Background

The human gut microbiota (HGM) consists of an astonishingly large number of phylogenetically diverse bacterial species. The healthy HGM has been extensively investigated notably by large research consortia across the globe, such as the Human Intestinal Tract (MetaHIT) [1] and the US Human Microbiome Project (HMP) [2]. Besides its core function that is to assimilate nutrients, such as plant carbohydrates [3] and glycans [4], the HGM plays other relevant biological functions for the human host, including bile acid metabolism [5], biosynthesis of short-chain fatty acids (SFCA) [6], biosynthesis of vitamins [5], immunomodulation [7, 8], and protection against pathogens [9, 10]. The HGM is constantly shaped by environmental or other external factors [11, 12] and host factors [13] and it changes through the different stages of life [14,15,16,17]. Deleterious changes in the HGM composition have been associated with some gastrointestinal disorders and diseases [18, 19] as well as systemic diseases [20], illustrating the intricate link between human health and HGM. This has led to major research efforts in developing biotherapeutic agents, such as prebiotics and health-promoting bacteria (probiotics), to re-shape dysbiotic HGM [21,22,23,24].

The infant HGM is unstable and has low microbial richness and diversity compared to the adult HGM [15, 16, 25]. In healthy adults, the gut is typically and predominantly colonized by two microbial phyla: Bacteroidetes and Firmicutes [26], whereas the infant gut is initially more abundant in Actinobacteria [27]. Underneath this over-simplified snapshot lie major temporal changes and events in microbial colonization, diversity, and dynamics in the infant gut [15, 16]. Microbial colonization and composition of the infant gut are influenced by multiple external factors, such as birth delivery mode [28, 29], antibiotic treatments [30, 31] or diet [32].

Infant botulism results from ingestion of spores of neurotoxinogenic Clostridia and subsequent spore germination into active neurotoxinogenic cultures in the gut [33, 34]. It is generally considered that small infants are susceptible to colonization because their immature gut microbiota are unable to outcompete neurotoxinogenic Clostridia, such as Clostridium botulinum, Clostridium butyricum and Clostridium baratii. In addition, other factors like bile acids or probiotic micro-organisms may impact spore germination, growth, toxin production or toxin potency of neurotoxinogenic Clostridia [35, 36]. Typically, most infant botulism cases are diagnosed in infants of less than 6 months of age [37]. Upon colonization of, presumably, the lumen of the large intestine by BoNT-producing Clostridia [38, 39], BoNT is produced in situ and intoxicates the host, resulting in flaccid paralysis known as botulism [40]. In adults, a similar condition (toxicoinfectious botulism) may be preceded by intestinal surgery, intestinal disorders, or antimicrobial treatment [41,42,43], all assumed to re-shape the gut microbiome and to provide a competitive edge for C. botulinum in a manner analogous to Clostridioides difficile infection. Clinical presentation, diagnosis, treatment, and prognosis of infant botulism have been well documented over the years [40, 44]. There are a number of well-established etiological factors that increase the risk of developing infant botulism, including diet (consumption of honey that may contain clostridial spores) [45, 46], constipation [47], environment (excess of dust particles) [45, 48] and geographical location (environmental spore load) [33, 47, 49].

Laboratory confirmation of infant botulism typically relies on detection of BoNT and/or isolation of neurotoxinogenic Clostridia from stool samples [44, 50]. The development of next-generation sequencing technologies has been instrumental in analyzing microbial composition of stool samples without the need to isolate and cultivate gut microbes. It provided significant insights into the association of the HGM with a number of diseases and disorders, such as metabolic disorders [51, 52], intestinal diseases [53, 54], and cancer [55, 56]. There is, however, scarce information on the role or impact of the HGM in the context of infant botulism. A single piece of work reported, among others, a lower relative abundance ratio of Firmicutes/Proteobacteria, and a higher relative abundance of Enterobacteriaceae in the fecal microbiota of infants diagnosed with botulism compared to the fecal microbiota of healthy infants [57]. There are no studies where the fecal microbiota of infants with botulism were followed over time to define the gut microbial context allowing C. botulinum to colonize, transiently persist, and be cleared from the infant gut. Understanding the relationship and dynamics between C. botulinum and gut microbiota gives novel insight into the ecology and epidemiology of infant botulism and provides novel perspectives to its prevention and treatment.

We earlier reported a botulism case of 3-month old infant and showed that, despite prompt clinical recovery, C. botulinum (Group I) type A persisted and produced BoNT in the infant gut for 27.7 weeks [58]. In contrast, in most infant botulism cases caused by C. botulinum type A, the median excretion of C. botulinum and the toxin was reported to be 5.9 weeks [59]. To further understand the exceptionally long colonization of the infant gut by C. botulinum in this particular case, we performed an extensive genomic analysis of C. botulinum isolates collected from the infant feces over time and identified possible pheno-genotypic adaptation traits of C. botulinum to the gut environment [60]. In the present work, we complemented our understanding of this infant botulism case by looking at temporal changes in the microbial signatures of the infant gut microbiota composition during the persistence and the clearance of C. botulinum from the infant gut. Based on well-established phylogenetic analytical tools, we observed a delay in the maturation of the infant fecal microbiota over time associated with a persistently high-level bifidobacterial population, and identified bacterial species possibly linked with the clearance of C. botulinum from the infant gut. To our knowledge, this is the first report where the gut microbiota composition of an infant diagnosed with infant botulism is examined temporally throughout various stages of the disease. It brings novel insights into the microbial ecological factors that may trigger intestinal botulism. While this case report depicts a single case, thus individual deviations in microbiota composition in other patients are likely, our report suggests that systematic large-scale approaches may help to identify patterns within the gut microbiota composition of infants associated with intestinal botulism, possibly leading to the development of preventive and therapeutic measures.

Materials and methods

Fecal samples

We analyzed 10 fecal samples collected over a period of 259 days from a case of infant botulism in Finland [58]. The infant was admitted to hospital at the age of 102 days and discharged at the age of 154 days. The first stool sample (infant age 122 days) related to a clinical stage when the infant displayed severe symptoms of botulism, whereas the very last sample (infant age 380 days) related to a stage where the infant had clinically fully recovered, had been discharged from hospital for more than 200 days, and tested negative for the presence of C. botulinum vegetative cells and spores and negative for BoNT in the feces [58]. All fecal samples analyzed in this work were collected after the infant had received both antibiotic treatment (ceftriaxone at infant age of 103–106 days) and antiviral medication (acyclovir at 102–108 days) (Fig. 1). The first three samples (infant age 122, 143 and 147 days) were collected at the hospital. All subsequent samples were collected at home, picked up and transported to the laboratory by a member of the lab and then stored in freezers. Informed consent was obtained from the patient’s parents.

Fig. 1
figure 1

Overview of the fecal sampling related to the presented infant botulism case. Only the first three fecal samples were collected at the hospital. Subsequent samples were collected at home by the parents. The infant age is indicated in brackets. FS, fecal sample

Full size image

DNA extraction, library preparation and sequencing

Fecal bacterial genomic DNA was extracted using the Quick-DNA Fecal/Soil Microbe Kits (Zymo Research, CA, USA) as per manufacturer’s instructions. DNA was quantified using NanoDrop™ 2000 Spectrophotometer (ThermoFischer Scientific, MA, USA) and stored at -20 °C. The hypervariable V3-V4 regions of the 16S rRNA gene were amplified using primers 341 F/758R and were further processed for library preparation using a modified protocol by Illumina and sequenced by Illumina HiSeq 2500 sequencer using HiSeq Rapid SBS Kit v2 (2 × 250 bases), as previously described [61,62,63]. Paired-end read sequencing using the Illumina MiSeq platform was performed at the Institute for Molecular Medicine Finland (FIMM, University of Helsinki, Finland).

Bioinformatics and statistical data analysis

Demultiplexed reads after adaptor removal by cutadapt [64] were processed using DADA2 [65], where truncLenF and truncLenR were set to 270 and 230, respectively, and reads with a number of expected errors higher than two were discarded. The forward and reversed reads were subsequently merged with a minimum overlap of 25 nucleotides to generate amplicon sequence variants (ASVs). Taxonomy was assigned to all ASVs using a pre-trained naïve Bayes classifier implemented in DADA2 (assignTaxonomy function with default settings) against the SILVA 138 reference database [66]. Species assignment was performed using DADA2 by exact string matching (addSpecies function with the argument “allowMultiple = FALSE”) against the SILVA v138.1 species assignment training database [67]. Of note, the ASV belonging to Clostridium botulinum was annotated as Clostridium sensu stricto 18 using the SILVA 138 database (Table S2), as confirmed by NCBI BLAST [68]. Principal coordinate analysis (PCoA) plot based on the Bray-Curtis dissimilarity was employed to visualize the differences in overall microbiota composition (β-diversity) between sampling points. Statistical significance of the difference in microbiota β-diversity between the hospital and home phases was tested using permutational multivariate analysis of variance (PERMANOVA; adonis2 function in the vegan package [69] with 999 permutations based on the Bray-Curtis dissimilarity). Microbiota α-diversity (observed richness, Shannon and inverse Simpson diversity indices) was estimated using the vegan package [69]. Statistical significance of the difference in microbiota α-diversity between the hospital and home phases was tested by calculating Tau-U, a non-overlap index designed for analysis of single-case research data [70]. P-values < 0.05 were considered significant for the analyses of microbiota α- and β-diversity. Given the single-case nature of this study compounded by the volatility of individual microbial taxa, we opted for visual and/or descriptive analysis for changes in specific microbial taxa over time to provide high-granularity information.

Results and discussion

Fecal microbiota composition during the course of infant botulism

We analyzed the microbial composition of 10 stool samples collected from an infant botulism case [58] over a period of 7 months, covering different stages of the disease. The metrics and statistics related to the 16S rRNA gene amplicon sequencing of the 10 samples are presented in Table S1. Additional metadata related to the fecal samples are available in our previous work [58]. The overall microbiota composition and within-sample diversity fluctuated over time, indicating that major changes occurred in the infant fecal microbiota composition during and after the course of the disease (Fig. 2). The overall microbiota compositions were similar in the fecal samples collected when the infant was symptomatic and treated at the hospital (102–154 days), as reflected visually in the PCoA plot showing clustering of the first three samples (PERMANOVA p = 0.006, 61% microbiota variation explained by hospital versus home; Fig. 2A). Microbiota α-diversity (observed richness, the Shannon and inverse Simpson diversity indices) was significantly lower during the hospital phase compared to the home phase (all Tau-U = 1, p = 0.017; Fig. 2B-D). This may be explained at least partly by the early antibiotic treatments, ceased oral food intake, and/or a more controlled environmental microbial load during the hospital phase compared to the home phase. In the absence of samples taken prior to the hospital stay, it remains unclear to what extent these factors impacted the infant gut microbiota. Interestingly, fecal sample 2 (infant age 143 days) appeared to be the “tipping point” in terms of diversity indices (lowest Shannon diversity index and inverse Simpson diversity index). Such tipping points have been described in adult microbiota in reflecting critical transitions with profound health implications [71].

Fig. 2
figure 2

Overview of the fecal microbiota composition of the infant stool samples over time. (A) Principal coordinate analysis (PCoA) plot of microbiota variation based on the Bray-Curtis dissimilarity matrix. Blue, samples collected at hospital; red, samples collected in the infant’s home. For each sample, the infant age (in days) was indicated as numbers. Sample collection phase (hospital vs. home) explained 61% of the microbiota variation (PERMANOVA p = 0.006). (B.-D) Microbiota α-diversity of the samples was significantly lower during hospitalization (all Tau-U = 1, p = 0.017) according to observed richness (B), Shannon diversity index (C), and inverse Simpson diversity index (D)

Full size image

The phylum-level composition of the infant fecal microbiota (Fig. 3) revealed that the phylum with the highest abundance across all samples was Actinobacteria, consisting of Bifidobacteriaceae and Eggerthellaceae. At the order level, Bacteroidales, Bifidobacteriales, Clostridiales, Coriobacteriales, Enterobacteriales, Lachnospirales, Lactobacillales, Oscillospirales and Peptostreptococcales were detected at all times in the infant gut, whereas other orders were intermittently present. We looked at the relative abundance of different families over time in the infant feces and compared to healthy infants with a normal gut microbiota development [15, 72, 73]. While the average relative abundance of Bifidobacteriaceae in healthy infants usually decreased to 10–20% by the end of the first year in life [72, 73], Bifidobacteriaceae remained at a high level over the age of one year in the infant botulism patient described (86.4%, infant age 380 days) (Fig. 3 and S1).

Fig. 3
figure 3

Family-level microbial composition of the infant fecal samples over time. All families shown here were detected at all times, except Bacteroidaceae, Veillonellaceae and Akkermansiaceae. Detection of C. botulinum by 16S rRNA gene amplicon sequencing is further detailed in Table 1. All samples were collected after antiviral and antibiotic treatment (administered at infant age 103–108 days). The lower part of the figure shows information related to the detection of C. botulinum in the fecal samples based on direct isolation of C. botulinum, PCR detection of bont/A and BoNT detection by mouse bioassay in the different samples and were published earlier [58]. Green dot, negative; orange dot, inconclusive; red dot, positive; white dot, not tested; grey dot, unspecific symptoms

Full size image

The relative abundance of Lactobacillaceae steadily decreased during the first year of life in our infant (Figure S1), as observed also in healthy infants [15, 73]. During the same period of time in our patient, however, Bacteroidaceae remained at low levels in all fecal samples (at most 3.21%), while in the healthy infant gut Bacteroidaceae established a larger population over time (an average of 15% of Bacteroidaceae in a Finnish infant cohort over the first year) [74]. In the present case, Lachnospiraceae and Ruminococcaceae marginally increased over time (Figure S1). Typically, Bacteroidaceae, Lachnospiraceae, and Ruminococcaceae outnumbered Bifidobacteriaceae by the age of 1 year in healthy infants [15, 73], which clearly contrasts with the fecal microbiota of the infant analyzed in our study. While levels of bifidobacteria were generally seen to decline over the first year of life in many studies, a recent large multicenter study suggested wide variation with some infants showing bifidobacteria-dominant enterotypes for longer periods, illustrating that there is a wide individual diversity [75]. Of note, Verrucomicrobiales that include Akkermansia sp. showed their highest levels (3.76%) in the last fecal sample collected from our infant, recovered from botulism by the sampling time. The genus Akkermansia serves as one of the biomarkers indicating a healthy gut condition (healthy mucus layer) [76]. Whether the emergence of Akkermansia sp. promoted clearance of C. botulinum from the infant gut, or developed as a consequence, remains to be understood. Overall, the data suggest that the maturation and development of the fecal microbiota was largely delayed in our patient with infant botulism, with a persistently high level of Bifidobacteriaceae over time in relation to healthy infants.

Clostridium botulinum persists at low levels in feces in the described case of infant botulism

Depending on the method used for the detection of C. botulinum in the infant feces, i.e. detection of bont/A by PCR, direct isolation of C. botulinum [58], or 16S rRNA gene amplicon sequencing (this study), 4 to 7 fecal samples were positive for C. botulinum (Fig. 3). Solely based on 16S rRNA gene amplicon sequencing (this work), the relative abundance of C. botulinum in the infant fecal microbiota was at its highest 0.023% in fecal sample 2 (infant age 143 days) and decreased over time, until being under the detection level (Table 1). This is in line with a previous study showing the relative abundance of C. botulinum in the feces of infants diagnosed with botulism to be < 0.001 to 0.01% [57]. These values indicate that neurotoxinogenic C. botulinum represents, at most, a marginal fraction of the gut microbiota in confirmed botulism cases. Yet, this appears sufficient for host intoxication and systemic paralysis, obviously due to the extremely high potency of BoNT. This illustrates how a very-low-abundant organism present in the gut can still have a consequent impact on the host health, so merely focusing on the most abundant taxa may be insufficient.

Table 1 The relative abundance (%) of C. botulinum 16S rRNA gene amplicon sequencing reads in the infant fecal samples over time
Full size table

As reported earlier [58], from the age of 154 to 245 days, the infant continued to excrete both neurotoxinogenic C. botulinum and BoNT, while not displaying clinical symptoms. This suggests that the infant may have developed mucosal immunity against BoNT over time, similarly to mucosal vaccines shown to prevent mucosal BoNT intoxication [77, 78]. From the age of 295 to 316 days, the clinically fully recovered infant may have excreted non-toxinogenic C. botulinum based on C. botulinum detection, isolation, and toxicity analysis by the mouse bioassay [58]. In addition to possible mucosal immunity to BoNT, the absence of clinical symptoms in this period of time could be explained by the pheno-genotype of the C. botulinum population evolving over time. Indeed, whole-genome sequencing of late C. botulinum stool isolates of the current case revealed the presence of multiple mutations in genes coding for the agr-2 quorum sensing system [60]. The agr-2 signaling system modulates neurotoxin production in C. botulinum strain ATCC 3502 in vitro [79], thus it is possible that a C. botulinum population with an impaired agr-2 quorum-sensing system remained in the infant gut and did not produce BoNT in the gut conditions. Interestingly, we also detected the presence of C. difficile in all fecal samples (up to 6.58%). Co-occurrence of the two species has been previously reported in other intestinal botulism cases [80,81,82]. It remains unclear if the higher relative abundance of C. difficile during the hospital phase than at home phase was due to nosocomial infection or contributed to the onset or course of infant botulism.

The low abundance of C. botulinum in botulism-confirmed samples may introduce a diagnostic challenge for 16S rRNA amplicon sequencing due to borderline detection sensitivity when investigating infant (intestinal) botulism cases. This warrants a dual approach where DNA-based detection of C. botulinum from fecal DNA samples could be conducted by using both 16S RNA amplicon sequencing (comprehensive analysis of the fecal microbiota composition) and real-time PCR (C. botulinum detection and diagnosis) in parallel (Fig. 3). Likely, the timing of fecal sampling is also critical, and C. botulinum appears to be more likely detected by 16S rRNA gene amplicon sequencing upon disease onset and at the time when symptoms are the most prominent. For longitudinal studies, therefore, multiple detection techniques should be preferred.

Bifidobacterial population structurally changed over time but remained at high level

The healthy adult gut is typically and predominantly colonized by the microbial phyla Bacteroidetes and Firmicutes [26], whereas the infant gut is more abundant with Actinobacteria [27]. Within Actinobacteria, Bifidobacteria have been shown to colonize and persist in the infant gut microbiota [83] and to play an important role in the development and maturation of the gut microbiota. They also have a protective role in the intestinal barrier function and contribute to immuno-modulation [84]. Bifidobacterium longum, Bifidobacterium breve, and Bifidobacterium bifidum are among the most prevalent species in the infant gut [27, 85, 86]. To understand their possible roles in infant botulism, we particularly examined the bifidobacterial populations present in the infant feces over time (Fig. 4). Overall, the bifidobacterial population of the infant gut remained at high levels for a longer time than observed in healthy infants [15]. Specifically, during the first year of life, the diet change from milk to solid and diverse food typically prompts a decline in Bifidobacteriaceae and Lactobacillaceae and an increase in Lachnospiraceae, Bacteroidaceae and Ruminococcaceae [15]. As indicated above, the persistence of a high bifidobacterial population in the infant gut suggests a delayed maturation of the infant fecal microbiota. The possible role of delayed maturation of the gut microbiota in the development, course, and recovery of infant botulism, and the relevance of early antibiotic treatments therein, warrants further investigation. On the same line, it remains unclear whether C. botulinum, BoNT, or host factors may have impacted the gut microbiota over time.

Fig. 4
figure 4

Relative abundance of the bifidobacteria population in the infant fecal samples over time. The cumulated relative abundance of B. breve and B. bifidum for each sample is indicated above the plot

Full size image

In terms of population structure, the bifidobacterial population in the diseased infant consisted of only few bacterial species as described for healthy infants [27]. While one bifidobacterial species dominated this population in the early fecal samples (Fig. 4), others such as B. breve and B. bifidum colonized the infant gut at a later stage in parallel with gradual clearance of toxinogenic colonization (Fig. 4). With existing data it is unclear if the colonization of the gut by B. breve and B. bifidum contributed to clearance of infant botulism, or if clearance of the toxinogenic colonization allowed B. breve and B. bifidum to thrive, and if and how other host or environmental factors may be involved. Since some strains of Bifidobacteria have been shown to inhibit the growth of C. botulinum in vitro [87], it is tempting to speculate that some Bifidobacteria species or strains may confer a protective effect against infant botulism by preventing the germination of C. botulinum spores and their colonization of the gut. Animal model-based assays will provide further evidence to support this hypothesis.

Concluding remarks

Here we aimed at further understanding of factors underlying an extremely long-term botulinum neurotoxinogenic colonization and different stages of infant botulism by looking at the infant gut microbiota composition over the course of the disease. We observed a delayed maturation of the infant gut microbiota with a persistently high bifidobacterial population and a low level of Lachnospiraceae, Bacteroidaceae and Ruminococcaceae over time compared to healthy infants where Bifidobacteriaceae usually decreased to a relative abundance of 10–20%, and Bacteroidaceae, Lachnospiraceae, and Ruminococcaceae outnumbered Bifidobacteriaceae by the age of one year. We suggest that the delay in the maturation of the infant gut microbiota may have explained the exceptionally long colonization and excretion of C. botulinum and BoNT in the infant gut. The relative abundance and population structure of Bifidobacteria is likely to play a central role during the course of toxinogenic colonization in infant botulism. B. breve and B. bifidum appeared to be temporally associated in the clearance of C. botulinum, highlighting the protective role of some bifidobacterial species against pathogens. Yet, it remains to be elucidated if and how B. breve and B. bifidum may interact with C. botulinum and if host or environmental factors are concomitantly involved. To conclude, this work provides valuable insights into the microbiota changes occurring during and after C. botulinum colonization in the infant gut. It is likely that other infant fecal microbiota composition signatures are also associated with infant botulism. Therefore, we advocate a more systematic analysis of the gut microbiota of infants diagnosed with botulism in an effort to further identify recurrent bacterial signatures associated with botulism and to develop prophylactic strategies and measures to prevent this severe disease in small children.