Targeting mutations predictably

The report by Popovic and colleagues in this issue of Blood, while a milestone on our way to personalizing immunologic cancer therapy, highlights the pitfalls in identifying specific peptides for immunotherapy.¹ 

Translocations are the archetype of a consistent genetic abnormality in cancer²  and arguably encode the best targets for cancer therapy. Fusion proteins encoded by these translocations are truly cancer-specific, shared between cancers of the same type or subtype and usually essential for cancer maintenance making them ideal targets for pharmacologic and immunologic intervention. Furthermore, although originally identified in leukemia, specific translocations are being identified in a wide variety of malignancies including some of those that are most aggressive. Pharmacologic approaches are exemplified by the drug imatinib targeted to the fusion protein of the 9;22 translocation in CML.^2,3  Remarkable successes but also clear limitations of this Abelson tyrosine kinase inhibitor have fostered explosive research activity leading to second- and third- generation drugs with increased therapeutic potential. But we still lack effective immunologic therapies for targeting fusion proteins. This lag is surprising because immunologic approaches (1) were proposed decades ago, (2) are conceptually solid, and (3) could complement pharmacologic approaches when they are not curative alone. However, we still lack adequate understanding of the restrictions and requirements for successful targeting. By showing proteasomal cleavage to be a critical predictor of therapeutic potential, Popovic et al¹  point in the right direction and give constructive guidance for immunologists trying to target mutant proteins.

Remarkable technologic advances allow us to obtain rapid and precise sequence information on a tumor of a given patient.⁴  While tumor-specific fusion proteins encoded by translocations are shared between cancers of the same type,²  all cancers also harbor multiple mutations specific for the patient's tumor. These mutations encode the so-called unique or individually distinct antigens⁵  that have been found by many independent investigations to be the strongest, referred to as tumor rejection antigens. But what can immunologists do with this sequence information? To be recognized on the surface of cells, a mutant peptide must (1) be 8-9 amino acids in length, (2) include the fusion point of the fusion protein or the mutant amino acid(s), and (3) bind in the cleft of the MHC molecules of the patient (see figure). In an approach often referred to as reverse immunology, the mutant sequences can be entered into 1 of several internet-based prediction tools (eg, http://tools.immuneepitope.org) that give a score predicting whether a given peptide could be a good target for T cells. These algorithms were derived from experimental knowledge of appropriate proteasomal cleavage, TAP transport, and/or MHC binding. But these predictive tools cannot presently replace empiric measurements. For example, it was not expected that the predicted HLA-A*0201 binding TEL-AML1 fusion peptide was not naturally generated by cells because of multiple proteasomal cleavage sites within the sequence of this fusion peptide.¹  Thus, the candidate peptide is probably not produced in useful amounts to serve as a target presented on MHC complexes on the surface of the acute lymphoblastic leukemia cells harboring this fusion protein. In contrast, proteasomal cleavage of the control proteins did result in peptides known to be expressed as peptide MHC-class I complexes.¹ 

Another important lesson to be learned from Popovic et al is that the relatively high affinity to the HLA-A*0201 molecule predicted by computer algorithms did not translate to biologic efficacy of either inducing T cells or serving as their targets. In fact, the peptide had to be anchored artificially to the MHC to accomplish these effects. Modifying the anchor residues of peptides to improve presentation of an otherwise poorly binding tumor peptide has become a common and effective approach for inducing T-cell responses to tumor peptides that are incapable of inducing a response because of low affinity to MHC. Often these T cells kill or release cytokines in vitro very effectively when cancer cells are artificially loaded with the unmodified peptide. But one always has to remember that (1) the actual cancer cell only makes the unmodified peptide and (2) even if it is made in sufficient amounts, poor binding to the HLA of the patient would not make it a clinically useful target. Certainly it is a dangerous assumption that T cells observed in vitro,⁶  whether or not induced with a modified peptide, will overcome the problem of poor affinity of the genuine peptide to the relevant MHC molecules in vivo. So far, these assumptions have neither been supported by clinical results nor experimentally by targeting tumors of clinically relevant size, that is, in tumors that have grown for at least 2 weeks and are at least 1 cm in average diameter or contain at least 10⁹ cells.⁷ 

Scientists have argued that during the long clonal evolution of cancer, the immunocompetent host selects against cancer-driving mutations that could serve as effective targets on the HLA molecules in each particular patient. Moreover, some tumor-specific mutations will not fulfill the molecular requirements to be properly processed and bound by the patient's MHC molecules. However, all cancers probably retain and express at least some powerful tumor-specific antigens, yet cancers fail to be rejected because the host is either tolerized to these antigens or the cancer cells became nonimmunogenic using other mechanisms that masked their antigenicity. Experimental studies using clinically relevant tumor loads can define what is needed in terms of affinity, amount, and specificity of antigen to achieve tumor eradication. Then, clinicians would be well served to select from the many potential tumor-specific target peptides only those for clinical studies that are properly processed, made in large amounts, and have high affinity for the relevant MHC of the patient. Popovic et al use excellent methods to examine appropriate processing of possible antigens,¹  and reliable experimental measurements of peptide-MHC affinities can be made rapidly and on a routine basis for an increasing number of specific HLA subtypes.⁸  These tools should be used to select the best targets from the multiple cancer-specific mutations a cancer expresses. Once identified, technology is certainly available to generate specific T-cell receptors in humanized mice or in vitro and these receptors could then to be used to transduce the patient's own T cells to selectively destroy the cancer. Cytolytic T cells specific for Epstein Barr virus (EBV)–encoded peptides are already used to eradicate far advanced chemo-, antibody-, and radiation-resistant EBV-positive cancers. This work on EBV-positive cancers gives us the hope that immunologic treatments will be developed that target effectively mutant proteins. In this way, immune therapy will complement or substitute for chemo-, radiation, and/or other forms of cancer therapy.