Sepsis is a dreaded, life-threatening condition that can occur when an infection spins out of control. Like a tsunami after an earthquake, sepsis occurs when an infection triggers a dysregulation of the immune system, which leads to widespread organ damage and even death.
The condition can result from nearly any kind of infection and afflicts tens of millions each year globally. Scientists don’t fully understand how it develops or how best to stop its progression in patients.
A new study by researchers at the Broad Institute of MIT and Harvard, Massachusetts General Hospital, and MIT offers new insight into what goes awry in the immune system during sepsis. By analyzing blood cells from COVID-19 patients who progressed to sepsis, they identified a cell type that impairs the body’s ability to detect and respond to the virus.
The work helps explain how the immune system is suppressed in severe cases of COVID-19 and other infections, and suggests that targeting these cells could one day improve the outcome for patients. The study appears in Science Translational Medicine.
“Sepsis is a major problem around the world, and no one knows how to correct it,” said Nir Hacohen, co-senior author on the new study, institute member at the Broad, and director of the Mass General Center for Cancer Immunotherapy and Broad’s Cell Circuits Program.
“The response to infection is complex, but work like this helps us to fill in the knowledge gaps and uncover new possibilities for treatment.”
In earlier work, the team analyzed individual immune cells in the blood of patients who had bacterial urinary tract infections, some of which resulted in sepsis. Patients with moderate urinary tract infections harbored a type of blood cell, dubbed MS1, which was even more abundant in septic patients. The MS1 cells were nearly absent from the blood of healthy control patients.
The researchers suspected that MS1 cells might also contribute to immune suppression during viral infections.
To discover what drives the production of the cells, they added blood plasma from patients with sepsis or severe COVID-19 to healthy bone marrow cells in a dish, and found that this combination generated new MS1 cells.
This indicated that some factors in the blood of severely infected patients – likely secreted immune molecules known as cytokines that were also identified in this study – drive the production of MS1 cells.
They next explored how MS1 cells alter the two main functions of the immune system: innate sensing of pathogens and the adaptive immune responses to them. The presence of MS1 cells generated from bone marrow prevented immune cells from sensing viral material.
In addition, activated T immune cells divided more slowly in the presence of MS1 cells, indicating a more sluggish adaptive immune response to pathogens. The results suggest that MS1 cells strongly suppress the immune system in severe cases of COVID-19, sepsis, and many other infections.
“Studying cells in bulk can only get us so far. By analyzing individual cells from patients in carefully-designed clinical studies, and following up with functional studies, we can uncover unique cell types with big effects on physiology, such as the MS1 cells,” said co-senior author Paul Blainey, core member of the Broad and associate professor in the Department of Biological Engineering at MIT.
The abundance of MS1 cells in the blood could potentially be useful as a clinical test to assess the severity and prognosis for patients undergoing treatment for sepsis. In addition, the researchers anticipate that future therapies to manipulate MS1 cells to change their influence on the immune response could improve the outcome for patients with sepsis, whether due to infection with COVID-19 or another pathogen.
epsis is a prevalent disease with high mortality that contributes to a large fraction of healthcare spending worldwide1. To date, no diagnostic biomarker nor targeted therapeutic agent for sepsis has proven useful or effective. This is likely attributable to substantial heterogeneity of disease due to multiple potential pathogens, sites of infection, individualized host immune responses and manifestations of organ dysfunction2–4.
Equally, there is limited insight into the cellular and molecular basis of sepsis-induced systemic immune dysregulation5–8. Prior host gene expression profiling studies relied on whole blood to characterize diagnostic or prognostic gene signatures9–12, an approach that aggregates transcriptomic signals from many different cell types, but may not detect signatures from rarer cells and does not identify cell type-specific disease signatures13. To overcome these limitations, we characterized the spectrum of immune cell states in the blood of septic patients using single-cell-resolved gene expression profiling.
scRNA-seq defines immune cell states in sepsis patients across multiple clinical cohorts
We performed scRNA-seq on PBMCs from septic patients and controls to define the range of cell states present in these patients, identify differences in cell state composition between groups, and detect immune signatures that distinguish sepsis from the normal immune response to bacterial infection (Figure 1). Our primary cohorts targeted patients with urinary tract infection (UTI) early in their disease course, within 12 hours of presentation to the Emergency Department (ED) (Figure 1b–e, Supplementary Table 1).
UTI was selected to minimize heterogeneity introduced by different infectious sites and maximize diagnostic clarity, since UTI can be reliably confirmed post hoc by urine culture. We included patients with UTI (clinical urinalysis with >20 WBCs per high-power field) as the primary infection both with and without signs of sepsis, and subsequently adjudicated the enrolled patients into UTI with leukocytosis (blood WBC ≥ 12,000 per mm3) but no organ dysfunction (Leuk-UTI), UTI with mild or transient organ dysfunction (Int-URO), and UTI with clear or persistent Methods); organ dysfunction (Urosepsis, URO) (patients with simple UTI without leukocytosis or signs of organ dysfunction were not enrolled. Our schema distinguishes transient versus sustained sepsis-related organ dysfunction, although both meet established criteria (Sepsis-2 criteria) for sepsis14.
Cohort definition and analysis strategy.
(a) Processing pipeline for blood samples used in this study. Total CD45+ PBMCs and enriched dendritic cells for groups of patients were labelled with cell hashing antibodies and loaded on a droplet-based scRNA-seq platform. Cells were demultiplexed and multiplets were removed based on calls for each barcoding antibody. (b) Schematic and number of patients for each cohort profiled in this study. (c) Age distribution of patients and controls analyzed in this study. (d) Time to enrollment from hospital presentation for each patient across all cohorts. Boxes show the mean and interquartile range (IQR) for each patient cohort, with whiskers extending to 1.5 IQR in either direction from the top or bottom quartile. (e) Barplots showing fractions of Gram-positive and Gram-negative pathogens for each cohort. (f) Analysis pipeline: cell states were identified via two-step clustering, and fractional abundances thereof were compared to find sepsis-specific states. Further signatures were derived from these states using differential gene expression and gene module analysis. These signatures were validated in external sepsis datasets via a combination of bulk gene expression deconvolution, direct mapping of gene signatures, and meta-analysis. Experiments were performed to identify surface markers, develop a model system for induction, analyze the epigenomic profile, and characterize the functional phenotype of the identified cell state.
We also profiled patients from two secondary cohorts from a different hospital: bacteremic patients with sepsis in hospital wards (Bac-SEP) and patients admitted to the medical intensive care unit (ICU) either with sepsis (ICU-SEP) or without sepsis (ICU-NoSEP). Inclusion criteria were the same for primary and secondary cohorts. These secondary cohorts included patients later in their disease course, enrolled at least 24 hours after initial hospital presentation and receipt of intravenous antibiotics. For comparison, we analyzed specimens from uninfected, healthy controls (Control). Our multi-cohort approach, spanning two hospitals and several clinical phenotypes, supports the generalizability of our results across different clinical contexts.
We profiled total CD45+ PBMCs (1,000–1,500 cells per patient) and LIN-, CD14-, HLA-DR+ dendritic cells (300–500 cells per patient) using a 3’ RNA tag sequencing approach. We multiplexed 6–8 samples per experiment using cell hashing, observing no major batch effects in our data (Methods, Extended Data Fig. 1).
We identified immune cell states by clustering the cells in two steps: low-resolution clustering to identify the major immune cell types (Figure 1f, Extended Data Fig. 2a–b, Extended Data Fig. 3), then sub-clustering each major cell type separately in a robust manner (Extended Data Fig. 2c–d, Methods). This approach identified 16 cell states that are found across numerous patients (n=31–69 per state) in different cohorts and processing batches (Figure 2a, Extended Data Fig. 2e–f).
Among these are transcriptional states of T, B, NK, and dendritic cells15–17, and importantly, four monocyte states (Extended Data Fig. 4,,5).5). We found four distinct monocyte groups: MS1, CD14+ cells characterized by high expression of RETN, ALOX5AP, and IL1R2 (Figure 2b); MS2, characterized by high expression of class II MHC; MS3, similar to non-classical CD16hi monocytes; and MS4, which is composed of the remaining CD14+ cells that express low levels of both class II MHC and inflammatory cytokines. We noted that some marker genes characterizing the MS1 state (Supplementary Table 2) had been previously associated with sepsis in studies measuring either serum protein or whole blood mRNA levels18–21.
scRNA-seq identifies sepsis-specific immune cell states and gene signatures.
(a) t-distributed stochastic neighbor embedding (tSNE) plots for each cell type (n = 32341, 7970, 9390, 58557, and 14299, cells for T, B, NK, Mono, and DC, respectively) colored by embedding density of cells from sepsis patients (Int-URO, URO, Bac-SEP and ICU-SEP; left) and cell state (right). (b) Select marker genes that are differentially expressed (false discovery rate [FDR] < 0.05, two-tailed Wilcoxon rank-sum test) in each cell state, when compared with other cell states within the same cell type. Color scale corresponds to z-scored, log-transformed mean gene expression counts for each cell state. TS, T cell states; BS, B cell states; NS, NK cell states; MS, monocyte states; DS, dendritic cell states; MK, megakaryocytes. (c) Fraction of total CD45+ cells across each patient type for total monocytes (left) and MS1 cells (right). In Control group, points for healthy controls that were follow-up samples from enrolled Leuk-UTI, Int-URO, and URO patients are indicated as black symbols and those for matched healthy control samples from an outside source are indicated as aqua symbols. FDR values are shown when comparing each disease state with healthy controls (two-tailed Wilcoxon rank-sum test, corrected for testing of multiple states). Boxes show the median and IQR for each patient cohort, with whiskers extending to 1.5 IQR in either direction from the top or bottom quartile. Sample size (n) for each cohort is indicated in Figure 1b. (d) Volcano plot showing differential expression analysis results (two-sided Wilcoxon rank-sum test) between MS1 cells from ICU-SEP and MS1 cells from ICU-NoSEP patients. Genes with log2 FC (fold-change) > 1 are highlighted in red, and the top 5 genes with the highest positive fold-changes are labeled. n = 2,153 and 1,442 cells from the 8 ICU-SEP and 7 ICU-NoSEP patients, respectively. (e) Box and swarm plots showing the mean expression (log2 unique molecular identifier [UMI] counts) of PLAC8 and CLU in MS1 cells for each patient from the ICU-SEP and ICU-NoSEP cohorts. Boxes show the median and IQR for each patient cohort, with whiskers extending to 1.5 IQR in either direction from the top or bottom quartile. (f-g) Scatterplots showing correlation between mean gene module usage in MS1 cells and SOFA scores for Int-URO and URO patients. Line and shadow indicate linear regression fit and 95% confidence interval, respectively. Significance of the correlations (Pearson r) were calculated with a two-sided permutation test, corrected for testing of multiple modules. SOFA, sequential organ failure assessment.
Expansion of a monocyte state, MS1, in the blood of sepsis patients
We analyzed the differences in abundances of these cell states across different patient phenotypes (Figure 1f). We found that the fractional abundances of cell states in the blood are strongly associated with the disease status of an individual (Extended Data Fig. 6a–b), whereas absolute abundances are less so (Extended Data Fig. 6c). The fractions of classical cell types vary significantly among the Control, Leuk-UTI, and sepsis (Int-URO, URO, Bac-SEP, and ICU-SEP) cohorts; however, this variation is less pronounced than that of cell states derived from our clustering (e.g., monocytes versus MS1, Figure 2c). Of note, MS1 cells constitute a significantly larger fraction of CD45+ cells in Int-URO and URO patients than in Control or Leuk-UTI patients and are also enriched in septic patients in our secondary cohorts (Bac-SEP and ICU-SEP vs. Control, FDR < 0.001). MS1 cells are also present at a slightly higher fraction in septic patients (Int-URO, URO, Bac-SEP, and ICU-SEP) than severely ill patients without infection (ICU-NoSEP, FDR = 0.27).
Given the expansion of MS1 in septic patients, we reasoned that analysis of gene expression signatures within MS1 cells may reveal useful clinical markers for sepsis and further insight into biological mechanisms. To find signatures that discriminate sepsis from critical illness without bacterial infection not distinguished by cell state abundance alone, we performed differential expression analysis specifically on MS1 cells from ICU-SEP and ICU-NoSEP patients (Figure 2d) and found two genes, PLAC8 and CLU, that discriminate these two patient populations (Figure 2e, Extended Data Fig. 7c). Whereas PLAC8 expression has been associated with sepsis in studies analyzing the bulk expression of blood cells22, CLU expression has not, perhaps due to its specific up-regulation in MS1 cells.
We analyzed co-varying genes among MS1 cells using non-negative matrix factorization and found five gene modules that are detected in more than half of our sepsis patients (Extended Data Fig. 7d–f, Supplementary Table 3). Of note, the module in MS1 cells corresponding to mitochondrial respiration (MS1-A; MT-ND4, MT-CO3, MT-ATP6) correlated significantly with disease severity in sepsis patients from our primary cohort (Int-URO and URO, FDR = 0.03; Figure 2f), supporting the link between alterations in energy metabolism and immunoparalysis in sepsis23.
In addition, a module of genes in MS1 related to anti-inflammatory and pro-resolving responses (MS1-B; S100A8, RETN, ALOX5AP, FPR2)24–26 correlates negatively with severity (FDR = 0.04) (Figure 2g, Extended Data Fig. 7g), consistent with a current model of sepsis wherein patients early in their disease have a heightened inflammatory state, but subsequently switch to an immunosuppressive state27.
reference link : https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7235950/
More information: Miguel Reyes et al, Plasma from patients with bacterial sepsis or severe COVID-19 induces production of suppressive myeloid cells from human hematopoietic progenitor cells in vitro, Science Translational Medicine (2021). DOI: 10.1126/scitranslmed.abe9599
