Molecular mutations may influence cytokine expression and critical illness subphenotyping in septic patients with Acute Myeloid Leukemia

Abstract Background: In acute myeloid leukemia (AML), mutations in IDH1/2, RAS-family, FLT3, and TP53 have all been associated with unique inflammatory changes in the tumor microenvironment that promote leukemogenesis and therapeutic resistance. However, the effects of these mutations on systemic inflammatory response to severe infection remain uncharacterized. We have previously shown that patients with any acute leukemia who develop sepsis exhibit a “hyperinflammatory subphenotype” characterized by elevated plasma cytokines, which may respond favorably to subphenotype-directed ICU interventions [Ronner et al. CCM 2025]. We hypothesized that cytokine levels, subphenotype expression, likelihood of respiratory and renal failure, and mortality would vary between molecularly defined AML subgroups. Methods Molecular Epidemiology of SepsiS in the ICU (MESSI) is an active prospective cohort study recruiting participants admitted to medical ICUs at the Hospital of the University of Pennsylvania. Patients must have a suspected infection and at least one organ failure to qualify for enrollment. Measurement of key inflammatory cytokines (IL-1RA, IL-1β, IL-6, IL-8, IL-10, sTNFR1) is performed by electrochemiluminescence on day-of-ICU-admission plasma samples. We evaluated a sub-cohort of MESSI patients with newly diagnosed or relapsed/refractory AML admitted between 2013 and 2019 (n = 107) and compared against a cohort of patients admitted during this same time without hematologic malignancy (n = 659). Using standardized mean difference (SMD) considering values ≥ 0.2 as indicative of meaningful difference, we compared individual cytokine levels, inflammatory subphenotype as assigned by a validated parsimonious algorithm, and major outcomes across AML with mutations in TET2, K/NRAS, FLT3, IDH1/2, TP53, or AML with none of these (“AML-NOS”). Multivariable logistic regression models controlling for age, sex, APACHE III, and baseline comorbidities were constructed to test associations between subgroups and likelihood of hyperinflammatory subphenotype expression, development of acute respiratory distress syndrome (ARDS), development of acute kidney injury (AKI), or death at 30 days. Results Compared to patients without hematologic malignancy, septic AML patients had meaningful elevations in all cytokines except sTNFR1. Consistent with prior findings, AML patients were more likely to require pressors on admission, develop ARDS, and express a hyperinflammatory subphenotype of critical illness. Using AML-NOS as our reference, we observed meaningful differences in cytokine expression profiles across all mutated groups. Low IL-10 was a feature of TET2 (median 9.5 vs 34.5 pg/mL, SMD 0.301) and FLT3 AML (11.32 vs. 34.5 pg/mL, SMD 0.418), while substantially low IL-10 distinguished the TP53 cytokine signature (8.8 vs 34.5 pg/mL, SMD 0.575). The RAS subgroup exhibited lower IL-1RA (5134.9 vs. 6295.1 pg/mL, SMD 0.405) and higher sTNFR1 (9689.9 vs. 7088.0 pg/mL, SMD 0.428). IDH1/2 AML exhibited higher levels of all cytokines except IL-10, with marked elevations in IL-1RA (9454.6 vs. 6295.1 pg/mL, SMD 0.460) and IL-8 (1691.7 vs. 668.1 pg/mL, SMD 0.422). The RAS, IDH1/2, and TP53 subgroups were distinguished by increased likelihood of hyperinflammatory subphenotype expression, incident ARDS, or incident AKI. However, in our logistic regression models only TP53 mutations retained a significant association with incident ARDS (OR 3.84, 95% CI 1.43 – 12.22, p = 0.013), while only RAS mutations retained a significant association with hyperinflammatory phenotype assignment (OR 3.93, CI 1.44 – 12.69, p = 0.012) and 30-day mortality (OR 3.54, CI 1.43 – 9.43, p = 0.008). Conclusion In this cohort of AML patients admitted to the ICU with sepsis, we observe distinct patterns of inflammatory cytokine expression, particularly in patients with IDH1/2, RAS family, and TP53 mutations. These three specific subgroups demonstrated a tendency towards hyperinflammatory subphenotype expression and incident respiratory and renal failure. Our analysis was limited by modest sample size; adjusted analyses are underpowered. Overall, these results suggest that septic immune responses may be influenced by the mutational profile of AML. As subphenotype-directed ICU management is being actively evaluated in clinical trials, these findings could ultimately inform a tailored approach to critical illness in AML patients. Replication of this analysis in a larger cohort is planned.