By William R (Bill) Strohl, BiStro Biotech Consulting LLC,  7-25-17

This is the first of several blogs that I will be writing for The Antibody Society on topics in the field of therapeutic antibodies and proteins that I find of interest.  The intention of these blogs is to stimulate thought and discussion about various aspects of therapeutic antibodies and related molecules, something that anyone reading on this website ought to appreciate.  For my first blog, I thought I would tackle an age-old (at least since the 1990s) discussion point concerning the use of transgenic mice producing human antibodies vs panning human libraries as preferred sources for fully human antibodies.  Virtually everyone with whom I have had the opportunity to discuss this topic has an opinion on it, some of which are rather strong.  That should make this all the more fun!  Moreover, 2017 appears, at least thus far, to be the year of the fully human antibody.  Thus far (to 7/23/17), 23 innovative drugs have been approved by the US Food and Drug Administration (FDA), seven of which are monoclonal antibodies.  Of those, six are fully human antibodies!

I’ll start out by reminiscing about antibody meetings in the timeframe of around 2000-2001, before any fully human antibodies were approved for marketing and commercial use.  In those days, scientists from the big four, i.e., Medarex and Abgenix on the transgenic mouse side, and Cambridge Antibody Technology (CaT) and Morphosys on the human antibody library side, would give talks at meetings, often back to back, on their platforms.  Of course, as would be expected, each group would expound on the positives for their particular approach, while finding negatives, or minimally providing neutral comments, for the others.  Coffee breaks and lunches at those meetings were typically filled with discussion about which approach, transgenic mouse or human antibody library panning, was better.  It was a fun time, watching, listening to, and participating in those discussions.  I’m sure that in many venues, vigorous discussions along the line of “Mouse vs Pan” still continue.

I will say up front that I don’t particularly have a favorite, because I see the potential use of both approaches.  Anecdotally, over my career I have been involved with many programs that were sourced from Balb/c mice and then humanized, transgenic mice producing human antibodies, and human antibody libraries, and I can’t say categorically that I’ve seen that a trend for “better antibodies” coming from one source or another.  It turns out that one of the worst behaved antibodies we made actually came from a Balb/c mouse, and it took intense phage-based engineering to improve the antibody into one that could potentially be developed.

First, though, to the statistics:  we all know that the first “fully human” antibody to be approved for commercial use, and the most valuable antibody on the market today, is adalimumab, marketed by Abbvie under the name of Humira®.  Approved by the US FDA in December, 2002, Humira® was used to treat nearly a million patients in 2016 and generated over US $16 billion in gross revenues.  Adalimumab was sourced from the CaT human antibody library using a technology called guided selection.

With the very recent approval of Janssen R&D’s guselkumab, there are now 75 antibodies and Fc fusion proteins approved by major regulatory agencies, according to my database.  Of these, 25 (one-third!) are fully human antibodies, 18 of which were derived from transgenic mice producing human antibodies and seven derived from human antibody libraries.  Additionally, there are currently 69 antibodies and Fc fusion proteins in Phase III clinical trials, 20 of which are fully human antibodies.  Of these 20 late clinical stage human antibody candidates, 10 are from transgenic mice, six were derived from human antibody libraries, and four are sourced directly from human B cells (i.e., the new kid on the block).  Combining the approved and Phase III numbers leaves us with 28 approved or late stage human antibodies sourced from transgenic mice, 13 sourced from human antibody libraries, and another four from human B cells.

From the data above, it is clear that transgenic mice have played a much larger role to date than antibody libraries in sourcing successful candidates that have either reached approval stage or late stage clinical trials.  That may or may not tell the whole story, however, as target selection, financing, collaborations, and clinical expertise of the four innovative companies mentioned above, dating back to the pre-2005 timeframe also would have a huge impact on those programs that ultimately were successful and represented by these data.  I do believe, however, that the overall numbers don’t lie, and that for many targets, including soluble cytokines or other serum proteins and single pass receptors with prominent exodomains, the combination of which make up the vast majority of antibody targets, immunization of a transgenic animal should lead rapidly to high quality, high affinity antibodies.  When it comes to isolating epitopes, certain infectious disease targets, and multi-pass membrane proteins, however, often times human antibody library display approaches may have an edge to obtain the right antibody to develop.

Historically, there has been a preconception, largely supported by various data, that antibodies derived from animals or mammalian cells were generally “better behaved” than antibodies derived from human antibody libraries (Jain et al., 2017).  Many papers have been written supporting various aspects of this argument, including the case against hydrophobic patches found in library-based antibodies, the case for phage-based sequence bias, the lack of mammalian cell “editing”, the need for in vitro affinity maturation, and so forth.  Some of these arguments may apply to some library-sourced antibodies, but likely not to all of them.  Many human antibody library antibodies in fact are derived from human antibody genes using PCR-based approaches.  Depending on the maturation level of the recovered variable sequences, one would expect that many of these structures might be more “animal like”, even if they have not gone through substantial in vivo editing.  Moreover, more and more library-based approaches are using eukaryotic cells (e.g., yeast, mammalian cells) for expression and display of their antibodies, which should improve the selection for antibodies with improved expression, folding, and solubility characteristics.

There are now more than a dozen transgenic animal platforms available today, ranging from the original Medarex HuMab mouse and the Abgenix Xenomouse, to transgenic rats, rabbits, chickens, and even cows producing human antibodies, so there is no shortage of potential transgenic animal platforms from which to derive human antibodies (Strohl, 2017).  Moreover, most companies using these approaches have moved away from low yield hybridomas to high yield and throughput B cell cloning approaches, substantially increasing the ability to obtain high quality clones from animals.

Likewise, for displayed antibodies, the field has progressed significantly from phage displayed scFv or Fab human antibody libraries from CaT and Morphosys, respectively, to libraries generated and panned in yeast, mammalian cells, and most recently, in recombinant mammalian cells in which the antibodies are matured “in cell”.  Thus, the range of display options, combined with FACS sorting and analysis, next generation sequencing, and tagging technologies, have significantly increased the ability to obtain large numbers of high quality antibodies from human antibody libraries.

The critical quality attributes of a human antibody panel are: (i) the ability to bind and neutralize key “functional” epitopes; (ii) high affinity and selectivity; (iii) proper biophysical properties (e.g., “well-behaved”, soluble, not aggregation-prone, biochemical stability); (iv) excellent expression as full-length antibodies in manufacturing cell lines; and (v) lack of immunogenicity when dosed in humans.  While virtually any of the approaches discussed above can ultimately end up with excellent attributes in each category, it is likely that, for most targets, antibodies sourced from transgenic animals will be more likely to achieve all the attributes more easily and quickly than those sourced from traditional human antibody libraries.  This may evolve, however, as improved cell-based maturation and panning approaches, such the HuTARG™ technology developed by Innovative Targeting Solutions in Vancouver, which couples mammalian display with cell-based V(D)J recombination and FACS sorting, may revolutionize how fully human antibody variable sequences are sourced in the future.

References:

Strohl WR.  2017.  Chapter 5.  Human antibody discovery platforms, pp. 115-160.  Protein Therapeutics, 2 Volume Set.  T. Vaughan, J. Osbourn, B. Jallal, R. Mannhold, G. Folkers, H. Buschmann, eds.  Wiley.  ISBN: 978-3-527-34086-6.

Jain T, Sun T, Durand S, Hall A, Houston NR, Nett JH, Sharkley B, Bobrowitz B, Caffry I, et al.  2017.  Biophysical properties of the clinical-stage antibody landscape.  Proc. Nat’l. Acad. Sci. USA 114:944-949.

Happy antibody hunting…

Bill Strohl, www.bistrobiotech.com