Emil W. Ciurczak, DoraMaxx Consulting06.12.18
The idea of process control for bioprocesses was introduced by Rick Cooley (Eli Lilly) for insulin production and purification. The quality control (QC) lab was constructed and the plant was, quite literally, built around the lab. Holes were drilled into the wall(s) where the fermentation tanks and purification columns were located. He eventually designed and built the first real-time process chromatography equipment for his own use. Since Lilly is a pharmaceutical entity, not an instrument company, he sought a vendor willing to supply similar equipment. [For those of you who haven’t had the “privilege” of bumping against the “not invented here,” a.k.a. “NIH” attitude of instrument companies, you have had a long, happy life.] Eventually, he persuaded Dionex (now part of Thermo-Fisher) to produce and sell them to him, allowing him to do the first biotech PAT work.
Please keep in mind that “real-time” for biopharma is nowhere near the time scales as for small molecule production—with the exception of API synthesis, but that is for another day. Where we look for millisecond scan/analysis speeds for tablet production, twenty minutes for a measurement/analysis in bioproduction is fine. Simply stated, a 12, 24, 36, or even 72-hour fermentation time scarcely needs data produced every second or two—even with inexpensive storage space for data, the retrieval headaches would be phenomenal. Five, ten or even 15-minutes between measurements is quite fine, especially considering that current “wet” chemical tests used in biotech analyses can take hours to glean the data necessary (i.e., time to add more nutrients, change stir rate, or quench the reaction to harvest the end product).
Back in the 1980s, when I suggested that fermentations could be followed by near-infrared, the “classically-trained” biochemists told me the spectra were too amorphous and they couldn’t see specific peaks and things like ammonium (nitrate) wouldn’t show up, etc. So, 30-years ago, I suggested a simple experiment; insert the probe and scan over time. When the scans no longer changed, the reaction had stopped and samples could be taken for “normal” testing. Why do this? Bioreactions are not as well defined and “normal” chemical reactions. Depending on the starting material, inoculation culture(s), speed of agitation, temperature variations, and such, the reaction could take +/- 25% to reach feed points or end points. Having an SOP that tells the operator to sample at set times can introduce problems: if the reaction is anaerobic, oxygen could enter; the operator could be exposed to the bacterium; and lastly, taking extra time-consuming assays is a waste of time and materials. Taking samples when your spectrometer signals is far smarter.
Obviously, much progress has been made since then. Numerous determinations have been generated and are used on a routine basis. A typical example of real time analysis is NIRS being used to monitor a bioreactor in situ during a CHO (Chinese hamster ovary) cell culture reaction. By inserting the proper probe, readings may be taken as often or seldom as empirical results dictate. Some simultaneous measurements include but are not limited to glutamine (Figure 1), lactate (Figure 2), glucose (Figure 3), ammonia, titer, methionine, and glutamate. Other reported predictions are pH, dissolved oxygen, biomass, and osmolity. Certainly, other tests could be developed, as well.
Figure 1. Glutamine Concentration: circular points = wet analysis, solid points = NIR prediction
Figure 2. Lactate Production: Solid points = NIR; circles = Wet Analysis
Figure 3. Glucose Concentration: solid points = NIR; white circles = wet analysis
What, then, is the purpose of spending all that money on real-time testing equipment NIRS, Raman, HPLC/UPLC (Ultra-high Pressure or Precision LC)? Well, for one reason, most contract manufacturing organizations (CMOs) and contract research organizations (CROs) do not have unlimited time, money, or personnel. While a larger organization—originator, often owned by a large company with deep pockets—may have “excess” capacity and a smaller number of products, CMOs are smaller, leaner, and have a high number of products to turn over in a set amount of time. The traditional manner of bioprocessing is both time and labor intensive, and not conducive to a smaller CMO starting a program. Add to that the expertise available to a typical traditional biopharma company.
What do I mean by “available expertise?” We may simply compare the disciplines in a traditional larger, established small molecule company with a macromolecular producing bio-pharma plant. The small molecule company has organic and analytical chemists, including R&D and QC; pharmacists, including formulators and clinicians; engineers; often an instrument group; and an array of instruments, including chromatography, spectroscopy, and physical measuring devices to call upon. The bioprocess companies? In most instances, they are run and staffed by people trained in the biological sciences: biochemists, microbiologists, bacteriologists, and so forth.
In the small molecule company, there are numerous steps from the initial synthesis of the drug candidate, through clinical testing, formulation, pilot plant to final production. Any number of technical disciplines “touch” the small molecule and any number of analytical techniques are used to measure these steps. To move from lab-based testing to process testing and control is simply a matter of upgrading existing instrumentation and software. The largest hurdle is management buy-in—convincing them of a real ROI—in most cases. The argument is, “We’re making a profit and been doing things this way for 50 years; why change now?”
In the macromolecular world, we have a completely different set of hurdles to jump. In a nutshell, homogeneity of backgrounds is the main problem. That is, at every level—discovery through final production—we have the same disciplines addressing every problem. The analysis of the first small amount of some bioproduct is, inevitably, the same as for a 1,000-liter reactor. As a consequence, there is no “advantage-of-scale” solutions sought. When a small molecule QC lab is confronted with dozens of lots of tablets to be analyzed, the manager/analysts seek alternate methods of analysis. Simply moving from HPLC to UPLC could shorten an individual analysis from 10-15 minutes to one-to-three minutes—a tenfold speed advantage is realized.
With traditional bioassays, the development time is seldom able to be sped up proportionately to a wet chemical analysis. When presented with an alternative technology, a small molecule chemist is more likely to at least be open to considering change. Traditional, or current, biological assays are, due to natural growth times, seldom able to be “hurried.” As a consequence, traditional bio-analysts have little experience with more rapid/alternative technologies. Framed in the picture of growth in an agar plate, technologies such as near-infrared, Raman, and such do not seem to be alternatives to be studied.
However, with the advent of rapid bio-assays, the “why bother when we still have to wait for the plates to develop’ ecus no longer holds water. At this point, allow me to emphasize one important point: any assay (prediction) performed by NIRS, Raman, or any other non-traditional method can also be confirmed since these technologies are non-destructive! Unlike traditional comparisons (e.g., a new HPLC method to existing one), where a sample is divided between two sample preps, then assayed (paired t-test), a spectroscopic method samples the exact same material as the traditional (compendial) method. BOTH may be run for confirmation for as long as necessary to appease quality assurance (QA) personnel.
Having stated that, why is it less specific for us to determine biomass by NIRS than measure the energy needed to turn the mixing blades? I have seen numerous cases where the power curve—energy needed to keep paddles at the same speed—is used to follow the growth of biomass. The power of Chemometrics, used with NIRS can be at least as accurate, while having the benefit of actual spectra to examine and file. Figure 4, for example, follows the titer of the CHO growth versus conventional measurements.
Figure 4. Following the Titer in a CHO growth medium
All these measurements mentioned so far are, of course, organic and are expected to have NIR spectra. What about inorganics? Figure 5 shows the determination of ammonia during the CHO reaction over 300 hours. While NH4+ doesn’t have a spectacular spectrum, its hydrogen bonding affects the overall spectrum of the mix, allowing for an algorithm to follow it in real time. While it does not have a characteristic spectrum, the resolution, low noise, and reproducibility of a good NIR system, the use of Chemometrics allows for the monitoring of things such as NH4+, dissolved O2, and the like.
Figure 5. Level of ammonia throughout a 300 hour (CHO) run
The next comment could well be, “Why bother speeding up the fermentation when downstream clean-up takes so long, anyway?” While that is a topic for a later column, let me say, “Spoiler alert: NIRS can be used here, as well.” For the time being, let’s just say that monitoring the bioprocess can allow for better sampling times and allows an operator to be aware of problems between formal sample times. As a bonus, a more rapid turnover of equipment optimizes its use, essentially allowing the production of more batches without adding hardware.
Emil W. Ciurczak
DoraMaxx Consulting
Emil W. Ciurczak has worked in the pharmaceutical industry since 1970 for companies that include Ciba-Geigy, Sandoz, Berlex, Merck, and Purdue Pharma, where he specialized in performing method development on most types of analytical equipment. In 1983, he introduced NIR spectroscopy to pharmaceutical applications, and is generally credited as one of the first to use process analytical technologies (PAT) in drug manufacturing and development.
Please keep in mind that “real-time” for biopharma is nowhere near the time scales as for small molecule production—with the exception of API synthesis, but that is for another day. Where we look for millisecond scan/analysis speeds for tablet production, twenty minutes for a measurement/analysis in bioproduction is fine. Simply stated, a 12, 24, 36, or even 72-hour fermentation time scarcely needs data produced every second or two—even with inexpensive storage space for data, the retrieval headaches would be phenomenal. Five, ten or even 15-minutes between measurements is quite fine, especially considering that current “wet” chemical tests used in biotech analyses can take hours to glean the data necessary (i.e., time to add more nutrients, change stir rate, or quench the reaction to harvest the end product).
Back in the 1980s, when I suggested that fermentations could be followed by near-infrared, the “classically-trained” biochemists told me the spectra were too amorphous and they couldn’t see specific peaks and things like ammonium (nitrate) wouldn’t show up, etc. So, 30-years ago, I suggested a simple experiment; insert the probe and scan over time. When the scans no longer changed, the reaction had stopped and samples could be taken for “normal” testing. Why do this? Bioreactions are not as well defined and “normal” chemical reactions. Depending on the starting material, inoculation culture(s), speed of agitation, temperature variations, and such, the reaction could take +/- 25% to reach feed points or end points. Having an SOP that tells the operator to sample at set times can introduce problems: if the reaction is anaerobic, oxygen could enter; the operator could be exposed to the bacterium; and lastly, taking extra time-consuming assays is a waste of time and materials. Taking samples when your spectrometer signals is far smarter.
Obviously, much progress has been made since then. Numerous determinations have been generated and are used on a routine basis. A typical example of real time analysis is NIRS being used to monitor a bioreactor in situ during a CHO (Chinese hamster ovary) cell culture reaction. By inserting the proper probe, readings may be taken as often or seldom as empirical results dictate. Some simultaneous measurements include but are not limited to glutamine (Figure 1), lactate (Figure 2), glucose (Figure 3), ammonia, titer, methionine, and glutamate. Other reported predictions are pH, dissolved oxygen, biomass, and osmolity. Certainly, other tests could be developed, as well.
Figure 1. Glutamine Concentration: circular points = wet analysis, solid points = NIR prediction
Figure 2. Lactate Production: Solid points = NIR; circles = Wet Analysis
Figure 3. Glucose Concentration: solid points = NIR; white circles = wet analysis
What, then, is the purpose of spending all that money on real-time testing equipment NIRS, Raman, HPLC/UPLC (Ultra-high Pressure or Precision LC)? Well, for one reason, most contract manufacturing organizations (CMOs) and contract research organizations (CROs) do not have unlimited time, money, or personnel. While a larger organization—originator, often owned by a large company with deep pockets—may have “excess” capacity and a smaller number of products, CMOs are smaller, leaner, and have a high number of products to turn over in a set amount of time. The traditional manner of bioprocessing is both time and labor intensive, and not conducive to a smaller CMO starting a program. Add to that the expertise available to a typical traditional biopharma company.
What do I mean by “available expertise?” We may simply compare the disciplines in a traditional larger, established small molecule company with a macromolecular producing bio-pharma plant. The small molecule company has organic and analytical chemists, including R&D and QC; pharmacists, including formulators and clinicians; engineers; often an instrument group; and an array of instruments, including chromatography, spectroscopy, and physical measuring devices to call upon. The bioprocess companies? In most instances, they are run and staffed by people trained in the biological sciences: biochemists, microbiologists, bacteriologists, and so forth.
In the small molecule company, there are numerous steps from the initial synthesis of the drug candidate, through clinical testing, formulation, pilot plant to final production. Any number of technical disciplines “touch” the small molecule and any number of analytical techniques are used to measure these steps. To move from lab-based testing to process testing and control is simply a matter of upgrading existing instrumentation and software. The largest hurdle is management buy-in—convincing them of a real ROI—in most cases. The argument is, “We’re making a profit and been doing things this way for 50 years; why change now?”
In the macromolecular world, we have a completely different set of hurdles to jump. In a nutshell, homogeneity of backgrounds is the main problem. That is, at every level—discovery through final production—we have the same disciplines addressing every problem. The analysis of the first small amount of some bioproduct is, inevitably, the same as for a 1,000-liter reactor. As a consequence, there is no “advantage-of-scale” solutions sought. When a small molecule QC lab is confronted with dozens of lots of tablets to be analyzed, the manager/analysts seek alternate methods of analysis. Simply moving from HPLC to UPLC could shorten an individual analysis from 10-15 minutes to one-to-three minutes—a tenfold speed advantage is realized.
With traditional bioassays, the development time is seldom able to be sped up proportionately to a wet chemical analysis. When presented with an alternative technology, a small molecule chemist is more likely to at least be open to considering change. Traditional, or current, biological assays are, due to natural growth times, seldom able to be “hurried.” As a consequence, traditional bio-analysts have little experience with more rapid/alternative technologies. Framed in the picture of growth in an agar plate, technologies such as near-infrared, Raman, and such do not seem to be alternatives to be studied.
However, with the advent of rapid bio-assays, the “why bother when we still have to wait for the plates to develop’ ecus no longer holds water. At this point, allow me to emphasize one important point: any assay (prediction) performed by NIRS, Raman, or any other non-traditional method can also be confirmed since these technologies are non-destructive! Unlike traditional comparisons (e.g., a new HPLC method to existing one), where a sample is divided between two sample preps, then assayed (paired t-test), a spectroscopic method samples the exact same material as the traditional (compendial) method. BOTH may be run for confirmation for as long as necessary to appease quality assurance (QA) personnel.
Having stated that, why is it less specific for us to determine biomass by NIRS than measure the energy needed to turn the mixing blades? I have seen numerous cases where the power curve—energy needed to keep paddles at the same speed—is used to follow the growth of biomass. The power of Chemometrics, used with NIRS can be at least as accurate, while having the benefit of actual spectra to examine and file. Figure 4, for example, follows the titer of the CHO growth versus conventional measurements.
Figure 4. Following the Titer in a CHO growth medium
All these measurements mentioned so far are, of course, organic and are expected to have NIR spectra. What about inorganics? Figure 5 shows the determination of ammonia during the CHO reaction over 300 hours. While NH4+ doesn’t have a spectacular spectrum, its hydrogen bonding affects the overall spectrum of the mix, allowing for an algorithm to follow it in real time. While it does not have a characteristic spectrum, the resolution, low noise, and reproducibility of a good NIR system, the use of Chemometrics allows for the monitoring of things such as NH4+, dissolved O2, and the like.
Figure 5. Level of ammonia throughout a 300 hour (CHO) run
The next comment could well be, “Why bother speeding up the fermentation when downstream clean-up takes so long, anyway?” While that is a topic for a later column, let me say, “Spoiler alert: NIRS can be used here, as well.” For the time being, let’s just say that monitoring the bioprocess can allow for better sampling times and allows an operator to be aware of problems between formal sample times. As a bonus, a more rapid turnover of equipment optimizes its use, essentially allowing the production of more batches without adding hardware.
Emil W. Ciurczak
DoraMaxx Consulting
Emil W. Ciurczak has worked in the pharmaceutical industry since 1970 for companies that include Ciba-Geigy, Sandoz, Berlex, Merck, and Purdue Pharma, where he specialized in performing method development on most types of analytical equipment. In 1983, he introduced NIR spectroscopy to pharmaceutical applications, and is generally credited as one of the first to use process analytical technologies (PAT) in drug manufacturing and development.
