Heterogeneity of surface CD19 and CD22 expression in B lymphoblastic leukemia

CD19- and CD22-targeting immunotherapy agents: tisagenlecleucel, blinatumomab, and inotuzumab, were approved for the treatment of relapsed/refractory B-ALL by the Food and Drug Administration in 2017.1 However, long-term responses are impacted by relapses.2, 3 Emerging data suggests that level of antigen expression before treatment may impact responses to immunotherapy.2 Currently, there are no guidelines for assessment of CD19 and CD22 prior to CD19- and CD22-targeted therapy. Correlative studies are needed to determine the relationship between antigen expression and response to targeted therapy. However, such studies are limited by lack of baseline quantitative data on CD19 or CD22 expression in pediatric B-ALL. Prior estimates of CD19- and CD22-dim/negative B-ALL vary from 0% to 20% depending on the cut offs used for “positivity” and the methodology.4-6 Correlative studies are also limited by variation in flow cytometry data from different time points and disparate laboratories. We address these questions in a large single institution cohort of B-ALL. The flow cytometry archives (2002-2016) at the Children's Hospital of Philadelphia (CHOP) were reviewed to identify cases of de novo and relapsed B-ALL (approved by CHOP Institutional Review Board, protocol 16-013199) in the pediatric and young adult age group (0-25 years). The flow cytometric instruments, panels and antibodies were relatively unchanged during this time period (see Supporting Information data for detailed flow cytometric methods). In total, 628 cases of B-ALL were identified, of which 80% were de novo, and 20% presented in relapse. Mixed lineage/phenotype leukemia was excluded and diagnosis of B-ALL was confirmed per the 2016 WHO hematologic malignancy classification. Patients treated with immunotherapy (blinatumomab, inotuzumab, and CAR T-cell therapy) were also excluded in order to determine baseline sCD19 and sCD22 expression prior to immunomodulation. Archived flow cytometric PDF files were re-examined to determine antigen expression. Antigen expression was evaluated on the brightest phycoerythrin (PE) channel for both CD19 and CD22. CD19 expression showed a large range of expression with early hematogones (B-cell precursors in the bone marrow) showing the lowest expression (low normal) and activated B cells in lymph nodes showing the highest expression (high normal). Mid-late stage hematogones, peripheral blood B cells and nonactivated B cells showed similar levels of normal CD19 expression. Antigen expression on blasts was classified as dim, low/high normal and bright based on comparison to normal B cells (Supporting Information Figure S1). Aberrant “dim” was defined as lower than “low normal”. Aberrant “bright” was defined as higher than “high normal”. sCD22 consistently showed a lower range of expression than sCD19. Even normal B cells can show dim expression of CD22. Hence, distinction between dim and normal was not possible for sCD22. Only cases in which the majority of the cells lacked sCD22 were considered as “negative”. No significant differences were noted in sCD19 and sCD22 levels between de novo and relapsed B-ALL. About 7% of cases were classified as dim for sCD19, 24% were low normal, 61% were high normal, and 8% were bright, highlighting the heterogeneity of CD19 expression in B-ALL (Figure 1A). Although CD19-negative leukemias have been rarely reported, the only B-ALL that were completely negative for sCD19 in our cohort were either mixed lineage/phenotype cases or relapses after immunotherapy. Even the CD19-dim cases mostly showed a single cluster of cells with variably low sCD19 expression rather than a distinct negative population (Supporting Information Figure S2C). In contrast to CD19, a large proportion of cases (22%) were predominantly negative for sCD22 (Figure 1B). Hence, sCD19 is widely expressed at a higher level than sCD22 in B-ALL. Complete or partial loss of CD19 or CD22 occurs in relapses after CD19 and CD22-targeted therapy.6, 7 Multiple mechanisms of target antigen loss have been described including mutations, loss of surface expression, alternative splicing and lineage switch.8 The occurrence of such cases is increasing with greater use of targeted therapy. It is not clear if they will respond to re-treatment with CD19 or CD22-targeted therapy. The threshold of antigen recognition is likely different for various immunotherapy agents and must be determined individually from their clinical trial data. Original flow cytometry list mode data files that were available on cases from 2007 to 2017 were re-analyzed to objectively assess antigen expression (Supporting Information Figure S2). Median fluorescent intensity (MFI) difference was calculated as MFI of the isotype control subtracted from the MFI of normal B cells or blasts. MFI difference of CD19-normal B-ALL and low/high normal B cells were not significantly different (Supporting Information Figure S2D). However, MFI differences of the B-ALL categories (low, normal and high) were significantly different; confirming the validity of our antigen classification system. Despite high initial responses to CD19 or CD22 targeted therapy, 30-60% of patients relapse with antigen-negative disease.2, 3 It is hypothesized that antigen-negative relapses arise from pre-existing antigen-negative blasts. Hence, we investigated whether there were low levels of CD19-negative events prior to any immunotherapy. Proportion of CD19-negative events was calculated from the left upper quadrant of CD10 vs. CD19 plots (Figure 1C, inset). Only discrete events that were greater than 1% and showed other aberrant blast markers were considered as CD19-negative blasts. About 17% of B-ALL cases showed sCD19-negative blasts even before any immunotherapy (Figure 1C). Correlative and genetic studies will be important in determining if these blasts are the origin of CD19-negative relapses. We then correlated dim sCD19 expression and lack of sCD22 with common genetic abnormalities seen in B-ALL (see Supporting Information Data for details of genetic studies). The proportion of genetic abnormalities seen in our cohort was consistent with their incidence in B-ALL, except for Ph-like B-ALL which were only characterized recently. KMT2A/MLL gene rearranged B-ALL were more likely to be sCD19-dim and sCD22-negative (Figure 1D). Lower CD22 site density was also noted previously in KMT2A rearranged B-ALL.9 KMT2A/MLL gene rearrangement is known to occur in an early myeloid/lymphoid progenitor10 with low expression of B cell antigens. Interestingly, dim sCD19 expression was weakly correlated with negative sCD22 (Spearman's rank correlation coefficient 0.29) suggesting that a subset of cases may be resistant to even CD19-CD22 dual targeting agents. There were no ETV6-RUNX1 translocated B-ALL that were dim for sCD19 suggesting that they are good candidates to either CD19- or CD22-targeted therapy. Intriguingly, hyperdiploid B-ALL was overrepresented in the CD19-dim group but underrepresented in CD22-negative group. We then analyzed TARGET data to determine levels of CD19 and CD22 RNA expression in B-ALL (method in Supporting Information data). About 18% of B-ALL showed lower expression of CD19 and 11% showed lower expression of CD22 than the RNA expression in normal bone marrow pro B cells. Hence, low expression of CD19 and CD22 can be demonstrated at the RNA level as well. Interestingly, CD22 RNA levels were higher than that of CD19 in the TARGET data set. This discrepancy with the surface expression data is likely due to greater cytoplasmic localization of CD22. Cytoplasmic CD22 rather than sCD22 is also the criteria used for assigning B cell lineage per the WHO classification for similar reasons. In conclusion, we semi-quantitatively classified baseline expression of sCD19 and sCD22 in a large cohort of pediatric B-ALL. We highlight the heterogeneity of antigen expression that may impact responses to targeted therapy. We developed and validated a classification method based on comparison to isotype controls and normal B cells that could be used in the correlation of antigen expression with responses and relapses after targeted therapy. We correlated dim CD19 and negative CD22 cases with common genetic abnormalities that will be useful in patient selection for targeted therapy. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.