Role of ASXL1 in Hematopoiesis and Myeloid Diseases

Abstract

Next-generation sequencing technology (NGS), including whole-exome or whole-genome sequencing and target gene sequencing, has allowed the molecular characterization of somatic mutation spectrums in hematologic diseases. Mutations in Additional sex combs-like 1 (ASXL1), a chromatin regulator, are identified in clonal hematopoiesis of indeterminate potential (CHIP), indicating ASXL1 mutations as early events in leukemogenesis. Not surprisingly, they occur at high frequency in myeloid malignancies and associated with poor prognosis. Therefore, understanding how mutant ASXL1 drives clonal expansion and leukemogenesis will serve as the basis for future development of preventative and/or therapeutic strategies for myeloid diseases with ASXL1 mutations. Here, we discuss the biology of ASXL1 and its role in controlling normal and malignant hematopoiesis. In addition, we review the clinical relevance of ASXL1 mutations in CHIP and myeloid diseases.

Introduction of ASXL1 gene and protein

Mammalian ASXL family genes (ASXL1, ASXL2 and ASXL3) are the mammalian homologs of Drosophila Additional sex combs Asx (1). Asx deletion leads to a homeotic phenotype characteristic of both Polycomb group (PcG, repressive complex associated with H3K27me3) and Trithorax group (TrxG, activating function associated with H3K4me3) gene deletions (1–3). Both Asxl1 and Asxl2 expression is virtually ubiquitous throughout embryogenesis and in adult tissues, whereas Asxl3 expression is more restricted and only detectable in lymph node, eye, lung, skin, brain, and pituitary gland (4).

The human ASXL1 gene is located on the chromosome 20q11.21 and encodes a 1541 amino acid protein (Figure 1) (5). ASXL1 contains an ASXN domain in the N-terminal region, an ASX homology (ASXH) domain in the N-terminal adjoining region, and a plant homeodomain (PHD) in the C-terminal region. ASXL family proteins share highly conserved ASXN, ASXH and PHD domains. The ASXN and PHD domains are putative DNA and histone binding domains, respectively. The ASXH domain (also referred to as DEUBAD, deubiquitinase adaptor) interacts with a partner protein BAP1 to confer deubiquitinase activity, leading to gene repression (6). At the endogenous level, truncated ASXL1 proteins resulting from ASXL1 mutations are rapidly degraded, and the ASXL1-BAP1 complex is undetectable (7). By contrast, overexpression of truncated ASXL1 increases the stability of BAP1 and enhances the deubiquitination activity of ASXL1-BAP1 complex. It is unclear whether overexpression of mutant ASXL1 recapitulates its function at the physiological level (8, 9).

Figure 1.

Schematic representation of structure of wildtype ASXL1 and mutant ASXL1.

In addition to BAP1, ASXL1 interacts with core polycomb repressive complex 2 (PRC2) components EZH2 and SUZ12, which are involved in the deposition of H3K27me3 histone repressive marks (7). The functions of the long stretch of amino acids between the ASXH and PHD domains have been poorly understood. A recent study revealed that the C-terminal intrinsic disordered region is important for the formation of nuclear paraspeckles. Deletion of this region disrupts these paraspeckles, leading to the attenuated repopulation capability of hematopoietic stem cells (HSCs) (10). Interestingly, ASXL1 mutations identified in myeloid diseases are predominantly located within this region, generating C-terminal truncated proteins (see below).

The functions of the other ASXL proteins are less known. ASXL2 has been shown to be essential for cardiac function and bone development (11–13). Recent studies revealed a high-frequency of ASXL2 mutations in acute myeloid leukemia (AML) patients bearing the RUNX1::RUNX1T1 (AML/ETO) fusion. Loss of Asxl2 in mouse leads to development of myelodysplastic syndrome (MDS)-like disease and promotes leukemogenesis driven by RUNX1::RUNX1T1 (14, 15). Unlike ASXL1 and ASXL2 mutations, ASXL3 mutations have not been detected in AML patients (16).

ASXL1 mutations in CHIP and myeloid diseases (clinal relevance)

Mutations in ASXL1 are identified in clonal hematopoiesis of indeterminate potential (CHIP) and significantly associated with smoking (17, 18). CHIP initially referred to the expansion of peripheral blood cells derived from hematopoietic stem cells (HSCs) with at least one somatic driver mutation in healthy elderly individuals (19–21). CHIP is strongly linked to aging and confers to an increased risk for blood cancers, non-hematological diseases (e.g. cardiovascular disease), and all-cause mortality (19–23). Although CHIP confers an approximately 10-fold increased risk to develop hematologic malignancies, such risk remains low (0.5-1% per year) (19). Therefore, it cannot explain the increased overall mortality associated with CHIP. A cause-specific mortality analysis revealed that non-leukemic mortality (e.g. cardiovascular diseases) in CHIP patients is higher than that due to blood cancers (21).

DNMT3A, TET2, and ASXL1 are among the most frequently mutated genes in CHIP. They are associated with initiation of acute myeloid leukemia (AML) and other myeloid diseases. Corroborating the human data, HSCs with Tet2 or Dnmt3a mutations robustly undergo expansion in transplant recipient mice (reviewed in (24)). Subsequently, CHIP was identified in patients previously treated for solid tumors and myeloid malignancy-associated CHIP mutations were also present in patients with lymphoid malignancies (24). The CHIP mutation spectrum in these patients is distinct from that in healthy individuals.

The selection and expansion of preleukemic-HSC clones precede the development of myeloid leukemia. Not surprisingly, ASXL1 mutations (and 20q deletion) are frequently identified in myeloid malignancies, in particular ~20% in MDS, ~45% in chronic myelomonocytic leukemia (CMML), ~10% in myeloproliferative neoplasms (MPNs), and ~20% in AML (25–28). Interestingly, ASXL1 mutations identified in CHIP are enriched around codons R404 (nonsense), Y591 (nonsense/frameshift), H630 (frameshift), and R693 (nonsense/frameshift). By contrast, ASXL1 mutations identified in myeloid diseases (including MDS and CMML) are predominantly frameshift mutations around codon G646 (G646: 18%; codon 630-660: 42%) (Figure. 2). Controversy has surrounded molecular testing of c.1934dupG p.Gly646fs ASXL1 variant. Its location within an 8 base-pair guanine mononucleotide repeat sequence made it suspicious for an artifact of PCR and/or sequencing rather than a true somatic mutation (29). However, the variant allele frequency of this mutation is >5% in many cases, arguing against PCR artefacts. Moreover, subsequent reports using NGS sequencing confirmed that the ASXL1 c.1934dupG is only detected in leukemia cells, but not in matched germline samples or healthy controls (30–33).

Figure 2. Distinct ASXL1 mutation spectrums in CHIP vs myeloid diseases.

This figure summarizes published datasets of CHIP in healthy elderlies and cancer patients and of patients with myeloid diseases (including MPN, MDS, CMML, and AML).

Clearly, the ASXL1 mutations around codon G646 are prevalent in myeloid diseases but much less common in CHIP. Similarly, the hotspot DNMT3A R882H mutation in AML is rarely seen in CHIP (34). We and others hypothesize that unlike majority of CHIP mutations that are fairly stable and less pathogenic in elderly patients, the hotspot ASXL1 and DNMT3A mutations represent pathological CHIP mutations with high risk for accumulating additional driver mutations and developing myeloid diseases. In support of this idea, CMML cells with hotspot ASXL1 mutations around G646 display distinct transcriptomic changes from normal BM cells and these changes are absent in CMML cells with non-hotspot ASXL1 mutations (35). Asxl1^−/− and Asxl1 G643Wfs (corresponding to human G646Wfs) knockin mice develop MDS and a fraction of them transform to myeloid leukemia (36, 37) (see below).

ASXL1 germline mutations and Bohring-Opitz syndrome

Bohring-Opitz syndrome (BOS) is a rare genetic disorder first reported by Bohring et al. in 1999, to describe four individuals with Opitz trigonocephaly (C)-like syndrome (38). BOS is a clinically recognizable syndrome characterized by facial dysmorphism, microcephaly, limb anomalies, postnatal failure to thrive, severe developmental delays and intellectual disability. To date ~100 cases have been described, almost half of which were molecularly confirmed to carry a heterozygous constitutive ASXL1 mutation, suggesting that constitutive mutations in ASXL1 are a major cause of BOS. Similar as CHIP and myeloid diseases-associated ASXL1 mutations, most of BOS-related ASXL1 mutations are de novo nonsense or frameshift mutation. Emma Bedoukiann and colleagues presented the first report of BOS caused by a pathogenic ASXL1 mutation inherited from a germline mosaic mother (39). Later, Karen Seiter and colleagues reported that a father and son were found to have the identical ASXL1 mutation (40), supporting the diagnosis of a germline mutation of ASXL1. Both of them developed AML without BOS symptoms. Therefore, how the same germline ASXL1 mutations cause different diseases remains unknown.

Biological function of Asxl1 (mouse work)

To evaluate the functions of Asxl1 in hematopoiesis and leukemogenesis, five different mouse models have been generated using different approaches (36, 37, 41–43). Conditional Asxl1 knockout mice were created to study loss of Asxl1 function in adult hematopoietic system (36). Asxl1^−/− bone marrow (BM) cells display increased number of HSCs and decreased re-plating capability as compared to wildtype (WT) cells. Upon transplantation, Asxl1^−/− BM cells show reduced reconstitution in young recipients. Deletion of Asxl1 leads to significant down-regulation of H3K27me3 due to loss of ASXL1-mediated recruitment of PRC2 key components, such as EZH2, to the chromatin (7, 36). In addition, a novel ASXL1-OGT(O-GlcNAc transferase) axis was identified to regulate H3K4 methylation in myeloid malignancies (44). ASXL1 interacts with HCFC1 and OGT and is stabilized via OGT-mediated O-GlcNAcylation. Disruption of this novel axis inhibits myeloid differentiation and H3K4 methylation(44). Consistent with the previous results, we reported that global H3K4me1, H3K4me3, and H3K27me3 levels were significantly decreased in Asxl1^−/− BM cells (45). Although global H3K27Ac level in Asxl1^−/− BM cells was comparable to that in control cells, H3K27Ac level was increased at specific gene loci.

Two transgenic overexpression models use different exogenous promoters (Rosa26 vs Vav1) to drive the transcription of different Asxl1 mutants (E635RfsX15 vs Y558X) (42, 43). Therefore, it is difficult to compare and interpret their results. Nonetheless, these transgenic overexpression models and in vitro overexpression studies (46) suggest that ASXL1 mutations may be dominant negative or gain-of-function. However, it is questionable whether these overexpression studies truefully reflects the physiology function of truncated ASXL1 proteins.

To overcome this problem, two groups independently generated Asxl1^tm knock-in mouse models (37, 41). In both models, the same Asxl1 guanine duplication was introduced into the endogenous Asxl1 locus, closely resembling patient-derived ASXL1 G646WfsX12 mutation. This hotspot frameshift mutation creates a truncated protein of 655aa (658aa in human) in contrast to the full length ASXL1 protein of 1514aa (1541aa in human). Studies with these two knock-in mouse models yielded highly consistent results, some of which are distinct from Asxl1^−/− data. For example, in comparison to WT cells, Asxl1^tm/+ BM cells exhibit reduced number of HSCs, increased re-plating capability, and largely comparable reconstitution in young recipients, suggesting that in addition to losing part of WT ASXL1 functions, Asxl1^tm instills some new functions. However, it remains unclear what epigenetic alterations this mutation causes and how this mutation could drive CH in humans.

Genetic interaction of ASXL1 with NRAS

ASXL1 mutations frequently coexist with other mutations, such as TET2 (47), RUNX1 (48), SETBP1 (49–51) and NRAS (25–28). Asxl1 loss in mice results in MDS that could transform to myeloid leukemia with age, suggesting that Asxl1 deficiency cooperates with additional mutations to induce myeloid leukemias.

ASXL1 mutations predict inferior outcomes in all myeloid diseases (26, 52, 53). They significantly co-occur with NRAS mutations in CMML (25–28). We showed that concurrent ASXL1 and NRAS mutations define a population of CMML patients with shorter leukemia-free survival compared to patients with ASXL1 mutation only (45). Corroborating these human data, we discovered that Asxl1^−/− accelerates CMML progression and promotes CMML transformation to AML (secondary AML, sAML) in Nras^G12D/+ mice. Although Nras^G12D/+; Asxl1^−/− (NA) model shares common genetic mutations with the published Nras^G12D/+; Ezh2^−/− (54) and Nf1^+/−; Asxl1^+/− (55) models, it displays distinct phenotypes and molecular mechanisms from the other two. Nras^G12D/+; Asxl1^−/− (NA) leukemia cells exhibited hyperactivation of MEK/ERK signaling and increased global level of H3K27Ac, a histone mark bound by bromodomain and extra-terminal domain (BET) proteins for gene transcriptional activation (45). NA-sAML cells were more immunosuppressive than NA-CMML cells and overexpressed all the major inhibitory immune checkpoint ligands, PD-L1/L2, CD155, and CD80/86 (45). Among them, overexpression of PD-L1 and CD86 correlated with upregulation of AP-1 transcription factors (TFs) in NA-sAML cells (45). An AP-1 inhibitor and shRNAs against AP-1 TF Jun decreased PD-L1 and CD86 expression in NA-AML cells. Once NA-sAML cells were transplanted into syngeneic recipients, NA-derived T cells were not detectable (45). Host-derived wildtype T cells overexpressed inhibitory immune checkpoint receptors, PD-1 and TIGIT, and displayed an exhausted T cell phenotype (45). Combined inhibition of MEK and pan-BET proteins led to downregulation of AP-1 TF expression, mitigation of the suppressive immune microenvironment, enhancement of CD8 T cell cytotoxicity, and prolonged survival in NA-sAML mice. Given the distinct phenotypes observed in Asxl1^−/− and Asxl1^tm/+ mice, it would be interesting to determine if Nras^G12D/+; Asxl1^tm/+ mice display similar phenotypes as NA mice and if the underlying mechanisms are distinct.

Treatment response of ASXL1 mutant leukemic cells

Recent studies revealed that patients with ASXL1 mutations are associated with distinct sensitivity to drug treatment. Hypomethylating agents (HMA) have been a standard treatment for CMML. A retrospective study of 177 CMML patients revealed that ASXL1 mutations predict a lower overall response rate to HMAs (azacitidine or decitabine) (ASXL1^mt 42% versus ASXL1^wt 60%, p = 0.02) (58). This clinical observation may be explained, at least partially, by the increased expression of anti-apoptotic gene BCL2 and elevated global cytosine methylation in ASXL1^mt leukemia cells (59). Not surprisingly, combined veneteoclax, a selective BCL2 inhibitor, and azacitidine effectively inhibit ASXL1^mt leukemia cell growth in vitro (59). This combination was approved by FDA to treat AML in 2018. It would be interesting to see if it treats ASXL1^mt CMML patients better than HMA alone.

In summary, ASXL1 hotspot mutations around codon G646 are prevalent in myeloid diseases but rarely identified in CHIP, suggesting that they are highly pathogenic and confers higher risk to develop myeloid diseases. The nature of these mutations remains elusive. While mouse genetic studies suggest that they are loss-of-function and gain-of-new function at the physiological level, overexpression studies in transgenic mice and cell lines indicate that they are dominant negative and gain-of-function. Additional rigorous investigations are needed to provide a definitive answer to this question. ASXL1 mutations are associated with poor prognosis in all myeloid diseases, perhaps due to the reduced response to current treatment options (e.g. HMA in CMML). We recently discover that concurrent ASXL1 and NRAS mutations define dismal outcomes in CMML patients. Correspondingly, Asxl1 loss cooperates with oncogenic Nras in mice to reprogram immune microenvironment and drive leukemic transformation. Our study provides a strong rational to develop combined targeted therapy and immunotherapy for treating leukemia patients with concurrent ASXL1 and NRAS mutations.