Operator training simulation for integrating cultivation and homogenisation in protein production

(1)

Operator training simulation for integrating

cultivation and homogenisation in protein

production

Inga Gerlach, Carl-Fredrik Mandenius and Volker C. Hass

Linköping University Post Print

N.B.: When citing this work, cite the original article.

Original Publication:

Inga Gerlach, Carl-Fredrik Mandenius and Volker C. Hass, Operator training simulation for

integrating cultivation and homogenisation in protein production, 2015, Biotechnology

Reports, (6), 91-99.

http://dx.doi.org/10.1016/j.btre.2015.03.002

Copyright: Elsevier

http://www.elsevier.com/

Postprint available at: Linköping University Electronic Press

(2)

Operator

training

simulation

for

integrating

cultivation

and

homogenisation

in

protein

production

Inga

Gerlach

a,b

,

Carl-Fredrik

Mandenius

b,

*

,

Volker

C. Hass

c

a

DepartmentofEnvironmental-andBio-Technology,HochschuleBremenUniversityofAppliedSciencesBremen,Neustadtswall30,28199Bremen,Germany

b

DivisionofBiotechnology/IFM,LinköpingUniversity,58183Linköping,Sweden

c

FacultyofMedicalandLifeSciences,HochschuleFurtwangenUniversityofAppliedSciencesFurtwangen,Jakob-Kienzle-Straße17,78054 Villingen-Schwenningen,Germany

ARTICLE INFO Articlehistory:

Received16January2015

Receivedinrevisedform2March2015 Accepted4March2015

Availableonline6March2015 Keywords:

Trainingsimulator

Recombinantproteinproduction Cultivation

Homogenisation Integratedprocess

ABSTRACT

Operatingtrainingsimulators(OTS)arevirtualsimulationtoolsusedfortrainingofprocessoperatorsin industryinperformingproceduresandrunningprocesses.Basedonstructuredmathematicalmodelsof theunitoperationsofabioprocessanOTScantrainaprocessoperatorbyvisualisingchangingconditions duringtheprocess,allowtestingoperatoractions,testingcontrollersettings,experienceunexpected technicalproblemsandgettingpracticeinusingprescribedstandardproceduresforaplant.

ThisworkshowsthedesignofanOTSwheretwosequentialstepsofarecombinantproteinproduction process, a fed-batch cultivationand ahigh-pressurehomogenisation, areintegrated. TheOTS was evaluatedonausertestgroupandshowedthattheOTSpromotedanddevelopedtheirunderstandingof the process,their capabilitytoidentify parametersinﬂuencingprocessefﬁciencyand theskills of operatingit.

ã2015TheAuthor.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/).

1.Introduction

Operatingtrainingsimulators(OTS)arevirtualsimulatorsfor

training of operators in industry or other human activities in

performingproceduresandrunningprocesses[22,17,5,2,18].

Well-knownexamples are OTS for runningprocess units,navigating

airplanes and ships, performing medical surgery and training

militarypersonnelinusingweaponsincombat[3,16,7,15,21].The OTSsimulatetheproceduresofthetechnicalsystemtobetrained onbasedonestablishedmathematicalmodels.ThetypicalOTShas aninterfacethat visualizesthetechnicalsystemand allowsthe traineetointeractwiththevirtualsystemthroughemulatedtools andgearssuchaspumps,valvesandcontrollers.Bythat,theOTS

provides a number of opportunities for training operators by

visualisingdifferentprocessconditions,tryingalternativeoperator actions,solvingtechnicalproblems,changingcontrollersettings andfollowprescribedstandardproceduresforaplant.

InparticularforbioprocessestheOTShavegreatpotentialby

visualizing critical and complex operating procedures in the

bioprocess.Sofar,feweffortsaremadetouseOTSinindustrial

trainingoreducation.Previousstudiesbyushaveshownhowthe OTScanefﬁcientlybeappliedtobiogasdigesters[4],bioethanol fermentation [9], and distillation [13] as well as recombinant proteinproductioninbioreactors[10].

ThesepreviouseffortsinOTSdevelopmenthavebeenmainly

focusedonsingleunitoperationofthebioprocesses(anethanol fermentor,a distillationtower,anaerobicrecombinantbacterial cultivation).However,goodperformanceofbioprocessoperators inindustryrequiresthatsequentialprocessunitscanbeoperated

efﬁciently and simultaneously. What happens in the ﬁrst unit

operation of a process sequence affects the subsequent unit

operationsinwaysthatrequireverygoodunderstandingofthe

process bytheoperatorsand skillsin runningitefﬁcientlyand withoutfaults.

Inthisstudyweshowhowthetwosequentialunitoperationsin

a bioprocess for production of a recombinant protein are

integrated in an OTS. The intracellular recombinant protein is

producedinafed-batchbioreactorfromwhichitssuspensionof

protein-containing cell is transferred to a subsequent

high-pressure homogeniser. The OTS with the two coupled process

unitsallowedtrainingofprocessoperatorsinhowtorunthe high-pressurehomogeniserinrelationtokeyconditionsandparameters

suchaspressure,ﬂowrateand numberofpasses.TheOTSwas

testedonausergroupforevaluatingtheirabilitytocomprehend

* Correspondingauthor.Tel.:+4613281000. E-mailaddress:cfm@ifm.liu.se(C.-F.Mandenius).

http://dx.doi.org/10.1016/j.btre.2015.03.002

2215-017X/ã2015TheAuthor.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4.0/).

ContentslistsavailableatScienceDirect

Biotechnology

Reports

(3)

thedynamicsoftheprocessandtotakeappropriateactionsduring operation.

2.Materialsandmethods

2.1.Microorganism

Escherichia coli strainHMS174(DE3) (Novagen,Madison,WI,

USA)transformedwithplasmidpET30a(Novagen)expressingthe

greenﬂuorescentprotein(GFP)mutant3.1(GFP-mut-3.1,Clontech, US),undercontroloftheT7/lacpromoteranda25bplacoperator

sequencewasusedforGFPproduction.

2.2.Cultivation

Cellswerecultivatedina10Linsitusterilizedbioreactor(Model

LMS 2002,Belach Bioteknik AB,Solna, Sweden)equipped with

standardinstrumentation.Theprocedure,analyticalmethodsand mediacompositionaredescribedin[12]and[9].

2.3.Celldisruption

HarvestedsuspensionoftwocompletedE.colicultivationswere mixed(10L)and storedat4C untiltheusein homogenisation

experiments.The cell suspensions were subsequently

homoge-nised using a Gaulin-Homogeniser APV LAB 40 (SPX Flow

TechnologyRosistaGmbH)withahorizontalthree-pistonpump.

The AVPhomogeniser had a maximum capacity of 45L/h at a

maximumpressureof1000bar.Theexperimentalset-upincluded

atankthatwasconnectedtothehomogeniserandaglassﬂaskfor collectingthehomogenate.Aftereachpassthetankwasre_ﬁlled

withthe homogenate and samples were taken for subsequent

analysis.Cellsuspensionsof5Lvolumeswerehomogenisedat300 and600bar.Thecellsuspensionwascooledto5–10_C_between_the

homogenisationpasses.

2.4.Fluorescencespectrophotometricanalysis

Two-millilitresamplesfromhomogenizationwerecentrifuged

at10,000gfor10minat4C.Foranalysingtheproteinrelease

the supernatant was collected and measuredin a ﬂuorescence

spectrophotometer(FluostarGalaxy,BMGLabtechnologiesGmbH,

Offenburg,Germany).Twohundredmicroliters of1:100diluted

(PBS buffer) supernatant were measured in well-plates at an

excitation/emission of485/520nm.Also, uncentrifugedsamples

fromthetwocultivationswereanalysedascontrols. 2.5.Modelformulation

ApreviouslydevelopedmodelforproteinexpressioninE.coli

[10]was modiﬁedinorder tosimplifyitsstructure,toimprove simulationofproteinexpressionandtofacilitateitsconnectionto amodelforhigh-pressurehomogenisation(seeSection3).Both

modelswereformulatedintheprogramminglanguageC++and

numericallysolvedusinga4thorderRunge–Kuttaalgorithmfor

dynamicsystems.Modelparameterweretakenfromliteratureor

adjustedbycomparingsimulationresultswithexperimentaldata. 2.6.Operatortrainingsimulator

The development of the OTS was carried out by using the

commercial software WinErs (Ingenieurbüro Schoop GmbH,

Hamburg, Germany; www.schoop.de/en/software/winers).

WinErs is a convenient software for development of process

control and automation and includes features for visualisation, datamonitoring, controlandsimulation ofindustrial processes.

TheWinErssoftwarewashereusedfordesigning thegraphical

userinterfaces(GUI),andcontrolschemesand functionsofthe

OTS.WinErswasalsousedforimplementationoftheC++models

asdynamic-linklibraries(DLL).Thein-andoutputsofthemodels whereconnectedtotheGUIstovisualizetheactualstatevariables indatatablesorhistorydiagrams.Bythat,theOTSreplacedthereal

bioreactor and homogenizer with models that mimic their

behaviourinthevirtualenvironmentoftheOTS.WiththeWinErs

M

PIC TIC QIC TIC

2 T

1 T

VIC TIC bioreactor

Fig.1.IntegratedprocessplantincludingabioreactorwithaproducttankandaHPHsystemconnectedtotwoholdingtanks(T1andT2)andaproducttank. 92 I.Gerlachetal./BiotechnologyReports6(2015)91–99

(4)

softwarethesimulationsintheOTScouldberuninreal-timeas wellasinacceleratedtime(1-,5-,10-or15-times).

2.7.VirtualOTStrainingandevaluation

ApplicabilityandfeasibilityoftheOTSwereevaluatedwitha

usertest groupwithbackgroundin bioengineering and similar

biosciences.Thetraineestookpartinaone-dayhands-ontraining

withtheOTSwheretheacceleratedmodeoftheOTSwasusedto

reduce the training time. Detailed written instructions were

available for the trainees beforehand. These instructions were

comparablewithastandardoperationprocedurefortheintegrated processandincludedthetypicalmanualactionsbyoperatorsand

the scheduling of the process sequence. The training was

subsequently assessed from questionnaires and tasks (see

Supplementarydata,3.)aswellaswithindividualinterviews. TheOTStrainingstartedwithvirtualfed-batchcultivationsfor

protein expression and separate homogenisation experiments

usingthevirtualhomogeniserandwascontinuedwithtrainingthe

integrated process with the cultivation and homogenisation

proceduresincludingvaryinginductionandfeedingtimesaswell

as comparing single-pass or continuous homogenisation at

differentpressures.

3.Homogenisermodeldevelopment

Thehomogenisermodelusedinthisstudyisbasedonamodel developedbyHetheringtonet al.[14]fortherelease ofsoluble

protein in a Manton–Gaulin homogeniser. Hetherington et al.

adaptedtheproteinreleasetoSaccharomycescerevisiaeasfunction oftheoperatingpressure(P)andthenumberofpassesthroughthe

homogenizer(N)as:

log Rmax

RmaxR

¼KNPa (1)

whereRmaxisthemaximumamountofsolubleproteinthatcanbe

released,Rthefractionofdisruptedcells,Ktherateconstantanda

a pressure exponent. The exponent a is speci_ﬁc for a certain

microorganismanddependsonthemicroorganism’srobustnessto disruption.Itisrelatedtothecellwallstructurebutmightalsobe inﬂuencedbygrowthconditions,growthrate,cellconcentration

and the temperature of the suspension as suggested in other

reports[11,8,19].Subsequentdevelopmentofthismodelhasbeen donebyAugensteinetal.;Saueretal.;Spidenetal.[1,24,19],and Chooniaetal.[6].

ForintegratingtheHPHwiththeprecedingcultivationstepwe modiﬁedEq.(1) todescribetherelease ofrecombinantprotein (cRPn)afterhomogenisation: cRPn¼cRPn;int 1:010:0KP a ð Þ (2) wherenreferstotank1(T1)ortank2(T2)oftheHPHsystem,cRPn

isthereleasedproteinconcentration(comparabletoR(Eq.(1))) and cRPn,int the intracellular recombinant proteinconcentration

(comparabletoRmax)(Supplementarydata,Eqs.S1:3andS1:4).

Table1

UserneedstobemetinthedevelopmentoftheintegratedOTS. Needs Description (1)Efﬁcientvirtualbioprocesstraininginrecombinant

proteinproduction

Proceduresofhowtocarryoutarecombinantproteincultivationincludingfermentationtechniquesand inductionaswellasasubsequentorcontinuoushomogenisation(downstream)shouldbetrainedwith highﬁdelityandwithinshortertimebyusingtheOTS.

(2)Efﬁcienttransferoffundamentalknowledgeabout high-pressurehomogenisation(HPH)

ThetransferoffundamentalknowledgeaboutHPHsuchaschangingpressureandtemperatureofthecell suspensionshouldbefacilitatedbyusingtheOTS.

(3)Efﬁcientvirtualtrainingincelldisruptionpropertiesof differentmicroorganismsduringHPH

Theanalysisofdisruptionpropertiesofdifferentmicroorganismsduringhomogenisationshouldbe accomplishedeasierandwithinshortertimebyusingtheOTS.

(4)Efﬁcienttransferofunderstandingofprotein denaturationeffectsduringHPH

Theunderstandingofproteinstabilityduringhomogenisationshouldtobereachedwithinshortertimeby usingtheOTS.

(5)Efﬁcienttransferofunderstandingforapplyingdifferent passstrategiesforHPH

Proceduresofhowtorunsingle-passandcontinuoushomogenisationstrategiesshouldbecomprehended easierandwithinshortertimebyusingtheOTS.

Fig.2.Comparisonbetweenexperimentaldatafromhigh-pressurehomogenisationruns(dots)andsimulation(sim.)withthehomogenisermodel(lines)showingthe releasedproteinconcentrationafterhomogenisationat:()300bar(30MPa)and(4)600bar(60MPa).Simulationresultsshowthereleasedproteinconcentrationintank1 (T1)andtank2(T2)foreachpassbytransferringthesuspensionfromT1throughthehomogenisertoT2andviceversa.Thetotalrecombinantproteinconcentrationfroma puresampleofthemixedculturesusedinthehomogenisationexperimentswasaround1.7g/L(&)(otherparametersaregiveninTable2).

(5)

The number of passes of the cell suspension through the homogeniser(NinEq.(1))wasincludedasshowninFig.1.

Thetotalreleaseofrecombinantproteinfromthehomogeniser iscalculatedfrominitiallyreleasedproteinintankT1orT2(cRPn,rel,

Supplementarydata,Eqs.S1:5andS1:6),releasedproteindueto homogenisation(cRPn,Eq. (2))diminishedbyproteindenatured

duringhomogenisation(cRPn,deg,Eq.(5)):

cRPHn;rel¼cRPn;relþcRPncRPHn;deg (3)

Theintracellularproteinconcentrationleftafterhomogenisationis calculatedfrominitialintracellularproteinintankT1orT2(cRPn,int,

Supplementarydata,Eqs.S1:3andS1:4)diminishedbyreleased proteinduetohomogenisation(cRPn,Eq.(2)):

cRPHn;int¼cRPn;intcRPn (4)

Thechangeofproteinconcentrationduetodenaturationduring

homogenisationisdescribedas:

cRPHn;deg¼ctn;relkd (5)

wherectn,relisgivenby:

ctn;rel¼cRPn;relþcRPn (6)

andkdisadenaturationfactordependingonthepressure:

kd¼1:010:0KdP b

ð Þ (7)

whereKdisthedenaturationconstantandbapressureexponent.

Augenstein et al. [1] assumed that protein denaturation

increases withpressure. Using Eq. (2) for deriving the release rateandEq.(7)forthespeciﬁcdenaturationratetheypredictedthe totalproteinyield.Moreover,Chooniaetal.[6]addedobservations ofdecreaseofproteinconcentrationathigherpressureandEngler andCampell[8]didobservationsondisruptionstresseffects.

Basedontheseﬁndings,wedeﬁnedtheparametersK,Kd,anda

inEqs.(2)and (7)andasafunctionofsuspensiontemperature

during homogenisation. For clariﬁcation these parameters are

writtenasK0,Kd0anda0(Supplementarydata,TableS1.2).Adouble

sigmoidalfunctionasdescribedbyGerlachetal.[10]isappliedin

which the suspension temperature is the argument x of the

function(Supplementarydata,Eq.S1:22).

Thehomogenisationtemperatureofthesuspensionisdeﬁned

as TH1 and TH2. TH1 and TH2 are modeled according to the

assumptionthatthetemperatureofthesuspensionincreasesby

2.5Cper10.0MPaofoperatingpressure[20]:

THn¼0:25 K

MPaPþTTn (8)

whereTTndeﬁnesthetemperatureofthesuspensionsfromtank1

ortank2.

Furthermore, we observed decreased ﬂow rates at higher

pressureasalsonotedbyChooniaetal.[6].Thus,alinearequation is used todescribethe changeof theﬂow rate(F) at different pressure(P):

F¼0:25Pþ45:0_hL (9)

Applied model parameters are deﬁned from own data where

45.0L/h is the maximum ﬂow rate without homogenisation

pressure.

Thedescribedkineticexpressionswereintegratedindynamic

massbalancesthatcanbefoundintheSupplementarydata.

4.Resultsanddiscussion

Inordertodeveloptheintegrated “Bioreactor-High-pressure-homogeniser_”OTS_fivecriticalneedsweresetupasgoalsforthe design (Table 1).The needs cover thecapability of the OTS to accomplish (1) efficient virtual interactionwith a recombinant proteinproductionprocess,(2)efficienttransferoffundamental knowledgeaboutHPHincludinginfluenceofpressuregradientand temperatureofthecellsuspension,(3)efficientvirtualtrainingin celldisruptionpropertiesofdifferentmicroorganismsduringHPH, (4) efficient transfer of understanding of protein denaturation

Table2

Parametersappliedinsimulations.

Fig.2 K a Kd b cRP0,int(g/L) cRP0,rel(g/L)

(a) 0.009 1.2 0.1E-05 0.0 1.65(30MPa) 1.4(60MPa)

0.34 (b) 0.01 1.0 0.2E-04 1.5 1.65 0.34

Fig.3.Simpliﬁedmodelschemeoftheproteinexpressionmodelshowingﬂuxesofsubstratetodifferentmetabolicpathways(anabolicandenergypathway),in-and reactivationprocessesofactive/inactivebiomass(qXiX,qXXi),themaintenancemetabolismthatisrelatedtothereactivationprocess,thedegradationofrecombinantprotein

andinactivebiomassandthemortalityrateaswellasthecompartmentstructureoftotalviablebiomass(Xt),deadbiomass(Xd)andreleasedbiomass(Xrel).qG–glucose

uptakerate;qSN–nitrogenuptakerate;qHAc–acetateuptakerate;qIPTG–IPTGuptakerate;Xact–activebiomass;XAA–aminoacidbiomass;RP–recombinantproteininXt;RPd–

recombinantproteininXd;RPrel–releasedrecombinantprotein;Xpl–plasmidbiomass;XS–structuralbiomass;Xi–inactivebiomass;II–intracellularIPTG.Themodelling

principleisshowne.g.,withqGtheglucoseﬂuxwithinthecelltowardsanabolicpathway(YXP)andenergy(ATP)producingpathway(1-YXP).Thetypesofarrowillustratethe

respectivereaction(legend).

(6)

effectsduringHPHand(5)efﬁcienttransferofunderstandingfor applyingdifferentpassstrategiesatHPH.Theseneedsguidedthe designoftheintegratedOTSintoafunctionalsupport.

4.1.Virtualhomogeniserdesign

The layout of the graphical user interface of the OTS was

designedclosetothepipingandinstrumentationdiagram(P&ID) oftheunitsoftheintegratedprocess.AsshowninFig.1,thecell suspensioncomingfromtheproteinharvesttankofthebioreactor ispumpedintoholdingtankT1oftheHPH.ThelayoutoftheHPH includestwostoragetanks,bythatallowingtheuseofdifferent passstrategiesthroughthehomogenizer,forexamplebytransfer ofthecellsuspensionfromholdingtankT1toT2orviceversa,orfor

continuoushomogenisation,byusingadditionalpumps.Sampling

portsallowcollectionofoff-linedatafrombothtanks.Theoff-line dataarepresentedtogetherwithon-linesignalsinseparate

sub-interface diagrams. Furthermore, cooling jackets control the

temperatureofbothholdingtanksandsensorsmonitor

tempera-ture(TIC), operating pressure(PIC)and tank volumes(VIC)for controllingtheprocess.TheseOTSfeaturesrelatetoNeed1aswell asNeed5inTable1.

Furthermore, by using the HPH-part of the OTS separately

(withoutconnectiontothebioreactor)theinitialconcentrationsof

intracellular and released recombinant protein into the cell

suspensionweredeﬁnedbythetrainees. DifferentHPH

experi-mentswerecarriedoutbychangingtheoperatingpressurethat

affectstheﬂowratecapacity(cf.Eq.(9)),therebyaddressingNeed 2.

Fig. 2 shows the simulation results for release of green

ﬂuorescentprotein(GFP) afterhomogenisationwithﬁvepasses

in comparison to data from experimental HPH runs. The

concentrationof GFPin thepuresample isaround1.7g/L.Two

differentsetsoftemperature-independentparameters(K,a,Kd,b)

areusedforadjustingthemodel(Table2).ThereleaseofGFPat 60MPaintheexperimentalrunisnotsigniﬁcantlyhigherthanthe release at 30MPa afterthe ﬁrstpass. Additionalpasses release

slightly more GFPat 30MPa whileat 60MPa theGFP is either

constantorshowatendencytodecrease.

Thelowerconcentrationobtainedat60MPamightbearesultof

the homogeneity of the culture in the experiments. Since the

concentration at 60MPa might be higher, a lower initial

intracellularrecombinantproteinconcentration(cRPO,int)wasused

(Fig.2a andTable2).Comparedtopreviousresultsfromothers

Fig.4.ComparisonbetweenexperimentaldatafromtwoE.colilaboratoryfed-batchcultivationsproducingrecombinantGFP(dots)andsimulationwiththeprotein expressionmodel(lines)showingrelevantbiologicaleffects.TheseeffectscanbespeciﬁcallytrainedwithintheOTS.Glucose(Glc,^),biomass(X,&),acetate(HAc,),GFP (4)concentrations,thevolume( )andtheappliedfeedrateproﬁles( )areshown.(a)Cultivationprocessincludingbatchandfed-batchphasewithinductionat7h. (b)Cultivationprocessincludingbatchandfed-batchphasewithinductionat6.5h.

(7)

(e.g.,Refs.[14,23,6])wherethereleaseofproteinincreaseswith pressure,oursecondsimulationat30MPashowslessreleasethan thedatafromtheexperimentalrun(Fig.2b).TheresultsinFig.2

showtheﬂexibilityofthemodel.BytuningtheconstantKandthe exponentathereleaseofrecombinantproteincanbeadaptedto

the applied HPH system or microorganism. Furthermore, the

denaturation constant Kd and the pressure exponent b can

compensatefordenaturationeffectsof therecombinantprotein

ifit is instable at thepressures and temperatures used in the homogeniser.

TotrainNeed3,fourmicroorganismsofdifferentsensitivityto celldisruptionwereincludedinthesimulations.Theconstraints

usedin thesesimulations werebasedonstudies showinghow

protein release and cell disruption depend on temperature,

pressureandnumberofpassesinthehomogeniser[23,19,6].In

particular the dependence of the type of microorganism and

suspensiontemperaturewas appliedwheretheparameterK,Kd

andaweredescribedasfunctionofsuspensiontemperatureusing asigmoidalfunction(Supplementarydata,Eqs.S1:22andS1:23, TableS1.2), yieldingK0,Kd0 and a0.Inaddition,when disrupting

proteincontainingmicroorganisms ofdifferentstability Need4 couldalsobeincludedintheOTS.

4.2.Integrationofhomogenisermodelwithamodiﬁedprotein expressionmodel

Thetwoprocessingstepswereconnectedbyusingtheoutputof theproteinexpressionmodelasinputforthehomogenisermodel (Fig.1).Intracellularrecombinantproteinwastakenasthesumof recombinantproteinintotalviablebiomass(Xt)anddeadbiomass

(Xd)whilereleasedrecombinantproteinwaspartofthereleased

biomass (Xrel). Both were used as inputs as initial protein

concentrationvalues(cRPO,int,cRPO,rel;Supplementarydata,Eqs.S

1:3and S 1:5) inthe HPHsystemwhen a certain volumewas

transferredintotank1(T1).

Theintegratedprocesswassimulatedinacceleratedtimemode allowingafastertrainingandtimefortestingseveralprocessing strategieswithinalimitedtime.ThisinparticularisrelatedtoNeed 1describedabove.

In this study a previous developed structured model for

recombinantproteinexpression[10]wasmodiﬁedforfacilitating itsintegrationwiththeHPH.Themodiﬁcationwasnecessaryto improvetheaccuracyofthesimulationandtosimplifyparameter adjustment,inparticularformodelingofsubstrateconsumption,

energy pathwayﬂux,recombinantproteinexpressionyieldand

effectsonexpressionduetostressresponse.Theschemeofthe proteinexpressionmodel isshown in Fig.3. Themodiﬁcations includedthemodellingof(1)substrateuptake,(2)compartmental

structure, (3) recombinant protein expression,(4) maintenance

metabolismand(5)diauxicgrowthonacetate.Aswepreviously

describedthesubstrateﬂuxwassplitintoseparatepathwaysfor anabolism(Yxp)andenergymetabolism(1Yxp)[10].Also,theuse

of a double sigmoidal function to describe a shift between

pathways,inhibitionofgrowthandoverﬂowmetabolismwasused formetabolicchangesintheOTS.Bythat,themodiﬁedmodeluses separateglucose(qG),acetate(qHAc)andnitrogen(qSN)uptakerates

that are split intoone anabolic and one energy pathway. This

simpliﬁestheparameterisationofthemodel.

Inthemodelthecompartmentsofthecell'sstructuralbiomass (XS),activeandinactivebiomass(Xact,Xi),plasmids (Xpl),amino

acidpool(XAA)andrecombinantprotein(RP)areportionsofthe

totalviable biomass (Xt).The aminoacidcompartment (XAA)is

included for describing stress response during recombinant

proteinexpression.Therecombinant proteincompartment (RP)

uses available amino acids when the cultureis induced,while

expressionrateandtherateforformationofXactarelimitedbythe

Fig.5. GraphicaluserinterfacesdesignedfortheintegratedOTS.Leftside:GraphicaluserinterfaceforvirtualrecombinantproteinproductionwithE.coliinabioreactor. Rightside:Graphicaluserinterfaceforvirtualhomogenisationusingasmall-scalehomogeniser.Smallersub-windowsshowingthedifferentconﬁgurationsettingsforthe homogenizer(leftside)andthepumpstationwheredifferentpumpscanbecontrolledmanually(rightside).

(8)

amino acid level in the total viable cells (XAA/Xt) (Fig. 3;

Supplementarydata, Eqs.S2:5 and S2:6). The rate of protein

expressionis related both tothe plasmidcopy number,

repre-sentedbyXpl,andtotheactivebiomass,Xact,afterinductionbyin

intracellularIPTG(Ii).Thedegradationofinactivebiomass(Xi)and

RP into XAA (e.g., degradation of misfolded and recombinant

protein) is described by a separate rate. The maintenance

metabolismisconnectedtothereactivationofinactivebiomass (Xi)toactivebiomass(Xact).Forinstance,whenglucoseisavailable

thereactivationrateishighandmoreactivebiomassisformedin thecell.Thus,theactivityofthecelldependsontheratioofactive biomasstothetotalviablebiomass(Xact/Xt).Whenthereactivation

rate(qXiX)islowerthantheinactivationrate(qXXi)Xactdecreases

whileXiincreases.Thereactivationrateaffectstheglucoseuptake

ratebecauseitisconnectedtotheenergymetabolismofthecell.

For example,activebiomasscomponents,suchas enzymesand

RNA, are required for performing metabolic reactions and the

activationofbiomassrequiresenergy.Whentheenergysourceis

depletedthesebiomasscomponentsbecomeinactivewhile still

beingpartoftheobservedbiomassinthemodel.

SubstrateuptakeratesdependontheRPlevelinthecells(RP/ Xt).Inordertoimprovethesimulationofstressresponseeffects

alsothereactivationratedependson(RP/Xt)(Supplementarydata,

Eq.S2:7).Accordingtothetransformationofglucosetobiomass

thetransformationofacetatetobiomasswasimprovedwithout

forming lactate and formate. Furthermore, the maintenance

metabolism was expanded for the growth on acetate (diauxic

growth).

Alsothecompartmentofdeadbiomass(Xd)wasspeciﬁedinto

native protein (Xact, Xi, XAA), structural biomass XS, plasmid

Fig.6. Graphicaluserinterfacesandsub-windowsdesignedfortheintegratedOTS.Viathegraphicaluserinterfaces(Fig.5)differentsub-windowscanbeopenedtocheck off-linedatafromthesamplestakensuchasbiomassandrecombinantproteinconcentrationandobserveon-linesignalssuchasdissolvedoxygentension(DO),temperatures andvolumesfrombothprocessunits.History1:on-linesignals(DO,stirrerspeed,gassingrateandfeedrate)fromthevirtualbioreactor;History2:on-linesignals(operating pressure,volumeoftank1(T1)andtank2(T2))fromthevirtualhomogeniser.

(9)

biomassXplandrecombinantproteinRPdtomodelthereleased

biomasscompartment(Xrel).ThisfacilitatesintegrationoftheHPH

assuming the whole cell suspension (including viable/dead

biomass,intra/extracellularrecombinantprotein)isprocessed.

Fig.4showssimulationswiththemodelincomparisontodata fromtwocultivationexperiments.Itshowstheconcentrationsof glucose,biomass,acetateandGFP.Inbothsimulationsthesame modelparameterswereappliedwhileonlytheinitialactiveand inactivebiomasscompositionwaschangedaswellasthefeedrate

andinductiontimeswereadjustedtotheexperiments.Whenall

initialglucosewasconsumedthefeedingwasstartedtocontinue growth.InductionwithIPTGwasdoneat6.5h(Fig.4a)and7h (Fig.4b).Inthesecondexperiment(4b)samplesweretakenuntil 8hcultivationtimewhilethesimulationisshownuntil14h.Both experimentshadaﬁnalGFPconcentration(26h)ofapproximately 2.0g/Lthatcorrespondtotheconcentrationmeasuredforthepure

sample for homogenization experiments (Fig. 2). The model

manages to simulate all on-line and off-line data from the

experimentalrunwithhighprecision.

4.3.Visualisationofthehomogeniseroperationanditsintegration withthebioreactorintheOTS

ThemodelsfortheintegratedHPHsystemwereimplementedon

theOTSplatformandtwouserinterfaceswithsub-windowswere

designedandlinked(Fig.5).Throughadditionalsub-windows off-linedatasuchasbiomassandproteinconcentrationaswellas on-linesignalssuchasdissolvedoxygentension,feedrate,stirrerspeed, operatingpressureandtemperaturesfromthevirtualprocesswere visible(Fig.6,History1and2).Theinterfacesvisualisedanimated pumpsandpipelinesofthehomogenisationunit(e.g.,single-pass fromtank1(T1)totank2(T2)wherepipelinesbecomegreenifthe useractivatesthevirtualhomogeniser).

4.4.Applicability

TheapplicabilityoftheintegratedOTSwasevaluatedintraining

with an inexperienced test group versus the needs deﬁned in

Table1.Afterthetrainingthetraineesansweredasetofquestions andsolvedtasksthatwererelatedtotheseneeds(see Supplemen-tarydata,3.).Theevaluationilluminatedtheefﬁciencyandtransfer ofunderstandingandskillsinthetraining(Table3).

QuestionsandtasksrelatedtoNeed1whichwerebasedon

on-line signals shown in graphs were solved satisfactory by the

majority of the trainees. Tasks related to Need 2 were solved

successfullybythemajority.ThetasksrelatedtoNeed3,where fourmicroorganismsofvaryingrobustnesstodisruptionwereto becombinedwithgraphsfromrunswithdifferentsettings,were lesssatisfactorilysolved.Fromobservationsandinterviewswith thetraineesitbecameclearthattheyunderstooddifferencesin disruptionpropertiesbutcouldnotapplythisonamicroorganism level.Thatprobablyindicatesthatrunningasimulationwithout

sufﬁcient pre-knowledge about cell physiology and structure

severelyhinderstransferofskills.TasksrelatedtoNeed4and5 werealsosolvedsatisfactorily.

Moreover, two critical prerequisites were necessary to pay

attentionto:

The requirement of having a correct pre-understanding of

engineering terminology and its precise deﬁnition through

properprofessionallanguagecommand

Not to underestimate the understanding of molecular and

physiologicalpropertiesofbiologicalsystems.

ApreviousevaluationbyusontheuseofthesametypeofOTS toolshashighlightedthetransfereffectconcerningactionrelated skills,e.g.,adjustmentofdissolvedoxygentensionand improve-mentofheat transferusing thestirrer[9].Inanotherstudywe

assessed psychological effects during OTS training such as

confusedness and abilityto recollectspeciﬁcevents afterwards

[10].Withtheobservationsfromthesepreviousstudiestogether withthisstudywithitsfocusonunderstandingand

conceptual-isation of the more demandingand advanced operator taskof

integratingtwoentirely differentunit operations,thebeneﬁtof

using OTS for training and transferring operator skills is

unambiguous. The OTS could be demonstrated in a one-day

simulationtraining.AllneedsdeﬁnedfortheOTS(Table1)were fulﬁlledbasedonthesoftwaredevelopment.

5.Concludingremarks

AnintegratedOTSwasdevelopedfortrainingoftheoperators’ skillsofrunningabioprocessconsistingofarecombinantprotein cultivationfollowedbyahigh-pressurehomogenisationstep.Two establishedmodelswereusedintheOTS:(1)astructuredmodel forproteinexpressionextendedwithaccumulationofthetarget proteinandacetateoverﬂow,and(2)amodelforcelldisruptionin ahigh-pressurehomogeniserextendedwithproteindenaturation. Bothmodelscouldeasilybeembeddedinashellstructureofan

OTSpreviouslydevelopedbyus.

TheOTSfortheintegratedprocesspresentedherewasableto

fulﬁll critical training needs when evaluated with untrained

volunteers.Bythat,itwasconvincinglyshownthatthedesignof theOTSallowedtransferoffundamentalunderstandingandskills

for operation of the integrated bioprocess sequence. Future

developmentoftheOTScouldincludeinfluenceofcell concentra-tionandspecificgrowthrateontheefficiencyofdisruptioninthe

homogenizer. Also, the integration of subsequent down-stream

steps,suchas centrifugation and chromatography,would beof

value.Finally,adaptiontootherbioprocesseswouldexpandthe utilityofOTSinthebiotechnologicalindustry.

Acknowledgements

TheauthorswouldliketothankJohanNorénandAndresVeide

at the Royal Institute of Technology, Stockholm, Sweden, for

valuableadvice onhomogenisationtechnique andfor accessto

theirpilotfacilityforHPH.WewouldalsoliketothankDr.Florian KuhnenattheHochschuleBremen,UniversityofAppliedSciences forthekindsupportinusingtheC++simulationsoftwarepackage.

Table3

EvaluationoftheOTSvs.userneeds.

Needs Evaluationa

(1)Efficientvirtualbioprocesstraininginrecombinantproteinproduction ++ (2)EfficienttransferoffundamentalknowledgeaboutHPH +++ (3)EfficientvirtualtrainingincelldisruptionpropertiesofdifferentmicroorganismsduringHPH + (4)EfficienttransferofunderstandingproteindenaturationeffectsduringHPH +++ (5)EfficienttransferofunderstandingforapplyingdifferentpassstrategiesforHPH +++

a

Successfully/agreeable+++;satisfactory++;lesssatisfactory+.

(10)

Furthermore, we thank Linköping University and Hochschule Bremen,UniversityofAppliedSciencesforﬁnancialsupport.

AppendixA.Supplementarydata

Supplementarydataassociatedwiththisarticlecanbefound,in theonlineversion,athttp://dx.doi.org/10.1016/j.btre.2015.03.002.

References

[1]D.C.Augenstein,et al.,Optimization inrecoveryof a labileintracellular enzyme,Biotechnol.Bioeng.16(1974)1433–1447.

[2]M.G.Balaton,L.Nagy,F.Szeifert,Operatortrainingsimulatorprocessmodel implementationofabatchprocessingunitinapackagedsimulationsoftware, Comput.Chem.Eng.48(2013)335–344.

[3]H.H.Bell,W.L.Waag,Evaluatingtheeffectivenessofﬂightsimulatorsfor trainingcombatskills:areview,Int.J.AviationPsychol.8(1998)223–242. [4]A.Blesgen,V.C.Hass,Efﬁcientbiogasproductionthroughprocesssimulation,

EnergyFuels24(2010)4721–4727.

[5]S.Brambilla,D. Manca,Recommended featuresof anindustrial accident simulatorfortrainingofoperators,J.LossPrev.ProcessInd.24(2011)344–355. [6]H.S.Choonia,S.D.Saptarshi,S.S.Lele,Releaseofintracellularb-galactosidase from Lactobacillus acidophilus and L-asparaginase from Pectobacterium carotovorumbyhigh-pressurehomogenization,Chem.Eng.Commun. 200 (2013)1415–1424.

[7]M.Díaz,D.Garrido,S.Romero,B.Rubio,E.Soler,J.M.Troya,Experienceswith component-oriented technologies in nuclear power plant simulators, SoftwarePract.Exper.36(2006)1489–1512.

[8]C.R.Engler,W.R.Campbell,DisruptionofCandidautiliscellsinhighpressure ﬂowdevices,Biotechnol.Bioeng.23(1981)765–780.

[9]I.Gerlach,V.C.Hass,S.Brüning,C.F.Mandenius,Virtualbioreactorcultivation foroperator trainingand simulation:application toethanol andprotein production,J.Chem.Technol.Biotechnol.88(2013)2159–2168.

[10]I.Gerlach,S.Brüning,R. Gustavsson,C.F.Mandenius,V.C. Hass,Operator trainingin recombinantproteinproductionusing astructuredsimulator model,J.Biotechnol.177(2014)53–59.

[11]P.P.Gray,P.Dunnill,M.D.Lilly,Thecontinuous-ﬂowisolationofenzymes,in:G. Terui (Ed.), Fermentation Technology Today, Society for Fermentation Technology,Japan,1972,pp.347–351.

[12]R.Gustavsson,C.F.Mandenius,SoftsensorcontrolofarecombinantEscherichia colifed-batchcultureproducinggreenﬂuorescentprotein,BioprocessBiosyst. Eng.36(2013)1375–1384.

[13]V.C.Hass,S.Kuntzsch,I. Gerlach,K.Kühn,M.Winterhalter, Towardsthe developmentofatrainingsimulatorforbioreﬁneries,Chem.Eng.Trans.29 (2012)247–252.

[14]P.J.Hetherington,M.Follows,P.Dunnill,M.D.Lilly,Releaseofproteinfrom baker’s yeast (Saccharomyces cerevisiae) by disruption in an industrial homogeniser,Trans.Inst.Chem.Eng.79(1971)142–148.

[15]J.-S.Ki,Interactivetrainingsimulatorforaerialworkingplatforminavirtual environment,Comput.Appl.Eng.Educ.19(2009)733–738.

[16]R.Kneebone,Simulationinsurgicaltraining:educationalissuesandpractical implications,Med.Educ.37(2003)267–277.

[17]A.T. Lee, Flight Simulation–Virtual Environment in Aviation, Ashgate PublishingLtd.,London,UK,2005.

[18]D. Manca,S.Brambilla,S.Colombo, Bridgingbetweenvirtualrealityand accidentsimulation for trainingof process-industry operators,Adv.Eng. Software55(2013)1–9.

[19]A.P.J. Middelberg, B.K. O’Neill, I.D.L. Bogle, A novel technique for the measurementofdisruptioninhigh-pressurehomogenization:studiesonE. colicontainingrecombinantinclusionbodies,Biotechnol.Bioeng.38(1991) 363–370.

[20]A.P.J.Middelberg,Microbialcelldisruptionbyhigh-pressurehomogenization, in:A.MohamedDesai(Ed.),DownstreamProcessingofProteins,Humana Press,Totowa,NJ,U.S.A,2000,pp.11–21.

[21]K.Murai,T.Okazaki,Y.Hayashi,Afewcommentsinvisualsystemsofaship handlingsimulatorforseapilottraining:trainingforenteringaport,Electron. Commun.Jpn.94(2011)10–17.

[22]G.Reinig,P.Winter,V.Linge,K.Nägler,Trainingsimulators:engineeringand use,Chem.Eng.Technol.21(1998)711–716.

[23]L.V.Saboya,etal.,EfﬁcientmechanicaldisruptionofLactobacillushelveticus, LactococcuslactisandPropionibacteriumfreudenreichiibyanewhigh-pressure homogenizerandrecoveryofintracellularaminotransferaseactivity,J.Ind. Microbiol.Biotechnol.30(2003)1–5.

[24]T.Sauer, C.W.Robinson,B.R.Glick,Disruptionofnativeandrecombinant Escherichiacoliinhigh-pressurehomogenizer,Biotechnol.Bioeng.33(1989) 1330–1342.