Feasibility study and direct extraction of endogenous free metallic cations combining hemodialysis and chelating polymer

In this article, we report the conception and the use of dialysis-based medical device for the extraction of metals. The medical device is obtained by addition in the dialysate of a functionalized chitosan that can chelate endogenous metals like iron or copper. This water-soluble functionalized chitosan is obtained after controlled reacetylation and grafting of DOTAGA. Due to the high mass of chitosan, the polymer cannot cross through the membrane and the metals are trapped in the dialysate during hemodialysis. Copper extraction has been evaluated in vitro using an hemodialysis protocol. Feasibility study has been performed on healthy sheep showing no acute toxicity througout the entire dialysis procedure and first insights of metallic extraction even on healthy animals.

. In sepsis conditions, over-production of reactive oxygen species and reactive nitrogen species is observed both in the blood circulation and in damaged organs 15 . Elevated free iron and transferrin saturation in serum have been recently associated with an increased mortality in patients with sepsis in a retrospective study on 1891 patients 16 . A dose dependent increase in the risk was found as the iron level increase 17 . Impact of iron has also been emphasized in chronic liver diseases with higher iron levels for patients with alcoholic liver diseases, nonalcoholic fatty liver disease and hepatitis C viral infection 18 . For acute-on-chronic liver failure, higher incidence of multiorgans failure and consequent mortality is observed for patients with dysregulated iron homeostasis 19 . Kidney is another organ at risk for iron dysregulation 20 . Acute kidney injury is a relatively common disease associated with high mortality; it can be seen in 5-7% of hospitalized patients and in 57% of patients in intensive care units. Clinical studies have shown that higher plasma concentration of catalytic iron is correlated with significantly greater risk of deaths 21 . Level of catalytic iron is an important factor during cardiac surgery. After surgery, patients with higher levels display greater odds of acute kidney injury, mortality and postoperative myocardial injury 22 . Rhabdomyolysis, a muscular disorder, can also lead to acute kidney disease and in the most severe forms to death for a large fraction of patients 23 . During rhabdomyolysis, myoglobin is massively released from the muscle tissues and is directly associated with increase of oxidative stress 24 .
All these examples emphasize the importance of extracting the excess of endogenous loosely bounds metals from blood in some patients. Even if chelating drugs are highly efficient to treat the majority of the metal overloads, in some specific cases, due to side effects or insufficient efficacy of chelating molecules, metal overloads can't be treated adequately leading to life-threatening situations in particular for acute treatment in intensive care units where release of redox active metals is a critical parameter. The aim of this article is to propose the concept of a safe medical device capable of extracting excess of toxic endogenous metals from blood with a particular focus on copper for the treatment of Wilson disease. It is based on a simple combination of hemodialysis (HD) and chelating polymer solution placed in the dialysis fluid (Fig. 1).
The chelating polymer is a reacetylated chitosan functionalized by DOTAGA ligands (Fig. 2A). The average molar mass is sufficient to prevent passage through dialysis membrane and DOTAGA functionalization offers the capacity to efficiently chelate excess free (or loosely chelated) iron, copper or zinc ions. Dialysis is performed in recirculation conditions in order to use a small volume of dialysate to avoid strong dilution of important molecules but permits to concentrate the metals by an important factor. Herein, we present the new dialysate formulation concept and we evaluate the safety and the efficacy for endogenous metal extraction on a large animal easily transposable to human: healthy sheep.

Results
Characterization and functionalization of chitosan. Functionalized chitosan (Chitosan@DOTAGA) has been synthesized by a two-steps process. The first step consisted in the reacetylation of initial chitosan by addition of pure acetic anhydride in a water/1,2-propanediol mixture 25 . The degree of acetylation is determined by 1 H NMR and confirmed by IR spectrometry and is close to 30 ± 2% (x = 0.3) (See Supporting Information Figs. S1, S2). The second step of the synthesis is the functionalization of acetylated chitosan by DOTAGA anhydride (1,4,7,10-tetra-azacyclododecane-1-glutaric anhydride-4,7,10-triacetic acid) that is directly added to the precedent solution of reacetylated chitosan in water/1,2-propane-diol. Then, purification was performed by tangential filtration (cut-off: 100 kDa) and the resulting modified chitosan was freeze-dried. Purity (98%) of the product was assessed by SEC (size exclusion chromatography) coupled with UltraViolet-Visible detection   2B). A retention time of 6.7 min is observed for polymer (Chitosan@DOTAGA) and a minor peak at 11.3 min is observed for ungrafted DOTAGA. The efficiency of functionalization was assessed by dosage, after addition of small amounts of Cu(II) on a DOTAGA-modified chitosan solution at 1 g L −1 . For each addition of copper, UV absorbance at 295 nm was plotted using Chitosan@DOTAGA peak area (Fig. 2C). Thus, hypothesizing that copper ions will only interact with DOTAGA moieties in acidic conditions (See Supporting Information Fig. S3 comparing Acetylated chitosan and Chitosan@DOTAGA), for a solution of 1 g L −1 of Chitosan@DOTAGA, a molar concentration of 0.35 mM of DOTAGA has been obtained corresponding to y = 0.072 (See Supporting Information). Using 1 H NMR, x = 0.28 has been determined on the reacetylated chitosan sample (See Supporting Information Fig. S1). The acetylation degree (x) is obtained through the ratio between the area of the peak at 3.1 ppm over the area in the region 2.9-4.1 ppm following the equation: The elemental analysis generally confirms the results found by HPLC-UV and 1 H NMR. The theoretical ratio C/N for chitosan, acetylated chitosan (acetylation degree 28%) and chitosan@DOTAGA is respectively 5.14, 5.54 and 5.46. The experimental results show a C/N ratio of 5.11, 5.96 and 5.47. The slight mismatch for the acetylated chitosan could be explained by the presence of traces of 1,2 propanediol found also in the 1 H NMR spectrum, which increase the total quantity of C.

Efficacy of metal extraction in vitro.
Copper ions extraction efficacy was first tested in vitro using a HF20 dialyzer with intermediate cut-off membrane (~ 10-20 kDa) in a hemodialysis protocol (Fig. 3A). The copper II solution was placed in the blood compartment at a concentration in copper of 100 ppb and different blood flow rates were tested (QB). In the dialysate compartment, Hemosol B0 alone was compared to Chitosan@ DOTAGA at a concentration of 1 g L −1 in Hemosol B0 and a fixed dialysate flow rate (QD) of 250 mL/h was used. For each tested flow rate, copper extracted per hour was significantly higher (p value < 0.001) in presence of Chi-tosan@DOTAGA (Fig. 3B). Median values for copper extracted per hour varied respectively from 29 to 38 µg/h and from 6 to 19 µg/h, in presence and in absence of Chitosan@DOTAGA. Copper extraction was confirmed by analysis of copper content in the effluent that is significantly higher for Chitosan@DOTAGA in comparison with   www.nature.com/scientificreports/ blood flow rate constant (QB = 50 mL min −1 ) was also tested (Fig. 3E). The presence of Chitosan@DOTAGA in Hemosol B0 leads to an increased extraction of copper at low and moderate QD (p value < 0.01 for QD = 50 mL/h and p value < 0.001 for QD = 250 mL/h) but no statistical difference is observed at high QD.
Feasibility and metallic extraction on healthy sheep. Prior to evaluation on animals, compatibility of Chitosan@DOTAGA with dialyzers membranes has been tested and integrity of membrane has been verified by Scanning Electron Microscopy (SEM) (See Supporting Information Fig. S5). After 4 h of dialysis with Hemosol B0 and Chitosan@DOTAGA on blood, no alteration of the membrane was observed. Further, hemodialysis has been performed on 4 sheep combining different dialyzers membranes and Chitosan@DOTAGA to evaluate the impact of the membrane on safety and metallic extraction efficacy (Fig. 4).
For these experiments, two high cut-off (HCO, 45-60 kDa) membranes (Theralite and SepteX) have been compared with standard dialyzer (POLYFLUX H6, 15-20 kDa cut-off membrane). The protocol of the experiment is described in Table 1. All sheep were submitted to a first hemodialysis treatment with Theralite and a second treatment either with Septex or Polyflux H6.
Hemodialysis (HD) was performed in dialysate recirculation conditions with sessions of 4 h and a volume of dialysate of 1 L (See Supporting Information Fig. S6). This protocol was applied to limit the diffusion process of blood components while maintaining the blood copper depletion thanks to copper sequestration in the Chi-tosan@DOTAGA polymer. The dialysis protocol was safe, each animal presented normal behaviors and stayed in healthy conditions post-dialysis (feed, drink, rumination and mobility). Blood analysis (n = 28) were performed one day before the first dialysis session (DAY 0), before and after each dialysis session and 4 (DAY 7) and 7 (DAY 10) days after the last dialysis session. The summary of the full blood analysis (cell blood count, liver enzyme and inflammation) is presented in Supplementary Tables S1 to S3. The evolution of blood parameters according to the samples type (before HD, after HD, N day(s) after HD) per dialysis membranes (Theralite, POLYFLUX H6, SepteX) have been tested by non-parametric Kruskal-Wallis tests, followed by Dunn tests as post-hoc when significant. Means, standard deviation and percentages of deviation from the "before HemoDialysis (HD)" samples were also computed for both "after HD" and "N day(s) after HD" samples. Figure 5 displays the results for a subset   Values "before HD" were melted with D0 values to consider the variability within sheep (See column HD with Theralite) or to mitigate the effect of the Theralite HD session in the analysis of both POLYFLUX H6 (sheep A/B, DAY 3) and SepteX (sheep C/D, DAY 4) HD sessions. Note that "before HD" samples of both POLYFLUX H6 and SepteX HD sessions were taken after HD with Theralite and one full resting day for all sheep. Values "after HD" correspond, for all membranes, to the sampling realised at the end of each corresponding HD session. Values "N day(s) after HD" correspond, for the Theralite membrane, to the "before HD" samples of both POLYFLUX H6 and SepteX HD sessions (one full resting day); and to the "DAY 7" and "DAY 10" samples for both POLYFLUX H6 and SepteX membranes (4 to 7 full resting day). The existence of a significant change in abundance with samples categories (before HD, after HD, N day(s) after HD) was tested with a non-parametric Kruskal-Wallis tests. Only the sodium, the albumin and the red blood cells were significantly different for the HD with Theralite membrane (p value < 0.05). www.nature.com/scientificreports/ of blood parameters. No significant differences were observed for the POLYFLUX H6 and SepteX membranes on this subset of data while sodium (mmol/L; p value = 0.034 for "after HD" versus "N day(s) after HD"), albumin (g/L; p value = 0.023 for "before HD" versus "N day(s) after HD") and red blood cells (10 12 /L; p value = 0.012 for "before HD" versus "N day(s) after HD") undergoes a significant but small change with the Theralite membrane. Total proteins (g/L; p value = 0.034 for "before HD" versus "N day(s) after HD" and p value = 0.025 for "before HD" versus "after HD") and red blood cells distribution width (%;p value = 0.049 for "before HD" versus "N day(s) after HD") also undergoes a small significant change with the Theralite membrane. With the POLYFLUX H6 membrane, the gamma glutamyl transferase (UI/L; p value = 0.024 for "before HD" versus "N day(s) after HD") and the monocytes (10 9 /L: p value = 0.04 for "before HD" versus "after HD") undergo a significant change, finally, with the SepteX membrane, only the gamma glutamyl transferase (UI/L; p value = 0.038 for "before HD" versus "N day(s) after HD") undergoes a significant change. Differences in the number and nature of significantly shifted parameters with the Theralite membrane compared to the POLYFLUX H6 and the SepteX membrane can be explained accounting several aspects (i) the "N day(s) after HD" samples were taken after only 1 resting day for the Theralite membrane in comparison with 4-7 days for both POLYFLUX H6 and SepteX membranes (e.g., albumin, total protein, red blood cells) (ii) the number of available data was higher with the Theralite membrane allowing a slightly better statistical treatment (e.g., sodium, Red blood cells distribution width: no more than 5% of deviation compared to "before HD", see Supplementary Table S2) (iii) the surface of the Theralite membrane is higher compared to both POLYFLUX H6 and SepteX membranes (2.1 m 2 versus 0.2 m 2 and 1.1 m 2 ) which favors the diffusion process. Indeed, when looking at the percentage of deviation from the "before HD" samples, both albumin and total protein decreased by 10 to 16% for all membranes in the "after HD" samples. After one resting day, no restauration in albumin and total protein can be observed with the Theralite membrane while after 4 to 7 days, both albumin and total protein were decreased by only 1 to 5% compared to "before HD" samples with both POLYFLUX H6 and SepteX membrane. The same observation was made with the red blood cells with a decrease of 7 to 11% recorded for all membranes in the "after dialysis" samples, no restauration observed after one resting day with the Theralite membrane, and no more than 4.5% decreased after 4 to 7 days in comparison with the "before HD" samples for both POLYFLUX H6 and SepteX membranes. The moderate decreased in red blood cells can be explained by residual volume of blood in haemodialyser lost on return, and via hemolysis which is frequent for patients treated with HD; it was accompanied with non-significant (but consistent across membrane) decreased of hemoglobin (5-8%) and hematocrit (8%). Loss of albumin (and thus total protein by passage through the membrane) was expected with both HCO membranes while not with POLYFLUX H6 membrane. It can be noted that 16% of albumin variation correspond to a change of only 4 g/L with a "before HD" value ranging from 23.8 to 25.5 g/L. According to Morgera et al. 26 who observed an increase of albumin loss when comparing 1000 mL/h to 2500 mL/h flow rate (from 410 to 950 mg/h which correspond to 1.6 to 3.8 g in 4 h), we can assume that the recirculation protocol limits the loss of albumin with HCO membranes (experimental dialysate flow rate = 6000 mL/h). Metallic extraction has been assessed for endogenous metallic cations (iron, copper and zinc) by measuring their quantities by ICP-MS in dialysate, but also in whole blood and plasma for the three different types of dialyzers membranes (POLYFLUX H6, Theralite and SepteX). In dialysate, samples were taken before dialysis, after 1, 2, 3 and 4 h of dialysis ( Fig. 6 and Table S4). For the three metals (Table S5), HCO membranes presented the same behavior with clear and important increase of the quantity of metal with time. In terms of efficacy, SepteX membrane can extract close to 0.2 mg of copper and 0.5 mg of iron on a healthy animal, in 4 h. On the other hand, POLYFLUX H6 membrane leads to negligible extraction for zinc and relatively small one for iron and copper. These results have been compared with metal analysis in total blood (See Supporting Information Fig. S6) and in plasma (See Supporting Information Fig. S7). For iron, only plasma measurements have been taken due to too large quantities of iron in blood cells. These trends are coherent with metallome study in dialysate especially for SepteX where an important decrease of metallic level is observed in plasma (See Supporting Information Fig. S8).

Discussion and conclusion
Extraction of metals is a requirement for pathologies where excess of metals is recognized to be one of the hallmarks of the pathology like Wilson disease or primary and secondary hemochromatosis. For these pathologies, the current treatments include copper chelators for Wilson disease or simple phlebotomy or iron chelators for hemochromatosis. These treatments are relatively well adapted for chronic pathology but cannot avoid neurological worsening for certain patients in Wilson disease and are also not adapted for rapid metal purification for patients that suffer burst of free metal liberation as it can occur in intensive care units due to secondary effects occurring at high concentration of chelates. On the other hand, dialysis is already currently used in intensive care units. Developing a medical device for rapid extraction of metals using dialysis can be easily adapted in the medical practice for those patients. Combining dialysis with chelating polymer is a new concept that is based on the chelation in the dialysate of free metals that leads to an overconcentration of metals in the dialysate that cannot go back to the blood circulation due to the size of the polymer and the cut-off of the membrane. It permits to have effective concentration of metals with recirculation avoiding using large volume of dialysate.
Chitosan has been chosen as a functional macrochelatant for this application because (i) it can be easily functionalized on the primary amines moieties, (ii) it is biocompatible, (iii) it is soluble at physiological pH after partial reacetylation (DA ~ 30%) and DOTAGA grafting, (iv) it can exhibit high molecular mass, sufficient not to pass through dialysis membrane and (v) it is affordable as clinical grade and can be obtained at large scale. DOTAGA (1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid) has been chosen as chelate for this application due to its high complexation constants for Cu 2+ , Fe 3+ and Zn 2+ (log β = 22.44, 29.4 and 20.52 respectively) 27 . For the functionalization on chitosan, DOTAGA anhydride has been chosen due to its additional reactive site to avoid losing a carboxylic acid group implicated in the coordination of metals 28 It has been shown in vitro that using standard dialyzer HF20, important copper extraction can be reached at moderate dialysate flow rate by increasing blood flow rate. A feasibility study has been performed on healthy sheep comparing 3 different types of dialyzers membrane: POLYFLUX H6, SepteX and Theralite (i.e. one standard dialyzer and two high cut-off membranes) and recirculation conditions for 1 L of dialysate. Sheep have been chosen, as large animal model, for future translation due to their easy taking care in dialysis experiments. Interestingly, no secondary effect has been observed on the sheep with no significant modifications of hematological, immunological or liver enzymatic parameters and the sheep are still living 12 months after the experimental protocol. During dialysis sessions of 4 h with classical POLYFLUX H6 dialyzer membranes, we found a modest extraction of 10-50 µg of free iron and free copper. Using HCO membranes, about ten times higher uptakes have been evidenced with about 200 and 500 µg copper and iron extracted respectively on healthy sheep that present only small quantities of free metals in comparison with diseased ones, that is encouraging for future application like Wilson disease and further clinical translation.

Material and methods
Synthesis of Chitosan@DOTAGA . Chemicals. Medical grade chitosan of low degree of acetylation (DA) from Alaska snow crab was purchased from Matexcel (Bohemia, NY, USA, reference number NAT-0030, https:// www. matex cel. com/p/ 30/ medic al-grade-chito san/). The weight-average and number-average molar masses (respectively Mw = 2.583 10 5 g/mol, Mn = 1.323 10 5 g/mol) were determined by size exclusion chromatography coupled with refractive index and multi-angle laser light scattering measurements 30 . The initial degree of acetylation of such raw chitosan was determined by 1 H NMR spectroscopy by the Hirai method 31 close to 4.5 ± 0.5%. DOTAGA anhydride has been furnished by CheMatech (Dijon, France). 1,2-propanediol and acetic anhydride were furnished by Sigma-Aldrich (France). Acetic acid was furnished by VWR (France).  www.nature.com/scientificreports/ the product is extracted and purified by tangential filtration with water for characterization of the intermediate product.
Functionalization of acetylated chitosan. 120 g of DOTAGA anhydride are added to 2 L of the precedent solution and the mixture is stirred for 16 h. Then, the solution is diluted by a factor 10 in milli Q water and purified by tangential filtration using a 100 kDa membrane (Sartocon® Slice 200, PES membrane) with a Sartoflow® Smart apparatus. Tangential filtration is used for the two other purification steps towards acetic acid at 0.1 M and milli Q water.
Characterization by HPLC-UV using SEC columns. HPLC-SEC-UV chromatograms were recorded with an Agilent-1200 HPLC apparatus with a DAD module. The SEC column used is a PolySep-GFC-P 4000 and a 0.1 M solution of acetic acid and 0.1 M ammonium acetate is used as a buffer. The operating temperature is 30° C and the absorption wavelength is 295 nm. The eluent flow rate is 0.8 mL/min.
Infrared spectroscopy. Fourier-transform infrared spectroscopy spectra were recorded using a Shimadzu IRAffinity-1 equipped with a DLaTGS detector, using the lyophilized powder as samples and recording in a range between 400 and 4000 cm −1 with 32 scans and a resolution of 0.5 cm −1 .
Nuclear magnetic resonance. The Preparation of Chitosan@DOTAGA and Hemosol B0 only dialysates. 1 L or 900 mL of Hemosol B0 were sampled manually from the 5 L reconstituted Hemosol B0 solution and introduced in a clean dialysis bag. 1 L of Hemosol B0 were used to conduct the experiment without Chitosan@DOTAGA and 900 mL of Hemosol B0 were completed with 150 mL of Chitosan@DOTAGA (7 g/L) to obtain a concentration of 1 g/L of Chitosan@ DOTAGA.
Experimental protocol. The dialysis protocol is summarized in Fig. 3A using a Prismaflex monitor and a HF20 set (0.2 m 2 ) as dialyzer. First, the dialysis flow rate is maintained constant at 250 mL h −1 and 4 blood rates are tested for 30 min each (20 mL min −1 , 40 mL min −1 , 80 mL min −1 and 100 mL min −1 ). For each flow rate, sampling of the Cu solution in duplicate at T0 (red sampling site), T10 min, T20 min and T30 min (blue sampling site). Sampling of the dialysate/effluent at T0 (Chitosan@DOTAGA only) in the dialysis bag and T20 min (yellow sampling site).
Then, the blood flow rate is maintained constant at 50 mL min −1 and 4 dialysate flow rates have been tested: (i) 5 min at 2000 mL/h: sampling of the Cu solution at T0 (red sampling site), T2.5 min, T5 min (in duplicate, blue sampling site); sampling of the effluent at T2.5 min (yellow sampling site), (ii) 7.5 min at 1000 mL/h: sampling of the Cu solution at T2.5 min, T5 min, T7.5 min (in duplicate, blue sampling site); sampling of the effluent at T5 min (yellow sampling site), (iii) 30 min at 250 mL/h: sampling of the Cu solution at T10 min, T20 min, T30 min (in duplicate, blue sampling site); sampling of the effluent at T20 min (yellow sampling site) and (iv) 60 min at 50 mL/h: sampling of the Cu solution at T0, T30 min, T60 min (in duplicate, blue sampling site); sampling of the effluent at T30 min (yellow sampling site). Both experiments were conducted with and without Chitosan@DOTAGA. www.nature.com/scientificreports/ Animals. 4 adult male sheep weighing 44 to 56 kg were included in this study. 14-day acclimation period was implemented before the study. Animals were fed with hay ad libitum and with alfalfa pellets and given free access to water. The day before the experiment each sheep was equipped under anesthesia (intramuscular injection of xylazine (0.1 mg/kg) and midazolam (0.2 mg/kg)) with a 11.5Fr double lumen catheter (Hemo-cath®, Medical Components, Harleysville, PA) inserted in the right jugular vein with the transcutaneous Seldinger technique (Pouzot-Nevoret and al, 2017). The day of the hemodialysis, two sheep were installed together in the dialysis room to respect animal welfare. At the end of the day, the animals were placed in facility in a common box. The physiological status pre, per and post dialysis session of each sheep was evaluated through the quantification of the food and drink ingested, and the observation of rumination mobility and behavior.
Hemodialysis protocol. The dialysis is performed in recirculation using the cascade dialysis prototype from Baxter to have manual control over all the parameters during the dialysis. The following parameters have been chosen for the hemodialysis: (i) blood flow rate of 100 mL min −1 , (ii) dialysate flow rate of 100 mL min −1 , (iii) dialysate volume of 1 L, (iv) concentration of Chitosan@DOTAGA in the dialysate of 1 g L −1 and (v) 4 h of dialysis. The dialysate solution used was Hemosol B0 supplemented with potassium at 4.5 mmol/L.
Metal extraction quantification. Blood samples were collected at day 0, 7 and 10 in the morning and day 1 and 3 before and after dialysis using Vacuette tubes (NH Trace Elements Sodium Heparin 6 mL; Greiner, Kremsmünster, Autriche). ICP-MS analysis were performed using a Perkin Elmer NexION2000 equipped with Syngistix software and a ESI SC-FAST Sample Introduction in Kinetic Energy Discrimination (KED) mode. The samples are prepared by dilution in an aqueous solution of HNO 3 1% v/v.
Statistics. Both blood analysis evolution according to samples type ("before HD, "after HD", "N day(s) after HD") and dialysis bag metals evolution according to time (hours) were tested by non-parametric Kruskal-Wallis tests (krustal.test from the vegan package, Oksanen et al. 32 ), followed by Dunn tests (kwAllPairsDunnTest from the PMCMRplus package, Pohlert 33 ) for multiples comparison when significant. The ggplot2 package (Wickham 34 ) was used for plotting the data along with the stat_compare_mean option (method = "kruskal.test", package ggpubr, Kassambara 35 ) to display statistical differences. All tests were performed on RStudio (v4.0.3, RStudio team 36 ).