The Genomic Landscape of Myeloid Neoplasms Evolved from AA/PNH

Up to 15% of AA patients (pts) treated conservatively with immunosuppression will evolve to myeloid neoplasia (MN), either MDS or AML, over a median time of 10 years regardless of response (0-18 years; n=238). The pathogenesis of MN secondary to AA is diverse and will often include antecedent clonal facilitating events that herald progression. Minor clones have been described in AA, some of which are not contributory to later evolution while other may result in subsequent progression. MDS evolution in inherited bone marrow failure (BMF) syndromes suggests that germ line (GL) alterations can be predisposing. In addition, progression to MN may reflect immune escape due to selection pressure e.g., through acquisition of HLA mutations. Here, we studied the molecular landscape of MN arising from AA, to better understand its pathogenesis and ultimately to develop measures of early detection, prevention, and therapeutic strategies.

Among 350 pts diagnosed with AA and PNH, 38 (11%) developed a secondary MN (sMN). Median age at AA/PNH diagnosis was 61 years (15-76). Almost all of pts who underwent transformation (89%) received a 1^st line treatment consisting of ATG+CsA in 85% of cases (ORR 59%; 21% CR and 38% PR) and 47% received more than one form of treatment, suggesting a lack/incomplete response or relapse. MDS was the most frequent diagnosis at evolution (77%), followed by AML (21%) and MPN (2%). Myeloid evolution was less common in pts with moderate AA (7% vs 14% in severe) or in the presence of a PNH clone (21% vs. 52% in non-progressors, p=.0003).

First we investigated GL alterations classified as Tier1 (9/38 pts) and Tier2 (11/38) based on their pathophysiological impact. Tier1 variants included NF1, CBLC, SBDS (n=2), and SAMD9L and overall were more frequently detected in del(7q) pts (76%, p=.0001). Tier2 included FA variants (BRCA2, FANCI, FANCD2; n=3). Of note, in sMN Tier1 variants were detected in 24% vs. 8% in non-evolved cases (p=.008) and none had concomitant Tier1/Tier2 configuration (0% vs. 9% in non-progressors, p=.05) or GATA2 variants.

Cytogenetic abnormalities were most frequent at the time of MN progression in 83% of cases, with chr. 7 alterations present in 47% of cases (-7, 35%; del(7q),12%), followed by complex karyotype (CK, 13%), involving chr.7 in 75% of cases. By comparison, -7/del(7q) are present in 7.5% of cases of our internal cohort of primary MN (p=.0001), but no differences in -7 and del7q distribution were seen.

A total of 148 somatic variants (myeloid and HLA panels) were found at the time of evolution in 34/38 sMN pts, with an average of 4.4 mutations/patient. ASXL1 (29% vs 14%, p=.02) and SETBP1 (15% vs 3%, p=.005) hits were more frequent in evolved cases while TET2 and TP53 mutations were less common as compared with pMN. Of note, sMN pts with CK harbored ASXL1 and TP53 mutations in 50% of cases. In a cross-sectional analysis of evolved cases studied at AA onset (n=17) and at myeloid evolution (n=35), somatic lesions in TET2, DNMT3A and ASXL1 genes were found in 5, 1 and 3 pts at baseline, respectively. If variants in TET2 and DNMT3A likely reflect antecedent CHIP, ASXL1 variants may have a role in driving myeloid progression as shown by the higher mutation rate in post AA cases. This hypothesis is further supported by the acquisition of subclonal chr7 abnormalities and by the overall higher clonal burden at sMN onset (median VAF 24% vs 43% respectively, p=.0001). When comparing pts with chr7 abnormalities with de novo counterpart, in sMN genes appeared most commonly mutated in ASXL1 (p=.02), SETBP1 (p=.0007), ETV6 (p=.02) and NF1 (p=.02), while TP53 mutations were less common. The intrinsic peculiarity of this -7/del(7q) sMN subset is also underlined by a different median survival time (12 vs 48 months in sMN vs pMN, respectively, p=.0002). The HLA mutational analysis available for 10 sMN cases showed the presence of somatic class I/II loci variants in 4/10 of progressors, including pts with chr7 abnormalities in 3/4 of cases. Of note, all class I HLA mutations were found in locus C. By comparison, in non-progressing AA pts HLA class I/II variants were found in 13% of pts.

Our results demonstrate that AA progression to MN has distinct molecular characteristics. The presence of HLA mutations suggests that immune escape or immune selection may play a role, while the presence of GL predisposition variants shows that they not only may facilitate AA but also clonal evolution as described from classic congenital BMF.