Inflammatory bone marrow signaling in pediatric acute myeloid leukemia distinguishes patients with poor outcomes

Bolouri, Hamid; Ries, Rhonda E.; Wiedeman, Alice E.; Hylkema, Tiffany; Scheiding, Sheila; Gersuk, Vivian H.; O’Brien, Kimberly; Nguyen, Quynh-Anh; Smith, Jenny L.; Alice Long, S.; Meshinchi, Soheil

doi:10.1038/s41467-022-34965-4

Download PDF

Article
Open access
Published: 23 November 2022

Inflammatory bone marrow signaling in pediatric acute myeloid leukemia distinguishes patients with poor outcomes

Nature Communications volume 13, Article number: 7186 (2022) Cite this article

2889 Accesses
2 Citations
17 Altmetric
Metrics details

Subjects

Abstract

High levels of the inflammatory cytokine IL-6 in the bone marrow are associated with poor outcomes in pediatric acute myeloid leukemia (pAML), but its etiology remains unknown. Using RNA-seq data from pre-treatment bone marrows of 1489 children with pAML, we show that > 20% of patients have concurrent IL-6, IL-1, IFNα/β, and TNFα signaling activity and poorer outcomes. Targeted sequencing of pre-treatment bone marrow samples from affected patients (n = 181) revealed 5 highly recurrent patterns of somatic mutation. Using differential expression analyses of the most common genomic subtypes (~60% of total), we identify high expression of multiple potential drivers of inflammation-related treatment resistance. Regardless of genomic subtype, we show that JAK1/2 inhibition reduces receptor-mediated inflammatory signaling by leukemic cells in-vitro. The large number of high-risk pAML genomic subtypes presents an obstacle to the development of mutation-specific therapies. Our findings suggest that therapies targeting inflammatory signaling may be effective across multiple genomic subtypes of pAML.

Feasibility of functional precision medicine for guiding treatment of relapsed or refractory pediatric cancers

Article Open access 11 April 2024

A single-cell atlas enables mapping of homeostatic cellular shifts in the adult human breast

Article Open access 28 March 2024

PERCEPTION predicts patient response and resistance to treatment using single-cell transcriptomics of their tumors

Article 18 April 2024

Introduction

In spite of its heavy toll on families and young lives, treatments for pediatric Acute Myeloid Leukemia (pAML) lag far behind other leukemias and the 10-year survival rate for children with pAML is <50%¹. This is due, in part, to the fact that pAML is a highly heterogeneous disease comprising many clinical, genomic, epigenomic, and transcriptional subtypes² (Supplementary Fig. 1). While genomic/precision medicine has delivered many successes over the past decade^3,4, development of targeted therapies for each of these rare, poor-prognosis pAML subtypes may be economically prohibitive. To address this challenge, we set out to identify dysregulated pathways that are shared across multiple pAML subtypes and mark poorer clinical outcomes. By spanning multiple disease subtypes, treatments targeting such pathways can overcome the cost-per patient barrier and deliver treatments to those most in need.

Inflammatory signaling is a hallmark of cancer^5,6. Accordingly, the pro-inflammatory IL‑6 signaling pathway is a well-established contributor to many cancers⁷. IL-6 synergizes with IL-3 to drive the proliferation of hematopoietic stem cells⁸, and increased bone marrow (BM) IL-6 levels have been shown to correlate with treatment resistance and poor outcomes in pediatric⁹ and adult AML^10,11 as well as in other blood malignancies^12,13. IL-6 signals through JAK1, JAK2, and STAT3¹⁴, and JAK2/STAT3 signaling is increased in the blood progenitor/stem cells of high-risk adult AML¹⁵. Consistent with a role for IL-6 signaling in treatment resistance, inhibition of JAK/STAT signaling sensitizes AML blasts to treatment with other drugs¹⁶, as well as reducing the proliferation rate of adult AML cells¹⁷. Moreover, the IL-6 neutralizing antibody Siltuximab increased the survival of mice engrafted with adult AML blasts¹⁸. Whether the IL-6 signaling pathway is activated in pAML leukemic cells, whether it is associated with specific subtypes of pAML, whether it has distinct roles in pediatric versus adult patients, and whether it is the key upstream driver of poor outcomes in pAML remain unknown.

To address these questions, we analyzed RNA-seq data from ~1500 pAML patients enrolled in 2 large-scale Children’s Oncology Group (COG) studies, and verified our findings through additional genomic, epigenomic, flow cytometry, and cytometry by time of flight (CyTOF) analyses on selected pAML samples. Here, we show that a subset of pAML patients with higher levels of IL-6 signaling in their pre-treatment bone marrow samples, have poorer 2-year event-free and overall survival, and higher inflammatory cytokine signaling via the IFNα/β, IL-1, and TNFα pathways. As such, combination therapies that counter these signal transduction pathways may help a relatively large (> 20%), genomically-diverse proportion of pAML patients avoid treatment resistance and poor outcomes.

Results

Higher IL-6 signaling occurs in > 20% of pAML patients and is associated with worse outcomes

Comparing bulk RNA-seq data from 1,275 pAML BM aspirates collected at diagnosis prior to treatment (COG study AAML1031, NCT01371981) to 45 age-matched healthy normal bone marrow (NBM) controls (Supplementary Fig. 2), we found that mRNA levels of IL-6 and its receptor subunit IL-6R are rarely higher in pAML than in NBM (Fig. 1a). However, IL-6 and IL-6R mRNA levels were bimodally distributed across the pAML cohort, suggesting 2 potentially distinct patient populations (Fig. 1b). Using Gaussian mixture modeling, we identified subsets of samples with higher expression of either IL-6 or IL-6R (red bars in Fig. 1b) and selected patients in the highest-expressing quartile in each higher-expression group as having unusually high levels of one or both genes (287 patients of whom 7 had high levels of both IL-6 and IL-6R, 22.5% of the cohort). We refer to this group as the “high-IL6/R” group. For comparison, we also selected a reference group of patients with below-median expression of both IL-6 and IL-6R across all patients in the cohort (“low-IL6/R” group, 306 patients, 24% of the cohort).

Fig. 1: Pediatric AML diagnostic samples with higher IL-6 or IL-6R expression have higher multi-cytokine signaling in the bone marrow (BM) and poorer 2-year outcomes than patients with low levels of bone-marrow IL-6 and IL-6R mRNA.

IL-6 is associated with chemotherapy and multi-drug resistance in diverse cancers¹⁹. Consistent with these findings, high-IL6/R pAML patients had significantly worse event free survival (EFS) and overall survival (OS) within 2 years of diagnosis (proportion of high-IL6/R patients alive at 2 years = 0.591. 95% Confidence Interval = (0.558, 0.677). The proportion of low-IL6/R patients alive at 2 years = 0.716. 95% Confidence Interval = (0.708, 0.808), Fig. 1c, 5-year plots in Supplementary Fig. 3). High- and low-IL6/R samples had similar percentages of blasts (means 62.0, 67.9 respectively, Fig. 1d), male-to-female ratios (1.1, 1.1 respectively), and AAML1031 Study Arm participation (Arm A 32% vs. 40%, Arm B 31% vs. 38%, Arm C 42% vs. 52% respectively, Fisher’s Exact Test P ~ 1), suggesting the observed differences are not confounded by these factors. However, high-IL6/R samples had significantly higher average expression levels of a proliferation-marker gene set (99 genes, Mann–Whitney U Test P < 1E−24, Supplementary Fig. 4).

To identify pathways and processes associated with higher levels of IL-6 and IL-6R in high- versus low-IL6/R samples, we carried out a series of gene set enrichment analyses contrasting these 2 patient groups (Fig. 1e and Supplementary Fig. 5). As expected, compared to low-IL6/R samples, high-IL6/R samples were highly enriched for an IL-6 signaling gene set (87 genes, Mann–Whitney U Test P < 1E−51). Unexpectedly, we found that, compared to low-IL6/R samples, high-IL6/R samples are also highly enriched for gene sets marking Type 1 interferon signaling (66 genes, Mann–Whitney U Test P < 1E−17), IL-1 signaling (149 genes, Mann–Whitney U Test P < 1E−37), and TNFα signaling (197 genes, Mann–Whitney U Test P < 3E−22).

Although IL-6, IFNα/β, IL-1, and TNFα signaling co-occur in approximately two thirds of all samples (Supplementary Fig. 6), we found that donor selection based on the expression levels of IL-6 and IL-6R was a better predictor of EFS and OS than sample selection using IFNα/β or IL-1 signaling gene sets (Supplementary Figs. 7–10), suggesting IL-6 signaling is the most likely driver of the observed poorer outcomes.

To verify the generality of our findings in an independent cohort, we used our previously published bulk BM RNA-seq data for 214 COG AAML0531 study (NCT00372593) participants². Applying the same selection criteria to this cohort as used for AAML1031, we identified 43 (20.1%) high-IL6/R and 59 (27.6%) low-IL6/R samples (Supplementary Fig. 11). Using the same gene sets as for AAML1031, we again found higher activity scores for IL-6, IFNα/β and IL-1 signaling (Mann–Whitney U Test P values <1E−12, <1E−5, and <1E−7 respectively, Fig. 2a–c. Supplementary Fig. 12). Five-year EFS and OS both exhibited a trend toward worse outcomes in high-IL6/R patients compared to low-IL6/R patients (Fig. 2d). Consistent with a role of for IL-6 in treatment resistance, two-year EFS and OS were significantly worse for high-IL6/R patients compared to low-IL6/R patients log-rank P = 0.028 and 0.014 respectively, with the proportion of high-IL6/R alive at 2 years = 0.512, 95% Confidence Interval = (0.392, 0.700), the proportion of low-IL6/R alive at 2 years = 0.738, 95% Confidence Interval = (0.633, 0.861)). Together, these findings suggest that higher levels of multi-cytokine inflammatory signaling in high- versus low-IL6/R pAML mark poorer 2-year outcomes in > 20% of pediatric AML.

**Fig. 2: Verification of findings in an independent cohort using the same gene sets and scoring as in the discovery cohort (Fig. 1, Supplementary Fig. 2B).**

High-IL6/R pAML samples have distinct immunological, transcriptomic, demographic, and genomic profiles

To better understand the etiology of multi-cytokine signaling in pAML, we sought to identify the key features of high-IL6/R pAML that set it apart from low-IL/R pAML. We started by asking whether the higher expression of inflammatory cytokine signaling in high- versus low-IL6/R pAML reflects broader immune function differences. Of 1793 immune genes downloaded from Immport.org 1506 are expressed in our RNA-seq data and collectively segregate high- and low-IL6/R pAML and NBM into distinct groups (Fig. 3a). Thus, there are genome-wide immune-state differences between high- vs. low-IL6/R pAML, and between these pAML subsets and NBM.

**Fig. 3: High-IL6/R pAML BM samples are highly distinct in terms of their immunological, transcriptional, and genomic characteristics.**

Using genome-wide differential expression analysis, we found transcriptional differences between high- and low-IL6/R pAML were not limited to the immune system. At a false discovery rate (FDR)-adjusted P value <0.05 and absolute fold-change > 0.25, out of a total of 51,573 measured transcripts, 15,609 (~30%) were differentially expressed between high-IL-6/R and low-IL6/R pAML, and 14,423 transcripts (~28%) were differentially expressed between high-IL6/R and NBM samples (Supplementary Data 1 and 2). Thus, transcriptional differences between high- and low-IL6/R pAML, and between high-IL6/R pAML and NBM, are pervasive and genome-wide.

To infer the key cancer and immune-associated genes underlying these differences, we used iterative feature selection from a collection of 4015 immune, epigenetic, and AML-related genes. We identified a set of 82 cancer and immune-related genes (Supplementary Data 3) that divide high- and low-IL6/R pAML samples into high-, medium-, and low-IL6/R clusters by unsupervised hierarchical clustering (clusters 1–3 respectively, Fig. 3B, Supplementary Fig. 13). More than 98% of Cluster1 samples are in the high-IL6/R group, while Cluster3 samples are highly concordant with our reference low-IL6/R samples (Supplementary Figs. 14–16). These transcriptionally-derived clusters revealed two surprising clinical characteristics of high-IL6/R pAML. Firstly, 48% of the donors in Cluster1 of Fig. 3b had clinically-identified chromosomal translocations impacting the MLL (KMT2A) gene, compared to only 7% in the low-IL6/R Cluster3. Consistent with this finding, high-IL6/R pAML BM samples are > 2-fold enriched for the French-American-British (FAB) M4 and M5 and WHO “AML, not otherwise categorized: Acute monoblastic/acute monocytic leukemia” and “AML, not otherwise categorized: Acute monoblastic/acute monocytic leukemia” classification categories (Fisher’s Exact Test P = 3.9E−20, odds ratio 3.76, 95% confidence interval: 2.8, 5.0, comparing high-IL6/R samples (n = 287) to all other AAML1031 donors (n = 1063)). In contrast, 43.8% of the low-IL6/R Cluster3 samples had no clinically-detected chromosomal abnormalities

High-IL6/R pAML patients were significantly younger than low-IL6/R pAML (median age at diagnosis = 8.4 and 11.7 years, respectively, Mann–Whitney U Test P value = 1.6E−05). 58.6% of the patients in Cluster1 are <10 years old and 41% are infants aged ≤ 3 years old. In contrast, only 16% of the low-IL6/R Cluster3 patients are infants (2-sided test of proportions P = 1.9E−08, Supplementary Fig. 17).

To establish whether high-IL6/R pAML samples have additional shared genomic aberrations, we analyzed targeted sequencing data (Supplementary Data 4, summarized in Fig. 3c) for the 181 patients in the high-IL6/R transcriptional cluster described above (Cluster1, Fig. 3b). Targeted sequencing data for a panel of 580 recurrent single nucleotide alterations and short insertion/deletions (indels), and 15 translocations previously identified by the NCI TARGET AML Project², revealed that 142 of 181 Cluster1 high-IL6/R samples (78.5%) had recurrent translocations. 91 samples (50.3%) had MLL (KMT2A) translocations, 55 (30.4%) had RAS point mutations, and 36 (19.9%) had FLT3 point mutations (Fisher’s Exact Test P values 9.6E−29, 0.004, and 0.004, respectively when compared to Cluster3). Furthermore, 77.8% of the FLT3 point mutations were demonstrated to be activating mutations by functional assays (Supplementary Data 4). Only 1 FLT3 internal tandem duplication (FLT3-ITD) co-occurred with an MLL-translocation. In contrast, among the Cluster1 samples without MLL-rearrangements, 21 (23.3%) had FLT3-ITD.

Consistent with the widespread occurrence of MLL gene-fusions in high-IL6/R samples, the mean expression of the MSigDB (gsea-msigdb.org/gsea/msigdb) “MLL-MLLT3 Fusion Target Genes” gene set was significantly higher in high-IL6/R pAML compared to low-IL6/R pAML (Mann–Whitney U Test P < 1E−44).

The patterns of co-occurrence and mutual-exclusion of the above recurrent somatic mutations divide the patients into distinct genomic groups. More than 93% of high-IL6/R pAML samples fall into one of 5 genomic subtypes (indicated by Group 1-5 color ribbons at the top of Fig. 3c). Patients in Groups 1 and 2 have chromosomal rearrangements resulting in the fusion of the MLL/KMT2A gene to one of 8 well-known MLL fusion partners. Group 1 patients additionally have RAS or FLT3 point mutations, both of which impact proliferation. The third high-IL6/R genomic group (indicated by the broken green bar at the top of Fig. 3c) is characterized by the co-occurrence RAS/FLT3 mutations with various pAML-associated chromosomal translocations, including chromosome 16 inversions, and RUNX1 and NUP98 fusions. Group 4 patients do not carry any recurrent large-scale chromosomal alterations, but have mutually-exclusive RAS or FLT3 point mutations, or FLT3-ITD. Finally, Group 5 marks 21 high-IL6/R patients (11.6%) with no recurrent somatic alterations in our targeted sequencing data.

Taken together, these findings suggest that high-IL6/R pAML, identified by a distinct transcriptional signature, is driven by specific and highly recurrent genomic alterations, exhibits broadly altered BM immune gene expression, and occurs primarily in younger children and infants.

Multi-cytokine signaling in high-IL6/R samples is receptor-mediated

We next asked whether the transcriptional signatures enriched in high- versus low-IL6/R pAML are driven by secreted cytokines or arise from intra-cellular gene regulatory interactions driven by genomic alterations. To address these questions, we performed perturbation experiments using high-IL6/R pAML samples.

To mimic in vivo conditions ex vivo, we co-cultured thawed and rested high-IL6/R pAML BM cells in transwells containing the HS-5 stromal cell line in the lower chamber. HS-5 cells replicate BM stromal cell expression patterns, including secretion of IL-1β, IL-6, KIT ligand, and the granulocyte/macrophage colony stimulating factors G-CSF, M-CSF, and GM-CSF^10,20,21. Cells were cultured with or without the JAK1/2 inhibitor Ruxolitinib (2μM), which has been shown to inhibit signaling by IL-6, IFNs, G/GM-CSF, and TNFα²². A 35-marker CyTOF panel (Supplementary Tables 1, 2), designed to measure eight proximal mediators of cytokine signaling in 23 BM cell types (see Supplementary Figs. 18–22), was used to assess proximal signaling 25 min post-stimulation. Additionally, we used bulk RNA-seq to verify downstream pathway activity 7 h post-stimulation.

CyTOF revealed that the leukemic cells of different high-IL6/R pAML genomic subtypes are derived from distinct hematopoietic developmental stages and have distinct signaling states (Fig. 4a, b). As detailed in Supplementary Figs. 23–26, in high-IL6/R samples with MLL gene rearrangements, leukemic cells resembled CD34- pre-monocytes. In contrast, in high-IL6/R samples with chromosome 16 inversions (Inv(16)) or a normal karyotype, the leukemic cells were less differentiated CD34+ common myeloid progenitors (CMP) and granulocyte macrophage progenitors (GMP). In spite of these differences, leukemic cells from all of these samples downregulated the levels of phosphorylated (i.e., activated) STAT1, STAT3, and ERK1/2 following treatment by Ruxolitinib (Fig. 4b, c, Supplementary Data 5). NF-κB, which may be active in unstimulated high-IL6/R samples (Fig. 4b) and is additionally activated by IL-1 and several other mechanisms²³, was not inhibited by Ruxolitinib (Fig. 4c).

Fig. 4: Phospho-CyTOF and RNA-seq suggest distinct cell-of-origin but common receptor-mediated signaling in high-IL6/R pAML genomic subtypes. All plots show data from 5 high-IL6/R pAML samples. Panel A additionally includes 5 NBM as indicated.

Consistent with our CyTOF findings, Gene Set Enrichment Analysis (GSEA) of bulk RNA-seq data from 5 MLL-rearranged high-IL6/R samples and 4 randomly-selected low-IL-6/R samples confirmed downregulation of IL-6, IFNα/β, and TNFα activity by Ruxolitinib in high-IL6/R samples (Fig. 4d). As expected, Ruxolitinib had a lower impact in low-IL6/R samples (Supplementary Fig. 27). Together, our RNA-seq and CyTOF data confirm that high-IL6/R pAML samples can transduce IL-6 (STAT3, ERK1/2), and IFNα/β (STAT1, ERK1/2) signaling, and that active JAK/STAT-mediated signaling is stronger in high- versus low-IL6/R pAML.

The genomic subtypes of high-IL6/R pAML upregulate distinct pro-leukemia genes

While high-IL6/R pAML samples are genomically diverse (Fig. 3c), ~60% of high-IL6/R pAML subjects have either an MLL-rearrangement with or without RAS/FLT3 point mutations, or an Inv(16) or NUP98 translocation with a RAS/FLT3 point mutation. To determine whether there are common factors driving the high-IL6/R phenotype, we focused on these genomic subtypes. We carried out a series of pairwise differential expression analyses (Supplementary Data 6) to identify genes that are (i) differentially expressed in a genomic subset of high-IL6/R pAML compared to low-IL6/R pAML and NBM, (ii) reported to be downstream of the associated genomic alterations, and (iii) reported to regulate the inflammatory signaling pathways co-activated in high-IL6/R pAML. Summary statistics are presented in Supplementary Data 6.

Compared to NBM and low-IL6/R pAML, high-IL6/R pAML samples carrying RAS/FLT3 point mutations and co-occurring Inv(16) or NUP98 gene fusions had significantly increased expression levels of the phosphatase DUSP6^24,25,26 (Fig. 5a, b), a known target of FLT3-ITD and RAS mutations^27,28 and a repressor of DNA damage response²⁹. Of note, DUSP6 has been reported to drive resistance to tyrosine kinase inhibitors in pre-B Acute Lymphoblastic Leukemia³⁰ as well as resistance to cisplatin treatment in solid tumors^31,32, and a recent pre-print³³ reports that elevated DUSP6 acts via the S6 protein to mediate resistance to JAK/STAT inhibition, leading to secondary adult AML. In agreement with this report, we find that untreated pAML samples with upregulated DUSP6 have higher levels of phosphorylated S6 (Supplementary Fig. 26).

**Fig. 5: Selected differentially upregulated genes in high-IL6/R pAML samples.**

In addition to DUSP6, high-IL6/R pAML samples with Inv(16) and RAS/FLT3 mutations had significantly upregulated levels of miR-221, a positive regulator of JAK/STAT and NF-κB signaling frequently upregulated in AML^34,35,36, whereas NUP98-rearranged high-IL6/R pAML had upregulated levels of STAT5A, a well-known hematopoietic regulator activated by more than 20 cytokines and growth factors in the BM³⁷.

In contrast to the above, DUSP6 is not upregulated in MLL-rearranged high-IL6/R pAML. Instead, in MLL-rearranged high-IL6/R samples with and without RAS/FLT3 mutations, we found CLEC11A (also known as stem cell growth factor (SCGF) and osteolectin)³⁸ and FLT3 are highly upregulated compared to NBM, and to a lesser extent compared to low-IL6/R pAML samples with matching genomic alterations (Fig. 5c). The role of FLT3 in pAML is well established³⁹. Self-secreted CLEC11A is required for the viability of multiple leukemia cell lines⁴⁰, can activate PI3K signaling⁴¹, and has been shown to increase the granulocyte/macrophage colony-forming abilities of G-CSF, GM-CSF, IL-3 and FLT3 ligand⁴². Compared to NBM, CLEC11A is expressed at higher levels in low- as well as high-IL6/R samples. However, CLEC11A binds integrin receptors^41,43, and 20 integrin alpha and beta subunit genes, as well as the ‘outside-in’ integrin signaling licensing genes TLN1 and KINDLIN3 are all more highly expressed in high- versus low-IL6/R samples (Supplementary Data 2). Thus, CLEC11A-integrin interactions may have much greater impact in high- compared to low-IL6/R pAML.

A recent analysis of data from The Cancer Genome Atlas (TCGA) suggests CLEC11A is epigenetically repressed in adult AML^44,45. We therefore checked the methylation status of CLEC11A in 68 MLL-rearranged high-IL-6/R pAML and 30 NBM samples for which we have assessed genome-wide DNA methylation levels using Illumina Infinium MethylationEPIC arrays. In our pediatric data, the ~1kbp CLEC11A upstream promoter region is strongly demethylated in MLL-rearranged high-IL6/R pAML compared to NBM, irrespective of the presence or absence of growth factor mutations (Fisher’s Exact Test P ~ 1.4E−05, Supplementary Fig. 28). Genome-wide, 7 CLEC11A-associated probes were differentially methylated with significantly higher methylation levels in NBM (all FDR-adjusted individual probe moderated T-test P values <0.01). Thus, in contrast to adult AML, CLEC11A is not epigenetically repressed in MLL-rearranged high-IL6/R pAML and is instead upregulated compared to NBM.

Taken together, the above findings suggest that distinct genes may contribute to poor clinical outcomes in different genomic subtypes of high-IL-6R pAML. Remarkably, all of these differentially upregulated genes converge on and regulate the inflammatory signaling pathways co-activated in high-IL6/R pAML.

Higher Toll Like Receptor (TLR) signaling in high versus low-IL6/R pAML

The alarmins S100A8 and S100A9 (hereon S100A8/9) and their receptor TLR4^46,47 are expressed at much higher levels in high-IL6/R pAML samples compared to low-IL6/R pAML samples (Fig. 6). Consistent with this finding, a recent study found that a cluster of 13 of 46 adult AML diagnostic BM samples (28%) had above-median basal expression of TLR4 and 8 and sharply higher expression levels of S100A8/9, IL-6, IL-1β, and TNFα following TLR4 stimulation with LPS⁴⁸.

**Fig. 6: TLR4 and S100A8/9 are expressed at higher levels in high-IL6/R samples compared to low-IL6/R, but not NBM.**

S100A8/9 expression levels in high-IL-6/R samples do not exceed levels in NBM and are thus not at supra-physiological levels that might be indicative of acute cellular stress. Higher S100A8/9 expression levels in high- versus low-IL6/R samples are also not due to differences in cellular stresses that might arise from sample processing delays⁴⁹ (Supplementary Fig. 29A). Rather, TLR4 and S100A8/9 expression appear to be downregulated in low-IL6/R pAML, and co-regulated in high-IL6/R pAML.

In agreement with previous findings in adults⁴⁶, in our pediatric samples, S100A8/9 are expressed in near-perfect correlation across both pAML and NBM samples (Pearson r = 0.98, 95% confidence interval: 0.975, 0.982, P < 2.2E−16, Supplementary Fig. 29B, C). In addition, S100A8/9 expression is highly correlated with the expression of TLR4 (Pearson r = 0.54, 95% confidence interval: 0.49, 0.61, P < 2.2E−16, Supplementary Fig. 29D, E). These findings suggest common regulatory mechanisms drive TLR4, S100A8, and S100A9 expression in the BM.

Across high- and low-IL6/R pAML samples, TLR4 expression is similarly highly correlated with IL-6 signaling activity (Pearson r = 0.56, 95% confidence interval = 0.49–0.63, P < 2.2E−16. Supplementary Fig. 30). Moreover, expression levels of the TLRs and their downstream target genes, the Interferon Response Factors (IRFs)⁵⁰ sharply segregate high- and low-IL6/R pAML samples (Supplementary Fig. 31), suggesting that TLR4 expression in high-IL6/R pAML samples is indicative of active TLR signaling in these samples. Taken together, these findings suggest S100A8/9-mediated TLR signaling in high-IL6/R pAML samples accompanies, and may contribute to, IL-6 and IFNα/β signaling activity in these samples.

S100A8/9 are transcriptionally activated by STAT3 and NF-κB^51,52,53 and in turn can act as transcriptional co-activators of NF-kB, AP-1 and STAT3^54,55. In addition, extracellular S100A8/9, released by activated, stressed, and dying neutrophils, monocytes, and macrophages, can bind TLR4⁴⁷ and activate IL-6, IL-1, and TNFα signaling^54,56. Thus, intra- and extracellular S100A8/9 can upregulate each other, and can form a self-reinforcing positive feedback loop that activates and maintains IL-6 signaling.

In adults, high expression of S100A8/9 has been shown to impact AML severity⁵⁷ and treatment resistance⁵⁸. In mice transplanted with AML-inducing cells overexpressing Hoxa9 and Meis1, S100a8 was found to block the differentiation of myeloid-lineage leukemic blasts, and anti-S100a8 antibody significantly improved survival⁴⁶.

Activated TLR signaling in high-IL6/R pAML may extend beyond TLR4. Compared to low-IL6/R pAML, high-IL6/R samples have significantly higher expression of 7 TLRs, as well as the TLR-activated interferon response factors (IRFs, Supplementary Fig. 32). In addition to IFNα/β signaling, TLR activation can also directly upregulate IL-6 signaling via increased AP-1 and NF-κB expression⁵⁹. In particular, high expression of TLR8, which is the most sharply increased TLR in high-IL6/R pAML, has been reported to increase serum IL-6 levels and drive multi-organ inflammation⁶⁰. High-IL6/R samples also have higher expression levels of myeloid and monocytic lineage markers compared to low-IL6/R pAML (Supplementary Figs. 33, 34). Of note, within monocytic and myelomonocytic pAML samples, IL-6 and IL-6R expression levels are much higher in high- versus low-IL6/R samples (Supplementary Fig. 35). Thus, differentiation status or sample composition differences do not explain differences in high- versus low-IL6/R cytokine signaling.

Taking the above observations together, we hypothesize that in high-IL6/R pAML, S100A8/9 and TLR expression may both activate and be activated by IL-6 signaling via NF-κB and AP-1.

Discussion

Approximately one-third of pAML patients who respond poorly to primary induction therapy die within three years of diagnosis⁶¹. IL-6 expression in the BM is associated with poor treatment response in both adult and pediatric AML, but its etiology has remained unknown. Using diagnosis stage BM samples from ~1500 pAML patients, we showed here that the IL-6, IFNα,β, IL-1/TLR, and TNFα pro-inflammatory signaling pathways are jointly activated in a subset of pAML patients with poor 2-year outcomes. After adjusting for age and common cytogenetic subtypes, high-IL6/R patients still have worse 2- and 5-year OS (Supplementary Figs. 36–38). Thus, factors beyond age and cytogenetics may contribute to the poor treatment response of this group.

Using a custom CyTOF panel, we showed that the genomic subtypes of high-IL6/R pAML have distinct cells of origin but share the same set of activated receptor-mediated inflammatory signaling pathways. Consistent with our findings, a previous functional screening study found that IL-6, IL-1 and GM-CSF signaling were jointly upregulated in a subset of adult AML patients with diverse genomic and disease subtypes⁶².

Remarkably, although each high-IL6/R genomic subtype is associated with high expression of a distinct set of leukemia-driving genes, all these genes activate the same set of signal transduction pathways. Specifically, the 3 gene pairs that are highly expressed in each high-IL6/R pAML subtype—namely DUSP6 and STAT5A in high-IL6/R pAML with NUP98 translocations, DUSP6 and miR-221 in high-IL6/R pAML with Inv(16) translocations, and CLEC11A and FLT3 in high-IL6/R pAML with MLL-rearrangements—all activate a common set of pathways: STAT-mediated signal transduction (STAT5A, miR221, FLT3), and PI3K/AKT-mediated signaling (DUSP6, CLEC11A). Another common feature of all high-IL6/R pAML genomic subtypes is the activation of NF-κB via both PI3K signaling and S100A8/9-mediated TLR signaling.

In addition to context-specific indirect regulation of subtype-specific genes, high-IL6/R associated genomic alterations are also known to directly activate the STAT, PI3K, MAPK, and NF-κB pathways. Most notably, activating RAS and FLT3 point mutations (which co-occur with MLL, NUP98 and Inv(16) translocations in high-IL6/R pAML) are well known to drive the above pathways. Additionally, the MLL protein can upregulate IL-6 transcription directly via interactions with a long non-coding RNA⁶³. It can also bind NF-κB target genes, increasing their expression synergistically⁶⁴. Interestingly, NUP98 fusion proteins physically interact with MLL and transcriptionally regulate many MLL target genes⁶⁵, and NUP98-fusion xenografts also upregulate IL-6 JAK/STAT3 signaling genes⁶⁶.

The fusion protein produced by the Inv(16) translocation has been shown to bind the DUSP6 gene and upregulate its transcription⁶⁷. In addition, MAPK/PI3K/NF-κB signaling is activated and plays an essential role in Inv(16) AML^68,69.

Signaling via S100A8/9 and TLR4 may be driven by distinct mechanisms in different high-IL6/R subtypes. In MLL-fusion AML, signaling via the TLR/IL-1 super-family enhances the chromatin occupancy and downstream effects of MLL-fusion proteins⁷⁰. In addition, the MLL protein (encoded by KMT2A) can facilitate NF-κB activity epigenetically⁶⁴, leading to higher IL-6 expression⁷¹. Consistent with these findings, human BM cells engineered to express the MLL-AF9 fusion protein upregulated TLR4 and S100A8/9⁷².

Two previous studies provide supporting evidence that in high-IL6/R pAML with NUP98 and Inv(16) translocations and RAS/FLT3 mutations, upregulated DUSP6 levels may drive TLR signaling and S100A8/9 transcription. In murine macrophages, DUSP6 was shown to facilitate TLR4/NF-κB driven production of IL-1β and IL-6⁷³. In addition, S100a8 and S100a9 were both highly upregulated in mouse BM cells engineered to co-express the Nup98-Nsd1 fusion and an activating RAS mutation⁷⁴, the same genomic alterations as in genomic Group 3 high-IL6/R pAML samples, which have upregulated DUSP6 expression.

In summary, in spite of their genomic and transcriptomic differences, high-IL6/R pAML samples are characterized by a shared set of co-activated signaling pathways (Supplementary Fig. 39). A model of these convergent regulatory interactions is presented schematically in Fig. 7. We note that, once established, such a system of interactions can become self-reinforcing via positive-feedback loops involving secreted S100A8/9, IL-6, IL-1, IFNα/β, and GM-CSF.

**Fig. 7: Schematic summary and model of the regulatory interactions reported.**

Upregulated IL-6, CLEC11A, DUSP6, and S100A8/9 have all been shown to contribute to AML disease severity and treatment resistance. Four additional lines of evidence support the hypothesis that the pathways activated by these genes are key drivers of poor clinical outcomes in high-IL6/R pAML.

First, the IL-6, CLEC11A, DUSP6, and S100A8/9 pathways are the most significantly impacted in terms of high- versus low-IL6/R differentially expressed downstream gene sets. Second, patient stratification based on the expression levels of IL-6 and IL-6R is a better predictor of clinical outcomes than models based on other genes, pathways, and gene sets, suggesting IL-6 signaling is a key contributor to the observed poorer outcomes. Third, the self-reinforcing positive feedback loops created by IL-6, IFNα,β, IL-1 and S100A8/9 signaling in leukemic cells (Fig. 7, Supplementary Fig. 25) can generate highly stable clonal leukemic cell populations that are more likely to become dominant during leukemogenesis. Fourth, the signaling pathways co-activated in high-IL6/R pAML are known to be regulated by the subtype-specific genes highly expressed in high-IL6/R pAML. Thus, multiple redundant pathways may converge-on and drive the activation of cytokine signaling in high-IL6/R pAML. Such “coherent Feed Forward Loop” network motifs have been shown to be a very common and robust mechanism for responses to perturbations^75,76.

Our study has a number of limitations. All data reported in this study were collected ex vivo from bio-banked samples. Additional regulatory interactions may exist in vivo. Currently, there are no syngeneic mouse models of high-IL6/R pAML, and experiments using fresh (<2 h from extraction) pAML BM samples raise many ethical and operational concerns. In the validation experiments reported here, we approximated the BM microenvironment by using transwells to co-culture thawed biorepository pAML BM samples with stromal HS-5 cells. This approach has two important limitations. First, several cell types present in the BM (e.g., granulocytes, bone tissue) are missing in our experiments. Second, the signaling events of interest occur at different timescales ranging from minutes to hours. Thus, our 25-minute assay provides only a qualitative confirmation of the signaling events and is not a quantitative measure of peak signal activity levels. We also note that the number of samples in our perturbation studies was small. Larger numbers of samples will be needed to confirm the generality of our findings.

We previously reported distinct age-related differences in the patterns of co-occurring and recurrent somatic genomic alterations in pAML². Here we showed that just over 1 in 5 pAML patients who are predominantly infants and children <10 years old have higher activity levels in multiple inflammatory cytokine signaling pathways and poorer treatment responses.

In another recent study⁷⁷, we report that infants with AML (defined as being ≤3 years old at diagnosis) also have a distinct BM immune profile. Compared to pAML in children older than 5 years, 95 of the top 200 genes with higher expression in infant-AML are lymphocyte-associated genes (FDR-adjusted Fisher’s Exact Test P ~ 7.7E−13), and 55 are in the Gene Ontology category “immune system process” (FDR-adjusted Fisher’s Exact Test P value ~ 2.6E−8). These findings raise the possibility that leukemic immune activity may be a common occurrence in younger children with pAML, and provide a potential clue to the previously observed increased risk of AML in patients with autoimmunity⁷⁸.

Taken together, our findings suggest BM leukemic inflammatory signaling is widespread in younger children with pAML, is driven by multiple groups of highly recurrent genomic alterations, and co-activates the JAK/STAT, PI3K, ERK1/2, and NF-kB pathways in all genomic groups. Thus, irrespective of their genomic subtype, high-IL6/R pAML may benefit from therapies that inhibit these co-activated pathways. Effective inhibitors of each of the pathways are already widely available, facilitating the development of combination therapies to inhibit multi-cytokine signaling in high-IL6/R pAML.

Methods

Pediatric AML biological samples were collected with informed consent (and in accordance with the Declaration of Helsinki) from patients diagnosed with de novo AML and enrolled on Children’s Oncology Group (COG) trials AAML0531 (NCT00372593), or AAML1031 (NCT01371981). Each protocol was approved by the National Cancer Institute’s central institutional review board (IRB) and the local IRB at Fred Hutchinson Cancer Center (Protocol 9950).

AAML1031 and AAML0531 RNA-seq

Total RNA was extracted from ficoll-enriched, viably cryopreserved samples from the COG biorepository using the AllPrep Universal Extraction Kit (Qiagen). Nucleic acids were quantified by NanoDrop (Thermo Scientific). RNA samples were tested for quality and integrity using the Agilent 2100 Bioanalyzer (Agilent Technologies). Ribosomal RNA (rRNA) species were removed from total RNA using the NEBNext rRNA Depletion Kit for Human/Mouse/Rat (NEB, E6310X). Strand-specific mRNA libraries were constructed and sequenced at the British Columbia Cancer Agency’s Genome Sciences Centre using an Illumina HiSeq2500 and 75 bp paired-end reads, as previously described².

Ex vivo cell stimulation

Frozen AML cells were thawed in media containing IMDM, 20% BCS, and 100 U/mL DNase I, and rested overnight in media containing 100 ng/mL each of SCF, IL-3, G-CSF, and GM-CSF. Rested cells were transferred to the upper chamber of a transwell with confluent HS-5 cells in its lower chamber with/without added Ruxolitinib. 25 minutes later, an aliquot of the cells was harvested for CyTOF analysis. For RNA-seq, the cells were harvested ~7 h after being transferred to the transwell. HS-5 cells were a gift from Dr. Beverly Torok-Strob (Fred Hutchinson Cancer Center).

SMART-seq V4 ultra-low-input RNA-seq

Total RNA (0.5 ng) was added to lysis buffer from the SMART-Seq v4 Ultra Low Input RNA Kit for Sequencing (Takara), and reverse transcription was performed followed by PCR amplification to generate full-length amplified cDNA. Sequencing libraries were constructed using the NexteraXT DNA sample preparation kit with unique dual indexes (Illumina) to generate Illumina-compatible barcoded libraries. Libraries were pooled and quantified using a Qubit® Fluorometer (Life Technologies). Sequencing of pooled libraries was carried out on a NextSeq 2000 sequencer (Illumina) with paired-end 59-base reads, using NextSeq P2 sequencing kits (Illumina) with a target depth of 5 million reads per sample. Base calls were processed to FASTQs on BaseSpace (Illumina), and a base call quality-trimming step was applied to remove low-confidence base calls from the ends of reads. The FASTQs were aligned to the GRCh38 human reference genome, using STAR v.2.4.2a (https://github.com/alexdobin/STAR/releases?page=4) and gene counts were generated using htseq-count (https://htseq.readthedocs.io/en/master/install.html). QC and metrics analysis was performed using the Picard family of tools (v1.134).

Mass cytometry/CyTOF

Details of the antibodies used are given in Supplementary Table 3. To ensure optimal performance, all signaling probes were tested at multiple titrations with BM and blood cells unstimulated and following 25 min of stimulation with IFNB, IL6, LPS, or TCR beads. Upon thaw, cells had high viability (mean 90%, range 78–99%, and retained good viability after overnight resting (mean 79%, range 64–72%). Dead cells were excluded from signaling analyses. There was no association of overall cell viability with the upregulation of signaling.

Dead cells were labeled by incubation in Cell-ID™ Cisplatin (Standard BioTools) solution (5 µM in PBS) for 5 min at 4 °C, then the reaction was quenched and cells washed with addition of an equal volume cold Media (RPMI + 25% FBS). Cells were then resuspended in Media and fixed by addition of and equal volume Fix I Buffer (BD Biosciences) and incubated for 15 min at 37 °C. Cells were next barcoded using Standard BioTools Cell-ID 20-Plex Pd Barcoding Kit, as per manufacturer instructions. After washing, all samples from an individual were combined, and then resuspended in surface staining cocktail in Maxpar® Cell Staining Buffer (CSB, Standard BioTools) for 20 min at room temperature. Combined samples were washed then fixed with 1.6% paraformaldehyde in PBS for 10 minutes at room temperature. Samples were next incubated in ice-cold Perm Buffer III (BD Biosciences) for 1 h at −20 °C, then washed with CSB and then resuspended in intracellular staining cocktail in CSB for 20 min at room temperature. Samples were again washed and fixed with 1.6% paraformaldehyde in PBS for 10 min at room temperature, then stored in Maxpar® Fix and Perm Buffer (Standard BioTools) containing 125 nM Cell-ID™ Intercalator-Ir at 4 °C until acquisition. On the day of acquisition, samples were washed with CSB and then cold ultrapure water, and kept at 4 °C. Immediately before acquisition, samples were resuspended in cold ultrapure water containing 1/5^th by volume EQ Four Element Calibration Beads (Standard BioTools) to at target concentration of 1 million cells/mL.

FCS files were normalized and randomized with CyTOF Software (version 7.0.8493) using Uniform Negative Distribution for the randomization and Median Bead Intensity with Passport EQ-P13H2302_ver2 without removing beads for the normalization.

Gating analysis was done using FlowJo version 10.7.1 (BD, flowjo.com). Samples were first debarcoded by gating on cells triple-positive for Palladium isotopes (masses 104, 106, 108, and 110), and individual samples were exported as separate FCS files.

Within each stimulation condition sample, cells were gated to exclude beads (Ce140−), isolate singlets (using Ir191 and Event Length), and capture live cells (Pt198−). Immune cells were identified as live cells that expressed high levels of CD45 but lacked the stromal cell marker CD73. The strategy for identifying stem-cell-like AML blasts was to negatively gate out other immune subsets (Supplementary Data 5). Immune cells were divided into NK cells (CD56+CD3−) and T cells (CD3+CD56−). Within the CD3-CD56- portion, B cells were identified as those expressing CD19 and/or CD20. Non-B cells (CD19-CD20−) were gated to exclude Granulocytes (CD66b+CD15+), and within the non-granulocytes, Monocytes (expressing CD14 and/or CD16) were identified. CD14-CD16- (non-monocytes) were then divided by expression of CD34. Within CD34+ cells, a CD38+ CMP/GMP subset was identified (CMP/GMP AML). Within the CD34− cells, a CD123-CD33+ subset was identified (CD34- AML).

For UMAP projection and probe intensity histograms, signal values were asinh transformed using the Bioconductor (bioconductor.org/) package “CytofKIT” (https://github.com/JinmiaoChenLab/cytofkit) with a co-factor of 5, and analyzed using the Bioconductor packages “flowCore” (https://bioconductor.org/packages/release/bioc/html/flowCore.html), “CytoML” (https://bioconductor.org/packages/release/bioc/html/CytoML.html), and “flowWorkSpace” (https://bioconductor.org/packages/release/bioc/html/flowWorkspace.html). UMAPs were generated using the Bioconductor packages ‘SingleCellExperiment’ (https://bioconductor.org/packages/release/bioc/html/SingleCellExperiment.html), “scMerge” (https://bioconductor.org/packages/release/bioc/html/scMerge.html) and “scater” (https://bioconductor.org/packages/release/bioc/html/scater.html).

Statistics

All analyses were performed using R (cran.r-project.org/) and Bioconductor (bioconductor.org/) packages. The analyses reported did not involve any custom statistical procedures, and did not involve any modifications to the libraries/packages used. Unless specifically noted, we used default method/function parameters. P values for all multiple-tests were adjusted using the FDR method unless otherwise stated. Bimodal distribution of IL-6 and IL-6R were modeled using the R package “mclust” (https://cran.r-project.org/web/packages/mclust/index.html). Differential expression analyses were performed using the Bioconductor “DESeq2” (https://bioconductor.org/packages/release/bioc/html/DESeq2.html), “edgeR” (https://bioconductor.org/packages/release/bioc/html/edgeR.html), and “limma” (https://bioconductor.org/packages/release/bioc/html/limma.html) R packages. Kaplan-Meier plots and associated P values were generated using the “survival” (https://cran.r-project.org/web/packages/survival/index.html) and “survminer” (https://cran.r-project.org/web/packages/survminer/index.html) R packages. Gene Set Enrichment analysis was performed using the Bioconductor package “enrichplot” (https://bioconductor.org/packages/release/bioc/html/enrichplot.html). Unsupervised hierarchical clustering heatmaps were generated using the “pheatmap” (https://cran.r-project.org/web/packages/pheatmap/index.html) and “ComplexHeatmap” (https://www.bioconductor.org/packages/release/bioc/html/ComplexHeatmap.html) R packages. Correlation tests were performed using the R function “corr.test” and the “rcorr” function in the R package “Hmisc” (https://cran.r-project.org/web/packages/Hmisc/index.html). DNA-methylation analyses were performed using the Bioconductor packages “minfi” (https://bioconductor.org/packages/release/bioc/html/minfi.html), “missMethyl” (https://bioconductor.org/packages/release/bioc/html/missMethyl.html), and “DMRcate” (https://bioconductor.org/packages/release/bioc/html/DMRcate.html) and the “IlluminaHumanMethylationEPICanno.ilm10b2.hg19” (https://bioconductor.org/packages/release/data/annotation/html/IlluminaHumanMethylationEPICanno.ilm10b2.hg19.html), and “IlluminaHumanMethylationEPICmanifest” (https://bioconductor.org/packages/release/data/annotation/html/IlluminaHumanMethylationEPICmanifest.html) annotation packages. Differential methylation results were visualized using the R package “Gviz” (https://bioconductor.org/packages/release/bioc/html/Gviz.html). All other visualizations were generated using the R “ggplot2” (https://cran.r-project.org/web/packages/ggplot2/index.html) package.

Study approval

All data presented here are derived from COG AAML1031 and COG AAML0531 samples. All patient samples were obtained by member COG institutions after written consent from the parents/guardians of minors upon enrolling in the trial. The study was overseen by the Institutional Review Board at Fred Hutchinson Cancer Research Center (Protocol 9950).

Inclusion and ethics

All data presented here are from samples collected independently for Children’s Oncology Group (COG) studies AAML1031 and AAML0531.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

The raw RNA-seq data for the studies reported here have been deposited in the Database of Genotypes and Phenotypes (dbGaP) under the study ID phs000465.v21.p8. Subject to the NIH Genomic Data Sharing Policy, the raw data are freely available to all researchers via https://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs000465.v21.p8. Processed data are available at the National Cancer Institute’s Genomic Data Commons (https://portal.gdc.cancer.gov/) under the TARGET-AML project. Selected clinical (e.g., age, EFS, OS, cytogenetic classification) and molecular features (e.g., KIT, RAS, NPM1, WT1, CEBPA, IDH1 mutations, and FLT3/ITD allelic ratios) were clinically available for patients included in the NCI/TARGET cohort and are included in the clinical data file available via the TARGET data matrix (ocg.cancer.gov/programs/target/data-matrix). Source data are provided with this paper.

References

Schulpen, M. et al. Increased survival disparities among children and adolescents & young adults with acute myeloid leukemia: a Dutch population-based study. Int J. Cancer 150, 1101–1112 (2022).
Article CAS PubMed Google Scholar
Bolouri, H. et al. The molecular landscape of pediatric acute myeloid leukemia reveals recurrent structural alterations and age-specific mutational interactions. Nat. Med. 24, 103–112 (2018).
Article CAS PubMed Google Scholar
McCarthy, J. J., McLeod, H. L. & Ginsburg, G. S. Genomic medicine: a decade of successes, challenges, and opportunities. Sci. Transl. Med. 5, 189sr184 (2013).
Article Google Scholar
Zhang, H., Klareskog, L., Matussek, A., Pfister, S. M. & Benson, M. Translating genomic medicine to the clinic: challenges and opportunities. Genome Med. 11, 9 (2019).
Article PubMed PubMed Central Google Scholar
Hanahan, D. & Weinberg, R. A. Hallmarks of cancer: the next generation. Cell 144, 646–674 (2011).
Article CAS PubMed Google Scholar
Kumari, N., Dwarakanath, B. S., Das, A. & Bhatt, A. N. Role of interleukin-6 in cancer progression and therapeutic resistance. Tumour Biol. 37, 11553–11572 (2016).
Article CAS PubMed Google Scholar
Johnson, D. E., O’Keefe, R. A. & Grandis, J. R. Targeting the IL-6/JAK/STAT3 signalling axis in cancer. Nat. Rev. Clin. Oncol. 15, 234–248 (2018).
Article CAS PubMed PubMed Central Google Scholar
Ikebuchi, K. et al. Interleukin 6 enhancement of interleukin 3-dependent proliferation of multipotential hemopoietic progenitors. Proc. Natl Acad. Sci. USA 84, 9035–9039 (1987).
Article ADS CAS PubMed PubMed Central Google Scholar
Stevens, A. M., Miller, J. M., Munoz, J. O., Gaikwad, A. S. & Redell, M. S. Interleukin-6 levels predict event-free survival in pediatric AML and suggest a mechanism of chemotherapy resistance. Blood Adv. 1, 1387–1397 (2017).
Article CAS PubMed PubMed Central Google Scholar
Hou, D. et al. Stromal cells promote chemoresistance of acute myeloid leukemia cells via activation of the IL-6/STAT3/OXPHOS axis. Ann. Transl. Med. 8, 1346 (2020).
Article CAS PubMed PubMed Central Google Scholar
Sanchez-Correa, B. et al. Cytokine profiles in acute myeloid leukemia patients at diagnosis: survival is inversely correlated with IL-6 and directly correlated with IL-10 levels. Cytokine 61, 885–891 (2013).
Article CAS PubMed Google Scholar
Harmer, D., Falank, C. & Reagan, M. R. Interleukin-6 Interweaves the Bone marrow microenvironment, bone loss, and multiple myeloma. Front Endocrinol. (Lausanne) 9, 788 (2018).
Article PubMed Google Scholar
Hemmati, S., Haque, T. & Gritsman, K. Inflammatory signaling pathways in preleukemic and leukemic stem cells. Front. Oncol. 7, 265 (2017).
Article PubMed PubMed Central Google Scholar
Hirano, T. IL-6 in inflammation, autoimmunity and cancer. Int Immunol. 33, 127–148 (2021).
Article CAS PubMed Google Scholar
Cook, A. M. et al. Role of altered growth factor receptor-mediated JAK2 signaling in growth and maintenance of human acute myeloid leukemia stem cells. Blood 123, 2826–2837 (2014).
Article CAS PubMed PubMed Central Google Scholar
Karjalainen, R. et al. JAK1/2 and BCL2 inhibitors synergize to counteract bone marrow stromal cell-induced protection of AML. Blood 130, 789–802 (2017).
Article CAS PubMed Google Scholar
Habbel, J. et al. Inflammation-driven activation of JAK/STAT signaling reversibly accelerates acute myeloid leukemia in vitro. Blood Adv. 4, 3000–3010 (2020).
Article CAS PubMed PubMed Central Google Scholar
Zhang, T. Y. et al. IL-6 blockade reverses bone marrow failure induced by human acute myeloid leukemia. Sci. Transl. Med. 12, https://doi.org/10.1126/scitranslmed.aax5104 (2020).
Jones, V. S. et al. Cytokines in cancer drug resistance: cues to new therapeutic strategies. Biochim. Biophys. Acta 1865, 255–265 (2016).
CAS PubMed Google Scholar
Graf, L., Iwata, M. & Torok-Storb, B. Gene expression profiling of the functionally distinct human bone marrow stromal cell lines HS-5 and HS-27a. Blood 100, 1509–1511 (2002).
Article CAS PubMed Google Scholar
Adamo, A. et al. HS-5 and HS-27A stromal cell lines to study bone marrow mesenchymal stromal cell-mediated support to cancer development. Front. Cell Dev. Biol. 8, 584232 (2020).
Article PubMed PubMed Central Google Scholar
Yeleswaram, S. et al. Inhibition of cytokine signaling by ruxolitinib and implications for COVID-19 treatment. Clin. Immunol. 218, 108517 (2020).
Article CAS PubMed PubMed Central Google Scholar
Hoesel, B. & Schmid, J. A. The complexity of NF-kappaB signaling in inflammation and cancer. Mol. Cancer 12, 86 (2013).
Article CAS PubMed PubMed Central Google Scholar
Ryan, M. B., Der, C. J., Wang-Gillam, A. & Cox, A. D. Targeting RAS-mutant cancers: is ERK the key? Trends Cancer 1, 183–198 (2015).
Article PubMed PubMed Central Google Scholar
Marchetti, S. et al. Extracellular signal-regulated kinases phosphorylate mitogen-activated protein kinase phosphatase 3/DUSP6 at serines 159 and 197, two sites critical for its proteasomal degradation. Mol. Cell Biol. 25, 854–864 (2005).
Article CAS PubMed PubMed Central Google Scholar
Bermudez, O., Marchetti, S., Pages, G. & Gimond, C. Post-translational regulation of the ERK phosphatase DUSP6/MKP3 by the mTOR pathway. Oncogene 27, 3685–3691 (2008).
Article CAS PubMed Google Scholar
Arora, D. et al. Expression of protein-tyrosine phosphatases in Acute Myeloid Leukemia cells: FLT3 ITD sustains high levels of DUSP6 expression. Cell Commun. Signal 10, 19 (2012).
Article CAS PubMed PubMed Central Google Scholar
Unni, A. M. et al. Hyperactivation of ERK by multiple mechanisms is toxic to RTK-RAS mutation-driven lung adenocarcinoma cells. Elife 7, https://doi.org/10.7554/eLife.33718 (2018).
Bagnyukova, T. V. et al. DUSP6 regulates drug sensitivity by modulating DNA damage response. Br. J. Cancer 109, 1063–1071 (2013).
Article CAS PubMed PubMed Central Google Scholar
Shojaee, S. et al. Erk negative feedback control enables pre-B cell transformation and represents a therapeutic target in acute lymphoblastic leukemia. Cancer Cell 28, 114–128 (2015).
Article CAS PubMed PubMed Central Google Scholar
Messina, S. et al. Dual-specificity phosphatase DUSP6 has tumor-promoting properties in human glioblastomas. Oncogene 30, 3813–3820 (2011).
Article CAS PubMed Google Scholar
Gao, Y. et al. Overexpression of DUSP6 enhances chemotherapy-resistance of ovarian epithelial cancer by regulating the ERK signaling pathway. J. Cancer 11, 3151–3164 (2020).
Article CAS PubMed PubMed Central Google Scholar
Kong, T. et al. DUSP6 mediates resistance to JAK2 inhibition and drives leukemic progression. bioRxiv 2021. https://doi.org/10.1101/2021.06.27.449656 (2021).
Liu, S. et al. A microRNA 221- and 222-mediated feedback loop maintains constitutive activation of NFkappaB and STAT3 in colorectal cancer cells. Gastroenterology 147, 847–859.e811 (2014).
Article CAS PubMed Google Scholar
Song, Q. et al. Role of miR-221/222 in tumor development and the underlying mechanism. J. Oncol. https://doi.org/10.1155/2019/7252013 (2019).
Rommer, A. et al. Overexpression of primary microRNA 221/222 in acute myeloid leukemia. BMC Cancer 13, 364 (2013).
Article CAS PubMed PubMed Central Google Scholar
Schepers, H., Wierenga, A. T., Vellenga, E. & Schuringa, J. J. STAT5-mediated self-renewal of normal hematopoietic and leukemic stem cells. JAKSTAT 1, 13–22 (2012).
PubMed PubMed Central Google Scholar
Yue, R., Shen, B. & Morrison, S. J. Clec11a/osteolectin is an osteogenic growth factor that promotes the maintenance of the adult skeleton. Elife 5, https://doi.org/10.7554/eLife.18782 (2016).
Tarlock, K. et al. Disease characteristics and prognostic implications of cell-surface FLT3 receptor (CD135) expression in pediatric acute myeloid leukemia: a report from the Children’s Oncology Group. Clin. Cancer Res. 23, 3649–3656 (2017).
Article CAS PubMed PubMed Central Google Scholar
Hiraoka, A. Leukemia cell lines require self-secreted stem cell growth factor (SCGF) for their proliferation. Leuk. Res. 32, 1623–1625 (2008).
Article CAS PubMed Google Scholar
Wang, M. et al. Molecular structure, expression, and functional role of Clec11a in skeletal biology and cancers. J. Cell Physiol. 235, 6357–6365 (2020).
Article CAS PubMed Google Scholar
Hiraoka, A. et al. Stem cell growth factor: in situ hybridization analysis on the gene expression, molecular characterization and in vitro proliferative activity of a recombinant preparation on primitive hematopoietic progenitor cells. Hematol. J. 2, 307–315 (2001).
Article CAS PubMed Google Scholar
Shen, B. et al. Integrin alpha11 is an Osteolectin receptor and is required for the maintenance of adult skeletal bone mass. Elife 8, https://doi.org/10.7554/eLife.42274 (2019).
Li, Y. et al. Detection of prognostic methylation markers by methylC-capture sequencing in acute myeloid leukemia. Oncotarget 8, 110444–110459 (2017).
Article PubMed PubMed Central Google Scholar
Zhao, X., Li, Y. & Wu, H. A novel scoring system for acute myeloid leukemia risk assessment based on the expression levels of six genes. Int J. Mol. Med. 42, 1495–1507 (2018).
CAS PubMed PubMed Central Google Scholar
Laouedj, M. et al. S100A9 induces differentiation of acute myeloid leukemia cells through TLR4. Blood 129, 1980–1990 (2017).
Article CAS PubMed Google Scholar
Frosch, M. et al. The myeloid-related proteins 8 and 14 complex, a novel ligand of toll-like receptor 4, and interleukin-1beta form a positive feedback mechanism in systemic-onset juvenile idiopathic arthritis. Arthritis Rheum. 60, 883–891 (2009).
Article CAS PubMed Google Scholar
Brenner, A. K. & Bruserud, O. Functional toll-like receptors (TLRs) are expressed by a majority of primary human acute myeloid leukemia cells and inducibility of the tlr signaling pathway is associated with a more favorable phenotype. Cancers (Basel) 11, https://doi.org/10.3390/cancers11070973 (2019).
Savage, A. K. et al. Multimodal analysis for human ex vivo studies shows extensive molecular changes from delays in blood processing. iScience 24, 102404 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Jefferies, C. A. Regulating IRFs in IFN driven disease. Front Immunol. 10, 325 (2019).
Article CAS PubMed PubMed Central Google Scholar
Nemeth, J. et al. S100A8 and S100A9 are novel nuclear factor kappa B target genes during malignant progression of murine and human liver carcinogenesis. Hepatology 50, 1251–1262 (2009).
Article CAS PubMed Google Scholar
Cheng, P. et al. Inhibition of dendritic cell differentiation and accumulation of myeloid-derived suppressor cells in cancer is regulated by S100A9 protein. J. Exp. Med. 205, 2235–2249 (2008).
Article CAS PubMed PubMed Central Google Scholar
Xu, K. & Geczy, C. L. IFN-gamma and TNF regulate macrophage expression of the chemotactic S100 protein S100A8. J. Immunol. 164, 4916–4923 (2000).
Article CAS PubMed Google Scholar
Simard, J. C. et al. S100A8 and S100A9 induce cytokine expression and regulate the NLRP3 inflammasome via ROS-dependent activation of NF-kappaB(1.). PLoS ONE 8, e72138 (2013).
Article ADS CAS PubMed PubMed Central Google Scholar
Song, R. & Struhl, K. S100A8/S100A9 cytokine acts as a transcriptional coactivator during breast cellular transformation. Sci. Adv. 7, https://doi.org/10.1126/sciadv.abe5357 (2021).
Sunahori, K. et al. The S100A8/A9 heterodimer amplifies proinflammatory cytokine production by macrophages via activation of nuclear factor kappa B and p38 mitogen-activated protein kinase in rheumatoid arthritis. Arthritis Res. Ther. 8, R69 (2006).
Article PubMed PubMed Central Google Scholar
Nicolas, E. et al. Expression of S100A8 in leukemic cells predicts poor survival in de novo AML patients. Leukemia 25, 57–65 (2011).
Article CAS PubMed Google Scholar
Karjalainen, R. et al. Elevated expression of S100A8 and S100A9 correlates with resistance to the BCL-2 inhibitor venetoclax in AML. Leukemia 33, 2548–2553 (2019).
Article PubMed Google Scholar
Zhu, J. & Mohan, C. Toll-like receptor signaling pathways–therapeutic opportunities. Mediators Inflamm. https://doi.org/10.1155/2010/781235 (2010).
Guiducci, C. et al. RNA recognition by human TLR8 can lead to autoimmune inflammation. J. Exp. Med. 210, 2903–2919 (2013).
Article CAS PubMed PubMed Central Google Scholar
McNeer, N. A. et al. Genetic mechanisms of primary chemotherapy resistance in pediatric acute myeloid leukemia. Leukemia 33, 1934–1943 (2019).
Article CAS PubMed PubMed Central Google Scholar
Carey, A. et al. Identification of Interleukin-1 by functional screening as a key mediator of cellular expansion and disease progression in acute myeloid leukemia. Cell Rep. 18, 3204–3218 (2017).
Article CAS PubMed PubMed Central Google Scholar
Fanucchi, S. et al. Immune genes are primed for robust transcription by proximal long noncoding RNAs located in nuclear compartments. Nat. Genet. 51, 138–150 (2019).
Article CAS PubMed Google Scholar
Wang, X. et al. MLL1, a H3K4 methyltransferase, regulates the TNFalpha-stimulated activation of genes downstream of NF-kappaB. J. Cell Sci. 125, 4058–4066 (2012).
CAS PubMed Google Scholar
Xu, H. et al. NUP98 fusion proteins interact with the NSL and MLL1 complexes to drive leukemogenesis. Cancer Cell 30, 863–878 (2016).
Article CAS PubMed PubMed Central Google Scholar
Cardin, S. et al. Human models of NUP98-KDM5A megakaryocytic leukemia in mice contribute to uncovering new biomarkers and therapeutic vulnerabilities. Blood Adv. 3, 3307–3321 (2019).
Article ADS CAS PubMed PubMed Central Google Scholar
Mandoli, A. et al. CBFB-MYH11/RUNX1 together with a compendium of hematopoietic regulators, chromatin modifiers and basal transcription factors occupies self-renewal genes in inv(16) acute myeloid leukemia. Leukemia 28, 770–778 (2014).
Article CAS PubMed Google Scholar
Simonis, A. et al. Disruption of CSF-1R signaling inhibits growth of AML with inv(16). Blood Adv. 5, 1273–1277 (2021).
Edwards, D. K. T. et al. CSF1R inhibitors exhibit antitumor activity in acute myeloid leukemia by blocking paracrine signals from support cells. Blood 133, 588–599 (2019).
Article CAS PubMed PubMed Central Google Scholar
Liang, K. et al. Therapeutic targeting of MLL degradation pathways in MLL-rearranged leukemia. Cell 168, 59–72.e13 (2017).
Article CAS PubMed PubMed Central Google Scholar
Minotti, R., Andersson, A. & Hottiger, M. O. ARTD1 suppresses interleukin 6 expression by repressing MLL1-dependent histone H3 trimethylation. Mol. Cell Biol. 35, 3189–3199 (2015).
Article CAS PubMed PubMed Central Google Scholar
Milan, T. et al. Epigenetic changes in human model KMT2A leukemias highlight early events during leukemogenesis. Haematologica https://doi.org/10.3324/haematol.2020.271619 (2020).
Zhang, F., Tang, B., Zhang, Z., Xu, D. & Ma, G. DUSP6 inhibitor (E/Z)-BCI hydrochloride attenuates lipopolysaccharide-induced inflammatory responses in murine macrophage cells via activating the Nrf2 signaling axis and inhibiting the NF-kappaB pathway. Inflammation 42, 672–681 (2019).
Article CAS PubMed Google Scholar
Mohanty, S. et al. Targeted inhibition of the NUP98-NSD1 fusion oncogene in acute myeloid leukemia. Cancers (Basel) 12, https://doi.org/10.3390/cancers12102766 (2020).
Milo, R. et al. Network motifs: simple building blocks of complex networks. Science 298, 824–827 (2002).
Article ADS CAS PubMed Google Scholar
Le, D. H. & Kwon, Y. K. A coherent feedforward loop design principle to sustain robustness of biological networks. Bioinformatics 29, 630–637 (2013).
Article CAS PubMed Google Scholar
Bolouri, H. et al. A B-cell developmental gene regulatory network is activated in infant AML. PLoS ONE 16, e0259197 (2021).
Article CAS PubMed PubMed Central Google Scholar
Kristinsson, S. Y. et al. Chronic immune stimulation might act as a trigger for the development of acute myeloid leukemia or myelodysplastic syndromes. J. Clin. Oncol. 29, 2897–2903 (2011).
Article PubMed PubMed Central Google Scholar

Download references

Acknowledgements

The authors are grateful to the children and families who provided samples for COG studies AAML1031 and AAML0531. The generation of multi-omic RNA-seq datasets for these cohorts was funded by the National Cancer Institute (U10CA98543, SM), National Institutes of Health (HHSN261200800001E, SM), Children’s Oncology Group Foundation (SM), and the Target Pediatric AML initiative (targetpediatricaml.org, SM). We thank Drs. Anne Hocking, Taylor Lawson, and Virginia Green for their help with manuscript preparation.

Author information

Jenny L. Smith
Present address: Research Scientific Computing, Seattle Children’s Research Institute, 818 Stewart Street, Seattle, WA, USA
These authors contributed equally: Hamid Bolouri, Rhonda E. Ries, Alice E. Wiedeman.
These authors jointly supervised this work: Hamid Bolouri, Soheil Meshinchi.

Authors and Affiliations

Center for Systems Immunology, Benaroya Research Institute, 1201 9th Ave, Seattle, WA, USA
Hamid Bolouri, Vivian H. Gersuk, Kimberly O’Brien & Quynh-Anh Nguyen
Clinical Research Division, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave N, Seattle, WA, USA
Rhonda E. Ries, Tiffany Hylkema, Jenny L. Smith & Soheil Meshinchi
Center for Translational Immunology, Benaroya Research Institute, 1201 9th Ave, Seattle, WA, USA
Alice E. Wiedeman, Sheila Scheiding & S. Alice Long

Authors

Hamid Bolouri
View author publications
You can also search for this author in PubMed Google Scholar
Rhonda E. Ries
View author publications
You can also search for this author in PubMed Google Scholar
Alice E. Wiedeman
View author publications
You can also search for this author in PubMed Google Scholar
Tiffany Hylkema
View author publications
You can also search for this author in PubMed Google Scholar
Sheila Scheiding
View author publications
You can also search for this author in PubMed Google Scholar
Vivian H. Gersuk
View author publications
You can also search for this author in PubMed Google Scholar
Kimberly O’Brien
View author publications
You can also search for this author in PubMed Google Scholar
Quynh-Anh Nguyen
View author publications
You can also search for this author in PubMed Google Scholar
Jenny L. Smith
View author publications
You can also search for this author in PubMed Google Scholar
S. Alice Long
View author publications
You can also search for this author in PubMed Google Scholar
Soheil Meshinchi
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

H.B., R.R., and A.E.W. contributed equally to this work. Writing the manuscript: all authors. Data generation and annotation: S.M., R.R., T.H., J.L.S., S.A.L., A.E.W., S.S., V.H.G., K.O., Q.A.N. Study design and analyses: H.B., R.R., A.E.W. Interpretation of analysis results: all authors.

Corresponding authors

Correspondence to Hamid Bolouri or Soheil Meshinchi.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Laila Poisson and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Peer Review File

Reporting Summary

Description of Additional Supplementary Files

Supplementary Data 1

Supplementary Data 2

Supplementary Data 3

Supplementary Data 4

Supplementary Data 5

Supplementary Data 6

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Bolouri, H., Ries, R.E., Wiedeman, A.E. et al. Inflammatory bone marrow signaling in pediatric acute myeloid leukemia distinguishes patients with poor outcomes. Nat Commun 13, 7186 (2022). https://doi.org/10.1038/s41467-022-34965-4

Download citation

Received: 12 February 2022
Accepted: 09 November 2022
Published: 23 November 2022
DOI: https://doi.org/10.1038/s41467-022-34965-4

This article is cited by

Alternative splicing and genetic variation of mhc-e: implications for rhesus cytomegalovirus-based vaccines
- Hayden Brochu
- Ruihan Wang
- Xinxia Peng
Communications Biology (2022)

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.