenge for most LBPs under development. We recently demonstrated high-level engraftment of Bifidobacterium longum subsp. infantis (B. infantis) in adults when co-dosed with a specific prebiotic, human milk oligosaccharides (HMO). Here, we present a cellular kinetic-pharmacodynamic approach, analogous to pharmacokinetic-pharmacodynamic-based analyses of small molecule- and biologic-based drugs, to establish how HMO controls expansion, abundance, and metabolic output of B. infantis in a human microbiota-based model in gnotobiotic mice. Our data demonstrate that the HMO dose controls steady-state abundance of B. infantis in the microbiome, and that B. infantis together with HMO impacts gut metabolite levels in a targeted, HMO-dependent manner. We also found that HMO creates a privileged niche for B. infantis expansion across a 5-log range of bacterial inocula. These results demonstrate remarkable control of both B. infantis levels and the microbiome community metabolic outputs using this synbiotic approach, and pave the way for precision engineering of desirable microbes and metabolites to treat a range of diseases.

KEYWORDS:

• Gut microbiome
• gut microbiota
• microbiome modulation
• Live biotherapeutic product (LBP)
• synbiotic
• Bifidobacterium
• B. infantis
• Human milk oligosaccharides (HMO)

Introduction

The use of live biotherapeutic products (LBPs) to manipulate gut microbiomes is a successful strategy for treating recurrent Clostridioides difficile infection (rCDI), as evidenced by the recent FDA approval of two different therapies, REBYOTA^TM (FDA STN 125,739) and VOWST^TM (FDA STN 125,757). LBPs, an emerging class of therapeutic microbes manufactured following rigorous safety standards,^1,2 are also currently under development for other therapeutic uses including treatment of inflammatory bowel and graft-versus-host diseases (e.g. Clinical Trial IDs: NCT06102213, NCT05370885, NCT04995653), and as a co-therapeutic for cancer immunotherapies (Clinical Trial IDs: NCT05122546, NCT04988841, NCT05354102, NCT03686202). Several studies have demonstrated that therapeutic success for rCDI is dependent on stable engraftment of the delivered microbes.^3,4 However, the determinants of successful engraftment are not well-defined^5–7 and many introduced bacteria do not stably engraft.^8–15 Furthermore, relative abundance of the engrafted microbes within the microbiome, regardless of how it has been measured and defined, may also influence efficacy.^3,16

The beneficial impacts of therapeutic microbes are thought to be largely driven by modulation of host-impacting metabolites, which are known to be important in health and disease.^17,18 Net metabolite output within the gastrointestinal tract is tied to microbial abundance and substrate availability. Inter-individual functional outputs of the microbiome, such as bile acid production and resistance to C. difficile colonization,^19,20 organic acid production with a corresponding decreased fecal pH,²¹ and cytokine responses,²² have been tied to the abundance of specific beneficial bacterial species, many of which are under investigation as candidate LBPs. In humans, modulation of bile acids and short chain fatty acids (SCFAs) by a therapeutic consortium occurred specifically after multiple doses that led to high levels of engraftment.²³ In another human study, administration of indigestible starches altered levels of branched-chain fatty acids and SCFAs only at the highest doses of starch.²⁴

Given that abundance, and not just presence, of a therapeutic microbe appears to be a key factor impacting the levels of efficacy-mediating metabolites or molecules,^3,4,16 the design of novel LBPs should incorporate strategies to maximize efficacy by optimizing abundance. This concept is analogous to the pharmacokinetic and structure-activity relationship analyses central to medicinal chemistry efforts in which potency, bioavailability, and metabolism of small-molecule drugs is optimized. However, given that LBPs are live organisms and that they primarily act through gastrointestinal sites, the types of pharmacokinetic analyses used for tracking the absorption and metabolism of small molecules or biologics are not directly applicable and are not widely applied during LBP development.²⁵ Recently, advances in whole metagenome sequencing-based strain tracking methods and quantitative systems pharmacology modeling approaches have improved LBP abundance data, generating what has been termed “cellular kinetic-pharmacodynamic” (CK-PD) data,^3,23,26 which link dosing strategy, microbial abundance, and metabolite production. These types of analyses have the potential to inform optimal LBP dosing strategies, provide a framework for evaluating interactions of the LBP within the microbiome and metabolome, and ultimately facilitate beneficial impacts of an LBP on patient health.

We aimed to generate CK-PD data for a novel LBP, Bifidobacterium longum, subsp. infantis (B. infantis), an infant-associated microbe selected for its ability to proliferate in the microbiome by efficiently internalizing and metabolizing human milk oligosaccharides (HMO).^27–29 We recently reported engraftment of B. infantis into both unperturbed and antibiotic-treated gut microbiomes of healthy adult volunteers when co-dosed with a mixture of HMO^30,31; dosing of the paired microbe and nutrient source is termed a synbiotic treatment. In unperturbed adults, fewer became engrafted with B. infantis when lower amounts of HMO were administered, suggesting that there may be a nutrient threshold for B. infantis engraftment.³⁰ Untargeted global metabolomics of fecal samples from unperturbed adults before and after treatment with B. infantis and HMO showed statistically significant changes in 305 metabolites, whereas in adults treated with HMO alone, only 86 metabolites were changed, suggesting that the synbiotic combination of B. infantis and HMO induces more consistent changes across treated individuals than does HMO alone.³⁰

In unperturbed adults treated with 18 g of HMO daily, B. infantis reached an abundance that averaged 5% but ranged from 0.004–25% across individuals.³⁰ In adults treated with antibiotics and each given the same dose of 18 g of HMO, B. infantis abundance ranged from 0.02–81%.³¹ This strongly suggests that individual microbiome communities are an important variable impacting abundance. For antibiotic-treated adults, even though B. infantis abundance varied between individuals, consistent changes in a number of therapeutically relevant metabolites, including lactate, acetate, indoles, and tyrosine metabolites, and robust impacts on the microbiome, characterized by increases in the lactate-utilizing, short-chain fatty acid-producing genus Veillonella, were observed.³¹ Antibiotic treatment has been suggested as a strategy to enhance engraftment and abundance of LBPs,³² but it can also disrupt other taxa in the microbiota and compromise patient health,^26,33,34 making it undesirable in general practice. Therefore, maximizing levels of abundance through other means, such as by dosing as a synbiotic or by altering the dosing strategy, may be critical for therapeutic impact, and this demonstrates the relevance of CK-PD data to facilitate optimization of B. infantis abundance across individuals in our approach.

In this study, we defined the impact of different HMO and B. infantis dosing regimens on bacterial abundance and metabolite levels using gnotobiotic mice colonized with healthy, unperturbed human stool microbiota and a highly sensitive and rapid qPCR assay for B. infantis. We found that the abundance of B. infantis was dependent on the dose of HMO administered, and that B. infantis expansion occurred in the presence of HMO independent of either the starting bacterial inoculum levels or the number of bacterial doses. Additionally, we observed distinct impacts on metabolite levels across groups that were administered different doses of HMO, suggesting that B. infantis abundance influences metabolite output. Together, these data demonstrate that for the synbiotic pairing of HMO and B. infantis, HMO creates a privileged niche for B. infantis replication and enables tunable control of abundance, opening the door to precision engineering of microbes and metabolites for specific therapeutic applications and individualized treatment of patients.

Results

B. infantis steady-state abundance is dependent on the administered dose of HMO

We previously demonstrated that B. infantis colonizes at higher levels in healthy adults when they are also treated with antibiotics,³¹ likely due to antibiotic-mediated clearing of microbiota niches. However, B. infantis is thought to proliferate in the microbiome due to a unique ability to metabolize HMO, thereby creating its own nutritional niche. We hypothesized that administration of different amounts of HMO would impact the size of the niche, and consequently the abundance of B. infantis. We therefore developed a gnotobiotic mouse model, in which germ-free mice were colonized with stool microbiota from a single human and maintained on a polysaccharide-free diet.³⁰ Mice colonized with stool from a healthy adult human were treated with B. infantis and a range of doses of HMO for three days, followed by continued dosing with HMO alone at different doses to support continued metabolism of B. infantis (Figure 1a). Fecal samples were analyzed for total bacterial abundance, as measured by 16S rRNA gene qPCR, and for B. infantis abundance using a subspecies-specific qPCR assay.³⁵ Total bacterial abundance did not significantly differ across the treatments (Figure 1b; no significant difference using a mixed effects model for repeated measures of log-transformed data), although it trended numerically lower at certain timepoints.

Figure 1. HMO dose determines steady-state abundance of B. infantis. (a) Gnotobiotic mice were colonized with healthy human adult stool #R01 and allowed to acclimate for 1 week, prior to treatment with 1 × 10⁸ CFU B. infantis and different amounts of HMO twice daily. Control arms included B. infantis treatment with vehicle (no HMO) and 32 mg HMO per dose without B. infantis. (b,c) 16S gene copy number, and B. infantis abundance in fecal pellets relative to total 16S gene copy number, were determined using qPCR. For samples with B. infantis signal at or below the limit of detection, values were imputed as one-half the lowest observed abundance. Data are from two independent experiments carried out under the same conditions. Data are geometric mean and geometric standard deviation, n = 4–8 per group. Significance between groups was calculated using a mixed effects model for repeated measures of log-transformed data with the Bonferroni-Sidak correction for multiple comparisons; asterisks represent statistically significant differences in the time versus treatment interaction factor for the indicated pair of conditions. (d) Individual traces for each mouse. (e) Steady state B. infantis abundance for groups receiving 1 mg HMO or higher per dose, determined as the geometric mean of values for days 12, 14, and 17 for each mouse in the indicated HMO dose groups. p-values were calculated for log-transformed data using one-way ANOVA.

During the initial phase of the experiment, when mice were actively dosed with B. infantis, abundance was not statistically different across all groups regardless of HMO dose, including control animals that did not receive HMO (Figure 1c,d, one-way ANOVA). After bacterial dosing ended, B. infantis abundance numerically declined to below the limit of detection in groups receiving 2 mg or less of HMO per dose, with kinetics not significantly different from those of animals receiving no HMO, suggesting that this was below the threshold necessary to enable HMO-dependent B. infantis engraftment. At intermediate doses of HMO (4 and 8 mg), B. infantis abundance declined for approximately 5 days, at a rate numerically slower than with no HMO, and plateaued at an abundance of 0.00008% and 0.0004% respectively (average across the last three timepoints). With either 12 or 16 mg of HMO per dose, B. infantis abundance did not decline, and plateaued at approximately 0.06%, a level slightly above that observed during the initial bacterial dosing period (0.015% across all mice); no significant difference was observed between these doses. In contrast, over the course of a week, abundance in mice receiving 32 mg of HMO per dose increased 200-fold, plateauing at approximately 3% of the microbiome for an additional week, and was significantly different from the 16 mg dose. Over the last week of the study, when abundance appeared to have plateaued in all groups, we observed that B. infantis increased as the amount of dosed HMO increased (Figure 1e). These results suggest that tuning of B. infantis abundance within a particular microbiome community may be achievable by tuning the dose of HMO.

B. infantis abundance, not just presence, influences the metabolite profile

We and others have previously demonstrated that B. infantis produces acetate and lactate via fermentation of carbon sources and produces indole-3-lactate from tryptophan.^30,31,36 We hypothesized that higher levels of B. infantis abundance would lead to higher production of these metabolites, which could be quantified in cecal contents (Figure 2). Levels of lactate, acetate, and indole-3-lactate were significantly higher in mice receiving B. infantis in combination with the highest dose of HMO (32 mg) compared to mice that did not receive HMO (Figure 2a). Significant increases in acetate and indole-3-lactate were also observed at the second highest HMO dose (16 mg), and lactate was numerically increased, although this change was not statistically significant. B. infantis-produced metabolites were numerically, but not significantly, increased in mice receiving intermediate doses of HMO (12 mg and 8 mg). This observation is particularly surprising for the group that received 12 mg, as the abundance of B. infantis was similar to that observed for the 16 mg HMO group. One possible explanation is that the number of animals in this group (n = 4) was not sufficient to detect statistically significant differences.

Figure 2. Metabolite analysis of cecal contents from mice treated with B. infantis and different doses of HMO. The indicated metabolites were quantified in cecal contents. (a) Heatmaps displaying the geometric mean of the metabolite quantification for each HMO dose group on day 17, log-transformed and normalized to that of the group receiving 0 mg HMO. Asterisks represent p-values compared to the 0 mg HMO group calculated using one-way ANOVA with Sidak’s posttest. (b) Geometric mean and geometric standard deviation of metabolites in cecal contents, and p-values between the indicated dose groups calculated by one-way ANOVA on all groups with Sidak’s multiple comparison test.

Significant changes in lactate and acetate, but not indole-3-lactate, were also observed upon treatment with 32 mg HMO alone compared to 0 mg HMO (Figure 2a). When we compared groups receiving 32 mg HMO with and without B. infantis (Figure 2b), we found significant differences for lactate and indole-3-lactate, supporting the idea that B. infantis robustly and specifically produces these metabolites, but not for acetate, suggesting that other members of this particular healthy microbiome utilize HMO to produce acetate.

Lactate and acetate production enables crossfeeding of other microbiome members that consequently produce propionate and butyrate.^30,31 Similarly, we observed increased propionate in the B. infantis and HMO treatment group relative to no HMO treatment (Figure 2a). Treatment with HMO alone also significantly increased propionate compared to no HMO (Figure 2a), and the level of propionate was significantly higher for treatment with HMO alone than treatment with HMO and B. infantis (Figure 2b). Treatment with HMO and B. infantis did not impact levels of butyrate, but treatment with HMO alone did significantly increase butyrate levels (Figure 2a), although interestingly there was no statistically significant difference in butyrate between groups treated with HMO with or without B. infantis (Figure 2b).

We also quantified the levels of indole-3-propionate, a metabolite that can be produced from indole-3-lactate by certain microbes and which may have host-beneficial impacts.^37,38 Indole-3-propionate was significantly increased by treatment with B. infantis and either 16 or 32 mg HMO, and was also increased during treatment with 32 mg HMO without B. infantis (Figure 2a), although to a statistically significantly lower extent than with B. infantis (Figure 2b).

B. infantis expansion rate is similar regardless of delivered inoculum or number of bacterial doses

Repeated dosing or varying the inoculum of LBPs has been suggested to increase the frequency of individuals that become engrafted, the abundance of the delivered microbes, and therapeutic efficacy for rCDI,^3,16,23,32 though one study found no impact of repeated dosing.³⁹ We therefore next employed our gnotobiotic model to evaluate the kinetics of B. infantis expansion while varying the initial inoculum or the total number of doses of bacteria administered (Figure 3a). Twice-daily doses of B. infantis (Figure 3i) ranging from 1 × 10³ to 1 × 10⁸ CFU were administered for either 1 day (2 total doses) or 3 days (6 total doses). In one study, HMO or vehicle dosing continued for a total of 14 days; in a second study, HMO dosing ceased after 10 days but sampling continued until 14 days. Total bacterial abundance, as measured by 16S rRNA gene qPCR, was not statistically significantly different by treatment (Figure 3b; no significant difference using a mixed effects model for repeated measures of log-transformed data).

Figure 3. B. infantis expands at a similar rate regardless of starting inoculum level. (a) Gnotobiotic mice colonized with healthy human adult stool #R02 were treated with the indicated CFU of B. infantis along with HMO (16 mg) or PBS twice daily. (b,c,d) 16S gene copy number, and B. infantis abundance in DNA extracted from fecal pellets relative to total 16S gene copy number, were determined using qPCR. Data are geometric mean and geometric standard deviation for mice receiving B. infantis with HMO (b,c) or with PBS (b,d); n = 8 for groups dosed over days 0–2 from two independent experiments (solid line HMO; dotted line PBS); n = 4 for groups dosed only on day 0 in a single experiment (hatched line HMO). For samples with B. infantis signal at or below the limit of detection, values were imputed as one-half the lowest observed abundance. Significance between groups was calculated using a mixed effects model for repeated measures of log-transformed data with the Bonferroni-Sidak correction for multiple comparisons; asterisks represent statistically significant differences in the time versus treatment interaction factor for the indicated pair of conditions. (e,f) Individual traces for each mouse dosed over days 0–2 from the data plotted in (c); 4 mice in each dose cohort continued receiving HMO to day 14 (e), and the other 4 received vehicle from days 11–14 (f). (g) Linear regression of log-transformed abundance data for all mice dosed over days 0–2 with HMO dosing through day 10. Timepoints for which B. infantis was below limit of detection in the 1 × 10³ or 5 × 10⁴ dose groups were omitted to improve regression. (h) The number of doublings of B. infantis per day for each group of mice was captured from the slope of the regression lines presented in (e). (i) Quantification of B. infantis inoculum for each dose in both studies.

As expected, the initial abundance of B. infantis during the bacterial dosing period varied according to the dose administered (Figure 3c, 3d). For mice receiving vehicle, B. infantis abundance numerically declined rapidly after the end of bacterial dosing (Figure 3d). However, with sustained administration of HMO, B. infantis abundance was maintained or increased in all mice, regardless of the bacterial dose administered or the initial abundance (Figures 3e, 3f). The two highest dose groups appeared to converge on a similar maximum abundance before the end of the study of approximately 1% (Figure 3e). In one experiment, B. infantis levels were monitored over time after cessation of HMO dosing and rapidly declined, as expected (Figure 3f).

We next performed regression analysis to determine the expansion rate of the B. infantis population (Figure 3g). Groups receiving 1 × 10,³ 5 × 10,⁴ or 2 × 10⁶ CFU B. infantis had similar doubling rates (Figure 3e,f,h; no statistically significant differences by one-way ANOVA with Tukey’s posttest for multiple comparisons), suggesting that rate of growth is independent of starting abundance. At the highest dose of 1 × 10⁸ CFU per dose, it is possible that the measured expansion rate is masked by a high passage of B. infantis through the gut during dosing, and that expansion slows when B. infantis approaches carrying capacity in the gut, perhaps due to a limited supply of HMO or other forms of inter-species competition. Indeed, the groups receiving 2 × 10⁶ CFU per dose also appear to have a slower rate of expansion after day 10.

We also examined the impact of repeated dosing with B. infantis. For the three highest dose groups, there were no statistically significant differences in the expansion curve between same-dose paired groups receiving 2 total doses or 6 total doses (Figure 3c; mixed effects model). However, in the groups receiving 1 × 10³ CFU per dose, the group that received 6 total doses displayed a significant, 2-day reduction in the time to achieve comparable abundance with the group that received 2 total doses, although there was no apparent difference in the rate of expansion.

Abundance of B. infantis varies in different healthy stools treated with the same dose of HMO

As shown in Figure 1, in mice colonized with stool #R01 and treated with B. infantis and 16 mg HMO, B. infantis abundance plateaued at 0.065% (average of the last three timepoints). As shown in Figure 3, mice colonized with stool #R02, sourced from a different healthy adult, and treated with the same B. infantis dose and HMO dose of 16 mg, B. infantis abundance exceeded 1% (Figure 3e). This comparison aligns with our previous observations in humans, where we observed steady-state abundances of B. infantis ranging from 0.004–81% in different engrafted individuals.^30,31

Discussion

Defining the effects of dosing permutations on engraftment of LBPs has been a suggested alternative for traditional pharmacokinetic (PK) studies,²³ and has the potential to inform dosing strategies²⁶ that may improve efficacy and decrease cost. Here we report two kinds of CK-PD studies, in which we tested the effect of varying the dose of each component of a synbiotic product, comprising B. infantis and its preferred carbon source HMO, on abundance and metabolite levels in gnotobiotic mice harboring a healthy human microbiota. We found that altering the dose of HMO led to corresponding changes in the abundance of B. infantis at steady state, and that the levels of lactate, acetate, propionate, indole-3-lactate, and indole-3-propionate were only altered at the highest doses of HMO tested. This synbiotic pairing of LBP and prebiotic substrate is expected to enable highly reproducible impacts on the microbiome and gut metabolites. Indeed, we previously observed consistent microbiome and metabolite changes (e.g. expansion of Veillonella due to crossfeeding by B. infantis-produced lactate) across multiple antibiotic-treated adult humans harboring high levels of B. infantis during treatment with HMO.³¹ Here, we also found that altering the B. infantis inoculum did not change B. infantis expansion rate in the microbiome (1–1.2 doublings per day) but did affect the time for B. infantis to reach steady-state abundance. Our data are consistent with the idea that HMO creates a privileged niche for B. infantis and that B. infantis expansion in the microbiome is limited primarily by access to carbon. Indeed, B. infantis is known to outcompete other microbes for use of HMO by selfishly importing HMOs and retaining degradation products intracellularly.^27–29

We found a 15-fold difference in B. infantis abundance across the two different donor stools used in our studies, when the same B. infantis and HMO dosing conditions were compared. This result suggests that the composition of a particular microbiota community impacts the size of the niche available to B. infantis, as we have previously observed in humans.^30,31 Although germ-free mice colonized with human stool microbiota are widely used as a model, it is important to note that retention of inoculated taxa varies widely across different inocula, and the distribution of retained taxa is generally distinct from that observed in the inoculum^40–48 due to physiological and sample handling factors; for example, differences in mouse and human gastrointestinal anatomy and diet, and storage time/temperature/oxygen exposure of fresh stool prior to cryopreservation. Additionally, microbial composition of the stool inoculum is typically determined via DNA sequencing, which detects DNA from both viable and non-viable cells and complicates measurements of retention of inoculated taxa. Despite these limitations of using germ-free mice to model human microbiota communities, we have previously observed microbiome-mediated recapitulation of human colonization phenotypes using our mouse model,³⁰ which supports the idea that the impacts of synbiotic dosing strategies observed here are likely to also apply to other communities in humans and mice.

Our synbiotic strategy enables LBP engraftment by using a prebiotic substrate to create a privileged nutritional niche for B. infantis in the absence of antibiotic pretreatment.³⁰ Successful engraftment remains a key challenge for LBPs, as it only occurs in a subset of human subjects,^{3,6,11,15,16,23,32,49} likely due to competitive exclusion by related microbes or limited resource availability.⁶ Additionally, LBPs comprising a consortium of microbes may require stepwise cooperation, where the success of initial colonizers is required to enable engraftment of all strains, and this has been demonstrated in mouse models.⁵⁰ To achieve engraftment in the absence of antibiotic conditioning, altering the bacterial dose regimen, by either increasing the dose or dosing repeatedly, has been suggested to increase the frequency of individuals that become engrafted and increase therapeutic efficacy for rCDI,^3,16,23,32 though one study found no impact of repeated dosing.³⁹ Here, we found that, in the presence of HMO, doses of B. infantis as low as 2 × 10³ total CFU were sufficient to enable expansion and engraftment. Our synbiotic strategy not only enables engraftment of B. infantis, but also allows for precise control of abundance by co-administration with different doses of HMO.

Other microbiome modulation strategies that rely solely on prebiotics may promote the expansion of particular taxa in humans and in mouse models, sometimes in a dose-dependent manner.^{24,30,51–57} Similarly, we observed increased production of acetate, propionate, and butyrate in mice treated with HMO alone, without B. infantis, suggesting that other members of the particular healthy microbiome used in this study may utilize HMO. Other members of genus Bifidobacterium, as well as members of Lactobacillus and Bacteroides, are known to secrete HMO-degrading enzymes, making monosaccharides available for consumption.^58–61 However, our previously published human data indicates that dosing with HMO alone did not lead to consistent, global changes in metabolite profiles across individuals, while dosing with B. infantis and HMO together did.³⁰ The number of significantly changed metabolites in adults receiving B. infantis and HMO together (305) compared with the number observed in adults receiving HMO only (86) suggest that the impact of HMO alone is not consistently strong across individual microbiomes. Indeed, prebiotics as sole agents are reliant on the preexistence of responding microbes, the presence of which likely varies to a high degree across individuals. Prebiotics useable by multiple species of bacteria are likely to have less impact on the abundance of any particular taxa and therefore minimal therapeutic relevance.^24,62–64 In contrast, highly targeted prebiotics support the expansion of only the microbes able to utilize that substrate.^63,65,66 For our synbiotic of HMO and B. infantis, there was a strong effect of HMO dose on B. infantis abundance and metabolites, suggesting targeted expansion of this species. This strategy enables exquisite control of B. infantis engraftment, because not only did dosing with HMO maintain B. infantis in the microbiome, but withdrawal of HMO led to B. infantis clearance.

In addition to supporting an increase in abundance of B. infantis, HMO supports B. infantis-produced metabolic products that cross-feed other bacteria in the microbiome, resulting in additional beneficial metabolites. Given that metabolites are key mediators through which LBPs are expected to impact the host, it is important to understand the relationship between LBP abundance and the desired metabolomic and therapeutic outcomes. In mice, high levels of particular taxa induced via administration of high doses of prebiotics have been shown to reduce measures of diversity and abundance of other potentially beneficial taxa,^24,56,57 raising the question of whether maximizing LBP abundance is the best therapeutic strategy or whether tuning LBP abundance to produce ideal levels of metabolites would be a better approach.

For B. infantis, the CK-PD data collected in this study support the idea that HMO dosing enables tunable control of abundance and metabolite levels within the microbiome. This feature could facilitate precision engineering of metabolites and microbes to create a microbiome profile for specific therapeutic applications. HMO-enabled tuning of B. infantis abundance may also allow for personalized dosing strategies to achieve optimal impact for individual patients harboring distinct microbiome communities or needing particular metabolite profiles. More broadly, we anticipate that CK-PD analyses have the potential to impact design of LBPs and improve the clinical application of microbiome-based therapeutics.

Methods

Bacterial culture

B. infantis PBI001 was propagated for mouse studies in Lactobacilli MRS Broth (BD Difco # 288130) or in media containing 20 g/L pea peptone, 5 g/L yeast extract, 2 g/L potassium phosphate dibasic, 5 g/L sodium acetate, 1 g/L sodium chloride, 20 g/L dextrose, and 0.1% polysorbate-80. Cultures were centrifuged, and cell pellets were resuspended in phosphate-buffered saline (PBS) with glycerol at a final concentration of 15%, aliquoted, and frozen at −80°C for long term storage. Viable colony-forming units (CFU) were determined by diluting and plating onto Brucella Blood Agar (Anaerobe Systems) and growing anaerobically at 37°C. In studies delivering a range of B. infantis doses, cryovials were thawed, diluted in 15% glycerol, and refrozen prior to thawing and use in animal studies.

Preparation of fecal slurries

Healthy adult stool was sourced from two self-reported healthy adults with a body mass index < 25 and no history of antibiotic use in the three months prior to collection. Stools were frozen within 4 hours of production. The stools were collected and processed using methods designed to remove food particulates while maximizing recovery of viable bacteria. To prepare stool slurries, frozen stools were anaerobically thawed, weighed, and combined with sterile PBS. Stool #R01 was simultaneously homogenized and large particulates were removed using a BagMixer 400 (Interscience) with a BagPage XR (280 µm porosity) filter bag. Stool #R02 was homogenized in a blender and centrifuged at 200 × g for two minutes to remove large particulates. The two methods for homogenizing and removing particulates are both recognized as appropriate methods for stool processing.^67–71 The resulting filtrate or supernatant was combined with glycerol in PBS for a final concentration of 15% glycerol (w/w). The stool slurries (8–10% stool w/w) were aliquoted and frozen at −80°C for experimental use. Viable CFU in aliquots of stool were determined by diluting and plating onto Brucella Blood Agar (Anaerobe Systems AS-141) and growing aerobically and anaerobically at 37°C.

Preparation of HMO concentrate

The HMO concentrate used in this study was manufactured by Prolacta Bioscience (Duarte, CA). Prolacta Bioscience makes human milk nutritional products from donated breast milk, intended for use by extremely premature infants in the neonatal intensive care unit and by term infants born with congenital malformations requiring immediate surgery. The milk is obtained from healthy donors who produce excess milk beyond the needs of their own infants. The donors are carefully screened for a variety of infectious diseases and lifestyle elements, such as drug use including nicotine, similar to what occurs in a blood bank. All potential donors provide attestation from their baby’s pediatrician that the baby will not be adversely affected if the mother donates her excess breast milk. Individual milk donations are also screened for a variety of pathogens of concern. All donors sign informed consent as part of their contract with Prolacta prior to donation and are aware that a portion of their milk may be used for the development of new products or other research purposes. Multiple lots of pooled human milk, from up to 200 donors per lot, were used to prepare the HMO product. Human milk permeate, a co-product of ultrafiltration used during the manufacture of commercial human milk-based human milk fortifier and ready-to-feed products, consists primarily of water, lactose, HMO and minerals. HMO concentrate was generated by subsequently removing lactose from the permeate and carrying out further ultrafiltration and pasteurization steps. The test product was supplied as a frozen liquid, stored at ≤-20°C, and thawed prior to administration.

Murine studies

Germ-free female C57BL/6 mice (age 5–6 weeks, Taconic) were divided into individual cages and inoculated with 0.1 mL of a 5% fecal slurry from a healthy human and maintained on a polysaccharide-deficient diet (Bio-serv, F5805–1, AIN-93 G, 68% glucose, irradiated 40–55 kGy).^72,73 After a 7-day stabilization period, animals were randomized based on the original group cage and orally gavaged with B. infantis PBI001 resuspended in 0.2 mL of either HMO or sterile phosphate buffered saline (PBS) every 10–14 hours over a period of one or three days, followed by administration of HMO or PBS twice daily until study end. Dosage of B. infantis and HMO for each study is indicated in the figures. Groups receiving different doses of B. infantis or HMO were dosed in ascending dosages. Fecal samples were collected from the bottom of the cage for each individual animal throughout the study and flash frozen for DNA extraction. At the end of the study, cecal contents were collected and snap-frozen for metabolite analysis. Individuals handling animals and samples were blinded to the dose of HMO or B. infantis being administered.

Animals were housed in a positive-pressure clean room with HEPA filtration (bioBubble) in solid-bottom micro-isolator cages (Innovive Innocage MVX6 with M-feed food hoppers), given sterile water ad libitum (Innovive Aquavive M-WB-300A), and provided with sterile 1/8” corn cob bedding (Teklad 7902). Environmental conditions were maintained at temperature 20–26°C, humidity 30–70%, with a 12/12 hour light/dark cycle. Health checks were performed at least twice daily and included observation of behavior and appearance. The number of animals per group was determined to be sufficient to detect changes in B. infantis abundance as small as 4.1-fold (#R01) or 2.7-fold (#R02) with a significance level of 0.05 with 80% power based on the test for the ratio of two means [log normal data])⁷⁴ using B. infantis abundance data from pilot studies in germ-free mice colonized with stools #R01 (coefficient of variation = 0.65) or #R02 (coefficient of variation = 0.27). Animal work was carried out in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All protocols were approved by the TransPharm PreClinical Solutions Institutional Animal Care and Use Committee.

DNA extraction from fecal samples and B. infantis-specific and 16S rRNA gene qPCR

Genomic DNA extraction from fecal pellets was performed using the ZymoBiomics 96 MagBead DNA Kit (Zymo Research, Inc), following the manufacturer’s instructions. The concentration of the extracted DNA was measured using a Nanodrop 8000 (Thermo Scientific). To quantify B. infantis or total bacterial copy number in DNA samples, qPCR was performed using 2 µl undiluted gDNA with primers specific to a B. infantis sialidase gene (5’-ATACAGCAGAACCTTGGCCT-3’; 5’-GCGATCACATGGACGAGAAC-3’; 5’-FAM/TTTCACGGA/ZEN/TCAC CGG ACCATACG-3IABkFQ), as previously described,³⁵ or using 3 ng gDNA as template with primers targeted to the 16S rRNA gene (5’-CGGTGAATACGTTCCCGG-3’; 5’-TACGGCTACCTTGTTACGACT T3’),⁷⁵ respectively.

Metabolite analysis of mouse cecal contents

Metabolite analysis was performed by Precion, Inc. (Morrisville, NC). Snap-frozen mouse cecal contents were thawed and 20–50 mg of material was transferred into a 2 mL cryotube containing three stainless steel 1/8” cone balls, and the exact weight of the sample was recorded. A solution of deuterium-labeled internal standards in water (50 μL) and 1.5 mL of methanol was added to the cryotube and the sample was homogenized by vortex mixing for 2 minutes. The resulting suspensions were centrifuged at 2,000 × g at 20°C for 10 minutes. Aliquots of the supernatants were transferred to a 96-well plate and derivatized using modified versions of published protocols (50 μL)⁷⁶ or (20 μL),⁷⁷ followed by liquid-chromatography mass spectrometry using an Exion UHPLC (Sciex) coupled to a 5500+ Triple Quadrupole Mass Spectrometer (Sciex) respectively in negative mode for lactate, acetate, butyrate, and propionate or in polarity switching mode for indole-3-lactate and indole-3-propionate. The peak areas of the respective parent to product ion transitions were measured against the peak area of the parent to product ion transitions of the corresponding labeled internal standards for the quantitated metabolites. Quantitation was performed with Sciex OS-MQ software (Sciex) based on fortified calibration standards prepared immediately prior to each run. Data were normalized to wet weight based on the exact wet weight of each sample.

Statistical analysis

Regressions and statistical analyses were performed using GraphPad Prism version 9.3.1 or higher for Windows, GraphPad Software, San Diego, California USA, www.graphpad,com. Significance values are noted as follows unless otherwise defined in figure legends: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Author contributions

ALR, JEB, GJM designed experiments. CMC developed methods. ALR, RS, OT performed experiments. ALR, SLB, and MLL analyzed data. ALR wrote the paper. AKS provided materials.

Acknowledgments

We would like to thank Shuning Rook Zheng, Richard Lavin, Aislinn Rowan-Nash, Chloe Autran, David Rechtman, Biranchi Patra, Shao-Hung Fred Tsen, Huiyu Jannela Xia, Kim Thu Tran, Tin Huynh, Thanhvan Vanessa Nguyen, Anita Wong, Jorge Ramirez, Gildardo Inzunza, Yesenia Martinez, and Mike Garelick for their support of and contributions to this study.

Data availability statement

The authors confirm that the data supporting the findings of this study are available within the article.

Disclosure statement

Prolacta Bioscience employees are also shareholders in the company. U.S. Patent No. 8,927,027, U.S. Patent Application No. 20200054035, International Application Pub. Nos. WO 2021/061991 and WO 2022/036225, and their corresponding family member patents and applications relate to aspects of this work.

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

This work was funded by Prolacta Bioscience.