Gut microbiome modulation during treatment of mucositis with the dairy bacterium Lactococcus lactis and recombinant strain secreting human antimicrobial PAP

Mucositis is an inflammatory condition of the gut, caused by an adverse effect of chemotherapy drugs, such as 5-fluorouracil (5-FU). In an attempt to develop alternative treatments for the disease, several research groups have proposed the use of probiotics, in particular, Lactic Acid Bacteria (LAB). In this context, the use of recombinant LAB, for delivering anti-inflammatory compounds has also been explored. In previous work, we demonstrated that either Lactococcus lactis NZ9000 or a recombinant strain expressing an antimicrobial peptide involved in human gut homeostasis, the Pancreatitis-associated Protein (PAP), could ameliorate 5-FU-induced mucositis in mice. However, the impact of these strains on the gut microbiota still needs to be elucidated. Therefore, in the present study, we aimed to characterize the effects of both Lactococci strains in the gut microbiome of mice through a 16 S rRNA gene sequencing metagenomic approach. Our data show 5-FU caused a significant decrease in protective bacteria and increase of several bacteria associated with pro-inflammatory traits. The Lactococci strains were shown to reduce several potential opportunistic microbes, while PAP delivery was able to suppress the growth of Enterobacteriaceae during inflammation. We conclude the strain secreting antimicrobial PAP was more effective in the control of 5-FU-dysbiosis.

Oncology treatments based on chemotherapy or radiotherapy are responsible for the occurrence of a gastrointestinal inflammatory condition known as mucositis [1][2][3] . Chemotherapy drugs, including 5-fluorouracil (5-FU), irinotecan and methotrexate, present high toxicity to cells with high proliferation rates, such as intestinal epithelial cells lining the gut mucosa. These drugs lead to apoptosis of those cells generating significant damage to epithelial barrier integrity, which allows bacterial colonization, invasion and consecutive triggering of inflammatory processes 4,5 . The scientific community believed that the gut microbiota would play a secondary role in the pathogenesis of the disease limited to aggravating conditions, such as bacterial translocation. However, recent studies have been investigating the role of commensal intestinal microbes under an ecological perspective where diverse organisms occupy niches that are essential for the development of mucositis 6,7 . The gut microbiota that colonizes the epithelial barrier of the intestine is composed mostly of bacteria which contribute to many functions of the host, while some are referred to pathobionts being capable of acquiring pathogenic characteristics under intestinal ecology disturbance 6 . Interestingly, it has been shown that germ-free mice are more resistant to 5-FU-induced mucositis, which reinforces the hypothesis that the microbiota is essential for the disease development 5,8 . Several pre-clinical and clinical studies have reported modifications in fecal microbiota diversity and composition following chemotherapy 9-12 . 5-FU causes disruption in the microbial community structure when compared to the noninflamed groups. There were no statistically signifcant variation on richness (Fig. 2), Shannon (Fig. 3), and evenness ( Fig. 4) among the groups, except LL group, which showed the highest richness when compared to Naïve (P = 0.014), LL5FU (P = 0.022) and PAP5FU (P = 0.028). However, the community structure of the naïve group was statistically different from the inflamed groups (5FU, LL5FU and PAP5FU). This dissimilarity was also observed when compared the LL and PAP to the inflamed groups (Table 1 and Table 2). Moreover, the group with inflamed animals that did not feed on probiotics (5FU group) showed highest values of the dissimilarity when compared to naïve, LL and PAP groups (R values in ANOSIM).
L. lactis NZ9000 and PAP are able to change the gut microbiota composition. There were no significant statistical differences in the phylum abundances between Naïve and LL group (Supplementary Figure 2); however, there was a significant decrease in the percentage of Actinobacteria (P = 0.003) in the animals fed with L. lactis expressing PAP when compared to the Naïve group. When analyzed at the OTU level, LL group had a lower abundance of the OTU identified as Clostridiaceae and higher abundance of Lactobacillales, Peptococcaceae, and RF39 than Naïve group ( Fig. 5 Figure 2. Microbial richness of the fecal microbiota among the groups. LL group showed an increase in richness when compared to Naïve (P = 0.014), LL5FU (P = 0.022) and PAP5FU (P = 0.028). Bonferroni, P-value <0,05.

Mucositis induced by 5-Fluorouracil promotes a broad modification of the OTUs abundances in mice fecal microbiota.
When the animals were inflamed, the abundance of Actinobacteria significantly decreased while the number of Verrucomicrobia increased when compared to Naïve (Supplementary Figure 2). Moreover, the OTUs Adlercreutzia, Anaeroplasma, Clostridium, Helicobacter, Lactobacillus, Odoribacter, Rikenella and Streptophyta significantly decreased when compared with the Naïve groups. At the same time, the following OTUs had the relative abundance increased in the inflamed animals: Akkermansia, Bilophila, Dehalobacterium, Desulfovibrio, Desulfovibrionaceae, Parabacteroides, Peptococcaceae, RF32, Rhodocyclaceae (

Discussion
Recently, the interest in using probiotics for alleviating intestinal inflammation in patients submitted to antineoplastic chemotherapy has increased 15,26,27 . Although several works demonstrate the protective effects of probiotics in preventing mucositis, their impact on the microbial community structure has been poorly explored. In this study, we evaluated the effects of two L. lactis strains with anti-inflammatory properties, NZ9000 and LL-PAP 24,28 , on the intestinal microbial community structure through a 16 S rRNA gene metataxonomic analysis in a mouse model of mucositis. Three major phyla constituted the predominant gut microbiota in all mice from all experimental groups, consistent with previous surveys done with mammalian subjects 29 . In addition, almost all the genera constituting the Mock communities were detected. Although this finding was expected, it gives us more reliability when comparing our data with other related studies. The 5-FU administration can cause a disturbance in the community structure, usually decreasing the richness and abundance of OTUs 30 . We were expecting to observe a similar effect in the inflamed animals, and despite the richness and diversity indices did not vary among the groups after 5-FU injection, we observed a significant alteration in the microbial community structure when comparing the noninflamed to the inflamed groups as indicated by the dissimilarity analyses. These findings suggest that different 5-FU regimens might cause disturbance states with particular structural traits.
At phylum level analysis, the proportion of Verrucomicrobia presented significant changes after 5-FU injection, corroborating with previous studies 31,32 . There are very few species belonging to this phylum found in the gut to date. The most dominant is Akkermansia muciniphila, a bacterium that scavenges mucins as a carbon and nitrogen source which has been inversely associated with obesity and diabetes, and presents protective activity in DSS-colitis in mice 33 . In the work of Kang and colleagues, extracted vesicles from A. muciniphila ameliorated inflammatory damage in the colon and reduced the expression of pro-inflammatory IL-6 stimulated by E. coli. Interestingly, we identified a genus-level OTU belonging to Verrucomicrobia, assigned as Akkermansia, being solely responsible for the increased proportion of Verrucomicrobia in the inflamed groups. The treatment with L. lactis presented an enrichment of Akkermansia compared to mice receiving 5-FU, although it was not statistically significant.
Actinobacteria was less enriched in the groups submitted to 5-FU injection compared to naive control. Intriguingly, this phylum was also decreased in healthy mice after treatment with L. lactis secreting PAP. Since Actinobacteria is almost exclusively formed by Gram-positive bacteria, its reduced abundance could be due to PAP anti-microbial affinity against Gram-positive bacteria as previously described 34,35 . However, there are controversies in the literature regarding the proportion of Actinobacteria in response to inflammation making it challenging to unravel its biological implications. For example, Bifidobacterium spp are considered as dominant bugs presenting anti-inflammatory properties, but other genera probably might play essential roles in the gut as well 10,36 . In our work, we identified two genera of Actinobacteria being significantly influenced during mucositis. A reduced proportion of Adlercreutzia was observed in mice receiving 5-FU suggesting its niche is essential for avoiding a dysbiosis state. The decreased abundance of Adlercreutzia has been previously reported in cases of colitis patients compared to control group, and in another clinical study, patients have shown a reduced proportion after chemotherapy submission 10 . Interestingly, this genus is currently formed by one species, A. equolifaciens, which produces an isoflavone metabolite, named equol, with anti-cancerous and anti-inflammatory properties 37,38 . Moreover, studies reveal that equol is exclusively produced by the intestinal microbiota 39 . The treatment with PAP did not cause any alteration at OTU level regarding Actinobacteria. The consumption of L. lactis culture did not seem to significantly increase the abundance of Adlercreutzia as well, but caused a reduction of another genus identified as Corynebacterium. In humans, these Gram-positive bacteria are commonly found on the skin, and some members of the genus are opportunistic pathogens when colonizing other sites of the body, such as the oral cavity. Commensal species of corynebacteria found in the gut are acquired from the mother's skin mainly in cesarian-section infants 29 . Opportunistic Corynebacterium spp. have also been isolated from oral mucositis patients 40,41 . It is possible that competitive exclusion activity from L. lactis in reducing the abundance of corynebacteria might affect the dysbiosis state. However, their role in the gut requires further investigation.
In a similar context, there is no previous report about Tenericutes traits as indicators of health state in mucositis. Unclassified RF39 and Anaeroplasma were found decreased in inflamed mice, treated only with 5-FU, while delivery of PAP was able to restore RF39 to normal levels.
The candidate phylum TM7 is a recently described subgroup of Gram-positive uncultivable bacteria initially found in different natural environmental habitats 42,43 . In our study, we identified a TM7 family-level OTU assigned as F16. Mice receiving 5-FU did not show alteration while treatment with L. lactis has caused significant enrichment of TM7/F16 at phylum-and OTU-level respectively. Controversially, Li and colleagues found a reduced abundance of TM7 proportion in mice treated with 5-FU in mice 32 . Although no definite correlation of TM7 has been associated with 5-FU-induced mucositis up to date, previous studies show TM7 OTUs has been associated with the pathogenesis of periodontitis 44 . Similarly, another study investigated TM7 in inflammatory bowel diseases (IBDs) suggesting it might play a key role in the development of inflammation 45 . Therefore, targeting TM7/F16 is of extreme importance to be investigated in further studies to improve or develop novel strategies for treating the disease.
No alteration was detected for Firmicutes, Bacteriodetes, and Proteobacteria at phylum-level analysis, possibly because they present the highest richness of OTUs being either down-or up-modulated. Perhaps they are more ecologically stable when compared to subdominant phyla such as Verrucomicrobia, TM7, and Actinobacteria, which presented significant changes after treatment with 5-FU or the Lactococci strains.
Bacteriodetes phylum comprised two members, Rikenella and Odoribacter, that were found decreased after 5-FU injection. Odoribacter spp are considered atypical opportunistic commensals because they produce butyrate and their presences are essential for preventing diseases such as hypertension though they may also contribute to intestinal abscesses 46 . Low levels of the Odoribacter population has been found in IBD patients 47 . However, the association of Rikenella with bad or good prognosis for inflammatory diseases has not been reported yet. Another Bacteriodetes, assigned as Parabacteroides, was found enriched in inflamed mice. Parabacteroides spp. are essential for digesting high-fiber diets that humans cannot process, and they tend to be missing from the gut of patients suffering from IBD 48 . L. lactis or PAP treatment in mice did not significantly affect any Bacteriodetes OTU either in healthy or inflamed mice, suggesting this phylum may present robustness against PAP inhibitory property and is less susceptible to L. lactis effects in the gut.
Firmicutes phylum is mainly formed by Gram-positive species of bacteria occupying several niches in the intestines, such as the production of Short-chain fatty acids (SCFAs) and trophic functions, although some are considered pathobionts as well 49,50 . Within Firmicutes, we verified that the majority of modulatory effects occurred in the Clostridialles, Lactobacillales and Bacillales. The Lactobacillales order, which is virtually formed by many species of bacteria with anti-inflammatory properties [51][52][53] , was found enriched in healthy mice that fed on L. lactis NZ9000 culture but seemed to be partially abrogated by 5-FU activity. The decrease of Lactobacillus corroborates with the study of Florez and colleagues, in which authors suggest that LAB species are more susceptible to 5-FU effects than other intestinal bacteria 54 . Although the treatment with L. lactis NZ9000 did not restore the abundance of Lactobacillus during mucositis, it caused an increase of the genus Lactococcus. Despite Lactococcus spp. are not usually considered to be commensal, this OTU was found in all groups, including mice that did not feed on the cultures containing live Lactococci strains.
The Lactobacillales order also contains opportunistic bacteria including Streptococcus spp and Enterococcus spp. In our work, unclassified Enterococaccea were found increased in inflamed mice treated with L. lactis NZ9000. In a recent study, Enterococcaceae dominance was associated with higher risk of neutropenia and diarrheal illness after chemotherapy treatment 55 . We were expecting a decrease of Enterococcaceae in mice treated with PAP, as we have previously demonstrated it was able to inhibit a representative commensal from this family, E. faecalis, in vitro, 24 . For unknown reasons, in the present study, we observed reduced levels of Enterococcaceae in PAP-treated mice, but it was not statistically significant when compared to the treatment with L. lactis NZ9000.
The Bacillales member Staphylococcus was decreased in mice consuming L. lactis wild-type strain. In humans, this genus comprises opportunistic commensals colonizing the skin and mucosal surfaces lining the nose and ear cavities. Studies suggest that parental transmission is the most common form for infants-gut colonization 56 . A study revealed that different Staphylococcus spp. strains had been isolated from the mouth of chemotherapy patients, presenting the ability to produce several staphylococcal enterotoxins 57 . These bacteria have also been reported to contribute to systemic infections during oral mucositis 58 . The representative species S. aureus caused 30 of 438 cases of bacteremia in neutropenic patients with cancer during a 10-year study period and septic metastases were more frequent in patients with S. aureus bacteremia, remaining as a significant cause of morbidity and mortality 59 . Therefore, our results imply a vital role for L. lactis in the prevention of Staphylococcus infection.
Our results suggest Clostridium was found depleted in mice submitted to 5-FU administration. The genus Clostridium comprises more than 200 species of bacteria in which some of these are pathogenic, but the majority is inoffensive. The representative pathogen is C. difficile, a Gram-positive bacteria that have been reported to be involved in IBD pathogenesis, but also in patients receiving antineoplastic chemotherapy 60 . Other Clostridiales bugs such as Dehalobacterium and unclassified Peptococcaceae was found overrepresented in mice injected with 5-FU. Although their role in the gut microbiome are unknown, the group of inflamed animals that consumed L. lactis NZ9000 culture restored the level of Peptococcacea and caused a reduction of the Lachnospiraceae population. Further studies are needed to provide possible clues about their biological implications. Another Clostridiales OTU which was suppressed by L. lactis consumption was Anaerotruncus, a recently described rod-like anaerobic bacterial genus belonging to Clostridiaceae family. The representative species is A. colihominis which have been isolated from human feces and associated with nosocomial bacteremia and to inflammatory traits in elderly subjects [61][62][63] . The expansion of Proteobacteria in the intestinal lumen, mainly Enterobacteriaceae has been consensually considered as a microbial signature of dysbiosis [64][65][66] . Among the Proteobacteria having increased numbers of mice receiving 5-FU, we identified three potential sulfate-reducing bacteria (SRB), Desulfovibrio spp., Bilophila spp. and unclassified Desulfovibronaceae. Increased levels of Desulfovibrionacea have also been found in ulcerative colitis [72,73] while Bilophila wadsworthia 67 have been isolated in clinical intestinal infections and bacteremia. Several studies suggest that SRB acquire sulfate by depolymerization and desulphation of host mucus glycoproteins such as mucins, which are secreted by goblet cells lining the gastrointestinal tract 68 . In this context, SRB might act in intestinal disorders by secreting metabolic end products such as hydrogen sulfide, which inhibits the production of SCFAs by other commensal bacteria and by promoting the formation of Reactive Oxygen Species (ROS) and Reactive Nitrogen Species (RNS) (Loubinoux et al., 2002). While ROS/ RNS may exacerbate the inflammatory process, they also serve as metabolic substrates for providing ATP for opportunistic Enterobacteriaceae in the gut 69 . The increase of Enterobacteriaceae family in the intestinal microbiota is associated with several intestinal disorders such as the Inflammatory Bowel Diseases and Colorectal Cancer. Moreover, recent studies reveal that the presence of Escherichia coli, the representative commensal species belonging to this family, can aggravate mucositis in mice 70 . Unexpectedly, in our study, feeding mice with L. lactis NZ9000 seems to favor the growth of this OTU in the inflamed mice. This result reiterate the importance of investigating the effects of probiotic strains in the gut microbiome as they may also stimulate the growth of SCIeNTIFIC REPoRts | (2018) 8:15072 | DOI:10.1038/s41598-018-33469-w undesirable bacteria. Previous studies have demonstrated that lactate, produced by LAB species can be used as an electron donor and may serve as a substrate for Enterobacteria 71 . In this context, as we did not observe significative augmentation of this OTU in healthy mice that also fed on L. lactis NZ9000 culture, we suggest lactate and the ROS/RNS generated by 5-FU activity in the mucosa might be acting synergically to provide fitness for Enterobacteriaceae 72 . In our previous work, we show that L. lactis NZ9000 producing PAP was able to preserve villous architecture of mice and increase Paneth cells activity in response to 5-FU inflammation 24 . Interestingly, in the present study our results show PAP delivery drastically inhibited the growth of the Enterobacteriaceae both in healthy and inflamed mice, suggesting a crucial protective role in the intestinal mucosa against the colonization of potential opportunistic Enterobacteria. Moreover, our study reinforces that PAP antimicrobial activity is not exclusively against Gram-positive bacteria. In an attempt to aggregate biological meaning to OTUs in which their role in the gut is not well established, we tried correlating the relative abundance with the metadata regarding inflammation markers that were assessed in our late work 24 . However, no significant correlation was obtained (data not shown).

Conclusions
This study was the first step in characterizing the effects of the L. lactis NZ9000 and PAP-secreting strain in the prevention of 5-FU-induced dysbiosis. We demonstrate that both Lactococci strains were able to prevent specific niches being occupied by microorganisms with potential implications in the prognostic of mucositis. We believe the data generated in the study will be of extreme importance for improving therapeutic strategies for treating the disease.

Methods
Bacterial strains and growth conditions. Lactococcus lactis NZ9000 strain harboring pSEC:PAP vector (LL-PAP) and L. lactis NZ9000 strain carrying pSEC vector without the open reading frame of PAP (LL), were grown in M17 medium (Difco) supplemented with 0.5% glucose (GM17) at 30 °C without shaking. The strains were selected by the addition of chloramphenicol (Cm, 10 μg/mL). For nisin-induced PAP expression, LL-PAP was cultivated until the optical density at 600 nm reached 0.6. Afterward, 10 ng/mL of nisin (Sigma) were added to the medium and cultures were maintained at 30 °C for 2 h. Immediately after incubation, bacterial cells were washed with saline solution by centrifugation at 12000 rpm for 10 minutes to eliminate residual antibiotic compounds. L. lactis NZ9000 or LL-PAP cells were then dissolved in M17 without the addition of antibiotics and transferred to feeding bottles before experimentation.
Animals and experimental treatment of the groups. Conventional female BALB/c mice between 6 and 8 weeks of age were obtained at Federal University of Minas Gerais (UFMG-Belo Horizonte, Brazil) and the Brazilian Ethics Committee on Animal Use (CEUA) approved the study. All mice were housed in cages in a controlled environment (23 °C, 12-/12-light/dark cycle with lighting), fed with standard chow diet, and provided with filtered water ad libitum before the experiment.
The animals were divided into six experimental groups (n = 4 in each group/cage), fed with standard chow diet and were administrated with 5 mL of filtered water or M17 medium containing 2.5 × 10 9 CFU/mL of the following of bacterial strains: L. lactis NZ9000 or L. lactis expressing PAP by continuous feeding for 13 days 73 . For the induction of mucositis, 300 mg/Kg of 5-Fluorouracil (Flaudfluor) was administered intraperitoneally to mice on day 10. All mice were euthanized on day 14, and stools samples were collected and kept at −80°C.
The first three groups consisted of noninflamed mice: (i) Control, injected with 0.9% saline on day 10 and daily administered with water; (ii) LL, fed with L. lactis NZ9000; and (iii) LL-PAP, fed with L. lactis expressing PAP. The following groups were composed by those mice with mucositis: (iv) 5-FU, receiving filtered water; (v) LL-5FU, fed with L. lactis NZ9000, and (vi) LL-PAP, fed with L. lactis expressing PAP. All experiments were done in three replicates, totalizing 12 animals per group. To determine the quality of metataxonomic method, two synthetic 16 S rRNA gene microbial communities (Mock Communities) of species with known genomes were used (Supplementary Document 1).
Bioinformatics analyses for taxonomic assignment. Fastq file with raw data of all barcodes (expects two barcodes with mock communities) were used in OTU classification pipeline derived from 16 S rRNA gene profiling data analysis of Brazilian Microbiome Project 76 . Briefly, the raw fastq file was processed to strip barcodes using Usearch package 77 . Then, quality filtering was performed including removal of truncated and low-quality sequences (Phred score smaller than 20). Next, sequences were submitted to dereplication, abundance sorting, singleton removal, OTU clustering (97% similarity), and chimera filtering using Vsearch 78 . Finally, pre-processed sequences were assigned taxonomically using QIIME requiring 97% of sequence similarity threshold against the Greengenes 13.8 database 79 . The two barcodes with mock communities were processed using the same steps.
Ecological analysis. The alpha diversity was estimated by richness, Shannon diversity and evenness index.
The diversity was estimated using Shannon (H′) index (H′ = −Σni/n ln (ni/n), where ni is the number of individuals in the taxon i and n is the total number of individuals), which is a heterogeneity index, influenced by both species richness and evenness. The evenness of species diversity was calculated using the Pielou formula: SCIeNTIFIC REPoRts | (2018) 8:15072 | DOI:10.1038/s41598-018-33469-w H′/H′max, where H′ = Shannon index and H′max = the possible maximum diversity of the number of species (S) present in the community, defined by the formula H′max = ln S. Rarefaction curves was performed to indicate if the sequencing depth was sufficient to wholly capture the diversity present using iNEXT package 80,81 .
Two different non-parametric analyses were used to determine the significance of differences among the groups: analysis of similarity (ANOSIM) and multiresponse permutation procedure (MRPP) 82 using Jaccard distance. A Bonferroni correction was applied to a p-value of 0.05 resulting in a significance level set at P = 0.0033. Ethics approval. The Protocol no. 366/2012, related to the present project is in agreement with the Ethical Principles in Animal Experimentation, adopted by the Ethics Committee in Animal Experimentation (CEUA/ UFMG), and was approved on 11/04/2013.