Research ArticleCancer

CD99 is a therapeutic target on disease stem cells in myeloid malignancies

See allHide authors and affiliations

Science Translational Medicine  25 Jan 2017:
Vol. 9, Issue 374, eaaj2025
DOI: 10.1126/scitranslmed.aaj2025

Stemming the tide of leukemia development

Acute myeloid leukemia and myelodysplastic syndromes are maintained by specific populations of malignant stem cells, and successful treatment requires the eradication of these disease-causing cells. Chung et al. identified CD99 as a marker expressed on the surface of leukemic stem cells but not normal hematopoietic stem cells, suggesting its potential as a therapeutic target. A monoclonal antibody against CD99 had promising preclinical effectiveness in xenograft models and was selective for malignant stem cells, paving the way for further development of this approach.

Abstract

Acute myeloid leukemia (AML) and the myelodysplastic syndromes (MDS) are initiated and sustained by self-renewing malignant stem cells; thus, eradication of AML and MDS stem cells is required for cure. We identified CD99 as a cell surface protein frequently overexpressed on AML and MDS stem cells. Expression of CD99 allows for prospective separation of leukemic stem cells (LSCs) from functionally normal hematopoietic stem cells in AML, and high CD99 expression on AML blasts enriches for functional LSCs as demonstrated by limiting dilution xenotransplant studies. Monoclonal antibodies (mAbs) targeting CD99 induce the death of AML and MDS cells in a SARC family kinase–dependent manner in the absence of immune effector cells or complement, and anti-CD99 mAbs exhibit antileukemic activity in AML xenografts. These data establish CD99 as a marker of AML and MDS stem cells, as well as a promising therapeutic target in these disorders.

INTRODUCTION

Acute myeloid leukemia (AML) and the myelodysplastic syndromes (MDS) are disorders arising in the hematopoietic system that share common molecular origins, with MDS patients showing a propensity to progress to AML. AML arises from immature hematopoietic cells and is composed of leukemic blasts organized in a developmental hierarchy reminiscent of normal hematopoiesis. At the apex of this hierarchy are leukemic stem cells (LSCs) that have the capacity to self-renew and differentiate into non–self-renewing progeny that comprise the vast majority of leukemic blasts (1, 2). LSCs appear to be largely resistant to conventional chemotherapy and are thought to serve as the reservoir of minimal residual disease that is responsible for disease relapse after initial treatment (3). Xenografts and mouse models of AML suggest that fully transformed LSCs arise not from hematopoietic stem cells (HSCs) but from more committed myeloid progenitors that acquire aberrant self-renewal (46). In contrast, MDS is initiated by neoplastic HSCs that fail to give rise to sufficient numbers of mature hematopoietic cells, resulting in bone marrow (BM) failure (79). True to their identity as disease stem cells, when transplanted into immunodeficient animals, MDS HSCs give rise to long-term grafts that recapitulate features of MDS (7, 10). Similar to AML LSCs, MDS HSCs are also highly resistant to standard therapies (8, 11). Thus, despite the likely differing cellular origins of AML and MDS, curative therapies for these malignancies must eliminate disease stem cells (LSCs or MDS HSCs, respectively), because they are likely the only self-renewing disease cells in the BM.

In efforts to identify potential therapeutic targets in AML, a number of groups have identified cell surface proteins preferentially expressed on AML LSCs compared to normal HSCs, including CD47 (12), CD44 (13), CD96 (14), TIM3 (15, 16), and CD123 (17). Although these antigens are present on LSCs, their relative expression among AML cells has not been shown to enrich for functional LSCs. In the context of MDS, aberrant HSC surface protein expression has not been carefully characterized, and therefore, no therapies targeting MDS stem cells on the basis of differentially expressed cell surface antigens have been described or evaluated.

Here, we demonstrate that CD99 expression is frequently increased on immunophenotypic AML LSCs and MDS HSCs compared to their normal hematopoietic counterparts, hematopoietic stem and progenitor cells (HSPCs). Furthermore, high CD99 expression can be used to identify and prospectively separate leukemic cells from nonleukemic cells in AML patient BM specimens, as well as to selectively enrich for functional LSC activity, as demonstrated by xenotransplantation assays. Ligation of AML and MDS stem cells with a monoclonal antibody (mAb) recognizing CD99 directly induces cytotoxicity in vitro. Moreover, anti-CD99 mAbs exhibit antileukemic activity in AML xenografts, demonstrating the potential of anti-CD99 mAbs as therapeutic agents.

RESULTS

MDS HSCs and AML blasts frequently exhibit high expression of CD99

To identify candidate cell surface proteins differentially expressed on MDS HSCs, we evaluated the transcriptomes of purified HSCs [lineage-negative (LN) CD38+CD34CD90+CD45RA] from MDS patients and age-matched controls (18), identifying 25 dysregulated transcripts encoding cell surface antigens. We used flow cytometry (FC) to validate cell surface expression of these antigens on MDS HSCs from 24 MDS patient specimens (table S1) as compared with cord blood (CB) HSC controls, identifying CD99 as the most frequently overexpressed among the antigens tested (83% of cases as defined by a >50% increase in mean fluorescence intensity, mean 6.35-fold increase; Fig. 1, A and B). Normal adult BM HSCs had cell surface CD99 expression at levels similar to CB HSCs (fig. S1). We previously demonstrated that CD99 transcripts are more highly expressed in LSC-enriched CD34+CD38 AML cells compared with normal BM CD34+CD38 cells (14), suggesting that CD99 protein is up-regulated on AML LSCs. Comparison of cell surface CD99 expression on unfractionated bulk leukemic blasts (CD45lowSSClow, representative gating shown in fig. S2) from 79 paired diagnosis/relapse AML samples (table S2) with expression on normal CB HSCs by FC confirmed increased expression of CD99 in 82% of diagnostic samples and 90% of relapse samples (average 8.34-fold increase; Fig. 1B).

Fig. 1. MDS and AML stem cells express high amounts of CD99.

(A) HSCs (LN CD34+CD38-CD90+CD45RA) from patients with MDS (n = 24) and normal CB controls (n = 24) were analyzed by FC for CD99 expression, (B) shown as fold change CD99 mean fluorescence intensity (MFI) on MDS HSCs compared with CB HSCs. Unfractionated blasts (CD45lowSSClow) from diagnostic (n = 39) or relapse (n = 40) AML specimens and normal CB HSCs (n = 18) were also analyzed by FC for CD99 expression, shown as fold change CD99 MFI on AML blasts compared with CB HSCs. The scatterplot shows the mean ± SEM. P values were calculated by paired t test. (C) CD99 expression was evaluated by FC on CD3CD19CD34+CD38 cells from AML specimen MSK AML-003. By immunophenotype, CD99-negative cells include HSCs and MPPs (LN CD34+CD38CD90CD45RA), whereas CD99-positive cells are almost entirely composed of LMPP-like cells (LN CD34+CD38CD90CD45RA+). (D) When these populations were plated in methylcellulose (750 cells in triplicate), normal myeloid/erythroid colonies formed only from the CD99-negative fraction. Error bars represent ±SEM of triplicates. CFU-E, colony forming unit–erythroid; BFU-E, burst forming unit–erythroid; CFU-GEMM, colony forming unit–granulocyte, erythrocyte, monocyte, megakaryocyte; CFU-GM, colony forming unit–granulocyte, monocyte; CFU-M, colony forming unit–monocyte; CFU-G, colony forming unit–granulocyte. (E) All CD99-negative–derived methylcellulose colonies lacked the heterozygous FLT3-ITD (internal tandem duplication) and NPM1 abnormalities present in MSK AML-003, but 14 of 33 sequenced colonies harbored a heterozygous DNMT3A R882P abnormality. CD99-negative sorted cells and xenografts lacked DNMT3A, NPM1, and FLT3 mutations, whereas all three abnormalities were present in CD99-positive sorted cells and xenografts. (F) Summary of the percentage of CD99-negative cell–derived colonies from 10 AML specimens that were positive for disease-associated mutations in the indicated alleles (which were confirmed to be present in the bulk fraction of each corresponding AML). Each data point represents a summary of results for each AML harboring a particular mutation (shown in detail in fig. S4). (G) CD99-negative cells (2500 cells) were transplanted into sublethally irradiated NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice (n = 3), with mice demonstrating human lymphomyeloid engraftment in the BM 10 months after transplantation. Representative FACS plots are shown from two CD99-low xenografts. (H) Sanger sequencing traces reveal the absence of DNMT3A R882P and NPM1 W288fs mutations in CD99-negative cell–derived xenografts and the presence of both of these abnormalities in CD99-positive cell–derived xenografts.

CD99 distinguishes between leukemic and nonleukemic hematopoietic cells

To determine whether differential CD99 expression can be used to identify and prospectively isolate LSCs from preleukemic or residual normal HSCs (19), we measured CD99 expression in the LSC-enriched CD34+CD38 fraction of AML (Fig. 1C and fig. S3A). Within this fraction, CD99+ cells expressed an antigenic profile consistent with lymphoid-primed multipotent progenitors (LMPPs; CD34+CD38CD90CD45RA+), the immunophenotype most highly enriched for LSCs in the majority of human AML (5), whereas CD99 cells exhibited an antigenic profile consistent with a mixture of HSCs and multipotent progenitors (MPPs) (CD34+CD38CD90+/−CD45RA) present in proportions similar to those observed in normal hematopoiesis (20). CD34+CD38CD99 cells from 10 independent AML specimens were fluorescence-activated cell sorting (FACS)–purified and grown in methylcellulose and demonstrated the robust myeloid colony formation characteristic of normal HSCs, but not observed with LSCs (Fig. 1D and figs. S3, B and C, and S4) (21). These colonies lacked the full complement of molecular genetic abnormalities identified in the corresponding bulk AML blasts, consistent with their derivation from residual normal or preleukemic HSCs (19, 22, 23) (Fig. 1, E and F, and figs. S3B and S4). Moreover, transplantation of FACS-purified CD99 cells into sublethally irradiated NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice (n = 3) resulted in lymphomyeloid human engraftment (Fig. 1G and fig. S5A), with engrafted cells lacking all leukemia-associated mutations (Fig. 1, E and H). In contrast, transplantation of FACS-purified CD99+ cells into NSG mice (n = 4) resulted in engraftment of a rapidly lethal myeloid leukemia (fig. S5, A to C) harboring all leukemia-associated mutations (Fig. 1E). Together, these data confirm the utility of CD99 in separating leukemogenic LSCs from functionally normal HSCs.

CD99 enriches for functional LSCs

In AML specimens with an identifiable LMPP-like population (CD34+CD38CD90CD45RA+, n = 69) within the CD99+ leukemic fraction, this LSC-enriched population consistently had higher expression of CD99 compared with more differentiated blasts, demonstrating intraleukemic heterogeneity of CD99 expression and suggesting that LSCs exhibit the highest expression of CD99 (Fig. 2A and fig. S2). Consistent with the chemoresistant nature of LSCs, CD99 expression was significantly higher on leukemic blasts at relapse when compared directly with paired specimens from the time of diagnosis (P = 0.01; Fig. 2B). To test whether CD99 enriches for LSC function among CD99+ leukemic blasts, we transplanted the highest and lowest CD99-expressing leukemic cells (top and bottom 10%, respectively) within the LSC-enriched LMPP-like fraction of a CD34+ primary AML specimen into sublethally irradiated NSG mice. Leukemia-initiating cell (L-IC) activity was only observed in mice transplanted with the top 10% of CD99-expressing LMPP-like cells, indicating that CD99 is not only highly expressed in AML but is also expressed in the highest amounts on functional LSCs (Fig. 2, C and D). Similar results were obtained when transplanting the highest and lowest CD99-expressing leukemic cells (top and bottom 15%, respectively) from a CD34 AML (Fig. 2E; gating strategy shown in fig. S6, A and B), a subset of AML for which a strategy to enrich for functionally defined LSCs has not been reported. To test whether CD99 can enrich for functional LSCs irrespective of CD34 and CD38 expression, we transplanted the highest and lowest CD99-expressing leukemic cells (top and bottom 10%, respectively) among bulk unfractionated leukemic blasts (CD45lowSSClow) from an AML with variable CD34 and CD38 expression, demonstrating a 10-fold increase in L-IC activity in the CD99-high fraction (Fig. 2F; gating strategy shown in fig. S6, C and D). When the CD99-high and CD99-low populations from both CD34+ and CD34 AMLs were cultured in vitro in liquid culture supplemented with cytokines, CD99-high cells demonstrated improved viability compared with CD99-low cells (fig. S7), a previously described characteristic of LSC-enriched blasts in AML (24, 25).

Fig. 2. CD99 expression enriches for functional LSCs.

(A) In AML specimens with an identifiable LMPP-like (LN CD34+CD38CD90CD45RA+) LSC-enriched population (n = 69), CD99 expression was higher in LMPP-like blasts compared with bulk unfractionated blasts. P value was calculated by paired t test. (B) CD99 expression was higher on bulk unfractionated AML blasts from relapse samples (n = 40) as compared with diagnostic samples (n = 39). Error bars represent ±SEM. P value was calculated by paired t test. (C) Gating strategy for purifying LSC-enriched LMPP-like cells from AML specimen UP31. From this fraction, the highest and lowest CD99-expressing blasts (top and bottom 10%, respectively) were FACS-purified to >95% purity and transplanted at limiting dilution into sublethally irradiated NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice. Sort gates were drawn on the basis of normal HSC, MPP, and LMPP populations in CB, as depicted in Fig. 1A. (D) Leukemic engraftment (defined as >0.1% detectable human cells in the BM >12 weeks after transplantation) was only observed in mice transplanted with “top 10%” CD99-expressing LMPP-like blasts. CI, confidence interval. (E) The highest and lowest CD99-expressing blasts (top and bottom 15%, respectively) from the CD34-negative AML specimen UP32 were FACS-purified to >95% purity and transplanted at limiting dilution into NSG mice. Leukemic engraftment was only observed in mice transplanted with “top 15%” CD99-expressing blasts. (F) The highest and lowest CD99-expressing blasts (top and bottom 10%, respectively) from the bulk (CD45lowSSClowCD45RA+) fraction of AML specimen MSK AML-003 were FACS-purified to >95% purity and transplanted at limiting dilution into NSG mice. L-IC activity was estimated to be 10-fold higher in the top 10% CD99-expressing blasts. (G to I) The top 10% and bottom 10% of CD99-expressing blasts within the LMPP-like fraction of eight independent primary AML specimens, as well as the bulk fraction from six of these AML specimens, were FACS-purified, and RNA sequencing was performed. Gene set enrichment analysis revealed in top 10% CD99-expressing blasts enrichment for LSC (2729) and HSC gene signatures (28, 30), as well as depletion of ribosomal gene transcripts (Kyoto Encyclopedia of Genes and Genomes) (55) and gene signatures associated with translation (Reactome) (56, 57). NES, normalized enrichment score; FDR, false discovery rate.

RNA sequencing of the top 10% and bottom 10% of CD99-expressing blasts within the LMPP-like fraction of seven primary AML specimens and the bulk fraction of a CD34 AML (postsort data shown in Fig. 2C and figs. S6 and S8) revealed 1558 differentially expressed genes (P < 0.05). Gene set enrichment analysis (26) revealed enrichment for LSC (2729) and HSC (28, 30) gene signatures in CD99-high blasts (Fig. 2, G and H). Furthermore, CD99-high blast–associated transcriptomes were depleted for ribosomal gene transcripts and gene signatures associated with translation (Fig. 2I), consistent with recent reports that both normal and malignant HSCs exhibit highly regulated rates of translation (31, 32). Because functional LSCs exist in populations outside of the LMPP fraction of AML (28, 33), we performed RNA sequencing of the top 10% and bottom 10% of CD99-expressing cells among bulk leukemic blasts from six of the eight primary AMLs initially tested (excluding one AML that was entirely CD34 and one that was entirely CD34+; postsort data shown in fig. S9) and found enrichment for the same LSC and HSC gene signatures, as well as depletion of translation-associated gene signatures (Fig. 2, G to I).

In six of the specimens confirmed to harbor recurrent AML-associated somatic mutations, we were able to assess for the burden of mutant transcripts in the CD99-high and CD99-low fractions. In three of the six specimens, the variant transcript frequency (VTF) was higher in the CD99-high fractions [MSK AML-002 (particularly in the LMPP fraction), MSK AML-005, and MSK AML-006], but there was no difference in two specimens (MSK AML-003 and UP31) and a decrease in VTF in one specimen (MSK AML-006) (fig. S10). Thus, in a subset of AMLs, CD99 expression may correlate in part with clonal heterogeneity.

Collectively, these data provide functional and transcriptomal evidence that CD99 is preferentially expressed on LSCs, identifying CD99 as an LSC-specific cell surface marker that can be used to selectively enrich for LSC activity in AML. Other LSC markers do not appear likely to share this feature, because they are expressed in equal or lower amounts on LSC-enriched fractions compared to more differentiated blasts (fig. S11) (1215, 17). Thus, these data validate CD99 expression as a means to purify LSCs, which should allow for more refined studies of this biologically and clinically relevant cell population.

Anti-CD99 mAbs are directly cytotoxic to AML and MDS cells

Because therapeutic targeting of CD99 with mAbs has shown promise in other human cancers such as Ewing’s sarcoma (34), we tested the ability of anti-CD99 mAbs to induce cytotoxicity in AML and MDS stem/progenitor cells in vitro. The anti-CD99 mAb clone H036-1.1 was cytotoxic to purified primary MDS CD34+ cells (Fig. 3A) as well as to LMPP-like AML blasts (Fig. 3B), AML CD34+ blasts (Fig. 3C), and bulk leukemic blasts (Fig. 3D). Notably, the only primary AML specimen that we tested that was resistant to anti-CD99 mAb treatment harbored a BCR-ABL translocation (Fig. 3C). In addition, when treated with lower concentrations of anti-CD99 mAb, MDS HSPCs and AML blasts with at least partial CD34 expression exhibited selective depletion of the least mature populations (either CD34+CD38 or CD34+) (Fig. 3, E and F, and fig. S12). Anti-CD99 mAbs were also cytotoxic to myeloid leukemia cell lines such as the AML-derived cell line MOLM13 (Fig. 3G). Anti-CD99 mAbs induced apoptosis, as confirmed by annexin V and activated caspase 3 staining (Fig. 3H and fig. S13), in the absence of immune effector cells or complement, consistent with a direct cytotoxic effect.

Because normal HSCs and endothelial cells express low and intermediate amounts of CD99, respectively (fig. S1), we tested whether cytotoxic anti-CD99 mAbs would affect their growth or survival. At concentrations toxic to AML and MDS cells, anti-CD99 mAbs had minimal effects on HSC growth in liquid culture supplemented with cytokines (Fig. 3I) and exerted no significant toxicity to human umbilical vein endothelial cells (HUVECs) (Fig. 3J). Ex vivo incubation of leukemic blasts from an AML patient previously shown to engraft in NSG mice (Fig. 2E and fig. S14) together with anti-CD99 mAb (clone H036-1.1) for 45 min before xenotransplantation completely abolished their engraftment capability (Fig. 4, A to C). To rule out a possible effect of anti-CD99 mAbs on LSC homing and to determine whether anti-CD99 mAbs can eradicate LSCs in vivo, we allowed NSG mice transplanted with leukemic blasts from the same AML specimen to engraft for 2 weeks and then treated them with 15 μg of anti-CD99 mAb (clone H036-1.1) or isotype control (Fig. 4A). One treatment with anti-CD99 mAb was sufficient to completely abolish AML engraftment, as measured in the BM at 5 months after transplantation (Fig. 4, B and C). When followed out to 11 months, none of the mice transplanted with AML cells treated with anti-CD99 mAb ex vivo or in vivo demonstrated engraftment. Conversely, treatment of normal CB HSCs with anti-CD99 mAb (H036-1.1) ex vivo or in vivo in the same manner did not have any detectable effect on engraftment in the BM or peripheral blood (PB) (Fig. 4, D and E).

Fig. 3. An anti-CD99 mAb is cytotoxic to MDS and AML cells.

(A) Incubation of purified CD34+ cells from MSK MDS-001 and MDS-002 with anti-CD99 mAb (clone H036-1.1) for 48 hours led to a marked decrease in cell number [median inhibitory concentration (IC50), 6.98 and 10.40 μg/ml, respectively]. Similar results were obtained with (B) CD34+CD38CD90CD45RA+ LMPP-like cells from AML specimens MSK AML-001 and UP31, as well as (C) CD34+ cells from MSK AML-004 (IC50, 10.40, 5.0, and 7.91 μg/ml, respectively). CD34+ cells from the BCR-ABL–positive AML MSK AML-005 were incubated with anti-CD99 mAb (H036-1.1) for 48 hours. The IC50 was not reached using mAb concentrations up to 35 μg/ml. (D) Incubation of bulk blasts (CD45lowSSClow) purified from UP32, UA8, UP4, UA16, UP34, and MSK AML-003 with anti-CD99 mAb (clone H036-1.1) for 48 hours resulted in a marked decrease in cell number (IC50, 26.24, 11.10, 31.40, 2.85, 16.81, and 26.74 μg/ml, respectively). (E) At the end of the incubation of MSK MDS-001 and MSK MDS-002 with anti-CD99 mAb, CD34 and CD38 expression was measured on remaining viable cells, demonstrating selective depletion of CD34+CD38 cells. Representative FACS plots are shown. P values were calculated by unpaired t test. (F) At the end of incubation of MSK AML-004, UP4, UA16, UP34, and MSK AML-003 with anti-CD99 mAb, CD34 and CD38 expression was measured on remaining viable cells, demonstrating selective depletion of CD34+CD38 cells (or CD34+ cells in UP4). Representative FACS plots are shown here and in fig. S12. P values were calculated by unpaired t test. (G) MOLM13 cells were incubated with anti-CD99 mAb (H036-1.1) for 48 hours, resulting in a marked decrease in cell number. (H) Incubation of MOLM13 cells with anti-CD99 mAb (H036-1.1, 20 μg/ml) induced apoptosis over the course of 36 hours. P values were calculated by unpaired t test. (I) Seven hundred HSCs (LN CD34+CD38CD90+CD45RA) purified from CB were incubated with anti-CD99 mAb (H036-1.1) for 48 hours. The IC50 was not reached using mAb concentrations up to 35 μg/ml. The sensitivity of MSK MDS-001 and MSK AML-001 to anti-CD99 mAb as shown in (A) and (B) is juxtaposed for comparison. (J) Similar results were obtained when HUVECs were incubated with anti-CD99 mAb (H036-1.1) for 48 hours. For (A) to (J), data points represent the means ± SEM of biological triplicates.

Fig. 4. Anti-CD99 mAbs exhibit antileukemic activity in vivo with relative sparing of normal HSCs.

(A) Schematic for combined ex vivo and in vivo anti-CD99 mAb (H036-1.1) treatment of AML specimen UP32. (B) Summary of engraftment in UP32 xenografts 5 months after transplantation (>0.1% threshold for human engraftment demarcated with dotted gray line). (C) Representative FACS plots of human engraftment in anti-CD99 mAb and isotype control–treated animals. (D) Schematic for combined ex vivo and in vivo anti-CD99 mAb (H036-1.1) treatment of normal CB HSCs (LN CD34+CD38CD90+CD45RA). (E) Summary of engraftment based on assessment of human CD45+ cells 2 months after transplantation and 1 month after antibody treatment. (F) Schematic for engraftment of primary AML specimens and in vivo treatment with anti-CD99 mAbs (H036-1.1 or 10D6). (G) Human leukemic chimerism in xenografts relative to pretreatment chimerism after 4 weeks of antibody treatment with the indicated anti-CD99 mAb clone or isotype. P values were calculated by Mann-Whitney U test. Representative FACS plots of human engraftment in the BM and PB before and after anti-CD99 mAb treatment. (H) Schematic for engraftment of mice with normal CB HSCs followed by treatment with anti-CD99 mAb (H036-1.1 or 10D6). (I) Summary of CB HSC-derived human engraftment in the BM and PB after 4 weeks of antibody treatment with the indicated anti-CD99 mAb clone or isotype. P values were calculated by Mann-Whitney U test. The scatterplots in (B), (E), and (I) show the means ± SEM.

To determine whether anti-CD99 mAbs have antileukemic activity against established AML, we transplanted NSG mice with four independent AML specimens and confirmed human leukemic engraftment in both the BM and PB (with the exception of specimen UP32, which demonstrated detectable engraftment in the BM, but not in PB) (Fig. 4F). Engrafted mice were treated for 4 weeks with the anti-CD99 mAb H036-1.1 (15 μg/ml three times weekly) or isotype control. Additionally, we generated another anti-CD99 mAb [clone 10D6, immunoglobulin G1κ (IgG1κ) isotype], which demonstrated direct cytotoxicity to AML cells (fig. S15), and we also treated engrafted mice with this mAb for 4 weeks (40 μg daily). Treatment with either anti-CD99 mAb (clone H036-1.1 or 10D6) reduced the leukemic burden in both the BM and PB (Fig. 4G). In contrast, treatment of NSG mice engrafted with normal CB HSCs with anti-CD99 mAbs (H036-1.1 or 10D6) had minimal effects on engraftment, with the exception of a modest reduction in PB chimerism with clone 10D6 (Fig. 4, H and I). Together, these results indicate that anti-CD99 mAbs preferentially induce death in AML disease stem cells, have a potentially large therapeutic window, and are a promising AML stem cell–directed therapeutic strategy.

Anti-CD99 mAbs activate SRC family kinases

To determine the mechanism of anti-CD99 mAb–induced cell death, we assessed the effect of anti-CD99 mAbs on SRC family kinase (SFK) activation because CD99 has been shown to negatively regulate SFK activation in osteosarcoma cells (35). We confirmed that CD99 negatively regulates SFK activation in AML cell lines, noting that short hairpin RNA–mediated knockdown of CD99 in the AML cell line MOLM13 induced SFK activation, whereas overexpression repressed SFK activation (Fig. 5, A and B). Cytotoxic anti-CD99 mAbs recapitulated the effects of CD99 knockdown, inducing rapid and robust SFK activation in both AML cell lines (Fig. 5C) and primary AML blasts (Fig. 5D). Pharmacologic inhibition of SFKs with the small-molecule inhibitor PP2 or dasatinib significantly attenuated anti-CD99 mAb–induced cytotoxicity in both AML cell lines (P = 0.0029; Fig. 5E) and primary AML blasts (P = 0.0037 and P = 0.019; Fig. 5F), confirming that anti-CD99 mAbs promote cell death in part by inducing SFK activation. Consistent with the deleterious effects of rapid SFK activation, we observed cell cycle arrest (Fig. 5G), as well as enrichment for gene expression signatures associated with DNA damage response, replication stress, and the unfolded protein response (Fig. 5H). The chronic myeloid leukemia blast crisis–derived cell line K562 was highly resistant to the cytotoxic effects of anti-CD99 mAbs (Fig. 5I and fig. S15). Similar to the resistant primary sample shown in Fig. 3C, this cell line harbors the constitutively active tyrosine kinase BCR-ABL, which promotes high basal activation of SFKs (36). Thus, resistance to anti-CD99 mAb–induced cytotoxicity is likely due to the ability of K562 cells to better tolerate acute SFK activation. To determine whether SFK activation is sufficient to induce cell death, we generated a constitutively active SRC mutant (Y530F) (Fig. 5J) and overexpressed it in anti-CD99 mAb–sensitive and mAb-resistant cell lines (MOLM13 and K562, respectively). MOLM13 cells exhibited a marked decrease in cell growth (Fig. 5K), cell cycle arrest (fig. S16), and increased apoptosis (Fig. 5L) in response to constitutive SRC activation, thereby phenocopying the effects of treatment with anti-CD99 mAb. In contrast, K562 cells demonstrated a modest decrease in growth and no increase in apoptosis (Fig. 5, K and L) or cell cycle arrest (fig. S16). Together, these findings suggest that anti-CD99 mAbs promote cell death by inducing SFK activation and oncogenic stress in myeloid leukemias that do not exhibit constitutive activation of SFKs.

Fig. 5. Anti-CD99 mAbs induce apoptosis by promoting SFK activation.

(A) MOLM13 cells were transduced with an shRNA targeting CD99. Western blot confirmed knockdown of CD99, which was accompanied by an increase in SFK activation, as measured by phosphorylated SRC (pSRC) (Y416). (B) MOLM13 cells were transduced to overexpress CD99 under a doxycycline-inducible promoter, and doxycycline (1 μg/ml) was added to the medium. Western blot confirmed overexpression of CD99 24 hours after addition of doxycycline, which was accompanied by a decrease in pSRC (Y416). (C) Incubation of MOLM13 cells with anti-CD99 mAb (clone H036-1.1, 20 μg/ml) induced rapid SFK activation. (D) Incubation of the indicated primary AML specimens with anti-CD99 mAb (clone H036-1.1, 36 μg/ml) induced rapid SFK activation. Numbers below Western blots represent densitometric quantification of displayed bands normalized to vector control (A and B), the 0-min time point (C), or isotype control (D). (E) MOLM13 cells were incubated with dasatinib (1 μM) or dimethyl sulfoxide (DMSO) for 8 hours before incubation with anti-CD99 mAb (H036-1.1, 5 μg/ml) for 48 hours. Anti-CD99 mAb induced a significant reduction in cell number that was partially rescued by dasatinib treatment. P values were calculated by unpaired t test. Bar graphs represent the means ± SEM of biological triplicates. (F) Primary AML blasts were incubated with PP2 (20 μM) or DMSO immediately before incubation with anti-CD99 mAb (H036-1.1, 36 μg/ml) for 48 hours. Anti-CD99 mAb induced a significant reduction in cell number that was partially rescued by dasatinib treatment. P values were calculated by unpaired t test. Bar graphs represent the means ± SEM of biological triplicates. n.s., not significant. (G) Incubation of MOLM13 cells with anti-CD99 mAb (H036-1.1, 20 μg/ml) for 36 hours results in a marked redistribution of cells from the G1 to the G0 or S/G2/M phases of the cell cycle. P values were calculated by unpaired t test. DAPI, 4′,6-diamidino-2-phenylindole. (H) MOLM13 cells were incubated with anti-CD99 mAb (H036-1.1, 36 μg/ml). After 24 hours, live cells (propidium iodide–negative) were FACS-purified, and RNA sequencing was performed. Gene set enrichment analysis revealed enrichment for gene signatures associated with DNA damage response (58), cell cycle arrest, replication stress, and the unfolded protein response (56, 59). (I) K562 cells were incubated with anti-CD99 mAb (H036-1.1) for 48 hours. The IC50 was not reached using mAb concentrations up to 17.4 μg/ml. The sensitivity of MOLM13 cells to anti-CD99 mAb, as shown in Fig. 3G, is juxtaposed for comparison. Data points represent the means ± SEM of biological triplicates. (J) A constitutively active SRC (Y530F) mutant was generated by site-directed mutagenesis. (K) MOLM13 and K562 cells were transduced to overexpress wild-type SRC or mutant SRC (Y530F) using a doxycycline-inducible system; SRC (Y530F) expression induced a greater decrease in cell growth in MOLM13 as compared with K562 cells. Growth curves show the means ± SEM of biological triplicates. P values were calculated by unpaired t test. (L) SRC (Y530F) expression induced apoptosis, as measured by activated caspase 3, in MOLM13 but not K562 cells. Bar graphs represent the means ± SEM of biological triplicates. P value was calculated by unpaired t test.

DISCUSSION

Both AML and MDS fulfill central tenets of the cancer stem cell hypothesis, whereby disease-initiating stem cells display the ability to self-renew and differentiate into non–self-renewing progeny that comprise most neoplastic cells (1, 79). Because disease-initiating cells are the only self-renewing cells in these diseases, therapies aimed at inducing durable remissions or cures must target these populations, which are resistant to current standard therapies (3, 8, 11). Previous studies have identified a number of antigens differentially expressed on AML LSCs (1217), but these markers are not specific for, or enriched on, functional LSCs (fig. S11). Examples of therapies that target cell surface antigens in AML include antibodies against CD47 to enhance phagocytic clearance of blasts (12), against CD44 to disrupt blast-niche interactions (13), and against CD123 to impair cytokine signaling (17). Numerous separate efforts to target LSCs with small-molecule inhibitors of self-renewal or survival pathways are also under way (37, 38). Despite these efforts, strategies to exploit specific AML LSC surface antigens have yet to be successfully translated to the clinic. Similarly, therapies directly targeting MDS stem cells have yet to be described, which is not surprising because very little information is available regarding aberrant antigen expression on MDS HSCs.

By examining a large number of primary AML and MDS patient specimens, we have identified CD99 as a cell surface antigen expressed in higher amounts on disease-initiating cells compared with normal HSPCs in most of the AML and MDS cases that we evaluated. Similar to a subset of previously defined AML-associated cell surface markers present on LSCs (12, 1416), differences in CD99 expression allow for the prospective separation of leukemic blasts from residual normal or preleukemic hematopoietic cells. This capability may eventually allow for effective purging of LSCs from autologous BM grafts that may be used clinically for therapeutic HSC transplantation. However, CD99 also exhibits the capability to enrich for functional LSCs within the leukemic blast population as demonstrated by limiting dilution xenograft assays, even when assessing the LN CD34+CD38CD90CD45RA+ LMPP-like fraction of AML, which is most highly enriched for LSC activity (5). This characteristic is not shared by previously described LSC markers and should allow for more refined studies of the functional and molecular features of LSCs in the future.

Given the high frequency of aberrant CD99 expression in our series of AML and MDS cases and the fact that these cases were not selected with previous knowledge of their cytogenetic and mutational profiles, our studies suggest that CD99 is likely to be an exploitable therapeutic target in most AML and MDS patients. The eradication of LSCs in vivo by anti-CD99 mAbs is likely due to the direct cytotoxic effect of the antibody, because IgM antibodies such as H036-1.1 have minimal capability to induce antibody-dependent cellular cytotoxicity (ADCC) (39), and NSG mice are deficient in hemolytic complement (40). Although we do not expect all myeloid leukemias to be sensitive to this single strategy, we speculate that it may be possible to enhance the therapeutic efficacy of anti-CD99 mAbs by combining their direct cytotoxic activity with other strategies, including conjugation with cytotoxic adducts (41, 42) or the derivation of bispecific antibodies to simultaneously target other leukemia-associated antigens or recruit cytotoxic T cells (43, 44). The robust and rapid activation of SFKs after anti-CD99 mAb binding suggests a mechanism of cytotoxicity via induction of oncogenic stress, as has been described for a number of oncogenes such as RAS (45), c-MYC (46), BCR-ABL (47), and SYK (48). Accordingly, overexpression of constitutively active SRC is sufficient to induce apoptosis in anti-CD99 mAb–sensitive MOLM13 cells, but not in K562 cells, which are relatively resistant to anti-CD99 mAbs and exhibit high basal activation of SFKs. One possible explanation for the relative resistance of K562 cells may be their lack of functional CDKN2A and TP53 (49), which are both critical mediators of oncogene-induced cell cycle arrest (45). Thus, although previous studies have suggested inhibiting the SFK pathway in AML as a potential therapeutic strategy (50), our study demonstrates that activation of SFKs may represent a therapeutic vulnerability in myeloid malignancies.

Although CD99 appears to be a robust marker of disease stem cells and a promising therapeutic target in AML, a number of unresolved issues remain. First, it is not clear whether CD99 is only up-regulated in fully transformed LSCs or whether it is also up-regulated in a subset of preleukemic HSCs. A number of recent studies have identified preleukemic HSCs that functionally give rise to normal hematopoiesis but may also harbor a subset of the full complement of somatic mutations, which are present in fully transformed leukemia (19, 22, 23). In our study, the CD99-negative fraction of a number of specimens completely lacked known preleukemic mutations, such as DNMT3A and IDH1/2 (19, 22, 23). Thus, it appears that CD99 may be up-regulated on preleukemic HSCs in some AMLs. Further characterization of CD99-positive and CD99-negative populations by comprehensive targeted sequencing and other approaches will be necessary to fully characterize the stage of leukemogenesis during which CD99 is up-regulated, as well as the mechanisms of its up-regulation. Identifying the spectrum of preleukemic HSCs that up-regulate CD99 will be important to assess whether anti-CD99 mAbs may also target this potential reservoir of preneoplastic cells that may reinitiate disease.

Second, although we demonstrate that immunophenotypically defined disease-initiating MDS HSCs (7, 9) express high amounts of CD99, it is unclear whether this high expression of CD99 is sufficient to prospectively separate MDS HSCs from residual normal HSCs. We were unable to assess this directly because of the limited cellularity of the MDS samples that we evaluated. Additionally, given the high burden of involvement of the HSC compartment by the MDS clone at diagnosis (>95 to 99% based on previous studies) (7, 8, 11), addressing this question may require evaluation of CD99 expression in the posttherapy and/or minimal residual disease setting, at which time MDS HSCs persist in detectable, but likely lower, quantities (8, 11). Single-cell transcriptomal/exome sequencing approaches may also have the potential to address this question by correlating differences in CD99 expression with somatic mutational status. Finally, it is unclear whether high CD99 expression enriches for disease-initiating function among MDS HSCs. We were unable to address this question experimentally because current MDS xenograft models require transplantation of relatively large numbers of cells (7, 10), thereby not allowing for limiting dilution analyses using samples with limited cellularity.

In sum, our evaluation of primary patient specimens and xenograft models reveals that CD99 is an AML LSC and MDS HSC marker that also identifies cells with functional disease-initiating capability, particularly in AML. Moreover, anti-CD99 mAbs effectively target AML LSCs, directly inducing their death in part by activating SFKs. These studies establish the use of anti-CD99 mAbs as a promising disease stem cell–directed therapeutic strategy and identify SFK activation as a molecular vulnerability in myeloid malignancies.

MATERIALS AND METHODS

Study design

Research objectives.

The primary objectives of this study were to characterize the frequency of expression of CD99 on disease-initiating stem cells in MDS and AML and to determine the potential of anti-CD99 mAbs as potential therapeutic agents. Our initial data confirmed our prespecified hypotheses that CD99 is frequently expressed on disease-initiating stem cells in MDS and AML and that anti-CD99 mAbs exhibit direct cytotoxicity to disease cells. On the basis of our initial data, we hypothesized that CD99 may also enrich for functional LSCs and that anti-CD99 mAbs may exert their cytotoxic effects by modulating cell signaling pathways regulated by CD99. Our secondary objectives were thus to validate CD99 as a cell surface marker that can be used to enrich for functional LSCs and to characterize cell signaling pathways perturbed by anti-CD99 mAbs.

Research subjects and study design.

For validation of CD99 expression, 24 MDS specimens and 79 AML specimens were analyzed (39 diagnostic specimens and 40 relapse specimens—for one patient, relapse specimens were available from two different time points). No patient samples were excluded from the analysis or presentation of results. Human AML and MDS specimens were obtained from patients at Memorial Sloan Kettering Cancer Center (MSKCC), the University of Pennsylvania, the University of Adelaide, and the University of Rochester with informed consent under institutional review board–approved protocols. Umbilical CB specimens were purchased from the New York Blood Center. Normal human BM specimens were purchased from AllCells Inc.

For limiting dilution analyses, at least 21 xenografts were analyzed for each AML specimen to ensure a large enough sample size for statistical comparison. Experimental groups (top 10% and bottom 10% of CD99-expressing LSCs) were transplanted into NSG mice in a randomized manner, with age and sex of animals balanced between groups. Animals that died within 1 week of transplantation were censored, but no other animals were excluded from the analysis. Outcome assessments were performed in a blinded manner (animals were assigned a random numerical code before analysis). In vitro cytotoxicity assays were performed on consecutive primary MDS and AML specimens based on availability and adequate cellularity, with no samples excluded from analysis or inclusion in results.

Experimental methods

FC and FACS.

All FACS and FC analyses were performed on a FACSAria II cell sorter (BD Biosciences). Analyses of leukemic blasts and LSCs in human AML specimens were performed using the following antibodies: CD19 (HIB29), CD3 (HIT3a), CD34 (581), CD45RA (HI100), CD47 (CC2C6), and CD44 (BJ18) from BioLegend; CD38 (HIT2), CD99 (3B2/TA8), CD123 (6H6), and TIM3 (F38-2E2) from eBioscience; and CD90 (5E10) from BD Biosciences. Given the expression of aberrant differentiation antigens on leukemic cells, we used an abbreviated lineage stain of CD3/CD19, with LN CD34+CD38CD90+CD45RA cells designated as HSCs, LN CD34+CD38CD90CD45RA cells as MPPs, and LN CD34+CD38CD90CD45RA+ cells as LMPPs.

Analysis of HSPCs in human MDS specimens and CB was performed using a full lineage stain, including CD2 (RPA-2.10), CD3 (HIT3a), CD4 (RPA-T4), CD7 (M-T701), CD8 (RPA-T8), CD10 (HI10a), CD11b (ICRF44), CD14 (TuK4), CD19 (CC2C6), CD20 (2H7), GPA (HIR2), and CD56 (B159), all from BD Biosciences. Definitions of HSCs, MPPs, and LMPPs/LSCs remained as described above.

Analysis of human and murine cells in xenograft experiments was performed using the following antibodies: mouse Ter119 (TER-119) and CD45.1 (A20) from eBioscience; human CD45 (2D1), CD99 (3B2/TA8), CD38 (HIT2), and CD33 (WM53) from eBioscience; and human CD34 (581) from BioLegend. Engrafted human cells were identified on the basis of their immunophenotype: mouse Ter119- and mouse CD45-negative, human CD45-positive, with confirmation of CD34, CD38, CD99, CD45RA, and CD33 expression depending on the immunophenotype of the transplanted AML.

Methylcellulose colony assays.

HSPCs (CD3CD19CD34+CD38) from primary AML specimens were double FACS–sorted to >95% purity and seeded in methylcellulose with myeloid/erythroid-promoting cytokines (MethoCult H4435; STEMCELL Technologies). Colony number and lineage were scored with an Olympus BX41 microscope after 14 days.

Genotyping of methylcellulose colonies.

Methylcellulose colonies were aspirated, and cells were resuspended in sterile water and heated to 95°C for 10 min. FLT3 mutational status was determined by polymerase chain reaction (PCR) using primers that yield a product of 329 base pairs for wild-type alleles and products of varying larger sizes for ITD alleles. Reactions were performed in 50-μl total volume containing 5 μl of template, 5 μl each of 10 μM forward primer 11F (5′-GCAATTTAGGTATGAAAGCCAGC-3′) and reverse primer 12R (5′-CTTTCAGCATTTTGACGGCAACC-3′), 1.25 U of Taq DNA polymerase, 50 mM KCl, 30 mM tris-HCl, 1.5 mM Mg, and 200 μM per deoxynucleotide. Thermocycling conditions were as follows: 30 s at 95°C, 30 s at 62°C, and 30 s at 72°C for 40 cycles. The presence of the BCR-ABL (e6a2) translocation present in MSK AML-005 was detected using the same reaction mixture with a forward primer (5′-GACTTCATTATCAGCTCAGAATGCACC-3′) and a reverse primer (5′-AGATACTCAGCGGCATTGCGG-3′). Thermocycling conditions were as follows: 30 s at 95°C, 30 s at 56°C, and 60 s at 72°C for 40 cycles. The mutational status of other genes was determined by Sanger sequencing of PCR fragments amplified from genomic DNA using the primers listed in table S3.

In vitro antibody incubation assays.

Cell lines were plated at a concentration of 2500 cells per 20 μl of RPMI 1640, 10% heat-inactivated fetal calf serum (Thermo Fisher), and penicillin (5000 U/ml)/streptomycin (5000 μg/ml) (Life Technologies) in a flat-bottom 384-well tissue culture–treated plate (Corning). Primary MDS, AML, and CB specimens were plated at a concentration of 700 to 2500 cells per 20 μl of StemSpan serum-free expansion medium (SFEM) (STEMCELL Technologies) supplemented with human LDL (40 μg/ml) (Sigma-Aldrich) and the following cytokines (PeproTech): Flt-3 ligand (100 ng/ml), stem cell factor (100 ng/ml), thrombopoietin (50 ng/ml), interleukin-3 (IL-3) (20 ng/ml), and IL-6 (20 ng/ml). HUVECs were plated at a concentration of 5000 cells per 20 μl of M199 medium (Thermo Fisher) containing 20% heat-inactivated fetal calf serum (Thermo Fisher), 15 mM Hepes (Sigma-Aldrich), endothelial cell mitogen (50 mg/liter) (Biomedical Technologies Inc.), heparin (50 mg/ml) (Sigma-Aldrich), 2 mM l-glutamine (CellGro), and penicillin (100 IU/ml)/streptomycin (100 μg/ml)/amphotericin B (250 ng/ml) (Life Technologies). For in vitro antibody incubation assays, we used anti-CD99 mAb clone H036-1.1 (Abcam) in sterile phosphate-buffered saline (PBS). H036-1.1 was added at the indicated concentrations up to 35 μg/ml, and cells were incubated for 48 hours at 37°C, 5% CO2. At the end of the treatment time course, cells were resuspended in buffer (PBS with 2% fetal calf serum) containing propidium iodide (100 ng/ml) (Sigma-Aldrich) and quantified using FC counting beads (BD Biosciences). The SFK inhibitor dasatinib (Selleck Chemicals) was resuspended in DMSO and used at a final concentration of 1 μM. The SFK inhibitor PP2 (EMD Millipore) was resuspended in DMSO and used at a final concentration of 20 μM.

Apoptosis assays.

Annexin V and caspase 3 staining were performed according to the manufacturer’s instructions (BD Biosciences). For cell cycle analysis, cells were stained with a LIVE/DEAD fixable dead cell stain (Life Technologies) according to the manufacturer’s instructions, followed by fixation and permeabilization (BD Cytofix/Cytoperm). Cells were then stained with Ki-67 (SolA15) from eBioscience for 30 min at room temperature, followed by 4′,6-diamidino-2-phenylindole (2 μg/ml, Sigma-Aldrich) at room temperature for 10 min.

Site-directed mutagenesis.

Mutant SRC (Y530F) was generated from a human SRC complementary DNA (cDNA) (CCSB-Broad Lentiviral Expression Library, Open Biosystems) using a QuikChange II XL Site-Directed Mutagenesis Kit according to the manufacturer’s instructions (Agilent Technologies) and the following primers: sense (5′-CCACCGAGCCCCAGTTCCAGCCCG-3′) and antisense (5′-CGGGCTGGAACTGGGGCTCGGTGG-3′). Successful mutagenesis was confirmed by Sanger sequencing, and the mutant SRC cDNA was cloned into a lentiviral expression vector (pLentiLox3.7_TetOn_st2) containing a doxycycline-inducible promoter. Lentiviral particles were generated using standard techniques (51). To induce ectopic wild-type SRC or SRC (Y530F) expression, doxycycline (50 ng/ml) was added to the medium.

Western blotting.

MOLM13 cells (5 × 106 to 10 × 106) were incubated with H036-1.1 or isotype control (20 μg/ml) for a 90-min time course. At each time point, cells were pelleted, washed in PBS containing sodium orthovanadate, and lysed using Pierce IP Lysis Buffer (Thermo Pierce) containing FOCUS ProteaseArrest (G-Biosciences) and Phosphatase Inhibitor Cocktail II (EMD Millipore). Western blotting was performed using anti-CD99 mAb clone EPR3097Y from Abcam, phospho-SFK (Y416) mAb clone D49G4, total SFK mAb clone 32G6, β-actin mAb clone 13E5, and HSP90 mAb clone C45G5 from Cell Signaling Technology. An electrochemiluminescent reagent (EMD Millipore) was applied to Western blots, and images were acquired using a GE ImageQuant LAS 4000 instrument (GE Life Sciences). Densitometry analysis was performed using GelQuant.NET image analysis software (http://biochemlabsolutions.com/).

Xenotransplantation assays.

All mice were housed in MSKCC animal facilities. All animal procedures were conducted in accordance with the Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committees at MSKCC. NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice 8 to 10 weeks of age were used for xenogeneic transplantation assays. FACS-sorted human cells were transplanted into sublethally irradiated NSG mice (185 cGy) via retro-orbital injection. Hu-CD34-NSG mice engrafted with human CB-derived HSCs (Jackson Laboratories) were also used for experiments shown in Fig. 4 (D and E). PB and BM analysis was performed by FC at the indicated times using the antibodies described above. Anti-CD99 mAb clone H036-1.1 or IgM isotype control was administered by retro-orbital injection, and anti-CD99 mAb clone 10D6 or IgG isotype control was administered by intraperitoneal injection.

RNA sequencing.

Primary AML specimens were FACS-sorted into TRIzol LS (Thermo Fisher). cDNA libraries were prepared using a SMARTer mRNA amplification kit (Clontech), and sequencing was performed using the HiSeq platform (Illumina) with 40 million paired-end reads per sample.

Statistics and data analysis

All FC data were analyzed using FlowJo (Tree Star). L-IC frequency estimations were calculated using a Poisson distribution probability calculator (L-Calc, STEMCELL Technologies; www.stemcell.com/en/Products/All-Products/LCalc-Software.aspx). RNA-sequencing data were aligned by Spliced Transcripts Alignment to a Reference (STAR) (52) to human genome hg37, and reads for each gene were counted by HTSeq (53). Differentially expressed genes were identified by DESeq2 (54), followed by gene set enrichment analyses (26). P values were calculated using two-tailed t tests (Prism 7, GraphPad Inc.). Estimated variation was taken into account for each group of data and is indicated as SE or SD in each figure legend.

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/9/374/eaaj2025/DC1

Fig. S1. CD99 is expressed in comparable amounts on adult BM HSCs and CB HSCs.

Fig. S2. CD99 is highly expressed on bulk AML blasts and further increased on LSC-enriched populations.

Fig. S3. CD99 expression distinguishes leukemic cells from residual normal or preleukemic HSPCs.

Fig. S4. CD99-negative CD34+CD38 cells from AML specimens lack the full complement of mutations present in bulk AML cells.

Fig. S5. CD99-positive CD34+CD38 AML cells engraft a lethal myeloid leukemia.

Fig. S6. CD99 expression enriches for functional LSCs.

Fig. S7. LSC-enriched AML cells with high expression of CD99 demonstrate improved survival in vitro.

Fig. S8. FACS purification effectively separates CD99-high and CD99-low LMPP-like cells from primary AML specimens.

Fig. S9. FACS purification effectively separates CD99-high and CD99-low bulk unfractionated AML blasts.

Fig. S10. CD99-high and CD-low AML blasts often exhibit differing frequencies of variant transcripts.

Fig. S11. Other described LSC markers are not enriched on LSC-enriched LMPP-like AML cells.

Fig. S12. Anti-CD99 mAbs selectively deplete CD34+CD38 or CD34+ cells from primary AML patient samples.

Fig. S13. Anti-CD99 mAb induces apoptosis in MOLM13 cells.

Fig. S14. Xenografted AML blasts express cell surface markers characteristic of human AML.

Fig. S15. Anti-CD99 mAb clone 10D6 is directly cytotoxic to MOLM13 cells.

Fig. S16. Expression of constitutively active SRC causes cell cycle arrest in anti-CD99 mAb–sensitive but not mAb-resistant cells.

Table S1. Clinical characteristics of primary MDS BM specimens.

Table S2. Clinical characteristics of primary AML specimens.

Table S3. Primers used for Sanger sequencing of CD99-negative methylcellulose colonies.

REFERENCES AND NOTES

  1. Acknowledgments: We thank F. Zhao, C. Miller, G. Kone, C. Li, J. Morowitz, G. In, L. Yu, and S. Zheng for technical support, as well as S. Rizvi and F. Weis-Garcia from the Antibody and Bioresource Core Facility at MSKCC for assistance with hybridoma maintenance and production of mAb clone 10D6. We thank M. Becker, R. D’Andrea, A. Brown, I. Lewis, and L. Bik To for help in accruing primary acute myeloid leukemia specimens. Next-generation sequencing protocols and sequencing were performed by the Integrated Genomics Operation at MSKCC. Funding: This work was supported by a Young Investigator Award from the Conquer Cancer Foundation of the American Society of Clinical Oncology, a U.S. Department of Defense Postdoctoral Fellow Award in Bone Marrow Failure Research (BM120096), and a Fellow Scholar Award from the American Society of Hematology (all to S.S.C.), as well as a Clinical Scientist Award from the Doris Duke Charitable Foundation, a Translational Research Program Award from the Leukemia and Lymphoma Society, and a Geoffrey Beene Grant from the Geoffrey Beene Cancer Research Center (all to C.Y.P.). Author contributions: S.S.C. and C.Y.P. designed the study. S.S.C., F.E.G.-B., R.L.L., M.C., and A.M.M. accrued primary human AML specimens. M.C. provided primary AML specimens with known engraftment potential. V.M.K. and S.S.C. accrued primary human MDS specimens. S.S.C. performed FACS of human samples and in vitro and xenograft assays. S.S.C. performed antibody treatment and biochemical studies. W.H. analyzed RNA-sequencing data. W.S.E. and M.T. performed biochemical studies. S.S.C. and C.Y.P. analyzed the data and prepared the manuscript with input from the other authors. Competing interests: C.Y.P., S.S.C., and M.T. are inventors on patent application (WO2016149682 A2) submitted by MSKCC that covers compositions and methods for targeting CD99 in hematopoietic and lymphoid malignancies. Data and materials availability: RNA-sequencing data have been deposited into the Gene Expression Omnibus (GSE86506).
View Abstract

Navigate This Article