Arsenolipids reduce butyrate levels and influence human gut microbiota in a donor-dependent way

Arsenolipids are organic arsenic species with variable toxicity. Accurate assessment of the risks derived from arsenic-contaminated seafood intake requires studying the interplay between arsenolipids and the human gut microbiota. This research used the in vitro mucosal simulator of the human intestinal microbial ecosystem (M-SHIME) to assess the effect of defined chemical standards of arsenolipids (AsFA 362 and AsHC 332) on a simulated healthy human gut microbiota (n = 4). Microbial-derived metabolites were quantified by gas chromatography and microbiota structure was characterized by 16S rRNA gene sequencing. A specific reduction in butyrate production (control = 5.28 ± 0.3 mM; AsFAs = 4.56 ± 0.4 mM; AsHC 332 = 4.4 ± 0.6 mM, n = 4 donors), concomitant with a reduction in the abundance of Lachnospiraceae UCG-004 group and the Faecalibacterium genus was observed, albeit in a donor-dependent manner. Furthermore, an increase in Escherichia/Shigella , Proteobacteria and Fusobacterium abundance was observed after arsenolipid treatments, depending on individual microbiota background. These alterations in microbial functionality and microbial community structure suggest a detrimental effect of arsenolipids intake towards the commensal gut microbiome, and consequently, on human health.


Introduction
The gut microbiota has complex interactions with arsenic (As), affecting its metabolism and toxicity whereas toxicant exposure can modify microbial community composition and function (Brabec et al., 2020;Lu et al., 2013Lu et al., , 2014)).Organic arsenicals (e.g., arsenolipids, arsenosugars) can be found in food, especially in sea products such as fish, shellfish, or algae, but they are historically considered less toxic than inorganic As (iAs).Specifically, arsenolipids are defined as lipid-soluble organoarsenic compounds, generally occurring in marine organisms, with arsenic-containing hydrocarbons (AsHCs) and arsenic-containing fatty acids (AsFAs) representing two major subgroups (Witt et al., 2017a).AsHCs have shown neurotoxic potential in vitro, exerting similar cytotoxicity and oxidative stress potential to trivalent iAs (Müller et al., 2018).Other arsenical species, including some arsenolipids, are more cytotoxic than trivalent iAs on human liver and bladder cells, and can cross the blood-brain barrier of Drosophila melanogaster model (Meyer et al., 2014(Meyer et al., , 2015a;;Niehoff et al., 2016), but their interplay with gut microbiota is still unexplored.Remarkably, it has been reported that AsHC is transferred to human milk (Stiboller et al., 2017;Xiong et al., 2020), with arsenic levels from 0.3 to 4.46 μg kg − 1 reported in breast milk, with variable percentages identified as AsFA and AsHC (Stiboller et al., 2017).In a recent study, around 3% of AsHC from a salmon fillet was recovered in breastmilk within 24 h, and despite this work only evaluated one volunteer, these results open the possibility of arsenolipids affecting infant developmental processes (Xiong et al., 2020).
Previous research showed that the gut microbiota is a modifying factor in iAs toxicity, influencing metalloid metabolism and effects at the host level.In human populations, the oral intake of iAs influences the gut microbiota, enriching pathogenic microorganisms and bacteria that produce volatile arsenic species, whereas it decreases the abundance of beneficial gut commensals (Brabec et al., 2020).Multiple studies have evaluated the effect of iAs on gut microbiota structure, but also on the metabolome of serum, feces and urine in animal models and humans (Lu et al., 2014;Chi et al., 2019;Karagas et al., 2022;Li et al., 2017;Wang et al., 2014Wang et al., , 2015;;Wu et al., 2018;Yang et al., 2021;Zhong et al., 2021).However, there is no information about the disrupting effect of arsenolipids on the human gut ecosystem.
Here, we used an in vitro approach to evaluate the effect of two different arsenolipids on gut microbiota structure (16S rRNA gene sequencing) and function (short-chain fatty acid profile), using M-SHIME, a dynamic multi-compartmental system mimicking both the luminal and mucosal microbial ecosystem from the human gut.Our results reinforce the need for incorporating the gut microbiota as a dual player in toxicity and risk assessment, on the one hand as a target organ and on the other as a metabolic organ capable of modifying the toxicity of ingested xenobiotics.

Mucosal simulator of the human gut microbial ecosystem (M-SHIME)
The M-SHIME containing sequential reactors representing the gastric, small intestinal, and colonic environments was used as detailed in Xiong et al. (2022a).Briefly, four separate M-SHIME reactors were respectively inoculated with fecal microbial slurries from four healthy donors and exposed for 70 h to a single dose of arsenolipids (an arsenic-containing fatty acid 362, AsFA 362 and an arsenic-containing hydrocarbon 332, AsHC 332, each 1.33 µmol).
Control reactors without arsenolipids were included in the run.At different time points (0 h, 6 h, 22 h, 46 h and 70 h for luminal compartment, and 22 h and 70 h for mucosal compartment), samples were obtained.Short-chain fatty acids (SCFA) were used as benchmarks of community activity, analyzed by gas chromatograpy.The microbial community of luminal and mucosal samples was taxonomically characterized by 16S rRNA gene amplicon sequencing.

SCFAs extraction and quantification
Short-chain fatty acids were used as benchmarks of community activity, and were collected from the luminal and mucosal compartment of the M-SHIME.Briefly, luminal samples (2 mL) were obtained and centrifuged (9509 × g, 10 min; Eppendorf) before SCFA analysis.Mucosal samples (1 g) were washed with phosphate buffer saline (PBS) to remove luminal bacteria and homogenized with 1 mL of PBS initially by pipetting and subsequently in a vortex at 3500 rpm (Eppendorf™ MixMate™, Eppendorf Nederland B.V., Belgium) for 30 s. C2-C8 fatty acids (including isoforms C4-C6) were measured in the supernatants by gas chromatography (GC-2014, Shimadzu®, The Netherlands) with a DB-FFAP 123-3232 column (30 m x 0.32 mm × 0.25 µm; Agilent, Belgium) and a flame ionization detector (FID).Liquid samples were conditioned with sulfuric acid and sodium chloride and 2-methyl hexanoic acid as internal standard for quantification of further extraction with diethyl ether.Prepared sample (1 µL) was injected at 280 ºC with a split ratio of 60 and a purge flow of 3 mL min-1 .The oven temperature increased by 6 ºC min -1 from 110 ºC to 158 ºC and by 8 • C min -1 from 158 • C to 175 • C where it was kept for 1 min.FID had a temperature of 220 ºC.The carrier gas was nitrogen at a flow rate of 2.49 mL min − 1 .
Total SCFA production was defined as the sum of the molar concentrations of acetate, propionate, butyrate, valerate, caproate, isobutyrate, isovalerate and isocaproate.

DNA extraction and Illumina library generation
Total DNA from luminal and mucosal samples was extracted using physical disruption with the bead beating method (Hernandez-Sanabria et al., 2010).Briefly, samples were thawed, manually homogenized, and centrifuged at 14,600 g for 5 min at 4 • C. The pellet was solubilized in mL of lysis buffer (100 mM Tris pH 8, 100 mM Na EDTA pH 8, 100 mM NaCl, 1 % (w/v) polyvinylpyrrolidone, 1% PVP40, and 2 % (w/v) sodium dodecyl sulphate) and transferred to a 2-mL microcentrifuge tube containing 0.3 g of zirconium beads (diameter, 0.1 mm).The cells were lysed in a Power Lyzer 24 (Mo Bio Laboratories, Carlsbad, CA, USA) for 3 min at 4800 rpm.DNA concentration and quality were verified based on the absorbance at 260 and 280 nm, using a DeNovix DS (Thermo Scientific, Waltham, USA).
The V3-V4 hypervariable region of the 16S rRNA gene was amplified using primers 341 F and 785 R. Illumina sequencing adapters and dualindex barcodes were added to the amplicon, using a limited-cycle PCR that included an initial denaturation step at 95 ºC for 3 min, 15 cycles of a denaturation step at 95 ºC for 30 s, an annealing step at 55 ºC for 10 s, an extension step at 72 ºC for 45 s, and a final extension at 72 ºC for min.Following, a clean-up step was performed using the AMPure XP beads (Beckman-Coulter, Krefeld, Germany) to remove free primers and primer-dimer species from amplicons.A second PCR to attach the specific Illumina multiplexing sequencing primers and index primers, was performed.Thermal cycling included an initial denaturation step at ºC for 3 min, 8 cycles of a denaturation step at 95 ºC for 30 s, an annealing step at 55 ºC for 30 s, an extension step at 72 ºC for 30 s, and a final extension at 72 ºC for 5 min.
These PCR products were verified by gel electrophoresis, purified using the Promega Wizard PCR clean-up kit (Promega, Madison, WI, USA) following the manufacturer's instructions and quantified with the QuantiFluor dsDNA System kit (Promega, Leiden, The Netherlands).High-throughput amplicon sequencing of the V3-V4 hypervariable region (Klindworth et al., 2012) was performed with the Illumina MiSeq platform according to the manufacturer's guidelines at LGC Genomics GmbH (Berlin, Germany).

Computational and statistical analysis
A DADA2 pipeline 1.8 was used for merging forward and reverse reads, quality filtering, sequence joining, and chimera removal (dada2 package v. 22.0.0)(Callahan et al., 2016) with the default parameters.Reads were trimmed at 240th nucleotide in both forward and reverse reads after quality visual inspection.Silva database v138 with the species level classification was included in the pipeline for the taxonomic classification (Quast et al., 2013).Resulted taxonomical tables were processed in Rstudio environment (RStudio Team 2020) using phyloseq package (McMurdie and Holmes, 2013) among others.
Alpha diversity indexes were calculated after rarefaction to the minimum reads (7283) using phyloseq package.Then, Kruskal-Wallis and pairwise Wilcox tests were performed to assess the differences between treatment groups.The effect of the treatment and other variables in the overall structure of the microbiota was analysed based on the Bray-Curtis distance and permutational multivariate analysis of variance (Adonis test).These potential differences in beta-diversity were visualized also by canonical Correspondence Analysis (CCA) and principal coordinate analysis (PCA) on the cantered log ration (CLR) transformed data at amplicon sequence variant level (ASV) using ggplots (Wickham, 2016), ggordiplots (Quensen, 2021), FactoMineR (Lê et al., 2008), FactoExtra (Kassambara and Mundt, 2020) and microbiome package (Lahti and Sudarshan, 2017).
Taxonomic tables were filtered for the differential abundant analysis removing those ASV with less than 3 reads in at least 10 % of the samples.Kruskal-Wallis test after CLR transformation with a false discovery rate (FDR) adjustment for multiple comparisons (referred as q in the text) was conducted to assess differences between donors and treatments.Significant differences between taxa using univariate analysis was performed using the MicrobiomeAnalyst online platform (Chong et al., 2020;Dhariwal et al., 2017).The 16S rRNA gene sequence data is available through NCBI Sequence Read Archive Database under project accession number BioProject ID PRJNA844548.

Arsenolipids modified the bacterial activity in the mucus and lumen compartments
Considering all 4 donors together, a different time-dependent effect of arsenolipids on short chain fatty acids was observed in luminal and mucosal compartment (Fig. 1).
When analyzed individually, we observed a different sensitivity to arsenolipids depending on the donor and gut compartment.Information on individual SCFAs profiles is presented in Fig. S1 (t = 70 h) and Tables S1-S4 (all time points).After 70 h, no significant differences in SCFAs profile were found between treatment in all four donors in the luminal compartment.In the mucosal compartment, the SCFA production increased after both arsenolipids exposure in donor 1 and donor 4 compared to the control, whereas AsFA 362 exposure inhibited the SCFA production in donor 2 and AsHC 332 exposure inhibited the SCFA production in donor 3 (Fig. S1).

Arsenolipids modulate gut microbiota structure in a donor-dependent manner
The authors want to highlight the descriptive nature of the results, due to the limited number of donors and the high interindividual variability.
Overall, arsenolipids exposure had no significant effect on general microbiota profile.In terms of alpha diversity indices, no differences were found between groups not in terms of diversity (p = 0.370) nor microbial richness measured as Chao1 index (p = 570).This is probably due to the high variability between donors (Fig. 2 A-B).Nevertheless, control samples showed slightly higher diversity than the other groups (Supplementary Fig. S2 A-B) with higher effect of the treatment in the samples from the mucus compared to the ones from the lumen, especially in terms of microbial richness (Chao1 index, Fig. 2A).Similarly, microbiota structure was no affected by arsenolipids exposure whereas donor (R2 = 0.53, p < 0.001), time (R2 = 0.16, p < 0.001) and compartment (R2 = 0.02, p = 0.022) were the main contributors to the differences observed in microbiota profiles (Adonis test based on Bray-Curtis distance).The interindividual effect of the donor was the main determinant grouping the samples.The donor effect was confirmed by the CCA plot (Fig. 2C; p = 0.001) and no significant model was observed when only the treatment is considered (Supplementary Fig. S2 C).
We then analyzed specific differences between groups at phylum and genus level.Considering mucus and luminal samples and arsenolipids treatments together, no significant differences were observed between the controls and the arsenolipid-treated samples.However, when both locations are studied independently, samples from the control groups showed a hardly significant enrichment in Firmicutes phylum (p = 0.052, q = 0.471) while lumen samples showed less divergence (Fig. 2D).
At genus level, the differences between donor did not enable to distinguish the potential effect of the arsenolipid treatments.However, Faecalibacterium genus (p = 0.038, q = 0.99) and Lachnospiraceae UCG-004 group (p = 0.050, q = 0.99) were reduced by arsenolipids treatment (Fig. 3A).The differences between genera in controls and treated samples seemed to be more obvious in the mucus location (Fig. 3B), where Senegalimassilia (p = 0.019, q > 0.99) and Faecalibacterium (p = 0.038, q > 0.99) genera were reduced in arsenolipid treatments compared to control samples.
The effect of arsenolipid exposure was time and compartment dependent in each donor.In donor 1, we observed an effect of AsFA at time 22 h, whereas this effect disappeared at 70 h.In contrast, treatment effect was observed at later time point (70 h) for both arsenolipids, suggesting an initial effect of arsenolipids in mucus (Fig. 4B).Peptrostreptoccocus was enriched in the control samples compared to arsenolipids (AF and AHC samples analyzed together) (p < 0.006, q = 0.456) and also when they were considered separately (Table 1).In contrast, Lachnospiraceae_UCG-004 group was enriched in control and AsFA treated samples but was less relatively abundant in AsHC samples (Table 1).Despite the lack of significance, Faecaliabacterium was reduced in mucus samples from donor 1 after arsenolipid treatments (Supplementary Fig. S3).Donor 2 showed higher distances between control and treated samples, in both compartments and studied time points, suggesting a higher sensitivity towards arsenolipids exposure at initial time points (Fig. 4B).Lachnospiraceae_UCG-004 group was reduced by both arsenolipids treatments together compared to control (p = 0.007, q = 0.587) (Table 1).In donor 3, the effect of arsenolipids was observed in lumen and mucus compartments at 70 h (Fig. 4B) and had less impact than in other donors, with only Parabacteroides genus showing a p < 0.05 (q = 0.966).Finally, microbiota from donor 4 displayed a clear response to arsenolipids, different to the other donors, with changes observed at 22 h and 70 h, especially after AsFA exposure (Fig. 4B).Lachnospiraceae_UCG-004 abundance was significantly decreased by both AsHC and AsFA treatment as compared to the control group (p = 0.004, q = 0.371), with the highest reduction observed for AsFA (Table 1).Other genera were also affected by the arsenolipids exposure in donor 4, especially after AsFA treatment, such as Faecalibacterium (p = 0.022, q = 0.218) and Subdoligranulum (p = 0.024, q = 0.218) which were enriched in both control and AsHC samples compared to the AsFA group and Escherichia/Shigella genus (p = 0.018, q = 0.218) which was enriched in AsFA treated samples.Interestingly, in this donor, samples from AsFA group showed a significantly higher relative abundance of Proteobacteria phylum (p = 0.025, q = 0.092) and Fusobacterium (p = 0.028, q = 0.092) genera and a reduction in Firmicutes phylum (p = 0.031, q = 0.092) (Table 1).
In general, Lachnospiraceae-UCG004 group was among the most prevalent genus affected by the arsenolipids treatment when both AsHC and AsFA are studied together or independently, with differences among donors (Table 1).

Discussion
This study showed for the first time that two different arsenolipids, AsHC 332 and AsFA 362, can induce a functional dysbiosis and affect key gut commensal levels in a donor-dependent way.
We used a single-dose approach to simulate the intake of arsenolipids via fish consumption in a single meal and investigated how this could affect the gut microbiota.The dosage of arsenolipids was calculated based on previous reports on AsHC and AsFA content in edible fish (Stiboller et al., 2017;Al Amin, 2018;Taleshi et al., 2014).The highest dose reported for AsHC was selected to estimate arsenic intake and AsFA was maintained at the same levels to allow for comparison.Assuming a worst-case scenario with a consumption of 150 g of skipjack tuna (Katsuwonus pelamis) (0.8 µg arsenolipids/g) (Stiboller et al., 2019) and 150 g of whitebait (0.241 µg AsHC/g) (Al Amin, 2018), arsenolipids intake could be around 160 µg.Considering between 50% of intestinal absorption from in vitro and human studies (Meyer et al., 2015b), 80 µg of arsenolipids could reach the colonic environment, which is in the range of our dose.Previous studies have reported lower levels of arsenolipid intake in Japanese populations and higher intestinal absorption (Al Amin, 2018;Schmeisser et al., 2006), but we assumed a high consumption of fish in a single dose to maximize the effect of the arsenolipids.Other concentrations and dosage strategies (e.g., arsenolipids present in fish oils or muscle, chronic exposure) should be tested to better understand arsenolipid interactions with food, gut microbiota and human health.Capability of an in vitro model to predict in vivo effects is highly influenced by the dose strategy (Groothuis et al., 2015).We evaluated the total nominal concentration of arsenolipids, but also the microbial cell-bound arsenic (Xiong et al., 2022a), which is considered the concentration at the site of action (gut environment) and on target cells (microbial cells).
Exposure to arsenolipids induced changes in SCFAs profiles in the M-SHIME, with variable trends depending on each individual.The high inter-individual variability might be linked to baseline microbiota differences, affecting arsenolipids biotransformation and distribution in different M-SHIME compartments (Xiong et al., 2022a).Concretely, we observed the highest effect of arsenolipids on microbial structure in donor 4, who was reported to have a higher retention of the toxicant in bacterial cells (Xiong et al., 2022a) and higher levels of steroid arsenolipids and arsenic-fatty esters derived from AsHC332 and AsFA362 metabolism.This observation suggests a dose-dependent effect based on the tropism of arsenic towards specific microorganism, which could contribute to a different sensitivity in developing detrimental effects.Additionally, metabolic potential of gut microbiota towards arsenolipids could generate novel molecules with unknown toxicity (Xiong et al., 2022a), which might have higher toxicity than parental compounds (i.e., bioactivation).Further studies are proposed to assess the toxicity of arsenolipid-derived metabolites on gut microbiota inhabitants and the host.
Time-course dependent analysis of different metabolites indicated a temporal effect from the AsFA and AsHC towards the microbiome with the highest impact observed at 22 h post exposure.This observation suggests that the microbial community and functionality recovers once the arsenic species are washed out.Due to the configuration of the M-SHIME system, arsenic levels in the bioreactors were reduced during the wash-out periods, mimicking the elimination of the toxicant after a single intake in vivo.However, continuous consumption of arsenolipidcontaining foods could chronically disturb SCFA production and permanently shift microbial structure and host metabolism, as previously suggested for iAs exposure (Gokulan et al., 2018).Previous research showed a good correlation between in vitro/in vivo chronic effect of low doses of chlorpyrifos on the SHIME system and rat gut microbiota, reporting similar trends in bacterial population changes and dysbiosis induction during periods of 30 and 60 days, respectively (Joly et al., 2013).
The effect of arsenolipids was significant in the luminal compartment.Butyrate was one of the most sensitive metabolites, showing a reduction in different donors and during the time course of the assay.Butyrate is a key mediator of host-microbe crosstalk, affecting host energy metabolism, intestinal barrier, and immune system homeostasis (Schulthess et al., 2019).It is a fuel source for coloncytes (Donohoe et al., 2011) and reduced levels of butyrate and/or butyrate-producing bacteria have been associated with different pathologies, including intestinal inflammatory conditions (Machiels et al., 2014) or chronic kidney disease (Wang et al., 2019).Consistently with our results, iAs exposure in a mice model (10 ppm, 4 weeks) affected the gut  Differences were assessed by Kruskal-Wallis test on the centered-log-ratio (CLR) transformed data.Those genera that showed a difference between conditions with a pvalue < 0.25 are reported.Microbiota data is expressed as median (interquartile range).
microbiome community and disturbed its metabolic profiles (Lu et al., 2014).Similarly, Chi et al. observed structural and functional changes in the gut microbiome after iAs exposure (100 ppb, 13 weeks) (Chi et al., 2017).Lu et al. revealed that iAs exposure in vivo induced changes in Faecalibacterium spp., which was also reduced by AsFA 362 treatment in this study in donor 4 (Lu et al., 2014).Faecalibacterium prausnitzii is one of the most studied butyrate-producers in the human colon, and it has been proposed as a bioindicator of human health (Ferreira-Halder et al., 2017).Faecalibacterium has been proposed as a gut commensal with a protective role against arsenic toxicity (Coryell et al., 2018(Coryell et al., , 2019)), whereas negative correlations have been found between F. prausnitzii levels and inflammatory bowel disease or colorectal cancer (Ferreira-Halder et al., 2017;Sokol et al., 2008Sokol et al., , 2009)).
We also observed consistent reductions in Lachnospiraceae_UCG-004 induced by AsHC and AsFA in all four donors.Lachnospiraceae is a microbial group with a positive correlation with concentrations of acetate, butyrate, and total SCFAs and it has been described as a common gut commensal with multiple enzymatic capabilities (Vacca et al., 2020).Concretely, Lachnospiraceae_UCG-004 was increased after an intervention study with prebiotic supplementation (i.e., chitin glucan) (Rodriguez et al., 2020), had decreased relative abundance in a meta-analysis of Parkinson patients and it was also related to constipation in the same cohort (Nishiwaki et al., 2020), suggesting a contributing role in ecosystem homeostasis and health.
Other key microbial members affected by arsenolipids exposure were classified as Escherichia/Shigella and Fusobacterium, mainly increased by AsFA 362 treatment.Members of these groups are described as potential pathogens and are related to pathological dysbiosis (Chen et al., 2022;Engevik et al., 2021;Gao et al., 2018;Zhu et al., 2018).
Remarkably, we also observed that AsFA reduced luminal valerate.Valerate has been shown to decrease vegetative growth of Clostridium difficile and patients receiving a fecal microbiota transplant have displayed restored valerate production (McDonald et al., 2018).Valerate was also positively correlated with gut microbial richness in healthy lean adults (Tap et al., 2015).It could therefore be considered a potential marker for arsenolipid-induced functional dysbiosis, togheter with butyrate.
In the mucosal environment, the increase of acetate at 22 h upon AsHC treatment could be explained by AsHC accumulation in mucus, thus providing an extra carbon source to the microbial community, but also due to a block of acetate conversion to butyrate by possibly inhibiting different enzymes (e.g., phosphate butyryltransferase, butyrate kinase, butyryl-CoA: acetoacetate CoA-transferase, and butyryl CoA: acetate CoA transferase (Si et al., 2018)) or by reducing acetyl coenzyme A production levels (Kulshrestha et al., 2014).Water-soluble arsenicals, including As(V), As(III), monomethylaresnic (MA) and dimethylarsenic (DMA) have been reported in mucus and bacterial samples of AsHC and AsFA treated M-SHIME (Xiong et al., 2022a).Trivalent arsenicals have a high affinity for sulfhydryl groups and can bind to reduced cysteines in peptides and proteins, inhibiting up to 200 enzymes (Shen et al., 2013).Acetate kinase and 3-hydroxy butyryl-CoA dehydrogenase have been reported at significantly lower levels in the gut microbiome of arsenic-exposed mice (Chi et al., 2017), supporting our hypothesis that arsenolipid-derived metabolites in the mucosal compartment could influence microbial enzymatic machinery for SCFA production.However, further mechanistic studies are required to provide direct evidences.
The shift between beneficial commensals and potential pathogens observed by arsenolipids exposure and the functional changes in gut microbiota should be further evaluated in human studies.Decrease in butyrate production ability and an inflicted dysbiosis on butyrateproducing bacteria can therefore be proposed as one of the potential mechanisms of arsenolipids toxicity and should be considered in risk assessment studies.In addition, arsenolipids exposure via contaminated food items in specific populations (i.e., inflammatory bowel disease patients) could worsen the pathological condition, already characterized by a dysbiosis, and might be considered in personalized nutrition approaches.
Of importance, AsHC is known to be transferred to human milk (Stiboller et al., 2017;Xiong et al., 2020).AsHCs massively altered the neuronal network, induced apoptosis in differentiated human brain cells, and exerted pronounced neurodevelopmental effects on pre-differentiated and fully differentiated human neurons (Witt et al., 2017a(Witt et al., , 2017b)).Our findings suggest another mechanism of action for AsHCs toxicity, likely affecting the establishment and development of the infant gut microbiota during a critical window of susceptibility.This requires further studies to evaluate potential risks from AsFA and AsHC exposure in sensitive populations that are currently not yet accounted for.
Arsenic toxicity has been demonstrated to be highly variable among individuals exposed to equivalent doses, attributed to genetic polymorphisms, epigenetics or microbiota-mediated processes (Bailey et al., 2013;Hernández and Marcos, 2008;Roggenbeck et al., 2021;Yin et al., 2017).We observed a significant effect of basal microbiota on arsenolipids effects, likely due to differences in arsenolipid distribution and accumulation in bacterial and mucosal compartments (Xiong et al., 2022a) that might induce different structural and functional shifts in the gut ecosystem.We cannot directly infer the consequence of these changes on human health, but propose the gut microbiota as a "target organ" of arsenolipid toxicity.
Limitations of this study are related to the reduced number of donors, single-dose, use of 16S rRNA sequencing instead of shot gun metagenomics and limited number of analysed metabolites.Further studies are proposed to include metagenomics and metabolomics analyses, long-term repeated-dose assays, and the incorporation of contaminated food matrices for better representing real-exposure scenarios.Human studies including arsenic-contaminated food items would provide a deeper understanding of detrimental effects of arsenolipids intake.
As strengths of the study, we used purified aresnolipid molecules to ensure the observed effect was caused by the toxicant, included the mucosal and luminal gut compartment with accurate characterization of bacteria, luminal and mucosal-bound arsenic (Xiong et al., 2022a) and provided novel information on potential arsenolipid effect on human gut ecosystem.To our knowledge, this is the first time reporting the effects of arsenolipids on human gut microbiota structure and function.Further, using personalized models with ex-vivo human samples from different donors provided useful information about interindividual responses to arsenolipids exposure.

Conclusion
This study demonstrated that arsenolipids could influence the structure and function of the gut microbiota, even after a single dose at environmentally relevant doses.These results provide novel data on the potential effect of arsenolipids on gut microbiota.The effect of arsenolipids on the simulated gut microbial ecosystem was donor-dependent, with alterations at a functional level, and in a lesser extent at structural level.These findings should be considered in risk assessment studies and we propose to include gut microbiota as a "target organ" of arsenolipids toxicity.Butyrate reduction and changes in key microbial commensals suggest a dysbiotic effect of AsHC and AsFA, with unknown consequences for human health.
Further research using a repeated-dose approach and food-matrix effect would be needed to evaluate chronic effects and confounding factors affecting the translation to a real-life scenario.Until obtaining more data, we propose the precautionary principle for arsenolipids exposure, reducing as much as possible the intake of contaminated food items especially in sensitive populations such as pregnant women, infants and diseased populations.Risk-benefit analysis of consumption of arsenolipid-containing foods, like fish or seaweed, should be further evaluated.

Fig. 1 .
Fig. 1.Effect of arsenolipids on short chain fatty acid profiles.(A) Dot plots represent mean ± standard error of the mean (SEM) of acetate, propionate, butyrate, valerate and total branched chain fatty acids (BCFA) in luminal compartment of control, AsHC and AsFA exposures at different time points of the experiment (n = 4 donors).Significant differences between the arsenolipid-treated samples and the control sample are marked with an x symbol.(B) Bar plots represent mean ± SEM of acetate, propionate, butyrate, valerate and BCFA in luminal compartment of control, AsHC and AsFA exposures (n = 20, corresponding to 4 donors and 5 time points per donor).(C) Bar plots represent mean ± SEM of acetate, propionate, butyrate, valerate, and BCFA, in mucosal compartment of control, AsHC and AsFA exposures (n = 4 donors).Significant differences between the arsenolipid-treated samples and the control sample are marked with an asterisk (* p < 0.05; ** p < 0.01).AsHC represents AsHC 332 exposure, AsFA represents AsFA 362 exposure.

Fig. 2 .
Fig. 2. Effect of arsenolipids treatment on the fecal microbiota.A-B) Effect of exposure on alpha-diversity of the fecal microbiota measured as Shannon (A) and Chao1 index (B).Significance of the differences were assessed by Mann-Whitney test.C) Canonical correspondence analysis (CCA) showing the distribution of the samples according to treatment.Donor and treatment were included in the model.D) Composition of fecal microbiota at phylum level according to arsenolipids treatment and compartment.

Fig. 3 .
Fig. 3. Effect of arsenolipids on the microbial genera relative abundance considering both compartments together (A) and only in the mucus compartment samples (B).Significance was assessed by Mann-Whitney test after CLR transformation.Only those genera with a p-value < 0.2 were plotted.

Table 1
Effect of different arsenolipid treatment on microbiota composition at genus level in each donor.Median values of selected taxa and interquartile range (IQR), in brackets.