Introduction

Horses have their specific bacterial pathogens, such as Streptococcus equi subspecies equi and Taylorella equigenitalis, responsible for strangles and contagious equine metritis, respectively [1]. Some infections can be caused by pathogens found also in human medicine, such as Pseudomonas aeruginosa and Enterobacteriaceae, in particular those resistant and/or virulent ones [2–5]. Enterobacter spp. is a ubiquitous Gram-negative bacillus, found in a variety of human clinical infections, especially in intensive care unit, on land and water, as well as in insects and animals; it is also considered as a phytopathogen [6]. The Enterobacter genus, firstly described in 1960, comprises 22 species, including six that form the Enterobacter cloacae complex (ECC)—E. cloacae, E. asburiae, E. hormaechei, E. kobei, E. ludwigii, E. mori, and E. nimipressuralis [7]—which are the most commonly found in human infections [8]. The six species of the ECC share at least 60% genomic homology and are divided into 12 clusters (based on average nucleotide identity analysis and hsp60 sequencing) and 18 clades [7–9]. Advances in sequencing technology have improved the resolution of Enterobacter species identification and led to reclassification of the genera Lelliottia spp., Pluralibacter spp., Kosakonia spp. and Cronobacter spp. formerly included in the Enterobacter genus, based on multilocus sequence analysis (MLSA) [9].

A study conducted in France from 2006 to 2016 showed that Enterobacter spp. accounted for 3.4% of equine bacterial infections, and was found mainly in pulmonary infections [3]. This study was continued from 2016 to 2019, and revealed that Enterobacter was found in 2.1% of samples suspected of causing infection in horses [10], suggesting that the prevalence of Enterobacter infection is stable over time and is not frequently found in equine infections.

Due to the expansion of antibiotic resistance, surveillance of circulating multidrug resistance (MDR) strains in human and veterinary medicine is crucial, especially in the equine sector, where humans and animals are in close relationship. Enterobacter are opportunistic pathogens that can display an MDR phenotype in particular through its ability to acquire mobile genetic elements and its adaptation to different environments [7]. Additional resistances can be acquired by plasmid encoding β-lactamases genes such as tem, shv and ctx-m [7]. Strains can also be resistant to aminoglycoisides, via the synthesis of drug-metabolizing enzymes such as N-6’-acetyltransferase-ln (aac(6’)-Ib) or aac(3)-IIa and ant(2′′)-Ia, which confer resistance to amikacin and gentamycin [11]. Genes such as aac(6’)-Ib-cr and qnr confer fluoroquinolone resistance by inactivating the antibiotic or protecting DNA gyrase from its action, respectively [11]. Chromosomal mutations can also be the cause of antimicrobial resistance. Enterobacter are naturally resistant to β-lactams notably penicillins and 1^st generation cephalosporin by producing the chromosomal encoded cephalosporinase ampC [12]. AmpC overexpression phenomena are responsible for increased resistance levels, notably through mutations in its repressor AmpD or in the regulator AmpR, which can convert it into a constitutive activator [13]. Enterobacter also modulates membrane permeability by modifying porin expression (OmpC) and antibiotic efflux pumps like AcrAB-TolC [7]. From 2006 to 2019, in France, a study observed some strains of Enterobacter spp. from equine infections resistant to streptomycin (32–64% of strains), tetracycline (37–56%) or amikacin (< 10%). In 2016, 26% of the strains were resistant to at least three classes of antibiotics (aminoglycosides, tetracycline and sulfonamide), a percentage that rises to 52% in 2019 [3,10].

Within the ECC, strains are capable of forming robust biofilms, particularly under nutrient-deficient conditions [14]. This ability is linked to their capacity to adhere to surfaces through the production of type I or III fimbriae (encoded by the fimA and fimH, or mrkB genes, respectively), as well as curli fibers (encoded by csgA, csgB, and csgD) [15]. Enterobacter spp. also possess a type VI secretion system (T6SS), which contributes to adhesion to eukaryotic cells during infection and mediates interbacterial communication within the biofilm [16]. In addition, Enterobacter species are motile via flagella-mediated swimming. Their ability to utilize siderophores also makes them highly competitive in nutrient-limited environments.

The aim of this study was to characterize the antimicrobial resistance profiles and the genomic characteristic of a collection of Enterobacter spp. found in equine necropsy specimens and suspected to be associated with the death. The genomic population structure, virulome, resistome and antimicrobial susceptibility profiles were investigated, revealing the presence of the genera Enterobacter spp., Lelliottia spp., Huaxiibacter spp., and Rahnella spp. among the necropsy samples. Notably, two E. hormaechei strains (ST114 and ST171), considered high-risk clones in human clinical medicine, were identified.

Materials and methods

Results

Species identification

Thirty-four strains have been previously characterized by the ANSES laboratory as Enterobacter spp., using phenotypic methods such as Gram staining (Gram-negative bacilli) and API20E galleries. From the CryoBeads storage tubes containing the isolates, strains were cultured using CPSE agar plates to ensure the purity of the samples (S1 Fig). Strains exhibiting turquoise-blue colonies characteristic of Enterobacter spp. were initially identified using MALDI-TOF as first-line identification method, which revealed the presence of Enterobacter spp., Lelliottia spp., Rahnella spp. (previously classified under Enterobacter spp.).

For 9/34 strains (26.5%) MALDI-TOF failed to identify the species and the genus. About half of the strains (15/34) were characterized as members of the Enterobacter genus, including E. cloacae complex (n = 5), E. ludwigii (n = 5), E. cancerogenus (n = 2), E. asburiae (n = 1), E. hormaechei (n = 1), and E. bugandensis (n = 1). A quarter of the strains were characterized as belonging to the genus Lelliottia (8/34), with two species which belonged to thisgenus: L. amnigena (7/34) and L. nimipressuralis (1/34). MALDI-TOF also identified two Rahnella aqualitis strains.

Due to the limitations of MALDI-TOF in species-level identification (score < 2), all strains were subsequently subjected to whole-genome sequencing to achieve accurate genomic identification. Strain identification using the rMLST allowed species-level assignment for all strains. Whereas MALDI-TOF identifies nine species types, rMLST identifies six species. Fig 1 and S4 Table. shows MALDI-TOF and sequencing identification of the 34 Enterobacter or formerly Enterobacter. Both methods showed 26.5% common identifications (9/34): five E. ludwigii, three L. amnigena and one E. hormaechei. Two Huaxiibacter chinensis were only identify using the rMLST approach (Fig 1). For the genus Enterobacter, MALDI-TOF identified six species (E. cancerogenus, E. asburiae, E. cloacae complex, E. hormaechei, E. ludwigii and E. bugandensis), while the genomic approach revealed the presence of two species (E. hormaechei and E. ludwigii). Regarding the nine strains that could not be identified by MALDI-TOF, rMLST revealed that four belonged to the genus Lelliottia and five to the genus Enterobacter. As for the other scenarios observed: i) among the seven strains of L. amnigena identified by MALDI-TOF, three strains were confirmed by rMLST but four were characterized as L. nimipressuralis; ii) two strains identified as Rahnella aqualitis by MALDI-TOF were classified as Rahnella asceris by rMLST; and iii) for the genus Enterobacter, MALDI-TOF identified six species: E. cancerogenus, E. asburiae, E. cloacae complex, E. hormaechei, E. ludwigii, and E. bugandensis, while the rMLST revealed only two species: E. hormaechei (n = 5) and E. ludwigii (n = 13). The five ECC strains identified by MALDI-TOF corresponded to three E. hormaechei and two E. ludwigii strains, as confirmed by rMLST. MLST typing was then performed to establish the STs of Enterobacter spp. (Fig 2). Twelve STs were found for E. ludwigii: ST374, ST895, ST896, ST1179, ST1271, ST1629, ST2129, ST2402, ST2738, ST2797, ST3312, and ST3314; and four STs for E. hormaechei: ST114, ST171, ST68, and ST1644. ST1629 and ST1644 were each detected in two strains. For the remainder of the study, we retained WGS-based characterization, as it allows for genomic discrimination within the Enterobacter cloacae complex strains as well as the characterization of newly classified genera such as Huaxiibacter spp. The rationale for this choice is detailed in the Discussion section. To resolve discrepancies between MALDI-TOF and rMLST results, we used the Mash tool to estimate genomic similarity (Mash distance) between our isolates and reference genomes from NCBI. The lowest distances (<0.05) consistently corresponded to the species identified by rMLST and these reference genomes, confirming their taxonomic assignment (S5 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Alluvial plot of comparison of identification of equine strains by MALDI-TOF and rMLST.

The same isolates were characterized by both methods. Colors were assigned based on rMLST identification: Enterobacter spp. (blue), Rahnella spp. (pink), Lelliottia spp. (green), and Huaxiibacter spp. (yellow).

https://doi.org/10.1371/journal.pone.0333701.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Phylogenetic tree of the collection strains associated with 17 reference genomes (in blue) downloaded from NCBI.

Sequence types (STs) are determined for the Enterobacter spp. genus, ND corresponds to non-determined STs. Colored dots represent the presence of antibiotic resistance genes against β-lactams (blue), aminosides (orange), macrolides (violet), quinolones (pink), fosfomycines (green), phenicols (yellow) and sulfonamides (turquoise). The heatmap shows antibiotic resistance phenotypes (dark squares: resistant, light squares: susceptible).

https://doi.org/10.1371/journal.pone.0333701.g002

Following the identification of the strains, we performed statistical tests (Fischer’s exact test) to determine whether there was a significant prevalence of certain species according to sampling period, sample type, breed (French Trotters versus Thoroughbreds), gender (male versus female), or life stage (fetus versus foal/yearling/adult). We observed that E. hormaechei was significantly more prevalent in horse necropsies at the “foal”, “yearling”, or “adult” stages compared to the “fetus” stage (p-value = 0.0046). We also observed that E. hormaechei is significantly more prevalent in samples from lungs than in other samples (p-value = 0.0017) (S6 Table).

Virulence genes

Seven families of virulence factors were found (Fig 3): genes involved in secretion system synthesis, flagellum synthesis, competition for nutrients in extracellular environment, tissue adhesion and invasion, antimicrobial resistance, transcriptional regulators and toxin-encoding genes. Genes involved in flagellum synthesis were the most widespread among the strains in the collection. By contrast, east1 and colicin/colicin IB genes encoding toxins were only found in two strains, E. ludwigii A6 and E. hormaechei B1, respectively. These genes have been acquired and are carried by a plasmid and are not commonly found in E. hormaechei. Enterobacter spp. have relatively conserved profiles between strains. However, E. hormaechei B2 (ST114) strain has additional genes involved in yersiniabactin synthesis (irp1, irp2, ybtP, ybtQ, ybtT, ybtU and ybtX). The two strains belonging to the genus Huaxiibacter spp. also exhibit similar arsenal of virulence genes as other genera. They both possessed others genes (aec18, aec23, aec25, aec30) involved in the synthesis of the T6SS. In contrast, they lacked certain genes associated with the same function (icm, imp, vipB, and sciN) that were present in some Enterobacter spp. They are also the only ones to carry the csgA gene, which encodes the major subunit of curli (adhesin). R. aceris E1 and E2 strains belonged to the genus with the fewest virulence factors. R. aceris E1 harbored only genes related to flagellar synthesis (fliN, fliY, cheB, and cheR), while R. aceris E2 possessed only two genes involved in T6SS synthesis, along with the transcriptional regulator rscB, which plays a role in regulating this process. For the genus Lelliottia spp., the virulome was more similar to that of the genus Enterobacter. However, the genes involved in flagella synthesis are less diversified. L. amnigena D6 was the strain with the most genes involved in this process (n = 9). Divergences appeared between the species L. amnigena and L. nimipressuralis. Indeed, L. amnigena possesses the csgF gene, while L. nimipressuralis has the csgB gene (coding for adhesin proteins). L. nimipressuralis strains lacked certain genes L. amnigena strains had, such as flgB, cheWZ, tcyJ (flagella synthesis), entE and entS (enterobactin), acrA (antibiotic efflux pump), and galF (transcriptional regulator).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Virulence genes found in equine isolates (coloured squares).

Genes involved in the synthesis of T6SS (pink), flagella (orange), competition for environmental nutrients (green), tissue adhesion and invasion (blue), antibiotic efflux (violet), synthesis of transcriptional regulators (yellow) and toxins (red). Boxed genes are plasmid-encoded.

https://doi.org/10.1371/journal.pone.0333701.g003

Discussion

This study allowed the analysis of a unique collection of Enterobacteriaceae strains obtained from horse necropsies and collected over an extended period (1997–2020). As these strains were suspected to be implicated in equine mortality, the characterization of their species, virulome, resistome, and antimicrobial resistance profiles is particular importance, especially within the One Health framework.

We first aimed to confirm the identification of the strains using mass spectrometry (MALDI-TOF), a technique commonly employed for microorganism identification in clinical microbiology. As MALDI-TOF failed to identify nine of the 34 strains, we used rMLST to complete and confirm the identification of strains at both the genus and species levels. This approach thus allowed us to identify several genera within the collection, including Enterobacter (previously identified using the API20E gallery), as well as Lelliottia spp., Huaxiibacter spp., and Rahnella spp.. As demonstrated by a study conducted in Denmark comparing identification methods, MALDI-TOF proves useful for the accurate characterization of commonly encountered strains such as K. pneumoniae and Staphylococcus aureus. However, this technique is less discriminating for strains within the E. cloacae complex and struggles to accurately identify less well-characterized strains at the species [28]. The unprecise assignments by MALDI-TOF compared to rMLST (26.5% of common identification) was notably due to the recent reclassification following advancements in sequencing techniques. For example, Enterobacter nimipressuralis and Enterobacter amnigena were reclassified in 2013 as L. nimipressuralis and L. amnigena based on sequencing of the gyrB, rpoB, infB, and atpD genes [9]. Similarly, the genus Huaxiibacter was newly described in 2022 [29]. The genus Rahnella differs somewhat; it was first proposed in 1979 and was reclassified in 2016 as belonging to the Yersiniaceae family instead of the Enterobacteriaceae family [30]. In comparison to the identifications made by rMLST, it was observed that the genera Lelliottia spp. and Rahnella spp. were consistently identified by MALDI-TOF, although species-level identification was not achieved. Conversely, 86% of Enterobacter spp. identified by MALDI-TOF were subsequently confirmed by rMLST, while 14% were reclassified by rMLST as belonging to the genus Huaxiibacter. In addition, the difficulty in identifying the genera Rahnella spp., Lelliottia spp., and Huaxiibacter spp. using MALDI-TOF is explained by the MBT 8468 MPS database (Bruker) employed, which contains only a limited number of reference spectra for these genera (seven spectra for Rahnella spp. and Lelliottia spp., and none for Huaxiibacter spp.).Taken together, of the two identification methods tested, WGS combined with rMLST analysis proved to be the most discriminating and particularly useful for population-level studies. Furthermore, identification by MALDI-TOF depends on database used: as ours is tailored specifically to human medicine, it may be inappropriate for the caracterisation of less well characterized species. However, MALDI-TOF remains a fully applicable method, especially when rapid diagnosis is required, as it provides a reasonably accurate genus-level identification, particularly for genera known to pose potential challenges.

MLST analysis of Enterobacter strains revealed a high diversity of STs, suggesting a lack of close genomic relatedness among the strains. ST1629 (E. ludwigii A4 and A13) and ST1644 (E. hormaechei B4 and B5) were each detected in two strains originating from different locations, indicating that these occurrences are unlikely to result from clonal dissemination. Moreover, these STs have not been previously described in the literature. Of particular note, two other E. hormaechei STs, ST114 and ST171, were also identified within the collection. These ST had been reported multiple times in various studies and is recognized as an international high-risk clone. ST171 was documented during an outbreak in a New York hospital, where it accounted for 50% of identified STs in 2014 [31], and its prevalence reached 28% of Enterobacter spp. in a hospital in Shandong, China, in 2023 [32]. This clone has also been detected in environments in contact with animals, such as zoological facilities in 2023 [33], and in companion cats in 2024 [34]. The E. hormaechei B1 ST171 isolate also carries the IncFIB plasmid, which has been identified in ST171 strains producing the KPC-4 carbapenemase [33]. The presence of IncFIB(K), IncFIB(pHCM2) and IncI1-I(Alpha) plasmids in E. hormaechei B1 could be responsible for carrying the acquired resistance genes blaCMY-2 and bla-TEM-1B. ST114 has also been frequently identified in both humans and animals. Its presence in animals is thought to reflect anthropogenic contamination of the environment [35]. ST114 strains recovered worldwide are extended-spectrum β-lactamase (ESBL) producers, commonly carrying blaOXA, blaSHV, blaCTX-M and blaTEM [36,37], and exhibit extensive resistance to 3rd-generation cephalosporins, like our isolate. ST114 is frequently associated with the IncHI2 plasmid [35], which was also detected in our isolate. This plasmid is typically responsible for ESBL gene carriage in most strains. However, due to the absence of complete plasmid sequencing in our study, we cannot confirm whether blaDHA-1 blaOXA-1, or blaSHV-12 (found in E. hormaechei ST114) are located on this plasmid. These various cases illustrate the ability of bacteria to become clonal population and to acquire mobile genetic elements carrying antibiotic resistance genes conferring multidrug resistance. However, Enterobacter spp. harbored the highest number of resistance genes and showed resistance to at least three classes of antibiotics in 26% of cases in 2016 and 52% in 2019 [3,10]. In our study, 22% of the strains from our collection were also resistant to at least three antibiotic classes, particularly E. hormaechei strains isolated between 2006 and 2017. Among the E. hormaechei strains, 80% (4/5) showed resistance to at least three classes of antibiotics, excluding β-lactams (including aminoglycosides, macrolides, quinolones, chloramphenicol, sulfonamides, and tetracyclines). For E. ludwigii, 8% (1/13) of strains were resistant to at least two classes of antibiotics. However, we observed one isolate with a high resistance potential: the E. hormaechei B2 isolate, corresponding to ST114. As this sequence type has been associated with human outbreaks, its presence in horses could result from human-to-animal transmission. However, no documented cases support this type of transmission, whereas animal-to-human transmission (zoonosis) is frequently reported [38].

Strains belonging to the same ST may still exhibit differences in their resistome and phenotypic profiles, likely due to environmental differences and genomic plasticity. The genus Enterobacter spp. displays a richer virulome, indicative of higher virulence and potential pathogenicity. For example, we detected the presence of the enteroaggregative heat-stable enterotoxin 1 (EAST1) and colicin, synthesized by E. ludwigii A6 (ST2797) and E. hormaechei B1 (ST171), respectively. The EAST1 was originally characterized in E. coli and is known to induce intestinal water secretion, leading to diarrhea in both humans and animals [39]. Colicin is a bacteriocin with bactericidal activity [40], providing a competitive advantage. It acts by forming pores in the bacterial cell membrane [41]. Its gene is carried by high-molecular-weight plasmids of the heterogeneous IncFI group [42]. Notably, the E. hormaechei B1 harbors two plasmids belonging to this group, suggesting that the colicin gene may be carried by one of them. We also observed the specific presence of the irp1, irp2, and ybtPQTUX genes in E. hormaechei B2, which are involved in iron chelation. The irp1 and irp2 genes are responsible for the synthesis of HMWP1 and HMWP2 (high-molecular-weight proteins 1 and 2) [43], while the ybt genes are involved in the transport of the yersiniabactin–Fe3+ complex, as well as in the regulation, maturation, resistance to self-intoxication, and recycling of yersiniabactin. In addition to providing a competitive advantage, yersiniabactin contributes to oxidative stress tolerance [44] and may serve as a marker of increased pathogenic potential. Genes involved in motility and adhesion are also widely distributed across the different genera. Motility and adhesion are key factors for colonization, environmental persistence, and virulence, making them central elements in pathogenicity. Genes associated with the T6SS are also widely found across the four genera. This secretion system is broadly distributed among Gram-negative bacteria and can be highly variable, particularly within the genus Enterobacter spp. [16]. The T6SS is involved in metal acquisition (Zn2+ and Mn2+), the response to stress induced by reactive oxygen species (ROS), and acidic pH conditions such as those encountered within macrophages [45]. Additionally, T6SS contributes to virulence by acting as an injector of effectors into eukaryotic cells. It has been shown to induce hemolytic activity and facilitate adhesion to epithelial cells during infection [16].

The genera Lelliottia spp., Huaxiibacter spp and Rahnella spp. are sporadically implicated in human infections such as septicemia in drug addicts or premature newborns [46,47], septic shock in immunocompetent individuals [48] or pyonephrosis [49]. However, no data have yet been published on its involvement in equine infections/mortality and limited data exist on their antimicrobial susceptibility profiles. Based on the collection, these different genera would account for a prevalence of presence in equine necropsy samples of 47%, while Enterobacter spp. would account for a prevalence of 53%. The genera Rahnella spp., Huaxiibacter spp. and Lelliottia spp. exhibited distinct resistomes and antibiotic resistance profiles. For example, Rahnella spp. and Lelliottia spp. have their own chromosomal β-lactamases (blaRAHN and blaLAQ-1, respectively). Overall, we observed that 50% (1/2) of R. aceris strains and 16.7% (2/12) of Lelliottia spp. strains were resistant to at least two classes of antibiotics (excluding β-lactams). These resistances involved macrolides (clarithromycin), aminoglycosides (amikacin and gentamicin), or fosfomycin. We also observed that H. chinensis strains were resistant only to certain β-lactams and clarithromycin, making these strains the least resistant to the antibiotics tested. Despite a limited resistome (with few or no known antibiotic resistance genes), Lelliottia spp., R. aceris, and H. chinensis nonetheless displayed antibiotic resistance, suggesting the involvement of alternative mechanisms such as efflux pumps, reduced membrane permeability, or chromosomal mutations. Conversely, other genotype–phenotype discrepancies were also observed. The presence of resistance genes such as fosA2, aac(6’)-Ib-cr, and qnrB2 in certain strains was not consistently associated with a resistant phenotype, suggesting a possible role of transcriptional regulation, conditional gene expression, or inactivating mutations. These findings highlight the importance of interpreting genomic data with caution and always correlating it with phenotypic evidence. Moreover, although the genera Huaxiibacter spp., Rahnella spp., and Lelliottia spp. are only sporadically associated with human clinical infections [46–49], they are predominantly found in the environment. Their presence in necropsy samples may suggest that they were part of the equine microbiota, possibly due to contamination of drinking water, for example. The potentially environmental origin of these strains could be at the root of their low resistance to antimicrobials and the absence of acquired resistance genes (low selection pressure). The presence of plasmids in some of these strains may then have other roles, such as promoting resistance to environmental stresses (tolerance, degradation, detoxification of heavy metals, for example) [50].

We observed that virulence profiles vary across genera and species. However, the same categories of genes—those involved in T6SS, flagellar synthesis, environmental resource competition, adhesion, AcrAB efflux pump synthesis, and transcriptional regulators— have been found in Lelliottia spp. and Huaxiibacter spp.. The R. aceris strains harbor a very limited virulome, suggesting either low pathogenic potential or the presence of virulence factors that have not yet been characterized.

Conclusion

This study, conducted on a unique collection of Enterobacteriaceae isolated from necropsied horses in Normandy over more than two decades, highlights an unexpected taxonomic and genomic diversity. Compared to MALDI-TOF, rMLST proved to be the most relevant approach for the accurate identification of strains, revealing the presence of recently described or reclassified genera such as Lelliottia spp., Huaxiibacter spp., and Rahnella spp.. This diversity underscores the limitations of conventional identification methods and reinforces the value of the whole-genome sequencing as a reference tool in veterinary microbiology. Genomic analysis revealed substantial heterogeneity in STs of Enterobacter spp., with no evidence of clonal transmission, yet the detection of high-risk clones known in human medicine (E. hormaechei ST114 and ST171) raises questions about potential interactions between animal, environmental, and human microbiota, within a One Health framework. The presence of antibiotic acquired resistance genes and plasmids suggests that E. hormaechei strains may act as reservoirs of mobilizable resistance determinants. In parallel, the variability of virulence profiles across genera (from highly pathogenic strains in Enterobacter spp. to less virulent genus like Rahnella spp.) reflects an evolutionary dynamic that remains poorly understood. Nevertheless, the detection of genes involved in T6SS, adhesion, motility, or iron metabolism suggests that these environmental strains could become opportunistic under certain conditions and potentially contributing to equine infections.

Supporting information

Acknowledgments

The authors would like to thank the University of Caen Normandy, Caen University Hospital (France), and the ANSES Laboratory in Goustranville for their institutional support. We also express our sincere gratitude to the molecular biology technicians Isabelle Renoir, Guillaume Leduc, and Anaïs Rousseau for their valuable assistance throughout the analyses, as well as to Nadia Coudray, head of the platform, for her technical support.