High BM plasma S100A8/A9 is associated with a perturbed microenvironment and poor prognosis in myelodysplastic syndromes

TO THE EDITOR:

Myelodysplastic syndromes (MDS) represent a heterogeneous group of malignant hematopoietic stem cell (HSC) disorders.^1-3 Recent advances in the immunome of the bone marrow (BM) microenvironment identified aberrant immune activation and proinflammatory signaling as vital drivers in MDS pathogenesis.^4-6 Among these complex networks, the S100A8/A9-toll–like receptor axis is a critical MDS phenotypes definer.⁷ 

The inflammatory proteins S100A8 and S100A9 often exist as a heterodimer under physiological conditions.⁸ In MDS, S100A8/A9 is synthesized and secreted by, among others, myeloid-derived suppressor cells (MDSCs), which play a central role in pathogenesis.⁹ The ligation of S100A8/A9 to TLR4 leads to NF-κB-mediated transcription and subsequent production of proinflammatory cytokines, and the induction of NLRP3 inflammasome,^5,10 which consequently drives pyroptosis of HSCs and an inflammatory milieu in the BM.^11,12 Although S100A8/A9 serves as biomarkers in various diseases,^13-15 its clinical implication in MDS is not fully deciphered.¹⁶ 

To investigate the clinical and microenvironmental relevance of S100A8/A9 in patients with MDS, we recruited 215 patients with MDS at the National Taiwan University Hospital and quantified BM plasma S100A8/A9 dimer levels by enzyme-linked immunosorbent assay (supplemental Method). Genomic DNA and mRNA were extracted from BM mononuclear cells and sequenced as previously described.^17,18 Methods for bioinformatic and statistical analysis are detailed in supplemental Methods. Patient characteristics are summarized in supplemental Table 1. The median age of the patients was 67.5 years. Over a median follow-up duration of 39.7 months, 64 patients (29.8%) progressed to acute myeloid leukemia (AML), and 93 patients succumbed to the disease. The National Taiwan University Hospital Research Ethics Committee approved the study (#201709072RINC). Informed consent was obtained in accordance with the Helsinki Declaration.

Correlation analysis revealed that the BM plasma S100A8/A9 protein levels significantly correlated with the mRNA expression in whole BM cell RNA sequencing (r² = 0.32 and r² = 0.31, respectively; supplemental Figure 1). We then explored the distribution of S100A8/A9 levels (supplemental Figure 2A) across disease subgroups. Patients with MDS/AML (MDS with 10%-19% blasts in the BM or peripheral blood) and concurrently mutated TP53 had higher S100A8/A9 levels than others (supplemental Figure 2B). Meanwhile, there was no difference in S100A8/A9 levels among patients in the International Prognostic Scoring System (IPSS) or the Revised IPSS (IPSS-R)³ subgroups (supplemental Figure 2C-F).

Because previous studies demonstrated the impact of genetic events on innate immune and inflammasome-signaling,⁵ such as the NF-kB pathway,^19,20 pyroptosis and β-catenin signaling,²¹ and NLRP3 inflammatory pathways,⁷ we examined whether S100A8/A9 concentrations differ across patients with various genotypes. Patients with ASXL1-mutation had significantly higher S100A8/A9 than those with unmutated ASXL1 (supplemental Figure 2G), whereas no difference was detected between patients with or without other epigenetic or splicing gene mutations (supplemental Figure 2G-H).

Next, we sought to investigate the relationship between mutant ASXL1 and S100A8/A9 levels. ASLX1 variant allele frequencies (VAF) significantly correlated with S100A8/A9 levels in 49 patients with ASXL1-mutation with available VAFs (Figure 1A). Hierarchical clustering suggested close associations among mutations in ASXL1, STAG2, RUNX1, EZH2, and ZRSR2 (Figure 1B). Interestingly, patients with ASXL1-mutation with concurrent abovementioned mutations (cluster 1) had a trend of higher S100A8/A9 than those without (cluster 2) (Figure 1C,D).

The 215 patients with MDS were subsequently divided into higher- and lower-S100A8/A9 groups with cutoff point of 7093 ng/mL determined by maximally selected rank statistics. There were no significant differences in the distribution of disease subgroups according to the International Classification Consensus (ICC), IPSS-R, and karyotypes between the 2 groups; but high-S100A8/A9 patients had a higher ASXL1 mutation rate (43.5% vs 21.4%) (supplemental Tables 1-3).

We then examined the impact of S100A8/A9 levels on patients’ survival. Higher-S100A8/A9 was associated with significantly inferior leukemia-free survival (LFS) and overall survival (OS) not only in the total cohort (Figure 1E), but also in the lower- and higher-risk subgroups based on the ICC and IPSS-R (Figure 1F-G; supplemental Figures 3-4). Time-dependent receiver operating characteristic curve analysis also suggested the potential for S100A8/A9 to supplement IPSS-R (Figure 1H). Moreover, despite having higher frequencies of mutant ASXL1, high-S100A8/A9 patients had poorer survival than those of the lower-S100A8/A9 group irrespective of ASXL1 mutation statuses (Figure 1I; supplemental Figure 5).

The prognostic implications of S100A8/A9 levels on survival were also demonstrated in patients carrying different karyotypes (supplemental Figure 6) or receiving different treatments (supplemental Figure 7). Remarkably, in 50 patients who received hypomethylating agent monotherapy, higher-S100A8/A9 retained strong discriminatory prognostic impact on LFS and OS (supplemental Figure 8). In multivariable analysis, we included parameters with a P value <.05 in the univariate analysis (supplemental Table 4), and hazard ratios were adjusted with treatments that well-stratified survivals (supplemental Figure 9). Higher-S100A8/A9 remained an independent adverse prognostic factor for LFS and OS (Figure 2A).

Considering that RNAseq was performed on whole BM MNCs, we referenced the single-cell data set of healthy controls²² to identify which cell types contributed to the differentially expressed genes. Curiously, 18 of the 20 most down-regulated genes (supplemental Table 5) in high-S100A8/A9 BM were regularly expressed by lymphocytes (supplemental Figure 10), implying different composition of lymphocytes in higher- and lower-S100A8/A9 BMs. We further adopted CIBERSORTx,^18,23 which infers the landscape of infiltrating immunocytes in the BM from gene expression profiles. Higher-S100A8/A9 was associated with significantly lower fractions of CD8 T-cells and activated natural killer (NK) cells (Figure 2B; supplemental Table 6).

Weighted gene coexpression network analysis revealed that the turquoise and blue modules were closely associated with lower- and higher-S100A8/A9, respectively (supplemental Figure 11). Corresponding to CIBERSORTx analysis, the 914 genes in the turquoise module were enriched in pathways involving NK- and T-cell functions, whereas the 509 genes in the blue module were enriched in pathways involving MDSCs in cancer immune escape and altered metabolism (Figures 2C,D).

To the best of our knowledge, this is the first study to significantly stratify the survival of patients with MDS based on S100A8/A9 levels. We also observed that patients with ASXL1-mutation had higher S100A8/A9 concentrations than their unmutated counterparts, corresponding with the increase in NADPH oxidase and ROS, TLR4 activation and pyroptosis.⁷ Fundamentally, the upregulation of S100A8/A9 can exert genotoxic stress in HSCs, thereby advancing the risk for AML transformation,²⁴ in accordance with our finding that higher-S100A8/A9 group had a shorter LFS. In a homogeneously treated lower-risk MDS cohort, S100A8/A9 expression in mesenchymal stem cells was correlated with p53 and TLR4 upregulation.²⁴ Furthermore, high S100A8/A9 concentrations doubled the risk of leukemic transformation and significantly reduced the time for AML transformation.

Comparisons of the transcriptomic data highlighted differences in functions and properties of immune cells between higher- and lower-S100A8/A9 BM and suggested a high-risk subentity that was not considered in current risk stratification in this heterogenous disease and required more attention. However, the lack of external validation, biological validation, and the assessment of the impact of other inflammasome components such as cytokines or chemokines is a major issue. Meanwhile, the demographics of our MDS population are more similar to Korean patients,²⁵ with a younger and higher-risk skewing when compared with that of western cohorts, suggesting that the extrapolation of our data may be compromised. Treatment heterogeneity could also potentially confound our analysis. Serial follow-up data after treatment will strengthen the prognostic power of S100A8/A9 and provide more insight into the changes in the BM microenvironment. In addition, our findings could be more granular and justified if cytometry by time of flight or single-cell multiomics approaches were adopted.

Despite above limitations, this study clearly showed that S100A8/A9 level was an independent poor prognostic factor in MDS, and higher S100A8/A9 in the BM intimated a perturbed microenvironment with enhanced MDSC signal and impairment in the functions and quantities of CD8⁺ T cells and NK cells. We propose that S100A8/A9 can be incorporated to the current risk stratification systems and prospectively assessed in clinical trials.

Acknowledgments: The authors acknowledge the service provided by Department of Laboratory Medicine, Department of Medical Research, and Division of Hematology, Department of Internal Medicine, National Taiwan University Hospital. The authors also appreciate Joanna Storer’s help with English editing. F.M.R.A. was supported by Cancer Research UK grant number C5759/A27412.

The work was supported by grants from Ministry of Science and Technology, Taiwan, project number: MOST 109-2314-B-002-221, 109-2314-B-002-222; and Taiwan Ministry of Health and Welfare, project number: MOHW109-TDU-B-211-134009.

Contribution: Y.-H.W. was responsible for data collection and management, statistical analysis, interpretation, visualization, literature research, and manuscript writing; F.M.R.A. and C.-Y.Y. assisted in bioinformatic analysis and visualization; C.-J.K, S.-C.Y., and P.-T.S. helped with sample preparation and processing; H.-A.H. and W.-C.C. were responsible for data collection and management; and C.-C.L. and H.-F.T. conceived and coordinated the study and revised the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Chien-Chin Lin, Department of Laboratory Medicine, National Taiwan University Hospital, No. 7, Chung-Shan S. Rd, Taipei City 10002, Taiwan; e-mail: lincc@ntu.edu.tw; and Hwei-Fang Tien, Department of Internal Medicine, National Taiwan University Hospital, No. 7, Chung-Shan S. Rd, Taipei City 10002, Taiwan; e-mail: hftien@ntu.edu.tw.