In vitro study of dual layer mixed matrix hollow fiber membranes for outside-in filtration of human blood plasma

(1)

ContentslistsavailableatScienceDirect

Acta

Biomaterialia

journalhomepage:www.elsevier.com/locate/actbio

In

vitro

study

of

dual

layer

mixed

matrix

hollow

ﬁber

membranes

for

outside-in

ﬁltration

of

human

blood

plasma

O.E.M.

ter

Beek

a,1

_,

_M.K.

_van

_Gelder

b,1

_,

_C.

_Lokhorst

a

_,

_D.H.M.

_Hazenbrink

b

_,

_B.H.

_Lentferink

b

_,

K.G.F.

Gerritsen

b

_,

_D.

_Stamatialis

a,∗

a (Bio)Artiﬁcial Organs, Department of Biomaterials Science and Technology, TechMed Centre, Faculty of Science and Technology, University of Twente,

Drienerlolaan 5, 7500 AE Enschede, the Netherlands

b Department of Nephrology and Hypertension, University Medical Center Utrecht, Heidelberglaan 100, 3584 CX Utrecht, the Netherlands

a

r

t

i

c

l

e

i

n

f

o

Article history: Received 6 August 2020 Revised 21 December 2020 Accepted 22 December 2020 Available online xxx Keywords: Hemodialysis Outside-in ﬁltration Mixed matrix membrane Protein-bound uremic toxins

a

b

s

t

r

a

c

t

Hemodialysismainlyremoves smallwater-solubleuremictoxinsbutcannoteffectively removemiddle moleculesand protein-bounduremictoxins.Besides,the therapyisintermittentleading tofluctuating bloodvaluesandfluidstatuswhichadverselyimpactspatients’health.Prolongedhemodialysis(with ad-equateanticoagulation)couldimprovetheremovaloftoxinsandthedevelopmentofportableand wear-ableartificialkidneyscould offermoreflexibilityinthe dialysisscheme.Thiswould enhancepatients’ overallhealth,autonomy,mobilityandflexibility,allowingpatientstoparticipateinsocialandeconomic life.However,thetimethatpatients’bloodisexposedtothedialyzermaterialislongerduringprolonged hemodialysis,and blood clots could obstructthe fiber lumen, resulting inadecrease ofthe effective membranesurfaceareaavailablefortoxinremoval.Theoutside-infiltration(OIF)mode,whereinblood flowsthroughtheinter-fiberspaceinsteadofthroughthefiberlumina,hasbeenappliedwidelyinblood oxygenatorstopreventfiberclotting,butnotinhemodialysis.

Inthisstudy, wepresentforthefirsttimethedevelopmentofamixedmatrixmembrane(MMM)for OIFofhumanbloodplasma.ThisMMMcombinesdiffusionandadsorptionandconsistsofapolymeric membranematrixwithactivatedcarbon(AC)particlesontheinsidelayer,andapolymericparticle-free layerontheouterfiberlayer.OurresultsshowthatinvitroMMMfibersforOIFdemonstratesuperior removaloftheprotein-bounduremictoxins,indoxylsulfateandhippuricacid,comparedtobothearlier MMMfibersdesignedforinside-outfiltrationmodeandcommercialhigh-fluxfibers.

1. Introduction

Patientswithend-stagekidneydisease(ESKD)need hemodialy-sis,tocompensateforthereducedexcretionofretentionsolutesby their kidneys,ifnodonororgan isavailable.However,thetherapy isintermittent(standardschemeis3× 4hours/week)resultingin inadequateremovalofwastesolutesandexcesswater,which con-tributestopoorwell-beingandhighmorbidityandmortality[1,2].

∗_{Corresponding author.}

E-mail addresses: o.e.m.terbeek@utwente.nl (O.E.M. ter Beek), M.K.vanGelder- 5@umcutrecht.nl (M.K. van Gelder), c.lokhorst@student.utwente.nl (C. Lokhorst), d.hazenbrink@live.nl (D.H.M. Hazenbrink), B.H.Lentferink@umcutrecht.nl (B.H. Lent- ferink), K.G.F.Gerritsen@umcutrecht.nl (K.G.F. Gerritsen), d.stamatialis@utwente.nl (D. Stamatialis).

1 Contributed equally

Clearanceofprotein-bounduremictoxinsisparticularlylow since onlythefreefractionisﬁlteredandtheiraccumulationinpatients with ESKD is associated withincreased cardiovascular morbidity [3,4].

Prolongingahemodialysissessionimprovestheremovalof tox-insfrompatients’ blood andmaythereby increase patients’ over-allhealth[5–7].Studieshaveshownthatextendingadialysis ses-sion from, for example, four to six or eight hours improves the removalofprotein-bounduremictoxins[7,8].Moreover,extended and/or nocturnal dialysis (up to 48 hours) – provided that ade-quate anticoagulationor rinsing is applied – could contribute to theimprovement ofboth uremictoxinsandquality oflife [9,10]. Portable or wearable artiﬁcial kidneys could facilitate prolonged therapywithimprovedtoxinremoval,butwouldalsoincrease pa-tients’ freedom in everydaylife andcontribute to decreasing the therapycosts[5,6,11,12].However,mostdialyzersarenotdesigned

https://doi.org/10.1016/j.actbio.2020.12.063

(2)

dialysis fluid (through fiber lumen)

blood (this study: plasma) (through intercapillary space)

(A)

(B)

blood (this study: plasma) (through fiber lumen)

dialysis fluid (through intercapillary space)

Fig. 1. Schematic overview of flow directions of the blood (or plasma) and dialysis fluid. (A) Conventional, inside-out filtration mode: blood flows through the fibers’ lumen and dialysis fluid flows through the intercapillary space in counter current direction. (B) OIF mode: dialysis fluid flows through the fiber lumen and blood flows through the intercapillary space in counter current direction.

forprolonged dialysisandthe appliedfiltration modeisnot suit-ableforlong-termtherapy.Infact,duringhemodialysis,theblood entersthemoduleandflowsintheboreofthefiberswhereasthe dialysis fluid flows inthe inter-fiberspace (seeFig. 1A).Thrombi orfibrinclotscouldobstructthefibers’lumenduringthetherapy. Asaresult,lessfibersofthedialyzerareopen and,therefore,less effectivemembranesurfacearearemainstofiltratepatients’blood [13].

To prevent fiber clotting, a new mode of filtration mode was introduced forblood oxygenators– the outside-in filtration (OIF) – and a new module design (mats with regularly spaced fibers for optimal blood flow and distribution) was proposed [14]. In the OIF mode, patients’ blood flows in the inter-fiber space, in-stead ofthroughthefiberlumen(Fig.1B).Theadvantagesofthis so-calledextraluminalflowtherapyincludelowerbloodboundary layerresistanceto masstransport(conventionalconfiguration: lift forces tendto movethe red bloodcellstoward thecenterofthe fiberlumen andawayfromthefiberwall,thereby thickeningthe plasma boundary layer and limiting the mass transport through the fiber wall) and maintenance of the effective membrane sur-face area [14,15]. Recently, Dukhin et al . [13] showed that com-mercially available dialyzers could be used for up to 100 hours when applied in OIF mode. Although the formation and deposi-tion of blood clots could not be prevented completely, a signifi-cantlylargermembranesurfacearearemained‘clot-free’compared with a dialyzerin conventional mode (i.e. blood flowingthrough the fiber lumen).Besides, inthe OIF mode, the blood clotswere only detected on the outside of the fiber bundle near the dia-lyzer’sbloodentrance,whereasasignificantnumberofthelumen ofthe fiberswereclotted intheconventionalinside-out filtration mode [13]. Yamashita et al . [16] also used commercial dialyzers withheterogeneousmembranestructuresintheOIFmodeand re-portedhighertoxinremovalthaninconventionalinside-out filtra-tion mode dueto toxinentrapment within thewedge-like mem-branepores.

In this work, we present, for the ﬁrst time, the development and applicationof a dual layer hollowﬁber mixedmatrix mem-brane (MMM) in OIF mode for achieving high toxin removal, in particular that of protein-bound uremic toxins. We hypothesize thathightoxinremovalcouldbeachievedduetotwo

complemen-taryinnovations,namely:(I)theOIFmode,allowingforprolonged dialysiswithlowerbloodboundarylayerresistancetomass trans-portand(II)theMMM,combiningdiffusionandadsorptioninone membrane.TheduallayerhollowﬁberMMMconsistsofalayerof apolymericmembranematrixwithactivatedcarbon(AC)particles, whichisincontactwiththedialysisﬂuid,andanhemocompatible polymeric particle-free layer,which isin contactwithblood. The particle-free polymeric layer preventsdirect contactbetween pa-tient’s blood andthe AC andit is responsible forthe membrane selectivity.The MMMposesanumberofadvantagescomparedto conventional therapies.Firstly, the useofrelatively small adsorp-tive particles increases the available surface area for the adsorp-tionof theuremictoxinswithouthighpressure drop,incontrast to bedadsorption column. Secondly, the useof adsorptive parti-clesincreasestheremovalofthetoxinsbykeepingthe concentra-tiongradientacrossthemembraneveryhigh,asmostofthetoxins areadsorbedtotheparticles.

Theproteinboundtoxinsareinfactdifficulttoremoveby cur-renttherapies,sincethedialyzers’cut-off ensuresretentionofthe proteins (e.g. albumin) thesetoxinsarebound to[7,17]. Their re-tentioninthebodycouldincreasetheriskofcardiovascular prob-lems andcould negatively affectthe body’s inflammatory system and metabolic function as well [7,8,18]. In fact, retention of the protein-bound toxin indoxyl sulfate (IS, ~90% protein binding) is associatedwithenhancing thecoagulation andinhibiting neovas-cularization [18]andretentionofhippuricacid(HA,~30%protein binding)isassociated withrenal tubular damage,proximal tubu-larinjury and inhibition of glucose utilization[18]. The MMM is quite advantageous forimprovedremoval oftheseprotein bound toxins.There,thealbumincannotreachthesorbentparticles, how-ever,thefreefractionofthetoxins(nonboundtoalbumin)passes theselectivemembranelayerandiseffectivelyadsorbedtotheAC particles leadingtofurther dissossiationoffree fractionoftoxins fromalbumininthebloodandthereforetohigherremoval.In ear-lierstudiesinvitro,wehaveshownthatthiscouldleadto3times higherremovalofPBTincomparisontocommerciallyavailableHD membranes [4,19–21]. Besides, we have already shown that the MMMhaslowfouling,notoxinadsorptiontothemembrane form-ingpolymerandexcellentbloodcompatibility[22].Applicationof an OIF mode may furtherimprove toxin removaldueto mainte-nanceoftheeffectivemembranesurfaceareaandminimizationof membranefoulingattheentranceofthemodule,asoccursinthe inside-outfiltrationmode.

Here, we investigate in detail the effect of the OIF applica-tion mode of the dual layer hollow fiber MMM (indicated as MMM-OIF) on toxin removal. First, the spinning parameters for thepreparationofMMM-OIFwere investigated.Second,we stud-ieduremictoxinremoval in vitro frombuffersolutionsandhuman blood plasma duringlong-term diffusive experiments. For study-ing, for thefirst time, the impact ofthe filter’s configuration on uremictoxin removal, the use ofa complex solution as(human) bloodisnotdesirable.Therefore,wehaveperformedexperiments with human plasma and compared the results to earlier studies with inside-out filtration (where human blood plasma was used aswell).Removal ofa small water-solublesolute, creatinine,and twoprotein-bounduremictoxinswithdifferentdegreesofprotein binding,IS(~90%proteinbinding)andHA(~30%proteinbinding), fromhumanplasmawasinvestigated.Theadsorptionoftoxinsto theACparticles isirreversibleandthereforethe‘adsorptivetoxin removal’wascalculatedindirectlyfromtheamountoftoxinsthat ismissingfromthebloodorbloodplasmaandtheamountthatis gainedinthepermeate(usuallythedialysissolution)[4,19–21]. Fi-nally,theperformanceoftheMMM-OIFwascomparedtothatofa conventionalhigh-fluxdialysismembrane,bothinOIFand inside-outfiltrationmode.

(3)

2. Materials and methods 2.1. Materials

The materials forfabrication of theMMM ﬁbers forOIF were similar to those used in earlier studies [4,20]. The MMM hol-low ﬁbermodules(effectivemembranesurfacearea:0.0016mm2_, length:9.5 cm,mass:450 μgcm−1) were preparedby dissolving polyethersulfone (PES) (ULTRASON, E6020P, BASF, Ludwigshafen, Germany) and polyvinylpyrrolidone (PVP) (K90, MW_≈360 000) (Fluka,Sigma-Aldrich,Germany)inN-methyl-2-pyrrolidone(NMP) (AcrosOrganics,Belgium).

Activated carbon (AC) particles (Norit A Supra EUR, Norit Netherlands BV, The Netherlands) were sieved in a 45-μm sieve (VWR, the Netherlands) and used in the MMM sorbent layer. Polyflux 2H modules (effective membrane surface area: 0.0016 mm2_, _length: _6.5 _cm, _mass: ₁₀₀ _μg _cm−1₎ _were _{reconstructed} fromlarge,commercialPolyflux2Hdialyzers (Baxter,The Nether-lands) containingfibers basedona polyarylethersulfone,PVP and polyamideblend.

Module housings were prepared by using polyethylene tubes (Bürkle, Germany) and push-in T-connectors (Festo, The Nether-lands) and the ﬁbers were potted into thesemodules with two-componentGriffoncombifastglue(Klium,TheNetherlands).

A Milli-Q puriﬁcation unit (Merck Millipore, Czech Republic) provided the ultrapure water necessary for spinning procedures and transport experiments. Phosphate-buffered saline (PBS) (pH 7.45, GibCo,United Kingdom) wasused todissolve bovine serum albumin(BSA,60kDa)andcreatinine(113Da),bothacquiredfrom Sigma-Aldrich(TheNetherlands).

Humanplasmawasobtainedfromhealthydonors(Sanquin,The Netherlands) in compliance with ethical guidelines. The protein-bound uremic toxins hippuric acid (HA, 179 Da, ~30% bound to albumin [23]) andindoxyl sulfate(IS, 213 Da, ~90%bound to al-bumin [23]) (Sigma-Aldrich, The Netherlands), were dissolved in humanplasma.

Dialysis ﬂuid was prepared as described by Geremia et al . [20] KCl 2 mM, NaCl 140 mM, CaCl2 1.5 mM, MgCl2 0.25 mM, NaHCO3 35 mM (Sigma-Aldrich, Germany) and glucose 5.5 mM (Life Technologies Europe BV, The Netherlands) were dissolved in ultrapure water (pH7.4). The following three eluents (Sigma-Aldrich,The Netherlands), eluentsAandB bothwithpH=3, con-sisting of 50mMammonium formate buffer andrespectively 10% and 90% methanol and eluent C that is acetonitrile, were pre-paredtodeterminetheprotein-bounduremictoxinconcentrations ofthehumanplasmaanddialysisﬂuid,usingreverse-phase high-performanceliquidchromatography(RP-HPLC,JASCO,Japan).

2.2. Hollow ﬁber preparation

For the preparationof the MMM hollow fibers for OIF (indi-cated hereasMMM-OIF), two polymer dope solutions were pre-paredfollowingtheprotocoldescribedinearlierstudy[4]: (1)15 wt% PES and7 wt% of PVP were dissolved in NMP for the pro-tective,outer layer and(2)14 wt% PESand1.4wt% ofPVP were dissolvedinNMPandACparticleswithaloadingof60wt%in re-lationtotheamountofPESwereaddedforthefibers’innerlayer. Next, both polymer dope solutions were mixedon a roller bank atroomtemperature.ABekiporSTAL315-μmfilter(Bekaert, Bel-gium)wasusedtofiltertheparticle-free,outerlayerpolymerdope solution.Subsequently,stainlesssteelsyringeswerefilledwiththe two polymersolutionsandtheywere allowedtodegasovernight. High-pressuresyringepumpsandaspinneretwereusedtoextrude themembranes(seeTable1forspinneretdimensionsandspinning conditions).

Table 1

Spinneret dimensions and spinning conditions for the fabrication of MMM-OIF.

Spinning parameters MMM-OIF

Inner diameter needle of spinneret 0.16 mm Outer diameter needle of spinneret 0.26 mm Inner diameter 1 st oriﬁce of spinneret _{0.46 mm}

Outer diameter 1 st oriﬁce of spinneret _{0.66 mm}

Inner diameter 2 nd oriﬁce of spinneret _{0.86 mm}

Inner layer, polymer dope pumping speed 1 mL min −1

Outer layer, polymer dope pumping speed 0.2 mL min −1

Bore liquid pumping speed (ultrapure water) 1. 2 mL min −1

Air gap 10 cm

Pulling wheel speed 5.5 m min −1

Thespinneret thatwasusedwasoriginally developedfordual layerMMMfibersforconventionalinside-outfiltration[4].Theslit ofthespinnerettocreatetheouter layerislargerthantheslitto createthe inner layer of the dual layer fiber, to obtain an outer layerwithACparticleswithsufficientsurfacearea [4,24].Forthis reason,itwasmoredifficulttoextrudethepolymersolutionwith ACparticlesthroughthissmallerinnerslitandobtainalayerwith ACparticles ontheinsideofthefiber.Thespinningofthehollow fibers was startedwith the polymer dope solution forthe inner layer.Assoonasthefirstpolymersolutionwasextruded,the wa-ter (boreliquid) pumpingwasstarted, atspeed twotimeshigher (2.4mLmin−1) thanthedesiredwaterpumpingspeed. This ‘ex-ertedforce’ wasnecessaryforextrudingthesolutionwiththeAC particles.Oncethe innerlayerof thefiberwiththeparticleswas formed,theboreliquid(water)pumpingspeed wasloweredto1. 2 mL min−1 and the polymer solution of the particle-free outer layerofthefiberwasextruded.Dry-wetspinning via immersion-precipitationwasappliedtofabricatetheMMMhollowfibers.The fiberswentthroughan airgap(Table1)andthendippedintothe coagulationbath.Fourwheelswereusedto directandgather the fibersand,atlast,thecollectedfiberswererinsedelaboratelyand storedinultrapurewater.

2.3. Hollow ﬁber characterization 2.3.1. Scanning electron microscopy (SEM)

A scanning electron microscope (Jeol JSM-IT 100 LV, InTouchScopeTM _software) _was _used _to _obtain _{cross-section} and surface images of the MMM-OIF and Polyflux 2H fibers. To obtain cross-section images, hollow fiber membranes were dried in the air overnight at room temperature, broken cryogenically inliquid nitrogenthe next day andfinally put into cross-section SEMsampleholders.Imagesfromtheinner andouter membrane surfacewere alsoobtainedafterairdryingofthefibersovernight at room temperature. Subsequently, the fibers were fixed to the SEMholderswiththeuseofdouble-sidedcarbontape(forthe in-nersurfacesamples,thefiberswere cutopenwitharazorblade). Last, samples were gold sputtered (Cressington 108 auto-sputter coater).

2.3.2. Membraneultraﬁltrationcoeﬃcient(Kuf)

TheMMM-OIFandPolyflux2Hmodules werewettedand pre-pressurized with ultrapure water, in dead-end, inside-out set-ting,attransmembranepressures (TMPs)of1500 mmHgand225 mmHg, respectively for 30 minutes. Then, the amount of water permeating across the fibers over time was measured in inside-outmodeatdifferentTMPs(n=3).FortheMMM-OIFfibers,TMPs of375,750,1125and1500mmHgwereapplied accordingtothe protocolof previous high-fluxMMM study[19]. Forthe Polyflux 2Hfibers,lowerTMPshadtobeapplied(75,150and225mmHg), since the specification sheetof the manufacturer, Baxter,

(4)

recom-mendsapplyingTMP≤600mmHg.Thefibers’ultrafiltration coeffi-cient (Kuf,mL h−1 m−2 mmHg−1) wascalculated asthe slope of graphoftheresultingflux(mLh−1m−2)versusTMP(mmHg).For all membranes,theultrafiltrationcoefficient(K_uf) wasdetermined prior to experiments forevaluation ofuremic toxin removal(see latersections).

2.3.3. Creatinine, HA and IS removal

For the creatinine removal experiments with the MMM-OIF fibers,the moduleswere connectedtothe Convergencecrossflow set-upand50mLofcreatininefeedsolutionand50mLofdialysis fluid were recirculatedthrough the system in OIF mode for four hours, seeFig.1B.The creatininefeedsolutionconcentration was 0.1gL−1inPBSsolution(closetothemeanuremiccreatinine con-centrationindialysis patients[25]) andthe dialysisfluidsolution was a PBS solution. The flow rates ofthe feed anddialysis fluid were settorespectively10mLmin−1 and5mLmin−1 tobe able to dotheexperiments ataTMPof0mmHg(diffusion mode). To quantifythecreatinineremoval,2mLsamplesofboththefeedand dialysisfluidweretakenattimepointst=0,1,2,3and4hours. Creatinineconcentrationsinthefeedanddialysisfluidwere deter-minedusingUV-VISspectrophotometry(230nm).

FortheHAandISremovalexperiments,themoduleswere con-nectedtotheConvergencecrossflowset-up.For24hours,50mLof humanplasmaspikedwith110mgL−1HAand40mgL−1IS (sim-ulating meanuremicconcentrationsindialysispatients[25]) and 100 mL of dialysis fluid were recirculated in counter current di-rection.ExperimentswereperformedwithMMM-OIFfibersinOIF mode(seeFig.1B)andPolyflux2HfibersbothinOIFandthe con-ventionalinside-outmode(seeFig.1).

During diffusive experiments (TMP=0 mmHg), flow rates through thefibers’lumen andinter-fiber spacewere1mL min−1 and10 mLmin−1, respectively,both inOIF andinthe inside-out filtration mode.Plasma anddialysisfluidsamples weretaken be-forestart (3mL)andafter0.5h,1h,2h,3h,4h,8hand24h(all2 mL)forestimationofHAandISconcentrations.

To determine theHA andIS concentrations ofall plasma and dialysisﬂuidsamples,reverse-phasehigh-performanceliquid chro-matography (RP-HPLC, JASCO, Japan) wasused. Furthermore, the freefractionofthetoxinsHAandISinthehumanplasmasolution wascheckedwiththeuseof‘notdenaturatedhumanplasma sam-ples’andRP-HPLCaftereachexperiment.Thisway,thepercentage oftoxin-proteinbindingatthestartwasalsochecked.Onlythe so-lutionswithapproximately30%and90%HA-andIS-protein bind-ing,respectively,wereincludedforanalysis.

All samples were diluted 4x with ultrapure water, de-proteinized via heat treatment (95°C, 30 minutes) and filtered (10 kDa filter, Ultracel-10, 0.5 mL sample volume, Merck) prior to RP-HPLC analysis (UV: 245 nm, fluorescence:

λ

ex = 272 nm,

λ

em =374nm).

For all toxins(creatinine,HA andIS), we estimatedthe dialy-sance, DLp and DLd (mL min−1 m−2), in the plasma(or PBS for creatinine) and inthe dialysis ﬂuid, respectively, following equa-tions1and2[23,26]: DLp= Xp t· Ae f f

(

Cp− Cd

)

(1) DLd= Xd t· Ae f f

(

Cp− Cd

)

(2)

where Xp is the amount of toxinsremovedfrom the plasma(or PBS forcreatinine)andXd istheamountoftoxinstransportedto thedialysisﬂuid(mg)afteracertaintimet(min).Aeffisthe mod-ule’seffectivesurfacearea(m2_)._C

p andCd (bothinmgmL−1)are the concentrationsofthetoxinsinthe plasma(orPBS for creati-nine)andinthedialysisﬂuid,respectively.Ifthereisnoadsorption

oftoxinsonthe membranesit isexpectedthat DLp / DLd =1.If thereis toxinadsorption on themembranes (i.e. MMM),it is ex-pectedthatDLp /DLd >1.Inconventionalhemodialysis,the dialy-sisfluidisconstantlyrefreshed(C_d ₌0),therefore,equation1and equation 2 can be simplified andcorrespond to the toxin clear-ancesnormalizedtothemembranesurfaceareaintheplasmaand dialysisfluid,respectively:

CLp= Xp t· Ae f f Cp (3) CLd= Xd t · Ae f f Cp (4)

Tocompareourresultswithliterature,thetoxinremovalbythe membranes wasexpressed as mgof toxin removed per effective surfacearea of the hollow ﬁbermodules (mg m−2) or as mgof toxinremovedpergofmembrane(mgg−1 membrane).

Finally,thetotal proteinconcentrationsof theplasma(diluted 40x withultrapure water) anddialysisfluid were determined (at 280nm)usingaNanoDrop-1000spectrophotometer(Fisher Scien-tific)andNanoDrop-1000software(V3.8.1).Then,theprotein siev-ingcoefficient (SCp) ofthemembraneswasdeterminedvia equa-tion5:

SCp=

Cpermeate

Cretentate

(5)

Cpermeate and Cretentate are the total protein concentrations

(mg mL−1) in the permeateand retentate solutions, respectively. SCp =0indicatesnoproteinpassageandSCp=1indicates unim-pededpassageofproteinsacrossthemembranes.

2.3.5. Statistics

One-wayANOVAwithpost-hocTukey’stestandorindependent samplesStudent’st-testswereused,asappropriate,tocomparethe resultsofMMM-OIFandPolyflux2Hfibers.Allanalyseswere per-formedusingGraphPadPRISM (version5.00softwarepackage).A Pvalue<0.05wasconsideredsignificant.

3. Results and discussion 3.1. Development of MMM-OIF ﬁbers 3.1.1. SEM

Inthiswork wedeveloped,forthefirst-time,duallayerMMM fibersforOIF (MMM-OIF). Toachievethis, we extrudedthe poly-mer solutions forboth the inner layer withthe ACparticles and theouterparticle-freelayerbyaspinneretthatwasoriginally de-signedforspinningduallayerMMMfibersfordialysisinthe con-ventional inside-out mode (membrane having inner particle-free polymeric layer andan outer layerwiththe ACparticles). At the start,a2-foldhigherboreliquid(ultrapurewater)pumpingspeed (than the desired speed) was necessary to extrude the viscous polymer solution with AC particles with extra force through the thinspinneretslit(designedforparticle-freesolutions).

Fig. 2 presents the SEM images of the cross-sections of the MMM-OIF fibers and the commercial Polyflux 2H fibers. The MMM-OIF fibers have large finger-like pores on the outer wall, in contrast to smaller finger-like pores on the fibers’ inner wall (Fig. 2A.1 and Fig. 2A.2). There is no distinct transitionbetween the particle-free layer andthe layer withthe AC particles (indi-catedby thewhitearrowsinFig.2A.3)andthetwolayersappear tobeinterconnected(indicatedbythedottedlineofFig.2A.3).

Fig. 2B presents the SEM images of the Polyflux 2H hollow fibers which have a spongy structure on the inside of the fiber, whereastheporesbecomebiggerandmorefinger-liketowardthe outside of the fiber. The overall dimensions of the Polyflux 2H

(5)

Fig. 2. SEM images of the cross-sections of: (A) MMM-OIF and (B) Polyﬂux 2H ﬁbers. The dotted line (A.3) indicates the transition between the particle-free outer layer and the inner layer with the AC particles. The white arrows (A.3) indicate the AC particles.

fibers are smaller compared with the MMM-OIF fiber (Table 2). Table 2 also compares the dimensions of the MMM-OIF and Polyflux 2H membranes totwo other MMM developed in earlier studies, a low-flux MMM fiber indicated as LF-MMM [4] and a high-fluxMMMindicatedasHF-MMM[19].BothLF-MMMand HF-MMM were developed for applicationin the inside-out filtration modeandtheirdimensionsarebiggerincomparisontothe MMM-OIFfiber(Table2).

Fig.3presentstheSEMimagesoftheinnerandoutersurfaceof theMMM-OIFandPolyflux2Hfibers.Onthe innersurfaceofthe MMM-OIF fibers, the ACparticles can be seen (indicated by the whitearrows,Fig.3A.2)well embeddedwithin thePES-PVP poly-mer matrix(seedottedwhitecircles,Fig.3A.2).Theouter surface oftheMMM-OIFseemstohavesmallporesbuttherearealso ar-eas where larger pores in the order of microns can be observed

(see Fig. 3C.2). Forthe Polyﬂux 2Hﬁber, the inner selective sur-face(Fig.3B)hassmallporeswhereastheoutersurfacehaslarger poresintheorderofafewmicrons(Fig.3D).

Forthe MMM-OIF, the solvent/ non-solvent system led to in-stantaneousdemixing [27] and a fastexchange resultedin small pores anda dense membranestructure that embeds the AC par-ticles (Fig. 3A). Moreover, an air gap of 10 cm was used before theextrudedMMM-OIFfiberwasimmersedinthewater contain-ingcoagulationbath.TheMMM-OIFfibers’particle-freeouterlayer wasexposedtowatervapor presentinthe airofthis10cmgap. However, the watercontent ofthe vapor isless thanthat of the boreliquid.Asaresult,theexchangebetweensolvent/non-solvent wassloweron theoutside ofthe fiberthan onthe inside ofthe MMM-OIFfiber.Therefore,theoutersurfaceoftheMMM-OIFfiber (Fig. 3C) shows small pores in the PES-PVP matrix,whereas the

(6)

Table 2

Dimensions of the MMM-OIF fibers, Polyflux 2H fibers and low-flux (LF) and high-flux (HF) MMM fibers of previous studies (applied in the inside-out filtration mode) [ 4 , 19 ]. Mean ± standard deviation is presented. MMM-OIF Polyflux 2H LF-MMM [ 4 ] HF-MMM [ 19 ] Inner diameter (μm) 329 ± 2 218 ± 7 450 669 ± 9 Outer diameter (μm) 518 ± 3 306 ± 5 586 984 ± 11 Wall thickness (μm) 92 ± 8 43 ± 1 68 160 ± 9 MMM layer thickness (μm) 78 ± 5 - 47 111 ± 4

Particle-free layer thickness (μm) 26 ± 6 - 21 49 ± 5

Fig. 3. SEM images of inner surface (A, B) and outer surface (C, D) of MMM-OIF fiber (left panels) and Polyflux 2H fiber (right panels). The white arrows indicate the activated carbon (AC) particles and the dotted white circles indicate the PES- PVP membrane matrix surrounding the AC particles (A.2).

MMM-OIF fibers’ inner PES-PVP surface showsno (visible) pores (Fig.3A).TheSEMimagesofFig.3thereforeindicatethatthe se-lectivelayeroftheMMM-OIFfibersispresentonthefibers’inner wall.

3.2. Hollow ﬁbers’ transport properties 3.2.1. Membrane’s K _uf

Fig.4presentsthewaterfluxversustheTMPacrossthe MMM-OIF and Polyflux 2H membranes. The MMM-OIF membrane has stable performance for TMPup to 1500 mmHg whereas Polyflux 2HmembraneshavestableperformanceuptoTMPof225mmHg; in fact when a TMPof 375 mmHg wasapplied the Polyflux 2H membranesweredamaged.Furthermore,noparticlelossfromthe MMM-OIF fiberswas observedduringthe experiments.From the slopesofthegraphsofFig.4AandB,theKufoftheMMM-OIFfiber wasestimatedtobe100mLh−1 m−2 mmHg−1 andtheKuf ofthe Polyflux2Hfiberwasestimatedtobe144mLh−1 m−2mmHg−1.

Table 3 compares transport properties of the MMM-OIF and Polyﬂux2HmembranestotheLF-MMM[4]andHF-MMM[19]of previousstudies.

3.2.2. Creatinine removal

The creatinine removal from PBS solution by the MMM-OIF ﬁbers was measured andcompared to previous studies withthe LF-MMM andHF-MMM (Table3). Fig.5 compares thetotal crea-tinine removalfrom PBS to that transportedto the dialysis ﬂuid intime,normalizedtotheconcentrationgradientacrossthe mem-brane.Therefore,theslopesofthepresentedgraphs,correspondto thedialysanceDLp andDLd,respectively(seeequation1,equation 2 and Table 3). If there is no adsorption of toxinson the mem-branesitisexpectedthat DLp /DLd =1.Ifthereistoxin

adsorp-0

10,000

20,000

30,000

40,000

50,000

0

50

100

150

200

250 h

L

m(

x

ulf

r

et

a

w

n

a

el

C

-1

m

-2

)

TMP (mmHg)

0 50,000

100,000

150,000

200,000

250,000

0

500 1000

1500

2000

h

L

m(

x

ulf

r

et

a

w

n

a

el

C

-1

m

-2

)

TMP (mmHg)

MMM-OIF

(A)

(B)

Polyflux 2H

Fig. 4. Clean water flux versus TMP graph of MMM-OIF (A) and Polyflux 2H (B) fiber. The clean water flux was determined in conventional inside-out mode. Mean ± standard deviation is depicted (n = 3).

tionon themembranes(i.e. MMM),itisexpectedthat DLp /DLd

>1.

Approximately 62% of the creatinine is transported from the buffer tothe dialysisﬂuid bydiffusion (Fig. 5).The fact thatDLp / DL_d ₌1.6also indicates thata signiﬁcant partofcreatinine re-movalbythe

MMM-OIFisduetoadsorption totheACsorbentparticles. Be-sides, both DLp and DLd are constant duringthe 4hours experi-ment,andthereisnoindicationofparticlesaturation.

Table3comparesthecreatinineremovalbytheMMM-OIFat4 hours tothat bythe LF-MMMandHF-MMM developedinearlier studies[4,19].Infact,thetotalremovalby theMMM-OIFﬁbersis muchhigherthanthatobtainedbytheLF-MMM andHF-MMMin

(7)

Table 3

Transport properties of the MMM-OIF, Polyflux 2H and dual layer low-flux (LF) and high-flux (HF) MMM fibers of previous studies [ 4 , 19 ]. Mean ± standard deviation is presented . MMM-OIFoutside-in mode Polyflux 2Hinside-out mode Polyflux 2Houtside-in mode LF-MMMinside-out mode [4] HF-MMMinside-out mode [19] Kuf (mL h −1 m −2 mmHg −1 ) 100 ± 19 144 ± 25 144 ± 25 3.35 78 ± 12 SCp (-) 0.00 ± 0.01 0.01 ± 0.01 0.01 ± 0.01 0 0.00 ± 0.02

Total creatinine removal (4h)

(mg m −2 ) _{3732 ± 915} _- _- ₂₅₇₉ ₂₈₂₅_[4] (mg g −1 membrane) _{86 ± 21} _- _- ₆₈ ₄₀ DLp creatinine (mL min −1 m −2 ) 200 - - - - DLd creatinine (mL min −1 m −2 ) 124 - - - - Total HA removal (4h) (mg m −2 ) _{1466 ± 209} _{1562 ± 263} _{1436 ± 82} _- _- ∗,# NS ∗ # (mg g −1 membrane) _{34 ± 5} _{103 ± 17} _{95 ± 5} _- ₁₂ _P_{= 0.0026} _P_{= 0.0009} DLp HA (mL min −1 m −2 ) 370 112 99 - - DLd HA (mL min −1 m −2 ) 42 80 101 - - Total IS removal (4h) (mg m −2 ) _{860 ± 112} _{374 ± 12} _{366 ± 16} ₃₆₇ ₂₅₂_[4] ∗_P_{= 0.0017} # P = 0.0097 ∗ # (mg g −1 membrane) _{20 ± 3} _{25 ± 1} _{24 ± 1} ₁₀ ₃ _P_{= 0.0354} _NS DLp IS (mL min −1 m −2 ) 145 63 49 - - DLd IS (mL min −1 m −2 ) 1 8 10 - -

Abbreviations: Membrane ultrafiltration coefficient (K uf ), protein sieving coefficient (SC p ), plasma dialysance (DL p ) and dialysis fluid dialysance (DL d ). ∗_{P < 0.05 MMM-OIF vs. Polyflux 2H inside-out filtration mode (mg m}−2 )

# P < 0.05 MMM-OIF vs. Polyﬂux 2H OIF mode (mg m −2 )

_{P < 0.05 MMM-OIF vs. Polyflux 2H inside-out filtration mode (mg g}−1 membrane) _{P < 0.05 MMM-OIF vs. Polyflux 2H OIF mode (mg g}−1 membrane)

0 20,000 40,000 60,000 80,000 0 50 100 150 200 250 Time (min) MMM-OIF removed from PBS

transported to dialysis fluid

l a v o m er e ni nit a er Cm L m( -2)

Fig. 5. Removal of creatinine from PBS by the MMM-OIF ﬁbers in time. The closed (black) data points represent the total removal (creatinine removed from the PBS) and the open (white) data points represent the creatinine transported to the dialysis ﬂuid compartment. Mean ± standard deviation is presented (n = 3).

the inside-out filtration mode, namely,the removalby MMM-OIF is3732mgm−2 after4hours, whereasthat bytheLF-MMM and HF-MMM were 2579mgm−2 [4]and2825mgm−2 [19], respec-tively.Thisimprovedperformanceisprobablyduetotherelatively high K_uf of the MMM-OIF fibers’ (Table 3) which contributedto achieving high diffusivity of creatinine across the membrane (to the dialysisfluid)butalsodueto thehighpartition ofcreatinine within thelayerwiththeACparticlesleadingtoenhanced creati-nineadsorption.Basedonthecreatinineremovalduring4hoursof dialysis,aMMM-OIFdialyzerwithaneffectivesurfacearea of ap-proximately 0.4m2 _(in _the_range_of_current_dialyzers_used_in_the clinic (0.4-2.6m2 _[28]_),_would_be_able_to_remove_the_daily creati-nineproductioninpatientswithESKD(1200± 600mg[19,25]).

3.2.3. Protein-bound uremic toxin removal

Inearlierstudies, we haveshownthat duallayer hollowfiber MMMs achieve superior removalof protein-bound uremictoxins comparedwithconventionaldialysismembraneswhen appliedin conventional inside-out filtration mode. Here, we investigate the combinedremovaloftheprotein-bounduremictoxins,HAandIS, withtheMMM-OIFfibersincomparisonto thePolyflux2Hfibers (bothinconventionalinside-outmodeandOIFmode).

Figs. 6 and 7 present the total removal of HA and IS from human plasmaby the membranes in comparisonto the removal transportedto the dialysisﬂuid, normalized to theconcentration gradientacrossthemembrane.Therefore,theslopes ofthegraphs correspondtotheHAandISdialysances,DLp andDLd,respectively (see equations 1 and 2). The total removal of HA andIS by the membranesafter4hoursisalsodepictedinTable3.Theremoval ofHAbyallmembranes(Fig.6)isgenerallyhigherthanthatofIS (Fig.7).Thiscouldbeexplainedbythepercentageofprotein-toxin binding (forHAthe free fractionis~70%, whereas forISthe free fractionis ~10%) andthehigher startingplasmaconcentration of HAcomparedtoIS(110mgL−1versus40mgL−1,respectively).

For the MMM-OIF, the ratio of dialysances DLp / DLd is very high(forHA:DLp/DLd=9andfor

IS:DLp/DLd=145)clearlyshowingtheimportantcontribution of adsorption to the removal ofthese toxins. Interestingly, while theDLp staysconstant forup to4hoursoffiltration, itdecreases afterapproximately8h(seeFigs.6Aand7A).Thisphenomenonis probablyduetogradualsaturationofthesorbentparticlesintime. Furthermore,thedropinHAremovalfromplasmabythe MMM-OIFafter4hours(Fig.6A)couldbeexplainedbypartialreleaseof someofthetoxinsthatwerealreadyboundtotheACparticles.It isalsoveryclearthatforHA,whichhasratherhighfreefractionin plasma(~70%),theremovaloccursbycombinedadsorptiontothe particles andfiltration to thedialysisfluid. ForIS,whichhas low free fraction inthe plasma (~10%), the removal ismainly due to adsorption ontheACparticles.Finally,estimation oftotalprotein

(8)

0 20,000 40,000 60,000 80,000 100,000 120,000 0 250 500 750 1000 1250 1500 m L m( l a v o m er A H -2) Time (min) MMM-OIF

removed from plasma

0 20,000 40,000 60,000 80,000 100,000 120,000 0 250 500 750 1000 1250 1500 m L m( l a v o m er A H -2) Time (min) Polyflux 2H, inside-out mode

removed from plasma

0 20,000 40,000 60,000 80,000 100,000 120,000 0 250 500 750 1000 1250 1500 m L m( l a v o m er A H -2) Time (min) Polyflux 2H, OIF mode

removed from plasma

transported to dialysis fluid (A)

(B)

(C)

Fig. 6. Removal of hippuric acid (HA) from human plasma in time by: (A) MMM- OIF, (B) Polyflux 2H, measured in inside-out mode and (C) Polyflux 2H, measured in OIF mode. The closed (black) data points represent the total removal (HA removed from the plasma) and the open (white) data points represent the HA transported to the dialysis fluid. Mean ± standard deviation is presented (n = 3).

intheplasmaanddialysisﬂuidafter24hoursindicatesthat~20% ofthe totalplasmaproteinsare adsorbedonthe membrane.This contributes, on one hand,to higher removalof HA andIS (since thoseareboundtoalbumin)butontheother hand,tolimited ac-cessibilityoftheACparticles.

ForthePolyﬂux2Hmembrane,theratioDLp/DLdofHAislow inbothmodes(1.4and1forinside-outandOIFmode,respectively, Table3) indicatingthat theHAremovalismainlybasedon diffu-sion asexpected. However, the ratioof DLp / DLd ofIS is rather highinbothmodes(8and5,forinside-outandOIFmode, respec-tively, Table3) indicating that there is alsosigniﬁcant IS adsorp-tion to the membrane. This hypothesis is supported by the total proteinmeasurements,whichshow that~60%and~40%ofthe

to-0 10,000 20,000 30,000 40,000 50,000 60,000 0 250 500 750 1000 1250 1500 m L m( l a v o m er SI -2) Time (min) MMM-OIF

removed from plasma

0 10,000 20,000 30,000 40,000 50,000 60,000 0 250 500 750 1000 1250 1500 m L m( l a v o m er SI -2) Time (min) Polyflux 2H, inside-out mode

removed from plasma

0 10,000 20,000 30,000 40,000 50,000 60,000 0 250 500 750 1000 1250 1500 m L m( l a v o m er SI -2) Time (min) Polyflux 2H, OIF mode

removed from plasma

transported to dialysis fluid (A)

(B)

(C)

Fig. 7. Removal of indoxyl sulfate (IS) from human plasma in time by: (A) MMM- OIF, (B) Polyflux 2H, measured in inside-out mode and (C) Polyflux 2H, measured in OIF mode. The closed (black) data points represent the total removal (IS removed from the plasma) and the open (white) data points represent the IS transported to the dialysis fluid. Mean ± standard deviation is presented (n = 3).

talplasmaproteins aremissingattheendoftheexperiments us-ingthePolyflux2Hfibersininside-outandOIFmode,respectively. Nevertheless,inallcases,theDLp andDLd remainratherconstant forup to4hours andsomewhatdeclineafter8hoffiltration (see Figs.6B,C,7B andC). Furthermore,theremovalofprotein-bound toxins such as IS by MMM-OIF fibers, undergoes faster kinetics compared tothe removalby Polyflux2H fibers,due tothe addi-tiveadsorptiontotheACparticlesoftheMMM-OIF.

Tocomparetheprotein-bounduremictoxinremovalresultsof the MMM-OIF with LF-MMM and HF-MMM of previous studies [4,19], the removal of HAand IS wasexpressed as mg m−2 and mg g−1 membrane (see Table 3). The removal of HA and IS by theMMM-OIFﬁbers(HA:~34mgg−1 membrane,IS:~20mgg−1

(9)

membrane)issigniﬁcantlyhigherwhencomparedtotheLF-MMM [4]

(IS: 10 mg g−1 membrane) andHF-MMM fibers [19] (HA: 12 mgg−1 membrane,IS:4mgg−1 membrane) measuredin inside-outmodeofpreviousstudies.Anexplanationforthiscouldbethat the MMM-OIFhavearelatively thinparticle-freelayer anda suf-ficiently thick layer with ACparticles, compared to the LF-MMM andHF-MMMfibers,improvingACparticleaccessibilityandmass transportacrossthemembranes.

Direct comparisonoftoxin removalby theMMM-OIFto other literaturestudiesisratherdifficultduetodifferencesintheapplied experimentalparameters(e.g. theuseofbloodvs.plasma,theflow rates ofblood/ plasmaanddialysis fluid, theeffective membrane surfacearea,theinitialtoxinconcentrations).Nevertheless,togive anindicationandtoputtheresultsoftheMMM-OIFfibersin clin-ical perspective, some comparison can be madewiththe clinical studyofEloot et al .[29].There,amuchlowerremovalofHA (ap-proximately 370 mgm−2) andIS(approximately 84mgm−2) af-ter 4hours ofdialysiswasreportedbycommercialPolyflux170H and 210H dialyzers in the inside-out filtration mode [29], com-paredwiththeremovalachievedhereby theMMM-OIFfrom hu-manplasma(HA:1466mgm−2 andIS:860mgm−2).

In addition, the dialysance of the MMM-OIF ﬁbers (Table 3) could be compared to the clearance of commercial dialyzers re-portedinliterature.Clinicalstudiesusingcommercialdialyzers re-portaHAclearanceintherangeof63-77mLmin−1m−2andanIS clearanceintherangeof14-19mLmin−1 m−2 [7,8,30],whilethe DLp ofthe MMM-OIFis muchhigher (HA: DLp = 370 mL min−1 m−2and

IS:DLp=145mLmin−1m−2).However,itisimportanttonote, thatdialysanceandclearanceareonlyequalwhenthedialysisﬂuid isconstantly refreshed[26].Sincethedialysisﬂuidofthepresent studywasconstantlyrecirculated,theclearancereportedin litera-turewouldbelowerthanthedialysanceestimatedhere.

4. Conclusion and outlook

DuallayerMMM hollowfibermembranesforoutside-in filtra-tion (OIF) demonstratesuperior removalof creatinine, HAandIS comparedwithconventionaldialyzersandduallayerMMMfibers in conventional inside-out mode, up to 24hours oftreatment in vitro . Future work will focus on implementing the new fibers in long-term full blood studies. There, the developmentof clot for-mation, blood-membranesurfaceinteractions andblood compati-bilitywill bestudied indetail inordertosupport thehypothesis forimprovedlong-termfiltrationofthemembrane.

Funding

This work was supported by Health~Holland KidneyPort [projectnumber40-43100-98-009].

Declaration of Competing Interest

Theauthorsdeclarethattheyhavenoknowncompeting ﬁnan-cialinterestsorpersonalrelationshipsthatcouldhaveappearedto inﬂuencetheworkreportedinthispaper.

Acknowledgement

I.Geremia(BioArtiﬁcialOrgans, UniversityofTwente)is grate-fully acknowledged forsetting up theprotocol forprotein-bound uremictoxinremovalfromhumanplasmawithRP-HPLC.

References

[1] C. Legallais , D. Kim , S.M. Mihaila , M. Mihajlovic , M. Figliuzzi , B. Bonan- drini , S. Salerno , F.A. Yousef Yengej , M.B. Rookmaaker , N. Sanchez Romero , P. Sainz-Arnal , U. Pereira , M. Pasqua , K.G.F. Gerritsen , M.C. Verhaar , A. Remuzzi , P.M. Baptista , L. De Bartolo , R. Masereeuw , D. Stamatialis , Bioengineering organs for blood detoxiﬁcation, Adv. Healthc. Mater. 7 (21) (2018) .

[2] M.K. van Gelder , I.R. Middel , R.W.M. Vernooij , M.L. Bots , M.C. Verhaar , R. Masereeuw , M.P. Grooteman , M.J. Nubé, M.A. van den Dorpel , P.J. Blankestijn , M.B. Rookmaaker , K.G.F. Gerritsen , Protein-bound uremic toxins in hemodialysis patients relate to residual kidney function, are not inﬂuenced by convective transport, and do not relate to outcome, Toxins 12 (4) (2020) .

[3] O. Deltombe , G. Glorieux , S. Marzouki , R. Masereeuw , D. Schneditz , S. Eloot , Selective transport of protein-bound uremic toxins in erythrocytes, Toxins 11 (7) (2019) .

[4] D. Pavlenko , E. Van Geffen , M.J. Van Steenbergen , G. Glorieux , R. Vanholder , K.G.F. Gerritsen , D. Stamatialis , New low-ﬂux mixed matrix membranes that offer superior removal of protein-bound toxins from human plasma, Sci. Rep. 6 (2016) .

[5] A. Davenport , V. Gura , C. Ronco , M. Beizai , C. Ezon , E.J.T.L. Rambod , A wearable haemodialysis device for patients with end-stage renal failure: a pilot study, Lancet 370 (9604) (2007) 2005–2010 .

[6] A. Davenport , Portable and wearable dialysis devices for the treatment of patients with end-stage kidney failure: wishful thinking or just over the horizon? Pediatr. Nephrol. 30 (12) (2015) 2053–2060 .

[7] S. Eloot , D. Schneditz , T. Cornelis , W. Van Biesen , G. Glorieux , A. Dhondt , J. Kooman , R. Vanholder , Protein-bound uremic toxin proﬁling as a tool to op- timize hemodialysis, PLoS One 11 (1) (2016) e0147159 .

[8] T. Cornelis , S. Eloot , R. Vanholder , G. Glorieux , F.M. Van Der Sande , J.L. Scheijen , K.M. Leunissen , J.P. Kooman , C.G. Schalkwijk , Protein-bound uraemic toxins, di- carbonyl stress and advanced glycation end products in conventional and extended haemodialysis and haemodiaﬁltration, Nephrol. Dialysis Transpl. 30 (8) (2015) 1395–1402 .

[9] M.P. Graham-Brown , D.R. Churchward , A.C. Smith , R.J. Baines , J.O. Burton , A 4-month programme of in-centre nocturnal haemodialysis was associated with improvements in patient outcomes, Clin. Kidney J. 8 (6) (2015) 789–795 . [10] M. Dam , P.J. Weijs , F.J. van Ittersum , B.C. van Jaarsveld , Physical performance

in patients treated with nocturnal hemodialysis-a systematic review of the ev- idence, BMC Nephrol. 20 (1) (2019) 317 .

[11] M.K. van Gelder , S.M. Mihaila , J. Jansen , M. Wester , M.C. Verhaar , J.A. Joles , D. Stamatialis , R. Masereeuw , K.G. Gerritsen , From portable dialysis to a bio- engineered kidney, Expert Rev. Med. Devices 15 (5) (2018) 323–336 . [12] J.C. Kim , F. Garzotto , F. Nalesso , D. Cruz , J.H. Kim , E. Kang , H.C. Kim , C. Ronco ,

A wearable artiﬁcial kidney: technical requirements and potential solutions, Expert Rev. Med. Devices 8 (5) (2011) 567–579 .

[13] S.S. Dukhin , Y. Tabani , R. Lai , O.A. Labib , A.L. Zydney , M.E. Labib , Outside-in hemoﬁltration for prolonged operation without clogging, J. Membr. Sci. 464 (2014) 173–178 .

[14] G. Catapano , A. Wodetzki , V. Baurmeister , Blood ﬂow outside regularly spaced hollow ﬁbers: the future concept of membrane devices? Int. J. Artif. Organs. 15 (6) (1992) 327–330 .

[15] E. Klein , The multidisciplinary approach to develop artiﬁcial organs, Blood. Pu- rif. 21 (4–5) (2003) 305–317 .

[16] A.C. Yamashita , T. Ono , N. Tomisawa , Veriﬁcation of physicochemical structures of dialysis membrane using reversal dialysis technique, Hemodialysis Int. 21 (2017) S3–S9 .

[17] J. Jansen , J. Jankowski , P.R. Gajjala , J.F.M. Wetzels , R. Masereeuw , Disposition and clinical implications of protein-bound uremic toxins, Clin. Sci. 131 (14) (2017) 1631–1647 .

[18] R. Vanholder , A. Pletinck , E. Schepers , G. Glorieux ,Biochemical and clinical impact of organic uremic retention solutes: a comprehensive update, Toxins 10 (1) (2018) .

[19] M.S. Tijink , M. Wester , G. Glorieux , K.G. Gerritsen , J. Sun , P.C. Swart , Z. Borne- man , M. Wessling , R. Vanholder , J.A. Joles , Mixed matrix hollow ﬁber membranes for removal of protein-bound toxins from human plasma, Biomaterials 34 (32) (2013) 7819–7828 .

[20] I. Geremia , R. Bansal , D. Stamatialis , In vitro assessment of mixed matrix hemodialysis membrane for achieving endotoxin-free dialysate combined with high removal of uremic toxins from human plasma, Acta Biomater. 90 (2019) 100–111 .

[21] D. Kim , D. Stamatialis , High ﬂux mixed matrix membrane with low albumin leakage for blood plasma detoxiﬁcation, J. Membr. Sci. 609 (2020) .

[22] I. Geremia , D. Pavlenko , K. Maksymow , M. Rüth , H.D. Lemke , D. Stamatialis , Ex vivo evaluation of the blood compatibility of mixed matrix haemodialysis membranes, Acta Biomater. 111 (2020) 118–128 .

[23] S. Eloot , Experimental and Numerical Modeling of Dialysis (PhD thesis), Ghent University, 2004 .

[24] M. Tijink , Membrane Concepts for Blood Puriﬁcation - Towards Improved Ar- tiﬁcal Kidney Devices (PhD thesis), University of Twente, 2013 .

[25] R. Vanholder , R. De Smet , G. Glorieux , A. Argilés , U. Baurmeister , P. Brunet , W. Clark , G. Cohen , P.P. De Deyn , R. Deppisch , B. Descamps-Latscha , T. Henle , A. Jörres , H.D. Lemke , Z.A. Massy , J. Passlick-Deetjen , M. Rodriguez , B. Stegmayr , P. Stenvinkel , C. Tetta , C. Wanner , W. Zidek , Review on uremic toxins: classiﬁcation, concentration, and interindividual variability, Kidney Int. 63 (5) (2003) 1934–1943 .

(10)

[26] J.F. Maher , Replacement of Renal Function by Dialysis: a Textbook of Dialysis, Springer Science & Business Media, 2012 .

[27] J. Mulder , Basic Principles of Membrane Technology, Springer Science & Busi- ness Media, 2012 .

[28] G. Catapano , J. Vienken , Biomedical applications of membranes, Adv. Mem- brane Technol. Appl. (2008) 489–517 .

[29] S. Eloot , W. Van Biesen , M. Axelsen , G. Glorieux , R.S. Pedersen , J.G. Heaf , Protein-bound solute removal during extended multipass versus standard hemodialysis Dialysis and Transplantation, BMC Nephrol. 16 (1) (2015) .

[30] N. Meert , S. Eloot , M.A. Waterloos , M. Van Landschoot , A. Dhondt , G. Glorieux , I. Ledebo , R. Vanholder , Effective removal of protein-bound uraemic solutes by different convective strategies: a prospective trial, Nephrol. Dialysis Transpl. 24 (2) (2009) 562–570 .