Skip to main content
NCBI home page
Cancer Science logoLink to Cancer Science
. 2026 Apr 28;117(7):1921–1934. doi: 10.1111/cas.70405

An Immune Dysfunction Signature Score Predicts Survival in MDS Patients: Insights From a Longitudinal, Multicenter Study

Yu‐Hung Wang 1,2,3, Carmelo Gurnari 4,5, Valeria Visconte 4, Arda Durmaz 4, Luca Guarnera 4, Wan‐Hsuan Lee 1,6,7, Chi‐Yuan Yao 1,6,8, Andres Jerez 9, Kristian Gurashi 2, Kiran Batta 2, Hsin‐An Hou 1, Wen‐Chien Chou 1,8, Mrinal S Patnaik 10, Jaroslaw Maciejewski 4, Daniel H Wiseman 2, Chien‐Chin Lin 1,8,11,✉, Hwei‐Fang Tien 1,12
PMCID: PMC13327062  PMID: 42046464

ABSTRACT

Myelodysplastic syndrome (MDS) and chronic myelomonocytic leukemia (CMML) are clonal myeloid neoplasms shaped by genetic lesions and immune dysregulation, both of which contribute to disease progression and poor outcomes. Existing prognostic systems, such as IPSS‐R and IPSS‐M, do not incorporate immune alterations. We assessed the biological and clinical relevance of an immune dysfunction signature (IDS) across multi‐center MDS and CMML cohorts. IDS scores, derived from bulk transcriptomic data, were significantly associated with inferior leukemia‐free and overall survival. In multivariable analyses, IDS retained independent prognostic value alongside IPSS‐R, IPSS‐M, and bi‐allelic TP53 inactivation. Incorporation of IDS into existing models improved prognostic discrimination and time‐dependent predictive accuracy. Longitudinal analyses revealed that rising IDS scores paralleled disease progression and acute transformation, whereas declining scores were observed in remission. Biologically, IDS‐high cases demonstrated reduced cytotoxic T‐cell activity, expansion of regulatory T cells, enrichment of primitive progenitor signals, and increased expression of checkpoint pathway genes. These findings were validated across multiple independent MDS cohorts and consistently reproduced in CMML, where IDS also stratified risk and tracked disease evolution. Finally, integration with drug response signatures from the BeatAMLv2 cohort suggested potential therapeutic vulnerability of IDS‐high cases to multikinase and NF‐κB pathway inhibitors. These results establish IDS as a robust and dynamic biomarker in MDS and CMML, with applications in refined risk stratification, longitudinal disease monitoring, and guiding personalized therapeutic strategies targeting immune dysfunction.

Trial Registration: #201709072RINC, #202109078RINB, #202207050RINB

Keywords: immune dysfunction signature, MDS, multicenter study, prognostication

An immune dysfunction score (IDS) provides independent prognostic information in MDS and CMML, complements current risk models, mirrors disease changes, and reflects underlying immune impairment. IDS may also highlight therapeutic sensitivities, supporting its use as a practical biomarker for risk stratification and clinical follow‐up.

graphic file with name CAS-117-1921-g006.jpg

Abbreviations

AML

acute myeloid leukemia

BM

bone marrow

CMML

chronic myelomonocytic leukemia

HMA

hypomethylating agents

IDS

immune dysfunction signature

IPSS‐M

Molecular International Prognostic Scoring System

IPSS‐R

Revised International Prognostic Scoring System

LFS

leukemia‐free survival

MDS

myelodysplastic syndrome

MNC

mononuclear cell

OS

overall survival

pDCs

plasmacytoid dendritic cells

ssGSEA

single‐sample Gene Set Enrichment Analysis

Tregs

regulatory T cells

1. Introduction

Clonal genetic alterations in hematopoietic cells have traditionally been regarded as central drivers in the development of myeloid malignancies, including acute myeloid leukemia (AML), myelodysplastic syndrome (MDS), and chronic myelomonocytic leukemia (CMML). However, recent findings highlight the complex interplay between malignant cells and the immune system, revealing immune dysfunction as a critical factor that fosters tumor survival and progression [1, 2, 3]. Throughout disease progression, substantial disruptions in immune cell populations and function can occur, including T‐cell dysfunction, impaired antigen presentation, increased regulatory T cells (Tregs), TP53‐mediated immunosuppressive hematopoietic stem cells, and enhanced inflammatory signaling (e.g., elevated IL‐17) within the bone marrow (BM) microenvironment [1]. These changes weaken immune surveillance and promote the clonal expansion of malignant hematopoietic cells, further advancing the disease.

While AML and MDS are often perceived as a spectrum of the same disease, their immune microenvironments exhibit shared immune dysregulation characteristics alongside distinct features. Lower‐risk MDS is typically associated with a more proinflammatory immune response that facilitates disease progression, whereas higher‐risk MDS and AML are characterized by a predominantly immunosuppressive milieu, including a notable expansion of immunosuppressive cell populations. Recent studies identified specific immune dysfunction signatures (IDS) that characterize immune exhaustion and senescence, particularly in CD8+ cytotoxic T cells [4, 5]. These signatures, derived from transcriptomic analyses, provide valuable insights into how the BM microenvironment fosters immune dysfunction and predicts patient outcomes. Notably, Rutella et al. [5] demonstrated that T‐cell senescence in AML is associated with poor overall survival (OS) and resistance to treatments such as immune checkpoint blockade and standard chemotherapy. The study revealed gene expression profiles linked to T‐cell exhaustion and senescence, which were strong predictors of adverse clinical outcomes, with IDSAML outperforming traditional prognostic markers, such as the European LeukemiaNet risk score.

While IDS has been studied in AML, its role in MDS remains unexplored. To address this gap, we investigated the significance of IDS in MDS and CMML to enhance our understanding of immune responses and provide a more precise approach to stratifying patients based on their risk of disease progression (Graphical Abstract).

2. Materials and Methods

2.1. Study Cohorts

The Taiwanese discovery cohort comprised 432 consecutive MDS patients from National Taiwan University Hospital with clinical, molecular, outcome, and diagnostic BM mononuclear cell (MNC) RNA‐seq data; 41 had paired follow‐up samples, including 27 treated with hypomethylating agents [HMA] at the time of follow‐up sample acquisition. Validation included 33 MDS patients from Cleveland Clinic [6] and reanalysis of public cohorts GSE97064 [7], GSE15061 [8], and GSE114922 [9]. For additional validation in CMML, RNA‐seq data were incorporated from Mayo Clinic (n = 27, BM MNC) [10], Christie NHS Foundation Trust (n = 33, CD34+) [11], Hospital Morales Meseguer (n = 25, BM MNC) [12], and GSE248596 [11]. A subset of Christie patients (n = 9) also had paired samples at acute transformation before treatment [13].

2.2. Genomic Studies

Cytogenetic analyses followed the International System for Human Cytogenetic Nomenclature. Next‐generation sequencing (targeted panels and RNA‐seq) was performed on BM samples from discovery and validation cohorts, as previously described [6, 10, 11, 12, 13, 14]. Methodological details are provided in the Methods S1.

2.3. The IDS Gene Set

The IDS gene set (Figure 1A) was derived from Desai et al. [4] They curated 50 immune inhibitory markers from the literature and showed their expression was broadly conserved across cell types, with higher levels in the immune‐exhausted cluster. Using scRNA‐seq datasets from van Galen et al. [15] and Petti et al. [16], they observed similar expression patterns.

FIGURE 1.

FIGURE 1

Open in a new tab

Immune dysfunction geneset and the biological relevance of immune dysfunction score (IDS), survival outcomes and Gene Set Enrichment Analysis (GSEA) in the discovery cohort of 432 myelodysplastic syndromes (MDS) patients. (A) List immune dysfunction signature gene set. (B) Bar chart of top enriched pathways from the GOBiologicalProcess2023 gene set library. The top 10 enriched pathways for the input immune dysfunction gene set are displayed based on the −log10(p‐value), with the actual p‐value shown next to each pathway term. The pathway at the top has the most significant overlap with the input query gene set. (C, D) Overall survival stratified by the Revised International Prognostic Scoring System (IPSS‐R) (C) and the Molecular International Prognostic Scoring System (IPSS‐M) (D). (E, F) Violin plots illustrating the distribution of IDS, calculated by single‐sample GSEA, across different IPSS‐R (E) and IPSS‐M (F) risk groups. (G) Volcano plot illustrating the differentially expressed genes between the IDShigh and IDSlow groups. Genes with adjusted p‐values ≤ 0.05 and |log2 fold change| ≥ 1 are considered significant. Genes from the immune dysfunction gene set are highlighted and labeled. (H) Representative GSEA plots highlighting pathway enrichment in MDS patients with low versus high IDS, where the cutoff was defined using the maximally selected rank statistics. FDR, false discovery rate; NES, normalized enrichment score.

2.4. Bioinformatic Analyses and Statistical Analyses

Single‐sample Gene Set Enrichment Analysis (ssGSEA, GSVA R package) was used to aggregate gene expression into sample‐level scores. IDS and drug response signatures were computed by ssGSEA, with drug signatures derived from the BeatAMLv2 cohort [17]. Pathway enrichment was performed with GSEA (v4.3.2) using ranked t‐statistics from differential expression. Cell‐type enrichment for 22 hematopoietic lineages (Table S1) was estimated with xCell [18].

2.5. Statistical Analyses

Patient stratification was performed using maximally selected rank statistics independently in each cohort. Categorical variables were compared by chi‐squared or Fisher's exact tests, continuous variables by Mann–Whitney or Kruskal–Wallis tests. Survival was analyzed by Kaplan–Meier with log‐rank testing and multivariable Cox models. Analyses were performed in GraphPad Prism (v10.0.2) and R (v4.3.1).

3. Results

3.1. Confirmation of the Biological Relevance of the IDS Genes

As the IDS gene set was originally derived and validated from single‐cell datasets, we first sought to confirm its biological relevance within our bulk RNA‐sequencing data (Figure 1A). Functional annotation using Enrichr, with the GOBiologicalProcess2023 gene set library, demonstrated significant enrichment in pathways governing the negative regulation of T‐cell activation and proliferation, immune response–inhibiting receptor signaling, and suppression of leukocyte and natural killer cell–mediated cytotoxicity (Figure 1B). These results accord with the proposed role of the IDS genes in mediating immune dysregulation in myeloid malignancies and are consistent with previous reports [4].

We next calculated IDS scores, using ssGSEA, for the 432 patients with MDS in the discovery cohort. Baseline characteristics are presented in Table S2; the median age was 67 years, with a predominance of male patients, consistent with the recognized demographic profile of MDS. After a median follow‐up of 22 months (range, 1–229 months), the median leukemia‐free survival (LFS) and overall survival (OS) were 34.0 and 42.4 months, respectively. Overall survival was appropriately stratified by both the Revised International Prognostic Scoring System (IPSS‐R; Figure 1C) and the Molecular International Prognostic Scoring System (IPSS‐M; Figure 1D). The distribution of IDS scores according to IPSS‐R and IPSS‐M categories is shown in Figure 1E,F.

Patients were further stratified into the IDShigh and IDSlow groups according to a cut‐off value derived from maximally selected rank statistics. Differential gene expression analysis identified 1151 upregulated and 34 downregulated genes in the IDShigh groups compared to the IDSlow group (Figure 1G; adjusted p‐value ≤ 0.05, |log2 fold change| ≥ 1). GSEA demonstrated that IDShigh BM cells exhibited downregulation of immune‐dependent pathways, including immune‐related pathways, such as inflammatory responses, T‐cell receptor signaling, T‐cell activation, antigen processing and presentation, and regulation of T‐cell‐mediated cytotoxicity (Figure 1H). These findings suggest a potential association between higher IDS scores and an immunosuppressed BM microenvironment.

3.2. Clinical, Molecular, and Prognostic Correlates of Higher IDS in MDS

A comparison of the clinical and laboratory features of the IDShigh and IDSlow groups is summarized in Table S2. Patients in the IDShigh group presented with a significantly higher proportion of BM blasts at diagnosis compared with the IDSlow group. Furthermore, notable differences were observed in the distribution of disease subtypes according to the fifth edition of the WHO classification and the IPSS‐M. The IDShigh group demonstrated a greater prevalence of very‐high‐risk MDS by IPSS‐M criteria, alongside a lower prevalence of low‐risk disease. No significant differences were detected in IPSS‐R–based risk stratification. At the molecular level, IDShigh status was strongly associated with the presence of ASXL1, SRSF2, and STAG2 mutations (p = 0.002, p = 0.002, and p = 0.005, respectively; Table S3).

In terms of prognosis, patients with higher IDS scores exhibited significantly inferior LFS and OS compared to those with lower IDS scores (16.5 vs. 61.8 months, p < 0.001; 22.3 vs. 73.3 months, p < 0.001; Figure 2A). Although IPSS‐R and IPSS‐M are widely adopted prognostic frameworks for MDS, neither incorporates gene expression signatures reflective of immune status within the BM microenvironment. We therefore assessed whether the IDS could provide prognostic discrimination within the IPSS‐R lower‐risk (very low [VL], low, and intermediate) and higher‐risk (high and very high [VH]) categories, as well as the IPSS‐M lower‐risk (VL, low, and moderate low [ML]) and higher‐risk (moderate high [MH], high, and VH) categories. The IDS retained significant prognostic utility for both LFS and OS across all subgroups, although its discriminatory capacity was modestly attenuated in the IPSS‐M lower‐risk cohort (Figure 2B,C).

FIGURE 2.

FIGURE 2

Open in a new tab

Survival outcomes of 432 myelodysplastic syndromes patients in the discovery cohort stratified by lower versus higher immune dysfunction score (IDS). Acute myeloid leukemia‐free survival (LFS) and overall survival (OS) are stratified by lower vs. higher IDS in the entire cohort (A), IPSS‐R subgroups (B), IPSS‐M subgroups (C), patients who received hypomethylating agent treatment or not (D), and patients who underwent hematopoietic stem cell transplantation (HSCT) or not (E).

The adverse prognostic impact of a higher IDS on both LFS and OS was also apparent across subgroups defined by treatment history. This effect was observed in patients who had received hypomethylating agents (the standard of care for MDS) as well as in those who had not (Figure 2D). Likewise, higher IDS was associated with inferior survival in patients who underwent allogeneic hematopoietic stem cell transplantation, the only potentially curative modality for MDS, and in those who did not (Figure 2E).

In multivariable analyses for both LFS and OS, a higher IDS emerged as an independent adverse prognostic factor after adjustment for age, IPSS‐R or IPSS‐M risk categories, and bi‐allelic TP53 inactivation. Specifically, higher IDS was associated with an increased hazard ratio (HR) for LFS (HR 1.6, 95% CI 1.2–2.1, p = 0.001) and OS (HR 1.6, 95% CI 1.2–2.1, p = 0.002) when adjusted for age, IPSS‐R, and bi‐allelic TP53 inactivation (Table S4); similar associations were observed when adjusting for IPSS‐M, albeit with slight attenuation. These findings prompted evaluation of whether IDS could augment existing prognostic frameworks. In Cox regression models for OS, the relative weight of higher IDS was comparable to that of IPSS‐R (IDS/IPSS‐R ratio 0.98) and substantial in relation to IPSS‐M (IDS/IPSS‐M ratio 0.7; Table S5). Accordingly, patients with higher IDS were upstaged from their original categories in newly constructed IPSS‐R/IDS and IPSS‐M/IDS systems, yielding six‐tier (VL, low, ML, MH, high, VH; Figure 3A) and seven‐tier (VL, low, ML, MH, high, VH, extremely high; Figure 3B) classifications, respectively. Both models effectively stratified LFS and OS (Figure 3C,D; Table S6). Incorporation of IDS improved time‐dependent AUC values (Figure 3E) and yielded higher C‐indices than the parent models (Figure 3F).

FIGURE 3.

FIGURE 3

Open in a new tab

Incorporating immune dysfunction score (IDS) enhances existing prognostic systems. (A, B) Sankey diagrams illustrating the re‐distribution of MDS patients from existing prognostic systems—IPSS‐R (A) and IPSS‐M (B)—to the integrated models, IPSS‐R/IDS and IPSS‐M/IDS. (C, D) Acute myeloid leukemia‐free survival (LFS) and overall survival (OS) are stratified by IPSS‐R incorporating IDS (C) and IPSS‐M incorporating IDS (D). (E) Time‐dependent ROC curve analysis demonstrates that adding IDS improves the performance of current prognostic systems. (F) C‐index comparison of the original IPSS‐R and IPSS‐M models with the new models incorporating IDS for predicting LFS and OS. EH, extremely high; MH, moderate high; ML, moderate low; VH, very high; VL, very low.

3.3. Validation of the Prognostic Implications of Higher IDS in Multiple MDS and CMML Cohorts

Across the MDS validation cohorts (Tables S7–S9), higher IDS was consistently associated with adverse disease features. In GSE97064 (Table S8) and GSE114922 (Table S10), higher IDS correlated with increased BM blast percentages, while in GSE97064 and GSE15061 (Tables S8 and S9) it was linked to higher‐risk MDS categories. By contrast, in the Cleveland Clinic cohort, no significant differences were observed between the IDShigh and IDSlow groups with respect to IPSS‐R or IPSS‐M distributions. Notwithstanding these inter‐cohort differences, the adverse prognostic implications of higher IDS were highly reproducible. In the Cleveland Clinic cohort, higher IDS was associated with significantly shorter LFS (27 vs. 94 months, p = 0.02) and OS (32 vs. 94 months, p = 0.016; Figure 4A). Similarly, higher IDS consistently predicted inferior prognosis across the three public MDS datasets (Figure 4B–D).

FIGURE 4.

FIGURE 4

Open in a new tab

Validation of the prognostic significance of immune dysfunction score (IDS) in myelodysplastic syndromes (MDS) and chronic myelomonocytic leukemia (CMML) cohorts. (A–C) Acute myeloid leukemia‐free survival (LFS) and overall survival (OS) in MDS stratified by lower versus higher IDS in validation cohorts from Cleveland Clinic (A), GSE97064 (B), and GSE15061 (C). (D) OS of MDS patients stratified by lower versus higher IDS in GSE114922 (LFS data unavailable in this cohort). (E–H) LFS and OS in CMML patients stratified by lower versus higher IDS in validation cohorts from Mayo Clinic (E), The Christie (F), Morales Meseguer Hospital (G), and GSE248596 (H). IDS cutoffs were determined using maximally selected rank statistics. (I) C‐index comparison of the original prognostication models with the new models incorporating IDS for predicting LFS and OS in MDS and CMML validation cohorts. CPSS‐Mol, CMML‐specific prognostic scoring system molecular; GFM, Groupe Francais des Myelodysplasies model; MMM, Mayo Molecular Model; NA, data not available.

In addition to the larger MDS cohorts, our analysis extended to smaller CMML cohorts from the Mayo Clinic, Christie NHS Foundation Trust, Morales Meseguer Hospital, and GSE248596 (Tables S11–S13). No significant differences in clinical features were observed between IDShigh and IDSlow groups across these cohorts, with the exception of lower hemoglobin levels at diagnosis in the IDShigh group within the Christie cohort. Despite the limited sample sizes, the IDS consistently exhibited strong discriminatory power for both LFS and OS (Figure 4E–H). Finally, applying the same integrative approach to combine IDS with established prognostic models for MDS (IPSS‐R and IPSS‐M) and CMML, including the Mayo Molecular Model [19], Groupe Francophone des Myélodysplasies model [20], and the molecular CMML‐specific prognostic scoring system [21], across multiple cohorts with available data, the enhanced models consistently demonstrated superior C‐indices compared to their parental counterparts, mirroring findings from the discovery cohort (Figure 4I). Collectively, these results highlight the biological and prognostic significance of immune dysregulation in both MDS and CMML, further supporting the IDS as a robust biomarker across these myeloid neoplasms.

3.4. Prognostic Role of IDS During Disease Follow‐Up

Among 41 MDS patients with paired RNA‐seq data, follow‐up samples were obtained during complete or partial remission in 6 patients, stable disease (SD) in 24, and disease progression in 11 (Table S14). We calculated IDS scores at follow‐up and compared them with those at diagnosis. Notably, patients with differing disease statuses at follow‐up exhibited distinct IDS dynamics. Specifically, those experiencing disease progression demonstrated an increasing trend in IDS, whereas patients in complete or partial remission predominantly showed decreasing IDS values (Figure 5A,B). These observations suggest that dynamic changes in IDS reflect evolving immune alterations within the BM throughout the clinical course. We further investigated whether integrating IDS measurements at diagnosis and follow‐up could enhance prognostic assessment. By summing IDS scores from both time points and stratifying patients into lower and higher IDSDiagnosis+Follow‐up groups, we found that patients with higher combined IDS exhibited significantly poorer outcomes in terms of LFS and OS from the time of follow‐up (Figure 5C).

FIGURE 5.

FIGURE 5

Open in a new tab

Correlation between dynamic changes of immune dysfunction score (IDS) and survival outcomes from follow‐up in myelodysplastic syndromes (MDS) and chronic myelomonocytic leukemia (CMML) patients with paired diagnostic and follow‐up RNA‐seq data. (A, B) Box plots displaying dynamic changes in IDS for MDS patients from the discovery cohort in complete remission (CR), partial remission (PR), stable disease (SD), and disease progression (PD) at follow‐up. *p ≤ 0.05. p values were computed using the Mann–Whitney test. (C) Acute myeloid leukemia‐free survival (LFS) and overall survival (OS) of 41 MDS patients stratified by the sum of IDS at diagnosis and follow‐up, divided into lower versus higher groups based on an IDS cutoff defined by maximally selected rank statistics. (D) Box plot displaying IDS changes in 9 CMML patients from the Christie cohort with acute transformation. (E) OS of 9 CMML patients stratified by IDS sum at diagnosis and follow‐up, categorized into lower versus higher groups using a maximally selected rank statistics cutoff.

In CMML, paired RNA‐seq data at the time of acute transformation were available for nine patients from the Christie cohort. All but one patient exhibited an increase in IDS, mirroring the pattern observed in MDS (Figure 5D). Similarly, integrating IDS scores at diagnosis and follow‐up demonstrated prognostic value, with a trend toward poorer outcomes in patients with higher combined IDSDiagnosis+Follow‐up scores; however, these prognostic observations should be interpreted with caution due to the very limited sample size. These findings suggest that IDS dynamics may have potential not only for monitoring disease progression and prognosis but also as an indicator of changes in immune status within the BM during follow‐up. However, validation in larger cohorts is required to substantiate these observations.

3.5. Biological Implication of Higher IDS

Accumulating evidence highlights the distinct roles of immune dysregulation affecting the broader immune cell landscape beyond the MDS and CMML clones themselves [22, 23]. To further explore this, we employed xCell to estimate and compare the signals of 22 distinct cell types (Table S1) in BM MNC RNA‐seq data from our MDS cohort. The IDShigh group exhibited consistently distinct profiles relative to the IDSlow group, with significantly stronger signals from primitive cell populations, including hematopoietic stem cells, multipotent progenitors, common myeloid progenitors, and megakaryocytic‐erythroid progenitors (Figure 6A, Table S15). Contrastingly, the IDShigh group showed notably diminished signals from key immune effector populations such as natural killer cells, CD8+ T cells, and classical dendritic cells (Figure 6A, Table S15). These observations reveal marked differences in the bone marrow cellular milieu between groups, indicative of a relationship between impaired immune competence and disease progression in MDS patients.

FIGURE 6.

FIGURE 6

Open in a new tab

Bone marrow (BM) cell signatures deconvoluted by xCell in patients with lower versus higher immune dysfunction score (IDS) and different disease status at follow up, and correlation between drug response signatures and IDS. (A) Representative box plots showing BM primitive and immune cell signatures in patients with lower versus higher IDS. (B) Representative box plots illustrating dynamic changes in BM cell signatures in patients with complete remission (CR), partial remission (PR), stable disease (SD), and disease progression (PD) at follow‐up. CMP, common myeloid progenitors; DC, dendritic cells; GMP, granulocyte macrophage progenitors; HSC, hematopoietic stem cells; MEP, megakaryocytic‐erythroid progenitors; NK, natural killer cells. ****p ≤ 0.0001, ***p ≤ 0.001, **p ≤ 0.01, *p ≤ 0.05. p values were computed using Mann–Whitney test. (C) Heatmap showing the correlation between drug response scores and immune dysfunction score across the discovery and validation cohorts. Drug response signature were calculated using single‐sample GSEA with gene sets derived from the BeatAMLv2 study.

Linking xCell‐derived cell type signals to clinical disease status at follow‐up, we identified a clear pattern: disease progression was accompanied by an expansion of primitive progenitor populations (e.g., hematopoietic stem cells and granulocyte–macrophage progenitors) alongside a reduction in immune‐relevant populations (e.g., natural killer cells and CD8+ T cells) (Figure 6B, Table S16). This evolving cellular landscape reflects dynamic alterations in the bone marrow microenvironment over the disease course. Our findings concur with previous studies employing alternative methodologies [24, 25, 26], which similarly report increased primitive progenitors and diminished immune responses, potentially reflecting disease advancement and immune evasion.

Finally, to investigate whether the IDS might provide insights into potential therapeutic avenues, we examined a drug response gene set derived from the BeatAMLv2 cohort [17]. By integrating these data, we assessed each patient's predicted sensitivity to a range of therapeutic compounds. This analysis aimed to explore potential correlations between IDS and drug response profiles, supporting hypothesis generation for personalized treatment approaches tailored to molecular features of individual patients. Correlative analysis identified 11 compounds whose response scores were consistently and significantly associated with IDS across all MDS and CMML cohorts (presented as the top 11 in Figure 6C). These included multikinase inhibitors targeting receptors such as IGF‐1R, insulin receptor, EGFR, VEGFR, PDGFR, and RAF, alongside inhibitors of NF‐κB signaling and MDM2. While direct evidence for the immunomodulatory effects of these agents in MDS and CMML remains limited, the association between IDS and drug response scores provides a rationale for further investigation.

4. Discussion

The altered immune landscape in MDS is defined by reduced cytotoxic T‐cell activity and expansion of immunosuppressive Tregs, facilitating immune escape and clonal expansion [1, 27, 28, 29, 30, 31]. Prognostic systems such as IPSS‐R and IPSS‐M do not account for immune‐related gene expression, limiting their capacity to capture this biology. Here we describe a novel immune dysfunction score (IDS), derived from inhibitory genes identified by single‐cell transcriptomics, which effectively distinguishes immune states in BM samples and stratifies MDS prognosis. Integrating IDS with established models improved predictive accuracy, indicating that immune dysfunction represents an additional, independent axis of risk.

While these findings highlight the potential clinical utility of IDS, it is also important to consider factors that may influence its interpretation in longitudinal analyses. As most follow‐up samples in the longitudinal MDS cohort were obtained after HMA therapy, IDS dynamics may reflect a combination of treatment‐related immunomodulatory effects and underlying disease biology. HMAs modulate several immune pathways, including T‐cell exhaustion [32, 33], antigen presentation [34, 35], and interferon signaling [36, 37, 38, 39], any of which may influence IDS values independently of disease trajectory. Even so, the observed IDS patterns, rising scores in patients with progression and falling scores in those achieving remission, are consistent with established immunological features accompanying the disease course in MDS. Disentangling treatment effects from intrinsic disease biology will require future studies incorporating treatment‐naïve follow‐up samples and comparative analyses across different therapeutic exposures.

Dynamic changes in IDS, particularly increases preceding disease progression, further support its potential as a longitudinal marker of evolving immune alterations. IDS was also prognostically relevant in CMML, consistent with shared immune dysregulation across myeloid neoplasms [3, 28, 40, 41, 42]. In CMML, the marrow microenvironment is shaped by clonal plasmacytoid dendritic cells (pDCs) [40, 42, 43, 44]. CD123+ pDC aggregates, enriched in RAS‐mutant CMML, contribute to immunosuppression by expanding Tregs and expressing checkpoints such as IDO. Elevated kynurenine levels, a marker of IDO activity, reflect systemic immune tolerance with altered T/NK‐cell balance [40], paralleling findings in MDS. Together with PD‐L1 upregulation, these mechanisms create an immune‐evasive milieu that sustains clonal survival and favors leukemic transformation [28, 45].

Identification of IDS in MDS also has profound implications for future therapeutic strategies. T‐cell dysfunction and exhaustion, key contributors to immune evasion in MDS, present opportunities for immune‐based treatments such as checkpoint inhibitors and adoptive T‐cell therapies, a field that has yet to achieve the transformative success seen in solid tumors [46, 47, 48]. Recent high‐profile clinical setbacks involving agents like pevonedistat, magrolimab, and sabatolimab (which targets TIM‐3 to restore T‐cell function) [49, 50, 51, 52, 53], highlight the pressing need for more effective disease‐modifying therapies in higher‐risk MDS. Our exploratory analysis leveraging BeatAMLv2 drug sensitivity data identified associations between IDS and predicted response to multikinase inhibitors (IGF‐1R, insulin receptor, VEGFR, PDGFR), MDM2 inhibitors, and NF‐κB pathway agents [54, 55], providing insights that may guide future research into targeted treatment strategies.

This study has limitations. IDS was primarily derived and validated from single‐cell datasets, and although strongly associated with immune inhibition, functional validation is needed. In the paired transcriptomes, most MDS samples had prior HMA exposure, which may influence results [28, 35], while CMML paired samples were limited to AML transformation and small in number. Lack of in vitro validation also restricts translational insight.

In summary, this study establishes IDS as a prognostically and biologically relevant metric in MDS and CMML, with potential utility in refining risk stratification and guiding personalized therapeutic approaches. The score should be evaluated in additional large MDS cohorts in future studies, and prospective evaluation, particularly in trials of immunomodulatory agents, will be essential to define its role in optimizing treatment selection and improving patient outcomes.

Author Contributions

Yu‐Hung Wang: conceptualization, methodology, software, data curation, investigation, validation, formal analysis, visualization, project administration, resources, writing – original draft, writing – review and editing. Carmelo Gurnari: data curation, validation, investigation, writing – review and editing. Valeria Visconte: data curation, validation, investigation. Arda Durmaz: data curation, investigation, validation. Luca Guarnera: data curation, investigation, validation. Wan‐Hsuan Lee: data curation. Chi‐Yuan Yao: data curation. Andres Jerez: data curation. Kristian Gurashi: methodology, data curation, software. Kiran Batta: data curation, supervision. Hsin‐An Hou: data curation, investigation, supervision. Wen‐Chien Chou: data curation, investigation, supervision. Mrinal S. Patnaik: investigation, supervision, data curation, writing – review and editing. Jaroslaw Maciejewski: data curation, supervision. Daniel H. Wiseman: data curation, investigation, validation, funding acquisition, writing – review and editing. Chien‐Chin Lin: conceptualization, methodology, data curation, supervision, visualization, project administration, resources, writing – review and editing. Hwei‐Fang Tien: data curation, investigation, validation, funding acquisition, project administration, resources, writing – review and editing, supervision.

Funding

This work was supported by grants from the National Science and Technology Council (NSTC 114‐2314‐B‐002‐167‐MY3), and the Taiwan Ministry of Health and Welfare (project number: MOHW112‐TDU‐B‐222‐124001). The Epigenetics of the Hematopoiesis Laboratory was funded by grants from The Oglesby Charitable Trust.

Ethics Statement

This retrospective study was approved by the National Taiwan University Hospital Research Ethics Committee and the institutional review boards of each participating hospital. Informed consent was obtained from all patients in accordance with the Declaration of Helsinki. All methods were performed in accordance with relevant guidelines and regulations.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: cas70405‐sup‐0001‐Supplementary Methods S1‐S5.pdf.

CAS-117-1921-s002.pdf (156.2KB, pdf)

Tables S1–S16: cas70405‐sup‐0002‐Tables S1‐S16.xlsx.

CAS-117-1921-s001.xlsx (273.6KB, xlsx)

Acknowledgments

We acknowledge the services provided by the Department of Laboratory Medicine, Department of Medical Research, Division of Hematology, and Department of Internal Medicine, National Taiwan University Hospital, Taiwan. Y.‐H.W., K.G., K.B., and D.H.W. were supported by grants from the Oglesby Charitable Trust. C.G. was supported by a grant from the Edward P. Evans Foundation.

Data Availability Statement

The data reported analyzed in this article can be accessed through the Gene Expression Omnibus (DOI: https://www.ncbi.nlm.nih.gov/geo/), including GSE97064, GSE15061, GSE114922, GSE248596, GSE223305, and GSE264117 or through reasonable requests from the corresponding author. All analyses were conducted using pre‐existing R packages and functions; no novel code was created for this study.

References

  • 1. Rodriguez‐Sevilla J. J. and Colla S., “T‐Cell Dysfunctions in Myelodysplastic Syndromes,” Blood 143, no. 14 (2024): 1329–1343, 10.1182/blood.2023023166. [DOI] [PubMed] [Google Scholar]
  • 2. Balandrán J. C., Lasry A., and Aifantis I., “The Role of Inflammation in the Initiation and Progression of Myeloid Neoplasms,” Blood Cancer Discovery 4, no. 4 (2023): 254–266, 10.1158/2643-3230.Bcd-22-0176. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3. Mangaonkar A. A. and Patnaik M. M., “Role of the Bone Marrow Immune Microenvironment in Chronic Myelomonocytic Leukemia Pathogenesis: Novel Mechanisms and Insights Into Clonal Propagation,” Leukemia & Lymphoma 63, no. 8 (2022): 1792–1800, 10.1080/10428194.2022.2056175. [DOI] [PubMed] [Google Scholar]
  • 4. Desai P. N., Wang B., Fonseca A., et al., “Single‐Cell Profiling of CD8+ T Cells in Acute Myeloid Leukemia Reveals a Continuous Spectrum of Differentiation and Clonal Hyperexpansion,” Cancer Immunology Research 11 (2023): Of1–Of18, 10.1158/2326-6066.Cir-22-0961. [DOI] [PubMed] [Google Scholar]
  • 5. Rutella S., Vadakekolathu J., Mazziotta F., et al., “Immune Dysfunction Signatures Predict Outcomes and Define Checkpoint Blockade‐Unresponsive Microenvironments in Acute Myeloid Leukemia,” Journal of Clinical Investigation 132, no. 21 (2022): e159579, 10.1172/jci159579. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Awada H., Durmaz A., Kewan T., et al., “Context‐Dependent Role of Trisomy 6 in Myelodysplastic Neoplasms and Acute Myeloid Leukemia: A Multi‐Omics Analysis,” Leukemia 38, no. 6 (2024): 1411–1414, 10.1038/s41375-024-02268-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Wang Y.‐H., Lin C.‐C., Yao C.‐Y., et al., “A 4‐Gene Leukemic Stem Cell Score Can Independently Predict the Prognosis of Myelodysplastic Syndrome Patients,” Blood Advances 4, no. 4 (2020): 644–654, 10.1182/bloodadvances.2019001185. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Mills K. I., Kohlmann A., Williams P. M., et al., “Microarray‐Based Classifiers and Prognosis Models Identify Subgroups With Distinct Clinical Outcomes and High Risk of AML Transformation of Myelodysplastic Syndrome,” Blood 114, no. 5 (2009): 1063–1072, 10.1182/blood-2008-10-187203. [DOI] [PubMed] [Google Scholar]
  • 9. Pellagatti A., Armstrong R. N., Steeples V., et al., “Impact of Spliceosome Mutations on RNA Splicing in Myelodysplasia: Dysregulated Genes/Pathways and Clinical Associations,” Blood 132, no. 12 (2018): 1225–1240, 10.1182/blood-2018-04-843771. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10. Carr R. M., Vorobyev D., Lasho T., et al., “RAS Mutations Drive Proliferative Chronic Myelomonocytic Leukemia via a KMT2A‐PLK1 Axis,” Nature Communications 12, no. 1 (2021): 2901, 10.1038/s41467-021-23186-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11. Wang Y.‐H., Lin C.‐C., Gurashi K., et al., “Prognostic and Therapeutic Implications of TP53 Expression in Chronic Myelomonocytic Leukemia,” Blood Cancer Journal 14, no. 1 (2024): 112, 10.1038/s41408-024-01087-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12. Hurtado A. M., Luengo‐Gil G., Chen‐Liang T. H., et al., “Transcriptomic Rationale for Synthetic Lethality‐Targeting ERCC1 and CDKN1A in Chronic Myelomonocytic Leukaemia,” British Journal of Haematology 182, no. 3 (2018): 373–383, 10.1111/bjh.15408. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13. Gurashi K., Wang Y.‐H., Somervaille T. C. P., Batta K., and Wiseman D. H., “Integrative Analysis of Blast Phase Chronic Myelomonocytic Leukemia Reveals Distinct Transformation Phenotypes Predictive of Drug Response,” Blood 140, no. Supplement 1 (2022): 213–214, 10.1182/blood-2022-168831. [DOI] [Google Scholar]
  • 14. Wang Y.‐H., Hou H.‐A., Lin C.‐C., et al., “A CIBERSORTx‐Based Immune Cell Scoring System Could Independently Predict the Prognosis of Patients With Myelodysplastic Syndromes,” Blood Advances 5, no. 22 (2021): 4535–4548, 10.1182/bloodadvances.2021005141. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15. van Galen P., Hovestadt V., Wadsworth Ii M. H., et al., “Single‐Cell RNA‐Seq Reveals AML Hierarchies Relevant to Disease Progression and Immunity,” Cell 176, no. 6 (2019): 1265–1281.e24, 10.1016/j.cell.2019.01.031. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Petti A. A., Khan S. M., Xu Z., et al., “Genetic and Transcriptional Contributions to Relapse in Normal Karyotype Acute Myeloid Leukemia,” Blood Cancer Discovery 3, no. 1 (2022): 32–49, 10.1158/2643-3230.Bcd-21-0050. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Bottomly D., Long N., Schultz A. R., et al., “Integrative Analysis of Drug Response and Clinical Outcome in Acute Myeloid Leukemia,” Cancer Cell 40, no. 8 (2022): 850–864.e9, 10.1016/j.ccell.2022.07.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Aran D., Hu Z., and Butte A. J., “xCell: Digitally Portraying the Tissue Cellular Heterogeneity Landscape,” Genome Biology 18, no. 1 (2017): 220, 10.1186/s13059-017-1349-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19. Patnaik M. M., Itzykson R., Lasho T. L., et al., “ASXL1 and SETBP1 Mutations and Their Prognostic Contribution in Chronic Myelomonocytic Leukemia: A Two‐Center Study of 466 Patients,” Leukemia 28, no. 11 (2014): 2206–2212, 10.1038/leu.2014.125. [DOI] [PubMed] [Google Scholar]
  • 20. Itzykson R., Kosmider O., Renneville A., et al., “Prognostic Score Including Gene Mutations in Chronic Myelomonocytic Leukemia,” Journal of Clinical Oncology 31, no. 19 (2013): 2428–2436, 10.1200/jco.2012.47.3314. [DOI] [PubMed] [Google Scholar]
  • 21. Elena C., Gallì A., Such E., et al., “Integrating Clinical Features and Genetic Lesions in the Risk Assessment of Patients With Chronic Myelomonocytic Leukemia,” Blood 128, no. 10 (2016): 1408–1417, 10.1182/blood-2016-05-714030. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22. Tentori C. A., Zhao L. P., Tinterri B., et al., “Immune‐Monitoring of Myelodysplastic Neoplasms: Recommendations From the i4MDS Consortium,” Hema 8, no. 5 (2024): e64, 10.1002/hem3.64. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23. Trowbridge J. J. and Starczynowski D. T., “Innate Immune Pathways and Inflammation in Hematopoietic Aging, Clonal Hematopoiesis, and MDS,” Journal of Experimental Medicine 218, no. 7 (2021): e20201544, 10.1084/jem.20201544. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24. Sallman D. A., McLemore A. F., Aldrich A. L., et al., “TP53 Mutations in Myelodysplastic Syndromes and Secondary AML Confer an Immunosuppressive Phenotype,” Blood 136, no. 24 (2020): 2812–2823, 10.1182/blood.2020006158. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25. Lamble A. J., Kosaka Y., Laderas T., et al., “Reversible Suppression of T Cell Function in the Bone Marrow Microenvironment of Acute Myeloid Leukemia,” Proceedings of the National Academy of Sciences of the United States of America 117, no. 25 (2020): 14331–14341, 10.1073/pnas.1916206117. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26. Josephine D. K., Martine E. D. C., Theresa M. W., et al., “Regulatory T Cells and Progenitor B Cells Are Independent Prognostic Predictors in Lower Risk Myelodysplastic Syndromes,” Haematologica 100, no. 6 (2015): e220–e222, 10.3324/haematol.2014.116657. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Zheng Z., Qianqiao Z., Qi H., Feng X., Chunkang C., and Xiao L., “In Vitro Deprivation of CD8(+)CD57(+)T Cells Promotes the Malignant Growth of Bone Marrow Colony Cells in Patients With Lower‐Risk Myelodysplastic Syndrome,” Experimental Hematology 38, no. 8 (2010): 677–684, 10.1016/j.exphem.2010.04.002. [DOI] [PubMed] [Google Scholar]
  • 28. Yang H., Bueso‐Ramos C., DiNardo C., et al., “Expression of PD‐L1, PD‐L2, PD‐1 and CTLA4 in Myelodysplastic Syndromes Is Enhanced by Treatment With Hypomethylating Agents,” Leukemia 28, no. 6 (2014): 1280–1288, 10.1038/leu.2013.355. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29. Sand K., Theorell J., Bruserud Ø., Bryceson Y. T., and Kittang A. O., “Reduced Potency of Cytotoxic T Lymphocytes From Patients With High‐Risk Myelodysplastic Syndromes,” Cancer Immunology, Immunotherapy: CII 65, no. 9 (2016): 1135–1147, 10.1007/s00262-016-1865-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30. Kordasti S. Y., Ingram W., Hayden J., et al., “CD4+CD25high Foxp3+ Regulatory T Cells in Myelodysplastic Syndrome (MDS),” Blood 110, no. 3 (2007): 847–850, 10.1182/blood-2007-01-067546. [DOI] [PubMed] [Google Scholar]
  • 31. Giovazzino A., Leone S., Rubino V., et al., “Reduced Regulatory T Cells (Treg) in Bone Marrow Preferentially Associate With the Expansion of Cytotoxic T Lymphocytes in Low Risk MDS Patients,” British Journal of Haematology 185, no. 2 (2019): 357–360, 10.1111/bjh.15496. [DOI] [PubMed] [Google Scholar]
  • 32. Hu Y., Zhang Y., Shi F., et al., “Reversal of T‐Cell Exhaustion: Mechanisms and Synergistic Approaches,” International Immunopharmacology 138 (2024): 112571, 10.1016/j.intimp.2024.112571. [DOI] [PubMed] [Google Scholar]
  • 33. Loo Yau H., Bell E., Ettayebi I., et al., “DNA Hypomethylating Agents Increase Activation and Cytolytic Activity of CD8+ T Cells,” Molecular Cell 81, no. 7 (2021): 1469–1483.e8, 10.1016/j.molcel.2021.01.038. [DOI] [PubMed] [Google Scholar]
  • 34. Sorrentino V. G., Thota S., Gonzalez E. A., Rameshwar P., Chang V. T., and Etchegaray J.‐P., “Hypomethylating Chemotherapeutic Agents as Therapy for Myelodysplastic Syndromes and Prevention of Acute Myeloid Leukemia,” Pharmaceuticals 14, no. 7 (2021): 641. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35. Wong K. K., Hassan R., and Yaacob N. S., “Hypomethylating Agents and Immunotherapy: Therapeutic Synergism in Acute Myeloid Leukemia and Myelodysplastic Syndromes,” Frontiers in Oncology 11 (2021): 624742, 10.3389/fonc.2021.624742. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. Wolff F., Leisch M., Greil R., Risch A., and Pleyer L., “The Double‐Edged Sword of (Re)expression of Genes by Hypomethylating Agents: From Viral Mimicry to Exploitation as Priming Agents for Targeted Immune Checkpoint Modulation,” Cell Communication and Signaling 15, no. 1 (2017): 13, 10.1186/s12964-017-0168-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37. Chiappinelli Katherine B., Strissel Pamela L., Desrichard A., et al., “Inhibiting DNA Methylation Causes an Interferon Response in Cancer via dsRNA Including Endogenous Retroviruses,” Cell 162, no. 5 (2015): 974–986, 10.1016/j.cell.2015.07.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38. Upadhyay P., Beales J., Shah N. M., et al., “Recurrent Transcriptional Responses in AML and MDS Patients Treated With Decitabine,” Experimental Hematology 111 (2022): 50–65, 10.1016/j.exphem.2022.04.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39. Lindblad K. E., Goswami M., Hourigan C. S., and Oetjen K. A., “Immunological Effects of Hypomethylating Agents,” Expert Review of Hematology 10, no. 8 (2017): 745–752, 10.1080/17474086.2017.1346470. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40. Mangaonkar A. A., Reichard K. K., Binder M., et al., “Bone Marrow Dendritic Cell Aggregates Associate With Systemic Immune Dysregulation in Chronic Myelomonocytic Leukemia,” Blood Advances 4, no. 21 (2020): 5425–5430, 10.1182/bloodadvances.2020002415. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41. You X., Liu F., Binder M., et al., “Asxl1 Loss Cooperates With Oncogenic Nras in Mice to Reprogram the Immune Microenvironment and Drive Leukemic Transformation,” Blood 139, no. 7 (2022): 1066–1079, 10.1182/blood.2021012519. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42. Lucas N., Duchmann M., Rameau P., et al., “Biology and Prognostic Impact of Clonal Plasmacytoid Dendritic Cells in Chronic Myelomonocytic Leukemia,” Leukemia 33, no. 10 (2019): 2466–2480, 10.1038/s41375-019-0447-3. [DOI] [PubMed] [Google Scholar]
  • 43. Mellor A. L. and Munn D. H., “IDO Expression by Dendritic Cells: Tolerance and Tryptophan Catabolism,” Nature Reviews Immunology 4, no. 10 (2004): 762–774, 10.1038/nri1457. [DOI] [PubMed] [Google Scholar]
  • 44. Munn D. H., Sharma M. D., Lee J. R., et al., “Potential Regulatory Function of Human Dendritic Cells Expressing Indoleamine 2,3‐Dioxygenase,” Science 297, no. 5588 (2002): 1867–1870, 10.1126/science.1073514. [DOI] [PubMed] [Google Scholar]
  • 45. Kang T. G., Lan X., Mi T., et al., “Epigenetic Regulators of Clonal Hematopoiesis Control CD8 T Cell Stemness During Immunotherapy,” Science 386, no. 6718 (2024): eadl4492, 10.1126/science.adl4492. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46. Garcia‐Manero G., Ribrag V., Zhang Y., Farooqui M., Marinello P., and Smith B. D., “Pembrolizumab for Myelodysplastic Syndromes After Failure of Hypomethylating Agents in the Phase 1b KEYNOTE‐013 Study,” Leukemia & Lymphoma 63, no. 7 (2022): 1660–1668, 10.1080/10428194.2022.2034155. [DOI] [PubMed] [Google Scholar]
  • 47. Chien K. S., Kim K., Nogueras‐Gonzalez G. M., et al., “Phase II Study of Azacitidine With Pembrolizumab in Patients With Intermediate‐1 or Higher‐Risk Myelodysplastic Syndrome,” British Journal of Haematology 195, no. 3 (2021): 378–387, 10.1111/bjh.17689. [DOI] [PubMed] [Google Scholar]
  • 48. Abaza Y. and Zeidan A. M., “Immune Checkpoint Inhibition in Acute Myeloid Leukemia and Myelodysplastic Syndromes,” Cells 11, no. 14 (2022): 2249, 10.3390/cells11142249. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49. Adès L., Girshova L., Doronin V. A., et al., “Pevonedistat Plus Azacitidine vs Azacitidine Alone in Higher‐Risk MDS/Chronic Myelomonocytic Leukemia or Low‐Blast‐Percentage AML,” Blood Advances 6, no. 17 (2022): 5132–5145, 10.1182/bloodadvances.2022007334. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50. Koss R. and Koss A., “Gilead to Discontinue Phase 3 ENHANCE Study of Magrolimab Plus Azacitidine in Higher‐Risk MDS,” Gilead To Discontinue Phase, 2023.
  • 51. Brunner A. M., Esteve J., Porkka K., et al., “Phase Ib Study of Sabatolimab (MBG453), a Novel Immunotherapy Targeting TIM‐3 Antibody, in Combination With Decitabine or Azacitidine in High‐ or Very High‐Risk Myelodysplastic Syndromes,” American Journal of Hematology 99, no. 2 (2024): E32–e36, 10.1002/ajh.27161. [DOI] [PubMed] [Google Scholar]
  • 52. Zeidan A. M., Giagounidis A., Sekeres M. A., et al., “STIMULUS‐MDS2 Design and Rationale: A Phase III Trial With the Anti‐TIM‐3 Sabatolimab (MBG453) + Azacitidine in Higher Risk MDS and CMML‐2,” Future Oncology (London, England) 19, no. 9 (2023): 631–642, 10.2217/fon-2022-1237. [DOI] [PubMed] [Google Scholar]
  • 53. Amer M., Zeidan Z. X., Sanz G., et al., “Primary Results of the Phase Iii Stimulus‐MDS2 Study of SABATOLIMAB + AZACITIDINE vs Placebo + AZACITIDINE as Frontline Therapy for Patients With Higher‐Risk MDS or CMML‐2,” HemaSphere 8, no. S1 (2024): e104, 10.1002/hem3.104. [DOI] [Google Scholar]
  • 54. Pellagatti A. and Boultwood J., “Hyperactivation of NF‐κB Signaling in Splicing Factor Mutant Myelodysplastic Syndromes and Therapeutic Approaches,” Advances in Biological Regulation 95 (2024): 101055, 10.1016/j.jbior.2024.101055. [DOI] [PubMed] [Google Scholar]
  • 55. Montalban‐Bravo G., Thongon N., Rodriguez‐Sevilla J. J., et al., “Targeting MCL1‐Driven Anti‐Apoptotic Pathways Overcomes Blast Progression After Hypomethylating Agent Failure in Chronic Myelomonocytic Leukemia,” Cell Reports Medicine 5, no. 6 (2024): 101585, 10.1016/j.xcrm.2024.101585. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Data S1: cas70405‐sup‐0001‐Supplementary Methods S1‐S5.pdf.

CAS-117-1921-s002.pdf (156.2KB, pdf)

Tables S1–S16: cas70405‐sup‐0002‐Tables S1‐S16.xlsx.

CAS-117-1921-s001.xlsx (273.6KB, xlsx)

Data Availability Statement

The data reported analyzed in this article can be accessed through the Gene Expression Omnibus (DOI: https://www.ncbi.nlm.nih.gov/geo/), including GSE97064, GSE15061, GSE114922, GSE248596, GSE223305, and GSE264117 or through reasonable requests from the corresponding author. All analyses were conducted using pre‐existing R packages and functions; no novel code was created for this study.

Articles from Cancer Science are provided here courtesy of Wiley

RESOURCES