Cell culture and fermentation scientists/engineers deal with the size of things they cannot see. Here's a cool website I came across that helps with the visualization.
http://learn.genetics.utah.edu/content/begin/cells/scale/
It uses Flash - not HTML5 - but it's worth about 30 seconds of your time.
As you zoom in, you'll see an antibody in there; it's crazy how small it is.
A week ago, the FDA released three documents to comply with the Biologics Price Competition and Innovation Act (BPCI Act) to create competition within the biologics space.
- Q&A for BPCI Act
- Scientific Considerations of Biosimilarity
- Quality Considerations of Biosimilarity
Why is the bankrupt U.S. government passing more laws to empower the FDA on this matter? Basically because the markets created biologics faster than the FDA was able to respond and this is essentially catch-up.
You see, unlike small biological molecules, large biological molecules (called "biologics") cannot be feasibly or economically synthesized with chemical reactions. Biotech companies differ from pharma companies because they genetically engineer microbes to secrete the complex biological molecules and then produce a lot of it with fermentation or cell culture.
The FDA has long held:
Quality cannot be tested into the product
I was at this IBC Conference once where a professor got up and said, "No one knows this to be more true than academia... each year we test our students more and more, and each year they don't get any smarter."
It has long been insufficient for drug companies to produce a drug that meets product quality specifications when substitute raw materials were used or procedures not followed during the manufacture of the drug.
Applying this rule to biologics, it would hold that anyone who doesn't have the original cell line used to manufacture the name-brand biologic would violate this FDA dogma and thus be forbidden from selling their biologic in the US markets.
In essence, there's no way for the FDA to allow biosimilars into the US markets without throwing everything they've been doing out the window. This lack of a regulatory pathway forbids biologics made by companies other than the original manufacturer to be sold in the U.S, thereby handing US biotechs monopolistic power.
This is why there's a BPCI Act in the first place. It is to force the FDA to create a pathway for generic biologics to enter the US markets and induce competition.
Watching the FDA on biosimilars has become a spectator sport for the wonkish Regulator Affairs folk. The folks over at Bioprocess Blog cover this much more thoroughly.
Suggested reading:
If you have about a minute and earphones, here's a really cool HTML5 website you ought to visit:
For the non-classical-musically minded, this is what the Naturalist guy in Master and Commander: The Far Side of the World was playing on his cello when they made it to the Galapagos Island.
There are a couple of cool things going on here:
- I can play this piece on the viola. I serenaded 3 girls at university with this piece.
- It's written in HTML5 - the future of web apps.
- It takes music played on 4 strings and 2 bow strokes and spreads it out to 8 strings and 4 points on rotating balls... taking a few complex processes and making it simple to visualize.
So basically, it combines a few personal and professional ideas and mashes it into something cool.
Here's you:
- You collect a ton of data from your large-scale cell culture/fermentation process.
- You're going blind Alt-Tabbing between Excel and JMP.
- You spend waaayyyyyyy too much time pushing around data and not getting answers.
And when you finally have the data the way you want it, your multivariate analysis tells you something like,
Final NH4+ (mmol) and Peak Lactate (g/L) correlate with Volumetric Productivity (mg/L/day).
Scientific curiosities are great for long-term process understanding, but when you're in the middle of a flagging campaign, manufacturing managers want to hear about immediate- and short-term actions they can take to meet the campaign goals.
The key to avoiding this career blunder (of presenting irrelevant work to your customers) is to select only actionable parameters for your main effects and interactions when building your multivariate analysis; in JMP, it looks something like:
In the above example, we can control inoculation density (Ini VCD) by extending the previous culture's duration. As well, a biologics license agreement may allow a window for executing pH shifts (VCD at pH Shift) as well when to feed (Cult Dur at Batch Feed). Actions that manufacturing can take by simple scheduling changes are ideal for putting into the multivariate analysis that deliver immediate solutions.
Constructing the main effects of your model by selecting actionable parameters is best for solving REAL manufacturing problems as well as for advancing your career as the person who finds the way to meet campaign goals.
At the core of mammalian cell culture process is reproducing the life-support systems that a mammal (like you) provides to their cells.
You - and by proxy, your cells - need oxygen, water, and nutrients. As well, you need warmth, the right pH balance and 'saltiness'. In addition to providing the nutrients and the environment, your body removes metabolic waste.
Large-scale cell culture is simply enabling a large stainless steel tank (called bioreactors) to support living mammalian cells. In reality, large-scale cell culture is filling a tank with sugar-water, spiking it with cells and letting it "stew."
Media
Cell culture is sometimes referred to as suspension cell culture because the cells are suspended in a nutrient-rich, pH-buffered liquid called media. Media is mostly water... extremely highly-purified water called "water for injection (WFI)." In it, you'll find nutrients like glucose or glutamine, sodium bicarbonate, trace amounts of metalic elements, folic acid and supplemental amino acids. Surfactant gets put in to help with the shearing (from the bubbles). The actual concentrations of these items are trade secrets; even if a firm outsources the media, the exact recipe is kept secret.
Sometimes, management-tampering happens right around now when a fully-defined media gets tested against the same media, except with peptones added in. Peptone is basically what you get when you take unused animal parts, grind it up, digest it with enzymes and make into powder. The idea here is, "Well, if it used to be a part of something that was once alive, then it ought to help support this here cell culture." Usually what happens is the undefined media outperforms the defined media and the peptone gets added in.
But fully-defined media is certainly the way to go if you're looking to reduce process variability: all your variables are known and controlled. Too many times, people blame variability in their undefined media as the source of their process variability. Some say that unknown nutrients is the secret sauce of peptones; others say it acts as a surfactant. Whatever it is - don't choose higher long-term process variability over short-term yields.
Environment
If you ever dump cells into water, you'll watch them sink to the bottom. In order to scatter the cells and make sure they are evenly distributed, you have to stir it with an agitator (impeller). The agitation also helps to evenly distribute:
- heat
- acid/base
- dissolved oxygen
The heat is added/removed by flowing hot/cold water around the outside of the bioreactor (called a jacket).
The acid is added by sparging carbon dioxide from the bottom of the reactor. As the CO2 bubbles its way to the top, it forms carbonic acid decreasing the pH. The base is added with carbonate as a fluid.
Oxygen is provided to the cells by sparging air/oxygen from the bottom of the reactor. As the air/O2 bubbles its way to the top, it is smeared into the media by the agitator, providing it to the cells.
The agitator speed is typically set by the power-to-volume ratio from dusty ChemE textbooks. In addition, the vessel is pressurized to ensure the direction of flow is away from the bioreactor.
Control systems are in place to maintain pH, Temperature, dissolved oxygen, pressure and agitation. Engineers have figured out how to control reactors since the beginning of time. The challenges of controlling bioreators is maintaining sterility since a cell-growth-promoting environment also promotes the growth of bacterial contaminants.
Cells
The final piece of cell culture is, of course, the cells. From a manufacturing standpoint, the cells are another ingredient... not much different than glucose solution in that it is added to the bioreactor to be stirred.
The genetically engineered cells are typically cryogenically frozen until they are needed. A vial of cryogenically cells is thawed (think Wesley Snipes and Sylvester Stallone in Demolition Man) into successively larger bioreactors. Typically at the 20L scale, the cells are maintained by "solera." Solera is when the cell density has increased near the maximum the media can support, a large fraction of the cell culture is siphoned off, leaving a smaller fraction. The larger fraction is used to inoculate another bioreactor while the smaller fraction gets fresh media to continue growing.
Recap
There's a lot more to cell culture than can be described in a blog post. There's seed cultures/selective media, inoculum cultures/growth media, production cultures/production media. There's fed-batch operations, there's perfusion. But in the end, cell culture processes are a lot like cooking. You put the ingredients: water, nutrients, cells into a bioreactor and then you wait until the cells have consumed the media and then you're done.
What's surprising is how maintaining large-scale biological systems actually is.
Drugs used in modern medicine are manufactured with mammalian cell culture.
Why Mammalian Cells?
Humans have been making therapeutic biological molecules using fermentation since Alexander Fleming sneezed into a Petri dish and discovered penicillan. But complex biological molecules, like antibodies, cannot be produced by microbes because microbes cannot do post-translational processing. Post-translational what?
Basically when our bodies construct a human protein from the things we eat, our cells will "sign" the newly created protein with a carbohydrate. This "signature" not only helps the protein fold, but also helps our bodies recognize the protein as self-made. Microbes do not have the machinery to do this post-translational glycosolation such that proteins made by them simply unfold and lose their therapeutic value.
What is Cell Culture?
Cell Culture refers to the cultivation of (typically mammalian) cells. The reason companies use mammalian cells is because the active pharmaceutical ingredient (API) in drugs are complex molecules that cannot be made with chemical synthesis and also needs post-translational processing. To make the API, firms genetically engineer cells to secrete the complex molecule. Cell culture is the art/science/process of growing mammalian cells in at a large-scale so the purified API can be sold.
Inoculum/Seed Cultures
Part 1 of 2 in cell culture is growing the cells to a sufficient amount. You'll hear people say, "Scale biomass" when referring to "seed" or "inoculum" cultures. Note: we've assumed the cells have been engineered and are stable. The goal is to get as many cells in as short a time as possible - i.e. get fast growing cells. To do this, you get some cells and put it in a large volume of nutrients at growth-promoting pH (somewhere near 7) and temperature (between 30 to 40 degC). The cells will grow, consuming nutrients and secrete metabolic waste (CO2 and lactic acid...etc). To keep the cells growing, you must replenish their nutrients or remove the waste - and companies do this by transferring the cells to a larger bioreactor... the classic "solution to pollution is dilution."
Production Cultures
Part 2 of 2 in cell culture is having the cells secrete the API. Afterall, your sales comes from the API, not the cells themselves. The cells get their final transfer into a bioreactor with nutrients that encourage the secretion of API. The bioreactor is typically called the "production bioreactor" and the nutritious fluid is typically called "production media." After the cells produce the API, the cells are typically discarded while the API is "harvested" for purification - a.k.a. "Downstream processing".
One Unit Operation
Many in the downstream world (Rob Caren) like to take pot-shots at cell culture scientists and engineers and ask, "How hard can this be? There's just one unit operation, right?" Sure. But at large-scale, very few things in cell culture are well understood, which is why armies of chemical engineers are hired to do multivariate data analysis on cell culture.
So in addition to providing this brief primer on mammalian cell culture, the reason I'm putting this out there is that the ultimate goal of the data acquisition is to enable multivariate analysis that confers process understanding... something which Zymergi engineers - given our first-hand experience supporting large-scale cell culture processes - is able to do.
Mark Graban has an excellent post on the Lean Blog on the blaming of the Stanford kicker for their 41-38 loss to Oklahoma State in the Fiesta Bowl. Holding with the dogma of lean, failure of a system (football team to win the game) is rarely a single root cause (the kicker).
Mark's post is entertaining as it is informative. Here is yet another great application of lean thinking:
http://www.leanblog.org/2012/01/blame-the-stanford-kicker-blame-the-kicker/
This one time - we were harvesting a production bioreactor with several million dollars worth of product - an operator left a valve open and 33% of product went to drain. The plant manager (VP) was fuming mad - rightfully so.
"I pay for all this monitoring software, how come no one could tell that a valve was open?"
It's a good question. I drifted off thinking that we didn't have a tag for every valve indicating whether or not it was open.
"We need a mass balance," he said to no one in particular.
And then I realized we could have a mass-balance since we have load cells on the bioreactor and harvest tank, as well as a flow totalizer on the centrifuge. In the case of biologics processing, we are dealing with constant density volumes, which means that a mass-balance is often the same as a volume balance. The pipes between the bioreactor and the centrifuge holds volume (called "hold-up"), as well, there is hold-up between the centrifuge and the harvest tank.
Plotted on the same trend using PI ProcessBook we get something like this:
From this perspective, we can easily determine if there is volume-loss between the centrifuge and the harvest unit by looking at the slopes: if the slopes are the same and the lines are parallel, we have no mass-loss. But it's hard to tell if there is any loss of mass between the bioreactor and the centrifuge since the line is downward sloping.
There's a trick in PI ProcessBook where you can simply reverse the Max and Min for the trace making the bottom axis the larger number and the top axis the smaller number:
This simply makes the top of the trend the small number and the bottom of the trend the larger number. Be sure to use Multiple Scales so only this tag is plotted upside-down. What you get is a trend where everything is sloping upwards:
This trend shows that the mass is balanced (i.e. no losses in the closed system).
But suppose there is a valve open on the line between the fermentor and the centrifuge... what does that look like?
In this case, we see that the slope of the totalized volume processed by the centrifuge is less than the slope of the fermentor. Literally, the rate of volume lost by the fermentor is greater than the rate of volume gained by the centrifuge... the difference is the amount gone down to drain.
Likewise, should there be losses just between the centrifuge and the harvest unit, the totalized volume of the centrifuge and fermentor volumes should match in slope, while the harvest unit should have a lesser slope.
PI ProcessBook is ideal for monitoring your process... especially if you need to know something in real-time (e.g. "Hey, there's a valve open and it shouldn't be").
In biologics manufacturing, there really is no excuse for having losses like these when we can monitor the entire system and prevent significant (in this case, seven-figure) losses in our operations.
Zymergi serves companies that run cell culture and fermentation processes. Nearly all our customers produce biologics... molecules that are synthesized by biological organisms - and all of them get their GMP plants inspected by the FDA.
Yesterday, FDAzilla published the rate the FDA hands out inspection observations called 483 for 2011. It's eye-popping; on average, they hand out 1 (Form) 483 every 50 minutes.
If you need more information on avoiding headache with the inspectional liability side of running cell culture/fermentation processes, go ahead and visit:
Say you are running a 2g/L product from a 10-day process at your 1 x 6000L plant, with strict orders from management to minimize downtime. This product is selling gangbusters, which means every gram you make gets sold, which means you've got to make the most of the 80-day campaign allotted for this product.
The volumetric productivity for the process is 2g/L/10days = 0.2 g/L/day.Running a 6000L capacity plant gives you
- 12 kilos every 10 days.
- 8 run slots given the 80-day campaign
- Maximum product is going to be: 96 kg for the campaign.
But suppose your Manufacturing Sciences team ordered in-process titer measurements and found that Day 8 titers were 1.8 grams per liter. 
Harvesting at day 8 means:
- 10.8 kilos every 8 days.
- 10 run slots given the 80-day campaign
- Maximum product is going to be 108 kg.
By harvesting earlier, you gain 2 additional run slots... during which time you can make 21.6 kg; but since you lost 1.2 kg/run for 8 runs totalling 9.6 kg, the net gain is 12 kgs.
There are a lot of assumptions here:
- Your raw material costs are low relative to the price at which you can sell your product
- Your organization is agnostic to doing more work (10 runs instead of 8).
It is difficult for plant managers to end a culture early to get 10.8 kgs when simply waiting 2 more days will get you 12 kgs. It rapidly becomes easy when you see how 2 run slots open up and you have the opportunity to make 21.6 kgs to make up for the lost product from ending the fermentation early... or rather, the point of maximum volumetric productivity.
