De Novo Myelodysplastic Syndromes in Patients 20-50 Years Old are Enriched for Adverse Risk Features

Abstract

Data concerning the treatment approach and clinical outcomes in younger patients with myelodysplastic syndromes (MDS) are lacking. Furthermore, published results from genomic profiling in the young adult MDS population are few. We identified patients aged 20-50 at diagnosis evaluated for de novo MDS at our institution over a 32-year period. Clinical information and results from sequencing panels were extracted for analysis. 68 eligible patients were found, including 32% with multilineage dysplasia and 29% with excess blasts-2 WHO subtypes. Revised International Prognostic Scoring System for MDS (IPSS-R) categorization had 47% high/very high-risk, and this classification held prognostic significance. The median overall survival was 59 months, and most patients (75%) underwent allogeneic hematopoietic cell transplantation (alloHCT). Thirty-four patients had mutational profiling; the most commonly mutated gene was TP53 and most commonly altered gene category was epigenetic regulators. Younger patients with de novo MDS represented a unique subset with high-risk disease features (adverse cytogenetics, higher R-IPSS) frequently observed along with alterations in TP53 and genes related to epigenetic and transcription pathways.

1. Introduction

Myelodysplastic syndromes (MDS) encompass a diverse group of clonal stem cell disorders. The defining disease characteristics include dysplastic bone marrow changes, cytopenia(s) of one or multiple cell lineages and clinical sequalae therein, and the potential for disease transformation to acute myeloid leukemia (AML). Epidemiologically, MDS is principally a disease of older adults, with most cases seen after age 60 and an increasing incidence with advancing age [1].

Recent studies have elucidated the molecular pathogenesis underlying MDS [2, 3]. These pivotal investigations have identified recurrent genetic mutations in many gene families implicated in the underlying disease biology in MDS, including transcription factors, cellular signaling pathways, RNA splicing, DNA methylation, and chromatin/histone modification. In one large such study [3], the most common mutations were in genes encoding for proteins involved in RNA splicing SRSF2, SF3B1, U2AF1, U2AF2, and ZRSR2 (64%). Another study sequenced 111 genes in 738 patients [2], also finding mutations in SF3B1 to be most common. Furthermore, mutational patterns differed when comparing results from patients with treatment-related MDS (t-MDS) and those with de novo disease [4].

Besides illuminating the fundamental biological changes underlying the development of MDS, results from mutational testing in contemporary practice can also hold critical prognostic [2] and therapeutic [5] implications for patients with MDS with the advent of rational genomically targeted treatment approaches. However, given the predominant pattern of disease incidence in older adult patients, these molecular studies are inherently skewed towards elucidating the pattern of genetic lesions present in older adult patients. Few studies [6-8] have examined molecular alterations present in younger patients with MDS, yet these have principally focused on germline variants and inherited bone marrow failure syndromes (BMFS) and/or predisposition towards MDS. Furthermore, other prognostic considerations such as cytogenetics or prognostic scoring indices have not been comprehensively explored or validated in younger adult patients with MDS.

We conducted a retrospective analysis of all patients with de novo MDS evaluated at our institution who were aged 20-50 at the time of diagnosis. Our objectives were to understand the clinical patterns of disease, explore prognostic factors, define treatment modalities used, and to explore the molecular features defining MDS in these patients, utilizing results from next-generation sequencing (NGS) testing panels.

2. Materials and Methods

2.1. Patients

We performed a retrospective query for all patients seen at our institution diagnosed from 1987 through 2019 with a diagnosis corresponding to any MDS subtype who were aged 20-50 at diagnosis. Choosing this age range, we sought to exclude 1) younger children/adolescents in whom MDS frequently exists in the context of an inherited BMFS and 2) the general, older adult population of patients with MDS with a median age of onset of 70-72 years [9]. We excluded patients with MDS-myeloproliferative neoplasm overlap syndromes including chronic myelomonocytic leukemia. To focus exclusively on de novo cases, we also excluded patients with a known or suspected underlying inherited BMFS and patients with documented exposure to benzene, chemotherapy, or radiation as these represent distinct entities from de novo MDS. The MDS subtype was assigned based on the Revised 2016 World Health Organization (WHO) Criteria [10]; cytogenetic risk profiles and prognostic risk groupings were designated at diagnosis according to the Revised International Prognostic Scoring System for MDS (IPSS-R) [9] with available data. This investigation was approved by the Institutional Review board at Memorial Sloan Kettering Cancer Center and all study practices were in accordance with the Declaration of Helsinki.

2.2. Sequencing Methods

Tumor samples (33) were analyzed with a commercial (Foundation Medicine, n=2; GenPath, n=1; Genoptix, n=1) or proprietary (n=29) CLIA-certified NGS assay interrogating a broad panel of genes implicated in myeloid neoplasia. When multiple sequencing tests were performed in the same patient, the earliest result was chosen for inclusion/analysis. Over time, our proprietary institutional sequencing panel [11, 12] evolved to include successively more genes, therefore different samples were analyzed with assays featuring different gene coverage (9 samples with 30 genes and 1 sample each with 49 and 400 genes; Supplement 1 details sequencing panel coverage). A subset of patients had not undergone sequencing, and in those cases, we retrieved archival tissue and performed sequencing according to a similar methodology to our most recent proprietary panel (MSK IMPACT [12]; 20 samples).

Sequencing on archival tissue samples was performed using a targeted panel designed to capture genes (Supplement) recurrently implicated in the pathogenesis of myeloid neoplasia [2, 13]. Samples were sequenced on Illumina HiSeq 4000 with 126 BP paired-end reads and an average target depth of 683.35x (range, 438X-1020X). The raw sequence data was aligned to the GRCh37 reference genome using BWA-MEM algorithm (v. 0.7.17) [14]. The data quality was assessed using FastQC (v. 0.11.5) [15]. Candidate single nucleotide variants were called using cgpCaVEMan (v. 1.7.4) [16], Mutect [17] version 4.0.1.2 and Strelka [18] (v. 2.9.1). For small insertions and deletions, we used cgpPindel (v. 1.5.4) [19], Mutect version 4.0.1.2, and Strelka (v. 2.9.1). These methods provide post-hoc filters that help to remove sequencing artifacts. Candidate mutations were annotated with VAGrENT version 3.3.0 [20] and Ensemble (v.91)-VEP(v.92) [21]. They were compared to the COSMIC (v. 81) [22] and Genome Aggregation Database [23] databases along with recurrence in a panel of normals to provide further information about the prevalence of each mutation in human cancer and normal populations. This information is used for the manual curation of each variant in order to classify it as pathogenic, likely pathogenic, or a variant of uncertain significance. To ensure the quality of reported somatic variants, the sequence of aligned reads was visualized for confirmation and interpretation using Integrated Genomics Viewer (v. 2.3.92) [24].

We assessed chromosomal level copy number changes based on NGS sequencing data using CNACS [25], which enables the detection of arm level and focal copy-numbers changes, regions of copy-neutral loss of heterozygosity (cnLOH), and is optimized to run in the unmatched setting. The selected high-confidence variants presented in this study are those that were identified as pathogenic or likely pathogenic. Categories of gene functional pathways/mechanisms were assigned based on published reports [2, 3] and publicly-available annotated genomic databases.

2.3. Statistical Analysis

A Kaplan-Meier method with left truncation was used to estimate the overall survival from time of diagnosis and log-rank test was used to determine the association between categories and overall survival. Effects of patient and disease characteristics on overall survival were estimated by univariate Cox proportional hazard model with P <0.05 being considered significant. A reverse Kaplan-Meier method was used to calculate duration of follow-up. For patients who underwent alloHCT, overall survival was estimated from time of transplantation by the Kaplan-Meier method. All analyses were done using the R statistical software.

2.4. Data Sharing Statement

Genomic data are available at GEO under accession number PRJNA757747

3. Results

3.1. MDS Disease Characteristics

A total of 148 patients were identified to match our study inclusion criteria; of these, 12 were excluded due to possible (clinical BMFS phenotype or strongly positive family history of myeloid neoplasia without formal germline sequencing performed) or confirmed (germline sequencing demonstrating a known pathologic variant) underlying BMFS, 9 for MDS-myeloproliferative neoplasm overlap syndromes, 49 for prior chemotherapy or radiation, 9 due to insufficient follow-up information, and 1 due to benzene exposure. Table 1 displays disease and demographic data for included patients (N = 68). Follow-up includes a median of 80 months from diagnosis. There was a near-equal distribution of sex; the most common MDS subtype was multilineage dysplasia (MLD) in 22 patients (32%). By cytogenetic risk category, there was a near-equal distribution of <intermediate risk compared to ≥ntermediate risk (N=34 or 52% versus N=32 or 48%), whereas by IPSS-R, only N=30 or 31% of patients were classified <Intermediate risk.

Table 1.

Patient Characteristics (N = 70)

Variable			N (%)
Age at diagnosis, years (median, range)			42 (22-50)
Sex, female			38 (54%)
WHO MDS Subtype [10]
	Single-lineage dysplasia		6 (9%)
	Multilineage dysplasia		22 (32%)
	Excess blasts 1		12 (18%)
	Excess blasts 2		20 (29%)
	Deletion 5q		4 (6%)
	Other*		4 (6%)
BM fibrosis/severity a
	None/none-mild		2 (40%)
	Mild/mild-moderate		23 (42%)
	Moderate/moderate-severe		7 (13%)
	Severe		3 (5%)
Cytogenetic risk category b
	Very good		1 (1%)
	Good		33 (50%)
	Intermediate		12 (18%)
	Poor		11 (17%)
	Very poor		9 (14%)
IPSS-R c
	Median (range)		4.5 (0-8.5)
	Prognostic Risk Group	Very low	6 (9%)
		Low	14 (22%)
		Intermediate	14 (22%)
		High	13 (20%)
		Very high	17 (27%)
Treatments d
	Allogeneic transplantation		52 (79%)
	Hypomethylating agent		35 (53%)
	Chemotherapy		28 (42%)
	Other		15 (23%)

^*Includes MDS-Unclassifiable, and RS-MLD

^aN = 56 patients with available information

^bN = 67 patients with available information; three patients harbored multiple sub-clonal cytogenetic populations and were categorized according to the poorest-risk subclone present

^cN = 64 with full IPSS-R information

^dN = 68 patients with clinical follow-up

3.2. Cytogenetic and Molecular Alterations

The most common single cytogenetic abnormality was deletion 5q in 13 patients, followed by trisomy 8 in 10 patients, and monosomy 7 in 6 patients. Figure 1 displays results from sequencing panels (33 patients), which occurred within 6 months of diagnosis/prior to treatment in 21/33 patients. The most commonly mutated gene was TP53 (8 mutations in 7 patients) followed by SF3B1 (6 patients). In 6 patients, no mutations were found (2 with limited panel testing, 3 with broad/expanded panel, and 1 with outside testing panels). As displayed in Figure 2, by category (defined in Supplement 2), there was a preponderance of genes related to epigenetic regulators (21 mutations) and transcription/DNA repair (17 mutations). Supplement 3 details sequencing results from in-house expanded panel on archival samples.

Figure 1: OncoPrint (N = 33 patients).

Each column represents an individual patient, characterized by sequencing panel, WHO MDS subtype, clinical status at time of sequencing. The middle panel displays the frequency of each gene. *8 mutations in 7 patients; **3 mutations in 2 patients

Figure 2: Mutations by Gene Category (N = 33 patients).

Gene categories are listed with the corresponding number of mutations in each category.

3.3. Therapeutic Approach

Among patients with available follow-up/treatment information (defined as ≥2 visits at our institution or 1 visit with subsequent treatment records, N = 66), the most common treatment modality utilized was alloHCT, in 52 patients (79%); 6 patients underwent a second alloHCT following disease relapse. Besides alloHCT, hypomethylating agents were commonly utilized (N = 35 patients, 53%). In 3 patients, anti-thymocyte globulin and cyclosporine A were given initially for possible aplastic anemia, but the diagnosis was ultimately felt to be hypoplastic MDS.

3.4. Survival Analyses

The estimated 5-year survival for all patients was 48% (95% CI: 38-63) and the estimated 5-year survival from time of alloHCT 54% (95% CI: 42-70) (Figure 3). With a median follow-up of 80 months (95% CI: 68.8-120), the median overall survival was 59 months (95% CI: 31.3-NR) among all patients. As displayed in Table 2 and Figure 3, cytogenetic risk categories (Good/very good versus Intermediate versus Poor/very poor) harbored prognostic importance (P = 0.0003). Patients with Poor/very poor cytogenetics compared to Good/very good had inferior overall survival: median 19.2 (95% CI: 15.3–43.8, HR 3.1, 95% CI: 1.51–6.35) versus 109.1 months (95% CI: 45.3.–NR). IPSS-R score at diagnosis (Low/very low versus Intermediate versus High/very high) also was prognostic (P = 0.001). Patients with TP53-mutated disease had inferior survival (P = 0.023). Patient sex, degree of bone marrow fibrosis, age, WHO MDS subtype, and use of alloHCT were not prognostic. In total, 28 patients progressed to AML, of which 11 occurred after alloHCT. Among the 51 patients who underwent alloHCT, the median follow-up time after transplantation was 77 (95% CI: 57.3-112.7) months and median overall survival was 74 months (95% CI: 18.4-NR).

Figure 3: Overall Survival.

The estimated 5-year survival (A) for all patients from diagnosis was 48% (95% CI: 38-63) and (B) from time of alloHCT 54% (95% CI: 42-70). (C) Cytogenetic risk (C, P = 0.0003) and IPSS-R score at diagnosis (D, P = 0.001) harbored prognostic importance.

Table 2.

Predictors of Survival

Variable		HR (95% CI)	P a
Sex			0.3
	Female	1
	Male	1.39 (0.71-2.74)
Age			0.2
	<35	1
	35-42	0.53 (0.18-1.59)
	42-47	1.58 (0.63-4.0)
	>47	1.50 (0.62-3.63)
WHO MDS Subtype			0.7
	Deletion 5q	1
	Single-lineage dysplasia	1.78 (0.16-19.7)
	Multilineage dysplasia	2.43 (0.36-18.6)
	Excess blasts 1	3.91 (0.49-31.1)
	Excess blasts 2	2.84 (0.36-22.3)
	Other	2.65 (0.24-29.3)
BM fibrosis/severity			0.9
	None/none-mild	1
	Mild/mild-moderate	0.95 (0.41-2.19)
	Moderate/moderate-severe/severe	1.17 (0.44-3.1)
Cytogenetic risk category			0.0003
	Very good/good	1
	Intermediate	0.42 (0.09-1.84)
	Poor/very poor	2.85 (1.51-6.35)
IPSS-R			0.001
	Very low/low	1
	Intermediate	2.63 (0.66-10.51)
	High/very high	6.63 (1.96-22.5)
TP53 Status *			0.023
	Wildtype	1
	Mutated	6.3 (1.03-38.36)
Allogeneic transplantation **			0.3
	No	1
	Yes	1.54 (0.69-3.44)

^aLog-rank

^*when assessed at diagnosis

^**time-dependent survival analysis

4. Discussion

Our retrospective study characterized a large cohort of patients with de novo MDS diagnosed between age 20-50 (median age, 42 years). We captured clinical/disease characteristics, cytogenetic alterations, molecular changes, treatment patterns, and clinical outcomes with a long period of follow-up. A substantial proportion of patients, 75%, underwent alloHCT, and overall outcomes were favorable despite adverse disease characteristics. Our report is the largest to characterize molecular alterations in this specific patient population. Our data prompt further confirmatory studies from other centers and/or cooperative groups to further characterize MDS disease features in these patients.

Examining disease characteristics, our results suggest a more unfavorable-risk patient population when compared to available general adult MDS cohorts, Vose and colleagues [26] validated the prognostic utility of the IPSS-R in a cohort of 380 patients with MDS with a median age of 71 years. Poor/very poor cytogenetics were present in 4% and 3% of patients respectively, compared to 17% and 14% in our cohort and carried prognostic significance. Furthermore, combined IPSS-R high/very high groups totaled 11%, compared to 47% at our center, and IPSS-R was also predictive of outcome, which had not been previously demonstrated specifically in this age category.

High-risk IPSS-R prognostic grouping was influenced by a preponderance of patients in the poor-prognosis excess blasts WHO disease subtypes (total 47%), which score highly in conferring adverse risk in this model, and these findings are also discordant with results in general adult MDS cohorts (for example, Voso et al. [26] reported total 27% for these same two groups). A previous single-center study of 51 adolescent/young adult patients aged 18-39 years showed a similar percentage of excess blasts (EB)1/EB2 cases compared to older adult MDS patients seen at their center [27]. The distribution of patients among cytogenetic risk categories in this study was similar to that in our cohort: by IPSS-R groupings, although similar values were seen for very low/low (30.1% versus 31% in our cohort), their cohort had a lower rate of high/very high-risk patients (37.3 versus 47%). As a referral center, there is the potential for bias towards such challenging cases, yet the magnitude of variance we have observed suggests the possibility of a true difference in disease patterns existing between age groups. Other patient/disease parameters we examined were not prognostic, including the degree of bone marrow fibrosis. This specific question has been examined previously in general adult MDS series, including by Ramos and colleagues [28], who found fibrosis grade 2 or higher to predict overall survival independent of age, performance status, and IPSS-R.

Our use of sequencing panels allows for preliminary molecular insights into MDS in younger adults. Extending our other findings of adverse-risk disease features, we found mutations in TP53 to be most common (8 mutations in 7 patients – 21% (95% CI: 9-39)). Although variability exists in the literature, the rate of TP53 mutation was 5-10% [3] in a study of 944 patients with MDS who underwent sequencing. We found mutated TP53 at diagnosis to be prognostic. Of these 7 patients, 5 harbored Poor/very poor cytogenetics and the remaining 2 were +del(5q). Results for other gene mutations aligned with adult populations, with enrichment for alterations in SF3B1, ASXL1, DNMT3A, TET2, and STAG2 following TP53 in frequency. Interestingly, we identified only one patient with mutated RUNX1, which is observed in 10-15% of adults with MDS. One previous study from the Fred Hutchinson Cancer Research Center explored the rate of germline mutation in inherited BMFS genes in younger patients undergoing alloHCT for MDS [6]. They found 13.6% of such patients harbored mutations, including in FANCA, MPL, GATA2, TP53, and RUNX1 among others. Pastor and colleagues [8] analyzed 50 children (maximum age 17 years) with “advanced-disease” (enriched for monosomy 7 and excess blasts subgroups) MDS and did not detect mutations in DNMT3A nor TP53. One possibility for differing mutational profiles is that older adult patients’ disease has arisen from age-related clonal hematopoiesis (CH) precursor states, which have distinct mutational patterns [29], whereas younger adult patients progress from a different precursor state or CH with a different mutational landscape.

Beyond specific genes, we may compare gene pathway alterations in our cohort with those seen in general adult MDS cohorts that have been molecularly defined. In contrast to our cohort with enrichment for mutations in genes involved in transcription/DNA repair, Haferlach and colleagues [3] in a general adult MDS population found a high frequency of mutation in genes involved in RNA splicing (64% of patients) followed by DNA methylation (45-50%). These same data were merged with two other data sets by Kennedy and Ebert [30] to confirm this pattern of gene mutations. Our cohort had 10 RNA splicing mutations, although SF3B1, which encodes for a subunit of the U2 small nuclear ribonucleoproteins complex involved in RNA splicing, was the most commonly mutated non-TP53 gene (N = 6 patients).

Given considerations around patient age/comorbidity and adverse disease features, two of the critical decision points around use of alloHCT for patients with MDS, the patterns of treatment we observed are unsurprising, with utilization of alloHCT in the majority (77%) of patients. Our cohort spanned a wide time period and during this period the availability of donor options expanded with more routine use of and comfort with unrelated, cord blood, or haploidentical donor sources, which may also contribute to the high use of alloHCT. The non-transplant group in our cohort likely included 1) patients with insufficiently remitted disease to allow pursuit of alloHCT and 2) patients with low enough risk disease to warrant a non-alloHCT-directed approach. Although we did not detect statistically different survival outcomes between these two patient groups, this finding likely reflects the existence of lower-risk patients with more indolent disease not pursuing an unnecessarily aggressive treatment approach.

The 53% survival rate unsurprisingly compares favorably to adult cohorts, likely due to fewer comorbidities and lower age, which are both broadly predictive of outcome following alloHCT. For example, an analysis of 1,333 patients with MDS aged >50 years through the European Group for Blood and Marrow Transplantation (EMBT) estimated a 4-year overall survival of only 31% [31]. Larger registry-based studies could validate our findings and further clarify the role for alloHCT for younger adult patients with MDS.

Our study characterized a unique patient cohort but has shortcomings that warrant attention. First, as this was a retrospective analysis including patients cared for over many years, there was notable heterogeneity with regards to the sequencing panel used (this evolved over time to include additional genes) and the timing of testing around treatment initiation. Furthermore, only a subset of patients had mutational testing, which was not explicitly used to identify inherited BMFS, limiting our ability to make definitive conclusions based on our results. Future analyses will ideally characterize groups of patients who undergo sequencing uniformly by specific assay and time point(s) of sequencing. Moreover, this shortcoming highlights the importance of implementing more comprehensive mutational profiling (both number of patients tested and number of genes included in testing panels) for younger patients with MDS beyond those genes that are highly recurrently mutated (e.g., TP53, SF3B1, TET2). Secondly, our analysis is retrospective in nature and occurred at a single center, which could bias findings towards challenging referral cases and a single geographic region and potentially limits applicability to all young adult patients with MDS. Future studies in this area could be multicenter in nature to garner a diverse array of cases.

In conclusion, younger adult patients with MDS often show more adverse disease features by cytogenetics and IPSS-R scoring. Mutations in TP53 and genes related to transcription/transcription factors predominate. Future studies in young adult patients with MDS should primarily address confirming our mutational profiling results, validating the prognostic significance of specific gene mutations, and further defining treatment patterns and clinical outcomes for this specific patient population.

Highlights:

• Younger patients with de novo MDS frequently exhibit high-risk disease features

• Most such patients are treated with allogeneic hematopoietic cell transplantation

• Epigenetic regulators and transcription/DNA repair genes were commonly mutated