Targeting of AML-Leukemic Stem Cells with Monoclonal Antibodies

The overall prognosis for acute myeloid leukemia (AML) patients remains poor, with the 5-year disease-free survival rate less than 40% for patients less than 55 years of age. The outlook for older patients is even more gloomy, with the likelihood of survival less than 10% for patients older than 65 years of age, or younger patients presenting with unfavorable cytogenetics [1]. The long-term survival rates for adult AML, especially for the higher risk groups, have not improved dramatically since development of the 3 + 7 daunorubicin/cytarabine regimen with high-dose cytarabine consolidation [2], despite the availability of new agents [3]. The underlying reasons for this dismal outcome remain unclear.

Acute myeloid leukemia is characterized by an accumulation of undifferentiated and functionally heterogeneous populations of cells. Leukemia stem cells (LSCs) are a subpopulation of AML cells that have self-renewal ability similar to normal hematopoietic stem cells (HSCs). LSCs are positioned at the apex of the AML cellular hierarchy and continuously replenish the more committed clonogenic leukemic progenitors. These progenitors subsequently produce rapidly dividing myeloid blasts, which present as the bulk of the disease. LSCs are functionally defined by their ability to re-establish a heterogeneous leukemic xenograft in immunodeficient mice, and are commonly found enriched in the CD34⁺/CD38^- fraction for the majority of AML samples [4], although LSCs have also been identified in the CD38⁺ fraction [5].

So how can we rationalize the clinical significance of LSCs, the existence of which is fundamentally an experimental concept? Recently, several studies suggest that LSCs may be central to post-treatment relapse and chemoresistance. In an AML xenograft model, cytarabine reduced overall AML engraftment in the bone marrow, but was less effective in reducing the number of CD34⁺/CD38^- AML cells in the osteoblast-rich endosteal region [6]. In addition, the relative quiescence of LSCs compared with other AML progenitor fractions implies that LSCs are more likely to be resistant to standard chemotherapy [7]. These data suggest that strategies that only target rapidly cycling blasts are unlikely to be curative in AML. Moreover, the clinical relevance of the LSC concept is reinforced by correlations between poor prognosis and either the ability of AML samples to engraft in immunodeficient mice [8] or high frequency of CD34⁺/CD38^- cells [9].

Effective therapeutic targeting of AML–LSCs requires the identification of characteristics that are distinguishable from normal HSCs. NF-κB is reported to be constitutively active in CD34⁺/CD38^-/CD123⁺ AML cells but not in normal CD34⁺/CD38^- cells. Inhibition of NF-κB via targeting proteasomes with MG-132, was effective at inducing preferential apoptosis of the leukemic cells [10]. In another recent study, STAT5 activation and expression of the MN1 and HOXA9 genes were reported to contribute to the ability of LSCs to initiate and sustain the disease [11]. Gene-expression arrays comparing LSCs and HSCs have also identified potential targets for therapeutic intervention [12], although the development of small molecule therapeutics that exploit these differences is in its infancy. By contrast, the differential cell-surface expression of various proteins and receptors between LSCs and HSCs has been relatively well characterized [13], and exploitation of these differences with monoclonal antibodies (MAbs) has the potential for immediate clinical benefit.

The IL-3 receptor α chain (CD123) is widely reported to be overexpressed on LSCs but not on normal HSCs [13–16]. High CD123 expression in AML was also associated with a poor prognosis [14,16]. A neutralizing MAb against CD123 inhibited IL-3-mediated survival of AML–LSCs in vitro, as well as homing, engraftment and serial transplantation of AML cells in immunodeficient mice, with lesser effects on normal hematopoietic cells [15]. This study also demonstrated that the innate immune system plays a critical role in mediating part of the antileukemic effects, raising the possibility that this, and other MAbs, can target LSCs by multiple mechanisms including antibody-dependent cellular cytotoxicity (ADCC). A humanized version of the anti-CD123 MAb is currently undergoing a Phase I clinical trial.

Gemtuzumab ozogamicin (GO) is an anti-CD33 MAb conjugated to the cytotoxic drug calicheamicin, and is already approved for the treatment of AML. GO showed limited benefit as a single agent against AML in older patients, although it was effective in extending relapse-free survival when used in combination with standard therapy in patients with favorable or intermediate cytogenetics [3]. While GO was not specifically developed to target LSCs, CD33 is expressed on LSCs in CD33⁺ AML, but not on HSCs [17]. However, another study showed that a population of CD34⁺CD33^- LSCs exists for most AML patients [18], which may explain the lack of definite effects of GO in the clinic. Another interesting lesson from the GO clinical experience is that it causes hepatic sinusoidal obstructive syndrome, presumably associated with adverse effects on CD33⁺ Kupffer cells [19]. This illustrates that the translation of MAb usage to the clinic will not be simple, but additional efficacy can be produced with appropriate combination.

CD47 is a transmembrane ligand for the signal regulatory protein a (SIRPα) receptor found on macrophages and dendritic cells. The interaction of CD47 with SIRPα inhibits phagocytosis. CD47 was reported to be upregulated on AML–LSCs relative to HSCs and higher expression in AML was associated with poor clinical outcome [20]. MAb-mediated disruption of the CD47–SIRPα interaction led to increased phagocytosis of AML–LSCs but not HSCs, and the antibody was effective at specifically inhibiting AML–LSC engraftment and serial transplantation in a newborn NOG model, which is more immunodeficient than the NOD/SCID mouse model [20]. Interestingly, this report suggested a lack of ADCC mediated by the CD47 MAb, in contrast to the CD123 [15] and FLT3 neutralizing MAbs [21]. This may have been due to the use of different immunodeficient mouse models. Another interesting observation from this study was that normal HSCs, despite expressing lower levels of CD47, evaded phagocytosis by an unknown mechanism [20], suggesting that normal and leukemic cells interact differently with the immune system.

Monoclonal antibodies against the adhesion-related molecules CD44 [22] and CXCR4 [23] have also demonstrated efficacy against AML–LSCs in NOD/SCID mouse models. The antileukemic effects were potentially mediated by preferentially disrupting the interactions between AML cells and the stromal survival microenvironment. MAbs against C-type lectin-like molecule-1 [24] and IREM-1 [25] are also being developed to similarly exploit the finding that those targets were overexpressed on LSCs but not on HSCs. These MAbs were not described as neutralizing antibodies, but exclusively relied on ADCC and complement-dependent cytotoxicity (CDC) to mediate their antileukemic effects [24,25].

So, what critical attributes should be considered when selecting appropriate targets for exploitation with MAb therapy? First, the protein should be significantly more highly expressed on LSCs compared with HSCs and other normal cells of the body – the high density of epitopes on LSCs and leukemic blasts can potentially mediate ADCC and CDC [15,21,24,25]. Second, an association with worse prognosis is desirable, but not essential. Third, neutralization of the target protein function by the MAb should preferentially inhibit growth or survival advantages to the cancer cells but not any normal cells. While the last point appears obvious, it is often overlooked.

In summary, while the role of LSCs in AML remains predominantly an experimental definition, chemoresistance and relapse are clinical realities and remain significant barriers to the long-term survival of patients. MAb therapies, apart from mediating their described LSC-targeting effects by neutralization, can also act via ADCC, CDC and antibody-dependent cell-mediated phagocytosis (ADCP), all of which can theoretically circumvent chemoresistance to deliver more durable remissions. However, MAbs are unlikely to completely replace chemotherapy in the very near future, and we need to understand how best to utilize them in combination with established therapy. Possible synergistic or antagonistic interactions between LSC-targeting MAbs and chemotherapy have not been rigorously tested in preclinical models, and this should be done so as a matter of priority to guide future clinical trials. Furthermore, the potential of combining different MAbs should also be examined, where multiple pathways such as disruption of adhesion and leukemic trafficking (CXCR4 [23], CD44 [22]) and inhibition of survival advantage (FLT3 [21], CD123 [15]) can be targeted simultaneously, while delivering additional antileukemic effects from ADCC, CDC and ADCP [20]. Combining different MAbs may also target a larger proportion of the heterogeneous AML population, overcoming possible clonal selection as well as evasion by epitope downregulation, as has been reported in lymphoma after treatment with CD20-targeting rituximab [26]. An AML patient surface immunophenotype is relatively cost-effective to characterize, raising the prospect of individualized therapy based on a selection of available MAbs. Most certainly, we are entering a new and exciting era in the struggle to improve outcome in adult AML.

Financial & competing interests disclosure

The authors wish to acknowledge funding support from Children’s Cancer Institute Australia for Medical Research and CSL Limited. RBL is a consultant for CSL Limited. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.

Additional information

Funding

The authors wish to acknowledge funding support from Children’s Cancer Institute Australia for Medical Research and CSL Limited. RBL is a consultant for CSL Limited. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.