Assessment of Pb(II), Cd(II), and Al(III) Removal Capacity of Bacteria from Food and Gut Ecological Niches: Insights into Biodiversity to Limit Intestinal Biodisponibility of Toxic Metals

Toxic metals (such as lead, cadmium, and, to a lesser extent, aluminum) are detrimental to health when ingested in food or water or when inhaled. By interacting with heavy metals, gut and food-derived microbes can actively and/or passively modulate (by adsorption and/or sequestration) the bioavailability of these toxins inside the gut. This “intestinal bioremediation” involves the selection of safe microbes specifically able to immobilize metals. We used inductively coupled plasma mass spectrometry to investigate the in vitro ability of 225 bacteria to remove the potentially harmful trace elements lead, cadmium, and aluminum. Interspecies and intraspecies comparisons were performed among the Firmicutes (mostly lactic acid bacteria, including Lactobacillus spp., with some Lactococcus, Pediococcus, and Carnobacterium representatives), Actinobacteria, and Proteobacteria. The removal of a mixture of lead and cadmium was also investigated. Although the objective of the study was not to elucidate the mechanisms of heavy metal removal for each strain and each metal, we nevertheless identified promising candidate bacteria as probiotics for the intestinal bioremediation of Pb(II) and Cd(II).

Although such LAB-mediated metal removal capacity was partly shown to be strain-dependent, scarce studies have explored species and strains diversity. Only few restricted types of LAB are generally analyzed in distinct and designed heterogenous studies among lactobacilli, enterococci and Weissella spp [41,42] together with dairy propionibacteria and bifidobacteria [38,43]. In addition, such properties were rarely sought in non-lactic acid bacteria, with a limited number of evaluations in few proteobacteria species (E. coli) and gut-isolated anaerobic bacteria (Akkermansia muciniphila, Faecalibacterium prausnitzii and Oscillibacter ruminantium single strain isolates) [33,44]. Other genera showing noticeable detoxification potentials in vitro, such as Pseudomonas, Stenotrophomonas or Bacillus, are not appropriate for intestinal compartment targeting. Only LAB such as L. plantarum, L. casei, L. rhamnosus and L. delbrueckii strains have been so far selected in vitro and confirmed to have detoxification abilities in vivo. Various selected food microbes can thus prevent the absorption of heavy metals in the gut (and dissemination in various tissues) and remove them upon defecation. Promising proofs of concept of efficacy were demonstrated in preclinical models of acute and chronic heavy metal toxicity in mice for lead [33,45,46], cadmium [47,48] and aluminum [49,50].
Cell surface associated compounds of probiotic lactobacilli sustain the strain-specificity dogma of strain's functionality [51]. The mechanism responsible for binding of metals to bacterial cell wall is highly suggested to depend on the huge variety of surface molecules of individual bacterial species and strain [52], including teichoic and lipoteichoic acids and peptidoglycan. In line, considering other binding site alike S-layer proteins, cell surface proteins and polysaccharides, we can question the distinct biosorbent properties of other Gram positive and Gram negative bacteria.
Here, we thus addressed the variability, among species and strains, by analyzing the ability of bacteria to remove the potential harmful trace elements lead, cadmium and aluminum in vitro. The purpose here is not to elucidate the mechanism of biosorption or bioaccumulation by plotting adsorption isotherms, but rather to compare the intrinsic aptitudes to cope with heavy metals among bacteria from the phylum Firmicutes, Actinobacteria and Proteobacteria. It comprises many LAB (99), several bifidobacteria (11), dairy propionibacteria (21), and cutibacteria (4), together with other gut-friendly bacteria such as Enterobacterales (90). This study thus aims to explore the strain diversity and the metal dependency of the overall toxic metal removal capacity of various food and gut bacteria. It also serves to identify the best candidates for preclinical assays and further veterinary and clinical applications.

Chemicals, reagents and instruments
Chemicals and reagents were purchased from Sigma-Aldrich Chemical (Saint-Quentin-Fallavier, France), unless otherwise stated. Ultrapure water corresponds to PURELAB Option-Q; Veolia Water (Antony, France). Ultraflex III MALDI-TOF/TOF instrument and Flex Analysis software were from Bruker Daltonik GmbH (Bremen, Germany).
Determinations of metal concentrations in diluted samples were performed using Inductively Coupled Plasma -Mass Spectrometry (ICP-MS) THERMO ICAP TM Qc (Thermo Scientific, Courtaboeuf Cedex, France).

Bacterial strains collections and culture conditions
A set of 225 bacterial strains of distinct origins was used in this study. Lactic acid bacteria (LAB) mostly came from the well characterized DSM and ATCC collections previously used for comparative genomics of lactobacilli and associated genera [53].
Most Escherichia coli strains belong to the ECOR standard reference strains of E. coli collection [55]. The later includes isolates from a variety of hosts and geographic regions, covering A, B1, B2, D, and E phylogroups, and were kindly provided by Dr Laurent Debarbieux (Institut Pasteur Paris). Other E. coli type strains or characterized as adherent-and invasive pathovars (AIEC) were described previously [56]. Strains of Serratia marcescens (Db10, JUb9, SM25, SM38 and SM45) were kindly provided by Dr Elizabeth Pradel [57]. Some cheese-derived Hafnia alvei strains (Gb01, E215, 920 and Grignon) were described elsewhere [58]. Finally, few bacterial strains (9 Bifidobacterium species, 4 Cutibacterium acnes, 2 Enterobacter, 2 Hafnia alvei and 5 Klebsiella) were sourced from historical clinical gut or fecal samples of human origin, food or as re-isolates from commercial probiotics products (Bb12 and Morinaga) belonging to the Faculty of Pharmacy of Lille (FPL) collection, University of Lille. Identification of those strains were determined by selective media and the species level was confirmed using matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) and are referenced by internal FPL numbers. were grown in LB (Luria Bertani medium) at 37 °C without shaking. Bacterial cultures duration ranged from overnight to 72h depending on the bacterial strain in order to reach the stationary phase.

Metal-removal capacity assays
(See also Preamble) Eight mL of stationary phase bacterial cultures were standardized at optic density (OD) 600 nm of 2.5 and washed twice in Ringer's solutions. Pellets were further suspended with 8 mL of the corresponding ion's metal solutions (Ringer, pH 7.0) at 25 ppm (PbCl2, AlCl2) or 1 ppm (CdCl2) and gently mixed using a rotary agitator (12 rpm), at room temperature for 1h. Samples were then centrifuged and washed twice before metal quantification by inductively coupled plasma mass spectrometry (ICP-MS). The pellets were suspended in 500 µL of 70% nitric acid, and heated at 98°C for 15 minutes. The samples were finally diluted in mQ water and further assessed by ICP-MS method. For each strain, a percentage of chelation/removal capacity is defined as the ratio of residual metal mass quantitated in the pellet toward the initial amount in the incubation medium. All assays were performed in triplicate, corresponding to three distinct bacterial cultures.

Statistical analyses
GraphPad Prism was employed for graph preparation and statistical evaluation. All analyses were performed by comparing experimental groups with their respective controls in the nonparametric, one-way analysis of variance (the Mann-Whitney U test) or a two-tailed Student t test, as appropriate (GraphPad Prism, version 6.0, GraphPad Software Inc, San Diego, CA, USA). Quantitative variables were quoted as the mean +/-standard error (SD). Data with p values ≤ 0.05 were considered to be significant.

Results
We first ensured that our methods were reliable and appropriate enough to screen distinct bacterial strains for their ability to remove selected metals. It has been previously shown that many factors may influence the levels of metal binding by bacteria, such as contact time, temperature, pH, and salt concentration of metal solutions, as well as washing buffers and inoculum size [37,60,61]. Thus, several key parameters were defined to mimic the gut environment unless fixed for convenience, e.g. temperature. The binding evaluation was thus done in physiological saline buffer (Ringer's solution), at a neutral pH (7.1) and at room temperature (22 +/-2 °C), in time-separated and bacterial culture triplicates to test reproducibility. Lead and aluminum concentration at 25 ppm were retained as a realistic dose to evaluate metal sequestration, whereas a lower dose of 1 ppm of cadmium was necessary to allow discriminant selection of strains.
Indeed, cadmium at 25 ppm was not appropriate to identify clear differences among bacteria and most strains had very low cadmium binding capacity in a first pre-screening step (data not shown), below 5% of removal capacity.
The screening of bacterial strains is based on the residual metal fraction robustly associated to the pellets. However, we also confirmed the stability and irreversibility of binding, in order to discriminate strains from weak to strong metal binding capacity for both lead and cadmium. In this aim, we performed two serial cycles of washes (consisting in two cycles of pellet resuspension in metal-free solutions followed by centrifugation). Indeed, the residual quantity of metal in the second and last wash samples were very low or undetectable and considered as negligible. This is illustrated by selected examples of strains with distinct binding efficiency on supplementary figure 1A and 1B. However, binding of aluminum to bacteria was more labile and appeared partly reversible after rinsing, although the test using 2 wash steps is quite reproducible (supplementary figure 1C). Consequently, aluminum binding is somewhat overestimated in our assay but allows to discriminate strains and may reflect the intrinsic capacity of bacteria to interact with the metal in physiolocal conditions, i.e. the gastrointestinal tract.

Lactic acid bacteria (LAB) exhibit variable lead removal capacities
Among the 99 individual lactic acid bacteria strains tested for their capacity to remove Pb(II) salts at 25 ppm, most (>2/3) were able to immobilize an average of 50% to up to 90% of this metal in solution ( Figure 1A) while others part showed moderate or quite weak biosorption potentials. When considering only the genus Lactobacillus, covering 65 distinct species and 76 strains, the removal capacity of lead ranged from 6% +/-2.5 to 92% +/-8.5. No particular consistency could be identified at the species level and some strains belonging to the same species could show extremely different properties such as L. acidophilus, L. casei, L. paracasei L. rhamnosus (Figure 2A). The two strains of L. fermentum were particularly efficient in lead removal, whereas 2 L. plantarum strains were unexpectedly poor lead biosorbers ( Figure 2B).
The lead removal capacity is also variable among strains from other genus comprised within the LAB.
Carnobacterium spp and Pediococcus spp, together with Leuconostoc and Fructobacillus members as well as the Weissella spp, demonstrated substantial lead removal properties ( Figure 2C). Such potential is not related to the intrinsic shape of bacteria (i.e. rods or cocci) because the 3 enterococci and 4 pediococci tested were quite good metal biosorbers (> 50%) while Lactococcus lactis and 4 distinct strains of Staphylococcus aureus were not (mostly < 30%) ( Figure 2D). Metal removal capacities were expressed as % of initial metal quantity in solution as mean +/-SD of three determinations. They are represented by colored heat maps as weak (0 to 25%), low (26 to 50), moderate (51 to 75%) and high (76 to 100%) performing bacterial strains, respectively in pale green, blue, orange and red, to remove lead.

Actinobacteria cover distinct lead biosorption potentials
Within the phylum of Actinobacteria, bifidobacteria exhibit very poor or extremely high lead removal capacities, depending on the species and related strains (Figure 1B), some strains reaching the value of nearly 90% whereas other could only bind 6.6% of the solubilized lead. This is not related to the species, as distinct B. longum strains may exhibit up to 10-fold higher binding capacities than others, i.e. 6.75% +/-0.9 versus 65.4% +/-7.2 (p < 0.001). Surprisingly, none of the Propionibacterium freudenreichii strains from dairy origins was able to alleviate the concentration of lead salts,

Bacteria-mediated cadmium removal capacity is phylum, genus and strain specific
As explained in the preamble, the dose of 25 ppm is not appropriate to discriminate bacterial strains with respect to binding of cadmium Cd(II). Nor is it relevant to mimic realistic contamination events. Here, we thus addressed the ability of several bacteria to lower cadmium salts from 1 ppm solution. Among the 95 LAB strains tested, most of them (around 90%) exhibited weak or low cadmium removal properties, below the value of 20% of binding ( Figure 3A).
Interestingly, few strains comprising 3 Pediococcus spp, a Carnobacterium divergens and to a lesser extent a single L.
rhamnosus and a Leuconostoc mesenteroides have cadmium binding capacities over 25% and up to 50% +/-15.7 for a P.
Considering the phylum of Actinobacteria, bifidobacteria are characterized by variable binding of cadmium, depending more on the strain than on the species. Indeed, B. breve strains removal capacity ranged from 6.2% +/-0.6 to 40.7% +/-6.7 (p<0.01) and B. longum strains from 3.6% +/-1.7 to 18.9% +/-5.8 (p<0.01) Most of the Propionibacterium freudenreichii strains were consistently weak cadmium biosorbers, whereas C. acnes strains were distinctly either low or moderate in their overall capacity to remove cadmium ( Figure 3B) (Figure 3C).

Bacteria-mediated aluminum removal capacity is also genus and strain dependent
As a matter of fact, Al(III) from aluminum chloride solution can also distinctly bind to bacteria, depending on their origin and phylogenic diversity. Aluminum removal capacity by LAB is discrepant from various species, ranging from 5 to 28% of 25 ppm solutions with an average of 14.8% +/-4.7 (Figure 4A). Similarly, the ability of bifidobacteria and propionibacteria to bind aluminum is quite weak and rarely exceed 10% (8.9% +/-2.8 and 9.4% +/-2.6, respectively).
Strains of Cutibacterium acnes were more efficient and showed moderate binding levels (means of 24.3% +/-4.7) ( Figure   4B). In contrast, representatives of Enterobacterales exhibited usually higher values for this phenotype, ranging from 12% up to 30% and a median value of 20.4% +/-4.7 ( Figure 4C). Few E. coli strains were particularly favorable, i.e.
ECOR37, ECOR40, ECOR50 and ECOR64, ranging from 25 to 30 %, whereas the overall Hafnia and Serratia strains seem to be less promising, below 15%. Metal removal capacities were expressed as % of initial metal quantity in solution as mean +/-SD of three determinations. They are represented by colored heat maps as weak (0 to 10%), low (11 to 20), moderate (21 to 30%) and high (over 100%) performing bacterial strains, respectively in pale green, blue, orange and red, to remove cadmium.   Metal removal capacities were expressed as % of initial metal quantity in solution as mean +/-SD of three determinations. They are represented by colored heat maps as weak (0 to 10%), low (11 to 20), moderate (21 to 30%) and high (over 100%) performing bacterial strains, respectively in pale green, blue, orange and red, to remove aluminum. cloacae  Because co-exposure to lead and cadmium may commonly happen due to their co-occurrence in food, water and environment, we also addressed the respective binding of both elements when these metals were mixed together. We thus compared the binding capacity for lead at 25 ppm, cadmium at 1 ppm and for the corresponding mixture (i.e. Pb 25 ppm and Cd 1 ppm) of a set of arbitrary selected 16 representative Gram positive ( Figure 5A) and 16 Gram negative bacteria ( Figure 5B). Interestingly, the intrinsic removal capacity of bacteria was fairly not influenced by the presence of the other metal, except for few strains (out of 32) showing rare but significant lowering of 20 to 40% of the initial baseline values (p < 0.05), irrespective of the metal considered. Figure 5. Ability of bacteria to remove lead and cadmium in pure or mixed solutions. Metal removal capacity was determined for lead at 25 ppm (black), for cadmium 1 ppm (grey), alone or as mixtures of both (25 ppm lead and 1 ppm Cd) for lead (hatched black) and cadmium (hatched grey). Values are expressed as % of initial metal quantity in solution as mean +/-SD of three determinations. * indicates the significant differences of mixture (p < 0.05) compared with corresponding metal alone. Here, we addressed the metal removal capacity in more than 200 bacterial strains, mostly lactic acid bacteria and associated genera, as well as representatives of gut enterobacteria inhabitants. We considered lead, cadmium, aluminum, and to a lesser extent, a mixture of lead and cadmium. Of note, all tested bacteria were able to survive at the corresponding doses of metals used, i.e. 25 ppm for lead and aluminum and 1 ppm for cadmium (data not shown) and as previously described elsewhere for several LAB strains [41]. Considering cadmium, the removal capacity of distinct strains was not correlated to the minimal inhibitory concentration (MIC) established at higher doses (data not shown).

Discussion
The later suggests that exploring metal tolerance is not appropriate for screening purposes, as also demonstrated for lead with various L. plantarum strains [62]. The viability of bacteria is so far not necessary to allow significant metal biosorption. Indeed, binding isotherms in the Langmuir model showed that the maximum binding capacity (Qmax) could either be significantly higher or lower in boiled or live forms for two probiotic strains (Lactobacillus rhamnosus and Propionibacterium freudenreichii) [37]. Others have demonstrated that dead and live bacteria had similar lead and cadmium binding capacity (33,37,40,43). However, some slightly higher removal efficiency of lead by living forms could be observed, owing to the occurrence of cell-specific intracellular metal accumulation [63].
Other strategies based on lactobacilli surface characteristics, such as hydrophobicity and electrostatic properties, failed to identify relevant selection criteria for lead and cadmium removal [64]. Yet, no rationale is established to select strains with high detoxification potential. In our study, we thus empirically characterized the removal metal capacity of bacterial living biomasses without a priori, considering the cross-species and strain diversity. We used gut friendly (nonpathogenic) bacteria, either originated from intestinal ecological niches or derived from food, mostly regarded as Generally Regarded as Safe (GRAS), because safety is essential for further in vivo applications. In contrast with many studies exploring the efficiency of metal removal within metal solutions in deionized water, we therefore used Ringer's solution at neutral pH, both for the binding assays and the washing of the pellets in order to achieve isotonicity close to biological conditions. Incubation time by bacteria and the distinct metals was set up at one hour to partly mimic the food transit time and the possible contact time within the gut.
Although LAB-mediated metal removal capacity has already been shown to be strain dependent, very few studies have explored the cross-species and strain diversity. Only few restricted types of LAB are generally analyzed and often in heterogeneous design studies among lactobacilli, enterococci and Weissella spp [41,42]. In line, data considering dairy propionibacteria and bifidobacteria are scarce and included a limited number of species and strains [38,43]. Studies that comprise proteobacteria are also poorly documented. We have here extended and confirmed the huge functional diversity throughout bacterium specimen and the distinct metals for Gram-positive and Gram-negative bacteria. Mechanisms of metal removal have been described elsewhere [52,65]. They include ion exchange, chelation, adsorption by physical forces and intracellular sequestration and are known to be strain-dependent. The role of hydroxyl (from the peptidoglycan), carboxyl and phosphate groups (from surface proteins) influenced by pH and specificity and abundance, is assumed to be a key determinant, together with the contribution of capsular polysaccharide for metal binding sites. Thus, the overall removal capacity of a single cell is complex and multifactorial.
Anyway, although many variables such as culture conditions, culture medium types and growth phase are involved, independently of the core and individual specific bacterial genes, the comparative genomic among lactic acid bacteriacan additionally be used help to identify specific genes amplifying or lowering factors for metal removal.
We found that LAB and bifidobacteria have generally moderate to high lead removal capacities, whereas dairy propionibacteria consistently have weak performances. Gram negative bacteria have almost low to moderate aptitudes to immobilize lead. We could identify good candidates toward lead among lactobacilli, and bifidobacteria, as previously described for L. sakei, L. delbruckii, L. fermentum and B. bifidum strains [45]. In line with our results, Weissella and Pediococcus spp were also described with high levels of lead removal [41]. To our knowledge, we first describe the promising potential of Carnobacterium spp. Out of 220 strains, only five were identified with high capacity to interact  [45,48], the few strains of L. rhamnosus and L. plantarum from our set of bacteria exhibited poor lead and cadmium lowering properties. Again, metal removal capacities are highly straindependent and yet cannot be generalized at the species taxonomic level. Interestingly, among the 20 strains of Propionibacterium, all are weak chelators for both lead and cadmium, although they demonstrated very strain-specific surface proteins and exopolysaccharide production abilities, related to various immunological functional properties [66].
Existing data on aluminum removal capacity by bacteria are quite rare and have only reported potentials for L.
plantarum and L. reuteri strains [50], exceeding the % of removal we could reach in our study (25%) in a similar experimental design. Noticeably, Enterobacteria demonstrated more consistent and higher capacities for aluminum removal than LAB and Actinobacteria.
Another interesting and promising result from our study is the overall maintenance of the removal capacity of a selected strain when both cadmium and lead are applied together. This encourages us to select strains with shared high lead and cadmium removal capacities. In line, synergistic issues to question the impact of mixture of strains on a single (or a mixture of) metal(s) have also to be explored [67].
The proof of concept that some bacterial strains which possess high in vitro metal removal capacities can also have these potentials in vivo in models of acute and chronic poisoning in mice [33,[45][46][47][48]50], rats [67] and humans [68] has been clearly demonstrated. The first screening step we described here needs to be further assessed in vivo considering the physiology of the gastrointestinal tract in the presence of other essential metals, trace elements and organic molecules. Experimental protocols in preclinical models must include negative controls by comparing strains with poor and high removal capacities, in order to ensure the intrinsic contribution of the selected bacteria as a detoxification tool. In order to develop probiotics for toxic metal removal, appropriate strains and even cocktails of several bacteria should clearly be evaluated. On the one hand, the interaction with other bacteria inhabiting the intestine has also to be considered, as the baseline role of resident commensal microbiota is a key factor [33]. On the other hand, distinct strategies used to modify the gut microbiota, including the prebiotics, may also interfere with bacteria and heavy metal equilibrium [69]. Because heavy metals influence the structure, the diversity, and the functionality of the gut microbiota [29] (including heavy metal sequestration), the bidirectional relationship of dysbiosis and heavy metals in various pathologies, and the interconnected use of probiotics with multipurpose functions, is highly complex [28] and will need to be integrated for a personalized medicine perspective [70].

Conclusion
Collectively, our results revealed the huge bacterial diversity in terms of ability to remove metal such as lead, cadmium, aluminum, or a mixture of lead and cadmium, in vitro. By exploring the cross-species and strain diversity of