Integrated reactor feeding system –amino acid feedback control via onlineHPLC for productivity gains

Abstract

In this Application Note we describe a fully integrated bioreactor system that auto-

mates all processes including reactor feeding. The automatic reactor sampling system

serves as an enabling control element that links the bioreactor and its reactor control

system to at-line and online assay instrumentation. The system has been successfully

demonstrated to permit fully automatic analysis and reactor feeding through an entire

fermentation process. This system has also demonstrated more than a 30% improve-

ment in product yield.

Author

George Barringer, PhD

David Lau, PhD

Groton Biosystems

Boxborough, MA

USA

Application Note

Biopharmaceuticals

2

Introduction

This study demonstrates the efficiencyand viability of a fully automatic reactorfeeding system to regulate essentialamino acid concentrations in a micro-bial fermentation. These studies wereconducted using an online amino acidHPLC system integrated with an ARS-MReactor sampling system, and a pro-

grammable reactor control system con-trolling amino acid feedstock pumps.

Baseline studies were conducted todetermine the amino acids that wereconsumed by the fermentation processand the rates of consumption.Subsequent studies were conducted inwhich those amino acids were fed tomaintain a desired concentration for

the entire course of the fermentation.Green Fluorescent Protein (“GFP”) wasused as a marker for specific productivi-ty measurement. Initial results indicatethat productivity is positively correlatedto maintaining appropriate levels ofessential amino acids. Future studiesare in progress to optimize the mediaand feedstocks.

GlucoseFeed

Pumps

OPC Client Data RequestAnalytical Data to Server

OPC Data Sent to Reactor Control System

HPLC

TargetAmino AcidFeed

Reactor Control System

Multi -BioreactorAutosampler

SanitaryInterface

Glucose Monitor

Fluid Connection Electronic Data TransferNot to scale

FilterModule

Cell CounterSample Collector

Third Party Data Server,for example, Data Historian

Automated feeding system for optimized bioreactor operation

Figure 1Integrated, automated bioreactor system.

3

Experimental Method

Fermentations were performed usingstock inoculation cultures and stockmedia. The cycle ran for 24 hours.Samples were extracted periodically bythe ARS-M through the course of thecycle and submitted automatically tothe HPLC for amino acid analysis bythe standard method. Data wasreduced to typical report form by theHPLC operating system software.Selected chromatogram data sets wereidentified by the user in a prerun setupprocess, then extracted automaticallyfrom the chromatogram report on eachsubsequent chromatogram assay bythe Groton HPLC Data Wizard forretransmission to the Reactor ControlSystem. The Reactor Control Systempump control algorithms reduced eachselected chromatogram concentrationdata point to a process value, com-pared that value to a desired set point

Figure 2Raw amino acid chromatograms – no feedback control.

value, then calculated a correct feedpump duty cycle for the next forwardinterval. This process iterated to theend of the fermentation cycle.

ResultsAmino Acid Baseline – BL21 E.ColiA baseline experiment was performedusing stock culture and stock media.Feeding was not performed in order tomap the consumption rates of aminoacids to determine potential single

essential amino acid feedstocks. Rawchromatograms from AAA assay areshown (Figure 2) at selected times inthe fermentation cycle.

Note the consumption of serine (Figure 3). Other amino acids alsoexhibited consumption or expressionduring the fermentation. Serine wasselected for the feeding evaluation(Figure 4).

0 hours

4 hours

24 hours

Conditions

Sample: Microbial (E.Coli) – rGFP-BL21 – casamino acid media

Sampling system: Groton Biosystems ARS-M440

Reactor: DASGIP 1.5 L Glass Bioreactor

Reactor interface: Groton Biosystems In-situ Probe and iRIM

Assay interface Groton Biosystems Agilent HPLC Interface with OPC Kit

Output assay: Agilent 1200 Series LC system – Agilent AAA Amino Acid MethodGroton HPLC Data WizardGroton Biosystems Sample Collector – GFP and Negative Control Samples

Process: 1 Day process

Serine

Serine

Serine

4

Serine feedback control – Bioreactorprofile with OPC, rBL21

Feedback control was enabled for thereactor via the Groton ARS-M AgilentHPLC with an OPC Interface Kit con-nected to the reactor control system.This kit acquires automatically selected chromatogram data from each chro-matogram, stores the data in an OPCserver, and retransmits the data onrequest from any OPC-enabled client. Inthis case, this is the Reactor ControlSystem enabled to programmaticallycontrol duty cycles for up to six feedpumps per reactor. The selected aminoacid for assay and feeding for thisdemonstration was serine (Figure 5).

0

0.05

0.1

0.15

0.2

0.25

0.3

50.0

60.0

70.0

80.0

90.0

100.0

110.0

2:24 PM 7:12 PM 12:00 AM 4:48 AM 9:36 AM 2:24 PM 7:12 PM 12:00 AM

pH2.

PV

[pH

], D

O2.

PV

[%

DO

], O

PC

.OP

C.S

erin

e [g

/L]

Timestamp

DO2.PV

OD

Serine

Serine

Figure 3No Feedback Control, rBL21; Selected process value plots from reactor control systemdO2 – Light Blue; OD – Red; Serine – Green. Derived from chromatogram data.

0

50

100

150

200

250

Am

ino

acid

mg/

L

300

350

400

450

500

0 5 10 15 20 25

AspGluSerHisGlyThrArgAlaValMet

PheTrp

IleLeuCys

Serine

Process time (hrs)

Figure 4No Feedback Control; Amino acid profile, rBL21; All amino acids above limit of detectionCasamino acid media. Derived from chromatogram data.

5

In this experiment manual sampleswere collected periodically (dark bluetrace in Figure 5) by extraction througha dip tube using a syringe. The sampleswere analyzed by standard techniqueswith the same HPLC and method usedautomatically by the ARS-M. Thesesamples were filtered according to thesame criteria used by the ARS-M auto-matic sampling system; cell free filteredto 0.22 µm and stored at 4 °C untilassayed. The automatically-collectedand processed serine data are shown inlight green (Figure 5). Each horizontalsegment of the trace represents theprocess value acquired by HPLC AAAassay as transmitted to the reactor con-trol system. The system was pro-grammed to acquire samples and per-form the AAA assay periodicallythrough the course of the entire fer-mentation cycle. Note the correlationbetween the automatic and the manual(“standard”) assay values (Figure 5).

The red trace at the bottom of thescreen represents the duty cycle tracefor the serine feed pump. The feedpump was set to a constant flow rate.The reactor control system proportion-ally turned the pump on and off toachieve an overall serine feed in unitsof mg/min of “on” time. Note that theduty cycle is inversely proportional toOPC serine process value; long whenconcentration is low, short when con-centration is high.

Figure 6 shows the time course of serine concentration through the entireprocess cycle.

Figure 5 Serine Feedback Control – Bioreactor Profile with OPC, rBL21; reactor control system process valueplots for selected parametersOPC Serine Process Value – Light Green; Manual Serine – Dark Blue; OD – Red; dO2 – Light BlueTemperature – Green; Serine Feed Pump Duty Cycle Modulated - Red

Figure 6All amino acids showing serine controlled at original and selected concentration of 250 mg/mL.

0

50

100

150

200

250

300

350

400

450

0 5 10 15 20 25

Am

ino

Aci

d m

g/L

Process time (hrs)

AspGluSerGlyThrArgAlaValMetPheIleLeuCys

0.0

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0

50

100

150

200

250

300

350

2:24 PM 7:12 PM 12:00 AM 4:48 AM 9:36 AM 2:24 PM 7:12 PM

pH2.

PV

[pH

], D

O2.

PV

[%

DO

], T

2.P

V [

°C],

FA

2.P

V [m

L/h]

, O

PC

.OP

C.S

erin

e [g

/L]

Time stamp

DO2.PVFA2.PVpH2.PVT2.PVSerine *(FC)OPC.OPC.SerineOD

OPC Serine

Offline Serine

* (FC) - Offline sample fromsample collector, collected

online

6

Figure 7Controlled feeding study measuring GFP titre.

rGFP-BL21 Serine FeedbackProductivity

An initial study was performed todemonstrate the correlation of struc-tured amino acid feeding to productivityof the fermentation process. Theselected E.coli strain expresses theGreen Fluorescent Protein (GFP) thatwas used as a marker for specific pro-ductivity in this study.

In this study, a control experiment wasperformed as described above to mea-sure GFP titre under control conditions(no feeding). This is shown in Figure 7.In the controlled feeding study GFP(dark blue) is shown to yield 33% moreproduct with continual feeding of oneamino acid, serine, but otherwise iden-tical culture conditions.

Conclusion

Strategic feeding of bioreactors undercontrolled conditions using real timefeedback control and intelligent algo-rithms is shown to provide an improve-ment in yield of desired products. The

yield of GFP, the marker protein trackedin this current study, increased by 33%through intelligent feeding of one aminoacid, serine, shown to be essential forthis E.coli strain.

Acknowledgement

Groton Biosystems acknowledges theMIT Bioprocess EngineeringLaboratory, Agilent, and DASGIP fortheir contributions to this study.

7

www.agilent.com/chem

© Agilent Technologies, Inc., 2010 May 1, 2010Publication Number 5990-5674EN

Groton Biosystems Publication Number MKT-196www.grotonbiosystems.com