MDS is a complex myeloid clonal hematopoietic disorder that presents with both myeloproliferative and ineffective hematopoietic components, resulting in cytopenia [7]. It has a high risk of transformation to AML [8]. Although not all patients progress to AML, many die of infection, bleeding and/or cardiopulmonary complications.

The pathogenesis of MDS is generally accepted to involve clonal expansion of hematopoietic stem cells caused by somatic mutations. Accumulation of driver gene mutations or chromosomal abnormalities also triggers AML [9]. More than 30 genetic mutations have been associated with MDS, which can be categorized into splicing factors (e.g., SF3B1, SRSF2, U2AF1, ZRSR2), DNA methylation (e.g., DNMT3A, TET2), histone modification (e.g., ASXL1, EZH2), cohesion components (e.g., SMC1A, SMC3, RAD21, STAG1 and STAG2), transcription factors (e.g., RUNX1), TP53 pathway and signal transduction (e.g., NRAS, KRAS, FLT3) [10, 11, 12].

The 5th edition of the World Health Organization (WHO) classification introduces the term myelodysplastic neoplasms to replace myelodysplastic syndromes. MDS is classified on the basis of genetic abnormalities and morphology. MDS with defining genetic abnormalities includes MDS with low blasts and isolated 5q deletion (MDS-5q), MDS with low blasts and SF3B1 mutationa (MDS-SF3B1), and MDS with biallelic TP53 inactivation (MDS-biTP53). Morphologically defined MDS includes MDS with low blasts (MDS-LB), hypoplastic MDS (MDS-h), and MDS with increased blasts (MDS-IB). MDS-IB can be further subdivided according to the number of blasts and whether fibrosis is combined [13].

According to the new International Consensus Classification (ICC) of myeloid neoplasms, MDS is divided into MDS without excess blasts, MDS with excess blasts and MDS/AML. MDS without excess blasts can be further subdivided into MDS with mutated SF3B1, MDS with del(5q) and MDS, not otherwise specified (NOS). MDS, NOS can be further subdivided according to the number of dysplasias. The subtype of MDS/AML emphasises the biological continuum between MDS and AML [7]. It should be noted that MDS, MDS/AML, and AML with mutated TP53 are grouped together as myeloid neoplasms with mutated TP53 due to similar aggressive behaviour [14].

Due to evolving diagnostic criteria and reclassification of MDS, it is difficult to describe the incidence of the disease in a consistent manner [15]. The incidence of MDS increases with age, and the most recent Surveillance, Epidemiology, and End Results (SEER) data report a median age at diagnosis of 77 [6]. Monitoring for disease relapse after MDS treatment can provide information on the efficacy of treatment and allow individualized adjustment of treatment plans to postpone disease progression, improve quality of life and prolong survival.

AML is a heterogeneous malignant disease characterised by rapid clonal expansion of aberrant myeloid progenitor cells [16, 17]. AML is the most common acute leukemia in adults and predominantly affects the elderly population. The WHO classification separates AML with defining genetic abnormalities from AML defined by differentiation [13]. Anderson et al. [18] performed an analysis of the age-standardized incidence of AML in 29 countries on four continents. They reported a median incidence of 2.28 cases per 100,000 people. Compared to patients with MDS, mutations involved in DNA methylation (e.g., DNMT3A), transcriptional dysregulation (e.g., RUNX1, CEBPA) and signalling (e.g., NRAS, KRAS, FLT3) are more common in patients with AML [19, 20].

Currently, the diagnosis of MDS and AML is mainly based on comprehensive analysis of cell morphology, immunology, cytogenetics and molecular biology (MICM) of samples obtained by bone marrow aspiration and biopsy. Chromosome banding analysis, fluorescence in situ hybridization (FISH), polymerase chain reaction (PCR) and next generation sequencing (NGS) are conventional methods of detecting genomic characterization [21].

DNA hypomethylating agents (HMAs), azacitidine and decitabine, have become the backbone of MDS and AML treatment [22]. Currently, for patients with high-risk or refractory AML and MDS, allogeneic hematopoietic stem cell transplantation (allo-HSCT) may be the only way to potentially achieve long-term survival. Therefore, early diagnosis, early treatment, timely monitoring of treatment efficacy and appropriate adjustment of treatment are particularly important in MDS and AML.

Considering the repetitive nature of bone marrow aspiration and biopsy during treatment and monitoring, pain experienced by the patient and bleeding at the puncture site are unavoidable [23]. Based on the disadvantages of traditional biopsy techniques, we urgently need a convenient and non-invasive means to aid diagnosis and treatment. Thus, the concept of “liquid biopsy” was born.

In 1869, the pathologist Ashworth discovered a class of cells whose morphology resembled that of tumours in the peripheral blood of patients dying of cancer. He named them circulating tumour cells (CTCs). CTCs are tumour cells that have been shed from a tumour lesion (primary or metastatic) into the bloodstream, and can be present as a single cell or in clusters. In 2010, Pantel et al. [24] first used the term “liquid biopsy” in a review of CTCs to describe the clinical applications of analyzing peripheral blood samples. The analysis of peripheral blood CTCs can screen for early tumours, detect tumour progression, facilitate the monitoring of response to therapy, study mechanisms of drug resistance and even be used to detect the emergence of distinct resistant clones and monitor patterns of early disease recurrence [25, 26, 27].

As the concept of liquid biopsy has been developed, circulating cell-free DNA (cfDNA), RNA and exosomes have become markers analyzed in liquid biopsy, particularly cfDNA. CfDNA refers to extracellular DNA found in body fluids such as plasma, serum and urine [28, 29, 30, 31]. In 2019, Alborelli et al. [32] conducted a study of cfDNA in 177 individuals, 63 of whom were diagnosed with lung or breast cancer and 114 of whom were healthy. They reported that the concentration of cfDNA in plasma ranged from 0–100 ng/mL in healthy individuals, which is lower than that in cancer patients. Cell-free plasma DNA is mainly derived from apoptotic or necrotic cells [33]. Normally, DNA from apoptosis is rapidly degraded by DNases [34]; however, in the case of acute injury, stroke, transplantation, infection, tumour, etc., as the uptake of apoptotic bodies is impaired or a large number of apoptotic cells are produced, the concentration of cfDNA in the plasma will increase [35]. In patients with tumors, DNA fragments released from tumour cells within tumour lesions and CTCs are important sources of cfDNA. The type of cfDNA derived from tumours is called circulating tumour DNA (ctDNA) (Fig. 1).

Fig. 1.

A schematic diagram describing the source and use of circulating tumour DNA (ctDNA). Peripheral blood is the most common source of liquid biopsies. Tumour cells release ctDNA by apoptosis, necrosis and secretion, which can be isolated and enriched from a blood sample. The information extracted from ctDNA can be broadly divided into quantitative information or genomic information for analysis.

CtDNA is derived from necrosis/apoptosis of tumour cells and CTCs, and from their exosomes. Mouliere et al. [36] reported that the cfDNA fragments at 167 base pairs (bp) can be released by apoptosis, maturation and turnover of blood cells. Mutant ctDNA has a maximum enrichment in fragments between 90 and 150 bp, with additional enrichment in the 250–320 bp size range. Mutated ctDNA is generally more fragmented than non-mutated cfDNA [36]. Peripheral blood from cancer patients contains ctDNA. However, this is diluted by much larger amounts of DNA shed from non-cancerous sources [37]. The half-life of ctDNA is 15 minutes to a few hours and can be cleared rapidly by the kidneys, liver and spleen.

The detection of ctDNA, which is present in very low concentrations in peripheral blood, requires high technology. With the development of PCR and the spread of NGS, highly sensitive detection methods have become feasible, making it possible to detect ctDNA. However, no method yet meets all the requirements for clinical applicability. Here, we list the commonly used methods for ctDNA detection in peripheral blood (Table 1, Ref. [38, 39, 40, 41, 42]).

The genetic and methylation information in ctDNA maintains a high degree of correlation with the tumour from which it was derived [32, 43, 44]. CtDNA can serve as a non-invasive pool for cancer detection and diagnosis by analyzing cancer-specific biomarkers [45, 46, 47]. In 2020, GRAIL published the results of three sub-projects of the Circulating Cell-free Genome Atlas (CCGA). Experiments proved that its methylation sequencing analysis technology of cfDNA can detect more than 50 cancer types simultaneously, with a specificity of $>$99%, and the prediction accuracy of the tissue of origin (TOO) is $>$90% [48].

Much research has shown that the detectable levels of ctDNA correlate with cancer stage and burden, and that patients who have undergone cancer resection may have decreased ctDNA levels [49, 50, 51, 52]. In addition, it is possible to analyze cancer evolution, genomes and transcriptomes using ctDNA [53]. CtDNA also has the potential to serve as a biomarker for therapy evaluation [54, 55]. Research has been conducted in various cancer types on prognostic assessment, MRD surveillance and detection of disease recurrence by serial ctDNA tracking, which has been shown to be more sensitive for detection of recurrence than traditional methods [56, 57, 58, 59, 60, 61].

Here we list some of the above references on the use of ctDNA in solid cancers (Table 2, Ref. [46, 47, 49, 53, 57, 58]).

Much research has been done on ctDNA associated with solid tumours. However, due to differences in tissue collection methods and observation of genetic abnormalities, there is a need for further research in hematopoietic tumours. Currently, ctDNA is being investigated in MDS and AML for the detection of tumour-specific mutations, evaluation of therapeutic efficacy, prediction of prognosis, detection of MRD and assessment of genomic heterogeneity. The use of cfDNA in MDS and AML is summarized below.

In 1994, Vasioukhin et al. [62] detected N-ras gene mutations in cfDNA, blood cells and bone marrow from 10 patients with MDS or AML using PCR. They showed that ras mutation information detected in cfDNA was not always detected in blood cells or bone marrow, suggesting that bone marrow biopsy or aspiration may not necessarily contain all the information about the malignancy. Plasma contains more comprehensive information about the tumour. Taking advantage of its easy availability, plasma can be used to detect and monitor myeloid disorders. Since then, more and more research has been carried out into the use of cfDNA in MDS and AML.

In 2004, Rogers et al. [63] compared the difference in loss of heterozygosity (LOH) between plasma ctDNA and bone marrow in MDS and AML patients with chromosomal abnormalities. They found that LOH was detected in 100% of cases in plasma ctDNA compared to 89% and 70% in bone marrow in MDS and AML, respectively. In the detection of residual disease, fifteen out of sixteen post-therapy samples showed complete concordance between LOH and cytogenetics. These data suggest that peripheral blood plasma is enriched for ctDNA and has the potential to replace bone marrow for the study of genetic abnormalities.

As the technology developed, in 2019, Zhao et al. [64] performed a gene detection study in 26 patients with MDS using a NGS platform with a 127-gene panel. The study strongly suggested high correlations between bone marrow tumour DNA and peripheral blood plasma ctDNA for both variant cells and somatic mutations. In addition, the frequency of gene mutations in ctDNA can, to some extent, predict and reflect disease status and monitor therapeutic response in MDS. Another study examined the molecular and cytogenetic profile of a cohort of 70 patients with MDS by NGS of cfDNA and compared the results with paired bone marrow DNA. It supported that analysis of cfDNA can serve as a promising strategy for performing molecular characterization, detection of chromosomal aberrations and monitoring of patients with MDS [65].

In 2020, Tiong et al. [66] used a single hybridization-based NGS assay of ctDNA and bone marrow to perform low-resolution digital karyotyping. They focused on the detection of sequence variants, copy number variation (CNV) and copy neutral loss of heterozygosity (CN-LOH) in AML patients. Their results showed that there was a high consistency in both sequence variant and CNV detection, supporting ctDNA as an alternative to bone marrow. Ruan et al. [67] detected mutations in ctDNA, bone marrow and peripheral blood mononuclear cells (PBMCs) of pediatric AML patients by NGS. The correlation of clinically significant mutations between ctDNA and bone marrow was 77%. This demonstrated that ctDNA can reflect the genomic information from bone marrow and is a reliable sampling method in pediatric AML. The other study also demonstrated that mutations were missed by both cfDNA and bone marrow analysis, particularly when the variant allele frequencies (VAF) were below 10%, suggesting that cfDNA and bone marrow may be complementary in the assessment and monitoring of patients with AML [68].

All of the above data demonstrate a high correlation between bone marrow and ctDNA results at the genetic level. At the epigenetic level, four CpG regions were identified in a study aimed at establishing a method to estimate MRD based on AML-associated DNA methylation (DNAm). The combinations of four CpG regions could reliably discriminate between healthy and newly diagnosed AML samples [69].

GRAIL used a targeted methylation- and cfDNA-based technology to develop a custom classification model for hematological malignancies and evaluated its performance for cancer detection and TOO prediction accuracy. They reported that the hematologic-specific classifier achieved the sensitivity of 45.8% and TOO prediction accuracy of 100% for myeloid neoplasms [70].

In summary, ctDNA enrichment can be found in peripheral blood plasma. Many experiments have shown that at the level of genetic and epigenetic analysis, peripheral blood ctDNA and bone marrow cells have a high degree of concordance. The use of ctDNA for genetic and epigenetic analysis is highly reliable. Importantly, ctDNA can be used as a tool to assist in the diagnosis of MDS or AML.

MRD refers to the small number of cancer cells that do not respond to treatment or are resistant to drugs and therefore remain in the body after cancer treatment. Although the amount of remaining cancer cells may be small, causing no signs or symptoms or even undetectable by conventional methods, they contribute to cancer recurrence. This suggests that for MRD-positive patients, ongoing monitoring for recurrence and early intervention may play a positive role in improving outcomes.

The MRD level of patients with MDS and AML is widely determined by various methods, including multiparameter flow cytometry (MFC), real-time quantitative polymerase chain reaction (qPCR) and, more recently, digital PCR and NGS. In 2010, Gao et al. [71] analyzed the integrity of cfDNA from 60 acute leukemia patients and 30 healthy controls. They used qPCR to amplify the $\beta$-actin gene and showed that DNA concentrations and DNA integrity were significantly higher in acute leukemia patients than in healthy controls. DNA integrity decreased significantly during complete response (CR) and increased during relapse. The data suggest that plasma DNA integrity may be a potential biomarker for monitoring leukemia progression and MRD. This study laid the foundation for ctDNA to monitor MRD in leukemia. Since then, the ultimate goal of researchers has been to use ctDNA-detected genetic mutations for diagnosis, treatment guidance, MRD monitoring and prognosis prediction. However, in Gao’s research [71], the number of samples, particularly those used to infer changes in DNA integrity from diagnosis to relapse, was too small to demonstrate sufficient accuracy. More work is needed.

Yeh et al. [72] collected serial bone marrow and plasma samples from 12 patients with MDS undergoing demethylation therapy in a phase 1 clinical trial. By analyzing serial bone marrow and matched plasma samples using digital PCR (dPCR), they demonstrated that ctDNA is comparable to bone marrow biopsy in the genomic heterogeneity of malignant clones and dynamically reflects tumour burden. CtDNA can be used to track both mutations and karyotypic abnormalities during therapy in MDS. The above conclusions not only prove the feasibility of using ctDNA to reflect the dynamic changes of genetic abnormalities in bone marrow, but also provide strong evidence for the use of ctDNA to assess efficacy.

Monitoring the evolution and type of mutations in ctDNA during and after treatment can be used to effectively monitor treatment response and assess the risk of relapse. Tong et al. [73] conducted a study of 31 patients with MDS and AML. All patients underwent bone marrow NGS and ctDNA testing at baseline. In 27 patients with ctDNA mutations at baseline who received demethylation therapy or chemotherapy, patients who achieved a complete response/complete hematological recovery (CR/CRi) at the end of treatment had a lower mean number of mutant VAF than those who did not achieve CR/CRi. Patients who achieved CR had a greater decrease in mutant VAF. The study also found that post-treatment ctDNA was negatively associated with longer progression-free survival (PFS) and overall survival (OS). Of the 10 patients who progressed from MDS to AML, seven subsequently acquired subclones containing FLT3 or NF1 mutations. In addition, new subcloning was detected in four patients 0.5–4 months prior to AML transformation, suggesting that genetic changes preceded morphological change.

For patients with high-risk and refractory MDS and AML, allo-HSCT is the only treatment. Methods to reduce relapse rates after transplantation are an urgent and important topic in clinical research. Researchers at the University of Tokyo conducted a study to identify patients at high risk of relapse in AML and MDS following allo-HSCT. Their results showed that the cumulative relapse rates of ctDNA-positive patients at 1 month and 3 months after transplantation were 56.6% and 63.0% higher than those of ctDNA-negative patients, and the 3-year OS decreased by 45.9% and 39.1% respectively [74]. This study demonstrated that driver mutation persistence in ctDNA could serve as a prognostic biomarker for MRD monitoring in AML/MDS patients undergoing first allo-HSCT.

In addition to gene-based MRD monitoring, ctDNA methylation-based MRD monitoring is also emerging. Iriyama et al. [75] used bisulfite pyrosequencing to perform a global methylation analysis based on the specific CpG sites of the LINE-1 promoter. They found that the methylation level of peripheral blood ctDNA decreased rapidly after azacitidine treatment. However, the study mentioned above, which proposed a method to estimate MRD based on AML-associated DNA methylation, showed poor accuracy in separating MRD-positive from MRD-negative samples by identifying four CpG regions [69]. This may be because the four AML-associated methylation regions selected in this study were from a small sample and were not highly representative of abnormal methylation regions. Increasing the sample size and the range of methylation detection regions, combined with genetic mutations, will make it possible to detect MRD-positive samples earlier and more accurately.

Although ctDNA cannot replace the gold standard of bone marrow sampling, it can be used as a complementary diagnostic tool for those MDS/AML patients without obvious clinical symptoms or who are unable to undergo traditional diagnostic procedures. It also has potential in the assessment of prognosis and monitoring of MRD.