The hype of science

When original and interesting research is distorted to garner additional attention, both the work in question and previous studies can be shortchanged. Here, I will describe a recent and notable case from the journal Science in which the perceived novelty and importance of a study were significantly enhanced.

In the 20 March, 2009, issue of Science, researchers from Genentech (Bostrom et al.) show that it is possible to select a new specificity for a therapeutic antibody with an established specificity while maintaining the original specificity.¹ The authors started with an antibody (Herceptin) specific for the tumor antigen, HER2, in clinical use for patients suffering from breast cancer. After mutagenizing the gene encoding the light chain variable domain of the antibody, they were able to select a variant antibody that had gained the ability to bind another clinically-relevant target protein, vascular endothelial growth factor (VEGF). The new antibody bound both antigens with relatively high affinities. This paper is a technologically impressive, interesting, and potentially medically important study. However, the authors' claims for the novelty of their results are seriously overstated. Disappointingly, the reviewers, editors, and commentators involved with the paper all failed to recognize or acknowledge the extent to which their claims misrepresented the prior understanding in the field.

Bostrom et al. begin their abstract by claiming, "The interface between antibody and antigen is often depicted as a lock and key, suggesting that an antibody surface can accommodate only one antigen." This statement actually combines two different assertions: 1) the notion that antibody recognition is widely believed to conform to the metaphor of lock-and-key binding, in which the antibody combining site (or paratope) binds the antigen without significant structural adjustment, as opposed to adaptive fit binding, in which the paratope binds the antigen in conjunction with structural changes of varying extent, and 2) the idea that an antibody can bind one and only one antigen. While adaptive fit binding is generally regarded as more likely to be associated with the ability of a single antibody paratope to bind effectively to two or more antigens, lock-and-key binding is not necessarily mutually exclusive with binding effectively to two or more antigens. I will address these two issues separately.

The thesis that each antibody can bind one and only one antigen, or even only one antigen and closely related antigens, has been disputed -- one could even argue refuted -- for many years. Even leaving aside the so-called polyspecific antibodies of the primary antibody repertoire that were first identified and characterized decades ago (reviewed, for example, in Ternynck and Avrameas),² there is still an overwhelming body of evidence suggesting that it is reasonably common for a single antibody (from a secondary or later immune response) to bind multiple antigens even when some of these antigens are lacking in obvious structural similarity. This perspective dates back at least 50 years as articulated by David Talmage in Science:³

-- The formation of a stable complex does not require a perfect fit between two complementary configurations. The heterogeneity of the antibodies com binding with a single haptene and the re actions of the same antibody with different haptenes provide experimental evidence for this premise.
-- A single globulin molecule may combine with a large number of different substances in the same sense that a master key may open a large number of different locks. There should be many different antigenic configurations structurally suited to combine with the same globulin if a lack of perfection in the fit in one area may be compensated by an increased binding energy elsewhere.(pp. 1645-1646, format edited for clarity)

But, the belief that antibodies could be multi-specific was not based on mere speculation. Numerous studies have demonstrated the ability of a single antibody to bind effectively to diverse antigens. For example, consider studies with the murine monoclonal antibody (mAb), HGAC 39, specific for group A streptococci. In the late 1980s and the 1990s, we demonstrated that this mAb could bind to: 1) the cell wall polysaccharide of group A streptococcus and the immunodominant monosaccharide component thereof, N-acetyl-glucosamine,⁴ 2) cytoplasmic proteins bearing O-linked N-acetyl-glucosamine,⁵5 3) multiple L- and D-amino acid peptides,^6-7 and 4) several different monoclonal anti-idiotopes, some of which competed with the monosaccharide hapten, N-acetyl-glucosamine, for binding to the paratope.⁸ There have also been many papers other than my own showing that multiple monoclonal anti-idiotopes can bind to antibodies of known specificity for non-immunoglobulin proteins.

The authors' claim that "...the few examples of such multi-specificity are limited to antibodies that bind small haptens (4),..." is difficult to reconcile with papers from other investigators in highly visible journals that definitively demonstrated the possibility of a single antibody binding two or more proteins. The most impressive papers in this context were published over a decade ago.9-10 Another relevant report was published more than five years ago.¹¹

The authors' claim implying that antibody recognition was typically assumed to correspond to lock-and-key binding is also undermined by a number of previously published papers, again including several high-profile examples, including some in Science.^12-15 These publications demonstrate, with structural or biophysical data, what is often regarded as the alternative to lock-and-key binding, i.e. binding via adaptive fit.

The misperceptions engendered by the paper itself may be facilitated by the accompanying commentary by Parren and Burton.¹⁶ The one-sentence summary at the top right of the first page of the piece ("An antibody is engineered to recognize two different proteins with high affinity, opening the door to improved combination therapies for cancers and infections.") is hard to justify given the evidence provided by Bostrom et al. (which was understandably limited to mouse models). To their credit, Parren and Burton acknowledge possible complexities in the clinical application of the dual-specificity monoclonal antibody generated by the Genentech group. A recent commentary in Nature cites new studies assessing the complex effects of therapy with the anti-VEGF antibody, bevacizumab that only serve to emphasize the uncertainties about the ultimate clinical value of a dual-specificity antibody that targets VEGF and another antigen (i.e., HER-2).¹⁷ Genentech's HER-2/VEGF antibody may well prove to be useful, but it will take extensive clinical testing in patients with different diseases to figure out how useful, in which conditions, and with what other therapies. It also remains to be seen if this approach can be successfully used to develop other antibodies specific for selected pairs of therapeutic targets.

The claim by Parren and Burton that "the prevailing one antibody-one antigen dogma" was overthrown by Bostrom et al. is disappointing. While Parren and Burton presumably understand the subtle distinction between the view that each antibody (whether related to therapy or not) can only bind one antigen (which was effectively refuted years ago, as demonstrated above) and the clinically-focused and (until now) justified belief that each therapeutic antibody can only have one intended molecular target, the statement quoted above may be interpreted by readers without extensive knowledge of the relevant literature to mean that Bostrom et al. deserve credit for the overturning the former (and much more fundamental) thesis.

Unfortunately, this episode is not the first (or second or third) time that I have encountered such failures to properly portray the history of a field in Science or Nature with the effect, whatever the intent, of boosting the apparent significance of a newly published study. Such deviations from ideal scientific practice unfairly damage investigators whose work and ideas are not cited or taken into account, and they are especially disappointing when the most prestigious and visible journals are involved. Perfection in such matters may be unattainable, but surely it should be possible to do better than is the case in the March 20 issue of Science.

Neil Greenspan is an immunologist and professor of pathology at the Case Western Reserve University School of Medicine.

References:

1. Bostrom J, Yu SF, Kan D, Appleton BA, Lee CV, Billeci K, Man W, Peale F, Ross S, Wiesmann C, Fuh G. Variants of the antibody herceptin that interact with HER2 and VEGF at the antigen binding site. Science. 2009 Mar 20;323(5921):1610-4.

2. Ternynck T, Avrameas S. Murine natural monoclonal utoantibodies: a study of their polyspecificities and their affinities. Immunol Rev. 1986 Dec;94:99-112.

3. Talmage DW. Immunological specificity, unique combinations of selected natural globulins provide an alternative to the classical concept. Science. 1959
Jun 19;129(3364):1643-8.

4. Greenspan NS, Monafo WJ, Davie JM. Interaction of IgG3 anti-streptococcal group A carbohydrate (GAC) antibody with streptococcal group A vaccine: enhancing and inhibiting effects of anti-GAC, anti-isotypic, and anti-idiotypic antibodies. J Immunol. 1987 Jan 1;138(1):285-92.

5. Turner, J.R., Tartakoff, A.M., and Greenspan, N.S. Cytologic assessment of nuclear and cytoplasmic O linked N acetyl glucosamine distribution by using anti streptococcal monoclonal antibodies. Proc. Natl. Acad. Sci. USA 87:5608 5612, 1990.

6. Harris, S.L., Craig, L., Mehroke, J.S., Rashed, M., Zwick, M.B., Kenar, K., Toone, E.J., Greenspan, N., Auzanneau, F.-I., Marino- Albernas, J.-R., Pinto, B.M., and
Scott, J.K. Exploring the basis of peptide-carbohydrate crossreactivity: Evidence for discrimination by peptides between closely related anti-carbohydrate antibodies. Proc. Natl. Acad. Sci. USA 94:2454-2459, 1997.

7. Pinilla, C., Appel, J.R., Campbell, G.D., Buencamino, J., Benkirane, N., Muller, S., Greenspan, N.S. All-D peptides recognized by an anti-carbohydrate antibody identified from a positional scanning library. J. Mol. Biol. 283(5):1013-1025, 1998.

8. Greenspan NS, Davie JM. Serologic and topographic characterization of idiotopes on murine monoclonal anti-streptococcal group A carbohydrate antibodies. J Immunol. 1985 Feb;134(2):1065-72.

9. Kramer A, Keitel T, Winkler K, Stocklein W, Hohne W, Schneider-Mergener J. Molecular basis for the binding promiscuity of an anti-p24 (HIV-1) monoclonal antibody. Cell. 1997 Dec 12;91(6):799-809.

10. Keitel T, Kramer A, Wessner H, Scholz C, Schneider-Mergener J, Hohne W. Crystallographic analysis of anti-p24 (HIV-1) monoclonal antibody cross-reactivity and polyspecificity. Cell. 1997 Dec 12;91(6):811-20.

11. Michaud GA, Salcius M, Zhou F, Bangham R, Bonin J, Guo H, Snyder M, Predki PF, Schweitzer BI. Analyzing antibody specificity with whole proteome microarrays. Nat Biotechnol. 2003 Dec;21(12):1509-12. Epub 2003 Nov 9.

12. Bhat TN, Bentley GA, Fischmann TO, Boulot G, Poljak RJ. Small rearrangements in structures of Fv and Fab fragments of antibody D1.3 on antigen binding. Nature. 1990 Oct 4;347(6292):483-5.

13. Rini JM, Schulze-Gahmen U, Wilson IA. Structural evidence for induced fit as a mechanism for antibody-antigen recognition. Science. 1992 Feb 21; 255(5047):959-65.

14. James LC, Roversi P, Tawfik DS. Antibody multispecificity mediated by conformational diversity. Science. 2003 Feb 28;299(5611):1362-7.

15. Jimenez R, Salazar G, Baldridge KK, Romesberg FE. Flexibility and molecular recognition in the immune system. Proc Natl Acad Sci U S A. 2003 Jan 7;100(1):92-7. Epub 2002 Dec 23.

16. Parren PW, Burton DR. Immunology. Two-in-one designer antibodies. Science. 2009 Mar 20;323(5921):1567-8.

17. Ellis LM, Reardon DA. Cancer: The nuances of therapy. Nature. 2009 Mar 19;458(7236):290-2.