Predictors of clonal evolution and myeloid neoplasia following immunosuppressive therapy in severe aplastic anemia.

Abstract

Predictors, genetic characteristics, and long-term outcomes of patients with SAA who clonally evolved after immunosuppressive therapy (IST) were assessed. SAA patients were treated with IST from 1989–2020. Clonal evolution was categorized as “high-risk” (overt myeloid neoplasm [meeting WHO criteria for dysplasia, MPN or acute leukemia] or isolated chromosome-7 abnormality/complex karyotype without dysplasia or overt myeloid neoplasia) or “low-risk” (non-7 or non-complex chromosome abnormalities without morphological evidence of dysplasia or myeloid neoplasia). Univariate and multivariable analysis using Fine-Gray competing risk regression model determined predictors. Long-term outcomes included relapse, overall survival (OS) and hematopoietic stem cell transplant (HSCT). Somatic mutations in myeloid cancer genes were assessed in evolvers and in 407 patients 6 months after IST. Of 663 SAA patients, 95 developed clonal evolution. Pre-treatment age >48 years and ANC >0.87×10⁹/L were strong predictors of high-risk evolution. OS was 37% in high-risk clonal evolution by 5 years compared to 94% in low-risk. High-risk patients who underwent HSCT had improved OS. Eltrombopag did not increase high-risk evolution. Splicing factors and RUNX1 somatic variants were detected exclusively at high-risk evolution; DNMT3A, BCOR/L1 and ASXL1 were present in both. RUNX1, splicing factors and ASXL1 mutations detected at 6 months after IST predicted high-risk evolution.

Introduction:

Severe aplastic anemia (SAA) is an immune bone marrow failure characterized by the destruction of bone marrow stem and progenitor cells by cytotoxic T lymphocytes, resulting in pancytopenia with consequent morbidity and mortality. SAA is effectively treated in most patients with either immunosuppressive treatment (IST) or allogeneic hematopoietic stem cell transplant (HSCT)(1). Six-month response rates to IST now approach 80% with the addition of eltrombopag (EPAG), a thrombopoietin mimetic, to IST combination therapy horse anti-thymocyte globulin (hATG) and cyclosporine (CSA)(2)(3). However, clonal evolution, defined as acquisition of a new cytogenetic abnormality or development of overt myeloid neoplasia, remains a significant long-term complication(4, 5).

Clonal hematopoiesis (CH), defined as an abnormal karyotype or presence of single-nucleotide polymorphisms or somatic variants in DNA, is present in most patients with SAA, but its presence does not predetermine development of hematologic malignancy(6–8). Copy number neutral loss of heterozygosity at the short arm of chromosome 6 including the HLA locus (6p CN-LOH) occurs in ∼ 10% of patients with SAA. 6p CN-LOH is implicated as an immune escape mechanism and is a diagnostic marker of immune AA(9, 10); its presence is not associated with evolution to a myeloid neoplasm(11). Somatic mutations in myeloid cancer genes occur in up to one third of SAA patients, particularly in BCOR/BCORL1, ASXL1 and DNMT3A(6). Mutations are found at small variant allele fractions (VAF) when compared to myelodysplastic syndrome (MDS). ASXL1 or DNMT3A mutations have been associated with reduced treatment response, myeloid neoplasia, and lower overall survival, while the presence of BCOR/BCORL1 or PIGA mutations predicts for higher response to IST(6, 12). Prior studies of mutational landscape after clonal evolution have identified RUNX1, ASXL1, splicing factors, CBL, and SETBP1 as enriched in contrast to non-evolved SAA(13, 14).

Clonal evolution remains the most feared long-term complication in SAA patients treated with IST. Risk factors previously reported include increased age(5, 15, 16), disease severity(17), very short telomere length(18, 19), type of therapy(15, 20), use of G-CSF(17, 21), the presence of HLA “risk” alleles or loss(22–24), and ASXL1(7). Clinical outcomes after clonal evolution differ depending on the chromosomal abnormality observed. Complex karyotype and monosomy 7 are considered “high-risk” and have been associated with leukemic transformation and decreased overall survival (OS). Conversely, deletion 13q and trisomy 8 associate with a good response to IST and a lower rate of leukemic transformation(8, 25, 26). Other less common isolated chromosomal abnormalities have less clear clinical implications due to rarity and lack of published data.

We sought to identify predictors for clonal evolution, particularly high-risk evolution, while examining long term outcomes and treatment strategies. Additionally, we further investigated the mutational landscape of clonal evolution between high and low-risk events.

Methods:

Patients with treatment naïve, relapsed, or refractory SAA treated with ATG or alemtuzumab-based IST at the NIH Clinical Center from 1989 to 2019 were included. Patients met Camitta criteria for SAA and underwent pre-treatment bone marrow examination with immunophenotyping and karyotype (Supplemental Table 1). Bone marrow examination after IST was performed routinely at predetermined trial timepoints or due to clinical suspicion of clonal evolution. Chromosome breakage testing to exclude Fanconi Anemia was performed for all patients <40 years. CLIA Inherited Bone Marrow Failure panel (University of Chicago) was performed in 178 patients (Supplemental Methods and Table 3) and research targeted NGS for IBMFS genes was performed in an additional 246 patients (Supplemental Tables 2+4) . No patients had pathogenic or likely pathogenic germline variants consistent with IBMFS.

Patients were enrolled on the following NIH clinical trials: NCT01623167, NCT00061360, NCT00001964, NCT00260689, NCT00944749, NCT00195624, and (Tisdale, The Lancet, 2000 and Rosenfeld, Blood, 1995). All patients or their guardians provided written informed consent.

A total of 663 patients were analyzed; excluded patients received alkylators (n=37) or had a baseline abnormal karyotype (n=4). Ninety-five patients developed either a new chromosomal abnormality or overt myeloid neoplasia after IST. Patients were divided into three groups: high-risk clonal evolution, low-risk clonal evolution, and no clonal evolution. High-risk clonal evolution was defined as the presence of an overt myeloid neoplasm with morphology as defined by the 2016 World Health Organization classification of myeloid neoplasms, including myeloproliferative neoplasm (MPN), myelodysplastic syndrome (MDS) or acute myeloid leukemia (AML); even in the absence of overt dysplasia or myeloid neoplasia, any chromosome 7 abnormality or complex karyotype was also considered high-risk. Any other isolated chromosomal abnormality (very good, good, and intermediate by R-IPSS) without morphological dysplasia or myeloid neoplasia was considered low-risk evolution, and those with a consistently normal karyotype were deemed not to have evolved. Characterization of low and high-risk evolution was based on previous literature in immune marrow failure and on the classification of high-risk cytogenetics in myeloid malignancies(8, 25–30).

Age, sex, baseline laboratory values prior to IST and at time of evolution, response at 6 months, relapse, type and number of IST received, presence of 6p CN-LOH and HLA loss, presence of somatic mutations at VAF ≥2% in myeloid cancer genes or BCOR/L1 found 6 months after IST, and exposure to EPAG were collected. Laboratory values included: absolute neutrophil count (ANC), absolute reticulocyte count (ARC), platelet count, and presence of paroxysmal nocturnal hemoglobinuria (PNH) clone ≥1% using peripheral blood immunophenotyping (patients from 2003 onwards). IST was defined as treatment with hATG, rabbit ATG (rATG) or alemtuzumab, and EPAG exposure as therapy for at least 12 weeks, either in the up-front or relapsed/refractory setting. HLA loss was assessed by either HLA flow cytometry plus deep nucleotide sequencing or deep sequencing alone as previously described(23). 6p CN-LOH was inferred using read counts of allele-specific single-nucleotide polymorphisms in HLA genes.

Error-corrected DNA-sequencing (ECS) for a panel of 52 genes associated with myeloid malignancies, CH, and constitutional disease (VariantPlex ArcherDx, CA; Supplemental Table 2) was used to analyze peripheral blood (granulocytes [n=26] or peripheral blood mononuclear cells [n=21] separated by Ficoll) or bone marrow (n=1) at time of evolution in 48 patients with available DNA. VAF (Supplemental table 5) was not directly compared between all patients due to different sample types. Somatic mutations in myeloid cancer genes and BCOR/L1 were also assessed at 6 months after IST in 407 patients. CLIA certified NGS for somatic variants was performed by Neogenomics (n=145) or Genomics Testing Cooperative (GTC) (n=16). Targeted research NGS was previously performed and published in NEJM (n=246)(6). Sequencing techniques are described in Supplemental Methods.

Statistics

Patients without a known IST start date (n=21) were excluded from analyses that required baseline data. Patients who developed >1 clonal evolution type (low-risk then high- risk) were considered low or high-risk depending on the specific analysis performed: low-risk for outcomes analysis (relapse, and OS) as this was the outcome of their first event, and high-risk for predictor analysis, as while they were initially low-risk, their ultimate outcome was high-risk.

Somatic mutations at 6 months were curated for predictor analysis; they included 40 genes recurrent in myeloid cancers and CH that were common to the 3 sequencing platforms (list in Supplemental Methods), including only mutations with VAF ≥2%. Somatic mutations were grouped according to risk profile to allow for predictor analysis (details in Supplemental Methods).

Summary statistics of baseline characteristics were calculated using median and range for continuous variables, and count and proportion for categorical variables. Univariable and multivariable analysis was performed using the Fine-Gray competing risk regression model(31, 32) for time to clonal evolution, with death taken as a competing risk. Cumulative incidence curves were used to display the probability of clonal evolution at timepoints. Classification and regression tree analysis(33) for time to clonal evolution was performed on continuous baseline variables to partition the data based on the best categorical cutoff. Survival probabilities were evaluated using Kaplan-Meier estimates and Cox Proportional Hazard models(34). All analyses were performed using R (4.0.3).

Results:

Baseline characteristics

Ninety-five of 663 (14%) SAA patients enrolled on IST protocols at the NIH CC developed clonal evolution (Supplemental Figure 1), including 59 patients classified as high-risk and 36 as low-risk; four high-risk patients initially developed low-risk evolution. Most high-risk clonal evolution patients had chromosome 7 changes (n=44/59; 75%); 39 with monosomy 7, four with 7p deletion and one inv(7). Others had complex karyotype (n=2), morphological myeloid neoplasm with normal karyotype (n=7) or unknown karyotype (n=3), MDS-EB2 with trisomy 8 (n=1), and AML with del 13q (n=1) and t(3;19;11) (n=1) (Supplemental Table 5). In the low-risk evolution group, the commonest cytogenetic abnormalities were deletion 13q (n=13), trisomy 8 (n=8) and deletion 5q (n=3) with other karyotypes seen at lower frequency (Supplemental Table 18).

Both median age at diagnosis and baseline ANC were significantly higher in those who evolved than in those who did not, while there was no difference between low-risk and high-risk patients (Table 1). There were no significant differences between groups in baseline platelet counts, ARC, or presence of a PNH clone ≥1%. Patients with high-risk evolution had more frequently been treated with >1 course of IST than had non-evolvers. Median follow-up time for the cohort from time of first IST to last follow up was 4.7 years (1700 days).

Table 1.

Baseline characteristics of all patients

	Clonal evolution	No clonal evolution	P value	High risk evolution	P value HR v No-CE	Low risk evolution	P value LR v No-CE	P value HR v LR
	N=93	N=549	(CE v No-CE)	N=57		N=36
Age (years, median) §	49 (6–79)	29 (2–82)	<0.001	50 (6–79)	<0.001	46 (7–78)	0.005	0.7
Female n, (%) £	40 (43)	240 (44)	0.99	21 (37)	0.39	19 (53)	0.38	0.19
Male n, (%)	53 (57)	309 (56)		36 (63)		17 (47)
Laboratory baseline (median)
ANC (×10 9 /l), (range) §	0.38 (0–1.93)	0.26 (0–1.76)	0.004	0.37 (0–1.81)	0.1	0.40 (0–1.93)	0.007	0.3
ARC (×10 9 /l), (range) §	20.2 (0.9–106)	14.3 (0–119)	0.1	16.3 (0.9–106)	0.53	21.1 (3.2–57)	0.054	0.4
Platelets (×10 9 /l), (range) §	8 (0–148)	9 (0–254)	0.77	9 (0–148)	0.93	8 (3–26)	0.55	0.62
PNH clone* n, (%) £	18 (19)	122 (22)	0.63	10 (18)	0.52	8 (22)	1	0.77
First IST n, (%)
hATG £ ¥	82 (88)	482 (88)	1 £	47 (82)	0.35 £	35 (97)	0.11 ¥	0.046 ¥
hATG+EPAG	24 (26)	154 (28)	0.75	11 (20)	-	13 (36)	-	0.12
rATG	8 (9)	51 (9)	__	7 (12)	-	1 (3)	-	__
Alemtuzumab	2 (2)	14 (3)	__	2 (4)	-	0 (0)	-	__
Treatment history, n (%)
Responder £	61 (66)	347 (63)	0.64	32 (56)	0.18	29 (81)	0.11	0.1
Non-responder	27 (29)	178 (32)	-	20 (35)	-	7 (19)	-	-
>1 IST £	25 (27)	102 (19)	0.09	18 (32)	0.03	7 (19)	1	0.3
Relapse £	31 (33)	129 (23)	0.06	17 (30)	0.37	14 (39)	0.06	0.5
EPAG therapy £	29 (31)	171 (31)	1	14 (25)	0.38	15 (42)	0.21	0.13

^*PNH data were unavailable in 276 patients

Patients without an IST start date (n=21) were not included in this table as baseline data were unavailable (including n=2 with evolution). Therefore, the baseline table includes 642 patients and 93 with clonal evolution. Patients who developed low risk evolution followed by high-risk evolution were classified as high risk.

Abbreviations: ANC, absolute neutrophil count; ARC, absolute reticulocyte count; PNH, paroxysmal nocturnal hemoglobinuria; IST, immunosuppressive therapy; EPAG, eltrombopag

^§Wilcoxon rank sum test;

^£Chi-square test;

^¥Fisher’s exact test

^__P value not calculated due to small size

Predictors of clonal evolution

The median time from IST to evolution was similar in high-risk and low-risk groups; 733 days in low-risk versus 784 days in high-risk. Older age at diagnosis and higher baseline ANC were identified as strong predictors of clonal evolution in univariable analysis. Using a decision tree, we agnostically analyzed data to stratify risk based on specific age and ANC cut offs. ANC 0.87×10⁹/L segregated for both overall and high-risk evolution versus no evolution (Supplemental Figures 2-3). Age >37 years segregated all clonal evolution and age >48 years segregated specifically for high-risk clonal evolution (Figures 1A-D). Other baseline characteristics including sex, ARC, platelet count, presence of PNH clone, CN-6p LOH, and HLA loss did not predict for evolution (Supplemental tables 11+12). Presence of a somatic mutation in a myeloid cancer or CH gene 6 months after IST with VAF ≥2% was predictive of both overall (p=0.02) and high-risk (p=0.01) clonal evolution. Choice of first-line IST, response at 6-months, or relapse were not predictive. High-risk clonal evolution was not significantly increased in patients who received EPAG versus those who did not (p=0.47) but there was a significant difference in all (low-risk and high-risk combined) clonal evolution (p=0.02) (Figure 2 C+D). Duration of EPAG treatment was not predictive of clonal evolution (Supplemental tables 11+12). The median time to high-risk clonal evolution in patients who received EPAG was shorter (371 days) compared with those who did not (1089 days) as reported previously(16).

Figure 1: Age and ANC as predictors for clonal evolution.

Cumulative incidence (CI) curves using Fine-Gray competing risk method with death as a competing risk. (A) CI for development of all clonal evolution in patients aged >37 years and (B) CI for development of high-risk clonal evolution in patients aged >48 years. (C) CI for development of all clonal evolution when baseline ANC >0.87×10⁹/L and (D) for high-risk clonal evolution when baseline ANC >0.87.

Figure 2: Overall survival after clonal evolution or HSCT, and cumulative incidence of clonal evolution with eltrombopag.

(A) Overall survival comparing high-risk and low-risk evolution from time of clonal evolution. (B) Overall survival from time of development of high-risk evolution comparing patients who underwent HSCT versus chemotherapy or supportive care using Kaplan-Meier survival. For all high-risk patients who underwent HSCT OS at 2 years was 72%, at 5 years was 54.3%, and at 8 years was 43.5% compared to 37.1%, 22.3%, and 7.4% respectively (C) Risk of all clonal evolution with eltrombopag (EPAG). There was a significant difference for development of any clonal evolution in patients who received EPAG (p=0.02). Cumulative incidence (CI) curves using Fine-Gray competing risk method with death as a competing risk. (D) Risk of high-risk clonal evolution with eltrombopag. There was no significant difference in development of high-risk clonal evolution in patients who received EPAG (p=0.47). Cumulative incidence (CI) curves using Fine-Gray competing risk method with death as a competing risk. EPAG exposure was defined as at least 12 weeks of EPAG therapy and included patients who received drug for treatment naive, relapsed, and refractory disease (n=200). Of these, 178 received EPAG with hATG and CSA for treatment naïve SAA. The median duration of EPAG exposure in those who were EPAG treated was 6 months (range 1–74). Patients were divided into exposure ≤3 months (n=69), >3 months and ≤6 months (n=99), and >6 months (n=32); none were predictive by univariate analysis. (Supplemental Table 11+12).

Patients aged >37 years at time of first IST had a significantly increased risk of developing clonal evolution (either high-risk or low-risk) that increased over time with a cumulative incidence (CI) of 11% by 2 years, 19% by 5 years, and 25% 8 years post treatment compared with patients ≤37 years (p<0.001) (Figure 1A). High-risk clonal evolution was significantly increased in patients aged >48 years at time of first IST, with CI of 7% by 2 years, 14% by 5 years, and 18% by 8 years (p<0.001) (Figure 1B). In comparison, patients aged ≤48 years had a lower CI of high-risk evolution at 3%, 4%, and 6% at 2, 5, and 8 years, respectively. Baseline ANC >0.87 ×10⁹/L predicted an even higher risk of clonal evolution; CI for all evolution was 17% by 2 years, 36% by 5 years, and 53% by 8 years (p<0.001) (Figure 1C) and for high-risk evolution was 11% by 2 years, 17% by 5 years, and 31% by 8 years (Figure 1D). Combined high ANC and older age (>48 years) predicted the greatest risk of developing high-risk evolution with a hazard ratio of 5.51 compared to 1.6 for high ANC and younger age, 1.91 for low ANC and older age. Combined low ANC and age <48 years was protective against developing high-risk evolution with hazard ratio 0.32. By multivariable analysis, age remained significant for any or high-risk clonal evolution (p=0.0005 and p=0.009 respectively), ANC was significant for any but not high-risk evolution, and the presence of any somatically mutated clone at 6 months approached significance (p=0.057) (Supplemental Tables 13-16) for high-risk evolution.

Overall survival

Overall survival from time of clonal evolution was significantly worse in the high-risk group compared to low-risk and non-evolver groups. In the high-risk group, OS was 37% by 5 years compared to a favorable survival in the low-risk group of 94% by 5 years (logrank p<0.001) (Figure 2A). Overall survival in the non-evolved group was significantly lower than in the low-risk group (p= 0.027), likely due to the non-evolved group containing patients who were transplanted or died early in their disease course (Supplemental Figure 4).

High-risk clonal evolution

Characteristics and standardized risk stratification at time of evolution

At time of evolution, ANC and platelet counts were significantly lower in patients with monosomy 7 (n=39) when compared with other high-risk patients (Table 2).

Table 2.

Characteristics and outcomes of high-risk evolvers

	Monosomy 7	Other chromosome 7 (n=5)	Other	P value
	(n=39)		(n=15)
Laboratory values at time of high-risk evolution (median)
ANC (×109/l)	0.64 (0–3.29)	1.77 (0.92–2.29)	1.76 (0.26–32.1)	0.002
ARC (×109/l)	52.7 (2.6–117)	56.7 (23.3–129)	50.3 (2.1–182)	0.78
Platelets (×109/l)	28 (3–251)	46 (9–108)	96 (9–1194)	0.03
Risk stratification
IPSS-R (median)	5 (3–8)	3 (2–5)	2.75 (2–7)	0.03
Marrow blast % (median)	2 (0–12)	1 (1–4)	5 (3–31)	0.14
Preceding low risk evolution, n (%)	3 (8)	0	1 (7)	__
WHO diagnosis at time of high-risk evolution, n (%)
MDS-SLD	2 (6)	0	0	__
MDS-SLD-RS	1 (3)	0	0	__
MDS-MLD	11 (30)	1 (20)	1 (7)	__
MDS-EB1	5 (14)	0	3 (20)	__
MDS-EB2	1 (3%)	0	1 (7)	__
AML	0	0	6 (40)	__
MPN/MDS	0	0	3 (20)	__
MDS-U	14 (39)	1 (20)	1 (7)	__
hMDS	2 (6)	0	0	__
Does not meet WHO criteria #	0	3 (60)	0	__
Outcomes, n (%)
MDS-EB or AML	19 (49)	0	10 (67)	__
HSCT	15 (38)	4 (80)	7 (47)	__
Death	28 (72)	1 (20)	7 (47)	__

^*Three patients in the monosomy 7 group had unknown marrow morphology

^#WHO criteria defined as dysplasia >10% in at least 1 lineage, or if absent the presence of an MDS defining cytogenetic abnormality with at least 1 cytopenia

Kruskal-Wallis p-value is used

^__p-value not calculated due to small sample size

Since patients with new chromosome 7 abnormalities were deemed high-risk evolution regardless of the presence of dysplasia or increased blast counts, the bone marrows from time of evolution were reviewed systematically by two dedicated hematopathologists (Supplemental Table 5). Of 39 patients with monosomy 7, 36 had material for review. Morphological dysplasia using WHO criteria was evident in 23 (64%) of patients; only 5 (14%) initially had excess blasts (≥5% marrow blast count). The median R-IPSS score was 5 at time of evolution with a median blast count of 2%. However, most patients (53%) had a final diagnosis of MDS-EB1 (n=10), MDS-EB2 (n=3) or AML (n=6) at last follow up, with median follow up time for the cohort of 594 days from development of monosomy 7. In the five patients with other (non-monosomy) chromosome 7 changes, only one had overt dysplasia at time of clonal evolution and none progressed to MDS-EB or AML with median follow up of 678 days.

Using univariate analysis, the presence of high R-IPSS or increased blast counts at evolution did not predict for OS in patients with monosomy-7 (Supplemental Table 17).

Impact of HSCT on HR clonal evolution outcomes

A total of 26 (44%) high-risk patients underwent HSCT with significantly improved OS compared to those who did not (Figure 2B) (P<0.01); the majority were patients with monosomy-7.

Fifteen (38%) patients with monosomy-7 underwent HSCT and 24 (62%) were treated with supportive care or chemotherapy. Overall survival was significantly better in HSCT group at 5 years (57 %) compared to medical therapy or supportive care (23%) (p<0.01) (Supplemental Figure 6). Median follow up from date of HSCT was 1401 days (3.8 years). Of note, the median age at SAA diagnosis was significantly younger in the HSCT group (19 years, range 6–64) when compared to the patients in non-HSCT group (60 years, range 19–79), which may confound interpretation of outcomes. Median baseline counts were similar between groups; ARC was 16.5 ×10⁹/L, ANC 0.244 ×10⁹/L and platelets 11 ×10⁹/L in the HSCT group compared to ARC 12.9 ×10⁹/L, ANC 0.37 ×10⁹/L and Platelets 8 ×10⁹/L in non-HSCT (p=0.9, 0.4, and 1, respectively). The median R-IPSS was 5.5 in the HSCT group and 4.75 in the non-HSCT group.

Four of five patients with non-monosomy chromosome 7 abnormalities underwent HSCT, one of whom died. Of 15 patients with non-chromosome 7 high-risk evolution, seven underwent HSCT; five were alive at last follow up and two had died, compared to three of the eight patients who did not undergo HSCT that were alive at last follow-up. Median follow up time from date of HSCT was 505 days.

Post-transplant complications, both short and long term (GVHD [n=2], infection [n=2], secondary malignancy [n=1] and VOD ([n=1])), were the commonest cause of death in HSCT patients (6/12 deaths; 50%) compared to disease progression (8/24 deaths; 33%) or infection (6/24 deaths; 25%) in those treated with supportive care or chemotherapy. Cause of death was unknown for 8 in non-HSCT group and 1 in HSCT group.

Low-risk evolution

Forty patients (6%) developed low-risk evolution, however 4 of these subsequently developed high-risk evolution (Supplemental table 18), leaving a total of 36 true low-risk cases. Median time to low-risk evolution was 728 days. Median blood counts at time of low-risk evolution were: Hb 11.9 g/dL, platelets 90 ×10⁹/L, ANC 1.82 ×10⁹/L and ARC 66.75 ×10⁹/L. OS was favorable in the low-risk group compared to both high-risk and non-evolved patients (Supplemental Figure 4).

High-risk evolution from low-risk

Four patients with low-risk evolution subsequently developed high-risk (11% of all low-risk evolvers); three initially had 13q deletion and one had t(12;13). Three developed monosomy 7, and one AML with persistent 13q deletion. Development of high-risk evolution was prolonged in two patients with 13q deletion at 1365 and 2319 days (3.7 years and 6.4 years) from the first abnormal karyotype, while the remaining patient with 13q evolved to high-risk evolution more quickly at 249 days. No patients experienced relapse between low-risk and high-risk evolution, but one had relapsed prior to development of 13q. HSCT was performed in two patients who were well at last follow up, while the two who did not undergo HSCT died of disease progression.

Relapse

Relapse was not increased in low-risk evolution patients compared with non-evolvers measured from time of IST, true even when 13q deletion was assessed as a subgroup (Supplemental Figure 7+8). Once low-risk clonal evolution occurred, seven (18%) patients subsequently relapsed with a median time to relapse of 174 days.

Somatic mutations

Somatic mutations at evolution

Twenty three out of 59 (39%) high-risk patients and 25/36 (70%) of low-risk patients had peripheral blood assessed for somatic mutations in myeloid cancer genes at time of evolution. Fourteen of 25 (56%) of the low-risk and 19/23 (82%) of the high-risk group had at least one mutation. RUNX1 (13 variants in 8 [35%] patients) and ASXL1 (13 variants in 10 [43%] patients) were the most frequently mutated genes in high-risk patients, and BCOR/L1 (14 variants in 8 [32%] patients) was the most frequently mutated in the low-risk group. RUNX1 and splicing factor mutations were present only in the high-risk group, while DNMT3A and ASXL1 were present in both groups (Figure 3A).

Figure 3: Clonal landscape of somatic mutations at the time of clonal evolution.

(A) Oncoprint showing type and number of overall somatic mutations in myeloid cancer or CH genes when comparing high-risk and low-risk patients, along with karyotype. (B) Clonal dynamics showing acquisition or loss of somatic mutations over time in five high-risk clonal evolution patients. Sample types as follows: For UPN1, UPN8, UPN9, and UPN47 all sequencing performed in granulocytes. For UPN32, PBMCs were sequenced at time of evolution and monocytes at 2 years prior.

Longitudinal data were available in five high-risk and eight low-risk patients. Three of five high-risk patients had acquisition or expansion of the RUNX1 clones at evolution. Small RUNX1 clones were present in two patients up to three years prior to high-risk evolution. Splicing factor mutations in three patients were present months prior to development of evolution and in two patients, clones remained stable both prior to and at evolution (Figure 3B). In low-risk patients there were no discernable patterns, but many ASXL1, DNMT3A and BCOR/L1 clones appeared stable over years (Supplemental Figure 17). No low-risk patients had RUNX1 or splicing factor mutations either at time of clonal evolution or on pre-evolution samples.

Somatic mutation data were available for 407 patients at 6 months after IST. One hundred and eighteen patients (29%) had a somatic mutation with 32 (27%) of those having >1 mutation. Patients had the following mutations: BCOR/L1 (n= 43), ASXL1 (n=35), DNMT3A (n=26), TET2 (n=3), Splicing factor genes (n=5), TP53 (n=5), RUNX1 (n=4), JAK2 (n=3); other infrequent mutations occurred in 20 patients. The presence of ASXL1, RUNX1 or splicing factor mutations predicted overall or high-risk clonal evolution (Figure 4A, B, D; Supplemental figures 10+11) but not low-risk evolution (Supplemental figures 15+16). BCOR/L1 and DNMT3A/TET2 were not predictive of clonal evolution (Figure 4C, Supplemental figures 12, 13, 14). When mutation groups that predicted (RUNX1/splicing + ASXL1) and did not predict (BCOR/BCORL1/TET2/DNMT3A) evolution were combined, CI of high-risk evolution was 17% at 5 years and 42% at 8 years in patients with RUNX1/splicing factors or ASXL1 compared to 10% for those with BCOR/BCORL1/DNMT3A/TET2 or 6% for those with no mutation.

Figure 4: Cumulative Incidence of clonal evolution according to somatic mutations at 6 months after IST.

Patients (n=407) were assessed for somatic mutations at 6 months after IST. (A) Rate of clonal evolution in patients with RUNX1/Splicing factor or ASXL1 compared to those with DNMT3A/TET2/BCOR, no mutation, and other mutations. Cumulative incidence of clonal evolution with RUNX1/splicing or ASXL1 was 12.8% at 2 years, 26.0% at 5 years, and 47.6% at 8 years. Rate of clonal evolution with DNMT3A/TET2/BCOR/L1 without other mutations is 6.0% at 2 years, 18.9% at 5 years, and 23.7% at 8 years. Patients with no mutation had high-risk clonal evolution of 8.1% at 2 years, 10.9% at 5 years, and 14.6% at 8 years. (B) Rate of high-risk clonal evolution in patients with RUNX1/Splicing factor or ASXL1 compared to those with DNMT3A/TET2/BCOR, no mutation, and other mutations. Cumulative incidence of high-risk RUNX1/splicing or ASXL1 was 10.5% at 2 years, 16.8% at 5 years, and 41.7% at 8 years. Rate of high-risk clonal evolution with DNMT3A/TET2/BCOR/L1 without other mutations is 4.0% at 2 years, 10.4% at 5 years, and 15.4% at 8 years. Patients with no mutation had high-risk clonal evolution of 3.7% at 2 years, 5.5% at 5 years, and 7% at 8 years. (C) Rate of high-risk clonal evolution in patients carrying only BCOR/L1 mutations as compared to non-mutant cases. (D) Rate of high-risk clonal evolution in patients carrying only ASXL1 mutations as compared to non-mutant cases. Cumulative incidence (CI) curves were drawn using Fine-Gray competing risk method with death as a competing risk.

Conclusions

In the present study, we identified older age at diagnosis, high pre-treatment ANC, and the presence of ASXL1 or RUNX1/splicing factor mutations 6 months after IST as strong clinical predictors of clonal evolution, using a large cohort of consecutively treated patients with deep clinical phenotyping. Cumulative incidence for those with high-risk predictors showed rates of long-term clonal evolution, including high-risk evolution, far exceeding those reported in the literature. Other variables such as other baseline blood counts, presence of a PNH clone, type of IST, CN 6p-LOH, and HLA loss were not predictive.

Recently published long-term results from our IST+EPAG trial did not find telomere length (<10^th percentile in lymphocytes) predicted clonal evolution; this was not re-analyzed in this study as no additional data were available(16). Previous publications from NIH have shown telomere length as a predictor of clonal evolution; these used a different laboratory assay (qPCR not flow-FISH) and were reported differently (by 1^st [shortest] quartile compared with <10^th percentile used in the recent cohort), making a direct comparison challenging. Likely, telomeres in the shortest range (<1% percentile; most equivalent to patients in the shortest quartile) predict for clonal evolution based on previous findings, however, we will require larger data sets to confirm.

HLA class I allele loss detected by flow cytometry or sequencing and HLA-B*14 genotype recently were associated with clonal evolution by our group. When HLA loss, high-risk genotype, and age >40 years were combined they strongly predicted for clonal evolution. In another recent paper, HLA class I loss alone was not associated with an increased risk of clonal evolution(24). HLA type II evolutionary divergence in AA has also recently been associated with clonal evolution (23, 35). In this updated analysis with longer follow up time HLA loss alone was not predictive of clonal evolution, supporting that it is not a risk factor independent of age; HLA genotype was not assessed.

While risk of clonal evolution with increased age was expected and has been previously reported, why high ANC predicts clonal evolution is unclear. Older age is associated with increased rates of both CH, and myeloid neoplasia, independent of an AA diagnosis(36, 37). Preserved ANC may reflect a survival advantage in myeloid progenitors or age-related myeloid skewed hematopoietic stem cells related to underlying clonality(38, 39). IBMFS may also be associated with a high rate of clonal evolution and patients may have more preserved counts. In our current cohort, IBMFS was excluded by genetic screening in nearly 2/3 of patients, however, 1/3 of patients were not genetically assessed.

The clinical implications of isolated non-chromosome 7 abnormalities have been unclear. Some published series suggested positive outcomes with deletion 13q and trisomy 8(8, 25). Our large study showed excellent long-term survival in patients with low-risk clonal evolution. Most did not progress to high-risk disease, and therefore isolated chromosomal abnormalities should not necessarily prompt diagnosis of MDS and can be expectantly managed. However, of those that did transform, all but one had preceding 13q deletion; these patients may require closer clinical monitoring than others with low-risk evolution. Prior studies have shown excellent response to IST in patients with 13q deletion suggesting an immune component to this chromosomal acquisition(28).

In contrast, monosomy 7 was strongly associated with progression to high-risk MDS or AML, even if morphological dysplasia was absent when the abnormal karyotype was first detected. Outcomes were so poor that discovery of monosomy 7 warrants prompt interventional therapy directed at the myeloid neoplasm, even with normal morphology. Outcomes in non-monosomy chromosome 7 abnormalities were superior, as has been previously reported in the primary MDS literature, but numbers in our study were low and the majority were pre-emptively transplanted(27). Optimal treatment of myeloid neoplasia following SAA is not well defined. Standard high dose chemotherapy, hypomethylating agents or HSCT are employed depending on the patient’s risk profile. HSCT is widely offered to eligible patients prior to the development of overt dysplasia or leukemia; this practice is supported by the better overall survival seen in patients in our study. However, patients who received HSCT were significantly younger than those who did not, which may also influence their long-term outcomes. No adjustment was made for age, donor source, and type of conditioning due to low numbers. Further prospective data will be needed to fully determine the best therapy in this group.

Eltrombopag use was not associated with an increased risk of high-risk evolution. Overall clonal evolution risk was increased, but given the good outcomes seen with low-risk evolution in our cohort this likely does not translate into clinically significant risk. In the refractory setting, clonal evolution has been reported in nearly 20% of patients at early follow up, but to date, rates have not been increased in the treatment naïve setting(2, 3, 40). Though high-risk clonal evolution was not increased, time to evolution was shorter with EPAG; further long-term data are required and patients on EPAG should continue to be monitored for clonal evolution, particularly in the first 6 months after drug initiation.

The genomic landscape differed between patients with low and high-risk clonal evolution. Somatic mutations were more frequent than in high-risk compared to low-risk evolution and RUNX1, splicing factors, and ASXL1 predominated. The low-risk group had mutations more typically associated with CH of indeterminate significance (CHIP) and immune AA, such as DNMT3A, ASXL1 and BCOR/L1(6). These findings suggest a different clonal landscape in patients with high-risk evolution that may underly the development of overt leukemia. Both RUNX1 and splicing factor mutations were seen only in high-risk patients and in some cases were detected prior to the clonal evolution event, suggesting a potential role as drivers. Prior evidence had not definitively shown that the presence of myeloid neoplasm gene somatic variants in SAA predicts for development of clonal evolution. ASXL1 was initially reported to be more common in patients who ultimately developed clonal evolution, but subsequent publications were not confirmatory(3, 6). BCOR/L1 was previously shown to have a protective effect on overall survival. In the current study, ASXL1 and RUNX1/splicing factor mutations at 6 months after IST strongly predicted for clonal evolution, both overall and high-risk. DNMT3A, TET2, and BCOR/L1 showed a similar risk to non-mutated patients. Of note, SETBP1 and CBL, previously found to be enriched in clonal evolution, were seen infrequently in our study.

RUNX1 has previously been associated with clonal evolution in AA(6, 13, 14) and been detected prior to development of MDS and AML while being absent in non-evolvers(13). RUNX1 variants are found in approximately 10% of standard MDS patients, but other variants such as TET2, SF3B1 (both almost absent in our cohort), ASXL1, splicing factors, and DNMT3A are more frequent. The predominance of RUNX1 in the post AA MDS setting, in 35% of patients, along with ASXL1 as the commonest mutations, may give future insight towards the pathophysiology of high-risk clonal evolution(41, 42). While the mechanisms underlying the role of RUNX1 in the development of MDS/AML remain unclear, RUNX1 variants have also been reported in up to 20% of Fanconi anemia patients who develop MDS and as such may be driven by selective pressure from underlying marrow failure(43, 44).

The current work reveals predictors based on readily available clinical information to determine risk of clonal evolution prior to therapy. Our data also suggest that the presence of RUNX1, splicing factors, and ASXL1 six months after IST are predictive of clonal evolution and should prompt more frequent monitoring. Risk stratification using these predictors may identify patients requiring close long-term follow up or in whom alternative therapies such as HSCT should be considered earlier in the disease course.