Molecular Signatures That Predict Response to Azacitidine Treatment for Myelodysplastic Syndromes

DNA hypomethylating agents (HMAs) constitute standard therapy for higher-risk myelodysplastic syndromes (MDS). However, the lack of reliable genetic markers predicting response to HMAs precludes an optimal use of these agents. Recently, several groups have reported marked efficacy of decitabine for TP53 -mutated myeloid malignancies. However, since decitabine is not approved for MDS in many nations, it is of great interest whether azacitidine (AZA) has similar effects on TP53 -mutated MDS. Thus, we investigated the clonal dynamics and molecular signatures that correlated with response to AZA in MDS.

We enrolled a total of 131 cases of MDS patients treated with azacitidine, of which 107 were registered from a prospective multicenter trial of AZA treatment for high-risk MDS patients, while the remaining 24 cases received AZA in off-trial settings. In all patients, a bone marrow specimen was obtained before (pre-therapy samples) and subjected to targeted-capture sequencing, which was designed to detect mutations in 67 known driver genes implicated in myeloid neoplasms as well as genome-wide copy number alterations (CNAs). Mutations and CNAs was also evaluated in samples obtained during and after AZA therapy (post-AZA samples).

Analysis of pre-treatment specimens revealed 408 driver mutations in 124 (95%) patients with a mean of 2.7 mutations/sample in 50 genes. At least one CNA was detected in 103 (79%) cases. Most frequently observed lesions included mutations in TP53, ASXL1, RUNX1, TET2, and SRSF2, and other high-risk CNAs, such as -7/7qLOH, 5qLOH, and 17pLOH. Among these TP53 lesions were found in 39 (38%) cases, of which biallelic lesions were suggested for 33 cases (85%), either by multiple mutations (N=11) or accompanying 17p LOH involving the TP53 locus (N=22). TP53 -mutated cases had a significantly lower number of other driver mutations (1.7 vs. 3.1, p<0.001) and a higher number of CNAs (9.6 vs. 2.1, p<0.001), compared with TP53 -unmutated cases. The best clinical responses during AZA treatment were 13 CR (12%) and 19 marrow CR (17%). No genetic lesions were significantly associated with response. However, CR were almost exclusively seen in TP53 -mutated cases (10/13, odds ratio=9.1, p<0.001). Furthermore, TP53 mutation was associated with a higher frequency of hematological improvement in erythroid lineage (HI-E) (56.7% vs. 21.7%, p=0.001). Clonal dynamics were successfully evaluated using post-AZA samples in 72 (55%) cases. For the 261 mutations detected in either pre- or post-AZA samples, the reduction in CNA-adjusted variant allele frequency (aVAF) showed a positive correlation with clinical response (p=5.59×10^-5, trend test), where aVAF in CR showed <0.20 in all but one case. Despite the larger aVAF reduction for TP53 mutations than other mutations before AZA treatment (0.59 vs 0.40, p=0.003), the reduction in clone size was more prominent in the TP53 mutations (ΔaVAF: -0.25 vs -0.019, p=0.004). TP53 mutation in pre-treatment samples was the only significant marker that predicted a reduction in tumor burden (q=0.008). However, the higher response rate in TP53 -mutated cases was not translated in to durable response with AZA alone, with a median duration of response of 217 days (range 10-783), although we identified two TP53 -mutated cases showing continuous remission for >3 years after receiving consolidation with transplant after AZA.

Our study revealed a significant positive association of TP53 mutations with favorable responses to AZA in MDS. However, the response was transient, where the development of effective post-AZA therapy after CR should be warranted, such as allogeneic stem cell transplantation to overcome this otherwise intractable disease.