Myeloid-Derived Suppressor-like Cells as a Prognostic Marker in Critically Ill Patients: Insights from Experimental Endotoxemia and Intensive Care Patients

Patients admitted to the intensive care unit (ICU) often experience endotoxemia, nosocomial infections and sepsis. Polymorphonuclear and monocytic myeloid-derived suppressor cells (PMN-MDSCs and M-MDSCs) can have an important impact on the development of infectious diseases, but little is known about their potential predictive value in critically ill patients. Here, we used unsupervised flow cytometry analyses to quantify MDSC-like cells in healthy subjects challenged with endotoxin and in critically ill patients admitted to intensive care units and at risk of developing infections. Cells phenotypically similar to PMN-MDSCs and M-MDSCs increased after endotoxin challenge. Similar cells were elevated in patients at ICU admission and normalized at ICU discharge. A subpopulation of M-MDSC-like cells expressing intermediate levels of CD15 (CD15int M-MDSCs) was associated with overall mortality (p = 0.02). Interestingly, the high abundance of PMN-MDSCs and CD15int M-MDSCs was a good predictor of mortality (p = 0.0046 and 0.014), with area under the ROC curve for mortality of 0.70 (95% CI = 0.4–1.0) and 0.86 (0.62–1.0), respectively. Overall, our observations support the idea that MDSCs represent biomarkers for sepsis and that flow cytometry monitoring of MDSCs may be used to risk-stratify ICU patients for targeted therapy.


Introduction
Sepsis is a heterogeneous syndrome defined as a life-threatening organ dysfunction caused by a dysregulated host response to infection [1].Despite major improvements in patient care, the incidence of sepsis is rising.The Global Burden of Disease Study estimated that sepsis affects around 49 million people and is accountable for 11 million deaths per year, representing close to 20% of all deaths worldwide.Moreover, almost half of patients surviving sepsis are re-hospitalized within a year.Hence, sepsis is a leading cause of critical illness and mortality worldwide [2].Sepsis is accompanied by immune alterations affecting Cells 2024, 13, 314 3 of 16

Measurement of Serum Mediators by Multiplex Bead Assay
The concentrations of IL-1RA, IL-6, IL-8, IL-10, CCL2, CCL3, CCL4, and TNF in the serum from healthy volunteers challenged with LPS were quantified using Luminex technology (Luminex Corporation, Austin, TX, USA).The concentrations of 45 mediators in the serum samples obtained from patients at their admission to the ICU (PIPOVAP study) were determined by the clinical laboratory of the Service of Immunology and Allergy, Lausanne University Hospital, using Luminex xMAP Technology as described [34].

Statistical Analysis
The data generated during this study, as well as detailed patient's demographics, are available in Supplementary Table S1.Baseline comparisons were performed using the chi-square exact test, Mann-Whitney U test, and Kruskal-Wallis test, as appropriate.The relationship between the percentages of cell population and clinical data (i.e., categorial variables) was assessed using the Mann-Whitney U test.The expression levels of PMN-MDSCs ≤ 10% and >10% of leukocytes and of CD15 int M-MDSCs ≤ 1.3% and >1.3% of leukocytes (cutoff value based on tertile) were categorized as high and low levels and were used to analyze mortality over time.Survival curves were compared using the log-rank test.Luminex data were analyzed using Spearman's rank correlation controlling for FDR (false discovery rate) using the Benjamini and Hochberg procedure.Mediators with a coefficient of correlation greater than 0.3 with a population of MDSCs are reported.Statistics and illustrations were achieved employing R v.3.6.0 (R Foundation for Statistical Computing, Vienna, Austria).Each dot represents an individual sample.*, p < 0.05; **, p ≤ 0.01; ***, p ≤ 0.001; ****, p ≤ 0.0001.

MDSC-like Cells in Healthy Subjects Challenged with Endotoxin
We used an experimental model of endotoxemia to delineate the impact of systemic inflammation on the expression of MDSCs since endotoxemia is a common feature of ICU patients.Eight healthy male volunteers (median age 23.5 years, interquartile range [IQR]: [22][23][24][25][26][27]) were challenged with 2 ng/kg of LPS (Table 1).Blood was collected before and up to 336 h after endotoxin administration (Figure 1A).leukocytes.PMN-MDSCs returned to baseline levels after 24 h (Figure 1D).
As expected, TNF peaked after 1 h, while IL-6, IL-8, CCL2, CCL3, CCL4 reached their maximum values after 2 h, and IL-1-RA and IL-10 reached their maximum after 3 h of endotoxin challenge (Figure 1F).So, proinflammatory cytokine response preceded MDSC accumulation in peripheral blood.The cytokines returned to baseline levels after 4-8 h, except for CCL2 and IL-1RA, which returned to baseline levels between 8 and 336 h.Overall, endotoxin administration induced a massive and transient accumulation of MDSC-like cells in the circulation, accompanied by elevated levels of cytokines.

MDSC-like Cells in Mechanically Ventilated ICU Patients
We set up a clinical study to test whether MDSC-like cells represent useful biomarkers in critically ill patients with a high probability of being infected during hospitalization.Thirty-three uninfected mechanically ventilated ICU patients (median age 65 [52-68] years) were studied (Figure 2A).The median ICU stay was 8.0 [6.0-14.0]days.Twenty-five patients (75.7%) developed an infection during their ICU stay (time to onset: 1 [0-2] day), consisting of ventilator-associated pneumonia (VAP) in fifteen patients (45.4%) and a non-VAP in ten patients (30.3%).Four patients (12.5%) died (days till death: 7.0 [5.5-9.3]days) (Table 1).Survivors and non-survivors had similar acute physiology and chronic health evaluation II (APACHE II) and sequential organ failure assessment (SOFA) scores (Table 1).The CRP values were high in ICU patients when compared to healthy subjects (normal values: <5 mg/L).In contrast, lactate levels in patients were close to normal values (<1 mol/L), suggesting that patients were not in acute circulatory distress/shock.Indeed, patients were selected without infection and with a likelihood of surviving the first 24 h.The blood was collected at the time of ICU admission in all patients and at ICU discharge in 17 patients.Flow cytometry analyses identified basophils, eosinophils, neutrophils, classical monocytes, intermediate/non-classical monocytes, lin + cells and DCs, as well as cell populations phenotypically reminiscent of PMN-MDSCs and M-MDSCs, which were further sub-clustered into CD15 intermediate (CD15 int ) and CD15 low M-MDSCs (Figure 2B,C).Counts and frequencies of neutrophils and monocytes did not differ between admission and discharge from the ICU (Figure 2D and Supplementary Figure S3).PMN-MDSCs decreased from 0.24 [0.05-1.6]× 10 6 cells/mL at ICU admission to 0.05 [0.01-0.13]× 10 6 cells/mL at discharge (p = 0.008), while M-MDSCs decreased from 0.13 [0.07-0.21] to 0.04 [0.02-0.11]× 10 6 cells/mL (p = 0.007) (Figure 2D and Supplementary Figure S3).The results were similar when the analyses were performed on paired samples from survivors.Among M-MDSCs, CD15 int M-MDSCs were more abundant than CD15 low M-MDSCs at ICU admission (0.059 versus 0.036 × 10 6 cells/mL, p > 0.05).Both populations decreased at discharge (0.024 versus 0.018 × 10 6 cells/mL).
PMN-MDSC and M-MDSC numbers at ICU admission were significantly higher than those measured in healthy subjects before the endotoxin challenge (12 and 6.5-fold, p = 0.0016 and p = 0.00016), indicating that the deterioration in underlying conditions was associated with higher levels of MDSCs.Yet, PMN-MDSCs at ICU admission were 9.1-fold lower than 8 h after endotoxin challenge (p = 0.019), while M-MDSC counts were comparable under the same conditions.This confirmed the power of LPS to stimulate the rise of MDSCs, particularly PMN-MDSCs.PMN-MDSC and M-MDSC levels at ICU discharge were comparable to those of healthy subjects (p = 0.17 and p = 0.67) (Supplementary Figure S4A,B).

MDSC-like Cells in Relation with Cytokines and Growth Factors Levels in ICU Patients
We performed correlation studies between cytokines/growth factors and MDSCs in blood collected at ICU admission (Figure 3).We quantified 23

MDSC-like Cells and Nosocomial Infections in ICU Patients
PMN-MDSCs and M-MDSCs at ICU admission (percentages and absolute counts) were not related to the development of hospital-acquired infection (HAI) and VAP (Figure 4A and Supplementary Figure S5A).Neither were they associated with the time to develop infection and the occurrence of sepsis or septic shock.The absence of an association between MDSCs and infection could reflect that MDSCs were elevated prior to ICU admission, while the time to develop infection was short.Furthermore, no other cell population (basophils, eosinophils, neutrophils, monocytes, monocytes, and lin + cells and DCs) was related to the development of infections.

Discussion
We report that MDSC-like cells rise sharply and transiently in the blood of healthy subjects challenged with endotoxin.In uninfected mechanically ventilated patients admitted to the ICU, the levels of CD15 int M-MDSCs correlated with the development of Gram-negative bacterial infections, whereas elevated levels of PMN-MDSCs or CD15 int M-MDSCs were associated with mortality.Therefore, MDSCs (or MDSC-like cells) may represent biomarkers for critically ill and infected patients [15,36].
The abundance of MDSCs in healthy subjects and in sepsis patients differed noticeably between studies [15,16,[19][20][21]23,[36][37][38][39][40].This is partly because the quantification of MDSCs via flow cytometry is influenced by factors including the type of sample (whole blood, freshly isolated or frozen PBMCs) and the immunophenotyping strategy [13,15,23,36,40].It is also possible that MDSCs progress phenotypically during sepsis.In our study, we used phenotypical standards [13].To minimize analytical variations, we collected whole blood in tubes containing lyophilized antibodies and used clustering of flow cytometry data to identify MDSCs.An advantage of whole blood is that it enables MDSCs and neutrophils to be quantified in a single sample.However, this is at the expense of using the low-density characteristic of PMN-MDSCs for discrimination purposes.While functional suppressive activity is the gold standard for defining MDSCs [11,12,17], phenotypical evaluation via flow cytometry is preferred in routine practice, and more so on whole blood than on density gradient-purified PBMCs.To improve immunophenotyping, the panel of antibodies could be enriched.For example, we could target lectin-type oxidized LDL receptor 1, which is expressed by human PMN-MDSCs in cancer patients and septic shock and severe coronavirus disease 2019 (COVID-19) patients [41,42].Another candidate is CD300ld, which was discovered in an in vivo CRISPR-Cas9 screen and was shown to be increased in PMN-MDSCs involved in tumor progression in mouse models, as well as in human tissues from colon and lung cancers [43].
In sepsis patients, single-cell RNAseq analyses of MDSCs have shown specific transcriptomes [22] but have failed to devise phenotypic markers, allowing for the unambiguous identification of MDSCs via flow cytometry.Furthermore, in-depth explorations of neutrophilic cells (i.e., including cells that should be PMN-MDSCs) via single-cell RNAseq suggested that neutrophils transit through states rather than durable subsets that are more easily traceable and that functional signatures transcend specific subsets of activation [44,45].A recent whole-blood single-cell multiomic atlas in sepsis patients identified immunosuppressive CD66b + neutrophils and features of emergency granulopoiesis with a higher frequency of immature neutrophils in a subgroup of patients with poor outcomes [46].Altogether, the most recent single-cell analyses indicate that it may be difficult to identify universal, stably expressed, phenotypical markers to track MDSCs.
Human experimental endotoxemia is the only model available to study the impact of a systemic inflammatory response in humans in a well-controlled setting.M-MDSCs showed an initial drop before a progressive increase, which reflects the transient monocytopenia observed in endotoxemia [47].Even considering this initial decrease, globally, MDSCs (re)increased quite rapidly after the LPS challenge.This is in line with mouse studies showing that MDSCs increased in blood, spleen, and liver 3 to 12 h after intraperitoneal or the intratracheal administration of LPS [48][49][50].In models of chronic sepsis induced by cecal ligation and puncture, MDSCs increased more gradually over several days [51,52], suggesting that the inflammatory burst has a deep impact on the generation of MDSCs.Indeed, the administration of LPS increased cytokines and chemokines that stimulate myelopoiesis and the development of MDSCs.Endotoxemia is common in ICU patients suffering from trauma, abdominal and cardiovascular surgery, as well as COVID-19 and bacterial sepsis [26,53].Thus, endotoxemia may contribute to an increase in MDSCs in uninfected and infected critically ill patients.
Our analytical strategy identified CD15 int and CD15 low M-MDSCs subgroups in ICU patients.CD15, known as Lewis X antigen, is expressed by granulocytes but also monocytes, macrophages, eosinophils, mast cells, and myeloid precursor cells.Interestingly, a study in the late 1990s described a subpopulation of whole-blood monocytes that resembles M-MDSCs with high SSC parameters and the robust expression of CD15 and ROS [54].Clustering did not identify the CD15 int M-MDSCs subpopulation in the endotoxin study, possibly due to the limited number of individuals analyzed or because this subpopulation is expressed more specifically in critically ill patients.Indeed, two recent studies applying unsupervised clustering of PBMCs and semi-automated analysis of whole blood in sepsis patients reported a dichotomous distribution of CD15 expression by what appears to be M-MDSCs [15,36].A so-called unconventional CD15 + CD11b + CD14 + CD33 + CD66b + HLA-DR − subset increased early during sepsis and returned to physiological levels in survivors [15].Likewise, we observed that CD15 int M-MDSCs were elevated at ICU admission compared to ICU discharge.The subdivision of M-MDSCs based on CD15 might be useful in stratifying critically care patients.Unfortunately, due to the design of our study, we were not able to cell sort and further characterize this subpopulation of M-MDSCs.
MDSCs increase in patients was probably initiated before their admission to the ICU.If so, we possibly did not capture the full range of mediators relevant to MDSC expansion in the blood collected at ICU admission.It would explain why only one association was statistically significant.The sample size combined with the large number of comparisons (45 mediators and four cell populations) is another factor that reduced sensitivity.Though not statistically significant after correction for multiple testing, positive associations involving IL6, CCL3, CCL4, CCL11, CXCL12, and HGF would be consistent with the impact these mediators have on the generation of MDSCs.For example, IL-6 plays a critical role in regulating the accumulation and activation of MDSCs.Moreover, MDSCs have been reported to express receptors for IL-6 (IL-6R), CCL3, CCL4, CCL11 (CCR2/3/5), CXCL8 (CXCR1/2), CXCL12 (CXCR4/7), and HGF (HGFR/Met receptor).Unfortunately, we were not able to quantify the expression of these receptors.MDSC levels did not correlate with nosocomial infections, which contrasts with other reports [19][20][21]23].However, high levels of M-MDSCs, particularly CD15 int MDSCs, correlated with the occurrence of Gram-negative bacterial infection.Additionally, M-MDSCs increased 10-fold during endotoxemia, and high levels of M-MDSCs have been associated with Gram-negative bacterial sepsis [19,37].Of note, on a per-cell basis, M-MDSCs are more potent immunosuppressive than PMN-MDSCs.Hence, a minor subpopulation of MDSCs may have a significant pathophysiological impact.Remarkably, patients who did not survive presented to the ICU with high levels of PMN-MDSCs and CD15 int M-MDSCs, while clinical severity scores were slightly lower compared to survivors.The stratification of the patients into those with high and low expression of MDSCs and ROC curve analyses revealed a good discriminative value of MDSCs.Accordingly, high levels of PMN-MDSCs and/or M-MDSCs at study inclusion correlated with mortality in patients with sepsis and COVID-19 and in non-surgical ICU patients.In addition, MDSCs play a role in establishing or maintaining a protracted immunosuppressive environment, contributing to chronic critical illness, secondary infections, and long-term morbidity and mortality in ICU patients [16,[19][20][21][22]33,39].Overall, the data suggest that MDSCs could be biomarkers for stratifying patients and selecting those who might benefit from targeted therapy.Preclinical studies have shown promising results with strategies to reduce MDSCs levels and increase T-cell function during sepsis, while phase II clinical trials targeting MDSCs in oncologic patients are in progress.
Our study has several limitations.We identified MDSCs through phenotypical and non-functional analyses.However, the immunosuppressive function of MDSCs isolated based on the expression of cell surface markers (through magnetic or flow cytometry cell sorting) has been reported in numerous studies [19,38].Of note, quantification of MDSC-like cells in blood via flow cytometry is currently the best option for clinical development, while it is not conceivable to introduce functional tests for immunosuppressive functions of MDSCs in routine.We used a stain/fix/freeze procedure before samples were analyzed.Although unusual, this method, developed by others (see, for example, [55,56]), permits the acquisition of all samples in a row, thus limiting the impact of possible fluctuations in instrument performance.The endotoxemia study was performed with males.The harmonization of the study group was intended, but it may be considered suboptimal as there is sexual dimorphism in the host response to microbial products.One should keep in mind that MDSCs are barely detected in the blood of healthy subjects.Therefore, it might be problematic to define a normal range in healthy people and challenging to stratify via flow cytometry ICU patients for targeted therapy.The limited sample size and the quantification of mediators at a given time point restricted the sensitivity of detecting associations between MDSCs and biological parameters in ICU patients.In addition, the predictive model of outcomes is based on a small number of patients who did not survive.Thus, larger cohorts should be studied to confirm that MDSC levels might be used to stratify patients, and to validate the association between elevated levels of PMN-MDSCs and/or CD15 int M-MDSCs with mortality and to characterize the functions of CD15 int M-MDSCs.

Conclusions
In conclusion, we show that MDSC-like cells were highly responsive to endotoxin challenge and that elevated levels of PMN-MDSCs and CD15 int M-MDSCs in blood correlated with the development of Gram-negative bacterial nosocomial infections and patient outcomes.Our observations support the idea that MDSCs may be used as biomarkers contributing to mortality prognosis and risk stratification of ICU patients for targeted therapy.

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/cells13040314/s1, Figure S1 S1: Data generated during this study.Table S1 contains all data and analyses, as well as detailed patient's demographics.Pages 1-4 are related to the endotoxin study, while pages 5-8 are related to the PIPOVAP study.

Figure 1 .
Figure 1.MDSC-like cells in the blood of healthy subjects challenged with endotoxin.(A) Study design.Eight healthy subjects were challenged with 2 ng/kg endotoxin.Blood was collected just before and 1-336 h after endotoxin infusion in DURAClone tubes and analyzed via flow cytometry.

Figure 1 .
Figure 1.MDSC-like cells in the blood of healthy subjects challenged with endotoxin.(A) Study design.Eight healthy subjects were challenged with 2 ng/kg endotoxin.Blood was collected just before and 1-336 h after endotoxin infusion in DURAClone tubes and analyzed via flow cytometry.(B) Relative expression levels of cell surface markers and forward and side scatter areas (FSC-A, SSC-A) of leukocyte populations.(C) t-SNE plots of leukocyte populations over time.(D) Absolute counts and percentage in leukocytes of PMN-MDSCs and neutrophils.(E) Absolute counts and percentage in monocytic cells of M-MDSCs, classical monocytes and intermediate/non-classical monocytes.(F) Concentrations of cytokines and chemokines.Graphs show medians with standard deviations.DCs: dendritic cells, Lin+: lineage (i.e., CD3, CD19 or CD56)-positive.* Values above the upper limit of quantification (IL-1RA at 3 and 4 h, panel (F)).

16 Figure 2 .
Figure 2. MDSC-like cells in the blood of mechanically ventilated ICU patients.(A) Study design.Blood was collected at ICU admission (n = 32) and at ICU discharge (n = 17) from mechanically ventilated ICU patients without infection and analyzed as described in Figure 1.(B) t-SNE plot of leukocyte populations.(C) Relative expression levels of cell surface markers and forward and side scatter areas (FSC-A, SSC-A) of leukocyte populations.(D) Absolute counts (millions of cells per mL) of neutrophils, lineage-positive (lin+) cells and dendritic cells (DCs), polymorphonuclear MDSCs (PMN-MDSCs), classical monocytes, intermediate/non-classical (inter/NC) monocytes, monocytic MDSCs (M-MDSCs), and CD15 low and CD15 intermediate (CD15 int ) M-MDSCs.Boxplots show medians and upper and lower quartiles, whiskers show the 5 to 95 percentiles, and dots show the individual values.

Figure 2 .
Figure 2. MDSC-like cells in the blood of mechanically ventilated ICU patients.(A) Study design.Blood was collected at ICU admission (n = 32) and at ICU discharge (n = 17) from mechanically ventilated ICU

16 Figure 3 .
Figure 3. Correlation matrix between MDSC-like cells, cytokines, chemokines and growth factors.Blood was collected from patients on admission to the ICU to quantify cytokines, chemokines and growth factors (see Materials and Methods).Correlations were calculated using Spearman's rank correlation, which was controlled for false discovery rate using the Benjamini and Hochberg stepup procedure.The correlation plot represents mediators with a correlation coefficient ≥ 0.3 or ≤ −0.3 with at least one population of MDSCs.* p < 0.05, ** p ≤ 0.01, *** p ≤ 0.001.The color scale ranges from blue for negative correlation to red for positive one.The size of the dots is proportional to the p value.

Figure 3 .
Figure 3. Correlation matrix between MDSC-like cells, cytokines, chemokines and growth factors.Blood was collected from patients on admission to the ICU to quantify cytokines, chemokines and

Figure 4 .
Figure 4. Proportions of MDSC-like cells in patients who developed or did not develop infections during their ICU stay.(A) PMN-MDSCs and M-MDSCs in patients who did (n = 25) or did not (n = 8) develop an infection.Infections were sorted into hospital-acquired infection (HAI, n = 10) and ventilator-associated pneumonia (VAP, n = 15).(B) PMN-MDSCs, M-MDSCs, CD15int M-MDSCs, and CD15low M-MDSCs at ICU admission in patients who subsequently developed Gram-negative or Gram-positive bacterial infections.Boxplots show medians and upper and lower quartiles, whiskers show 5 to 95 percentiles, and dots show individual values to the p value.

Figure 5 .Figure 5 .
Figure 5. MDSC-like cells at ICU admission in relation to patient's outcome.(A) Leukocyte populations in ICU survivors (n = 28) and non-survivors (n = 4).(B) CD15 int and CD15 low MDSCs in survivors and non-survivors.Boxplots show medians and upper and lower quartiles, whiskers 5 to 95 percentiles, and dots individual values.The association of MDSCs with survival was assessed using the Mann-Whitney U and the Kruskal-Wallis tests.(C) Kaplan-Meier survival curves based on low and high levels of PMN-MDSCs (≤ 10% and >10% of leukocytes).(D) Kaplan-Meier survival curves based on low and high levels of CD15 int M-MDSCs (≤1.3% and >1.3% of leukocytes).(E) Kaplan-Meier survival curves based on low and high levels of PMN-MDSCs and/or CD15 int M-MDSCs.(F) Receiver operating characteristic (ROC) curves of PMN-MDSCs and CD15 int M-MDSCs for mortalint : Gating strategy of flow cytometry data to exclude debris, doublets and CD45-non-hematopoietic cells (A), univariate histogram plots of cell surface marker expression by whole blood cells from the endotoxemia study (B) and ICU study (C) study, and back-gating of samples from ICU study (D); Figure S2: Leukocyte counts in the blood of healthy subjects infused with endotoxin; Figure S3: MDSC-like cells in non-infectious mechanically ventilated ICU patients; Figure S4: Comparison of MDSC-like cells levels between healthy subjects challenged with endotoxin and ICU patients; Figure S5: Absolute counts of MDSC-like cells in relation with the development of infection in mechanically ventilated ICU patients; Table

Table 1 .
Characteristic of healthy volunteers and patients.