Editorial theme | Molecular diagnosis and prognosis in MDS

The standard diagnostic method for myelodysplastic syndromes (MDS) is the World Health Organization (WHO) schema, while prognosis is given using the soon to be updated, International Prognostic Scoring System (IPSS). Although both provide discrimination of risk to help guide treatment decisions, the multifaceted outlook of MDS pathology encourages further detail, possibly through the integration of molecular genetics.

In this editorial theme piece on diagnosis in MDS, we summarize key points from two sessions on diagnosis of MDS, presented at the European School of Haematology (ESH) 8th Translational Research Conference: Myelodysplastic Syndromes. Torsten Haferlach¹ and Jaroslaw Maciejewski² discussed the viability and benefits of integrating molecular diagnostic/prognostic markers for MDS. We summarize their presentations below.

The benefit of molecular genetics in MDS diagnosis and prognosis

Genetic information can be used to identify unexplained cytopenia with discrete dysplasia and normal karyotype. Patients with clonal cytopenia of unknown significance (CCUS) who have ≥1 mutation have a significantly increased probability of evolution to myeloid neoplasms.³ A previous study has proven that it is possible to discriminate genes that are highly predictive of this evolution versus genes with low predictive capacity.³

Occurrence of genetic/molecular aberrations in MDS

Studies in patients with MDS indicate a high mutation burden that has the potential to aid in differential diagnosis. Table 1 shows the key information from studies cited by Haferlach.¹

Table 1. Studies showing a high mutation burden in patients with MDS*

ICUS, idiopathic cytopenia of unknown significance; IPSS, International Prognostic Scoring System; LFS, leukemia-free survival; MDS, myelodysplastic syndromes. *Adapted from Haferlach1
Reference/Year	Patients/ Genes examined (N)	Somatic mutation burden	Notes
Baer et al.4/ 2018	756/24	85% of patients with myeloid neoplasms; 34% in patients with ICUS.	MDS: patients with 82% of genes mutated; MDS possible: those with 52% mutated; MDS not visible: those with 38% mutated; Unclear: those with 30% mutated; Reactive change: those with 25% mutated.
Papaemmanuil et al.5/ 2013	738/111	78% (33% harbored cytogenetic abnormalities).	As the number of driver mutations increases, the rate of LFS decreases, which became significant when increasing from 4–5 driver mutations to ≥6 (p < 0.0001). This correlation was also true for IPSS low-risk patients.
Haferlach et al.6/ 2014	944/104	89.5% (31.4% had abnormal karyotype).	Also found to be important for prognosis. The discriminative power of determining risk (low, intermediate, high, and very high) when using 14 genes alone was comparable to that of the IPSS schema.
Malcovati et al.3/ 2017	683/40	86% of patients with myeloid neoplasms; 39% in patients with ICUS; 21% of patients with other cytopenia.	—

Impact of single genes in differential diagnosis

Genes like SF3B1 are known to be important at diagnosis.⁷ It has been demonstrated that luspatercept treatment led to better independence from red blood cell transfusions in patients with this mutation.⁷ Furthermore, it has been proposed by Malcovati et al.⁸ to include a specific subtype of MDS with mutated SF3B1. Haferlach¹ shared data from his lab, stating that when correlating the status of SF3B1 mutations with ring sideroblasts, more than 75% of patients with >15% ring sideroblasts had a SF3B1 mutation. Also, one third (33%) of patients with 5–14% ring sideroblasts carried the mutation.¹

However, Maciejewski² described the limitations of using molecular mutations for diagnosis by highlighting problems with variant allele frequency (VAF) resolution when differentiating between monoallelic and biallelic cells. The same VAF can be explained by various allelic inactivation (e.g., a VAF of 20% can be 40% monoallelic cells, 20% biallelic cells).²

Prognosis

When considering prognosis, a previous study⁹ showed that an increased mutation number was associated with both reduced overall survival (OS) (≥5 mutations; p < 0.0001) and leukemia-free survival (LFS) (≥5 mutations; p < 0.0001). Furthermore, when OS was stratified by the IPSS score proceeded by incorporation of a new molecular risk score, patients who were upstaged to lower risk scores had shorter OS while those who were downstaged to higher risk (intermediate, high, or very high) had improved OS.⁹

Another study¹⁰ outlined the impact of common clinical features, such as bone marrow blasts, chromosomal abnormalities, gene mutations, and demographics on OS. Finally, a study presented at the 63th American Society of Hematology (ASH) Annual Meeting and Exposition, by Bernard et al.¹¹, showed an analysis of the Molecular IPSS (IPSS-M) schema in a multivariable model—with all 16 genes tested having an impact.

Dominant and secondary mutations

Maciejewski² evaluated the validity of adding molecular prognostic markers to clinical tools and how to improve resolution of mutational information. Also important for prognostic schemas is identifying whether a mutation is a dominant (founder) or secondary (sub clonal). Primary mutations will have different impacts compared to the same gene mutated sub-clonally. This is true for some, but not all mutations. Both primary and secondary mutations can also have a positive or negative impact on survival.

In a study example, Maciejewski² discussed clonal hematopoiesis of indeterminate potential (CHIP). The mutation frequency in CHIP and MDS were compared. Table 2 shows how genes were classified into CHIP-related, undetermined, and non-CHIP-related.

Table 2. Relationship between CHIP and CHIP-MDS*

CHIP, clonal hematopoiesis of indeterminate potential; MDS, myelodysplastic syndromes. *Adapted from Maciejewski2
Classification	Dominant mutation
CHIP	PPM1D JAK2
CHIP-related MDS	DNMT3A TET2 ASXL1
Undetermined	TP53 SRSF2 SF3B1
Non-CHIP-related	U2AF1 RUNX1 STAG2 ZRSR2 EZH2

Patients with MDS were grouped into CHIP-related or non-CHIP-related based on combinations of dominant/ancestral mutations. CHIP-derived prognosis was found to be favorable (p = 0.03) for survival compared with de novo MDS.²

Morphological correlates

Morphological features should correlate with genetic features. For example, rare anemia with ring sideroblasts correlates with SF3B1 mutations. In the data discussed,¹² this correlation was tested using common parameters for MDS morphological profiles. To deal with the complexity, the researchers subclassified by using an unbiased approach with machine learning. The results were validated using an external data set and three morphological profiles with strong genetic correlations were identified (Table 3).¹²

Table 3. Correlations of morphological profiles and genetic signatures*

Mgk, megakaryocyte; MPN, myeloproliferative neoplasms  *Table from Nagata et al.12
Morphological profile	Genetic signatures
Trilineage dysplasia Pancytopenia No MPN feature	TET2mut, SRSF2wt
No mutation
Trilineage dysplasia Anemia + thrombocytopenia Monocytosis	TET2mut, SRSF2wt
TET2mut, SRSF2mut
Heterogenous, SF3B1wt
Erythroid + Mgk dysplasia Anemia Elevated Mgk	SF3B1mut, JAK2mut

Subclassification of risk

Maciejewski² discussed a study analyzing a total of 6,788 cases of primary and secondary acute myeloid leukemia (AML) which were molecularly profiled; four major molecular clusters were identified (low risk, intermediate-low risk, intermediate-high risk, and high risk). Both primary AML and secondary AML were almost evenly distributed in three of the clusters, excluding the low-risk cluster (where primary AML was overrepresented). When identifying primary/secondary AML via molecular signatures, there was poor subclassification accuracy.¹³

Finally, a new technique using unsupervised machine learning was used to identify molecular clusters based on similar molecular makeup in 3,598 cases of MDS/secondary AML.² Clusters were grouped by risk relationship (low risk, intermediate-low risk, intermediate-high risk, high risk, and very-high risk). The survival probability was calculated for these risk categories and there was a significant difference in OS probability.¹¹

Summary

In summary, genetic aberrations are an important diagnostic and prognostic tool for MDS and can provide further discrimination of risk, which will help guide patient-specific treatment decisions and provide new opportunities for targeted therapies. Maciejewski² highlighted several techniques to improve resolution including mutation clonality, copy number changes, microduplication/deletions plus clonality, germline variants, ancestral vs sub-clonal mutations, impact of rare hits, and CHIP-derived vs de novo mutations.

Conclusion

Integration of molecular prognosis and diagnosis can add benefit to clinical features used in standard tools; however, further improvements to mutational resolution will improve this benefit.