High Throughput Bioprocessing Systems

By Joe Qualitz, PhD

Conceptually, a high throughput bioprocessing system is any single system that enables more than one culturing or fermentation process to be carried out in parallel. Ideally, such a system provides a sufficient degree of process information, but does so with less cost and less labour for each experiment compared to conventional laboratory methods (flasks, bioreactors, etc.).
In the last decade, increased efforts have been applied toward the development and utilization of high throughput systems. Prior to that time, most fermentation and cell culturing processes, while at the heart of bioprocess development, utilized technology little different from what was first used in the development of penicillin over a half-century ago.

The Need

In recent years, a number of compelling factors have reinforced a growing need for high throughput systems:

The completion of the human genome sequence — while a remarkable achievement in itself — is only the tip of the iceberg in terms of the technological efforts needed to fully realize the benefits promised by this research. Still needed is characterization of the identified genes — that is, the determination of what turns them on and what proteins they express. This effort requires cloning and expressing the targeted genes in a suitable host and then evaluating them in a large number of different environmental conditions. This is a daunting task — given the many tens of thousands of identified genes — without some means of testing the various sets of conditions in a rapid, systematic and largely automated manner.
l     Much angst, finger pointing and black market profiteering, all stemming from the country’s inability to provide adequate supplies of flu virus vaccine this past year, dramatically reinforced the recognition that reduced research and development times for such products are no longer a nicety but a necessity.
l     The ever-growing threat of biological weapons being used by a terrorist group, here or elsewhere, mandates increased development of ways to respond more rapidly and effectively to such an eventuality. Clearly, any methods and devices that can dramatically reduce the development cycles of responsive drugs and vaccines would be most welcome to researchers.

Even without these recent developments, there would still be substantial motivation to design and develop high throughput systems. Bioprocess research and development laboratories at major pharmaceutical and biotechnology companies typically conduct more than 100,000 glassware experiments and several thousand bioreactor experiments per year for each drug developed. While glassware systems, particularly shake flasks, are ubiquitous in such work — because of their relatively low cost, their high degree of parallelism, and their comparatively low labour requirements — these systems contain little instrumentation and thus provide minimal tracking and control of process parameters. Even an important process parameter such as pH is commonly measured off-line in these systems, and the process of doing so is both labour-intensive and potentially inaccurate, as pH changes during the sampling process.
On the other hand, small lab-scale bioreactors (1 to 3 l) are adequately instrumented, but are both expensive and labour-intensive. Hence, the number of experiments performed in such devices must be, by necessity, limited. While mathematical modelling and analysis of the bioprocess has been successfully used to decrease the number of experiments needed for process optimization, a huge number of experiments is still required to establish the needed process parameters.
A typical drug-development cycle is shown in Figure 1. As shown, the total time to bring a product to market is extremely long, and the cost is enormous. Moreover, the development phase of a new cell-based drug is performed in a pressure-cooker environment. During this period, the R&D team must decide what constitutes optimum culture conditions: even small errors can result in enormous increases in production costs over time. Also, unless allowable variances in cell culturing parameters can be fully explored and documented, there remains the real possibility that batches of product may eventually end up being needlessly discarded because of small excursions from culturing conditions that are, in reality, of no consequence.

Criteria for High Throughput Design
The compelling criterion driving the design of an effective high throughput system is that it yield a substantially lower cost and elapsed time per experiment. It is important to note that the costs of an experiment must include all of its components: equipment, facilities, media and other consumable resources, and, perhaps most importantly, labour. (One might also argue, justifiably so, that time-to-market should be included as a separate cost component, but that parameter is difficult to quantify and is at least partially included in facilities costs.)
There is also a second criterion: the data generated through the use of the system must be relevant to the ultimate goals of the research or development process. Stated another way, a system that does not provide a path to scale-up may be an interesting research tool, but perhaps not be very useful in the development process.
Approaches to High
Throughput Design
There are several approaches to high throughput design. Plate readers and similar devices allow the opportunity to study a large number of bioprocesses simultaneously and require minimal media costs and labour, instead utilizing robotics or other automation techniques to carry out their tasks. Until recently, however, such systems could measure only one or possibly two process parameters, and provide virtually no means of bioprocess control. Moreover, the volumes of the individual cultures preclude repeated sampling and off-line testing. Newer systems promise more capabilities, but even with these systems (which tend to be quite expensive compared to conventional culturing equipment) the question remains of how best to scale them up to the large-capacity systems required for production (generally in the form of stirred tank bioreactors, although bag reactors are also coming into use).
A second approach — typified by systems offered by Dasgip AG (Juelich, Germany), Infors AG (Bottmingen, Switzerland), Sartorius BBI Systems GmbH (Melsungen, Germany), etc. — uses the simple concept of ganged stirred tank reactors. These systems incorporate a number of vessels ranging from three or four to a dozen or more, in sizes ranging from a fraction of a litre to several litres. While their functionality is the same as that which would be obtained by buying and simultaneously using an equivalent number of individual lab-scale reactors, these packaged systems have the advantage of using less bench space than the individual reactors would. Also, the packaged vessels share lab utilities, as well as a common control system. Perhaps most importantly, these systems provide a high throughput capability that uses an almost universally accepted technology, and as a result, the data generated by these systems can be readily used in the scale-up process. Finally, their volumes are sufficient to allow repeated sampling without disruption of the process.
The drawbacks to these systems, however, stem from the fact that they are using essentially decades-old technology, albeit repackaged advantageously. While bench space is conserved, parallel processing is facilitated, and scale-up procedures are clear, system labour requirements are quite high: setup, calibration, and tear-down/cleaning times are virtually the same as they would be for multiple experiments in an equivalent number of individual reactors. The primary reason for this is that the systems use conventional probe-type sensors: each vessel has its own set which must be individually calibrated, a very time-consuming procedure that still does nothing to mitigate reactor-to-reactor variances in readings due to differences caused by inherent probe inaccuracy and drift. Also, the need to physically allow for these probes in the vessels limits the smallest practical vessel size, even if (often expensive) miniature probes are used. Thus, volumes of required media (and consequently, supply costs) differ little from that required by the multiple use of small lab reactors.

New Opportunities
Advances in sensor technology over the past dozen years have suggested another approach to high throughput design. Innovative, fluorescence-based sensors that utilize one or more high intensity LEDs to illuminate “peel-and-stick” sensing patches or foils (with the resultant response captured and analysed) have been utilized successfully in a variety of culturing apparatus, including flasks, bioreactors, bag reactors etc., for measurement of DO, pH, pCO2 and other parameters.
As an example, a system for measuring DO can be described as follows: a modulated, properly filtered light source (blue LED) can be used to illuminate a selectively treated patch immersed in a fluid. Because of the chemistry used in the preparation of the patch, the light it reflects will differ from the incident light in wavelength, phase and intensity, with the latter parameters dependent on the DO level in the immersion fluid. The reflected light, in the red spectrum, can then be captured using a common and inexpensive photodetector, again using appropriate filtering. Either intensity or phase can then be used as the measured parameter, with phase shift (electronically more difficult to measure, but leading to a more robust system that’s substantially less affected by background noise, differences in ambient light, etc.) being preferred. Measuring instruments for pH, pCO2, and other parameters, while utilizing different chemistries and somewhat different measurement techniques, can also be offered using physically similar designs and apparatus.
Because there are no probes entering the culture vessel with these devices, even a stirred-tank vessel can (and has, in practice) been reduced in size to volumes of 10 to 30 millilitres, with further reductions possible. This is because the illuminated foil area — the system component that must be resident in the process vessel — need not be more than a few millimetres in diameter. In fact, in can be far smaller if fibre optics are used as light source and receiver. Of course, the use of fibre optics has disadvantages in terms of lower signal-to-noise ratios, possible signal attenuation/variance due to bending in fibre optic paths and so forth, but they enable the use of this technology in systems with working volumes measured in a few millilitres or less.
In addition, standard calibration procedures are rendered obsolete. The sensors are electronically stable and subject to very little drift with age. An initial instrument calibration, which in itself takes less than a minute, need only be repeated every few months.
Consequently, use of these fluorescence-based systems provides the opportunity to simultaneously reduce media costs (small volume), bench space (small footprint), and labour (calibration of the sensors can be done in a matter of minutes, rather than hours). In addition, since the sensors are non-invasive, they can be readily multiplexed with simple automation techniques so that one set of sensors can be used for multiple reactors, ensuring consistency of measured readings from reactor to reactor.
Utilizing readily available means to move the individual vessels over fixed sensors, the only limitation on the number of reactors that can be serviced by one set of probes is the sampling time of each parameter (currently several seconds per measurement). All reactors of a 12-reactor system (Fig. 2) can be measured for pH and DO in less than three minutes — and potentially in far less time with refinement of the sampling techniques. Also, if shorter sampling systems are required for certain applications, a second or third (or more) sensing platform can be added to proportionally decrease the system sampling intervals with nominal increase in sensor cost and virtually no increase in setup and calibration times.
A crucial question, of course, is whether the data generated in miniature reactors by these sensors is at all relevant. That is, can this data be used in scale-up to larger vessels. Figures 3 and 4 show great promise that it can.
These figures indicate that, except for lag times due to temperature changes, the data in 3-ml cuvette tracks that measured in the lab-scale reactor. At volumes exceeding 3 ml, one would expect the correlations to be even closer. This issue is currently being explored at a number of sites using the 30-ml vessels shown in Figure 2.

Conclusions
The need for more efficient, less time-consuming methods of facilitating drug and vaccine development is mandated for a variety of reasons, with or without threats of vaccine shortages or biological warfare. The criteria for such systems include reduction of real cost per experiment, and relevance to the subsequent issues of scale-up and production.
A variety of methods for achieving these goals is currently being explored, and most are already in use at select sites. The comparative advantages and disadvantages are fairly apparent, but the fundamental question of “Which is best?” is not at all clearcut. Potential users will make their own decisions, based on individual goals and requirements, and tolerance to new technologies.

Joe Qualitz, PhD, is president and CEO of Fluorometrix Corp. of Stow, Mass. and Baltimore, Md., and has over three decades of experience in the design and development of products for the biotech, medical and other industries.