Splicing-Mediated Antigen Escape from Immunotherapy for B-cell Malignancies

Antigen escape, the process by which an epitope on a cancer cell targeted by an anticancer immunotherapy is no longer recognized by the immune system, has emerged as a major cause of resistance to a diverse array of immunotherapy agents. Mechanisms of antigen escape have been heavily studied in the context of B-cell malignancies such as B-cell acute lymphoblastic leukemia (B-ALL), mature B-cell lymphomas, and multiple myeloma because of the ease of obtaining malignant cells from patients with many of these cancer types as well as the multitude of successful B-cell antigen–targeted therapies. For example, a number of studies have elucidated mechanisms of acquired resistance to anti-CD19 (1–3) and anti-BCMA chimeric antigen receptor (CAR) T cells (4).

Downregulation of an antigen targeted by immunotherapies has been documented to occur by a wide array of means. These include mutations at the level of DNA that alter protein expression, sequence, or plasma membrane localization (1, 2); changes in RNA splicing of the transcript encoding the epitope (1, 2); and alteration in the cellular lineage with consequent change in cell surface marker expression (ref. 3; Fig. 1). More recently, masking of an epitope by unintentional transduction of a CAR vector (5) to a tumor cell and even transfer of the cell surface antigen from the tumor cell to neighboring immune cells (a process known as trogocytosis; ref. 6) have been documented to result in antigen escape (Fig. 1).

Despite these insights, our understanding of precise mole-cular mechanisms of antigen escape is still evolving. One challenge to this effort has been that cell surface protein or mRNA measurements alone may not identify changes in the exact protein epitope that serves as a therapeutic target. Changes in the cell surface protein can even occur despite a lack of detectable mutations within protein-coding exons. To this end, Zheng and colleagues tackled the question of whether the mRNA encoding CD22 may be altered as a means of resistance to anti-CD22–targeted immunotherapies (7).

CD22 is a Siglec family lectin whose expression pattern is very similar to CD19, where its presence on the cell surface begins at the pre–B-cell stage, persists on mature B cells, but is lost at the level of plasma cells. Studying mechanisms of resistance to CD22-targeted agents is important, as there are currently two FDA-approved CD22 antibody–drug conjugates (ADC). These include inotuzumab ozogamicin, which is FDA approved for relapsed pediatric and adult B-ALL, as well as moxetumomab pasudotox, which is FDA approved for relapsed hairy cell leukemia. Given that CD22 expression is restricted to B cells, there are a number of ongoing efforts to develop anti-CD22–directed CAR T cells for a number of B-cell malignancies (8).

CD22 is an inhibitory coreceptor that promotes apoptosis and serves to restrict B-cell proliferation. Interestingly, prior work had identified that deletion of exons encoding the intracellular domains of CD22 required to promote apoptosis can drive pediatric B-ALL disease development (9). However, prior to this study by Zheng and colleagues, molecular causes for resistance to anti-CD22 immunotherapies were obscure. For instance, a clinical trial of CD22-targeted CAR T cells in B-ALL by Fry and colleagues noticed diminished CD22 cell surface expression in patients who developed resistance to this CAR (8), but no mutations or altered splice isoforms of CD22 could be identified to explain the change in CD22 cell surface expression and resistance.

CD22 has 14 exons, and the canonical translational initiation codon is located in exon 2. Evaluation of the mRNA isoforms of CD22 in short-read conventional RNA sequencing data from untreated pediatric B-ALL cells by Zheng and colleagues identified a wide array of CD22 alternative spliced isoforms, and the distribution of these isoforms within B-ALL cells differed greatly with those in normal primitive and mature B-cell subsets. Interestingly, CD22 appeared to have more isoforms within pediatric B-ALL cells than other B-ALL immunotherapy targets such as CD19, CD20, or CD79B.

Two isoforms of CD22 appeared to be more highly enriched in B-ALL cells than normal B-cell subsets and were found to have therapeutic importance. First, CD22 isoforms lacking exons 5 to 6 (CD22 Δex5–6) were found to result in a CD22 protein that localizes to the cell surface but is resistant to killing by HA22 CAR T cells, whose CAR is based on the exon 5–recognizing RFB-4 antibody. The authors impressively developed custom isoform-specific antibodies to rigorously confirm the presence of CD22 protein lacking exons 5 to 6.

In addition to CD22 Δex5–6, the authors identified CD22 isoforms that lack exon 2 (CD22 Δex2). Interestingly, skipping of exon 2 does not allow usage of a later translational initiation site, and CD22 Δex2 isoforms fail to result in CD22 protein expression. This latter point was confirmed using epitope-tagged constructs of this isoform where the tag could not be detected. B-ALL cell lines expressing this isoform of CD22 alone were as insensitive to inotuzumab as CD22 knockout versions of the same cell lines. Moreover, antisense morpholinos, which reduce exon 2 inclusion in the CD22 transcript, decreased CD22 cell surface expression, and cells that exclusively express CD22 Δex2 fail to respond to inotuzumab.

Beyond these elegant mechanistic studies, Zheng and colleagues also identified several children with B-ALL who relapsed to treatment with inotuzumab whose cells were characterized by CD22 protein downregulation and elevated expression of the CD22 Δex2 isoform. Interestingly, these inotuzumab-resistant CD22 isoforms are present to some degree in patients with B-ALL prior to inotuzumab therapy. Whether this is due to a subpopulation of B-ALL cells with inotuzumab-resistant CD22 isoforms or a mixture of CD22 isoforms within individual B-ALL cells will be very interesting to discern in the future. Nonetheless, this finding of preexisting drug-resistant isoforms suggests the possibility that some patients may be predestined to become inotuzumab resistant. This splicing-based mechanism of therapeutic relapse could be screened for prior to treatment.

In addition to preexisting presence of inotuzumab-resistant CD22 isoforms, there was also evidence of acquired mutations within genomic DNA impacting the RNA splicing of CD22 and enforcing skipping of exon 2 within patients with acquired resistance to therapy. Whether mutations at the level of DNA impacting RNA splicing underlie all cases of patients with CD22 Δex2 or CD22 Δex5–6 escape variants remains to be elucidated.

Given the interest in moving antigen-targeted immunotherapies—including inotuzumab, the CD19 bispecific T-cell engaging antibody blinatumomab, and anti-CD19 and CD19/CD22 dual targeting CAR T-cell therapies—to the first line of therapy for patients with B-ALL, studies such as this aimed at identifying mechanisms of antigen escape will be critical. In addition, it will be interesting to understand the expression of these CD22 splice isoforms in adults with B-ALL and patients with other B-cell malignancies. These data also argue for the need of more comprehensive genomic studies including both DNA- and RNA-based sequencing to characterize response to targeted immunotherapies moving forward. In parallel, it will be very interesting to perform unbiased functional genomic screens of cell surface immunotherapy targets to systematically identify antigen escape variants.