Scale up, look sharp – why analytics in large-scale cell therapy manufacture is so difficult

CAR-T and other cell and gene therapies represent an exciting new paradigm in how we approach previously untreatable diseases. But as ‘living drugs’ where the cell is the product, QC / analytical testing is more important than ever to ensure safety and efficacy. Analytics can be thought of in terms of:

• Process analytics – collecting data on variables that affect the quality of the cells, like temperature, pH, dissolved oxygen, and metabolite concentration, throughout the process to analyse and refine the process
• Product analytics – collecting data through or at the end of the process to analyse and characterize the product (in this case the cells), including cell count, cell viability, and flow panels

As per GMP guidelines, analytics that are critical to the manufacturing process and not measured automatically, ‘Critical Quality Attributes’ and ‘Critical Process Parameters’, are measured by the QC team. While batched QC analysis is achievable in traditional pharma manufacture, it isn’t optimal for production of complex advanced therapies. In order to scale production of cell and gene therapies and reduce the all important vein-to-vein time developers need to:

• Limit people-intensive manual measurements and recording which are laborious and error-prone
• Minimise manual sample and data exchanges between manufacturing and QC teams

The obvious solution would then be to have online measurements and sample analysis. But how can this be achieved across multiple bioreactors without significant capital outlay? Fitting an array of different analytical technologies all measuring different things on each and every bioreactor is expensive (usually being restricted to just a few bioreactors at a time), and the unfortunate truth is that current technologies can barely automate pH and dissolved oxygen measurement reliably. Automated cell count is often only an estimate, only a handful of early-stage technologies are looking at measuring online cell viability, and no one can do label-free flow cytometry at scale.

“From an engineering perspective, we knew we had to think about things differently to solve the challenge of making online analytics a reality as we designed the Cyto Engine platform. So we came up with the concept of mutualising the analytics we integrated, i.e. one set of analytical tools that can make measurements on multiple bioreactors. We also looked to build in flexibility on what we can integrate  – that allows us to add to the system as analytical technology improves as well as allowing customization on what analytics come built in,” says James Davies, VP of Engineering at MFX.

What’s more, the mutualisation contributes negligible CapEx, making manufacturing cost-effective in the long-term.

Why successful cell therapy processes need to scale both ways

When it comes to cell therapy manufacturing, much of the industry is focussed on optimizing large scale manufacture – whether that’s scaling up larger batches for allogeneic therapies, or scaling out more batches for autologous therapies. This makes sense if the only issue in delivering cell therapies was the lack of manufacturing capacity. But with over a third of clinical stage cell therapies reporting issues ranging from safety and efficacy to CMC, and batch failure rates reportedly in the low double digits at commercial scale, understanding the whole development process is crucial.

To gain this insight Process Development (PD) teams often use Design of Experiment (DoE), varying process parameters in a combinatorial way to understand the impact on cells and determine critical process parameters. But there’s a problem.

Most cell culture tools are either designed for research OR manufacturing, meaning a process in a research tool won’t translate through to a larger volume manufacturing process and vice versa. Therefore, PD teams need to decide whether to continue with (often) manual research tools or jumping to (usually) automated manufacturing tools. And there’s pros and cons with each:

Optimizing a process with research tools

Pros

• Experiments can be kicked off quicker
• Reagent costs can be kept low
• Several experiments can be run in parallel

Cons

• Manual interventions generate variability
• There’s a lack of process data
• There’s a lot of re-work needed to translate the process for large-scale manufacturing

Optimizing a process with manufacturing tools

Pros

• Future proofs the process for large-scale manufacturing
• Can capture more data than manual processes

Cons

• High costs to run – one experiment can cost up to $30k in reagents because of the large volumes
• Can only run one experiment per device
• In-process analytics are basic at best

So what’s the solution?

We asked Lindsey Clarke, the latest member of the MFX team what was on her wish list.

“We need cell culture tools that can translate seamlessly – you won’t believe how many times I’ve been asked for this over the last decade. So many potential therapies work great at the research bench when you’re only thinking about getting the best cell, but then you try and scale them up and the biology doesn’t like it or you’re limited by what you can do at GMP, its improving as more tools companies realize this but it’s by no means seamless, especially when it comes to what you grow your cells in”

And what does it look like?

“An ideal research tool for me would be a perfectly scaled-down version of the manufacturing tool, capable of automatically running several dozen experiments, requiring very little reagents, and outfitted with live in-process analytics, so you can really understand the biology that’s happening when you change parameters.”

At MFX, we’ve been working on making this ideal world a reality. Our Cyto Engine™ performs automated cell culture in a research setting by multiplexing the exact same microfluidic chambers used in automated manufacturing, with each chamber carrying out a distinct experiment. This way, the cells are exposed to the same environment, no matter the scale. 

Is innovation in automation enough to drive cell therapy manufacturing optimization?

Moving away from manual handling in large-scale cell therapy manufacturing is undisuptably the next step in optimization. Manual operations are often tricky to replicate between different operators and sites. Ensuring reproducibility is challenging and very expensive, especially in light of the lengthy training needed for GMP manufacturing. With the 20% Compound Annual Growth Rate (CAGR) of the cell and gene therapy industry, we’re likely facing a shortage of these skilled operators in the future. On top of all of this is human error- people unavoidably make mistakes.

So is automation the answer?

Several companies are currently developing automated bioreactors for cell therapy manufacturing. Previously, the approach to automation involved simply taking traditional cell culture vessels such as bottles or flasks and adding some level of automation to them – such as tubes for automated media change. Nowadays more sophisticated devices focus on improved systems that are purposely built for automation, such as cartridges or integrated valves and chambers.

This may remove the cost of manual operators, but is it enough to revolutionize the whole manufacturing process? Here’s 3 reasons why automation alone won’t cut it.

Number 1 – cells don’t care about automation

“Cells need to thrive in the culture system they are being grown in, regardless of whether this is automatic or manual”, says James Kusena, VP of Bioprocessing and Applications.

Currently, cell batches that don’t meet the quality threshold can reach up to 10% during manufacture. And that’s for commercial-stage batches that come from years of optimization. Automation won’t fix this.

Number 2 – automation is not process control

Automation allows for feedback loops but it doesn’t fix quality of measurement. “You don’t know if what you are measuring on one end reflects what is happening on the other,” explains James, “and when you try to course-correct your process, you might course-correct one end and make the other worse. This leads to very heterogeneous cell populations, the opposite of what cell therapy needs”. Once again, automation won’t fix this.

Number 3 – scientists need to be able to go up or down the ladder

Automation won’t fix the scale-down issue James explains; “While scaling up is often the focus and the ultimate goal for translating to large-scale manufacture, when scientists need to optimize the process scaling down is necessary to save on costs and allow large datasets to be gathered”. Therefore, we need bioreactors that allow scaling down and this is independent of whether they are automated or not.

Instead of focusing on automating a slightly improved version of cell culture setups, we need to start from the ground up and build sustainable systems to really revolutionize cell therapy manufacturing. At MFX, we are developing the Cyto Engine™, a microfluidics-based bioreactor where cells can thrive, be controlled precisely and, thanks to highly parallelized microfluidics, be truly scaled up and down. And yes, it is automated. 

Footprint size matters in cell and gene therapy manufacturing

Cell and gene therapies have demonstrated their extraordinary, curative potential in the last decade. They’ve had the biggest impact in liquid tumors, but with over 1000 clinical trials in the advanced therapy space underway, cures for a range of diseases are on the horizon.

The latest ground breaking news in the space is the European Marketing Authorization for Kite’s Yescarta as a second line treatment for B-cell diffuse lymphoma. Until now, CAR-T therapies have only been approved for refractory patients, with patients having to undergo at least 2 lines of treatment before being eligible. But with the increased focus on getting cell and gene therapies to market, the expansion of approved therapies to wider patient populations, and the use of these therapies earlier in the treatment pathway, manufacturing capabilities need to be drastically improved to meet the oncoming demand.

This opens up a plethora of problems. Aside from the commonly discussed issues; the lack of qualified manufacturing personnel, supply chain issues with reagents and plasticware, the high batch failure rates plaguing manufacturing processes, and the logistics of delivering cell therapies to patients; there simply isn’t enough space for large-scale cell and gene therapy manufacturing. So how big is this problem? Let’s do a rough breakdown:

1 clean room = ~150sqm

à This can run 10 simultaneous batches with the best automated platforms on the market

For a typical CAR-T process with a team working full day and night shifts:

1 batch = 2 weeks

So in 1 year: 10 batches x 2 weeks x 12 months = 240 batches

Currently CAR-T manufacturing represents less than 1% of clean room utilization and the number of patients treated with cell therapies every year is in the thousands. But the addressable population for just CAR-T therapy is going to be ~2 million patients in the next 5-10 years. Which will require:

8,000 clean rooms = 1,200,000 sqm

That’s equivalent to the current total cleanroom capacity worldwide.

In reality, this space should be doubled to factor in space for QC labs, storage space, offices etc, and including the manufacture of non-CAR-T cell therapies this requirement could even be tripled. This substantial need for more space is good news for clean room suppliers. But the space and construction requirements, and the expensive equipment needed to run these clean rooms to regulatory standards isn’t so good for cell and gene therapy developers, patients or the environment.

At MFX we are developing the Cyto Engine™, an automated cell therapy manufacturing platform that can produce up to 100 batches per square meter. That’s more than 30x the number  of batches per square meter than current manual processes, and more than 4x the number of batches per square meter than current automated processes.

How to ensure a consistent cell expansion process? Don’t start at the end

Cell culture has a culture of measuring endpoints. Vessels used to expand cells (like multi-well plates and flasks) are essentially mystery boxes in the days or weeks of culture. It’s only at the point of cell harvesting that most researchers measure and analyse their cells, which means all the information that they need to know – what goes on inside the vessels during expansion, what effect the environment is having on the cells, is the expansion process consistent – remains an enigma.

One of the issues facing sampling during culture is the size of the vessels. For T-25 flasks or multi-well plates, removing enough cells or medium to give accurate readings will disrupt the culture and substantially affect results. For mid-range vessels like large flasks or bags this is less of a problem, but only sampling once or twice per process operation (as is usually the case) doesn’t accurately reflect cell behavior over time.

The next big issue is the analytical equipment needed. Most labs have flow cytometers, cell counters, microscopes, and PCR machines. But few have more advanced equipment like metabolite analyzers, and on-line, real-time measurement capabilities don’t exist for standard research vessels. Sensors that continuously monitor parameters like pH and dissolved oxygen can be found on (very) expensive automated stir tank bioreactors, but when was the last time you saw a T-75 hooked up to a computer?

To understand cell behavior and ensure a consistent expansion process, measuring the parameters that can affect cells during culture is crucial. Existing indicators like color changing medium may provide rough estimations of metabolite build up in a vessel, but oftentimes metabolites accumulate to detrimental levels way before the medium turns yellow, with cell health and growth being hindered by amino-acid deficiencies. This will only become apparent after weeks of tedious pipetting when the cells are analyzed, and if this happens in the same run, researchers will be left none the wiser about why part of their experiment failed.

A solution to this problem is affordable online measurement tools that are intuitive, account for common problems like calibration and drift, and collect data into central, usable formats. But these don’t currently exist.

At MFX we’re developing the Cyto Engine™ – automated cell research and manufacturing platforms with online media and cell measurements. We’re working with the best sensor companies in the industry to integrate their technology and making it easy to account for drift and calibration. Data is aggregated and available from any device at any time, and our partnerships with the best visualization and data mining software providers means the Cyto Engine™ give real, valuable insights into cell and process data and maximize the information that each experiment provides.

New technologies, old methods- Why 99% of cell culture vessels will never scale

“We’re entering a new frontier in medical innovation with the ability to reprogram a patient’s own cells to attack a deadly cancer.”

That’s what Scott Gottlieb, commissioner of the FDA said when the first cell therapy Kymriah® was approved in the US in 2017. Since then, only 23 cell and gene therapy products have been approved by the FDA. Why are so few of these life-changing therapies reaching the market? And when they do, why are they so hard to access? Sure, these new technologies require stringent testing through clinical trials before approval. But one of the biggest issues restricting access to approved therapies- the expensive price tag- is largely due to the inadequate manufacturing methods currently available.

So why can’t existing manufacturing methods scale-up? Let’s look at the development process.

Research typically starts with a multi-well plate. These are good because they’re cheap, don’t consume a lot of reagent and cells, and have a small footprint. But the process is largely manual, which brings in a host of issues from manual handling and makes it hard to monitor what is going on in the wells. And soon, something with a larger surface area is needed to expand the cells in. A T-flask for example.

It may seem obvious to say that a flask has a very different geometry to a multi-well plate, but this change of vessel means cells will be exposed to a different environment. Nutrients, metabolites, paracrine factors, and gasses will all diffuse differently, while seeding, agitation and harvest methods will introduce new shear stress and temperature fluctuations. All these factors will ultimately have an effect on cell density, viability, and phenotype. And now, adjusting the process to redirect cell phenotype towards their original state is going to require more space and reagents.

Once pre-clinical testing is done, it’s time to scale-up for Phase I using a stirred tank bioreactor, culture bag, or even an automated platform that genty rocks cells. All the factors in the previous jump from plates to flasks are multiplied here, so cells are going to look COMPLETELY different. Although there’s more real-time monitoring methods available with bioreactors, even more time, effort, and money will need to be spent optimizing expansion.

“We talked to over 100 cell therapy developers during our development of the Cyto Engine and on average, they will spend 1-2 years translating their process from a research scale (flask) to a manufacturing scale (stirred tank, bag or otherwise)” says Antoine Espinet, CEO & Co-Founder of MicrofluidX, “It really got us thinking –  how many patients could have been treated in that time and how much money is being spent on all this?”

There is huge unmet need for truly scalable cell culture platforms that can translate cell and gene therapies from research to manufacturing scale, keeping consistency of the final product throughout. That’s why we’re developing the Cyto Engine, combining scalable bioprocessing, complete integration of online PATs, and powerful data visualization and analysis.