Fecal virome transfer improves proliferation of commensal gut Akkermansia muciniphila and unexpectedly enhances the fertility rate in laboratory mice

ABSTRACT

Probiotics are intended to improve gastrointestinal health when consumed. However, the probiotics marketed today only colonize the densely populated gut to a limited extent. Bacteriophages comprise the majority of viruses in the human gut virome and there are strong indications that they play important roles in shaping the gut microbiome. Here, we investigate the use of fecal virome transplantation (FVT, sterile filtrated feces) as a mean to alter the gut microbiome composition to lead the way for persistent colonization of two types of probiotics: Lacticaseibacillus rhamnosus GG (LGG) representing a well-established probiotic and Akkermansia muciniphila (AKM) representing a putative next-generation probiotic. Male and female C57BL/6NTac mice were cohoused in pairs from 4 weeks of age and received the following treatment by oral gavage at week 5 and 6: AKM+FVT, LGG+FVT, probiotic sham (Pro-sham)+FVT, LGG+Saline, AKM+Saline, and control (Pro-sham+Saline). The FVT donor material originated from mice with high relative abundance of A. muciniphila. All animals were terminated at age 9 weeks. The FVT treatment did not increase the relative abundance of the administered LGG or AKM in the recipient mice. Instead FVT significantly (p < 0.05) increased the abundance of naturally occurring A. muciniphila compared to the control. This highlights the potential of propagating the existing commensal “probiotics” that have already permanently colonized the gut. Being co-housed male and female, a fraction of the female mice became pregnant. Unexpectedly, the FVT treated mice were found to have a significantly (p < 0.05) higher fertility rate independent of probiotic administration. These preliminary observations urge for follow-up studies investigating interactions between the gut microbiome and fertility.

KEYWORDS:

• Fecal virome transplantation
• Gut microbiome
• Fertility
• Probiotic engraftment
• Lacticaseibacillus rhamnosus
• Akkermansia muciniphila

Introduction

During the last decade, it has become commonly accepted that gut microbiome imbalances (dysbiosis) play important roles in the etiology of a number of diseases^1–3. Probiotics are suggested as a tool to restore gut microbiome balance^4,5 and are defined as live microorganisms that, when ingested in adequate amounts, confer a health benefit to the host⁶. However, traditional probiotics, mainly lactobacilli and bifidobacteria, have generally no or only modest influence on the gut microbiome composition⁷. So-called next-generation probiotics, like Akkermansia muciniphila, have recently been suggested to alleviate gut microbiome associated malfunctions^8–10. Persistent beneficial effects are challenged by the difficulties of the administered bacteria to become a permanent and adequately abundant member of the densely populated gut microbiome^11,12.

Mounting evidence suggests that the gut viral community plays a pivotal role in shaping the composition of the gut microbiome^13,14. The gut virome is predominated by prokaryotic viruses¹⁵, including bacteriophages (phages), which are viruses that attack bacteria in a host-specific manner¹⁶. A transfer of sterile filtered feces (containing phages, but no intact bacterial cells) from a healthy donor have shown to successfully treat recurrent Clostridioides difficile infections in human recipients¹⁷. Other studies using sterile filtered feces have reported to alleviate symptoms of type-2-diabetes mellitus and obesity in mice¹⁸ and to prevent the development of necrotizing enterocolitis in preterm piglets¹⁹. These changes in phenotype may be driven by a phage-mediated modulation of the gut microbiome^18–22. The inherent nature of the predator–prey relationship between bacteria and phages²³ likely explains how phages are able to play an important role in modulating and maintaining the gut microbiome²⁴. Temperate phages may increase the fitness of the infected bacteria by transferring auxiliary metabolic genes^25,26 and thereby affecting the competition landscape in the gut. When transferring the fecal viral components, a significant change in the bacterial diversity and composition was observed, with the bacterial GM-component of the recipients becoming more like the gut microbiome of the donors, in all the above-mentioned cases. We will refer to this approach as fecal virome transplantation (FVT).

Many clinical trials are investigating the potential of using fecal transfer (fecal microbiota transplantation, FMT) to treat various diseases^20,27–30. However, FVT has the advantage over FMT that no bacteria are transferred and thereby decreases the risk of transferring bacteria-associated infectious diseases²¹. We hypothesized that initial phage-mediated modulation of the existing bacterial landscape in the gut microbiome using FVT would improve the enteric engraftment and abundance of the subsequently administered probiotic bacteria (A. muciniphila or Lactocaseibacillus rhamnosus GG).

Results

Here, we investigated the potential of improving persistent gut colonization of the next-generation probiotic bacterium A. muciniphila in lean recipients by initial modulation of the gut microbiome landscape with fecal virome transplantation (FVT) donor material originating from an A. muciniphila rich gut microbiome. The bacterium L. rhamnosus GG was included as a commercially available representative. We did the following considerations to maximize our chances for evaluating a successful enteric engraftment of A. muciniphila: (i) C57BL/6NTac (B6N) mice were selected since previous experience has shown a low relative abundance of A. muciniphila³¹. A. mucinphila YL-44 strain was used as probiotic treatment in this study due to its enteric origin from the genetically closely related C57BL/6J (B6J) wildtype mouse, and (iii) the FVT virome represented a gut virome from mice donors with a high relatively abundance (>6%) ofA. muciniphila ³¹. Fecal samples from three different time points were included to investigate the level of probiotic engraftment and gut microbiome changes over time: baseline, 6 days after 2^nd intervention, and at termination. See Figure 1 for the experimental design of the animal model.

Figure 1. Experimental setup of the animal model. In total 24 male and 24 female C57BL/6NTac mice (4 weeks old) were divided into six groups: LGG+FVT, AKM+FVT, Pro-sham+FVT, LGG+Saline, and AKM+Saline, control (Pro-sham+Saline). The Saline consisted of SM buffer and Pro-Sham of Intralipid. The mice were administered 1 M sodium bicarbonate prior oral gavage of FVT/Saline to protect the viral community against the acidic environment in the stomach. The day after, the mice were inoculated with probiotic solutions of LGG/AKM/Pro-sham suspended in Intralipid which constituted the 1^st inoculation. The same procedure was repeated as the 2^nd inoculation one week after. The mice were fed ad libitum low-fat diet (LFD) for the entire study (6 weeks) until termination at age 9 weeks. Fecal samples from baseline, after intervention, and termination were analyzed in this study. Abbreviations: Lacticaseibacillus rhamnosus GG = LGG, Akkermansia muciniphila = AKM, fecal virome transplantation = FVT, Pro-sham = probiotic sham.

FVT enhanced the abundance of natural occurring A. muciniphila strains

Our hypothesis was that initial modulation of the gut microbiome landscape driven by the FVT would lead to an increase in the abundance of A. muciniphila and/or L. rhamnosus strains after probiotic administration. However, we did not observe any significant effect of the FVT on the exogenous (probiotic) A. muciniphila or L. rhamnosus strain abundance after intervention nor at termination (Figure 2AB). Instead, we assessed whether FVT had the potential to increase the enteric endogenous (native) A. muciniphila and L. rhamnosus strains without exogenous (probiotic) administration. When comparing mice that were provided FVT but no A. muciniphila as probiotic, with mice that neither were provided FVT nor A. muciniphila (Figures 2A & 2C), we observed at termination that FVT had increased (p < 0.05) the abundance of, what would be expected to be, native A. muciniphila strains, since A. muciniphila was not administered as probiotic. No effect by FVT was observed with regard to enhancing native L. rhamnosus abundance (Figures 2B & 2D). The abundance of A. muciniphila strains at baseline was significantly lower (p < 0.05) compared to termination in AKM+FVT and Pro-sham+FVT mice, while tending lower (p < 0.1) in the LGG+FVT mice as well (Figure S1).

Figure 2. qPCR with L. rhamnosus and A. muciniphila specific primers were used to assess the abundance of gene copies per gram feces over the time span of baseline, after intervention, and termination. A) the development of the A. muciniphila abundance, B) L. rhamnosus abundance. C) the abundance development of native A. muciniphila strains for mice receiving FVT and no probiotic A. muciniphila (Pro-sham+FVT and LGG+FVT) compared to not receiving FVT or A. muciniphila (Lgg+saline and control). D) the abundance development of native L. rhamnosus strains for mice receiving FVT and no L. rhamnosus (Pro-sham+FVT and AKM+FVT) compared to not receiving FVT or L. rhamnosus (Akm+saline and control). Abbreviations: Lacticaseibacillus rhamnosus GG = LGG, Akkermansia muciniphila = AKM, fecal virome transplantation = FVT, Pro-sham = probiotic sham.

This indicated that the FVT had improved the growth conditions of naturally occurring A. muciniphila strains. The AKM+Saline mice had similar A. muciniphila abundance as the FVT groups and thereby indicated that exogenous (probiotic) administration of A. muciniphila also led to enteric colonization. Additional experiments were performed to rule out that the FVT initially contained any A. muciniphila strains (Figure S2). The undiluted sterile filtered donor feces (used for FVT) were incubated in 96 h on nonselective agar plates from which eight colonies appeared. Cell morphology of these colonies was imaged with phase-contrast microscopy, and subsequently screened with A. muciniphila specific primers in both a PCR and qPCR assay. Neither colony morphology, cell morphology, PCR nor qPCR indicated any traces of A. muciniphila in the applied donor FVT (Figure S2).

FVT leads to a reduction in bacterial diversity of the gut microbiome component

FVT significantly (p < 0.05) decreased the bacterial Shannon diversity index in the LGG+FVT and Pro-sham+FVT mice at termination (9 weeks of age) when compared with the AKM+Saline, LGG+Saline, and the control mice (Figure 3A). Whereas the Shannon diversity index of the AKM+FVT mice remained unchanged compared to the control mice, hence suggesting that A. muciniphila may have counteracted the decrease in the Shannon diversity that was associated with the FVT treatment (Figure 3A). The initial bacterial diversity at baseline was similar between all groups (Figure 2A). The most abundant genus in all groups at all time-points was Lactobacillus (Figure S3). The FVT-associated differences in the bacterial Shannon diversity index were not reflected in the bacterial composition analysis (Bray-Curtis dissimilarity), since no significant differences were observed between treatments at all three timepoints (Figure 3B). Probably due to the state of pregnancy, the sex of the animals (male vs female) showed significant (p < 0.001) differences in their bacterial composition at termination (Figure S4). Differential abundance analysis showed that Candidatus Arthromitus (segmented filamentous bacteria) and A. muciniphila amplicon sequence variants (ASVs) were significantly (p < 0.05) increased in relative abundance at termination in mice receiving FVT compared to the other groups (Figure 3C). Thus, supporting the FVT-mediated enhancement of A. muciniphila abundance measured by the qPCR analysis. The administration of the probiotic A. muciniphila significantly increased (p < 0.05) the relative abundance of Ruminococcus gnavus (Figure S5).

Figure 3. Gut bacteriome analysis of A) the bacterial diversity (Shannon diversity index) and B) PCoA plots of the bacterial composition (Bray-Curtis dissimilarity) at baseline, after intervention, and termination. C) relative abundance of ASVs with significant differential abundance (p < 0.05) between FVT treated mice and mice not receiving FVT, summarized at genus level. Abbreviations: Lacticaseibacillus rhamnosus GG = LGG, Akkermansia muciniphila = AKM, fecal virome transplantation = FVT, Pro-sham = probiotic sham, ASV = amplicon sequence variant.

Probiotic and FVT intervention may have changed the viral gut microbiome profile

The viral Shannon diversity index at termination (9 weeks of age) was affected by FVT (p = 0.032) as well as the administration of the probiotics A. muciniphila (tendency, p = 0.07) and L. rhamnosus (p = 0.006) when compared to the control mice (Figure 4A and Figure S6). The effects of probiotics A. muciniphila (p = 0.025) and L. rhamnosus (p = 0.014) were also reflected on the viral composition at termination (Figure S7). The sex of the animals appeared to influence (p < 0.05) the viral community composition across all time points (Figure S4). The donor FVT virome consisted of more than 90% Microviridae viruses and was markedly different in both viral diversity and composition (Figures 4A & 4B) compared to the recipient gut virome (Figure S8). Differential abundance analysis was performed using both the predicted bacterial hosts and raw viral taxonomy at termination (9 weeks of age). These analyses showed a significant increase in the relative abundance of predicted hosts belonging to the taxa Lachnospiraceae, Parabacteroides, and Bacteroides in FVT treated mice (Figure S9), and an increase in the relative abundance of Petitvirales (likely Microviridae) when comparing with mice not receiving FVT (Figure 4C). It could be speculated that the elevated level of Petitvirales in the FVT treated mice was driven by the highly Microviridae abundant (>90% of relative abundance) donor virome (Figure S8).

Figure 4. Gut virome analysis of A) the viral diversity (Shannon diversity index) and B) PCoA plots of the viral composition (Bray-Curtis dissimilarity) at baseline, after intervention, and termination. C) relative abundance of viral contigs with significant differential abundance (p < 0.05) between FVT treated mice and mice not receiving FVT, summarized at family level. Abbreviations: Lacticaseibacillus rhamnosus GG = LGG, Akkermansia muciniphila = AKM, fecal virome transplantation = FVT, Pro-sham = probiotic sham.

The viral Shannon diversity at baseline of the AKM+FVT and Pro-sham+FVT mice were significantly lower (p < 0.05) compared to control (Figure 4A), while the diversity of the remaining treatment groups was similar to the control mice. This initial variance also tended to be reflected (p < 0.084) on the viral composition at baseline, but these differences were diminished at termination (Figure 4B). The extent to which the above-mentioned inter-group differences were associated with the baseline variance of viral diversity and composition was not clear.

A. muciniphila affects expression of a gene involved in mucin production and limits an inflammatory response associated to FVT

The ileum tissue was investigated for changes in the expression levels of genes associated with inflammatory responses. Mice with the highest abundance at termination of the mucin-degrading A. muciniphila (Figure 2A) had either significantly (p < 0.05, AKM+Saline and AKM+FVT) or tended toward (p < 0.1, LGG+FVT) lowered gene expression of the Muc1 gene compared to control mice (Figure 5A). The Muc1 gene is involved in transmembrane mucin production³². Whether this affected the mucin production in the mice was not clear. However, neither histological staining of mucus in the ileum or the expression of the Muc2 gene (involved in mucin production by goblet cells³²) indicated differences in the mucin production (Figure S10). We also measured lipopolysaccharide (LPS) levels in serum as a measure of bacterial antigens that cross the intestinal barrier and thereby also the integrity of the epithelial mucus layer. No clear changes were observed of the LPS levels in the treatment groups when compared to control (Figure S11A), which indicated that the intestinal barrier was not compromised due to FVT or probiotic administration. However, the mice provided A. muciniphila as probiotic appeared with a tendency of less LPS in the serum (Figure S11C) and the level of LPS content was affected (p = 0.03) by the sex of the animal (Figure S11B). The expression of nine genes that are involved in inflammation and as a response to infection (Clc2, Ccr10, Ctla4, Cxcl1, Il1b, Il4, Il6, Retnlb, Timp1) was significantly (p < 0.05) elevated in Pro-sham+FVT compared to control mice (Figure 5A,J). Additionally, two genes (Ffar2 and Ffar3) for short-chain-fatty-acid receptors involved in both energy homeostasis and intestinal immunity were respectively increased (Ffar2, tendency, p = 0.06) or decreased (Ffar3, p = 0.01) in the Pro-sham+FVT compared to control mice (Figure 5K,L). Excluding the Pro-sham+FVT mice, the nod2 gene was lowered in expression in all treatment groups compared to control (Figure 5M).

Figure 5. The ileum tissue was investigated for changes in the expression levels of genes associated to inflammatory responses at termination. In A) to M) the expression levels of these selected genes are shown, and the different treatment groups are compared with the control mice. The relative abundance of total number of T cells N) and cytotoxic T cells (CD8+ T cells) P) were measured. abbreviations: “*” = (p < 0.05), “.” = (p < 0.1), Lacticaseibacillus rhamnosus GG = LGG, Akkermansia muciniphila = AKM, fecal virome transplantation = FVT, Pro-sham = probiotic sham.

The administration of A. muciniphila or L. rhamnosus along with FVT seemed to counteract this inflammatory response, since the expression of none of the above-mentioned 11 genes were changed in the AKM+Saline, AKM+FVT, or LGG+FVT compared to control mice (Figure 5A,L).

Fluorescence-activated cell sorting (FACS) was performed to evaluate the presence of selected immune cells in the mesenteric lymph node at termination. The FVT treated male and female mice expressed a significant (p = 0.043) decrease in the total number of T cells (CD3⁺ lymphocytes) (Figure 5N), while mice provided only LGG had a significant decrease (p = 0.01) in the level of cytotoxic T cells (CD8⁺ out of CD3⁺ T cells) (Figure 5O). Neither the level of cytotoxic T cells in the mesenteric lymph node or LPS levels in serum (Figure S11) were affected by the FVT treatment, suggesting that the increased expression of inflammatory genes in the ileum tissue, isolated from the Pro-sham+FVT mice, was not a broad intestinal response. The levels of different subsets of dendritic cells (CD11c⁺, CD86⁺CD11c⁺, CD11b⁺CD11c⁺, CD103⁺CD11c⁺) were not significantly affected by the sex of the animals, probiotics (AKM/LGG), or FVT.

Increased fertility rate following FVT

The pregnancy status and fertility rate (number of fetuses or born pups) were evaluated for each female mouse (Figure 6 and Figure S12) due to the natural consequences of co-housing male and female mice in cages. Unexpectedly, the FVT treated female mice (Pro-sham+FVT, AKM+FVT, and LGG+FVT) exhibited a significant increase in both fertility rate (p = 0.014, Figure 6A) and pregnancy status (p = 0.025, Figure 6B) compared to controls. These observations were independent of the administered probiotics. It was also observed that none of the female AKM+Saline mice (n = 4) became pregnant (Figure S12).

Figure 6. Bar plots of the fertility rate (number of pups) and pregnancy rate. A) The observed number of pups (born or as fetuses) based on a linear model (y~FVT+probiotics). B) The mean distribution of the of the binary event of pregnancy. Only the 23 female mice that received either FVT/Saline along with probiotic solutions of AKM/LGG/Pro-sham was included in the statistical analysis that was based on a generalized logistic regression model. Abbreviations: Lacticaseibacillus rhamnosus GG = LGG, Akkermansia muciniphila = AKM, fecal virome transplantation = FVT, Pro-sham = probiotic sham, “*” = (p < 0.05).

Discussion

Here, we investigated the potential of using phage-mediated “gut microbiome modulation” to improve the engraftment of two different probiotics (A. muciniphila; AKM and L. rhamnosus GG; LGG) in lean mouse recipients over a time span of 5 weeks. Species specific qPCR analysis showed that the fecal virome transplantation (FVT) treatment at termination (9 weeks of age) had significantly increased the abundance of naturally (native) occurring A. muciniphila strains, compared to non-FVT and non-AKM treated mice. The control and LGG+Saline mice also increased their A. muciniphila abundance over time which likely can be explained by regular gut microbiome maturation^33,34.

The applied FVT originated from a donor gut microbiome community that allowed a relative abundance of A. muciniphila >6% while no L. rhamnosus was detected amongst the 16S rRNA gene amplicons³¹. Other studies have reported that a phenotype can be transferred along with FVT^17–20. Thus, the transfer of a fecal phageome that allowed a relatively high abundance of A. muciniphila in the donors may modulate the gut microbiome of the recipient into a composition and conditions that favors the fitness of naturally occurring (native) A. muciniphila strains. For example, by phages targeting competitors to A. muciniphila or by stimulating mucus production directly or indirectly through gut microbiome modulation. In contrast, the FVT treatment did not enhance the abundance of native L. rhamnosus strains.

FVT significantly decreased the bacterial Shannon diversity in the LGG+FVT and Pro-sham+FVT mice at termination compared to the control and the other treatment groups (Figure 3A). A. muciniphila may have counteracted this tendency since the bacterial diversity of the AKM+FVT was unaffected and that Akkermansia spp. in some etiologies play a role in affecting the gut microbiome³⁵. Assuming that a gut microbiome composition can be transferred along with the FVT, the decrease in bacterial diversity may be associated with very low viral diversity and markedly different viral composition in the transferred FVT donor virome compared to the recipients. It could therefore be hypothesized that a low viral diversity may also favor a low bacterial diversity, due to the inherent link between phages and their bacterial hosts²³. With opposite sign, this may fit with a previous study where the viral diversity of the fecal donor virome was higher than the recipient virome and FVT resulted in increased bacterial diversity in the recipient mice compared to non-FVT treated mice¹⁸. More than 90% of the relative abundance of the FVT donor virome represented Microviridae viruses, which was in accordance with the relative abundance of genetically identical B6N mice in a previous study³¹. Furthermore, PhiX (a Microviridae used to spike the metavirome sequencing) sequences were excluded from the analysis, thus the high relative abundance of Microviridae was not a technical artifact. It should be noted that the initial viral diversity and composition at baseline appeared to vary within the treatment groups.

Both immune cell counting and the expression levels of genes involved inflammatory responses were measured to evaluate potential safety issues associated with FVT from lean donors to lean recipients. Here especially the Pro-sham+FVT appeared with significantly increased gene expression levels compared to control mice (Figure 5) in the following pro- and/or anti-inflammatory related genes; Ccl2³⁶, Ccr10³⁷, Ctla4³⁸, Cxcl1³⁹, Il1b⁴⁰, Il4⁴¹, Il6⁴², Retnlb⁴³, Timp1⁴⁴, Ffar2^45,46, whereas Ffar3⁴⁵ and Nod2⁴⁷ decreased. Interestingly, the combination of AKM+FVT counteracted the elevated expression of these genes, which may be explained by previously suggested synergistic effects of combining probiotics and phages⁴⁸. These indications of inflammatory response in the ileum tissue were not reflected by more systemic measures as immune cell counts in the mesenteric lymph node or the LPS levels in the blood serum (Figure S11). The female mice had increased LPS levels relative to the male mice, which may be explained by a low-grade inflammation in the third trimester⁴⁹. Hence, suggesting that the FVT did not activate a broad intestinal immune response, but rather a more a local response due to the presence of foreign viral particles^50,51.

The administration of A. muciniphila significantly decreased the expression of the Muc1 gene that codes for a membrane-tethered mucin with signaling properties^32,52, but not Muc2, which codes for the major secreted intestinal mucin that provides the bulk of the mucus layer³². Indeed, the ileal mucin density assessed by histochemical staining was not affected by A. muciniphila administration. Hence, this gene expression pattern of Muc1 might reflect a local signaling interaction between A. muciniphila and epithelial cells with no effect on the overall structure of the intestinal mucus layer. A. muciniphila is a mucin-degrading bacteria and has been suggested to have a beneficial impact on human health, of which is linked to the regulation of the mucus thickness and gut barrier integrity⁵³. No treatments were associated to elevated serum LPS levels, while mice administered A. muciniphila as probiotic tended to have lower LPS level in serum (Figure S11).

The FVT treated mice were unexpectedly associated with increased fertility rate (Figure 6A) and pregnancy status (Figure 6B). The study was not designed to investigate fertility rates which also are reflected by the group sizes. Both the male and female mice were treated with FVT; thus, the basic premises of the experimental setup make it challenging to evaluate if the increased fertility rate and pregnancy status were due to improved sperm quality of the males and/or improved conditions for fertility in the female mice. Emerging evidence suggest that infertility should be added to the list of gut microbiome associated diseases^54–56, and the importance of validating this link to the gut microbiome is emphasized by infertility being estimated to affect up to 15% of couples world-wide^57,58. Our observations are in line with other studies suggesting a link between fertility and the GM, e.g. the demonstrated improvement of spermatogenesis with fecal microbiota transplantation from healthy donors⁵⁹, as well as, impairment of spermatogenesis with fecal microbiota transplantation from donors with a dysbiotic gut microbiome⁵⁴. With regard to females, links have been suggested between maternal obesity, gut dysbiosis, and inflammation⁵⁶. New results have discovered a markedly increase in the abundance of Bacteroides vulgatus in the gut of polycystic ovary syndrome (PCOS) individuals, that through deconjugations of bile acids in the liver affects the interleukin-22 (IL-22) levels and ultimately the fertility⁵⁵. Although additional experiments need to be conducted, it cannot be ruled out that the transfer of A. muciniphila in the AKM+Saline mice led to similar cascading events that might have decreased the fertility of the male and/or female mice (Figure 6).

A water-oil emulsion solution based on Intralipid was used as sham (control) for the probiotics. A daily intake of intralipid has been shown to affect glucose kinetics in rats⁶⁰ and reported to have preventive effects on cholestasis in pigs, which was reflected as changes in metabolic and GM profiles⁶¹. In this study, all mice were only administered small volumes of Intralipid twice. A potential effect associated with the Intralipid on the mouse phenotype would be expected to be equally distributed between the treatment groups as well as diminished at the time of termination.

While there were good indications that phages were a key component in the observed effects associated with the FVT treatment, it remains possible that metabolites or entities with a molecular size above 30 kDa (size cutoff of applied centrifugal filter) may have contributed to the observed effects. These molecules could for instance be metabolites from lactobacilli (cell wall muramidases)⁶², and Akkermansia (pasteurized cell cultures)¹⁰, bacteriocins with antimicrobial properties affecting the GM composition^63–65 or extracellular vehicles which have been shown to affect immune regulation during pregnancy^66,67, and may be involved in the etiology of inflammatory bowel diseases^68,69. However, considering a long-term colonization of donor phages in FMT studies^70–72, phages being associated with the treatment outcome of recurrent C. difficile infection^20,72, no reported effects of heat-treated FVT controls^22,73, and studies reporting beneficial effects of FVT in different etiology regimes^{17–19,22,74}, it suggests that the viral component of FVT constitutes an important role. A guideline was recently proposed to ensure reproducible clinical studies using FMT, and these recommendations may be transferrable for future FVT studies as well⁷⁵.

Conclusively, we here demonstrate that FVT increases the abundance of, what is expected to be, naturally occurring A. muciniphila strains in the recipient mice. The bulk and undefined nature of fecal viromes prevents any direct use as a commercial product, but our results highlight the potential of using phage-mediated changes of the gut microbiome as a supplement to probiotics to enhance the growth of healthy commensals that outside the body are defined as probiotics. An unexpected event of increased fertility rate and pregnancy status was associated with the FVT treatment, which urge for additional studies specifically designed to clarify our observations.

Methods

Bacterial strains

The commercially available probiotic bacterium Lacticaseibacillus rhamnosus GG (LMG 18,243, former Lactobacillus rhamnosus⁷⁶) was included along with Akkermansia muciniphila YL-44 (DSM 26,127), as a representative of next-generation probiotics⁷⁷.

Preparation of inocula of L. rhamnosus (LGG) and A. muciniphila (AKM) for transfer to mice

L. rhamnosus GG (LGG) and A. muciniphila YL-44 (AKM) were both handled and incubated anaerobically as described previously⁷⁸. AKM was incubated in Gifu Anaerobic Medium (GAM, HyServe, cat. No. 5422) and LGG in de Man Rogosa Sharpe broth (MRS, Merck, cat. No. 69966) in broth or agar plates containing 1.5% agar. In brief, GAM or MRS broth were boiled prior to distribution in Hungate tubes (SciQuip, cat. No. 2047–16125), and subsequently flushes with 100% N₂ with an anaerobic gassing unit (QCAL Messtechnik GmbH) for at least 3 min per 10 mL. Both liquid and solid media contained 0.02% (w/v) 1,4-dithiothreitol (Merck, cat. No. DTT-RO) and 0.05% (w/v) L-cysteine (Merck, cat. No. 168149) as reducing agents and 1 mg/L resazurin as oxygen indicator. All media were autoclaved (121°C for 20 min). Anaerobic handling of cultures was performed in an anaerobic chamber (Model AALC, Coy Laboratory Products) containing ~93% (v/v) N₂, ~2% H₂, ~5% CO₂ at room temperature (RT), and agar plates were incubated in an anaerobic jar (Thermo Scientific, cat. No. HP0011A,) along with an anaerobic sachet (Thermo Scientific, AnaeroGen cat. No. AN0035A). Incubation of tubes as well as plates was performed at 37°C. For preparing the probiotic solutions, a single bacterial colony was inoculated to the growth medium and incubated until the stationary phase was reached after 24 h for LGG or 72 h for AKM. This was followed by a 2% (v/v) culture incubation until the exponential phase was reached after 12 h for LGG or 48 h for AKM. The bacterial concentrations were measured with an optical density at 600 nm (OD₆₀₀) with Genesys 30 Visible spectrophotometer (Thermo Scientific, cat. No. 840–277000) mounted with a test-tube holder (VWR, cat. No. 634–0911). To ensure high bacterial loads in the probiotic inocula, the bacterial cultures were upconcentrated 40× by centrifugation at 4,450 × g for 30 min at RT under anaerobic conditions and resuspended in anaerobic intralipid. Intralipid was used to protect the viable bacterial cells against the acidic environment in the mouse upper gastrointestinal tract⁷⁹ and an oil-water emulsion solution was made by mixing the resuspended bacterial cultures with a 3-way stopcock (Braun, Discofix cat. No. 409511). Pure Intralipid was used as the probiotic sham (Pro-sham). Small single-use vials of the probiotic solutions were prepared for each mouse to minimize the introduction of oxygen when administering the probiotics. The probiotic solutions were freshly prepared for both the 1st and 2nd inoculations which explain the variances in the bacterial colony forming units (CFU)/mL. Phase contrast microscopy images were taken to check for contamination on the cell morphology level (Figure S13). The total CFU transferred to each mouse at 1st inoculation was LGG: 2.8 × 10⁸ CFU and AKM: 3.8 × 10⁸ CFU and at 2nd inoculation LGG: 5.5 × 10⁹ CFU and AKM: 2.0 × 10⁹ CFU.

Preparation of donor virome

Fecal viromes were extracted from intestinal content from mice (low-fat diet fed male C57BL/6NCrl and C57BL/6NRj mice) that previously³¹ was found with a relative abundance of A. muciniphila above 6% and to exhibit inter-vendor variance in their gut microbiome profiles^31,80. The titer of the applied FVT virome was approximately 5.4 × 10⁹ virus-like particles (VLP)/mL (Figure S14 and Table S1) for both 1st and 2nd inoculations and was evaluated by epifluorescence microscopy stained by SYBR Gold (Thermo Scientific, cat. No. S11494) as previously described (dx.doi.org/10.17504/protocols.io.bx6cpraw). The total VLPs transferred to each mouse per inoculation was 8.0 × 10⁸ VLPs. SM buffer (NaCl 200 mM, MgSO₄·7 H₂O 16 mM, Tris-HCl 100 mM, pH 7.5) was used as viral sham (Saline).

Animal study design

In total 48 C57BL/6NTac mice at 4 weeks old (Taconic) were included in the study (representing 24 males and 24 females). They were ear tagged upon arrival and divided into six groups: LGG+FVT, AKM+FVT, Pro-sham+FVT, LGG+Saline, AKM+Saline, and control (Pro-sham+Saline) (Figure 1). The saline consisted of SM buffer and Pro-sham consisted of Intralipid (Fresenius Kabi, Intralipid 200 mg/mL) that were used to suspend the probiotic bacteria. The mice were housed in open transparent cages with a wire lid (1290D Eurostandard Type III, Scanbur A/S, Karlslunde, Denmark) with access to bottled tap water ad libitum, and the cages were enriched with bedding, cardboard housing, tunnel, nesting material, felt pad, and biting stem (Brogaarden, respectively, cat. No. 30983, 31000, 31003, 31008, 31007, 30968). The mice were fed ad libitum chow diet (Brogaarden, Altromin 1324) during the entire 6 weeks of the study. Health monitoring of animals was performed without revealing any pathogens according to FELASA guidelines⁸¹. Cages were changed weekly. The mice were housed in male-female pairs (in total 24 cages) to evaluate the effect of sex of the animals on the interventions, increase animal welfare by eliminating aggression between co-housed males, and by consequence also allowed natural mating behavior. After a week of acclimatization, the mice were inoculated orally using a pipette with 50 µl 1 M bicarbonate solution (Merck, cat. No. S5761) that 5 min later was followed by oral gavage with 0.15 mL FVT/Saline solutions (FVT/Saline, n = 24). The following day the mice were inoculated orally by gently using a pipette with 100 µL probiotic solution of AKM/LGG/Pro-sham (AKM/Pro-sham, n = 16 and LGG/Pro-sham, n = 16), which constituted the 1^st inoculation. The same procedures were repeated in the 2^nd inoculation a week after. The mice were weighted, and fecal samples were taken at several timepoints during the study, amongst other at baseline, 6 days after 2nd intervention, and at termination. The fecal samples were stored at −80°C. One female mouse (representing the AKM+FVT group) was sacrificed following the first probiotic inoculation due to suffering. At termination (9 weeks of age), the remaining 47 mice were anesthetized with a hypnorm/midazolam mixture. Both hypnorm (VetaPharma, cat. No. P736/005) and midazolam were mixed with sterile water in a ratio of 1:1 (Braun, cat. No. 353 0418). Retro-orbital sampling of blood was performed after effective anesthetization with sterile heparinized hematocrit capillary tubes (Merck, cat. No. BR749321). The blood was centrifuged at 6,000 × g at 4°C for 10 min and the supernatant was carefully sampled as serum. The animals were euthanized by cervical dislocation. The mesenteric lymph node was sampled in ice-cold PBS and 2 cm of the distal ileum was sampled in two pieces and snap frozen in liquid nitrogen and stored in −80°C. Surgical equipment used for tissue and fecal sampling during terminal procedures was sterilized between each animal. Pups, both born and in utero were counted and euthanized by decapitation. The study was approved by the Danish Competent Authority, The Animal Experimentation Inspectorate, under the Ministry of Environment and Food of Denmark, and performed under license No. 2017-15-0201–01262 C1–3. Procedures were carried out in accordance with the Directive 2010/63/EU and to the Danish law LBK Nr 726 af 09/091993, and housing conditions as earlier described³¹.

Gene expression assay

Ileum (1 cm) pieces were transferred to tubes (Mpbio, cat. No. FastPrep 50-76-200) including 0.6 g glass beads (Sigma-Aldrich, cat. No. G4649), 600 µl lysis binding solution concentrate (Invitrogen, cat. No. AM1830) and 0.7% β-mercaptoethanol (Sigma-Aldrich, cat. No. M6250) and homogenized on the FastPrep-24™ Classic Instrument (Mpbio) with 4 × (45 sec at speed 6.5 m/s) runs. The homogenate was centrifuged at 16,000 × g and the supernatant was frozen at −20°C for at least 24 h before purification of RNA using the MagMax Express Magnetic Particle Processor (Applied Biosystems) using manufacturer’s instructions (Invitrogen, cat. No. AM1830). RNA purity and concentration were assessed using DeNovix DS11 Fx+ Spectrophotometer (DeNovix) and intact 18S rRNA and 28S rRNA bands were visually inspected on a 1.4% agarose gel. cDNA was synthesized from 500 ng total RNA with the High-capacity cDNA reverse Transcription Kit (Applied Biosystems) in a reaction volume of 20 µl following manufacturer’s recommendations. RT controls were prepared without the reverse transcriptase enzyme and cDNA samples was diluted 8× after synthesis. High throughput qPCR was run on duplicates on the Biomark HD system (Fluidigm Corporation) on 2 × 96.96 IFC chips on pre-amplified cDNA duplicates using manufacturer’s instructions with minor adjustments as previously described⁸². The majority of the primers in the Ileum primer panel was previously published¹⁸. Eighty-one primer assays (74 candidate genes, 7 reference genes, 1 gDNA control assay) (see Supplementary file 1 for the full list) were present with one product and had a sufficient efficiency between 75% and 110%. In addition, the MVP1 gene assay was included to control for gDNA contamination⁸³. qPCR data were analyzed as previously described and normalized to the four most stable reference genes: Sdha, Tuba, Pgk1, Ppia^18,82. A log2 fold change threshold was set to 0.5 and the FDR p-value to 0.05 for gene expressions to be included in the analysis.

Cell isolation and flow cytometry (FACS)

Directly after euthanasia of the mice, the mesenteric lymph node was placed in ice-cold PBS. Single-cell suspensions were prepared by disrupting the lymph node between two microscope glasses and passing it through a 70 μm nylon mesh. After washing and resuspension, 1 × 106 cells were surface stained for 30 min with antibodies for Percp-Cy5.5 conjugated CD11c, PE-conjugated CD86, APC-conjugated CD11b, and FITC-conjugated CD103 (all antibodies were purchased from eBiosciences) for the detection of tolerogenic dendritic cells (DCs). For the detection of T cell subsets, 1 × 106 cells were initially surface stained for 30 min with FITC-conjugated CD3, PercP-Cy5.5-conjugated CD4, and APC-conjugated CD8a (ebiosciences), then fixate and permeabilized with the FoxP3/Transcription Factor Staining Buffer Set (ebiosciences), and finally stained for 30 min with PE-conjugated intracellular forkhead box P3 (FOXP3) (ebioscience). Analysis was performed using an Accuri C6 flow cytometer (Accuri Cytometers).

Lipopolysaccharide (LPS) content in blood serum

Lipopolysaccharide (LPS) content was measured on serum collected retro-orbitally from the mice before euthanization using Lonza Pyrogene Recombinant Factor C Endotoxin Detection Assay (Lonza) as previously described⁸⁴. In brief, LPS was diluted 1:1000 and procedures were carried out by following the manufacturer’s instructions, except an additional step where the samples were heated in 70°C for 10 min prior being added to the wells of the microplate. The microplate was measured at time zero and at 1 h of incubation along with the catalyzing reagents. The concentration was calculated in EU, as the values obtained at the second reading subtracted from the values obtained at the first reading.

Ileum histology

The ileum tissue was immediately after sampling frozen in liquid nitrogen. The frozen tissue was embedded in Tissue-Tek optimal cutting temperature gel (Sakura, Torrance, CA, USA) and sectioned on a freezing microtome. Mounted 2 µm sections were stained with Alcian Blue-periodic acid-Schiff (AB-PAS) to visualize mucin glycoproteins. Microscopic images captured at 10× magnification were subject to image analysis using ImageJ (National Institutes of Health, Bethesda, MD, USA). Images were color deconvoluted into AB and PAS channels respectively, and relative staining area estimated after signal thresholding. AB signal-to-background was too poor to yield useful results, but manual inspection of sections ensured a high degree of overlap between AB and PAS staining patterns. Hence, relative periodic acid-Schiff area was reported as a measure of intestinal mucin density.

Pre-processing of fecal samples

Fecal samples from three different timepoints were included to investigate gut microbiome changes over time: baseline, after intervention, and at termination. This represented in a total of 142 fecal samples from the C57BL/6NTac mice. Separation of the viruses and bacteria from the fecal samples generated a fecal pellet and fecal supernatant as earlier described³¹.

Quantitative real-time PCR for measuring probiotic density

The bacterial density of AKM and LGG in the fecal samples was estimated using quantitative real-time polymerase-chain reaction (qPCR) with species specific primers (AKM_Fwd: 5’-CCT TGC GGT TGG CTT CAG AT-3’ and AKM_Rev: 5’-CAG CAC GTG AAG GTG GGG AC-3’⁸⁵ and LGG_Fwd: 5’-GCC GAT CGT TGA CGT TAG TTG G-3’ and LGG_Rev: 5’-CAG CGG TTA TGC GAT GCG AAT-3’⁸⁶) purchased from Integrated DNA Technologies. Standard curves (Table S2) were based on a dilution series of total DNA extracted from monocultures of AKM and LGG. The qPCR results were obtained using the CFX96 Touch Real-Time PCR Detections System (Bio-Rad Laboratories) and the reagent SsoFast EvaGreen Supermix with Low ROX (Bio-Rad Laboratories, cat. No. 1725211), and run as previously described⁸⁷.

Bacterial DNA extraction, sequencing and pre-processing of raw data

The Bead-Beat Micro AX Gravity kit (A&A Biotechnology, cat. No. 106–100 mod.1) was used to extract bacterial DNA from the fecal pellet by following the instructions of the manufacturer. The final purified DNA was stored at −80ºC, and the DNA concentration was determined using Qubit HS Assay Kit on the Qubit 4 Fluorometric Quantification device (Invitrogen). The bacterial community composition was determined by Illumina NextSeq-based high-throughput sequencing of the 16S rRNA gene V3-region, as previously described³¹. Quality-control of reads, de-replicating, purging from chimeric reads and constructing amplicon sequence variants (ASVs) were conducted with the UNOISE pipeline⁸⁸ and taxonomically assigned with Sintax⁸⁹ (not yet peer reviewed). Taxonomical assignments were obtained using the EZtaxon for 16S rRNA gene database⁹⁰. Code describing this pipeline can be accessed in github.com/jcame/Fastq_2_zOTUtable. The average sequencing depth after quality control (Accession: PRJEB52388, available at ENA) for the fecal 16S rRNA gene amplicons was 47,526 reads (min. 6,588 reads and max. 123,389 reads).

Viral DNA extraction, sequencing and pre-processing of raw data

The sterile filtered fecal supernatant was concentrated using Centriprep filter units (Merck, cat. No. 4311 and cat. No. 4307). This constituted the concentrated virome. Due to a permanent stop in the production of Centriprep filter units at the manufacturer, we were forced to use residual stocks of filter size 30 kDa (cat. No. 4307) for 42% of the samples (Table S3). The samples were all centrifuged at 1500 × g at 15ºC until approx. 500 µL concentrated virome sample was left, the filter was removed with a sterile scalpel and stored along with the concentrated virome at 4ºC. Viral DNA was extracted, multiple-displacement amplification (MDA, to include ssDNA viruses), and Illumina NextSeq sequencing data generated as previously described³¹. The average sequencing depth after quality control (Accession: PRJEB52388, available at ENA) for the fecal viral metagenome was 209,641 reads (min. 21,580 reads and max. 510,332 reads. The raw reads were trimmed from adaptors and the high-quality sequences (>95% quality) using Trimmomatic v0.35⁹¹ with a minimum size of 50nt were retained for further analysis. High quality reads were de-replicated and checked for the presence of PhiX control using BBMap (bbduk.sh) (https://www.osti.gov/servlets/purl/1241166). Virus-like particle-derived DNA sequences were subjected to within-sample de-novo assembly-only using Spades v3.13.1⁹² and the contigs with a minimum length of 2,200 nt, were retained. Contigs generated from all samples were pooled and de-replicated at 90% identity using BBMap (dedupe.sh). Prediction of viral contigs/genomes was carried out using VirSorter2⁹³ (“full” categories | dsDNAphage, ssDNA, RNA, Lavidaviridae, NCLDV | viralquality≥0.66), vibrant⁹⁴ (High-quality | Complete), and checkv⁹⁵ (High-quality | Complete). Taxonomy was inferred by blasting viral ORF against viral orthologous groups (https://vogdb.org) and the Lowest Common Ancestor (LCA) for every contig was estimated based on a minimum e-value of 10e⁻⁵. Phage-host prediction was determined by blasting (85% identity) CRISPR spacers and tRNAs predicted from >150,000 gut species-level genome bins (SGBs)^96,97 (⁹⁷, not yet peer reviewed). Following assembly, quality control, and annotations, reads from all samples were mapped against the viral (high-quality) contigs (vOTUs) using the bowtie2⁹⁸ and a contingency-table of reads per Kbp of contig sequence per million reads sample (RPKM) was generated, here defined as vOTU-table (viral contigs). Code describing this pipeline can be accessed in github.com/jcame/virome_analysis-FOOD.

Bioinformatic analysis of bacterial and viral DNA sequences

Initially, the RPKM normalized dataset was purged for viral contigs which were detected in less than 5% of the samples, but the resulting dataset still maintained 99.5% of the total reads. Cumulative sum scaling (CSS)⁹⁹ was applied for the analysis of β-diversity to counteract that a few viral contigs represented a majority of count values, since CSS have been benchmarked with a high accuracy for the applied metrics¹⁰⁰. CSS normalization was performed using the R software using the metagenomeSeq package¹⁰¹. Α-diversity analysis was based on raw read counts and statistics were based on ANOVA. R version 4.01¹⁰² was used for subsequent analysis and presentation of data. The data are uploaded as supplementary data (https://osf.io/tm2a5/). The main packages used were phyloseq¹⁰³, vegan¹⁰⁴, deseq2¹⁰⁵, ampvis2¹⁰⁶ (not yet peer reviewed), ggpubr¹⁰⁷, mctoolsr (https://github.com/leffj/mctoolsr/), and ggplot2¹⁰⁸. Β-diversity was represented by Bray Curtis dissimilarity and statistics were based on PERMANOVA. A linear model (y~FVT+probiotics+sex), similar to ANOVA, was applied to assess the statistically differences between the treatment groups of gene expression, bacterial abundance, immune cell counts, fertility rate, LPS content, and PAS staining area. Two treatment groups were applied in the model; a FVT group (levels: control and FVT) and a probiotic group (levels: control, AKM and LGG). The sex of the animal was added as an additional factor, except for fertility outcome analysis where only females were included. For binary outcomes, a generalized logistic regression model was applied.

Author contributions

TSR, CMJM, AKH, and DSN conceived the research idea and designed the study; TSR, CMJM, MRD, LSFZ, performed the experiments TSR, CMJM, MRD, RRJ, LSFZ, JLCM, AB, CHFH, LHH, AKH, and DSN performed laboratory and data analysis; TSR wrote the first draft of the manuscript. All authors critically revised and approved the final version of the manuscript.

Acknowledgments

We thank the animal caretakers Helene Farlov and Mette Nelander at Section of Experimental Animal Models (University of Copenhagen, Denmark) for taking care of the animals during the study and assisting with the animal handling.

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability statement

Raw sequencing data can be accessed at ENA with project ID: PRJEB52388 (https://www.ebi.ac.uk/ena/browser/).

Supplemental data

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19490976.2023.2208504

Additional information

Funding

Funding was provided by the Danish Council for Independent Research with grant ID: DFF-6111-00316 “PhageGut” (www.phagegut.ku.dk).