Flow cytometric analysis of erythroid precursors and mutational signatures of lower risk myelodysplastic syndromes identify responders to erythroid stimulating agents

Anemia in low-risk myelodysplastic neoplasms (LR-MDS) derives from defective maturation and apoptosis of erythroid precursors [1] and its first-line (1L) treatment relies on erythropoiesis stimulating agents (ESAs) in case of serum EPO < 200 U/L. Still, 30–40% of patients selected with a high probability of response are resistant [2]. The knowledge of MDS mutational landscape has paved the way for molecular classification (WHO 2022 [3], ICC [4]) and IPSS-M prognostic risk stratification [5], but their role in ESA response has not been completely elucidated.

The availability of novel agents like Luspatercept, which targets late erythroid precursors, and Imetelstat has in fact prompted this need of precision in order to avoid ineffective and costly treatments [6, 7].

Since ESAs act on early erythroid precursors [8], we exploited multiparametric flow cytometry (MFC) to analyze baseline erythroid subpopulations [9] from ESA treated LR-MDS patients. Results were correlated with ESA response and validated in an external cohort.

Composition of erythroid subpopulations and ESA response were also correlated with both MDS classification and IPSS-M risk categories. We believe that integrating MFC and molecular information at baseline would improve the decision-making for 1L treatment in anemic LR-MDS.

Bone marrow (BM) cells from LR-MDS (n = 87 from the learning cohort and n = 54 from the validation cohort) were analyzed with MFC and targeted next generation sequencing (t-NGS, n = 83 and 44, respectively) analyzed presence of somatic mutations in 35 genes before ESA treatment. We evaluated by MFC the repartition of erythroid (ery) early precursors: ery-HPCs/CD34⁺cells, ery-CD117⁺/CD117⁺ and ery-CD117⁺/ery (see Suppl. for gating and definitions) and correlated with ESA response. In addition, age matched healthy BM controls (HC) were analyzed in both cohorts (n = 6 and 8, respectively).

Finally, patients were stratified according to WHO 2022 and IPSS-M score (see supplementary methods).

All patients received ESAs and after 12 weeks erythroid response was evaluated by IWG 2018 revised criteria [10], defining long term response (Long-R) as ≥24 months.

Median age was 75 years with male prevalence (M: F = 1.7) and only 8.7% of cases were transfusion dependent (TD) at baseline (median 2 red blood cell units/8 weeks). sEPO was <200 U/L in most patients from both cohorts (72.4 and 91%, Table 1), with mean serum EPO of 235 U/L and 78 U/L (p = 0.03) in the validation cohort.

Table 1 Baseline characteristics of LR-MDS patients in the two cohorts.

Overall response rate (ORR) in the learning cohort was 53.6% (median duration of response (DOR = 23 months) and among them, 55.7% were long-R. In the validation cohort, ORR was 54.7% (median DOR = 18 months) (27% were long-r).

In terms of erythroid precursors, ery-CD117⁺/117⁺ were significantly increased in long-R (median 23%) compared with non-responders (non-R) (19%, p = 0.019), and the same was observed for ery-CD117⁺/ery (6.4 vs 4.0% respectively, p = 0.018 (Fig. 1A). However, more mature erythroid precursors were not differently distributed between long-R vs non-R (Supplementary Fig. S1). Similar results were obtained in the validation cohort (Supplementary Fig. S2).

Fig. 1: Erythroid precursors composition and mutational pattern identify luspatercept responders.

A Repartition of CD34⁺ and CD117⁺ erythroid precursors in LR-MDS (n = 97) according to ESA response duration and compared with age-matched healthy BM controls (n = 6) from the learning cohort. Ery-HPC/CD34⁺ are CD34⁺ erythroid committed precursors among total CD34⁺ BM cells, erythroid CD117⁺ cells (ery-CD117⁺) are expressed as fraction of total CD117⁺ myeloid progenitors (ery-117⁺/CD117⁺) and total erythroid cells (ery-117⁺/ery). Long-r = long responders (>24 months), short-r = short responders (<24 months) and non-r = non-responders, ctrl = healthy controls. B Percentages of ery-HPC/CD34⁺, ery-CD117⁺/ CD117⁺ and total erythroid cells among total nucleated BM cells (ery/TOT.) across WHO 2022 categories (n = 83) and in age-matched healthy BM controls (n = 6) from the learning cohort. MDS-LB RS = MDS with low blast count and ring sideroblasts (presence of >15% RS without SF3B1 mutation and BM blasts < 5%). C Percentages of immature erythroid cells in MDS non-SF3B1 cases (n = 50) according to ESA response duration and compared with age-matched healthy BM controls (n = 6) from the learning cohort. D ROC analysis to identify thresholds of IPSS-M score (left) and CD117⁺ erythroid cells (right) discriminating ESA responders from non-responders. The graphics report the thresholds with the best combination of sensitivity and specificity (in brackets). AUC = area under the curve. Binomial multivariate analysis based on these thresholds is reported in the bottom part. OR = odd ratio, 95%CI = 95% confidence interval. E Genetic clusters for MDS non-SF3B1 (A, n = 50) and MDS-SF3B1 (B, n = 38) from the learning cohort. In both groups, clustering is based on the type of mutated gene (35 genes) and corresponding VAF expressed in log₂ scale within each type of ESA response (LR = long response, SR = short response, NR = non-response). In the upper part of the graphic, clinical information is also provided: sex (M = male, F = female), sEPO ranges (U/L), Transferrin saturation ranges (Tf, %), Ferritin ranges (ng/mL) and transfusion dependence (TD, yes/no). The legend on the right reports the threshold and colors used for each variable. *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001.

To capture MDS heterogeneity in terms of erythroid maturation, MFC analysis was evaluated across WHO subtypes (n = 83, Fig. 1 B). Total erythroid cells were decreased in MDS del(5)q (n = 12) compared to other MDS in both cohorts (Fig. 1B and Supplementary Fig. S3 for the validation cohort), confirming the association of erythroid hypoplasia in this MDS subgroup [11]. In MDS del(5)q, ery-CD117⁺/ery were significantly increased in responders compared to non-R (Supplementary Fig. S4).

Moreover, MDS-SF3B1 (n = 33) exhibited an expansion of CD34⁺ and CD117⁺ erythroid precursors compared to other MDS categories and HC group (Fig. 1B and Supplementary Fig. S3 for the validation cohort). However, despite the expansion of the immature erythroid compartment in MDS-SF3B1, both ery-HPC/CD34⁺ and ery-CD117⁺ were not increased in MDS-SF3B1 responders (Resp) vs non-R cases (Supplementary Fig.S5).

Given this finding, we focused on MDS non-SF3B1 per WHO 2022 classification (n = 50): not only CD117⁺ erythroid cells, but also ery-HPC/CD34⁺ cells were significantly increased in long-R vs non-R (median 22 vs 10,8% respectively, p = 0.003) (Fig. 1C). Similar results were obtained in the validation cohort (Supplementary Fig. S6). Finally, left-shifted erythropoiesis (higher ratio of early precursors/late precursors) was observed in MDS non-SF3B1 long-R (mean ratio =0.3) compared to non-R (mean =0.16, p = 0.018) (Supplementary Fig. S7). In terms of IPSS-M, long-R (n = 25) and short responders (short-R, n = 21) showed significantly lower scores (mean −1.12 and −0.64, respectively) compared with non-R (n = 37, mean −0.32, p = 0.004 and p = 0.009, respectively), while there was no difference in terms of IPSS-R score (Supplementary Fig. S8). Similar results were found in the validation cohort (long-R, n = 12, mean −1.14 vs non-R, n = 17, mean −0.46, p = 0.028).

As previously observed [5], a fraction of cases in both cohorts transitioned from lower IPSS-R categories to IPSS-M moderate high (12.2 and 6.8%), high (9.7 and 6.8%) and very high (2.4 and 2.3%) risk groups (Table 1 and Supplementary Fig. S9). These upstaged cases were either ESA non-R (78 and 57.2%) or short-R (22 and 42.8%). Multivariable analysis integrating percentages of immature erythroid populations, IPSS-M score and clinical variables (i.e., Hb, MCV, RDW, iron status, TD and sEPO) showed that only a lower IPSS-M score (OR 0.45, 95% CI 0.17–1.03, p = 0.075) and higher percentage of ery-CD117⁺ /Ery (OR 1.31, 95% CI 1.01–1.82, p = 0.071) positively correlated with ESA response in MDS non-SF3B1 (Supplementary Table S3).

Based on these results, IPSS-M score of 0.25 and percentage of 10.6% of CD117⁺ erythroid precursors showed good sensitivity (0.89) and specificity (0.96), respectively, in discriminating MDS non-SF3B1 Resp vs. non-R (Fig. 1D). This result is consistent with previous finding, showing positive correlation between increased fraction of CD117⁺ erythroid precursors and ESA response [12].

Finally, binomial logistic regression analysis confirmed our observations (IPSS-M < 0.25 O.R. 15, CI 95% 2.15–54, p = 0.018 and ery-CD117⁺/ery<10.6% O.R. 0.06, CI95% 0-1.17, p = 0.043) (Fig. 1D). By applying these cut-offs in both cohorts, cases with IPSS-M score <0.25 and Ery-CD117^+/ery >10.6% were all responders, mainly long-R (78 and 100%), whereas LR-MDS with IPSS-M score >0.25 and %Ery-CD117^+/ery <10.6 were either non-R (86 and 50%) or short-R (50 and 50%, Supplementary Fig. S10).

On the other hand, in MDS-SF3B1 only female sex (p = 0.04) and sEPO (p = 0.06) were associated with ESA response in univariable, but not multivariable analysis (Supplementary Table S4).

Finally, to investigate mutational profiles, hierarchical cluster analysis of mutated genes and corresponding VAF was performed according to ESA response and separating MDS-SF3B1 (n = 38) from MDS non-SF3B1 (n = 50) (Fig. 1E, F).

In MDS non-SF3B1 (Fig. 1E), TET2^WT cases (n = 17) were enriched in female responders (67 and 70% respectively) with low sEPO (90% <200 U/I). On the other hand, non-R cases (30%) were mainly males with sEPO >200 U/L (60%). TET2^MUT cases (n = 22) displayed a male preponderance (75%), high ORR (68,2%), low sEPO (<200 U/L in 90%) and absence of iron overload (90% with Tf <45%). TET2^MUT non-R cases (n = 7) were all males with sEPO >200 U/L (75%), iron overload (Tf >45% in 80%), high TD rate (40%) and harbored additional poor prognostic mutations (i.e. EZH2, ASXL1 or IDH2).

Finally, U2AF1^MUT and RUNX1-STAG2^MUT cases (n = 4 and 5 respectively) were all non-R cases, predominantly males (80 and 100% respectively), TD (50 and 60% respectively) and the latter also showed very high sEPO (>500 U/L in 60%).

Regarding MDS-SF3B1 cases, 63% of them harbored SF3B1^K700E variant of whom 54% were responders (Fig. 1F). SF3B1^K700E mutated non-R cases (n = 11) were mostly males (82%) with either DNMT3A or TET2 co-mutations.

On the other hand, SF3B1^non-K700E non-R cases were enriched in females with sEPO <200 U/L (80 and 100%, respectively) and with higher SF3B1^non-K700E VAF compared to responders (mean 43 vs 26%, p = 0.02). A comprehensive graphic and a table recapitulating these associations are presented in the supplementary file (Supplementary Fig. S11 and Table S5).

In conclusion, our results suggest that long-term ESA response in MDS non-SF3B1 is associated with the increase of BM CD34⁺ and CD117⁺ erythroid cells, indicating that erythropoiesis could be persistently sustained by ESAs when EPO sensitive erythroid precursors are expanded. Erythroid maturation was evaluated across distinct 2022 WHO categories, revealing a link between MDS-SF3B1 and immature erythroid cell expansion, though not correlating in this subtype with ESA response. The lack of this correlation might be explained by the effects of SF3B1 mutant causing abnormal immature erythroid proliferation [13], mitochondrial iron accumulation and replication stress during erythropoiesis [14, 15], hindering maturation despite precursor expansion.

Our study also reveals that IPSS-M score was higher in ESA non-R and that MDS non-SF3B1 non-R were more frequently mutated in poor prognostic genes (i.e., IDH2, EZH2, STAG2, U2AF1 and RUNX1), stressing the relevance of specific somatic mutations for erythroid response.

Taken together, we propose a model integrating IPSS-M scoring and fraction of immature CD117⁺ erythroid cells to predict response in ESA eligible, non-TD, and non-SF3B1 LR-MDS. The application of this model can practically be of straightforward support for the choice of 1L treatment in SF3B1^negative non-TD LR-MDS, i.e., ESA vs luspatercept, agents characterized by different target cell populations.