The non-driver mutation landscape in different stages of PMF
Comparison of the baseline clinical and laboratory characteristics of patients classified in different stages are summarized in Supplemental Table 4. Driver mutations were distributed as follows: JAK2V617F ( 55.8%, n = 144), CALR Exon9 (24%, n = 62), MPLW515 ( 3.5%, n = 9), double mutations (CALR Exon9 and MPLW515) in 1 patient(0.4%). 42(16.3%) patients did not have any driver mutations (triple-negative). The median variant allele frequency (VAF) of driver mutations in different stages of PMF was shown in Supplemental Fig. 3A-C, only the VAF of JAK2V617F had a significant increment between Pre-PMF and PMF-AP/BP patients (Median of VAF: 38.3% vs. 50.9%, P = 0.031), but no difference between overt-PMF and PMF-AP/BP (Median of VAF: 50.9% vs. 45.4%, P = 0.333 ). The distribution and the number of non-driver mutations in different stages of PMF were shown in Supplemental Fig. 3D-G.
Mutations in the ASXL1 (31.1% vs. 10.1%, P = 0.001) and U2AF1 (13.7% vs. 1.4%, P = 0.003) were more frequent in Overt-PMF compared with Pre-PMF (Fig. 1A). Moreover, mutations in ASXL1 (51.1% vs. 31.1%, P = 0.013), SRSF2 (26.7% vs. 11.2%, P = 0.009), RUNX1 (22.2% vs. 5.6%, P = 0.001), SETBP1 (17.8% vs. 7.5%, P = 0.039), NRAS (15.6% vs. 3.1%, P = 0.002) and EZH2 (13.3% vs. 5%, P = 0.049) were significantly more frequent in PMF-AP/BP compared with Overt-PMF (Fig. 1B). From the whole course of the disease, the frequency of ASXL1 mutations significantly increased during the progression (10.1% vs. 31.1% vs. 51.1%) (Fig. 1C). However, there was no significant change in the VAF of ASXL1 mutations in different stages (Median of VAF: 42.1% vs. 30.6% vs. 37.8%) (Supplemental Fig. 3H). As expected, the frequency of HMR mutations also significantly increased during the disease progression (14.9% vs. 40.4% vs.70.5%) (Supplemental Fig. 3I).
Non-driver mutation enrichment in different stages of PMF
To clarify the relationship between non-driver mutations and disease progression, we looked at the enrichments of major non-driver mutations in different stages of PMF.
In univariate comparison, ASXL1(OR = 3.99, 95%CI 1.71∼9.33; P = 0.001) and U2AF1(OR = 10.76, 95%CI 1.42∼81.53; P = 0.005) mutations were significantly enriched in Overt-PMF compared to Pre-PMF. Mutations in five genes, including ASXL1(OR = 2.32, 95%CI 1.18∼4.55; P = 0.013), SETBP1 (OR = 2.69, 95%CI 1.02∼7.64; P = 0.039), RUNX1 (OR = 4.83, 95%CI 1.82∼12.76; P = 0.001), NRAS (OR = 5.75, 95%CI 1.73∼19.10; P = 0.002) and SRSF2 (OR = 2.89, 95% CI 1.27∼6.58; P = 0.009) were significantly enriched in PMF-AP/BP compared to Overt-PMF (Supplemental Fig. 3J and Table 1).
Table 1
Univariate and multivariate analysis of enrichment of non-driver mutations in different stages of PMF.
Overt-PMF vs. PMF-AP/BP |
Univariate analysis | Multivariate analysis |
Gene | Odds ratio | 95%CI | P value | Odds ratio | 95%CI | P value |
ASXL1 | 2.32 | 1.18 ~ 4.55 | 0.013 | 2.88 | 1.36 ~ 6.10 | 0.006 |
SETBP1 | 2.69 | 1.02 ~ 7.64 | 0.039 | 2.35 | 0.81 ~ 6.81 | 0.114 |
RUNX1 | 4.83 | 1.82 ~ 12.76 | 0.001 | 5.11 | 1.66 ~ 15.74 | 0.004 |
NRAS | 5.75 | 1.73 ~ 19.10 | 0.002 | 4.06 | 1.05 ~ 15.60 | 0.042 |
SRSF2 | 2.89 | 1.27 ~ 6.58 | 0.009 | 1.87 | 0.73 ~ 4.79 | 0.196 |
Pre-PMF vs. Overt-PMF |
Univariate analysis | Multivariate analysis |
Gene | Odds ratio | 95%CI | P value | Odds ratio | 95%CI | P value |
ASXL1 | 3.99 | 1.71 ~ 9.33 | 0.001 | 3.65 | 1.54 ~ 8.62 | 0.003 |
U2AF1 | 10.76 | 1.42 ~ 81.53 | 0.005 | 9.21 | 1.20 ~ 70.69 | 0.033 |
To exclude the effects of the co-occurrence of non-driver mutations that might confound the result, we performed a multivariate analysis. In the comparison between Overt-PMF and Pre-PMF, ASXL1 (OR = 3.65,95%CI 1.54∼8.62; P = 0.003) and U2AF1 (OR = 9.21,95%CI 1.20∼70.69; P = 0.033) mutations were still significantly enriched in Overt-PMF. Meanwhile, in the comparison between PMF-AP/BP and Overt-PMF, only RUNX1 (OR = 5.11,95%CI 1.66∼15.74; P = 0.004), NRAS(OR = 4.06,95%CI 1.05∼15.60; P = 0.042) and ASXL1 (OR = 2.88,95%CI 1.36∼6.10; P = 0.006) mutations were strongly enriched in PMF-AP/BP (Fig. 1D and Table 1). These results indicate that ASXL1 mutations might play a critical role in myelofibrosis progression and blast phase evolution during the whole course of PMF progression.
ASXL1 mutations in the AP/BP transformation of PMF
As mentioned above, ASXL1 mutations play a critical role both in the progression from Pre-PMF to Overt-PMF and through Overt-PMF to PMF-AP/BP. To evaluate the status of ASXL1 mutations in different stages, we analyzed 17 patients who had at least two-time points serial samples during disease progression, including 4 patients who progressed from Pre-PMF to Overt-PMF and 13 patients who progressed from Overt-PMF to PMF-AP/BP (Table 2).
Table 2
The mutations of 17 paired serial samples from PMF patients.
Patient | Overt-PMF phase | VAF | Accelerate/Blast phase | VAF |
UPN688 | JAK2V617F | 91.4 | JAK2V617F | 95.8 |
| ASXL1 | 45.3 | ASXL1 | 45.5 |
UPN809 | JAK2V617F | 14.2 | JAK2V617F | 17 |
| ASXL1 | 11.8 | ASXL1 | 17.7 |
| TET2 | 4.2 | TET2 | 2.4 |
| NF1 p.Q1724X | 11.6 | NF1 p.Q1703X | 19.6 |
| DDX41 | 23.2 | DDX41 | 15.5 |
| ETNK1 | 5.7 | ETNK1 | 11.7 |
| EZH2 | 11.4 | EZH2 | 23.6 |
| | | ETV6 p.S139Yfs*14 | 2.4 |
| | | ETV6 p.R369W | 12.4 |
| | | SETBP1 | 1.8 |
| | | TP53 | 11.1 |
| | | TET2 p.F1287S | 2.2 |
UPN236 | JAK2V617F | 45.4 | JAK2V617F | 37.31 |
| ASXL1 | 43.6 | ASXL1 | 34.41 |
| IDH2 | 47.8 | SRSF2 | 2 |
| SRSF2 | 45.5 | RUNX1 | 15.91 |
| BRIP1 | 48.5 | NOTCH2 | 45.59 |
UPN045 | CALR | 50 | CALR | 14 |
| ASXL1 | 5.7 | ASXL1 | 2.3 |
| | | NF1 | 32.3 |
UPN927 | CALR | 49.7 | CALR | 38.9 |
| ASXL1 | 41.9 | ASXL1 | 44.5 |
| SETBP1 p.D868N | 12.81 | SETBP1 p.D868N | 43.4 |
| EZH2 | 87 | EZH2 | 81 |
| SETBP1 p.D868Y | 26.5 | ETV6 p.N85Kfs*5 | 41.4 |
| ETV6 p.R105X | 28.7 | NRAS | 40.6 |
UPN578 | CALR | 48.4 | CALR | 14.8 |
| SF3B1 p.K700E | 14.7 | SF3B1 p.K700E | 34 |
| SF3B1 p.K666N | 4.7 | SF3B1 p.K666N | 3.5 |
UPN546 | MPLW515K | 50 | MPLW515K | 50 |
| SETBP1 | 44.97 | SETBP1 | 48.8 |
| SUZ12 | 40.31 | SUZ12 | 55.4 |
| | | NRAS p.G12V | 8.8 |
| | | NRAS p.G12D | 8.7 |
| | | NRAS p.G12A | 26.8 |
| | | SRSF2 | 58.1 |
| | | RUNX1 | 16.7 |
| | | BCOR | 48.9 |
UPN834 | CALR | 40 | CALR | 16.4 |
| PLCG1 | 39.1 | PLCG1 | 9.7 |
| FAT1 | 49.8 | CBL | 70.2 |
| | | CSMD1 | 2.9 |
| | | RUNX1 | 60.8 |
UPN853 | JAK2V617F | 6.82 | JAK2V617F | 35.8 |
| CCND3 | 5.6 | CCND3 | 31 |
| SETBP1 | 47.16 | RUNX1 | 38.6 |
UPN511 | MPLW515K | 50 | MPLW515K | 50 |
| SRSF2 | 51.4 | SRSF2 | 45.8 |
| | | ETV6 | 12.2 |
| | | DIS3 | 45.3 |
UPN705 | JAK2V617F | 80.2 | JAK2V617F | 94.6 |
| SRSF2 | 48.9 | SRSF2 | 44.8 |
| RUNX1 | 47 | RUNX1 | 49.1 |
| TET2 | 5.5 | | |
UPN667 | JAK2V617F | 44.5 | JAK2V617F | 56 |
| SF3B1 | 27.9 | SF3B1 | 33.1 |
| | | ASXL1 | 41.3 |
| | | GNAS | 37.1 |
| | | SETD1B | 39.9 |
UPN663 | JAK2V617F | 2.1 | JAK2V617F | 43.5 |
| SRSF2 | 52.2 | SRSF2 | 48.4 |
| | | ASXL1 | 31.4 |
Patient | Pre-PMF phase | VAF | Overt-PMF phase | VAF |
UPN490 | JAK2V617F | 39.9 | JAK2V617F | 36.9 |
| DNMT3A | 48.7 | DNMT3A | 39.7 |
| SF3B1 | 47.4 | SF3B1 | 42.7 |
| TP53 p.K132R | 5 | TP53 p.K132R | 1.6 |
| KMT2D | 50.6 | TP53 p.E258K | 2.2 |
| FGFR3 | 46.9 | | |
| NOTCH1 | 36.2 | | |
UPN755 | ATG2B | 47.4 | | |
UPN127 | ZMYM3 | 50.3 | ZMYM3 | 52.6 |
UPN312 | PRF1 | 45.7 | | |
| PDGFRB | 50.9 | | |
Firstly, to explore how ASXL1 mutations promote the progression from the chronic stage to the AP/BP stage, we analyzed clonal evolution in serial samples collected from Overt-PMF to PMF-AP/BP. Excepting newly acquired ASXL1 mutations at the AP/BP stage, all ASXL1 mutations that occurred in the Overt-PMF phase kept relatively stable with less than 10% of fluctuation (Fig. 2A). Contrary to ASXL1, RUNX1 mutations (4 in 5, 80%), RAS pathway mutations (7 in 8, 87.5%) and TP53 mutations (1 in 1,100%) were more likely to be freshly acquired during the transformation from Overt-PMF to PMF-AP/BP (Fig. 2B and Supplemental Fig. 4A and 4B). Interestingly, SRSF2 mutations tended to decrease clone size during disease progression (Supplemental Fig. 4C). Then, to answer the question that why AP/BP transformation did not accompany with obvious ASXL1 mutations expansion, we analyzed the co-mutations in PMF-AP/BP patients who had ASXL1 mutations at the chronic stage. We found that ASXL1 mutations were more likely to co-occur with RAS pathway mutations (NRAS 20% and NF1 50%) and ETV6 (50%) (Fig. 2C). Moreover, ASXL1 mutations even cannot be significantly removed by hypomethylating agent (HMA) therapy in 3 PMF-AP/BP patients (Fig. 2D), but co-occurred RUNX1 and SETBP1 mutations burden were obviously decreased (Supplemental Fig. 4D).