Dear Editor,

Emerging evidence supports that ASXL1 mutation in myeloid neoplasia leads to aberrant protein gain-of-function rather than loss-of-function as initially thought [1]. Thus, a reassessment of the potential biologic relevance of the site of mutation in ASXL1 in patients with myeloid disease is warranted. To date, a large-scale comparison of patients with the c.1934dupG (p.G646Xfs*12) hotspot mutation vs. those with other ASXL1 mutations has not been performed. In addition, ASXL1 has a role in stabilizing the cohesin complex by means of a cohesin-binding motif (CBM) [2], and patients with CBM mutations have been insufficiently characterized. Thus, we sought to further characterize the clinicopathologic and genetic features of patients by ASXL1 mutation site using a large clinical dataset.

We retrospectively analyzed a cohort of 6,043 adults with a documented or suspected myeloid neoplasm and at least one mutation identified by FoundationOne®Heme testing between January 1, 2014, and August 15, 2021. The gene panel and coverage (Supplementary Table 1), as well as the sequencing methods (Supplementary Methods), have been previously published [3, 4]. The research-consented patient database was queried for specimens assigned a diagnosis category of acute myeloid leukemia (AML), myelodysplastic syndrome (MDS), myeloproliferative neoplasm (MPN), or MDS/MPN overlap. Logistic regression analysis confirmed significant genetic differences between AML, MDS, MPN, and MDS/MPN categories typical of each myeloid phenotype [5], supporting separation into these four groups for study purposes (Supplementary Fig. 1).

Pathogenic ASXL1 mutations were identified in 1,414 patients, occurring in 18% of AML and 26% of chronic myeloid neoplasms (Supplementary Table 2), and most occurred in the final exon (Fig. 1A). Twenty-eight (2.0%) patients had multiple ASXL1 mutations, and ASXL1 was the sole mutated gene in only 52 patients (3.7%). The most common ASXL1 mutation was c.1934dupG (p.G646Xfs*12), and this was the sole or dominant ASXL1 mutation in 520 cases (37%, collectively referred to as ASXL1c.1934dupG hereafter, Supplementary Fig. 2). The remaining 894 patients (63%) had mutations at other sites in the ASXL1 gene (ASXL1other).

Fig. 1: Mapping of ASXL1 mutations.
figure 1

A Lollipop plot of ASXL1 mutations by codon sites in all myeloid neoplasms (n = 1414). B Frequencies of ASXL1c.1934dupG and ASXL1other co-mutation with individual genes by phenotype (genes mutated in ≥5% of cases are shown). Gene names with frequency differences with a p-value < 0.05 are shown. ASXL1 Ensembl gene IDENSG00000171456.12; Ensembl transcript ENST00000375687.4 HARE-HTH HB1, ASXL, restriction endonuclease helix-turn-helix domain (codons 11–82), ASXH ASX homology domain (codons 237–361), Cohesin cohesin-binding motif (codons 401–587), Exon 13 codons 573–1541, PHD plant homeodomain zinc finger (codons 1500–39).

Full size image

There were no significant differences in age, sex, or genomic ancestry between ASXL1c.1934dupG and ASXL1other. We noted slightly fewer ASXL1c.1934dupG mutations in patients with MDS (ASXL1c.1934dupG:ASXL1other 0.48:1) compared to AML (0.65:1, p = 0.03) and MPN (0.60:1, p = 0.01) and those in whom ASXL1 was the sole mutation (Supplementary Fig. 3A). However, these trends may be due to our VAF-based reporting thresholds (see Supplementary Methods), as ASXL1 VAFs were lower in singly mutated patients and those with MDS (Supplementary Fig. 3B). Across all ASXL1-mutated patients, STAG2 mutations were more likely to be seen with ASXL1c.1934dupG (21% vs. 16%, p = 0.02), whereas SETBP1 mutations were more commonly co-mutated with ASXL1other (15% vs. 10%, p = 0.01).

More prominent differences emerged within phenotypic subsets (Fig. 1B). For instance, STAG2 mutations were strongly associated with ASXL1c.1934dupG in MPN, with an ASXL1c.1934dupG:ASXL1other ratio of 9:1 (p < 0.01). In contrast, two genes in AML, TP53 and SETBP1, had a significantly higher co-mutation rate with ASXL1other (TP53: 11% vs. 3% in ASXL1c.1934dupG, p < 0.01; SETBP1: 14% vs. 7%, p = 0.04). As such, we sought to determine whether other specific ASXL1 mutations were associated with TP53 or SETBP1 co-mutation in AML (Supplementary Fig. 4). Mutations at codon 693 were significantly more frequent in cases of ASXL1mutTP53mut AML compared with ASXL1mutTP53wt AML (22% vs. 3.5%, respectively, p < 0.01). Similarly, the p.R404* variant was more often seen in ASXL1mutSETBP1mut AML (7.3% vs. 0.05% of ASXL1mutSETBP1wt AML, p = 0.01), though the total number of mutations in each group was small (n = 3, n = 2, respectively). These results confirmed that the c.1934dupG variant occurs in a similar patient population to other ASXL1 variants. However, subset analysis supported that ASXL1c.1934dupG and ASXL1other may be associated with some phenotypic and co-mutational tendencies.

The ASXL1 protein directly interacts with core cohesin proteins between amino acids 401 and 587 [2], a region denoted the cohesin-binding motif (CBM) [6]. Thus, we also sought to characterize ASXL1 CBM mutations to investigate for potential relationships to cohesin gene mutations (Supplementary Fig. 5). Overall, only seven patients with CBM mutations had cohesin co-mutations (Fig. 2A). Mutations consisted of either STAG2 (n = 6) or RAD21 (n = 1) and occurred in all disease phenotypes. VAF data suggested cohesin mutations were subclonal to CBM mutations in three cases. Notably, in 5 of 6 CBM/STAG2 co-mutated cases, the CBM mutation clustered within a 21-codon region between amino acids 491 and 512.

Fig. 2: ASXL1 cohesin-binding motif mutations.
figure 2

A Lollipop plot of ASXL1 mutations in all patients with cohesin gene co-mutations within and just beyond ASXL1’s cohesin-binding motif (CBM). B Frequencies of co-mutation with individual genes in those with CBM and non-CBM ASXL1 mutations. VAF variant allele fractions. STAG2 VAFs in males were halved given this gene’s location on the X chromosome.

Full size image

The overall frequency of CBM mutations in cohesin-mutated patients (7/435, 1.6%) was significantly lower than in cohesin-wild type patients (76/979, 7.8%, p < 0.01). When restricting analysis to the ASXL1mut cohort, the proportion of CBM mutations was also lower in cohesin co-mutated cases (n = 7/244, 2.9%) than in ASXL1-mutated patients overall (n = 84/1414, 5.9%, p = 0.06). In these patients, cohesin mutations became more frequent just beyond the CBM, as we found that the frequency of cohesin co-mutation in the 7-codon region after the CBM (15%) was similar to the rate found throughout the remainder of the ASXL1 gene (18%). This further supported some degree of exclusivity between cohesin and CBM mutations. Lastly, we investigated ASXL1mut patients for differences in co-mutation frequencies of non-cohesin genes based on CBM mutation status (Fig. 2B) and found that cases with CBM mutations were significantly enriched for SETBP1, EZH2, or CUX1 co-mutation (each p < 0.05).

To our knowledge, our cohort represents the largest number of patients (n = 1414) with ASXL1-mutated myeloid disease analyzed to date. A reappraisal of the spectrum of ASXL1 mutation sites is indicated given the recent discovery that ASXL1 mutations result in a stable, truncated gain-of-function protein, which promotes leukemogenesis via deregulation of BAP1 [1], a key component of the polycomb repressive deubiquinase complex [7]. It also appears that ASXL1 mutation site has effects on downstream epigenetic changes. For instance, Binder et al. identified heterogeneous expression profiles in chronic myelomonocytic leukemia patients with different ASXL1 mutations [8]. In their study, the two patients with the most distal ASXL1 truncation sites at codons 695 and 957 had gene expression profiles intermediate between those patients with the c.1934dupG mutation and those who were ASXL1-wild type [8].

Few patients in our cohort had biallelic ASXL1 mutations, which is consistent with the gain-of-function mechanism resulting from ASXL1 mutation. Our data also confirmed the c.1934dupG occurs in a patient population with similar demographic and clinicopathologic features as non-c.1934dupG mutations, further supporting its pathogenicity in myeloid disease [9, 10]. However, in AML, patients with non-c.1934dupG ASXL1 mutations more frequently harbored TP53 and SETBP1 mutations, both of which are associated with adverse-risk in myeloid disease. It currently remains unclear if ASXL1 mutation site has an effect on patient outcomes. While in AML, patient outcomes appear similar between those with c.1934dupG mutations and non-c.1934dupG mutations [11], in primary myelofibrosis, c.1934dupG mutation is associated with worse overall survival [12, 13]. Notably, we found a very strong predilection for STAG2 co-mutation in c.1934dupG-mutated MPNs. Overall, results of our mutation site analysis raise the possibility that site of ASXL1 truncation may confer selective pressures for acquisition of different driver mutations. Thus, future study is warranted to investigate for potential prognostic impacts based on ASXL1 mutation site, and specifically with the presence of TP53, SETBP1, and STAG2 co-mutations.

While ASXL1’s role in chromatin modification is well-established, the gene also appears to function in cohesin complex stabilization. Li et al. [2] demonstrated direct protein-protein interactions between a cohesin-binding motif (CBM) in the ASXL1 protein and the gene products of RAD21, SMC1A, SMC3, and STAG2. Scarpa et al. subsequently analyzed a large AML cohort and found that, of 18 patients with an ASXL1 CBM mutation, none had a mutation in any cohesin gene [6]. In our study, CBM mutations were generally—but not entirely—mutually exclusive with cohesin mutations, occurring at a significantly lower rate than in cohesin-wild type patients. Independent of cohesin mutation status, we found that EZH2, SETBP1, and CUX1 mutations were significantly more likely to co-occur with CBM mutations than non-CBM ASXL1 mutations. These collective findings suggest cohesin co-mutations are selected against in patients with CBM mutations and that alternative non-cohesin mutation pathways may drive leukemogenesis when the CBM is compromised. Further investigation into the biologic implication of CBM mutations is warranted, however, as it is unknown if alterations in the CBM—which occurs before the final exon—result in protein loss-of-function or gain-of-function.

Important limitations in our study were the lack of available comprehensive data and rigorously defined pathologic diagnoses for the cohort. Thus, it is inevitable that some patients would not meet strict World Health Organization diagnostic criteria and that our diagnosis categories may be more clinically heterogeneous. Thus, while our results may not be representative of all patients with myeloid disease, this cohort nonetheless allowed for the unique opportunity to explore the genetics of a large number of patients with ASXL1 mutation.

In summary, our data from a large clinical cohort suggests that ASXL1 mutation site may be biologically relevant in patients with myeloid disease. As targeted treatment options emerge for patients with ASXL1 [14] and cohesin [15] mutations, further study is warranted to assess for potential therapeutic and prognostic implications in patients with these co-mutations.