Schaepper 2009 Online 2ndAuthor

8/8/2019 Schaepper 2009 Online 2ndAuthor

http://slidepdf.com/reader/full/schaepper-2009-online-2ndauthor 1/17

REVIEW

Application of microbioreactors in fermentation processdevelopment: a review

Daniel Schäpper & Muhd Nazrul Hisham Zainal Alam &

Nicolas Szita & Anna Eliasson Lantz & Krist V. Gernaey

Received: 16 March 2009 /Revised: 30 June 2009 /Accepted: 6 July 2009 /Published online: 2 August 2009# Springer-Verlag 2009

Abstract Biotechnology process development involvesstrain testing and improvement steps aimed at increasingyields and productivity. This necessitates the high-throughput screening of many potential strain candidates,a task currently mainly performed in shake flasks or microtiter plates. However, these methods have somedrawbacks, such as the low data density (usually onlyend-point measurements) and the lack of control over cultivation conditions in standard shake flasks. Micro- bioreactors can offer the flexibility and controllability of bench-scale reactors and thus deliver results that are morecomparable to large-scale fermentations, but with theadditional advantages of small size, availability of onlinecultivation data and the potential for automation. Current microbioreactor technology is analyzed in this review paper, focusing on its industrial applicability, and directionsfor future research are presented.

Keywords Microbioreactor . Fermentation.Bioprocess monitoring. Microfluidic suspension culture

Introduction

Microbioreactor technology offers the potential to developdisposable and miniaturized versions of bench-scale bio-reactors for carrying out fermentation experiments. Such ascaled down version of a bioreactor would then ideallycombine the ease of handling of shake flask operationswhile preserving the online sensing and control capabilitiesof stirred-tank bench-scale bioreactors. Microbioreactor technology thus offers academia and industry the capacityto acquire real-time experimental data via cheap and high-throughput experimentation under well-controlled condi-tions [1].

Bench-scale bioreactors typically operate with a volumeof 0.5 – 5 L. Such bioreactors are the most versatileexperimental tool available for studying fermentation processes. They enable cultivations in batch, fed-batch,and continuous modes under a broad range of environmen-tal conditions. Indeed, temperature, pH, and — at least for aerobic fermentations — dissolved oxygen concentrationsare typically quantified on-line with well-established probes, and each variable can be maintained close to adesired set point or set point profile with a suitablecontroller. In addition, bench-scale bioreactors usually offer online measurement of acid and base consumption used for pH control, substrate consumption, off-gas composition(especially CO2 ), and sometimes also biomass concentra-tion [2].

Microbioreactors with working volumes of between 50to 800 μ L have been realized (see Table1). Due to their small size, they offer a number of cost-reducing advantages

D. Schäpper : M. N. H. Z. Alam: K. V. Gernaey (* )Department of Chemical and Biochemical Engineering,Technical University of Denmark,2800 Lyngby, Denmarke-mail: [email protected]

M. N. H. Z. AlamDepartment of Bioprocess Engineering,

Faculty of Chemical and Natural Resources Engineering,Universiti Teknologi Malaysia,Johore, Malaysia

N. SzitaDepartment of Biochemical Engineering,University College London,London WC1E 6BT, UK

A. Eliasson LantzDepartment of Systems Biology,Technical University of Denmark,2800 Lyngby, Denmark

Anal Bioanal Chem (2009) 395:679 – 695DOI 10.1007/s00216-009-2955-x



for assessing biological processes [3-5]. These includelower running costs due to the reduced amount of substrateand utilities such as water or power consumption needed per experiment, and reduced space requirements for paralleloperation due to their small footprint. Furthermore, compared to bench-scale bioreactors, disposable microbioreac-tors also reduce the labor and effort required to preparefermentation experiments; for example, they do not have to be cleaned and sterilized, and there is the potential toautomate experiments.

The scope of this review is restricted to microbioreactorsthat allow fermentations with microorganisms growing insuspension, since the majority of industrial fermentationsare based on this type of process. Moreover, in this review,we will only consider microbioreactors with an operatingvolume smaller than 1 mL, which are designed to mimictypical bench-scale reactors. Therefore, the term“ microbioreactor ” will exclusively refer to this type of reactor throughout the manuscript. Larger reactors have already been looked at in other reviews [4-6].

This review starts with a concise overview of thematerials most commonly used for microbioreactor fabri-cation. Once the reactor is established, the key requirementsto obtain a successful fermentation experiment are then anadequate temperature and supply of nutrients, sufficient mixing, and — for aerobic fermentations — a sufficient supply of oxygen. The second part of the review thereforefocuses on mass and heat transfer in microbioreactors. Inthe third part of the review, the state of the art in on-linemeasurement techniques for microbioreactors is highlight-ed, and methods currently used to control the mainmicrobioreactor variables — temperature, pH, dissolved ox-ygen concentration — are presented. The final part of thereview consists of an extended discussion of microbior-eactor technology considering the needs of the fermentationindustry, as well as the outlook for future developments.

Materials

Microbioreactor prototypes are typically made from poly-methylmethacrylate (PMMA) and polydimethylsiloxane(PDMS) [7-11]. These are the most commonly usedsubstrates for the fabrication of microbioreactor prototypesdue to their low material cost, easy handling, biocompati- bilities, nontoxicities, good solvent and chemical compati- bilities, and durabilities. Additionally, they also exhibit highoptical transmission in the visible wavelength range, whichfacilitates optical measurements [12, 13]. Moreover, usingPMMA and PDMS substrates, it is possible to fabricatetwo- or three-dimensional (2D or 3D) microfluidic geom-etries using a rather straightforward and relatively cheapmicromachining process [14, 15]. Whilst the bonding (or structuring) of multiple layers of PMMA is normally done T

a b l e 1

C u r r e n t m i c r o b i o r e a c t o r p l a t f o r m s a n d t h e i r b a s i c c h a r a c t e r i s t i c s

S o u r c e

V o l u m e

( μ L )

O r g a n i s m a n d

f e r m e n t a t i o n m o d e

M i x i n g

M i n i m i z a t i o n

o f e v a p o r a t i o n

k L a

( h − 1 )

O D

m a x .

O D m e a s .

D O

m e a s .

C o n t r o l

p H m e a s .

C o n t r o l

T e m p e r a t u r e

r e g u l a t i o n

[ 1 0 ]

5 – 5 0

E .

c o

l i , b a t c h

D i f f u s i o n

H u m i d a i r

6 0

8

T r a n s m i t t a n c e

O p t o d e

N o n e

O p t o d e

B u f f e r e d

H e a t e d c h a m b e r

[ 2 7 ]

5 0

S .

c e r e v i s i a e , b a t c h

S t i r r e r o n a x i s W a t e r r e p l e n i s h m e n t

f r o m e l e v a t e d r e s e r v o i r

-

6 . 8 7


O p t o d e

N o n e

O p t o d e

-


[ 9 ]

1 0 0

E .

c o

l i , b a t c h

P e r i s t a l t i c

O x y g e n f e e d

h u m i d i f i e d

5 0 0

4 0


O p t o d e

O x y g e n c o n c .

i n m i x e r

O p t o d e

F l u i d i c

i n j e c t i o n

F o i l h e a t e r o n

b a s e p l a t e

[ 2 3 ]

2 5 0

E .

c o

l i , b a t c h

S h a k e n

s t e e l b e a d

S e a l e d

1 5 0

2 . 5


N o n e

N o c o n t r o l ,

e l e c t r i c a l

g e n e r a t i o n

I S F E T

E l e c t r . C O

2

g e n e r a t i o n

T h e r m i s t o r a n d

r e s i s t i v e h e a t i n g

[ 7 , 8 ]

1 5 0

E .

c o

l i [ 7 ] , b a t c h ,

S .

c e r e v i s i a e [

8 ] ,

b a t c h

S t i r r e r o n a x i s W a t e r r e p l e n i s h m e n t

f r o m e l e v a t e d r e s e r v o i r

2 0 – 7 5

6 [ 7 ] ,

7 [ 8 ]


O p t o d e

N o n e

O p t o d e

B u f f e r e d


[ 5 0 ]

7 0 0

E .

c o

l i , b a t c h

S h a k e n

S e a l e d

-

-

S c a t t e r e d l i g h t N o n e

N o n e

O p t o d e

F l u i d i c

i n j e c t i o n

I n c u b a t o r

680 D. Schäpper et al.



using replication methods like injection molding or hot embossing [14, 15], the fabrication of PDMS-based reactorsmerely involves a few casting and curing steps [14, 16].Pressing a PDMS slab between two PMMA layers (seeFig. 1 for an example) with metal screws or clamps is analternative way to ensure watertight seals in multilayeredreactors consisting of PMMA-PDMS substrates [11]. Their low material cost coupled with their cheap fabrication processes are ideal characteristics for low-cost mass production and rapid prototyping. This also allows for thefabrication of disposable microbioreactors — a useful ad-vantage, as disposable reactors prevent sample contamina-tion and reduce the effort involved in reactor preparationsignificantly.

An important advantage of PDMS substrates over other thermoplastic polymers like PMMA, polyetheretherketone(PEEK) or polycarbonate (PC) is that PDMS enables theintegration of microfluidic devices such as microvalves,micropumps, and mixers as an integral part of the reactor [9, 17]. While the integration of microfluidic devices maycomplicate the reactor design, it — more importantly —

decreases the cost of the entire microbioreactor setup because fewer macroscale devices are needed to drive thesystem. Furthermore, PDMS also exhibits high permeabil-ity to oxygen and carbon dioxide, which makes it highlysuitable for cell-based systems [14]. Despite all of theseadvantages, PDMS also suffers from several drawbacks.One drawback is that the gas permeability of PDMS canlead to unwanted levels of evaporation [11, 14]. Evapora-tion is discussed further later on in this review (see thesection on aeration and evaporation). PDMS also swells inmost organic solvents [14], although this is not really anissue for standard microbial fermentations because most fermentation media are prepared using water as a solvent.However, the swelling could become a problem for novelreactor designs where a solvent is added to the fermentation

broth; for example, to remove product or inhibitorycompounds from the water phase in an attempt to increasethe productivity [18, 19].

Silicon and glass can also be used as substrates for thefabrication of microbioreactors, but these are rather difficult to work with. Moreover, the fabrication techniques used for these substrates are expensive, time consuming, and requireaccess to clean-room facilities [16].

Mass and heat transfer in microbioreactors

Adequate mass transfer and heat transfer are required inorder to achieve successful fermentation experiments, andthis of course also applies to fermentation in microbior-eactors. Therefore, this part of the manuscript focuses onestablished mixing and aeration concepts in microbioreac-tors, the importance of minimizing evaporation, and micro-fluidic pumping for different modes of operation.

Mixing

Mixing is one of the most important operations in asubmerged cultivation: the mixing quality directly influen-ces the distribution and suspension of substrate and micro-organisms in the reactor, and thus affects both the growthcharacteristics (e.g., the availability of oxygen andnutrients, removal of waste products, etc.) and the qualityof the measurements. Also, optimal mixing ensures an eventemperature profile throughout the reactor. Mixing becomesmore difficult as the scale of the reactor decreases, as it becomes impossible to achieve the turbulent conditionsassociated with good mixing. This is due to the smallReynolds numbers that can be achieved, which in turn can be explained by the small characteristic lengths [20]. Inaddition, the influence of the volume boundaries becomes

Fig. 1 a Longitudinal section of a microbioreactor consisting of alternating layers of PDMS and PMMA clamped together with screws.b Solidmodels of the different layers with a PMMA waveguide.c Solid models of the different layers with a PDMS waveguide [7]

Application of microbioreactors in fermentation process development: a review 681



much more apparent due to both the molecular forces present and the much higher proportion of cells growing onthe walls compared to, for example, a bench-scale fermen-tor. If the system also contains particles (e.g., cells), thenthe mixing problem is further accentuated: the mixingsystem must then also provide some updraft that lifts the particles, and dead zones (with insufficient mixing) must beavoided even more rigorously, as they would lead, for example, to the local accumulation of particle heaps on thereactor floor.

Optimal mixing is a compromise between ensuringhomogeneous conditions and efficient mass and heat transfer and avoiding damage to the cells, and thus affectsthe quality of the achieved growth. Key here is a trade-off where mixing is“ as good as necessary” while being “ asgentle as possible.” The severity of the mixing problemdepends very much on the type of microorganism used(Saccharomyces cerevisiae cells will, for example, sediment much quicker thanEscherichia coli cells, but they are quiterobust; also, different microorganisms can have different sensitivities to concentration gradients), on the reactor geometry (shape, such as the ratio of height to diameter),and on the mixing scheme chosen.

Mixing can be achieved in microfluidic bioreactors byactive methods that utilize moving parts to disturb the fluidand passive methods that have no moving parts [21, 22].Passive mixing can also be subdivided into molecular diffusion and chaotic advection. In molecular diffusion, thenatural Brownian motion of the molecules is utilized. Goodmixing is then achieved with large interfacial areas, largegradients, and high diffusion coefficients. Large interfacialareas can be achieved by, for example, laying many thinlayers of the fluids to be mixed upon each other, similar toan alternating stack of differently colored paper sheets.Thus, the same molecular diffusion rate becomes a moreeffective mixing operation. Chaotic advection further enhances mixing by stretching, folding, and breaking upthe fluid streams. This is achieved by forcing motiontransverse to the flow direction upon the fluid, byfabricating obstacles that are placed in the channel for instance.

When active mixing is used, it is important to take intoaccount the fact that microbioreactors often consist of completely filled volumes without any headspace. Shakingthe entire reactor will only move the small liquid volume asa whole; very little mixing will be induced. Therefore, insuch reactors, there is a need to actively induce the mixinginside the chamber of the microbioreactor. This will also aidin the retention of microbial cells in suspension. Mimickingthe agitation methods of larger-scale reactors, many microbioreactors include a stirrer bar mounted on and revolvingaround a rigid vertical post [7, 8] (Figs. 2 and 3). These bars can also potentially be microfabricated to mimic the

shapes of larger-scale impellers. Whilst these systems createdefined liquid movement in the reactor volume, theyrequire very precise fabrication of both the impeller andalso the axis needed to mount the impeller. The utilizationof this type of microimpeller cannot guarantee that there areno dead zones in the reactor: the corners between thevertical walls and the horizontal floor are typically the most critical points. A simpler solution has been achieved byreplacing impellers with steel beads [23], thus avoiding theneed for precise manufacturing. However, the fact that anoptical density of only 2.5 was reported by Maharbiz et al.[23] may indicate that mixing with steel beads does not work properly for high cell concentrations.

Another active mixing approach involves moving the boundaries of the reactor. Lee et al. [9] arranged a series of inflatable air cushions in the ceiling of the reactor chamber.Periodic inflation of these cushions created peristalticmovement of the liquid in the reactor. Cultivations donewith E. coli successfully demonstrated that high celldensities can be achieved. Zanzotto et al. [10] fabricatedshallow microfluidic bioreactors with operating volumes of between 5 and 50μ L. Due to the shallow design of thereactor chamber and the motility of theE. coli cells used,diffusion was efficient enough to support growth. Thefermentation results of Zanzotto et al. [10] compared wellwith those obtained from 500 mL bench-scale reactors.

Passive mixing requires some kind of flow of the broththrough a microchannel, for example by recirculating thefermentation broth in and out of the reactor chamber.Molecular diffusion can be enhanced with, for example, parallel lamination, where the broth stream is split intovarious substreams and is later rejoined, preferably in adifferent order so that different streams meet each other.While this will mix the fluid, the channel length will bequite substantial (depending of course on fluid velocity,etc.). Chaotic advection can significantly reduce channellength and thus increase mixing efficiency.

One possibility here is to use passive mixing structuressuch as the staggered herringbone mixer [24] that impose atransverse swirling motion upon the fluid (see Fig.4).Another possibility is to increase the recirculation flowvelocity such that the fluid motion in the reactor chamber isno longer truly laminar but approaches transition phaseconditions, which in turn reduces diffusion lengths. Thiscan, for example, be implemented with a system of threeserially connected reactors where the broth is pumped backand forth between the two outer chambers using the middlechamber as a mixing chamber [25]. Mixing can then further be enhanced by arranging the inlet/outlet ports eccentricallysuch that the fluid is forced into a swirling motion in themain reactor chamber, which further reduces diffusionlengths and thus increases the quality of mixing. Whilst passive mixing elegantly removes the need for some kind




of active mixing in the chamber, it creates additional fluidicrequirements, such as pumps to move the liquid through the passive mixing elements and possibly valves to separate therecirculation stream from the dilution flow (for example), asneeded for continuous cultivations. Also, as with activemixing, great care must be taken to prevent dead zoneswhere particles could sediment.

While many of these methods have been shown to createsufficiently thorough mixing for the fermentation of micro-organisms, mostly E. coli , they often require delicatemanufacturing. Whilst a micro-impeller, for example,requires very precise manufacturing of both the impeller itself and the vertical axis that supports and guides theimpeller, a system with moving boundaries [9] requireselaborate pneumatic installations. Both systems are alsotechnically feasible for mass fabrication. They will,

however, result in higher production costs than amechanically simpler solution. Therefore, ideally, it would be desirable to find a simple and cheap mixing solutionthat provides sufficient mixing for a wide range of microorganisms.

Aeration and evaporation

Aeration is relevant for aerobic fermentations, and isachieved in most if not all microsystems through a thin polydimethylsiloxane (PDMS) membrane [7-10]. Thismethod allows for the diffusion of both oxygen as well asoff-gases at sufficiently high rates whilst also ensuringcontinuous sterility of the fermentation broth. Additionally,if other parts of the reactor are also fabricated from PDMS,handling of the fragile membrane then becomes very easy.

Fig. 2 Schematic of a microbioreactor setup complete with actuation, fluidic connections ( solid lines ), and optical fibers (dashed lines ) for thesensing of OD, DO, and pH [48]




Indeed, the membrane can be directly bonded to other PDMS elements, making the membrane an integral part of another reactor component.

Unfortunately, the membrane also allows for the diffu-sion of other molecules such as water vapor out of thereactor [26]. With evaporation rates on the order of 5 μ L h−1 (water at 37 °C [27]), a 100 μ L reactor wouldcompletely dry out after 20 h. Apart from increasing theconcentrations of cells, substrate, and products in theremaining liquid, the gas bubbles that appear in the reactor due to evaporation may well disturb any online measure-ments in the microbioreactors, rendering them useless.

This, therefore, calls for solutions that either prevent or minimize the evaporation of water or rely on continuous

replenishment of the evaporated volume. In most cases thefirst approach is utilized by placing the whole reactor into ahumidified chamber, thereby reducing evaporation out of the reactor to negligible amounts [7]. This, however, makesdirect access to the reactor more difficult, and also requiresthat all connections to and from the reactor lead into thechamber. Another solution is to passively replenish theevaporated water by connecting an elevated water reservoir to the reactor, which places the reactor under a slight overpressure, as demonstrated in [7]. Whenever water evaporates, the same amount automatically flows in fromthe elevated reservoir. This method efficiently keeps thewater volume in the reactor constant, although it increasesthe risk of contamination too. Also, this solution is difficult to install if the cultivation needs to run continuously or infed-batch mode, since it creates an input/output port withundefined characteristics. Therefore, in continuous or fed- batch operation, preventing evaporation instead of replen-ishing water seems to be the best solution.

Fluidics (pumping mechanism, connections)

Fermentations can be run in a variety of operating modes,ranging from batch fermentations to fed-batch to continu-ous cultivations. Batch cultivations are (from the fluidic point of view) the easiest to handle as it is only necessary tocope with the evaporation of culture liquid. In continuousand fed-batch cultures, on the other hand, it is necessary tohandle the evaporation and also to control the in- andoutflow of liquid to the reactor. If passive replenishment of the evaporated water is used, then some kind of valve isneeded to ensure that the inflow of fresh culture medium isnot simply pushed out through the water replenishment port. On the other hand, if the evaporated water is to bereplenished actively (e.g., with a pump), then the evapora-tion rate must be measured precisely, as any difference between the water replenishment rate and the evaporation

Fig. 3 Photographs of the mixing of phenol red dye in a microbioreactor using a stirring speed of 180 rpm [74]

Fig. 4 Effect of a passive mix-ing element (staggered herringbone mixer). One mixing cycleconsists of two sequentialregions of ridges; the directionof asymmetry of the herringbone

switches with respect to thecenterline of the channel fromone region to the next [24]




rate will result in the dilution or concentration of the culture broth. For continuous culture microbioreactors (chemostat operation, constant reactor volume), the obvious solution isto control the water replenishment rate such that the liquidflow rate in the output equals the feed flow rate.

External syringes or peristaltic pumps are usually used to pump fresh culture medium (see Fig.2 for an example) intothe reactor, which then also passively (due to the fixedreactor volume) pushes the same amount of culture brothout of the reactor in the case of a continuous reactor withconstant volume [8]. This is an easy solution to use in thelab; however, it poses some problems with respect to parallelization. Using one pump for multiple reactors may be a solution, but this will lead to equal flow rates for allreactors unless more elaborate flow-splitting systems areused.

One solution is to use a system of micropumps andmicrovalves [17] operated by pressurized air (Fig.5). Thisallows for individual flow rates for each reactor, but alsorequires a number of air connections (including switchingand flow metering circuitry) for each reactor, which addscomplexity to each reactor in terms of fabrication require-ments and installation difficulty. This type of installationhas only been applied in a perfusion reactor system [17],although there is nothing that fundamentally prevents itsapplication to suspension systems.

Onboard micropumps have been investigated by variousresearchers [17, 28, 29]; Lee et al. [9] also proved their

applicability to fermentation microbioreactors by injectingminute amounts of acid/base into the reactor chamber.Micropumps can also be used as a driving mechanism for recirculation flow, thus making peristaltic or syringe pumpsobsolete. The use of micropumps may on the one hand bean important cost-reducing factor, and on the other hand it also makes the multiplexing of microbioreactors mucheasier. This in turn would make mixing schemes without stirrer bars more attractive.

Fluidic connections between the macro and the microworld have been investigated in great detail, and thesestudies have spawned a large number of solutions [30].Though most of these solutions were generally intended tosolve macro-to-micro world interfacing issues for micro-fluidic devices, some of them have been adapted (directlyor with slight modification) to microbioreactor systems. For microbioreactors consisting of multiple PMMA-PDMSlayers, fluidic connections are typically established on thesolid layer (PMMA) of the reactor, as this allows for adefined fluidic interconnect. In the simplest approach, aneedle that can be connected directly to a syringe or tubingis glued into a channel in a PMMA layer. Whilst this typeof connection is easily established and only requiresmaterial that is typically present in any laboratory, thedisadvantage of this approach is that it only allows for connections in the plane of the PMMA layer (in most casesthis is horizontal). Additionally, there is a significant risk of glue/epoxy seeping into the needle and thus clogging it. A

Culture area

Reversed reservoir

Medium reservoir

Waste reservoir

PBS reservoir

Micropump

Microvalve

Microchannel

Air inlet

Heater Sensor

A

20 µm

150 µm

Magnification of area A

Fig. 5 Schematic illustration of an automatic cell culture system.

It consists of a cell culture area,four micropumps, four microcheck valves, microchannels,reservoirs, two heaters and amicro temperature sensor [17]




more advanced solution uses O rings embedded in aPMMA layer. Rigid tubes can be stuck through these(slightly too small) O rings, thus achieving tightness [31].This method allows for repeated plugging and unpluggingof the connection in an easy manner. The connecting tubewill then stand normally to the PMMA plane and can reachto any desired depth in the microbioreactor. It requires twosealing interconnects to make it tight: the surface betweenthe tube and the O ring (mostly defined by how oversizedthe needle is), and that between the O-ring and the deviceitself (largely defined by the axial or radial space into whichthe O ring is pressed). This connection type obviouslyrequires precise fabrication of the O rings and thecorresponding PMMA layers and thus skilled fabrication personnel. If the reactor is designed to be placed into aholder or docking station, the fluidic tubes can be rigidlymounted on the holder. Surface-contact gaskets then seal between the holder and the clamped reactor [9]. Thissolution is very elegant, although quite inflexible. As such,it is only applicable to systems that have fixed outer dimensions and fluidic connection locations. A middle pathcan be pursued by using specialized connectors (e.g., fromLee, Upchurch, and Vici Jour) for a standard tube-nut assembly, as illustrated in Fig.6. This type of connectionhas a very small and defined dead volume and may even provide a plug-and-play solution for fluidic connections of microbioreactor systems. It is, however, relatively costly.Finally, for a PDMS-based reactor, the simplest solutionwould be to establish a needle – diaphragm interconnect where the needle is pushed through the PDMS itself, so that the contact pressure between the needle exterior and theenclosing PDMS then ensures adequate tightness (Fig.7).This may offer an easy-to-use solution for microbioreactor systems, but misalignment or improper interconnect sealing

may compromise the leak-free operation of the reactor. Asystem with a docking station would also be feasible for areactor with a finalized design. In this case, the holder would also encompass needles that pierce the reactor material when the reactor is placed in the docking station.

These systems all have their specific (dis)advantages andwork fine in an appropriate lab environment, but they areseldom applicable to low-cost single-use systems due toeither price or integration difficulties. Consequently, as yet there is no single state-of-the-art solution which solves most problems satisfactorily. However, Kortmann et al. [32] haverecently proposed a clamping system for lab-on-a-chipdevices that focuses primarily on the fluidic and electricalconnections. A system of spring elements for eachconnection allows for precise adjustment of the sealing pressure. This system addresses many of the abovemen-tioned problems, and in the future it may well become thesolution of choice for microbioreactors.

Sensing and control

While mass and heat transfer challenges must be resolvedin order to enable fermentations, sensing [33] and controlover the fermentation are also crucial to obtaining mean-ingful results. Indeed, the growth and production rates of the microorganisms usually depend strongly on pH andtemperature. Control over these variables is therefore arequirement for performing fermentation experiments under relevant conditions that lead to high productivity. This part of the review therefore focuses on the key fermentationvariables — temperature, pH, and dissolved oxygen — whilst also considering some other measurements that couldgreatly increase the amount of information per experiment.

Fig. 6 Fluidic connections with a standard tube-nut assemblymounted on a rigid part of the microbioreactor

Fig. 7 Needle – diaphragm interconnect where the needle is pushedthrough the PDMS layer into the reactor chamber




Temperature

In microbioreactors, temperature is typically measured bymeans of thermistors or resistance temperature detectors(RTDs) [9, 23, 25, 34]. Thermocouples are used in somesetups [7, 8, 10, 27]. RTDs are sensors that have aresistance that varies according to the temperature; RTDsmade from platinum are preferably used (e.g., Pt 100 or Pt 1000 sensors; these have nominal resistances of 100 and1000 Ω at 0 °C, respectively) [35]. RTDs are very cheapand are commercially available in bulk quantities. Sincethey are mass-fabricated for many industries in relativelysmall sizes (5×2×1.3 mm), this type of sensor can easily beembedded into a microbioreactor to precisely measure thetemperature of the microbioreactor content [34]. They arefairly accurate (to ~0.1 °C) [36], and are able to operatereliably for long periods of time [34]. Moreover, RTDs canalso be connected to an external circuit with connectinglead wires or wire bonded onto a printed circuit board for online temperature measurement and control [23, 36]. Suchfeatures are crucial to parallel operation at various operatingtemperatures. Ideally, for precise temperature measurement,the temperature sensor should be located close to themicrobioreactor chamber. However, in several cases it wasreported that temperature was measured at other parts of thesetup. Lee et al. [9] controlled the temperature of themicrobioreactor by measuring the temperature of the base plate holding the microbioreactors. Vervliet-Scheebaum et al. [37] integrated the Pt 100 sensor into a flow-throughmeasuring device to measure the temperature of the liquidthat was circulating in the microbioreactor system. Zanzottoet al. [10], Szita et al. [7], Boccazzi et al. [27], and Zhang et al. [8] measured the temperature of the circulated heatedwater that was used to control the temperature of themicrobioreactors or that of the enclosing chamber.

Whilst temperature measurement is relatively easy toachieve in microbioreactors, temperature control is a rather challenging task due to the high surface area to workingvolume ratio (S/V). For instance, a conventional bench-scale bioreactor with a height-to-diameter (H/D) ratio of 2would have an S/V ratio of about 0.9 [38]. In contrast,microbioreactors with a chamber diameter of 10 mm and aheight of about 2 mm would give a S/V ratio of about 1500[8]. This implies that heat transfer in microbioreactor systems is large and very rapid compared to conventionalsystems. One also needs to bear in mind that heat loss bynatural convection is significant and inevitable whenworking at such a small scale. Another factor that affectsthe efficiency of temperature control in microbioreactorsis the location of the heating element. It is necessary to place the heating element in a position where it will not create a large temperature gradient [34]. This is particularlyimportant for microbioreactors fabricated from poly(methyl

methacrylate) (PMMA) and polydimethylsiloxane (PDMS), because these materials have very low thermal conductiv-ities of about 0.2 W m−1 K −1 [39] and 0.17 W m−1 K −1

[40], respectively.Temperature uniformity in the microbioreactor chamber

is another critical issue when establishing a good temper-ature control system for a microbioreactor. There areseveral techniques that can potentially be used to probetemperature uniformity in the reactor chamber. First, onecould place several temperature sensors at different spotswithin the reactor chamber to evaluate the temperaturedistribution. Second, thermal distribution patterns in thereactor chamber can also be examined by monitoring heat images using a thermo or infra-red camera [41, 42]. Finally, based on a model describing mass and heat transfer in amicrobioreactor, one could also opt to simulate a three-dimensional (3D) temperature distribution profile in thereactor chamber using, for example, the finite element method [8, 34], which is now a standard engineering toolimplemented in several software packages. As a rule of thumb, however, one should, in our opinion, assume that amixing system that can keep cells in suspension will also provide for adequate temperature uniformity.

Several alternatives are available to efficiently controlthe microbioreactor temperature. The simplest method is toconduct the experiment in a temperature-controlled room[37] or in an incubator [43]. Vervliet-Scheebaum et al. [37]reported that the temperature variation can be kept to within1.5 °C of the desired set point in a temperature-controlledroom. Although the temperature can be accurately con-trolled with this method, and it also allows for the operationof many microbioreactors in the same space at the sametemperature, it is of course not feasible for the paralleloperation of microbioreactors at different operating temper-atures, such as when screening for the optimal growthtemperature of a strain. Another option is to link the deviceto a water bath and circulate thermostated water through themicrobioreactor chamber base [7, 8, 10, 27]. The maindrawback of this method is the limitation it imposes on the parallel operation of reactors at different temperatures.Also, the additional fluidic system increases the risk of aliquid leak leading to failure of the temperature controlsystem. Temperature control in microbioreactors can also be accomplished by integrating electrical microheaters [9,23, 34, 36]. Maharbiz et al. [23] and Krommenhoek et al.[36] incorporated a microheater onto a printed circuit boardthat was adhered to the base of the microbioreactors. Themicroheater was positioned next to the thermistors such that tight temperature control of the microbioreactor content could be achieved. Lee et al. [9] implemented a commer-cially available on-off controller (Minco, Fridly, MN, USA,CT325TF1A5) to control the temperature of the base platethat served as the microbioreactor holder. The micro-




bioreactor temperature was controlled by maintaining thetemperature of the copper base plate at a constant levelusing a foil heater. Petronis et al. [34] embedded twoelectrode heaters into the side walls of the microbioreactor chamber to create an even heat distribution across thechamber. Having an integrated heater in the microbioreactor setup is by far the most preferable method of temperaturecontrol in microbioreactors, as it is simple and cheap andallows for parallel operation at different temperatures, provided that some thermal insulation preventing thermalcrosstalk is in place. However, as the reactor chamber is not thermally insulated, the high S/V ratio makes a well-functioning temperature control loop necessary.

pH

Standard pH probes are too bulky and are not feasible for use in microbioreactors. The most commonly appliedminiature pH sensors are optical sensors based on fluores-cence sensor spots (Fig.2) or “ optodes” [7, 8, 10, 27, 35,44, 45] and solid-state, ion-sensitive, field-effect transistor (ISFET) pH sensor chips [11, 23, 36, 46]. Many microbioreactor operators seem to prefer the optical sensors dueto their noninvasive nature and ease of integration [7 – 10,27, 47, 48]. Since optodes do not require any referenceelement to perform pH measurements and are relativelycheap (EUR 15 per piece), they are good alternatives for disposable microbioreactors [7, 8, 10, 27]. Optical sensingis performed through either fluorescence intensity or fluorescence lifetime measurements [7, 8, 10, 27, 47].However, the fluorescence sensor spots suffer from the photobleaching effect, reducing their lifetimes, and have arather narrow dynamic pH range, typically 2 – 4 pH units[47]. Lifetime would not be an issue for single-use reactorsas the reactor packaging could be lightproof and thecultivation itself typically does not run long enough for this to be an issue. Optical pH sensors have a measurement accuracy of about 0.01 pH units with a response time (t 90 )of less than 90 s. Furthermore, the operating temperaturefor optodes ranges from 0 to 50 °C. Most pH sensor spotshave a measurement range from pH 4 to 9 and a nonlinear response [47]. This nonlinearity is apparent as a lowsensitivity of the sensor signal to pH changes at both endsof the measurement range, which makes pH control therevery difficult. In the middle of the measuring range, thesensor spots have an almost linear response, with asensitivity of around 10° phase angle per pH unit. Incontrast, ISFET pH sensors cover a wider pH range(typically from pH 2 to 12 [49]) and have a linear responsethat is similar to that of a standard pH probe (Nernstiansensitivity of about 58.2 mV per pH unit at 20 °C [49]).ISFET pH sensors have a measurement accuracy of about 0.01 pH units with response time (t 90 ) of less than a second.

They can also work across a wide temperature range(−45 °C to 120 °C). Though ISFET pH sensors have a broader dynamic measurement range and greater sensitivitythan pH sensor spots, the ISFET pH sensors also have somedrawbacks. This includes measurement drift (linear andreproducible in due course, thus allowing for data compen-sation) and sensitivity to surrounding light [11, 23, 49].Moreover, the ISFET pH sensor chip requires a referenceelectrode (i.e., Ag/AgCl reference electrode) to perform the pH measurement [11, 23, 36], meaning that the microbioreactors have to be designed such that the part with theintegrated ISFET pH chip is reusable in order to keep thecost per microbioreactor at an acceptable level. Practically,this is typically done by encapsulating the ISFET pH sensor chip on a printed circuit board before integrating the pHsensors into the microbioreactors [11, 23, 36]. Despite somelimitations, both sensors have been shown to provide rapidand precise pH measurements over a long period of time[7-11, 23, 27, 36, 45-48]. It can thus be concluded that real-time pH measurement in microbioreactors, similar to the pH monitoring feature of lab-scale bioreactors, is fullyfeasible.

While a number of reliable pH sensors are available for online monitoring of pH, pH control in microbioreactors isstill in the development phase. Early attempts to control pHin microbioreactors were accomplished by either a bufferedsystem [7, 8, 10, 27, 37] or by intermittently injecting baseor acid [9, 36, 48, 50]. The use of buffer to control pH inmicrobioreactors is the simplest way of controlling pH. Nevertheless, the use of a buffered system is not alwayssufficient to maintain a constant pH level: pH buffers havea limited buffering capacity and can only compensate for acertain number of ions before losing their resistance to pHchanges. This limitation was apparent in a stirred,membrane-aerated microbioreactor study where the pHdropped by nearly 2 pH units due to acidification duringthe fermentation process [10]. The limit on the buffer capacity can, however, be extended in a continuous culturemicrobioreactor system. Wu et al. [51] showed in a perfusion reactor system that continuous feeding of afreshly buffered nutrient medium enabled the reactor pHto be kept constant to within ±0.02. It is expected that asimilar effect would be visible in a continuous culturemicrobioreactor with cells in suspension, assuming that the buffer strength is adapted to the substrate concentration.

Intermittently injecting base or acid is another alternativeapproach to pH control in microbioreactors. It is a direct adaptation of a bench-scale bioreactor pH control method.Zhang et al. [48] and Lee et al. [9] showed that the decreasein the pH value due to the metabolic activity of the growingmicroorganisms can be compensated for by intermittentlyinjecting a base solution. Lee et al. [9] managed to reachhigher cell densities when the pH was maintained for a




longer period at the desired set point. However, this methodis limited by the microbioreactor chamber volume. Theaddition of an excessive volume of acid or base dilutes theculture broth, which leads to uncertainties in concentrationmeasurements [50]. A strong base (or acid) can be used toreduce the added volumes, but this may lead to local high(or low) pH values, particularly if mixing is poor. Such pHdeviations can locally stress the growing cells, leading tolower average growth rates for the cultivation as a whole. Asolution to this problem would be gaseous pH control, the practical feasibility of which has been proven by Isett et al.[44], and it has been implemented in an industrial system[52] in a 24-well 4 – 6 mL plate where NH3 /CO2 are spargedthrough the individual wells. Maharbiz et al. [23] developedan in situ electrolytic gas generation where CO2 gas can be precisely dosed from underneath the reactor chamber through a thin, semipermeable silicone membrane. Howev-er, Maharbiz et al. [23] are yet to test the effectiveness of CO2 gas generation for pH control with a real culture. Themethod applied seems promising, but the formation of bubbles due to gas diffusion is not desirable in microbioreactors. De Jong [11] demonstrated that the pH of aSaccharomyces cerevisiae fermentation can be controlled by dosing CO2 gas and NH3 vapor. However, this methodneeds further development, as the microbioreactor that wasfabricated had poor mixing and the response time withrespect to pH control was rather slow (the pH took severalhours to return back to its desired set point). Although bothMaharbiz et al. [23] and De Jong [11] are yet to optimizetheir methods, they have certainly underlined the potentialof introducing gases through a membrane to control pH inmicrobioreactors.

Cell concentration

In microbioreactor experiments, one typically uses theBeer – Lambert absorption law [53] to estimate the celldensity in the microbioreactors online via an opticalmeasurement. The absorbance measured is normally corre-lated to either the dry weight or the number of cells per volume. This enables the visualization of the cell growth inreal time. An online system is necessary because samplingis typically not possible in such small working volumes,and analytical techniques other than optical density (OD)measurements are difficult to apply. Online monitoring of the cell density in microbioreactors is normally realized viaoptical probes (Fig.2). Light from light-emitting diodes(LED) is guided into the microbioreactors with opticalfibers, sent through the reactor chamber, and then guided toa photodetector [7-10, 23, 27]. The wavelength for micro-organisms is normally chosen to lie within the range of visible light (400 – 700 nm); for example,S. cerevisiae hastraditionally been measured at 600 nm. Many state-of-the-

art microbioreactors implement this technique for ODmeasurements [7-10, 23, 27]. Most devices send in light from the bottom of the reactor and collect the transmittedlight at the top [7, 8, 10, 27], although one device works theother way round [9]. In either case, the resulting path lengthis in the range 0.3 – 2 mm. Optical fibers for ODmeasurement can also be positioned on the side of themicrobioreactor chamber, resulting in a horizontal opticalmeasurement path. This approach can provide a longer optical path length because the lateral dimensions of microbioreactors are normally larger than their heights.This can be useful for improving the measurement accuracywhen operating at low biomass concentrations.

The fixation of the optical fibers in the microbioreactor is important as it determines the accuracy and sensitivity of the OD measurement. Misalignment of the fibers will leadto a loss of transmitted light which will hinder themeasurement. The OD measurement depends on both the path length and the intensity of the incoming light, where both parameters can be adjusted individually. A shorter pathlength will allow for the measurement of higher-density broths, as will more intense illumination, and vice versa.Depending on the mixing method chosen, care must betaken to avoid collisions between, for example, the mixing bead(s) and the fibers. In the case of single-use reactors,care must be taken to ensure that the fibers can only beattached to/inside the reactor at one single defined position,as variations here will change the measured absolute valuesof transmitted light, which in turn will make calibrationnecessary for every single reactor.

Though OD measurement via optical probes has provento be the most feasible way of estimating cell density inmicrobioreactors, gas bubbles in the reactor chamber caninterfere with OD measurements. OD measurement viaoptical probes is also sensitive to ambient light, but this can be circumvented by utilizing modulated light or by filteringthe OD signal using a lock-in amplifier. It is also important to mention that the OD measurement measures the total cellconcentration, since it cannot distinguish viable cells fromdead cells.

Near-infrared technology (NIR) also allows for theonline optical measurement of cell concentration [54]. Asa complete NIR setup is expensive and requires elaboratesignal processing, industrial NIR turbidity measurement systems are now available [55] at a reasonable price,although they are still significantly more expensive than asimple transmission measurement system.

Alternatively, cell density can also be estimated inmicrobioreactors by means of impedance spectroscopy, as proposed by Krommenhoek et al. [36, 49]. This method,which is commonly applied to larger-scale reactors, appliesan alternating current (AC) electrical field to the culture andmeasures the cell conductivity as a function of frequency.




As only cells with intact membranes are able to polarizecharge and hence contribute significantly to the conductiv-ity, this method only measures live cells. Krommenhoek et al. [36, 49] integrated this impedance sensor onto amultisensor chip and placed it underneath the microbior-eactor chamber to measure biomass concentration. Here,the lowest biomass concentration that can be measured is1 g L−1 , which corresponds to a relative conductivitychange of about 0.1%.

Dissolved oxygen concentration

Oxygen is one of the substrates in an aerobic fermentation.Thus, it is crucial that this variable is controlled properly to prevent oxygen-limiting conditions during the fermentation process [2]. In industrial-scale fermentations, the dissolvedoxygen concentration is normally controlled, often evenwith dynamic set-point profiles during, for example, fed- batch operation. Providing a sufficient oxygen supply inmicrobioreactors is a challenge due to the low solubility of oxygen in water, which is only worsened by laminar flowand difficult mixing circumstances. To supply sufficient oxygen in a bubble-free system, a high interfacial area isrequired in order to improve the interfacial mass transfer area to volume ratios of the microbioreactors [9, 10, 56].Luckily, the high S/V ratios of microsystems can provide alarge interfacial area. Still, efficient transport of oxygen-richfermentation broth away from the membrane is alsoimportant, and can be achieved by providing sufficient mixing (see above). Another method of achieving sufficient oxygen supply is to increase the O2 content in the gas phase, as this will significantly increase the saturationconcentration of oxygen in the fermentation broth.

The dissolved oxygen concentration in microbioreactorsis typically measured by optical sensors (Fig.2) with theuse of fluorescence sensor spots [7, 8, 10, 27, 43, 48].Optodes for dissolved oxygen are based on the quenchingof fluorescence by oxygen [57]. Even though they can workfor the entire range of saturation values, from 0 to 100%,their sensitivity is optimal at low concentrations [57], whichis the relevant range for fermentations. Just like the optodesfor pH measurement, they can be manufactured in smallsizes, are easy to integrate, insensitive to ambient light,relatively cheap, and nonreactive. Using these oxygenoptodes, microbioreactors can also be made disposable[7, 9, 10, 23, 27].

The dissolved oxygen concentrations in microbioreactorscan also be measured by an electrochemical sensor such asthe ultra-microelectrode array (UMEA). The UMEA is anamperometric biosensor and measures oxygen concentra-tion based on the electrochemical reduction of oxygen [36].Its working principle is similar to that of the Clark oxygenelectrode [2]. Such a sensor has been used in a bench-scale

reactor cultivation of Candida utilis [36]. The sensor delivered results that were very similar to those of thereference sensor. It is small enough to fit into a 96-well plate, which clearly makes it a candidate for implementa-tion in a microbioreactor system.

For microbioreactors with a low working volume (e.g.,50 – 300 μ L), where bubbles are not desirable and bubblesparging is therefore not feasible, oxygen supply is oftenachieved by membrane aeration. Several strategies have been developed for the oxygenation of microbioreactors.Maharbiz et al. [23] developed in situ electrolytic gasgeneration to supply of oxygen into a microwell-basedmicrobioreactor. Oxygen was allowed to diffuse into thereactor chamber through an orifice that was laminated witha thin gas-permeable membrane. With this method, oxygentransfer rates as high as 40 mmol O2 L−1 h−1 can beachieved. For comparison, values of up to 200 mmolO2 L−1 h−1 have been reported for bench-scale reactors[55]. However, since the system of Maharbiz et al. wasaerated from the bottom of the reactor chamber, the aerationmembrane bulged upward due to positive pressure from theelectrolysis chamber. Furthermore, there was immense bubble formation in the system because the magnetic beadsused for mixing were not strong enough to break the bubbles forming on the aeration membrane surface.Implementing a mixing system that more efficiently scrapesthe bubbles from the membrane surface may in this casealleviate the formation of large bubbles.

Alternatively, aeration in microbioreactors can also beachieved by integrating a gas-permeable membrane into thetop side of the microbioreactor chamber [7-10, 27, 48]. Inthis method, the gas stream is typically pushed into a gaschamber which is separated from the reactor contents by athin PDMS membrane that allows the diffusion of smallmolecules from the gas phase (e.g., oxygen) into the liquid phase. Since diffusion is merely based on surface aeration,no bubbles were observed in the reactor chamber. Thestandard measure of the oxygen transfer efficiency is thevolumetric oxygen transfer coefficient k L a (h−1 ), which iscomposed of the mass transfer coefficient k L (m/h) and theinterfacial areaa (m2 /m3 ). Islam et al. [58] compared theoxygen mass transfer coefficients in five different volumesranging from microwell plates to 75 L stirred-tank reactors,and then performed cultivations at matchingk L a values. For high k L a values (247 h−1 ), they report very similar fermentation kinetic profiles and yields across scales, andactually identifyk L a as being one of the most important factors during scale-up. Zanzotto et al. [10] developed amicrobioreactor with a shallow reactor chamber (300μ m)to shorten the diffusion distance (distance from the gas-liquid interface to the bottom of the reactor chamber). Intheir design, a 100μ m thick PDMS membrane was bondedonto the top of the reactor chamber for oxygenation. The




k L a obtained in the Zanzotto et al. [10] microbioreactor is,however, rather low (k L a ~72 hr −1 ), as a comparison withlarger-scale reactors shows [1]. This is because no activemixing scheme was realized and oxygen transfer was basedmerely on the diffusion of air from the atmosphere. On theother hand, Lee et al. [9] demonstrated that, under perfect mixing conditions and tight oxygen control, ak L a value ashigh as 500 h−1 could be achieved in their microbioreactors.They implemented a proportional – integral (PI) controlalgorithm to control the dissolved oxygen concentration inthe microbioreactors. They also showed that oxygenation at various oxygen concentrations or with compressed air coupled to integrated peristaltic mixing tubes can accuratelymaintain the dissolved concentration at a set point that is upto about 30 to 40% of the saturation value.

CO2

In bench-scale reactors, the CO2 content in the off-gas isoften measured. However, the amounts of off-gas producedin a microsystem are very small, which makes another typeof sensor necessary. Ge et al. [59] and Liebsch et al [60]have implemented an optical and thus noninvasive CO2

sensor based on a fluorescent dye applied onto a siliconemembrane. Similar sensors spots are also available commer-cially (PreSens, Germany); these use sensing equipment similar to the abovementioned dissolved oxygen and pH spots.

Spectroscopy

Spectroscopy may provide the means to estimate theconcentrations of various analytes, such as glucose, acetate,and lactate, with online optical methods. These systemsmeasure a whole wavelength spectrum for one singlemeasurement. Appropriate signal processing and the eval-uation of the data with chemometric models then allow theanalyte concentrations to be estimated. These basic charac-teristics make these measurement systems interesting but they also make their application more complicated. Eachmeasurement system, such as near-infrared (NIR) [54, 61]or Raman spectroscopy [62-64], has its own advantages anddisadvantages that must be evaluated with the specificapplication in mind. Raman spectroscopy, for example, isadvantageous due to the low interference from water andthe flexibility in choice of wavelengths [62], but it is moreexpensive and more laborious than NIR, which is a morewell-developed technique for fermentation [54, 61].

Discussion

The construction and operation of well-controlled microbioreactors have been demonstrated, and there is a great

potential for their application within process and straindevelopment labs in the fermentation industry. With theincreasing drive towards a bio-based economy where bioresources are replacing hydrocarbons as feedstocks,harnessing microorganisms for chemical and energy pro-duction becomes more important. In the US, the market share of bio-based chemicals is expected to grow from 5%in 2002 to 12% in 2010, and this growth is expected tocontinue at a significant pace [65]. Furthermore, a recent vision paper published by the Industrial Biotechnologysection of the European Technology Platform for Sustain-able Chemistry [66] foresees that up to 30% of the rawmaterials used by the chemical industry will originate fromrenewable sources by 2025.

In every industrial biotechnology process, strain improvement programs for increased yields and volumetric productivities are crucial to economical viability. High-throughput screening is needed to identify interesting,improved production strain candidates among a largenumber of mutants. Microbioreactors provide an interestingalternative to the shake flasks or microtiter plates common-ly used for screening today. They allow for the continuousmeasurement and control of various fermentation parame-ters, such as pH or dissolved oxygen, which are not that straightforward in shake flasks or microtiter plates. Thus,establishing suitable control loops on the microbioreactor platform implies that the conditions in the microbioreactor will be much more similar to those in larger-scaleoperations than they presently are for experiments instandard shake flasks or microtiter plates. Another area of application for microbioreactors could be investigations of process and media optimization, where a range of con-ditions can be tested in less time and at a lower cost inmicrobioreactors than in shake flasks or bench-scalereactors.

In order to be able to apply them to the abovementionedapplications, microbioreactors must be disposable, cheap,allow for tight control over the fermentation, and giveresults that are upscalable. If all of these conditions can befulfilled, microbioreactors will undoubtedly be used asstandard experimental platforms for fermentation researchin the future. As mentioned before, cost reductions should be obtained on the one hand through the considerablereduction in size and on the other through the use of polymer materials and established polymer processingmethods to construct disposable reactors.

As shown in this review (Table1), many enablingtechnologies for this vision of a single-use reactor alreadyexist. The most important culture variables — temperature, pH, cell density, and dissolved oxygen content — can bereadily measured at the micro scale using methods that are both noninvasive and disposable. One future area of research should be to extend the measurement ranges of




sensor systems that already exist; pH is the most obviousexample. Most cultivations of yeast and filamentous fungiare currently run at pH 4 – 6, which corresponds to the rangewhere pH optodes are rather insensitive. pH sensors that can cover the full range needed for cultivations at a cost similar to that of pH optodes would greatly increase their applicability and usage.

It should be emphasized once more that control over thefermentation conditions is of crucial importance to thefuture acceptance of microbioreactors as a useful andversatile experimental platform. Methods for controllingtemperature, pH, and dissolved oxygen in microbioreactors[9, 23, 34, 36, 44, 48, 50] do exist, and it has also beendemonstrated that several control loops can be combined ina single microbioreactor [7], such that the experimentalwork in microbioreactors can be performed under cultureconditions that are fully comparable to cultivation con-ditions in bioreactors that operate at industrially relevant scales. However, in comparison to bench-scale reactors, thelimited ability to manipulate essential reactor variableswithin the small working volume typical of microbioreac-tors is an engineering challenge, and therefore continuedresearch on the implementation of simple but well-functioning control loops in microbioreactors is a must.At this point, the capabilities of microbioreactor technologycould be extended further by including substrate feedingcontrol strategies [67], such that, for example, advancedfeeding strategies typical of fed-batch operation [68, 69] —

the most widespread mode of operation for industrialfermentations — could be investigated at the micro scale. It should also be kept in mind that the integration of all thesecontrol loops into a limited reactor volume will be acontinuous challenge, and typically a tradeoff will have to be made between the ability to control many variables inthe microbioreactor and the need to keep the cost per reactor at an acceptable level.

Provided the reactor design is in place, the next important step is to prove upscalability. Only if the resultsof, for example, a ranking experiment in the microbior-eactor can directly be transferred to bench-scale or evenfull-scale reactors will the system be a valid candidate for industrial applications. A number of articles already report that results obtained in microbioreactors compare well toresults obtained at larger scales [10, 70]. Future researchshould maintain its focus on this topic, since the acceptanceof microbioreactor technology as a useful and versatileexperimental platform will need the support of successfulexamples where scalability of results has been demonstrat-ed convincingly. Also interesting from a scalability per-spective is the incorporation of additional sensors that provide more information on the metabolism or microor-ganism performance than achieved with standard sensors.As mentioned above, spectroscopic methods provide an

appealing option since biomass, substrate, and product concentrations may be determined. Interestingly, for aerobicfermentations, the implementation of spectroscopic meth-ods may be easier in microbioreactors with bubblelessaeration than in bench-scale reactors, where sparging of gas bubbles disturbs the measurements. Other interestingsensors are lab-on-a-chip devices for biosensing different biological molecules such as DNA, proteins, or smallmolecules [71]. Assuming that a few relevant biomarkersthat are correlated to the stress level or production level of the cells can be identified for a certain process, theinclusion of a chip analyzing these biomarkers in themicrobioreactor design would provide detailed insight intothe performance of the strain. This would open up the possibility of gaining in-depth knowledge about the processstrain behavior in microbioreactors, and would thus providevaluable information for scaling up.

Clearly, parallelization is also an important issue. Inorder to be able to run high-throughput experiments (e.g.,comparing different culture conditions or strains), paralle-lization is a necessity. This not only requires the potentialfor parallel sensing, but ideally it also demands aninfrastructure that enables the reactors to be run under different conditions (e.g., pH or temperature). Whilst someresearchers and companies [7, 23, 44, 52, 64] have proventhe basic feasibility of multiplexed reactors, current systemsdo not actively support full control over different operatingconditions. For example, temperature may be the same for all reactors [7], or either pH or DO can be controlled, but not both together [52]. Additionally, current systemsnormally do not allow for continuous operation.

One major criterion here is the separation of “ fixedmachinery” from “ single-use reactor.” Or in other words,questions of major importance in microbioreactor designinclude: how many sensors, actuators, etc., should beincluded in the reactor itself? What should be included inthe supporting machinery? How should the interfaces beestablished? The answers to these questions greatly impact on the reactor design, as well as on the microbioreactor operating cost per fermentation experiment in particular.Current systems usually solve the sensing question byincluding fluorescent sensor spots in every single reactor well, and then use one single measurement/evaluation unit for every measurement type (e.g., DO, pH, etc.) [7, 52].This measurement/evaluation unit is then sequentiallyapplied to the individual wells (see Fig.8 for a practicalexample). This of course calls for some machinery and/or electronics that can precisely position the sensor onto thefiber or switch from well to well in some other manner.However, with modern robotics this is a minor obstacle.

Controlling pH and DO for every individual well callsfor either an elaborate system of tubes and valves, or againfor a suitable positioning mechanism — in this case for a




dosing system — that also ensures adequate tightness.Individual temperature control of every well obviouslyrequires a heating system in every reactor well. Possibly ageneral minimum temperature could be adjusted atmospher-ically such that the individual wells only have to be heatedto their individual destined temperatures. This necessitatesan incubator-type enclosure, but makes the temperatureeasier to control due to the lower thermal losses.

In order to make the whole microbioreactor system ascheap as possible, the system should be flexible; in other words, it should only seek to satisfy the requirements of each individual application (e.g., to only provide pH controlif it is needed), as stripping out unnecessary featuressignificantly reduces the complexity of the system.

Whilst one focus must be the complexity and thus cost of the system, the other should be the ruggedness and ease-of-use required in industrial applications. As current systems [52] prove, it is possible to reduce the complexityenough to allow the fabrication of a convenient system.Further work here may be directed at downsizing and providing more individual control over every reactor.

Last but not least, the importance of software aspectsshould not be underestimated. If a microbioreactor platformis employed to acquire real-time experimental data viacheap and high-throughput experimentation under well-controlled conditions, an immense amount of data will begenerated. The trend here is for the amount of online datagenerated for each cultivation to increase in the future; for example, when online spectroscopic methods are imple-

mented on the microbioreactor platform too. Handling suchdata does not pose any problem in terms of computing power or data storage capacity. It should be emphasizedthough that manual processing of these data is simply not an option. Instead, microbioreactor platforms should besupported by advanced software for online visualizationand interpretation of the online data. Perhaps most importantly, the software will need to allow for the efficient storage, retrieval, and documentation of experimental dataso that the user can work efficiently with a microbioreactor setup. This documentation functionality is also especiallyrelevant to process development in the pharmaceuticalsector. In addition, existing algorithms relying on multivar-iate statistical methods should, for example, be imple-mented as standard to keep track of the state of thefermentation experiment [72]; ideally, abnormal situationsshould be detected and logged so that this type of information can be presented to the experimentalist in acondensed form for each experiment. Spectroscopic data —

which will also be standard microbioreactor measurementssooner or later — will need to be converted to usefulestimates of the analytes of interest in an online fashion,meaning that chemometric methods will have to beimplemented in the software as well. Last but not least,online data can be further enhanced by making use of software sensors to estimate important cultivation parame-ters that are otherwise difficult to measure online. For example, a software sensor application was recently devel-oped to detect physiological stress in the growing cells [73].

Fig. 8 Multiplexed microbioreactor setup with four stirred microbioreactors, indicating sequential measurements of OD, DO, and pH [7]




Conclusions

A review of existing microbioreactor technology confirmsthat microbioreactors can be applied for fermentationexperiments under well-controlled conditions of pH, temperature, and dissolved oxygen concentration. The use of polymer materials to construct microbioreactors shouldresult in cheap and disposable (single-use) reactors that are ideally suited to fermentation screening experiments.For microbioreactors to become generally accepted in thefermentation industry, future research should focus on: (1)the further development of cheap and disposable onlinesensors, including extending the usable range of existingsensors such as pH optodes and extending the number of variables that can be measured online; (2) the implemen-tation and demonstration of controlled substrate feedingstrategies in microbioreactors, similar to fed-batch opera-tion at the industrial scale; (3) scalability, in other words,the ability to transfer experimental results obtained inmicrobioreactors to pilot or production scale for a broadrange of microorganisms; 4) parallelization, or morespecifically the development of smart technical solutionsto perform a series of experiments under different con-ditions in a range of parallel microbioreactors; 5) softwareto support the documentation of experiments and to handlethe storage, online interpretation, and retrieval of themassive amounts of experimental data generated by parallelcultivations.

Acknowledgement The Ph.D. project of Daniel Schäpper isfinanced by the Novozymes Bioprocess Academy.

References

1. Weuster-Botz D, Hekmat D, Puskeiler R, Franco-Lara E (2007)Adv Biochem Eng Biotechnol 105(205):247

2. Shuler ML, Kargi F (2002) Bioprocess engineering: basicconcepts. Prentice Hall, Upper Saddle River

3. Lye GJ, Ayazi-Shamlou P, Baganz F, Dalby PA, Woodley JM(2003) Trends Biotechnol 21(1):29 – 37

4. Kumar S, Wittmann C, Heinzle E (2004) Biotechnol Lett 26:1 – 105. Betts JI, Baganz F (2006) Microb Cell Fact 5:216. Micheletti M, Lye GJ (2006) Curr Opin Biotechnol 17:611 – 618

7. Szita N, Boccazzi P, Zhang Z, Boyle P, Sinskey AJ, Jensen KF(2005) Lab Chip 5:819 – 8268. Zhang Z, Boccazzi P, Choi HG, Perozziello G, Sinskey AJ, Jensen

KF (2006) Lab Chip 6:906 – 9139. Lee HLT, Boccazzi P, Ram RJ, Sinskey AJ (2006) Lab Chip

6:1229 – 123510. Zanzotto A, Szita N, Boccazzi P, Lessard P, Sinskey AJ, Jensen

KF (2004) Biotechnol Bioeng 87(2):243 – 25411. De Jong J (2008) Application of membrane technology in

microfluidic devices (Ph.D. thesis). University of Twente, Twente12. Zhao YG, Lu WK, Kim SS, Ho ST, Marks TJ (2000) Appl Phys

Lett 77(19):2961 – 296313. Fleger M, Neyer A (2006) Microelectron Eng 83:1291 – 1293

14. Becker H, Gärtner C (2008) Anal Bioanal Chem 390:89 – 11115. Tsao CW, Devoe DL (2009) Microfluid Nanofluid 6:1 – 1616. Mcdonald JC, Whitesides GM (2002) Acc Chem Res 35(7):491 –

49917. Huang C-W, Lee G-B (2007) J Micromech Microeng 17:1266 –

127418. Anvari M, Khayati G (2009) J Ind Microbiol Biotech 36:313 – 31719. Zautsen RRM, Maugeri-Filho F, Vaz-Rossell CE, Straathof AJJ,

van der Wielen LAM, de Bont JAM (2009) Biotechnol Bioeng102:1354 – 1360

20. Purcell EM (1977) Am J Phys 45(1):3 – 1121. Hessel V, Löwe H, Schönfeld F (2005) Chem Eng Sci 60:2479 –

250122. Hardt S, Schönfeld F (2007) Microfluidic technologies for

miniaturized analysis systems. Springer, New York23. Maharbiz MM, Holtz WJ, Howe RT, Keasling JD (2004)

Biotechnol Bioeng 85(4):376 – 38124. Strook AD, Dertinger SKW, Ajdari A, Mezić I, Stone HA,

Whitesides GM (2002) Science 295:647 – 65125. Li X, van der Steen G, van Dedem GWK, van der Wielen LAM,

van Leeuwen M, van Gulik WM, Heijnen JJ, Krommenhoek EE,Gardeniers JGE, van den Berg A, Ottens M (2008) Chem Eng Sci63:3036 – 3046

26. Berthier E, Warrick J, Hongmeiy Y, Beebe DJ (2008) Lab Chip8:852 – 85927. Boccazzi P, Zhang Z, Kurosawa K, Szita N, Bhattacharya S,

Jensen KF, Sinskey AJ (2006) Biotechnol Prog 22:710 – 71728. Olsson A, Stemme G, Stemme E (1996) Sens Actuators A

137:14329. Jeong OC, Park SW, Yang SS, Pak JJ (2005) Sens Actuators A

123:45330. Fredrickson CK, Fan ZH (2004) Lab Chip 4:526 – 53331. Perozziello G (2006) Doctoral thesis. Technical University of

Denmark, Lyngby32. Kortmann H, Blank LM, Schmid A (2009) Lab Chip 9:1455 – 146033. Vojinović V, Cabral JMS, Fonseca LP (2005) Sens Actuators B

114:1083 – 109134. Petronis S, Stangegaard M, Christensen CBV, Dufva M (2006)

Biotechniques 40:368 –

37635. Maiti TK (2006) IEEE Sens J 6(6):1454 – 145836. Krommenhoek EE, van Leeuwen M, Gardeniers H, van Gulik

WM, van den Berg A, Li X, Ottens M, van der Wielen LAM,Heijnen JJ (2008) Biotechnol Bioeng 99(4):884 – 892

37. Vervliet-Scheebaum M, Ritzenthaler R, Normann J, Wagner E(2008) Ecotoxicol Environ Saf 69:254 – 262

38. Gimbun J, Radiah ABD, Chuah TG (2004) Food Eng 64:277 – 28339. Assael MJ, Antoniadis KD, Wu J (2008) Int J Thermophys

29:1257 – 126640. Shin YS, Cho K, Lim SH, Chung S, Park SJ, Chung C, Han DC,

Chang JK (2003) J Micromech Microeng 13:768 – 77441. Liu L, Peng S, Niu X, Wen W (2006) Appl Phys Lett 89:22352142. Yamamoto T, Nojima T, Fujii T (2002) Lab Chip 2:197 – 20243. van Leeuwen M, Heijnen JJ, Gardeniers H, Oudshoorn A,

Noorman H, Visser J, van der Wielen LAM, van Gulik WM(2009) Chem Eng Sci 64:455 – 45844. Isett K, George H, Herber W, Amanullah A (2007) Biotechnol

Bioeng 98:1017 – 102845. PreSens (2009) Noninvasive pH probes.http://www.presens.de/

products/brochures/category/sensor-probes/brochure/non-invasive- ph-sensors.html. Accessed 9 March 2009

46. Microsens (2009) Ion-sensitive field-effect transistor.http://www.microsens.ch/products/chemical.htm. Accessed 9 March 2009

47. John GT, Goelling D, Klimant I, Schneider H, Heinzle E (2003) JDairy Res 70:327 – 333

48. Zhang Z, Perozziello G, Boccazzi P, Sinskey AJ, Geschke O,Jensen KF (2007) J Assoc Lab Automat 12(3):143 – 151


http://www.presens.de/products/brochures/category/sensor-probes/brochure/non-invasive-ph-sensors.html



http://www.microsens.ch/products/chemical.htm









49. Krommenhoek EE, Gardeniers JGE, Bomer JG, Li X, Ottens M,van Dedem GWK, van Leeuwen M, van Gulik WM, van der Wielen LAM, Heijnen JJ, van den Berg A (2007) Anal Chem 79(15):5567 – 5573

50. Buchenauer A, Hofmann MC, Funke M, Büchs J, Mokwa W,Schnakenberg U (2009) Biosens Bioelecton 24:1411 – 1416

51. Wu MH, Huang SB, Cui Z, Cui Z, Lee GB (2008) BiomedMicrodevices 10(2):309 – 319

52. Applikon Biotechnology (2009) Micro 24 bioreactor.http://www.applikon-bio.com/index.php?option=com_content&view=article&id=80&catid=43&Itemid=17. Accessed 9 March 2009

53. Jensen KH, Alam MN, Scherer B, Lambrecht A, Mortensen NA(2008) Opt Commun 281:5335 – 5339

54. Cervera AE, Petersen N, Eliasson Lantz A, Larsen A, GernaeyKV (2009) Biotechnol Progr (in press). doi:10.1021/bp.280

55. Sartorius BBI Systems (2009) Sartorius FUNDALUX II turbiditysensor. http://www.sartorius-stedim.com/index.php?id=1999.Accessed 9 March 2009

56. Villain L, Meyer L, Kroll S, Beutel S, Scheper T (2008)Biotechnol Prog 24:367 – 371

57. Papkovsky DB (1995) Sens Actuators B 29:213 – 21858. Islam RS, Tisi D, Levy MS, Lye GJ (2007) Biotechnol Bioeng 99

(5):1128 – 113959. Ge X, Kostov Y, Rao G (2004) Biotechnol Bioeng 89(3):329 – 33460. Liebsch G, Klimant I, Frank B, Holst G, Wolfbeis OS (2000) Appl

Spectrosc 54(4):548 – 55961. Scarff M, Arnold SA, Harvey LM, McNeil B (2006) Crit Rev

Biotechnol 26:17 – 39

62. Lee HLT, Boccazzi P, Gorret N, Ram RJ, Sinskey AJ (2004) VibSpectrosc 35:131 – 137

63. Cao E, Firth S, McMillan PF, Gavriilidis A (2007) Catal Today126:119 – 126

64. Ferstl W, Klahn T, Schweikert W, Billeb G, Schwarzer M,Loebbecke S (2007) Chem Eng Technol 30(3):370 – 378

65. Biomass R&D, Technical Advisory Committee (2006) Vision for bioenergy and biobased products in the United States.http://www. brdisolutions.com/default.aspxAccessed 9 March 2009

66. European Technology Platform for Sustainable Chemistry, Indus-trial Biotechnology Section (2005) Homepage.http://www.suschem.org. Accessed 9 March 2009

67. Riesenberg D, Guthke R (1999) Appl Microbiol Biotechnol51:422 – 430

68. Åkesson M, Hagander P, Axelsson JP (1999) Biotechnol Tech13:523 – 528

69. de Maré L, Cimander C, Elfwing A, Hagander P (2007)Bioprocess Biosyst Eng 30:13 – 25

70. Sartorius BBI Systems (2009) Biostat Cultibag.http://www.sartorius-stedim.com/index.php?id=4387. Accessed 7 March 2009

71. Sathuluri RR, Yamamura S, Tamiya E (2008) Adv Biochem EngBiotechnol 109:285 – 350

72. Lennox B, Montague GA, Hiden HG, Kornfeld G, Goulding PR (2001) Biotechnol Bioeng 74:125 – 135

73. Sundström H, Enfors S-O (2008) Bioprocess Biosyst Eng 31:145 –

15274. Zhang Z, Szita N, Boccazzi P, Sinskey AJ, Jensen KF (2005)

Biotechnol Bioeng 92(2):286 – 296


http://www.applikon-bio.com/index.php?option=com_content&view=article&id=80&catid=43&Itemid=17



http://dx.doi.org/10.1021/bp.280

http://www.sartorius-stedim.com/index.php?id=1999

http://www.brdisolutions.com/default.aspx


http://www.suschem.org/











http://dx.doi.org/10.1021/bp.280




Documents

Schaepper 2009 Online 2ndAuthor