The low-hanging fruit is long gone from the bioprocessing tree and even the horse chestnuts in the upper branches are harder to reach. When biotech companies meet in Carlsbad, CA, for the IBC Life Sciences' “Antibody Production and Development” conference in March of this year, they will grapple with the question of how best to raise the quality of their product and at the same time boost quantity as demand for therapeutic antibodies spirals higher and higher.
The large doses that humanized antibodies demand for effective therapy, combined with the expanding markets for anticancer and anti-inflammatory antibodies, puts more and more pressure on companies to increase their product yields, says Scott Camberg of Human Genome Sciences(www.hgsi.com) “Typical Mab production in NSO cells employs a multifeed fed-batch strategy,” Camberg explains, “with the result that culture viability is below 40% as a result of nutrient depletion, metabolic buildup, and cytokine release upon cell death. While this approach still results in yields in the .5–5 g/L range, it may be accompanied by poor product quality due to degradative enzyme activity.”
In fact, 70% of an upstream production protocol goes toward expanding the cells for inoculation into the production reactor, with numerous opportunities for failure due to contamination or equipment malfunction. Other chances for breakdown result from the amount of time required for cell-mass buildup in the production vessel prior to initiation of the feeding schedule. These long run times are bedeviled by deviation and failure as conditions in the vessel become suboptimal, and when higher titers are achieved, demands are exerted on the downstream side of the equation, given that yields in the gram per liter range put pressure on the clarification and purification process.
So Camberg and his colleagues decided to eliminate, or at least minimize, these negative process aspects. Their proposal was based on developing a single seed train for the entire project, rather than for each production run, along with a shortened production process. This allows more harvests in the same amount of time and a higher cell viability upon harvest, with reasonable titers on the downstream side. Camberg’s team performed a series of bench scale and scale-up runs employing this format.
According to Camberg, the sequential runs were highly reproducible, and one seed train was successfully used to provide repeating production runs. Moreover, the production process decreased from 12 days to six, and the cost analysis yielded a 33% decrease in cost per gram simply by lowering the time between harvests from seven days to four.
“The ability of the cell line, process optimization, and the nature of the project should be the determining factors for production-process length, as well as the number of runs per seed train expansion,” Asserts Camberg.
The implementation of the process changes can contribute to lower cost per gram, greater ease of manufacture, and improved efficiency of facility usage. So the lesson is that minor improvements to protocol can produce benefits as substantial as those accrued through much more high-tech interventions.
Automated Optimization of Media
Robotic cell culture workstations are now developed to a point where they permit high-throughput screening of cell lines and culture conditions, a point that has not been lost on scientists at PD-Direct™, Invitrogen’s (www.invitrogen.com) integrated services for streamlined process development. Brian J. Horvath, R&D scientist, discussed the company’s use of the Microlab STAR workstation, engineered by Hamilton Robotics (ww.hamiltonrobitics.com), which has been adapted for cell culture as Cellhost.
The system offers a number of innovative features, including a platelifter adapted for tilting cell culture plates to allow complete removal of media and a robotic arm adapted to simulate manual movements.
Pipetting is performed by monitored air-displacement technology, which uses disposable tips without the need to change the hardware. In order to handle the disposable tips, Hamilton employs the CORE Technology, comprising an O-ring on the pipetting channel that is compressed and subsequently expanded into a groove present in the tips and needles. The mechanism releases itself to drop the tip off without dispersing the aerosol.
The plates are robotically transferred to 37°C CO2 incubators equipped with barcode readers and then monitored and fed automatically. The entire station is enclosed within a laminar air flow housing with automated UV decontamination.
Horvath combined the Hamilton technology with SimCell™ from BioProcessors (www.bioprocessors.com). The SimCell technology uses miniaturized bioreactors that can mimic the conventional controlled and monitored benchtop systems in a 150–1000-µL volume. The technology utilizes microfluidics, permeable films, membranes, and advanced optical measurement systems, generating scalable results on micro-engineered solutions. The combination of Hamilton and SimCell technologies is made available contractually to clients of PD-Direct and will initially be used for media development, process optimization, platform process creation, and clone selection.
PD-Direct has also introduced high-throughput technologies into earlier stages of process development through the addition of a ClonePIX™ instrument for screening and picking clonal cells into single wells. This technology, manufactured by Genetix (www.genetix.com), avoids human error and variation in pipetting.
The ClonePix automatically images, selects, and picks mammalian cell colonies and is compatible with a wide range of cell types, says Genetix. Screening and selection of colonies uses a number of parameters, including size, roundness, and neighboring proximity.
Hanne Bak, Ph.D., a staff engineer in preclinical manufacturing and process development at Regeneron (www.regeneron.com), discussed the use of continuous disc stack centrifugation in the company’s platform purification protocol. “We process a large number of different proteins,” she said, “and they run the gamut of shapes, sizes, and molecular weights, all the way from less than 15kD to over 200 kD.”
For their programs, the company needed a versatile and robust harvest technology for its CHO-based platform expression systems. The high concentrations of cells and proteins require a clarification process that will remove cells, debris, and sub-micron size particles that can plug downstream chromatography steps.
The Regeneron team uses a disc-stack technology, which ensures mechanical separation of different liquid and solid phases on a rapid, continuous basis. The interior of the centrifuge contains special plates, referred to as the disc stack, which provides additional surface area, greatly accelerating the process. The separation process is gentle, with shear cut to a minimum.
“We see a large number of different molecules in the course of a year,” Dr. Bak continued. “Many of these are pilot projects for internal research involving multiple-gram quantities of proteins where speed of delivery is the key. Therefore the continuous disc stack centrifugation technology has proven to be ideal for our needs.”
Dr. Bak concluded by stating that she and her colleagues were able to rely on their platform harvest and clarification steps without performing time and resource-intensive development or optimization work.
Better Downstream Processing
Improved yields of proteins at the upstream processing end have driven a demand for better performance at the downstream component. Uwe Gottschalk, Ph.D., vp of purification technologies at Sartorius (www.sartorius.com), has investigated membrane chromatography as an alternative to conventional packed column chromatography for protein purification.
Most antibody purification protocols involve the use of protein A columns as the primary capture mechanism, which typically can generate a 98% pure product. The subsequent treatment of the antibody calls for polishing and removal of DNA, endotoxins, proteases, and viral contaminants. But, conventional packed-bed chromatography requires columns with large diameters to permit high-volumetric flow rates and avoid a process bottleneck.
Because packed columns are configured for speed and not binding capacity, they will not be utilized to their maximum capability. In recent years, Q membrane chromatography has been re-evaluated for process scale due to the increased demands of the industry. Stacked 15-layer Q membranes have been successfully used for large-scale antibody production and viral removal.
For characterization studies, Dr. Gottschalk and his group evaluated different scale-down models, which were redesigned and carefully examined using different conditions. Scale-down modeling is essential in the biotech industry for establishing the performance of a process, but often yields a poor representation of the real life situation.
Since the original Sartobind Q 75 membrane generated higher backpressure, two new scale down devices, Sartobind Q 125 (3.5 mL) and eventually Sartobind Q 40 (1 mL) were designed in collaboration by Sartorius and Joe Zhou, Ph.D., of Amgen (www.amgen.com). The device mimics the liquid flow path found in the larger-scale module and represents an ideal scale-down tool for validation studies.
Using this approach, a Mab-processing capacity of 3 kg/M2 (11 kg/L) was achieved, which represents one membrane absorber for the polishing of each 10,000 L of bioreactor capacity. Because of their high cost and disposable nature, the Sartobind Q membranes do not economically outperform the Q column until all the categories, including hardware investment, labor, and utility costs, are factored into the picture.
According to Dr. Gottschalk, the disposable Q membrane chromatography devices are robust and simple to use, as they require no column packing or cleaning validation. Moreover, he says they appear to be an optimal technology for contaminant removal, and the amount of buffer required is less than 5% of that required for packed-bed chromatography. These studies validate Q membrane chromatography as a viable alternative to Q column chromatography for the polishing involved in late-stage antibody production.
“We like to think of ourselves as artists,” says Dr. Gottschalk. “Michelangelo claimed that he liberated a beautiful statue from its surrounding marble, and we think of bioprocessing in the same way, so we use purification technology to free an antibody from its contaminating viruses and detritus.”
Massaging Gene Sequences
“We founded Geneart in 1999 as a way of overcoming the high cost of outsourcing gene sequences for optimization,” states Marcus Graf, Ph.D., co-founder and operations manager of Geneart (www.geneart.com). “We were interested in building anti-HIV DNA vaccines, but the native sequences were too dangerous. Once we’d built the systems, we decided to offer gene design as a service, and the company grew on this foundation.”
A major challenge in reaching maximum antibody productivity in transfected cell lines is a suboptimal gene sequence. This may be the result of suboptimal codon usage, sequence repeats, cryptic splice sites, or excessive GC content. Moreover, even with optimized transcription, the amino acid sequence may be sub-optimal because it results in improper folding, protein instability, or a lethal, suicidal protein.
In order to deal with all these drawbacks, Dr. Graf and his colleagues developed a proprietary software, Gene Optimizer™, designed to produce the best possible sequence. The program takes into account the fact that codon choice for the various amino acids is nonrandom for different species and errant choices can greatly affect the expression rate of the protein.
“Screening all of the workable sequences for the best one would be impossible,” adds Dr. Graf. “The astronomical number of possibilities is greater than the number of stars in the universe, so we designed an algorithm that moves through the sequence using overlapping cassettes. It optimizes a string of three codons, then ratchets forward by one codon, optimizing the next three and so forth.”
The final sequence may be quite dissimilar from the wild type with the result that large improvements in quality and quantity of the final protein can be realized. For example, a redesigned GM-CSF protein expressed in human cells experienced a 50% increase in the amount of protein produced.
“We have seen a few cases in which the redesigned gene did not have higher expression rates, yet these were rare exceptions, and we have never encountered a redesigned gene with a lower expression rate,” Dr. Graf says.