INTRODUCTION BY BIOGEN’S ANDY WEISKOPF AT THE CGT CMC STRATEGY FORUM

The following are the introductory comments by Biogen Regulatory Affairs Director Andy Weiskopf at the CASSS 2017 CMC Strategy Forum on the “Manufacturing, Quality and Regulatory Considerations for Cell and Gene Therapies.” The timing of the forum, the forum co-chair explained, reflected the “critical mass of activity happening in the development, manufacture and quality assurance of these products.”

Just a couple of brief comments before we begin about the topic for this year’s summer forum, “Manufacturing, Quality and Regulatory Considerations for Cell and Gene Therapies.” Now for some of you, who are familiar with CASSS, you probably associate the CASSS family programs with protein biologics. But the fact is CASSS has been committed to developing, programming and fostering dialogue for all modalities of biologic products, whether they be vaccines, cell therapies, or gene therapies.

I am going to ask you to climb into a way-back machine back to January of 2005 to kind of set the scene here. This was the idea of a technologically advanced cell phone at the time and this, of course, was the time when Donald Trump was still a reality TV show star.

In January of that year, the CMC Strategy Forum held its first program on a topic pertaining to advanced therapy medicinal products. It was a one-day program, a lot release and characterization testing of a live virus vaccine and gene therapy products. Then, three years later, there was another one-day forum devoted to analytical characterization of gene therapy. So this is something that certainly the CMC strategy forum has been touching upon throughout the years and throughout other CASSS meetings – particularly the bioassay meeting and WCBP – and has always been ingrained into the dialogue that we like to have.

Obviously, a lot has changed since 2005. Just to take a look at what has happened in this space. This is not even counting the hundreds of INDs and programs that are still in clinical development. This is not counting tissue-based products that have been developed and approved.

I think the thing that is most striking about the milestones that you see on this slide is how many of these have happened just in the past two years: ● Imlygic, the first approved oncolytic gene therapy for cancer ● Strimvelis in May 2016, a cell therapy for ADA skit. And then of course this year, three milestone BLAs: ● one from Spark for inherited retinal disease, and then ● the two CAR-T programs from Kite and most notably Novartis, of course. Those of you have not been living in a cave last week heard the news about the successful advisory committee outcome. For those who have been involved in biologics product development for a long time, there is a sense of deja vu all over again. This kind of slow trickle that is now accelerating into increased successes with this modality feels an awful lot like where we were 20 to 25 years ago with protein biologics. That is why having you here for the next two days and having this meeting comes at such a great moment. Because these are not one-off success stories. There is a real critical mass of activity happening in the development, manufacture and quality assurance of these products, that the time has come for us to have a conversation. [Celgene Biologics Executive Director Sid Advant], [CASSS Executive Director Stephanie Flores] and I would like to recognize the members of our scientific organizing committee who you see listed on the bottom half of this slide.

These are the folks who really helped us keep on target with the topics that you are going to be seeing that you are going to see covered over the next two days, helped us get in touch with the right speakers, made sure we are asking the right questions. We want to thank them. During the breaks, if you see these folks and you are really enjoying the program, please go up to them and thank them. As I noted, we were very excited to see the final count of folks who registered: ● 168 attendees pre-registered, which is a record for the CMC strategy forum, and I think reflects the excitement of the enthusiasm in this field ● 14 regulators preregistered ● 65 different companies represented, and ● we have representatives from eight different countries.

NIST’S ANNE PLANT ON MEASUREMENT ASSURANCE FOR RMATS

At the summer 2017 CASSS cell/gene therapy CMC strategy forum, NIST Division Chief Anne Plant highlighted the challenges of product characterization and measurement science in the RMAT arena and the initiatives to address them that are underway at NIST, and through the public/private partnerships with whom it is engaged. The normal disclaimer that the presentation represents the views of the speaker and not necessarily that of his/her organization is not included.

I want to start by saying that I am going to show a bunch of data in here that is the work of a lot of different people at NIST – mostly in our division, which is the biosystems and biomaterials division, but then also from the information technology laboratory.

Andy, thank you so very much for giving us a good overview of how fast this field is changing…. One of the things that is so exciting is that it has been a long time coming and really things are just exploding in this field as I am sure that most of you being here are well aware.

It is really an exciting time, and it is also a time that is disruptive in a lot of ways, because we are talking about therapies now that are curing, and not just treating diseases. But then also, the field has been expanding. If we talk about regenerative medicine in general – I am drawing on some of the material that is compiled by the Alliance for Regenerative Medicine – this field of regenerative medicine really includes a whole lot of entities.

People are starting to talk about this as an ecosystem of capabilities. The Alliance for Regenerative Medicine includes hundreds of companies that both create therapies, but then also support those productions with ancillary materials and equipment and instrumentation. It is a very complex space with a lot going on.

Challenges of Product Characterization and Measurement Science But of course, similar to the protein therapeutics world, the most important thing is, how do you characterize your product? Everybody is familiar with the idea of quality attributes, and that is one of the things that this implementation strategy has in common with protein therapeutics. There are a lot of things in common, which really are helping this industry develop. But there are some things that are more challenging or a little orthogonal to the common ideas in protein therapeutics. While we have this list of quality attributes, and everybody can attest to those and sort of understand them, in the case of cell-based therapy products, when you are actually talking about using cells, there is a lot of challenges that are a little different from just using cells as the manufacturing conduit for making a protein. First of all, these therapies can be very diverse in terms of their characteristics of what they are supposed to do. Some of the quality attributes, like identity, or like biological activity, are pretty tough to get your hands around because, in fact, for many of these products the exact mechanism of action is not really well understood. In a protein, you might have a target that you understand. In this kind of field, some of these products may be doing very diffuse things, and you are not really exactly sure of what they are doing, or even in the case of CAR-T, for example, you know what they are doing, but you don’t really know what are the biological characteristics of the product that are going to be the most successful clinically. Ferreting that out is a source some consternation. So the mechanism of action is a big gap in this area. Assessing the qualities that are essential for good clinical response is really challenging. Of course, characterization of product is really essential on a number of different levels. Knowing the quality attributes that make for a good product is really important, but it is also has a lot of impact in the manufacturing process and how to deal with the manufacturing process. Many of these therapies are still at the stage of preclinical development or very early stage clinical development. And so the method of manufacturing for a small clinical trial or in the discovery process is going to be really different from what you have to do further down the line when you start thinking about having a really marketable product.

That makes the characterization of product really essential. Because there is great regulatory implications of changing your manufacturing process. If you do not have really good measurements and the ability to trace your measurements back to what they were at some earlier time when you thought you understood your product characteristics, if you do not have those measurement when you get down here, then you do not really know how to characterize that product or what might have changed when the manufacturing process itself changed. If do you do not have that connection, that traceability, you basically can go back to clinical trials. So that is really important and the reality is that with these products you can just guarantee that there are going to be changes in the manufacturing process as you go forward. One of the things that has been fun to watch in this field as it has progressed over the past five to eight years, is this idea that not only do you have to have a product that has some characteristics and some clinical responsiveness, but you also have to be able to manufacture it consistently and make a profit – get reimbursed and make money off of it. Well, once that thought process really started to get into peoples’ heads, a lot of things started changing in terms of the sophistication of how we think about these processes. And also, the fact that you really have to think now about cost of goods – at the beginning, people didn’t care about that so much, but this is a huge issue. Because the cost of goods is large and perhaps supply is uncertain, you can just imagine that there are going to be lots of changes over time. And you really have to think about how to be prepared for that. In our speak – in a measurement laboratory like NIST – we think about comparability as being a key word for what you have to understand about your measurements so that you can compare your measurements today with your measurements sometime in the future. One of the ways that we do this, particularly at a place like NIST, is that we make reference materials – that is one of the things that NIST does. You can put a lot of chemistry in bottles, but it is really hard to put biological activity in a bottle or impossible really. Even if you think you have an RPE preparation that you feel really confident about, how do you translate those characteristics to something traceable to the next product that you have? It is very difficult. These are real big challenges. Complex Biology Part of the challenge is the biology is just complex – we are dealing with really complex things.

● So for me, one of the biggest questions when I think about this is what are the in vitro metrics that you prediction of the in vivo response? This is a big black box in many applications. ● Part of that is just due to the gaps in fundamental understanding and the fact that the biology is complex. There is not a linear response from point A to point B in the body. ● There are lots of variables – lots of analytical variables, and, of course, lots of clinical variables. ● One of the conundrums, particularly in manufacturing, is that many of the products are autologous in nature. In other words, the starting material is a patient’s cells. Given the fact that every patient is going to be different, and every patient is probably going to be different from a normal healthy individual, you might have to adjust manufacturing processes in order to accommodate those differences in the patient sample. Maybe some patient cells are not as robust to a process or maybe they do not grow at the same rate, so your manufacturing processes are going to change. Those are huge variables that have to be thought about up front. ● Of course, these materials have a dynamic nature. ● And as in most things in biology, there is really no ground truth to base a reference material on. So reference materials are really challenging in this space.

Measurement Science At NIST, we try to engage practitioners in a conversation – not immediately about reference materials – but rather about measurement assurance. When I use the word measurement assurance, basically, I am talking about what are those pieces of data that you have to have surrounding your numbers that give you confidence in those data and how to interpret those data. Because in the end, you need that confidence in order to make a competent decision. There are some really fundamental issues around the measurement science that are quite intriguing for those of us who do measurement science every day. Because even asking what is the measurand? Are you measuring a cell? Are you measuring a population of cells? Are you measuring the presence of a cell surface receptor? Are you measuring the numbers of cell surface receptors on any particular cell? Are you measuring the distribution of cell surface receptors on cells across the population? When you start thinking about it, it gets really complicated. Then again, you have all the other things that you have to worry about, like assay variables etc., that you worry about under all conditions and robustness of the measurement. One of the other things that comes up is this idea of dispersion of biological variability. And that idea that you are mostly working with a distribution of responses all the time just adds another element of how do we even measure this – what should we be measuring, what should we be looking at? Of course, you cannot know biological variability unless you really have a good handle on the technical measurement variability of the assay. The bottom line is that anything that you measure, you are hoping is going to be biologically relevant, because that is why you are spending the time getting a good assay to measure it. But that is something that a lot of times you can’t know until you have really good validated measurements that you can correlate with clinical responsiveness. So it is kind of a chicken and egg thing. These ideas about how you qualify and validate your measurements: These are things that probably everybody in the room is very familiar with – precision, accuracy, robustness, characteristics, sensitivity and specificity of the assay, and understanding what the dynamic range and response function of the assays are. These are all standard issues that everybody deals with in any kind of critical measurement that you make. And of course, they are important from the preclinical stage right up to the product.

Unfortunately, one of the things that has happened in the cell therapy area, I think maybe because it has been such an academic endeavor for so long, is that FDA guidance that says you do not really have to demonstrate a validated assay until Phase III clinical trials, has sort of left people thinking that they do not really have to validate their assays until they get there, which of course is not a good way of thinking about it. You certainly want to have a really good basis of confidence in your measurements and what you are measuring at the preclinical stage when it is actually relatively easy, so that you can then have a good basis for correlating when you get to clinical responsiveness.

What NIST is Focused On I want to give you a couple of things of what we are trying to do here. Let me go back and just say this, that in order to sort of get to here – in this very complicated space – one of the ways we deal with that at NIST, is that we try to think what are pre-competitive technologies, what are the fundamental things that most everybody -- either tools that they are using or things that they are trying to measure – that we can think about being more robust and try to make them more robust. So, I am going to give you some examples of that. One we have had for a long time, is a program in flow cytometry. We are inching our way, and I think the flow cytometry community is inching its way, toward working flow cytometry into being a technique that is actually quantitative. I think that anybody who uses flow cytometry is sensitive to the fact that if you are really good, you can get somewhat consistent data on a single platform in your laboratory, but every time that you go to a difference instrument or god forbid, that you go to a different laboratory, or you try to compare data from different laboratories, often there is a very poor basis of comparability. So, one the things that NIST has done is that we have a series of referenced fluorophores that are served by for amount of florescent molecule in the solution. But now we are looking at using these things to help calibrate and help provide comparability between instruments.

Everybody uses calibration beads. But every manufacturer’s calibration beads are different. They are not comparable with one another, as it turns out. What we are working on now through a consortium of bead manufacturers is to trace these beads properties to this reference material, so that beads made by one manufacturer should be now comparable to beads made by another manufacturer. You should be able to draw that traceable link. This is an early first step. A lot of work is yet to be done.

One of the other things in flow cytometry is producing this in an international effort with NIBSC and others to produce to lyophilized cells that have known numbers of receptors on them, and this can be used as a traceable material.

One of the other things that NIST spends time on is design of experiment. This is an example of an MTS assay, a metabolic activity assay, across different laboratories. This was an international effort that after design of experiment, there was a lot of concordance. For this lab that was out of spec, one could understand why by looking back at the protocol.

But to get from here, where everything was completely different, to get to here, where now you could say that you could produce an LD 50 and it sort of meant something relative to somebodies else’s LD 50, took an awful lot of work. All of these spots on this plate are different experiments. Most of them are controls, and very few of them are actually sample. But that is what it takes to get there.

One of the things that surprised us a lot is when we started talking about companies who were thinking about cell therapies was something very fundamental, which is cell counting and how difficult that is. I have talked with folks at a major protein therapy manufacturing facility, who that said that when they were just counting CHO cells, they knew what they were doing. When they started thinking about counting lymphocytes or some other cell therapy product, all of a sudden, they realized they did not know how to count cells.

So, the idea that you get discrepancies between methods is clearly unacceptable, because you are left saying to yourself, ‘well what is the right answer?’ And you do not want to be off by 10, 20 or 30%.

One of the things that NIST has done in collaboration with Lonza and with our information technology laboratory was to develop a statistical model that is based on an assumption that there is going to be a linear relationship between dilution and numbers of cells counted, and a very stringent statistical design for dilution, in order to be able to evaluate statistically one method compared to another method in terms of accuracy and precision.

This is an example of using experimental design and statistical analysis to get around the problem of having a reference material, because really it is very hard to think about what is going to reference material that is going to applicable for all kinds of cells, and all kinds of instruments, and applicable under all conditions. So a method like this gives you confidence without providing something that is absolutely knowable, which is of course impossible.

One of the other things that you would like to think was knowable, but it turns out is impossible, is we spent a lot of time on imaging and trying to understand how we might use machine learning to pull out features of images in order to give us ideas about evaluating characteristics that would be informative of clinical responsiveness for example for release assays.

One of the things that after we started doing this with some expert laboratories, we realized that experts are not reliable. Experts are not perfect at classifying and you cannot make a robust classification algorithm based on an expert or even two experts probably. So how do you deal with this stuff. It is really hard. This is something that is unpublished yet and we are still in the process of trying to figure that out.

But one of the things that we are also doing then is looking at technologies that we can develop that might help us address this problem of really what are the characteristics that are meaningful to measure, and can we assess the biomarkers that we are assuming are important. And so this is one kind of experiment where we looked at stem cell colonies. These are embryonic stem cells that have a GFP associated with the OCT 4 site – OCT 4 being a pluripotency marker.

In order to say what are the characteristics in the phase contrast image of a stem cell colony like this that might give us information about pluripotency, can we test a pluripotency marker that we think we know about with another pluripotency marker like the phase-contrast characteristics, image characteristics of these colonies, and can they be the same?

By doing time dependent imaging, we can ask what happens to this colony over time. If it started out as one thing, did it end being pluripotent in the end? How does it compare to other colonies? How do you do QC on a plate of colonies where you have a lot of diversity of colonies, many different colonies? How do you put numbers on these things?

One of the things with our ITL collaborators that we did was we took these data sets that were very large, about a terabyte of data altogether, and they developed for us methodologies for using technologies such as google maps in order to stitch all these images together, look at them all at one time, hone in on particular colonies that we might be interested in, and then be able to get data out of any place in this image that we were interested in.

One of the important things about this is to be able to generate visualization of data so that you can look at what is happening over time with respect to a number of different parameters about these colonies. There is a lot of potential for this kind of approach to say, ‘this is a goodlooking colony here and it seems to be pluripotent because it is expressing GFP, but it is certainly did not look great at the beginning,’ and you might have thrown it away at some earlier stage of the process.

One of the things that we noticed is not all colonies are producing GFP homogeneously and what do you with that information? Another thing that this analysis was really helpful for is allowing us to find colonies that we can see by phase contrasts that are not producing GFP that are not pluripotent. And those kinds identifying rare events is really critical. How else do you define rare events if you do not want a pluripotent cell in your preparation? For example, how do you know that you have eliminated all the pluripotent cells?

So these are fundamental problems. One of the things that this kind of analysis allows us to do is to say, ‘if we can do these kinds of measurements on a test system – understand what the probability is of events that we don’t want to have happen or events that we do want to have happen – then we have a better idea of how we have to sample, how much do we have to sample, in order to find those rare events.

Community Efforts to Address Problems

These things are really difficult. There is a lot of community activity going on. I just want to give you a snapshot of that. NIST has been very active in having workshops and developing consortia. In fact, some of the output of this is a consortium that is being developed right now on gene editing and the measurements that are needed for that. This is a public private consortium of course.

One of the other things that is happened is the 21st Century Cures Act. While this had a lot of impact on FDA processes, it also has required NIST and FDA to work with others to develop a coordinated response to prioritizing and developing standards.

The Standards Coordinating Body is a group that has come out of the Alliance for Regenerative Medicine. We have an MOU with them. I think they will be a very important element for helping to coordinate activities and standards.

Then, one of the other things that has happened – a lot has been going on – is the development of the public private partnerships. This one is funded by NIST. It is called NIMBL, the National Institute for Innovation in Manufacturing Biopharmaceuticals. It is a public/private partnership that is well funded by the federal government and by private industry to bring together different kinds of entities in the manufacturing of advanced therapies. A lot of collaboration and cooperation is going on here. I am sure that many of you in the room are aware and working on this.

And this is just my summary slide. This is how NIST interacts with others. We have our laboratory programs that are very highly integrated into industry engagement and engagement with other agencies, including FDA. A lot of this takes place within the standards leadership framework.