In 2018 there were a total of 906 regenerative medicine companies in operation globally, operating over 1000 active clinical trials, 70% of which deployed vectors in the products under investigation.  Along with AAV, lentivector is a common vector of choice in FDA approved gene and cell therapy products. In 2018 two FDA approved CAR-T cancer cell therapy products that utilize lentivectors in their manufacture achieved expanded market approval, and one gene therapy based on an AAV vector achieved a first market approval in the US, leading to increased demand for the manufacturing of these vectors.
The current vector manufacturing capacity crunch
Novartis’ Kymriah CAR-T therapy is a prime example of an FDA and EMA approved personalized cell therapy product manufactured using lentivectors. Edison Investment Research has forecast that under current manufacturing conditions Oxford Biomedica will supply Novartis with lentivector batches priced at $1.5M per batch with an independent analysis of the cost of goods predicting that the lentiviral vector component per dose of Kymriah accounts for $5,000, indicating that under 300 patients can have their treatment manufactured using a single batch of lentivector.   Given the incidence of the approved indications for Kymriah in the US and EU this equates to at least 20 batches of lentivirus supplied to Novartis per year for a single cell therapy product. Gilead’s approved CAR-T, Yescarta, also uses lentivector during manufacture and is produced by a non-disclosed contract manufacturing organization (CMO). Supply demands will be similar.
In addition, excluding approved products, given current global clinical trial enrollment targets,  there is a need for production at scale, complying with current good manufacturing practice (CGMP), of 42,000 doses (Assuming that 70% of the 59,575 total targeted enrollment of patients in current - 2018 year end - regenerative medicine clinical trials worldwide require a vector to be used in the manufacture of their therapy) of vector to meet current trial demand. Cell and gene therapy vectors, however, are not limited to lentivirus; AAV, adenovirus, plasmids, transposable elements, and other non-viral approaches to delivery such as nanoparticles, electroporation and microfluidic devices are also of growing importance. Vector selection is application specific and must be made on the basis of cost/benefit and feasibility analysis. For example, virus-free microfluidic approaches may become cost effective for ex vivo cell therapies, however for the delivery of gene therapies inside the body, it is not a suitable approach.
Challenges and progress in viral vector bioprocessing
Gene and cell therapy companies typically outsource the production of viral vectors to contract manufacturing organization (CMOs), however there are only a handful of these companies available with the expertise and capacity needed to supply CGMP grade vectors at scale. This is known as the vector manufacturing capacity crunch. Certain gene therapy companies have taken vector manufacture in-house in response. Given demand for vectors, this also provides them with a commercial CMO revenue stream before the achievement of their own market approved products.
RELATED ARTICLE: Is cell therapy manufacturing a supply chain challenge at its core?
Vector manufacturing for cell and gene therapies is a significant portion of the overall production costs. With viral vectors economies of scale can be realized by switching production from 2D cell factories, or flatbed bioreactors to suspension culture bioreactors at scales on the order of 200L+ per batch. Suspension culture can generate 15-fold greater yield than adherent at CGMP 200L scale. This depends on the availability of vector production optimized suspension culturable cells such as HEK293T utilized for lentivirus, the historically popular adenovirus vector, and some AAV serotypes. The suitability of HEK293T is not limited to these vector types.
Technical and strategic issues in bioprocessing and CMC controls
Other considerations for viral vector manufacturers include whether or not to generate stable producer cell lines. Transient transfection systems allow immediate and rapid production, which is important if a client is in the process development phase with rapidly changing requirements and small order requirements, however transient transfection does not scale economically, and is inherently variable, two big negatives for CGMP manufacture of clinical grade product. Stable producer cell lines are ideal here with a potential for 5-fold yield enhancement,  however large master cell banks are a requirement, and measures must be taken to generate new banks with similar or higher titre producer lines as master bank stocks run out, if they are to supply an approved product with indefinite supply requirements. This includes meeting comparability standards.
A role for automation and machine learning
Manufacturing is essentially automating an optimized fixed process to generate quality product reproducibly, therefore the role of robotics, online monitoring, and process analytical technologies (PAT) that allow for real-time monitoring of bioreactor conditions apply both to process development and quality assurance during final manufacture.
RELATED ARTICLE: Cell therapy process development: why and when?
The ability to monitor bioreactors inline and collect data wirelessly without direct sampling and operator intervention provides a significant cost savings. Inline monitoring also allows a system to remain closed, preventing contamination and fluctuations in temperature and pH. Cell viability is critical for cell therapy. Realtime inline monitoring populations in bioreactors using a flowcytometry method. Data from such outputs can be monitored by software algorithms that make real-time adjustments to keep bioreactor conditions optimized for cell viability. Automatic data collection also assists in monitoring batch-to-batch comparability, for which software can synthesize data from an entire manufacturing process. Virus titre determination by qPCR and ELISA can also be automated, reducing variation and standardizing assay performance.
The input of data scientists is critical during the design of automated data collection systems and development of appropriate algorithms. Consideration of data servers and information security is also important. Critical quality attributes for vector manufacture can be discovered using machine learning approaches. It can be used to optimize or monitor processes where real-time data is generated and compared to historic datasets. This could be particularly useful for making process adjustments with minimal impact on comparability. Recently major collaborations have been enacted between big data analytical firms such as Microsoft and leading vector manufacturing companies aiming to improve process operations and ultimately yields through analysis and input of historical batch manufacturing datasets to machine learning algorithms. 
Automation of stable cell line generation can significantly reduce timelines and increase screening capacity, improving from manual methods taking approximately 12 weeks for up to 200 clones, to automation of 8 weeks for up to 3000 clones, in the case of lentivirus. This automation includes robotic liquid handers, cell imagers, and specialized software to control the robotics and process the experimental data to select high yield clones.
The future of lentiviral vectors
The choice between lentivectors and AAV is application specific with AAV requiring fewer safety measures and lighter regulatory monitoring. Although in a clinical setting the phenomenon has never been observed, and lentivectors are designed so that under normal circumstances generation of replication competent vectors is impossible, experience with lentivector therapies in the clinic is limited and the FDA requires proof that products produced with lentivectors have no replication competent particles associated with them before lot release. The FDA also recommends, but does not enforce that patients are monitored for replicating lentivirus yearly for 15 years after treatment. Lentiviral vectors are also potential mutagens. By contrast AAV vectors have no such FDA requirements or and fewer safety concerns, however lentivectors have certain efficacy advantages over AAV including significant transgene capacity, 10kb vs 4.5kb with AAV, the ability to permanently modify dividing cells, such as T-cells or stem cells and no pre-existing immunity.
As clinical experience accumulates the FDA may loosen its regulatory requirements for lentivectors as third-generation ‘minimal’ lentiviral vectors, which contain only approximately 10% of the wild type virus genome, have key safety features such as starter cultures generated using 4 separate plasmids, minimizing the chances of generating replication competent virus through recombination. In addition, the delivery platforms often incorporate two minimal vector systems, one based on the equine infectious anaemia virus (EIAV), which cannot infect human cells, and the other being the human immunodeficiency virus-1 (HIV-1), which is the basis of lentivector efficacy in human cells. Recombination of these elements would yield non-infectious particles.
The Transgene Repression in vector Production (TRiP) system is a recent innovation that increases specific lentivector yields and particle purity. The TRiP system enables production of toxic transgenes as well as increasing the yields of non-toxic payloads. For example, the use to TRiP can increase yields of CAR-T lentivectors by 30-fold from a given bioreactor batch, resulting in reduced batch cost per patient. 
Optimised vector production systems in response to the viral vector capacity crunch
A present the answer to the viral vector capacity crunch is to improve the efficiency of existing manufacturing processes such as enabling bioreactor suspension cultures scale up through further cell line development and optimization, vector innovations that improve yield but also transgene expression efficiency once inside the target cell type such as with codon optimized promoters, and enhancer elements. This may be another area where machine learning could be particularly useful, as the potency of these elements is often guided by structure which is difficult to infer from sequence alone. Machine learning is already more accurate at protein structure prediction than trained structural biologists in head to head challenges. Automation provides new entrants to the vector CMO space, to rapidly compete with the pioneers in the space. As supply increases and automation brings production costs down the capacity crunch will ease.
Promising emerging non-viral approaches
Looking longer term viral vectors are here to stay, 70% of vectors in clinical trials are of viral origin, however the original viral element may be small. The most advanced vectors are already 90% synthetic. This is likely to increase further. Companies are developing fully synthetic targeted nanoparticles for drug delivery, and this can include plasmid encoded genes, RNA, transposons, or gene editing protein complexes such as TALEN, CRISPR for example. The advantage of such systems is the freedom of cell targeting that is provided from the design perspective.
Technologies such as electroporation and microfluidic devices that create transient entry pathways for drugs or genes to enter a cell without the need for the vector itself to induce its internalization potentially make vector manufacture as simple and cheap as chemically synthesizing the transgene as one would at R&D scale. This is particularly useful for personalized treatment of rare genetic disorders where one device can deliver different DNA sequences for multiple conditions, thus the value comes from the device rather than the vector. Each rare disease would no longer be “niche” but part of a single rare disease market served by a competing set of DNA transfer devices. It is worth noting that these drug delivery technologies have already been transferred out of research labs and into pioneering start-ups, some of which have already received investment from large pharmaceutical companies.
The Sleeping Beauty transposon is a notable new non-viral type of vector to have entered gene and cell therapy clinical trials. DNA transposons are native DNA elements that are able to change their position within the genome by virtue to the transposase enzyme. By harnessing the transposase gene and necessary inverted repeats in bi-component synthetic plasmids it is possible to transfer genes of interest to the genome from the transfected plasmid. This offers a promising non-viral mode of genome manipulation that can be delivered by nanoparticles, electroporation or microfluidic devices. 
In summary, there is a pressing need for vector production for approved cell and gene therapies as well as some 42,000 doses for planned or operational clinical trials. This need is currently mostly met be viral vector production which is being optimized through process optimization aided by big data. Viral vectors are here to stay in the gene and cell therapy industry, however promising alternatives for delivering non-viral plasmid DNA are under development such as new electroporation devices and microfluidic systems.
By David Orchard-Webb, PhD, Freelance Consultant and Medical/Biotech Writer
- Lambert J. Introduction and Industry Update. Alliance for Regenerative Medicine’s 9th Annual Cell & Gene Therapies State of the Industry Briefing. January 2019. https://alliancerm.org/ arm-event/sotibriefing/
- Wilkinson D., and Jana S. Oxford BioMedica Corporate update; Validation achieved, growth expected. Edison Investment Research Limited. May 2018. https:// www.edisoninvestmentresearch.com/? ACT=18&ID=21055&LANG=
- Kleutghen P., et al. Drugs Don’t Work If People Can’t Afford Them: The High Price Of Tisagenlecleucel; ESTIMATES OF PATIENT DEMAND FOR KYMRIAH. Health Affairs. 2018;Blog: DOI: 10.1377/hblog20180205.292531
- Jones P. Key engineering challenges in the biomanufacturing of lentiviral viral vectors. In Advancing Manufacture of Cell and Gene Therapies VI. Dolores Baksh, GE Healthcare, USA Rod Rietze, Novartis, USA Ivan Wall, Aston University, United Kingdom Eds, ECI Symposium Series, 2019; http://dc.engconfintl.org/ cell_gene_therapies_vi/89
- Terry M. Oxford Biomedica Teams With Microsoft to Use AI to Improve Gene Therapies. BioSpace. March 2019; https:// www.biospace.com/article/oxford-biomedica-teams-withmicrosoft- to-use-ai-to-improve-gene-therapies/
- Hudecek M. et al. Going non-viral: the Sleeping Beauty transposon system breaks on through to the clinical side. Critical Reviews in Biochemistry and Molecular Biology. 2017;52: 355– 380. https://doi.org/10.1080/10409238.2017.1304354