Commensal Clostridia in the preterm gut as reservoirs of antimicrobial resistance: susceptibility profiles, and resistance genes

Abstract

The gut microbiome of preterm infants is highly vulnerable to perturbations. Members of the class Clostridia are among the first anaerobes colonizing the preterm gut, yet their ecological roles and antimicrobial resistance (AMR) properties remain poorly understood. We characterized 98 Clostridia isolates from fecal samples of preterm infants, spanning 17 species and 11 genera. Isolates were identified by MALDI-TOF and 16S rRNA sequencing, colonization levels were quantified, and antimicrobial susceptibility was assessed by disk diffusion and E-test. Resistance determinants were screened by PCR and sequenced. We focused on Clostridia that were present at low colonization levels (mean 5.3 log₁₀ CFU g⁻¹ of feces). While most isolates were susceptible to amoxicillin–clavulanic acid, imipenem, and metronidazole, resistance to tetracycline (12%), clindamycin (35%), and cefotaxime (35%) was observed. Distinct species-specific resistance included linezolid (Clostridium argentinense), chloramphenicol (Clostridium innocuum), and tigecycline (Paeniclostridium sordellii), and one Robinsonella peoriensis isolate displayed vancomycin resistance. The detection of tet and erm genes corresponded with phenotypic resistance, while β-lactamase activity was uncommon. Although colonizing at low levels, these findings highlight the ecological significance of rarely studied commensal Clostridia and their contribution to the neonatal resistome, acting as underappreciated reservoirs of AMR genes during a critical window of microbiome assembly.

Rare gut clostridia in preterm infants are important reservoirs of antibiotic resistance, revealing how low-abundance bacteria shape early microbiome development.

Introduction

Anaerobic bacteria colonize the digestive tract of newborns early after delivery and have been isolated from various types of neonatal infections, including neonatal aspiration pneumonia, bacteremia, conjunctivitis, omphalitis, and necrotizing enterocolitis (Brook 2010, 2024). Members of the class Clostridia are spore-forming, obligate anaerobes with wide ecological distribution across soil, sediments, and animal hosts (Rainey et al. 2009). They are among the earliest anaerobes to colonize the infant gut, detectable within the first few weeks after birth (Milani et al. 2017). They persist until the age of 1 year in formula-fed infants but are largely underrepresented in breastfed infants after weaning (Albenberg and Wu 2014). Colonization patterns differ between infants and adults, with neonates showing a higher prevalence of cluster I species, while clusters IV and XIVa dominate later in life (Derrien et al. 2019). In preterm neonates, Clostridium perfringens, Clostridium butyricum, Clostridium paraputrificum, Clostridioides difficile, and Clostridium neonatale are frequently detected asymptomatically (Ferraris et al. 2012, Roze et al. 2017). However, anaerobes account for 1.8%–12.5% of neonatal cases of bacteremia, with Clostridium species—primarily C. perfringens—responsible for one-third of these (Brook 2010).^. Some species, including C. butyricum, C. perfringens, C. neonatale, and C paraputrificum, have been implicated in neonatal necrotizing enterocolitis (NEC), underscoring their clinical relevance during early life (Schonherr-Hellec and Aires 2019). C. butyricum has also been associated with rare cases of type E infant botulism (Fenicia et al. 2002, Abe et al. 2008, Shelley et al. 2015).

Despite their importance as early colonizers and occasional pathogens, many Clostridia remain poorly characterized in terms of their ecological roles and functional traits. Next-generation sequencing has provided broad insights into infant gut composition and resistome profiles (Bargheet et al. 2023, Trosvik et al. 2024), yet the phenotypic and genetic resistance properties of unusual Clostridia are underexplored. This gap is particularly relevant in preterm infants, where low-abundance taxa may act as hidden reservoirs of resistance genes with potential to influence microbial ecology and therapeutic outcomes.

Here, we investigated a collection of fecal isolates from preterm infants to examine the distribution, colonization levels, antimicrobial susceptibility, and genetic resistance determinants of rarely studied Clostridia. Our findings highlight the ecological significance of these taxa as underappreciated reservoirs.

Materials and methods

Bacteria isolation and identification

All strains included in the present study are part of our laboratory strain collection [Laboratoire de Microbiologie, U1139 (FRPM), Faculté de Pharmacie, Université Paris Cité, France]. Strains were isolated from the fecal samples obtained from preterm infants (aged 21–158 days; Table 1) who were hospitalized in 21 French NICUs and enrolled in previous cohorts conducted between 2003 and 2016 (Roze et al. 2017, Couturier et al. 2022, Aires et al. 2023). Strain isolation was performed as previously described (Ferraris et al. 2012). Briefly, fecal samples were homogenized in brain–heart infusion broth using an Ultra-Turrax T25 (Fisher-Bioblock, Illkirch, France), diluted in peptone water, and plated on sulfite-polymyxin-milk selective agar medium at 10⁻², 10⁻⁴, and 10⁻⁶ dilutions using a WASP apparatus (Don Whitley Scientific, UK). The media were then incubated for 48 h at 37 °C under anaerobic conditions (H₂: CO₂: N₂, 10 : 10 : 80, v/v/v) in an A35 anaerobic workstation (Don Whitley Scientific). Bacterial counts were expressed as colony-forming units (CFU) g⁻¹ of feces. Strains were conserved in brain–heart infusion medium containing 20% (v/v) glycerol at −80 °C. Species identification was performed using matrix-assisted laser desorption ionization-time of flight mass spectrometry (Microflex spectrometer, Bruker Daltonics S.A.) and PCR amplification and sequencing of the 16S rDNA gene, as previously described (Bouvet et al. 2014). All bacterial liquid cultures were performed in TGYH broth (tryptone 30 g L⁻¹, glucose 5 g L⁻¹, yeast extract 20 g L⁻¹, and hemin 5 mg L⁻¹) for 48 h at 37 °C under anaerobic conditions.

Table 1.

Clostridia isolates included in the study (n = 98).

Class	Family	Genus	Cluster classification (Collins et al  1994; Lawson and Rainey 2015)	Previous strain name	New strain name	N	Level of colonization Log10CFU g−1 of feces [min—max]	Mean age of infant (days)
Clostridia	Clostridiaceae	Clostridium	I	Clostridium tertium		11	5.4 [3,3–8,6]	140
				Clostridium argentinense		4	6.4 [5.3–7.5]	20
				Clostridium baratii		5	5.5 [4.0–6.2]	94
				Clostridium disporicum		2	5.8 [5.3–6.3]	226
			-	Clostridium saudiense		1	4.0	26
			XVI	Clostridium innocuum		29	4.5 [3.3–7.2]	229
		Hungatella	-	Clostridium hathewayi	Hungatella hathewayi	7	5.4 [3.6–7.3]	217
	Lachnospiraceae	Enterocloster	-	Clostridium aldenense	Enterocloster aldenensis	1	5.3	451
				Clostridium bolteae	Enterocloster bolteae	2	4.6 [4.0–5.3]	398
		Lacrimispora	XIVa	Clostridium celerecrescens	Lacrimispora celerecrescens	5	4.7 [3.3–7.0]	126
		Mediterraneibacter	-	Ruminococcus gnavus	Mediterraneibacter gnavus	13	6.4 [5.3–7.4]	310
		Robinsoniella	XIVa	Robinsoniella peoriensis		7	6.4 [4.0–8.7]	201
	Oscillospiraceae	Flavonifractor	III	Clostridium orbiscindens	Flavonifractor plautii	3	6.3 [5.3–7.5]	263
	Peptostreptococcaceae	Paeniclostridium	XI	Clostridium sordellii	Paeniclostridium sordellii	4	5.7 [3.8–7,5]	108
		Paraclostridium	XI	Clostridium bifermentans	Paraclostridium bifermentans	1	5,7	NA
		Terrisporobacter	XI	Clostridium glycolicum	Terrisporobacter glycolicus	2	3.9 [3.0–4.8]	130
Erysipelotrichia	Coprobacillaceae	Thomasclavelia	XVIII	Clostridium ramosum	Thomasclavelia ramosa	1	4.7	52

NA, not avalable.

Antimicrobial susceptibility

Antimicrobial susceptibility testing was performed using the disk diffusion method according to the EUCAST V1.1 recommendations on Brucella agar medium supplemented with 5% (v/v) sheep blood, 1 µg mL⁻¹ of vitamin K1, and 5 mg mL⁻¹ of hemin (EUCAST. 2025). The bacterial inoculum was 1.0 McFarland, and the plates were incubated in an anaerobic atmosphere at 35 ± 2C for 20–44 h. The antibiotic discs that were tested included amoxicillin, amoxicillin-clavulanic acid, piperacillin, piperacillin-tazobactam, ertapenem, imipenem, cefoxitin, tigecycline, chloramphenicol, moxifloxacin, metronidazole, linezolid, and vancomycin (bioMérieux, Marcy l’Etoile, France). The minimum inhibitory concentration (MIC) determinations for clindamycin, tetracycline, and cefotaxime were performed using the E-test strips according to the manufacturer’s instructions (bioMérieux). To enable a comparison of strain susceptibility levels, the breakpoints used in our study are those for anaerobic bacteria from the EUCAST V1.1 recommendations (EUCAST 2025), unless they are absent. In that case, the breakpoints are from the “CA-SFM—Comité de l’Antibiogramme de la Société Française de Microbiologie” 2013 (CA-SFM 2013G10). β-lactamase testing was detected in all amoxicillin-resistant and/or cefotaxime-resistant isolates using the nitrocefin chromogenic assay as recommended by the manufacturer (Sigma Aldrich Chimie, Saint-Quentin-Fallavier, France).

tet and erm genes

Genomic DNA was purified from 98 isolates using an Instagen kit (Bio-Rad) and used as a template for the PCR amplification of the tet(M), tet(W), tet(O), tet(Q), erm(B), and erm(Q) genes, as previously described (Ferraris et al. 2024). Sanger sequencing of the tet and erm amplicons was performed in both strands (GenWiz, Leipzig, Germany), and the sequences analyzed using the NCBI BLAST^® blastn suite (Altschul et al. 1997) (last accessed 07/01/2025).

Nucleotide sequences

Partial nucleotide sequences of tet and erm genes were deposited in the NCBI GenBank database: tet(O) accession number PX048391 and PX048392; tet(M) accession number PX048393; tet(W) accession number PV980114 and PX048398; tet(32) accession number PX048394; tet(O/32/O) accession number PV980115 and PV980116; erm(B) accession numbers PX048395 to PX048397; and erm(Q) accession numbers PV980117 to PV980119

Statistical analysis

XLSTAT version 2014.5.03 was used for statistical analysis. The Mann–Whitney U test was used to compare the observations between two groups. Hierarchical clustering was performed using R v4.3.2 to produce distance-based dendrograms and alluvial plot. Significance was set at P < 0.05.

Results and discussion

Isolate selection and colonization levels

From 461 fecal samples collected from 397 preterm infants hospitalized across 21 French NICUs (2003–2016), we recovered 452 non-redundant Clostridia. The most frequent species were C. difficile (29%, n = 115), C. butyricum (20%, n = 80), C. perfringens (19%, n = 74), C. neonatale (15%, n = 59), and C. paraputrificum (7%, n = 26), consistent with previous observations (Ferraris et al. 2012). For the present analysis, we focused on the less-studied non-dominant fraction of Clostidia, selecting 98 isolates representing 17 species and 11 genera (Table 1). These isolates were obtained from infants aged 21–158 days and exhibited low colonization densities (mean 5.3 log₁₀ CFU g⁻¹ of feces). Because our sampling overlaps recent taxonomic revisions (Lawson and Rainey 2015), Table 1 lists both current and historical names to avoid confusion. Many of these taxa are considered intestinal commensals yet can behave as opportunists in compromised hosts implicated in various infection cases including bacteremia, bloodstream, peritonitis, joint infections, traumatic wound infections, and extra-intestinal infections (Randazzo et al. 2015, Hall et al. 2017, Mormeneo Bayo et al. 2020, Cherny et al. 2021, Karpat et al. 2021, Milosavljevic et al. 2021, Wilton et al. 2022, Yoon et al. 2022, Cai et al. 2023, Martínez de Victoria Carazo et al. 2023, Ioannou et al. 2024). Information about these species is scarce in the pediatric population. Our results provide original data on these species as commensals of the gut microbiota of infants in early life. Their low abundance but broad trait diversity positions them as potential functional reservoirs during early microbiome assembly.

Disk-diffusion antibiotic susceptibility and resistance clustering among isolates

Comprehensive disk testing showed uniform susceptibility to amoxicillin–clavulanic acid, imipenem, and metronidazole across all isolates examined (Supplementary Table S1). Two isolates—Paraclostridium bifermentans and Clostridium saudiense—were susceptible to all tested antibiotics. Species-specific resistance phenotypes were observed: linezolid resistance in Clostridium argentinense; chloramphenicol resistance in Clostridium innocuum; and tigecycline resistance in Paeniclostridium sordellii. Notably, one R. peoriensis isolate exhibited vancomycin resistance. Additional findings included resistance in Lachnoclostridium celerecrescens to amoxicillin, piperacillin, and ertapenem; in Clostridium tertium to piperacillin (9/15), piperacillin–tazobactam (14/15), and ertapenem (1/15); and in P. sordellii and Mediterraneibacter gnavus to ertapenem (n = 1 and n = 2, respectively) and moxifloxacin (n = 4 and n = 8, respectively). Robinsonella peoriensis additionally showed moxifloxacin (n = 7) and cefoxitin (n = 1) resistance; single resistant isolates were observed for Faecalicatena plautii (cefoxitin) and Clostridium disporicum (moxifloxacin).

Hierarchical clustering of resistance profiles (Fig. 1) revealed two principal species clusters: one comprising broadly susceptible taxa and another containing species with resistance to moxifloxacin, cefoxitin, and ertapenem. Antibiotic clustering indicated resistance consistent with overlapping resistance. Vancomycin, chloramphenicol, metronidazole, and amoxicillin–clavulanic acid antibiotics formed a distinct, largely susceptible group. These patterns reinforce the species-specific nature of resistance and align with the phenotypic profiles obtained by disk testing. For several taxa observed intrinsic resistance matched established anaerobe susceptibility data (e.g. C. innocuum, Enterocloster boltae, Enterocloster aldenensis, Hungatella hathewayi, and Terrisporobacter ramosa) (Dubreuil et al. 2020, EUCAST 2025). Together, these findings expand the resistance information for lesser-studied anaerobes, highlighting both conserved susceptibility to key therapeutic agents and rare, species-specific phenotypes such as vancomycin-resistant R. peoriensis.

Figure 1.

Antibiotic resistance clustering across species. Heatmap shows the proportion of resistant isolates by species and antibiotic. Color scale indicates the proportion of resistant isolates. Hierarchical clustering (Euclidean distance and complete linkage) revealed two principal main species groups—one broadly susceptible and one with elevated resistance. AMO, amoxicillin; AMC, amoxicillin-clavulanic acid; PIL piperacillin; PTZ, piperacillin-tazobactam; ETP, ertapenem; IPM, imipenem; FOX, cefoxitin; CHL chloramphenicol; LZD, linezolid; MTR, metronidazole; MXF, moxifloxacin; TGC, tigecycline; VAN, vancomyin.

MIC distribution patterns across species

Given prior reports of tetracycline, clindamycin, and cefotaxime resistance in Clostridium spp. (Ferraris et al. 2012, Dubreuil et al. 2020), we determined MICs for all 98 isolates. Overall resistance rates were tetracycline 12% (n = 12), clindamycin 35% (n = 34), and cefotaxime 35% (n = 34). MIC ranges and MIC₉₀ values were tetracycline 0.125–48 mg L⁻¹ (MIC₉₀ = 48 mg L⁻¹), clindamycin 0.036–>256 mg L⁻¹ (MIC₉₀ >256 mg L⁻¹), and cefotaxime 0.125–>32 mg L⁻¹ (MIC₉₀ >32 mg L⁻¹) (Tables S3–S5). Species-specific resistance patterns were observed. For tetracycline, most isolates were susceptible; however, C. innocuum, E. aldenensis, F. plautii, L. celerecrescens, and P. sordellii displayed resistance (MIC > 8 mg L⁻¹) (Fig. 2A, Table S3). Cefotaxime resistance was widespread (Fig. 2B), with MIC > 32 mg L⁻¹ for C. innocuum, C. tertium, E. aldenensis, E. boltae, H. hathewayi, L. celerecrescens, R. peoriensis, and T. glycolicus (Table S4). In contrast, M. gnavus and P. sordellii exhibited comparatively low MICs. Clindamycin MICs were variable (Fig. 2C), with resistance (MIC > 4 mg L⁻¹) observed in C. baratii, C. innocuum, C. tertium, E. aldenensis, L. celerecrescens, P. sordellii, R. peoriensis, and T. ramosa  (Table S5). Taken together, these MIC distributions underscore pronounced lineage-specific differences in susceptibility, paralleling the resistance clustering observed in Fig. 1. This suggest that even taxa present at low abundance contribute to the overall resistome signal, reinforcing the ecological significance of species-level variation in anaerobic resistance profiles.

Figure 2.

MIC distributions for selected antibiotics by species. Boxplots show MIC distributions for (A) tetracycline, (B) cefotaxime, and (C) clindamycin. Dashed lines mark clinical resistance breakpoints. For cefotaxime, the number of strains with a MIC > 32 mg L⁻¹ is as follows: n = 10, C tertium; n = 1, E. aldenensis; n = 2, E. boltae; n = 7, H. hathewayi; n = 4, L. celerecrescens; and n = 6, R. peoriensis. Full MIC ranges values are provided in Supplementary Tables S2–S4.

β-lactamase production

Despite β-lactam resistance in several isolates, nitrocefin testing detected β-lactamase activity in only two isolates: E. boltae and L. celerecrescens. Previous work reported β-lactamase production in 65% of E. bolteae isolates (Warren et al. 2006). To our knowledge, β-lactamase in L. celerecrescens has not been previously described. Resistance by production of β-lactamases has been described for C. butyricum, C. perfringens, C. difficile, C. clostridioforme, T. ramosa, and rare strains of C. botulinum (Dubreuil et al. 2020). In this study, the paucity of nitrocefin-positive isolates suggests alternative β-lactam resistance mechanisms (e.g. altered penicillin binding proteins, decreased permeability) (Dubreuil and Odou 2010, Dubreuil et al. 2020, Oliveira Paiva et al. 2025).

Tetracycline resistance determinants (tet)

The presence of the tet genes, which encode tetracycline ribosomal protection proteins, is one a common genetic determinant associated with tetracycline resistance (Roberts 2003). In the present study, PCR detected tet genes in 24/98 (26%) isolates (Table S6). Of these isolates, 12 were tetracycline-resistant (MIC>8 mg/L), while 13 were susceptible (MICs ≤ 8 mg/L) (Table S3). These data suggest that tet genes confer different levels of tetracycline resistance. The MICs of the 12 tetracycline-susceptible isolates carrying a tet gene ranged from 0.25 to 8 mg L⁻¹. This range was 1.5 to 4 times higher than the MICs of the 73 tetracycline-susceptible isolates without tet (MICs ranging from 0.016 to 2 mg L⁻¹).

Sequencing of the 1200 bp PCR products of the 24 tet-positive isolates identified tet(M) (n = 8), tet(W) (n = 4), tet(O) (n = 8), tet(32) (n = 2), and tet(O/32/O) (n = 2) (Table S6). Among tetracycline resistant isolates (n = 12), 11 carried a tet gene by PCR: 6 (50%) tet(M), 3 (25%) tet(W); 1 (8%) tet(O), and 1 (8%) tet(O/32/O). Alternative resistance mechanisms may be involved for the remaining tetracycline-resistant strain (e.g. target-site mutations, efflux, or tet determinants not targeted by our primers) (Roberts 2003, Ferraris et al. 2012). Among tetracycline susceptible but PCR tet-positive isolates (n = 13), 7 (54%) carried tet(O), 2 (15%) carried tet(M), 2 (15%) carried tet(32), 1 carried (8%) tet(W), and 1 (8%) carried tet(O/32/O). Several genes were identified in species without prior records in public databases, extending the known distribution of these determinants. In terms of genotype–phenotype relationship, isolates carrying tet(M) or tet(W) showed significantly higher MICs than those with tet(O) (P = 0.008 and P = 0.039, respectively), indicating allele-specific functional impacts. Our results show that the presence of diverse tet families and hybrids across rare lineages supports multiple acquisition routes and variable resistance potential among early life gut Clostridia colonizers. The presence of tet in low-abundance taxa strengthens the case that they may act as reservoirs within the neonatal resistome.

Clindamycin resistance determinants (erm)

The presence of the erm genes encoding 23S rRNA methyltransferases has been reported to be frequently responsible for Clostridium spp macrolide resistance (Roberts 2003). In the present study, among 34 clindamycin-resistant isolates (MIC > 4 mg L⁻¹) (Table S5), 10 carried an erm(B) gene and 5 carried an erm(Q) gene (Table S6). Sequencing the PCR products confirmed the identification of erm(B) (711-bp) and erm(Q) (309- bp) genes (Table S6). As with tet, we identified erm genes in species not previously represented in databases. In agreement with previous reports (Roberts 2003, Ferraris et al. 2012), several clindamycin-resistant isolates lacked detectable erm genes, suggesting alternative mechanisms (e.g. target-site mutations, efflux, or erm determinants not targeted by our primers). As for tet, the coexistence of erm(B) and erm(Q) across diverse clostridial lineages indicates multiple evolutionary solutions to macrolide–lincosamide resistance in early-life communities.

Distribution of tetracycline and macrolide resistance genes among species

Clostridium innocuum showed the highest frequency of tet and erm genes carriage (17%), with tet(M) or erm(B) detected in 41% of strains (Fig. 3, Table S6). Dual tet/erm carriage occurred primarily in C. innocuum [tet(O/32/O) + erm(B)], L. celerecrescens (tet(W) + erm(B)), and M. gnavus (tet(O) + erm(Q)) consistent with their tetracycline and clindamycin MICs (Fig. 2, Tables S3 and S5). Hierarchical clustering of gene presence grouped taxa according to gene diversity, with tet(W) and tet(O), forming distinct subclusters (Fig. 3). The distribution pattern suggests both phylogenetic structuring and horizontal gene transfer events among closely related species. Overall, the genomic profiles parallel the MIC and resistance clustering data, confirming that tet and erm genes explain the observed species-specific resistance phenotypes. The coexistence of tet and erm genes across diverse clostridial lineages indicates multiple evolutionary solutions to resistance in these early-life communities.

Figure 3.

Distribution of tetracycline (tet) and macrolide (erm) resistance genes across species. Heatmap showing the proportion of isolates carrying each resistance gene (columns) by species (rows). Color scale indicates the proportion of positive strains per gene and the number of strains.

Association of resistance gene carriage with colonization levels, host age, and periods

Resistance gene carriage did not significantly influence bacteria colonization levels (P > 0.05) (Fig. 4A). However, the presence of tet genes was associated with host age. Individuals carrying tet genes were significantly older than those without resistance genes (P = 0.00076) (Fig. 4B), suggesting that resistance acquisition reflects transmission over time rather than differences in colonization efficiency. To further quantify this trend, we assessed the relationship between host age and overall resistance load of tet and erm genes per isolate. A significant positive correlation was observed (Pearson r = 0.37, P = 0.0001), indicating that the number of resistance genes per isolate increases with host age. This supports the hypothesis of gradual enrichment of resistance determinants in the gut microbiota over time. To explore temporal patterns, isolates were mapped onto the four sampling periods (Fig. 4C). Isolates without detectable tet or erm genes were identified in all periods and represented the largest proportion overall. Isolates carrying only tet genes or only erm genes were also recovered throughout the study periods. Those harboring both tet and erm genes were less frequent but likewise observed in multiple time periods. Together, these patterns suggest that the different resistance gene profiles have been present in the neonatal gut community over an extended time span, compatible with recurrent introductions and/or persistence.

Figure 4.

Relationship between resistance gene type, colonization level, and host age. (A) Colonization levels by resistance gene category: none, tet, erm, or both. No significant differences were observed among groups. (B) Host age by resistance gene type. Individuals carrying tet genes were significantly older than those without resistance genes. (C) Temporal distribution of resistance genes in strains. Alluvial plot shows the relationship between resistance gene category and sampling period. Each ribbon represents an isolate and is colored by sampling period.

Limitations

Our genotypic analysis focused on a restricted panel of tetracycline and macrolide resistance genes (tet(M), tet(W), tet(O), tet(32), tet(O/32/O), erm(B), and erm(Q)). As a result, less common or divergent determinants, as well as novel genes not targeted by our primers, may have gone undetected. This restricted panel probably contributes to the observation of phenotypically resistant isolates without the corresponding gene detected, as additional alleles or alternative mechanisms may be involved. Likewise, the genetic basis of linezolid resistance in the C. argentinense isolate and vancomycin resistance in the R. peoriensis isolate remains unknown, because we did not investigate cfr/cfr-like genes, van gene clusters, or other linezolid and vancomycin resistance determinants. Future studies combining WGS with detailed phenotypic testing will be essential to fully resolve the diversity and mobilization potential of resistance in these early-life Clostridium isolates.

Conclusion

Rare, low-abundance members of the class Clostridia colonizing the preterm gut carry a diverse set of antimicrobial resistance phenotypes and determinants despite modest densities. We document species-specific resistance and a vancomycin-resistant R. peoriensis isolate, while most taxa remained uniformly susceptible to amoxicillin–clavulanic acid, imipenem, and metronidazole. Genotypically, tet and erm families were widespread across uncommon clostridial lineages, with allele-specific effects on MICs, and β-lactamase activity was rare but included, to our knowledge, a first report in L. celerecrescens. The recurrence of similar tet/erm gene profiles across distinct sampling periods suggests that these resistance reservoirs are maintained over time rather than arising from a single transient cohort. This is of interest in the context of the increasing prevalence of antimicrobial resistance in anaerobes (Boyanova et al. 2015), as well as the potential for antibiotic resistance to be transferred.

Together, these findings indicate that rarely studied Clostridia function as underappreciated antimicrobial resistance reservoirs during a critical window of microbiome assembly in preterm infants. Incorporating such low-abundance taxa into ecological models of resistome development and into NICU surveillance frameworks should improve predictions of community resilience and inform antimicrobial stewardship (e.g. caution with agents showing ≥35% resistance such as clindamycin and cefotaxime). Future work should place these determinants in their genomic context, track persistence and transmission longitudinally, and test transfer potential experimentally to quantify their contribution to early-life antimicrobial resistance dynamics. Recognizing and monitoring these commensal reservoirs will sharpen both ecological understanding and clinical risk assessment in vulnerable neonatal populations.

Author contributions

Johanne Delannoy (Data curation, Formal analysis, Visualization, Writing - original draft, Writing—review & editing), Laurent Ferraris (Methodology, Validation, Visualization, Writing—review & editing), Chantal Labellie (Investigation, Validation, Visualization), Léa Dupire (Methodology, Resources, Validation), Denissa Ilavska (Methodology, Resources, Visualization), Marie-José Butel (Formal analysis, Validation, Writing—review & editing), Frédéric Barbut (Formal analysis, Visualization, Writing—review & editing), and Julio Aires (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing).

Conflict of interest

None declared.

Funding

No funding.