Differences in NK cells

ViSNE clustering of NK cells (defined within the non-tumor compartment by CD45^hiCD3^-CD14^-CD19^-CD56⁺ events) identified three distinct NK populations (figure 1A). These populations were defined by their surface and cytosolic immune markers as (1) immature NK cells (CD56^hiCD16^-CD57^-), (2) early cytotoxic NK cells (CD56^dimCD16⁺CD57^-Granzyme⁺), and (3) mature CD57+NK cells (CD56^dimCD16⁺CD57⁺granzyme⁺) (figure 1B). The immature NK cells expressed higher levels of NKG2D and CD27 as previously described19 while CD57+NK cells expressed higher amounts of Tbet and HLADR in addition to granzymeB and lacked expression of CD27.19 Differences in NK cell subsets were most evident in the Hispanic group. Hispanic patients had higher proportions of CD57⁺ mature NK cells when compared with other groups, (42%±3% vs 33±2%; p=0.006) (figure 1C). These differences were most apparent within SR patients with Hispanic SR patients having greater numbers of CD57+NK cells compared with B and NHW SR patients (mean H SR 44±5% vs 25±2% p=0.001) (figure 1D–E).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Changes in NK cells bone marrow mononuclear cells obtained at diagnosis (Hispanic=29, black=18, NHW=38) were stained with metal conjugated antibodies and data acquired using Helios mass cytometry machine. The data were normalized using the calibration beads and analysis performed using Cytobank software and VISNE algorithm. Concatenated files with equal numbers of cells from individual samples were used to visualize the data. (A) Density plot of NK cells from H, B and NHW patients showing three distinct subsets of NK cells (NK1; CD56^hiCD16^-CD57^-GZMB^-), NK2; CD56^DimCD16⁺CD57^-GZMB⁺ and NK3; CD56^DimCD16⁺CD57⁺GZMB⁺). Numbers denote percentages of NK cells in the specific cluster. (B) Heatmap showing expression of various markers in the NK subsets. (C) Density plot showing distribution of NK cells within the 3 distinct clusters by race/ethnicity and disease risk (standard risk; SR high risk; HR). Numbers in the plot denote percentages of NK cells in the specific cluster. (D) Dot plot showing CD57+NK cells as percent of total NK cells in all Hispanic patients compared to black and NHW(others) as well as in SR and HR Hispanic as compared to SR and HR blacks and NHW(others) . Figure shows mean and standard error of the mean. Figure shows mean and SE of the mean, B, black; H, Hispanic; NHW, non-Hispanic white.

Differences in myeloid cells

The myeloid cell compartment was defined within the non-tumor compartment by CD45^hiCD3^-CD56^- cells that were positive for either CD11c or CD33. ViSNE clustering of myeloid cells identified five distinct clusters (termed M1-M5) based on patterns of cell surface markers (figure 2A). This included a CD14+CD16- classical monocyte population (M3), CD11c+DR low population(M4), CD33+C-kit+population (M2), CD11b/11c+CD15+population (M5) and a CD11c+CD16+DR^hi inflammatory/non classical myeloid population (M1) (figure 2B). M1 myeloid population also expressed higher levels of PD-L1 as well as HLA-E (figure 2B). Further analysis showed that NHW SR patients have significantly lower proportions of CD16+DR+ myeloid cells (M1) compared with other groups (mean NHW SR 3±2% vs Others 17%±3% p=0.02) (figure 2C–D).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Changes in myeloid cells bone marrow mononuclear cells (Hispanic=29, black=18, NHW=38) were analyzed using single cell mass cytometry or CyTOF as described above. Figure shows VISNE visualization of the myeloid cells. (A) Density plot of myeloid cells from H, B and NHW patients showing five distinct subsets of myeloid cells (M1–M5). (B) Heat map shows expression of different markers that help distinguish five distinct myeloid clusters (M1; 16+DR^hi non classical/inflammatory myeloid cells; M2 34+immature/progenitor myeloid cells, M3. CD14+classical monocytes, M4 CD11C+16- myeloid cells and M5 CD11C+CD11B+16+DR- myeloid cells). (C) Density plot showing proportion of myeloid cells by race/ethnicity and risk. Numbers denote percentages of myeloid cells in the specific cluster. (D) Dot plot showing proportions of CD16+DR+ myeloid cells as percent of total myeloid cells in individual SR white patients compared with all others. Figure shows mean and SE of the mean. B, black; H, Hispanic; HR, high risk; NHW, non-Hispanic white; SR, standard risk.

Differences in T cells

ViSNE analysis of T cells revealed four CD8 and two CD4 clusters progressing from CCR7/CD45RA+naïve cells to CD57+CD45RO-RA+terminally differentiated cells (online supplemental figure 2A, B). These subsets revealed expected changes in transcription factors and surface/cytosolic markers (online supplemental figure 2B), but in contrast to findings with innate immune cells, these clusters were comparable between racial/ethnic groups (online supplemental figure 2C).

Differences in immune checkpoints, transcription factors and co-stimulatory genes

Both CD4 +and CD8+T cells expressed PD-1, TIGIT as well as LAG-3 as major inhibitory immune checkpoints, while TIGIT was the dominant inhibitory checkpoint on NK cells and molecules in the TIGIT axis (CD112, CD155) were also prominent on myeloid cells (online supplemental figure 3). The expression of inhibitory checkpoints on CD4/CD8+T cells was comparable between racial/ethnic (R/E) cohorts (online supplemental figure 4). CD4/CD8+T cells also did not differ in terms of costimulatory molecules (CD27, CD28, 4-1BB) and key transcription factors (TCF1, T-bet, Eomes) (online supplemental figures 5 and 6). While checkpoints (LAG3, PD-1, TIGIT and DNAM-1) were comparably expressed on NK cells from different R/E cohorts (online supplemental figure 7A), NK cells from Hispanic patients expressed lower proportion of NKG2D, consistent with lower proportion of immature NK cells in this cohort19 (online supplemental figure 7B). Myeloid cells showed comparable expression of molecules on TIGIT axis and HLA-E in different R/E cohorts (online supplemental figure 8). Myeloid cells from SR-NHW cohort however had lower expression of PD-L1, consistent with prior data showing lower proportion of inflammatory myeloid cells in this cohort (online supplemental figure 9). Together, these data suggest that the prominent differences based on race/ethnicity are in the innate immune system.

These studies use a high-dimensional single cell proteomic analysis to identify several race/ethnicity-based differences in the immune microenvironment in children with B-ALL. Surprisingly, major race/ethnicity-based differences were found in the innate immune marrow non-tumor compartment. These differences are of interest as both NK and myeloid cells have been implicated in the immune-pathogenesis of B-ALL and therapies specifically targeting innate immunity are now entering the clinic.16 17 These differences may therefore not only contribute to poor disease-free survival in Hispanic and Black children, but also impact outcomes following treatment with emerging immune therapies.

In recent studies, alterations in NK cells have been linked to worse outcome in pediatric B-ALL.16 20 Single cell sequencing studies have also characterized altered transcriptional profiles in NK cells in B-ALL consistent with NK dysfunction.14 However, these initial studies were based on B-ALL samples from the Children’s Oncology Group tissue bank and samples from non-white racial/ethnic groups were underrepresented. The finding that Hispanic children had increased proportion of CD57 +mature NK cells is therefore of interest. This is a distinct subset of human NK cells with lower proliferative potential and decreased ability to secrete interferon gamma, but higher cytolytic function, that may expand in the setting of chronic viral infections.19 The mechanism underlying differences in NK differentiation in Hispanic children needs further study but may be related in part to differences in killer inhibitory receptor genotypes or environmental exposures in these children.21 In prior studies, changes in NK cells were more prominent in children with AML than in B-ALL.14 Therefore, it would also be of interest to evaluate if the race/ethnicity-associated differences in NK cells that we observed in B-ALL are also present in children with AML. Race/ethnicity related differences in NK biology may also have clear implications for the emerging application of NK-based therapies in leukemia.

These data have several implications for testing novel immune approaches to treat B-ALL in children. The finding that TIGIT and LAG-3 are prominently expressed by immune cells in children with B-ALL suggest that further studies to better understand the role that these clinically targetable immune checkpoint molecules play in the leukemic microenvironment should be explored. Combination studies involving TIGIT and LAG-3 blockade have recently shown promise in adult cancer and should be formally tested in pediatric leukemia.22 Our finding that black and Hispanic children have an increase in distinct populations of non-classical activated myeloid cells is also of interest, as this specific population was recently linked to leukemic progression in both patient specimens and preclinical models.17 In preclinical models, macrophage-targeted approaches led to improved outcome.17 Therefore, strategies targeting the myeloid compartment may be particularly useful in Hispanic or black children. NK-based immune therapies have already shown promise in adult cancers and are rapidly moving to the clinic.23 Our data suggest that racial/ethnic differences in NK biology may impact the translation of these strategies in the clinic, including in children.

Strengths of the current study are evaluation of specimens from patients with racially diverse representation obtained from a well-characterized biorepository and use of high-dimensional approaches validated in prior studies.14 24 25 Weaknesses include a relatively small sample size, although to our knowledge, this dataset represents one of the largest cohorts of non-white children analyzed at diagnosis of B-ALL with high-dimensional approaches. In prior studies, we have shown that risk status has a major impact on immune microenvironment.14 While the clinical features of different racial/ethnic cohorts were comparable, there could be yet unknown differences in cancer genetics in these cohorts. Nonetheless, these data need to be confirmed in another independent dataset, and as such should be viewed as hypothesis generating. Together, our novel finding is that there are surprisingly substantial differences in the innate immune compartment in Hispanic and black children with B-ALL. These findings implore the urgent and unmet need to include higher proportions of these children in current studies of immune microenvironment and emerging immune therapies, particularly those that specifically target innate immunity.