1 

 
Masters by Research (M.Sc.) 

2021-2023 

 
Project Title: Cell landscape of immunosuppressive reprogramming during fatal 

sepsis and severe-COVID-19 through scRNA-SEQ 

 

 
by: Alexis Garduno 

Student ID: 18338856 

 
 
Discipline of Clinical Medicine, School of Medicine 
Trinity Centre for Health Sciences, 
St James's Hospital  
 
 
Primary Supervisor:  
 
Prof. Ignacio Martin-Loeches (Consultant in Intensive Care Medicine at St James's Hospital 
and Clinical Full Professor at Trinity College, Dublin, Ireland) 
 
Funding:  
This work was primarily supported by the SFI (Science Foundation Ireland), Grant Number: 
20/COV/0038.  
 
 
 
 
 
 
 
 
 
 
 



 
 

 

 

2 

X
Alexis Garduno

Declaration concerning plagiarism 
 
I declare that this report has not been submitted as an exercise for a degree at this or any other 
university and it is entirely my own work. I have read and I understand the plagiarism provisions in 
the General Regulations of the University Calendar for the current year, found at: 
http://www.tcd.ie/calendar 
 
I have also completed the Online Tutorial on avoiding plagiarism ‘Ready, Steady, Write’, located at 
http://tcd-ie.libguides.com/plagiarism/ready-steady-write 
 
 
 
Date: October 15, 2023 
 
 

 
 
 
 

 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 

 
 
 

Signature: 

http://www.tcd.ie/calendar
http://tcd-ie.libguides.com/plagiarism/ready-steady-write


 
 

 

 

3 

Abstract  
 

It is evident that some patients with sepsis and septic shock are admitted to hospitals late in their 
illness, contributing to poor outcomes and high mortality rates across all age groups worldwide. 
Current diagnostic and monitoring procedures often rely on delayed and inaccurate clinical 
identification, leading to suboptimal treatment decisions. Following the initial "cytokine storm," 
sepsis is frequently associated with immune system paralysis, highlighting the need for precise 
immunological subtyping for therapy. In sepsis, the immune system is activated to produce 
interleukins, and endothelial cells express increased levels of adhesion molecules. These processes 
lead to changes in circulating immune cell proportions, including a decrease in regulatory cells and 
an increase in memory and cytotoxic T cells. These alterations have lasting effects on CD8 T cell 
phenotype, mHLA-DR expression, and microRNA dysregulation. 

Aim: To investigate the pathophysiology of sepsis (bacterial and viral) and the immunosuppressive 
reprogramming of immune cells using an integrated immunological and -omics model approach. 
 
Methods: Peripheral blood mononuclear cells (PBMCs) were isolated using the Sepmate procedure. 
Single-cell 3′ cDNA libraries were generated using the BD Rhapsody protocol. Our cohort from St. 
James's Hospital included 16 participants, encompassing both mild and severe cases of viral and 
bacterial sepsis. Libraries were sequenced with 150 bp paired-end reads on the Illumina HiSeq 
platform, achieving an average of 180 million reads per sample. Subsequent analyses included 
publicly available datasets to assess PBMC reprogramming in sepsis and COVID-19 compared to 
healthy donors, controlling for batch effects and the single-center study design. 

Findings: Critically ill COVID-19 patients exhibited a distinct immunosuppressive endotype, leading 
to inflammatory dysregulation even in moderate disease forms. Significant differences in CD8 naive 
T cells, CD4 naive T cells, and CD4 memory T cells were observed between COVID-19 and mild 
bacterial sepsis. However, severe infections (critical COVID-19 and bacterial septic shock) displayed 
shared immune patterns, including upregulated transcriptome profiles in B cells, classical 
monocytes, CD4 and CD8 naive T cells, and natural killer cells. Genes associated with myeloid-
derived suppressor cell (MDSC) proliferative and immunosuppressive functions, such as 
CYBB/NOX2, S100A8/9, and RETN, were enriched in sepsis datasets. Transcriptomic similarities 
between the sepsis cohort and healthy controls were more pronounced at post-sepsis day 21, 
compared to those in septic shock. Cell cycle signature analysis from an in-silico atlas (11 datasets) 
highlighted the predominance of blood leukocytes (B, T, and NK cells) in COVID-19 severity and 
disease trajectory. 

Conclusion: Significant gene expression differences were identified based on infection etiology in 
mild cases (COVID-19 or bacterial sepsis). However, severe infections (critical COVID-19 and 
bacterial septic shock) shared immune profiles related to both adaptive and innate responses, 
irrespective of the cause. These findings may support the implementation of co-adjuvant therapies 
and interventions aimed at preventing the severe disease forms associated with high global 
mortality rates. 

 
 
 
Keywords: single cell transcriptomics (scRNA-SEQ); critically ill patients; SARS-CoV-
2; sepsis; septic shock; mild disease; peripheral blood mononuclear cells (PBMCs); 
immunosuppression; endotypes; immunomodulation therapies, cell reprogramming  

 
 



 
 

 

 

4 

 

Objectives:  
 

1. Utilize scRNA-seq pipelines to characterize the dysregulated immune response 
resulting from aberrant activation of innate and adaptive immunity, stratified by 
severity in both critically ill COVID-19 (viral sepsis) and bacterial-induced sepsis. 

2. Assess the relative proportions of immune cell reprogramming among PBMCs in 
sepsis and COVID-19 compared to healthy donor groups using public reference 
datasets to validate findings from our internal dataset. 

3. Identify immunosuppressive cell state trajectories and late-stage reprogramming of 
MDSCs using Monocle 2, which may indicate treatment response and correlate with 
the worst clinical outcomes. 

 
 

 

Hypothesis:  

H1: In both mild and severe COVID-19, there will be an increase in neutrophil transcriptional 
signatures and a reduction in non-classical monocyte reprogramming. Systemic cytokines 
may contribute to enhanced myelopoiesis during severe bacterial sepsis and COVID-19. 

H2: Unlike in cancer and other inflammatory diseases, myeloid-derived suppressor cells 
(MDSCs) will retain immunosuppressive transcriptional signatures throughout late-stage 
sepsis, characterized by persistent proliferation and functional activation. 

H3: In severe COVID-19, there will be persistent dysfunction in distinguishing NK cell subtypes, 
with strong associations to disease stage, the most critical clinical conditions, and treatment 
regimens, compared to mild disease groups. 

 
 
 
 
 

 

 

 

 

 

 

 

 
 
 
 
 



 
 

 

 

5 

Research questions:  
 

1. Can an integrative transcriptomic analytics approach enhance our understanding of 
immunosuppressive cell reprogramming during critically ill COVID-19 (viral sepsis) 
and bacterial-induced sepsis, in comparison to mild disease states? Furthermore, how 
does this reprogramming relate to the worst clinical outcomes, treatment regimens, 
respiratory support needs, and overall patient prognosis? 
 

2. How can the identification of early and late cell reprogramming trajectories inform 
immunomodulatory strategies and optimize the realignment of host immunity in the 
treatment of severe bacterial sepsis and COVID-19? 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 



 
 

 

 

6 

DECLARATIONS  
 
I declare that, unless otherwise stated, all work presented in this thesis is my own. Several 
aspects of the study relied upon collaboration where part of the work was conducted with or by 
others. 
 
Patient enrollment in the “Sepsis Immunosuppression in Critically Ill Patients” study, and 
sample pre-processing was performed by Alexis Garduno at St James Hospital. Written 
informed consent from patients/and or assent from their next-if-kin was also obtained.  
 
BD Rhapsody scRNA-SEQ: PBMC isolation, Targeted mRNA and AbSeq PCR Amplification Kit, 
library preparation, and analysis was performed by Alexis Garduno (Trinity College Dublin). 
Computational pipelines, and QC metrics were performed by post doc, Gustavo Sganzerla 
Martinez (Dalhousie University).  
 
Public reference in-silico scRNA-SEQ atlas analysis) performed by Alexis Garduno (Trinity 
College Dublin).  
 
Other preliminary experiments: Luminex MagPix platform: Human Sepsis Magnetic Bead 
Panel (96-Well Plate Assay) experiments were performed by Alexis Garduno (Trinity College 
Dublin). 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 



 
 

 

 

7 

ASSOCIATED PUBLICATIONS  
 

 
Garduno, A.; Cusack, R.; Leone, M.; Einav, S.; Martin-Loeches, I. Multi-Omics Endotypes in ICU 
Sepsis-Induced Immunosuppression. Microorganisms 2023, 11, 1119. 
https://doi.org/10.3390/microorganisms11051119      
 
Garduno, A.; Martinez, G.S.; Ostadgavahi, A.T.; Kelvin, D.; Cusack, R.; Martin-Loeches, I. Parallel 
Dysregulated Immune Response in Severe Forms of COVID-19 and Bacterial Sepsis via Single-
Cell Transcriptome Sequencing. Biomedicines 2023, 11, 778. 
https://doi.org/10.3390/biomedicines11030778 

 
Martinez G, Garduno A, Mahmud-Al-Rafat A, Ostadgavahi AT, Avery A, e Silva SD, Cusack R, 
Cameron C, Cameron M, Martin-Loeches I, Kelvin D. An artificial neural network classification 
method employing longitudinally monitored immune biomarkers to predict the clinical 
outcome of critically ill COVID-19 patients. PeerJ. 2022 Dec 12;10:e14487. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

https://doi.org/10.3390/microorganisms11051119
https://doi.org/10.3390/biomedicines11030778


 
 

 

 

8 

CONTENTS  
 
 
Statement of Plagiarism        2 
Abstract (Specific aims and objectives)      3 
Objectives and Hypothesis        4 
Research questions        5 
Declarations          6 
Associated Publications        7 
 
Contents          8 
Abbreviations          9-10 
Introduction and Literature review       11-23 
Materials, Methods, and Techniques learned     24-40 
Findings          41-51 
Discussion         52-53 
Conclusion         54 
References          55-60 
 
Appendices          61 
 

• Garduno, A.; Cusack, R.; Leone, M.; Einav, S.; Martin-Loeches, I. Multi-Omics Endotypes 
in ICU Sepsis-Induced Immunosuppression. Microorganisms 2023, 11, 1119. 
https://doi.org/10.3390/microorganisms11051119      
 

• Garduno, A.; Martinez, G.S.; Ostadgavahi, A.T.; Kelvin, D.; Cusack, R.; Martin-Loeches, I. 
Parallel Dysregulated Immune Response in Severe Forms of COVID-19 and Bacterial 
Sepsis via Single-Cell Transcriptome Sequencing. Biomedicines 2023, 11, 778. 
https://doi.org/10.3390/biomedicines11030778 
 

• Martinez G, Garduno A, Mahmud-Al-Rafat A, Ostadgavahi AT, Avery A, e Silva SD, 
Cusack R, Cameron C, Cameron M, Martin-Loeches I, Kelvin D. An artificial neural 
network classification method employing longitudinally monitored immune 
biomarkers to predict the clinical outcome of critically ill COVID-19 patients. PeerJ. 
2022 Dec 12;10:e14487. 
 

 
 
 

 
 
 
 
 
 
 
 
 
 
 

https://doi.org/10.3390/microorganisms11051119
https://doi.org/10.3390/biomedicines11030778


 
 

 

 

9 

Abbreviations  
 
Ab   Antibody 
Abseq   BD Rhapsody cell surface protein sequencing 
APACHE II  Acute Physiology and Chronic Health Evaluation  
AKI  Acute Kidney Injury 
ADT   Antibody-derived tags 
APC   Antigen presenting cell 
ARDS   Acute respiratory distress syndrome 
ASGR1  Asialoglycoprotein receptor 1  
AST  aspartate aminotransferase 
GBTM  group-based trajectory modeling  
CI   Confidence interval 
cMono   Classical monocyte 
CCI  Charlson Comorbidity Index 
CITE-seq  Cellular indexing of transcriptomes and epitopes by sequencing  
CRP   C-reactive protein 
CRRT  continuous renal replacement therapy 
DAMP   Damage associated molecular pattern 
DA   Differential abundance 
DC   Dendritic cell 
DE   Differential expression 
DEG   Differentially expressed genes 
DGE   Differential gene expression 
EDTA   Ethylenediaminetetraacetic acid 
GEX   Gene expression 
GO   Gene Ontology 
GOBP   Gene Ontology biological processes 
GWAS   Genome wide association study 
FDR   False discovery rate 
FC   Fold change 
HD  Hemodialysis  
HLA   Human leukocyte antigen 
ICU   Intensive care unit 
IL   Interleukin 
IRF   Interferon regulatory factor 
IVIG  intravenous immunoglobulin  
LFC/LogFC  Log fold change 
MALS  macrophage activation-like syndrome  
MDSC   Myeloid derived suppressor cell 
MHC   Major histocompatibility complex 
MOF  Multi-organ failure  
ncMono  non-classical monocyte 
DE-NRGs necroptosis-related genes  
NK cell  Natural killer cell 
PBMC   Peripheral blood mononuclear cell 
PAMP  Pathogen associated molecular pattern 
PCA  Principal component analysis 
PCR  Polymerase chain reaction 
qSOFA  Quick sequential organ failure assessment  
QC  Quality control 
RNA   Ribonucleic acid 
RNA-SEQ  RNA-sequencing 



 
 

 

 

10 

ROC  receiver operating characteristic 
RRT  renal replacement therapy 
SARS-CoV-2  severe acute respiratory syndrome coronavirus 2  
scRNA-SEQ  Single-cell RNA-sequencing 
SICM  Sepsis Induced Cardiomyopathy 
SIRS   Systemic Inflammatory Response Syndrome  
SOFA   Sequential organ failure assessment 
SRS  Sepsis response signature 
TF  Transcription Factor  
TLR   Toll-like receptor  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 



 
 

 

 

11 

1. Introduction and Literature review  
 

 
About 2% of all hospitalized patients develop sepsis [1-3], making it a major public 

health concern throughout Europe and the rest of the world. More than 25 million people are 
diagnosed with sepsis each year, despite international treatment standards. Repeated 
hospitalizations during sepsis indicate the existence of severe immunosuppression and are 
significantly associated with poor clinical outcomes. Primary sepsis readmissions and chronic 
critical illness (CCI) after severe sepsis account for more U.S. healthcare expenditures than any 
other medical cause [4]. For decades, the precise mechanism behind sepsis-induced 
immunodeficiency has remained unknown. 

The diagnosis of sepsis relies heavily on the detection of living bacteria and the 
subsequent identification of the organism responsible for infection. Existing screening methods 
have little sensitivity, and the findings are not available promptly. The risk of rapid resistance 
evolution due to a single mutation in a target determining gene is high. This calls for a de-
escalation strategy predicated on empirical targeted treatment, sensitivity to antibiotics, and 
host response to infection, in addition to the discovery of new chemical scaffolds or binding 
sites that can be targeted. Culture, susceptibility profile, organ-specific infection goal 
attainment, and clinical stability should determine optimal loading dosage antibiotic duration.  

The idea that sepsis is fatal due to unregulated inflammation was reevaluated after 
more than 120 phase II and Phase III studies of anti-inflammatory and anti-cytokine therapies 
failed. There were almost one hundred times lower levels of TNF- and IL-1 in the blood of septic 
patients compared to endotoxin models, according to studies. Despite successful pre-clinical 
testing, there are not many FDA-approved immunologic therapeutic therapies for sepsis used 
in trials that have shown a survival benefit [5-6]. 

When doctors are confronted with sepsis, it is frequently unknown what is causing the 
infection. Current methods for identifying and characterizing microorganisms necessitate 
waiting 2 to 6 days for pathogen detection. This is a critical issue because patient mortality 
increases by 7.8% every hour for the first 6 hours if no intervention is provided. By the time a 
patient receives their diagnosis, they may have already suffered tissue damage, organ failure, 
and death. As shown in (Figure 1), infection causes an initial cytokine-mediated host 
inflammatory response characterized by an interaction between pattern recognition receptors, 
PAMPs, and DAMPs that is associated with immune cell reprogramming, transcriptomic 
changes, epigenetic modification, and metabolic dysfunction. In individuals with sepsis, 
inflammatory responses can be severe, resulting in tissue damage, endothelial cell dysfunction, 
and organ failure. Many pathways have been shown to participate in myocardial dysfunction 
caused by sepsis in the heart, for example. Complements and DAMP can contribute to 
mitochondrial malfunction in the presence of ER stress, resulting in calcium excess and 
electrophysiological abnormalities in the heart [7]. 

By comparing the levels of cytokines, immune modulators, and other endothelial 
mediators between a control group and critically ill non-septic patients, septic patients, and 
septic shock patients, Beltrán-García et al., indicated that at early stages neutrophils and other 
components of the innate immune system aid in the fight against infection in the early stages 
of sepsis. Septic and septic shock patients, on the other hand, were shown to have an impaired 
adaptive immune system. Both pro-coagulation and immunosuppressive responses evident in 
septic shock patients. The levels of IL-6 and pyroptosis-related cytokines (IL-18 and IL-1) were 
highest in individuals with septic shock. Moreover, septic shock has been shown to reduce 
endothelial function because of an increase in the production of adhesion molecules such s-
ICAM1 and E-Selectin. 
Antibiotic treatment initiated within the first few hours has been shown to significantly 
improve prognosis. Even more challenging is the detection and treatment of sepsis, as 
expected complication rates are 32.7%, with 'failure to rescue' - FTR - (risk of death after a 
curable complication) estimated at 6.8% only in post-surgical patients [1,8]. Frequent 
contamination (>50%) leads to confusion and inaccurate therapy, as it is difficult to determine 



 
 

 

 

12 

whether a positive blood culture is due to real bacteremia or a false positive induced by 
contamination. 

It is difficult to treat sepsis because of (i) a lack of patient trajectory assessment 
systems for a personalized transcriptomic-driven approach to treatment, (ii) paucity of host 
diagnostic indicators for antimicrobial-resistant gram-negative bacteria (iii) prolonged time to 
diagnosis, and (iv) inconsistent immune eligibility and outcome criteria. To better understand 
the mechanisms of sepsis, its phenotyping, and patient trajectories, it is important to integrate 
multi-level (epidemiological, genetic, metabolic, systemic, and cellular level) and multi-modal 
data (medications, vital signs, demographics, biomarkers, and other clinical elements) data 
using state-of-the-art machine-learning solutions. Both pro- and anti-inflammatory insults can 
be mitigated by blocking certain inflammatory regulatory mechanisms as seen with the reduced 
Spns2/S1P signaling that has been indicated to contribute to immunosuppression in the late 
stages of sepsis. 

 
Fig. 1 Schematic diagram of immune hemostasis imbalance in sepsis highlighting infection triggers 
leading to initial cytokine-mediated host inflammatory response and reprogramming of transcriptional 
changes. Created with BioRender.com. 
 

 
1.1 Defining sepsis standard care and causative pathogens  
 

For decades, microorganisms that cause infection were thought to have a vendetta against 
us, causing high mortality from uncontrolled prolonged inflammation, as observed in the case 
of sepsis. In critically ill patients, accumulating evidence suggests that it is our response to their 
presence that causes multiorgan dysfunction syndrome (MODS) and related damage. Our host 
reaction and innate immunity endanger us more than the pathogen harm.  As shown in severe 
COVID-19 patients with acute respiratory distress syndrome (ARDS), innate immunity memory 
has been defined as a functional reprogramming in which our innate immune cells respond 
following encounter with a pathogen or viral stimulus. This response can result in one of two 
outcomes: reprogramming resulting to increased (trained immunity) or decreased (immune 
tolerance) response to a secondary stimulus [9]. 

Sepsis was originally described as "an immune system gone haywire" in writings that 
focused on the idea of uncontrolled inflammation [10]. Group A strep (a flesh-eating bacteria), 

https://biorender.com/


 
 

 

 

13 

toxic shock syndrome, and meningococcemia are all clinical examples. To gain an upper hand, 
pathogens try to weaken critical functions of the host immune system. With a present 
predisposition of Gram-positive bacteria and an increased clinical and epidemiological 
significance of fungal sepsis, the frequency of identified microorganisms in sepsis/septic 
shock has varied throughout time. Staphylococcus aureus and Streptococcus pneumoniae are 
the most regularly isolated pathogens among Gram-positive bacteria, whereas Escherichia coli, 
Klebsiella, and Pseudomonas spp. are the most frequently identified pathogens among Gram-
negative bacteria. Candida spp., which is commonly seen in immunosuppressed or neoplastic 
patients undergoing long-term treatment with chemotherapeutic and immunosuppressive 
medications, plays the most significant role among fungal infections associated with the 
disease [10]. The lungs and airways account for 43% of sepsis infections, followed by the 
urinary tract at 16%, the abdominal cavity at 14%, the head at 14% (in cases of FUO), and other 
sites/causes at 13% [11]. 

The Children's Hospital Affiliated to Zhengzhou University's NICU admissions' 
microbiological data was mined from the Laboratory of Microbiology's database from January 
1, 2015, to December 31, 2022. Two distinct forms of neonatal sepsis: early-onset sepsis (EOS), 
which develops in the first 72 hours of life, and late-onset sepsis (LOS), which develops later 
were assessed at measures of disease progression. Results identified a total of 679 bacterial 
strains found in 631 newborns, with 543 coming from the blood and 136 from the cerebrospinal 
fluid. Of the 121 strains identified in EOS, the most common were CoNS (33.88%), Klebsiella 
pneumoniae (23.97%), and Escherichia coli (3.26%). There were 67 (55.37%) MDR 
microorganisms in patients with early-onset septicemia. There were 558 different types of 
bacteria and other microbes recovered from LOS patients; the most common were CoNS 
(37.10%), Klebsiella pneumoniae (19.71%), and Escherichia coli (15.05%). There were 332 
(59.50%) MDR microorganisms in patients with late-onset septicemia. MDR was prevalent 
among CoNS (76.21%), CRE-KP (66.91%), and MRSA (33.33%) isolates [12]. 

The study highlighted the urgent need to identify and implement effective preventative and 
therapeutic methods by revealing an alarming incidence of MDR strains isolated from newborn 
sepsis. Treatment options for staphylococcal infections include vancomycin and teicoplanin, 
while colistin can be used for MDR Gram-negative bacteria. Recent research led to the Sepsis-
3 consensus, which defines the systemic syndrome as an imbalanced host immune response 
to an invasive infection that can lead to fatal organ failure. Sequential organ failure assessment 
(SOFA) components have been used to predict in-hospital mortality in ICU patients with sepsis 
and evaluate model performance.  Six factors—respiratory, cardiovascular, hepatic, coagulation, 
renal, and neurological—calculate the SOFA score [13]. Sepsis is diagnosed by an infection-
induced 2-point SOFA score rise. In the absence of hypovolemia, septic shock, a subtype of 
sepsis, is diagnosed when hypotension continues after vasopressors and the patient's blood 
lactate level is greater than 2 mmol/L despite appropriate fluid resuscitation. Whereby in this 
scenario the patient develops vascular-circulatory dysfunction together with profound immune 
response (IR).  Current CV-SOFA total is 4 [14].  

A recent study Pletz et al., evaluating in-hospital mortality of patients with severe blood 
stream infections highlighted that survival rates for sepsis patients have been drastically 
lowered thanks to advances in critical care treatment [15]. The majority of patients who died 
did so after (presumed) effective completion of antibiotic therapy. In patients who were 
followed for an extended period of time after a severe infection, mortality was significantly 
higher in the months following the infection.  

Fluid (crystalloids) replacement and vasoactive agents (e.g., norepinephrine (NE)) are used 
to maintain mean arterial pressure above 65 mmHg and reduce the risk of fluid overload in 
sepsis/septic shock management, which is difficult and involves multiple pathophysiological 
factors. Vasopressin, not epinephrine, should be given with NE to regulate pressure in refractory 
shock. Tidal volume should be decreased from 10 to 6 mL/kg if mechanical ventilation is 
needed. Heparin and glycemic control prevent venous thromboembolism. Proton-pump 
inhibitors, sodium bicarbonate, and other therapies may be used, however their usefulness is 
debated. We can now draw major conclusions about sepsis and septic shock treatment in 
recent years [11].  



 
 

 

 

14 

The role of corticosteroids for sepsis patients is becoming more clear, mainly due to COVID-
19 pandemic data. Meta-analyses show that low-dose corticosteroids reduce hospital stays, 
shock, and organ failure.[16]. Vasopressors and hydrocortisone may impair immunocompetent 
septic patients with a transcriptome profile [17]. There was a correlation between the 
transcriptomic profile at the beginning of septic shock and the response to corticosteroids. 
Corticosteroids significantly increased mortality in people with the immunocompetent SRS2 
endotype compared to those given a placebo. The SRS endotype interacted with treatment 
assignment (hydrocortisone vs. placebo; P = 0.02). Individuals with the SRS2 phenotype had a 
higher risk of dying from taking hydrocortisone (odds ratio = 4.6; 95% confidence range = 1.5-
14.4) [17]. Moreover, these results emphasize that corticosteroids may prolong the lives of 
severe COVID-19 [18] and non-COVID ARDS patients [19].  

Current theories assert immunoparalysis often leads to recurrent infections and "super-
infections” [20]. This has led to the strong rational the modulation of host immunity may lead to 
improved survival in sepsis as illustrated in (Figure 2). It is widely known that sepsis has the 
highest fatality rate in patients with weaker immune systems (elderly, co-morbidities). Immune 
boosting therapeutic strategies may reduce secondary infections and hospital readmissions, 
which are major sources of morbidity and mortality (40 percent of deaths occur after day 30). 
Improving immunity can defend against a wide range of opportunistic infections (MDR bacteria 
and fungi), which are a growing cause of death. Numerous additional hypotheses exist 
regarding the pathogenesis of immunosuppression in sepsis, such as impaired leukocyte 
recruitment and decreased cell surface protein expression.   

About 40% of septic patients and up to 64% of septic shock patients develop acute kidney 
injury (AKI), which is associated with a higher risk of death [184-186]. Absolute indications for 
dialysis (such as severe metabolic acidosis, refractory fluid overload, electrolyte imbalance, and 
uremic consequences) sometimes need RRT in septic AKI. It is still unclear, however, whether 
continuous RRT or IHD is the superior modality for optimum RRT in septic AKI [11]. The timing 
of RRT initiation remains controversial due to conflicting findings from high-quality RCTs and 
meta-analyses. There is no significant difference in overall mortality at 90 days between 
patients who had undergone early vs. delayed RRT, according to the only randomized controlled 
trial (RCT) involving septic patients with AKI to date [11]. The CRTSAKI Study (Continuous RRT 
Timing in Sepsis-Associated AKI in ICU) is currently underway, comparing early vs. delayed RRT 
techniques in terms of the outcomes of patients with septic AKI in the ICU (Chen et al., 2021). 
 

Immune activation, coagulation, and endothelial damage complication features have been 
identified due to the varying onset dates (D0, D3, D7, D14). Clinical consequences include shock, 
renal failure [21], liver dysfunction [22], and ventilator-associated pneumonia (VAP) typically 
attribute to these secondary or super-infections [19]. Late-stage sepsis can cause chronic 
critical illness as a long-term consequence. A mal-adaptive host response leading to prolonged 
inflammation, T cell fatigue, increase of suppressor cell activities, and protein catabolism [23], 
quick recovery [21], and early multi-organ failure (MOF) leading to mortality [24-26]. It has been 
demonstrated that survivors of septic shock have a unique circulating leukocyte transcriptome 
due to this third mechanism. Comparison of total leukocytes in whole blood on day 1 of 
systemic inflammatory response and day 14 from sepsis survivors who rapidly recovered 
revealed differential expression of 185 unique genes [27] despite similar gene expression 
patterns shown in CCI patients and those who survived. 

Some of the genes in our study, such as BLK, BAG6, FOXO4, and ERF, have been 
associated to a negative character trait [27-29]. However, on day 14, an inflammatory response 
and poor prognosis were associated with upregulation of these genes. The genes ATG12, EHD1, 
NACC1, and SLC7A5, which drive autophagy, T cell differentiation, and stem cell self-renewal 
and maintenance, were likewise altered at day 14 following sepsis and CCI [27]. Implications 
for our knowledge of the molecular mechanism behind the rapid recovery of some CCI patients, 
as opposed to others who may need immunomodulation and carefully timed treatment after 
sepsis, are substantial. 

 
1.2 Role of cellular events in the pathophysiology of sepsis and septic shock  



 
 

 

 

15 

 
 

Most sepsis immunity experiments focused on neutrophils and monocytes rather than T 
cells [30]. The CD4 helper T cell conducts the immune system like an orchestra [31]. Oncologic 
"superstars" checkpoint inhibitors that reverse T-cell weariness are in most immunotherapy 
trials, highlighting the necessity of targeting T cells for an effective immune response. T cells 
are so powerful that anti-PD-1 and anti-PD-L1 drugs can treat over 10 different tumor types. 
Interleukin-7 stimulates CD4, CD8, MAIT, and γδ T cells, which is being tested in several cancer 
related trials. This sets the stage for sepsis which can also benefit from a combination drug 
therapy model [31-32].  

A mouse model showed in Chen et al., that late-stage sepsis expressed the long noncoding 
RNA HOTAIRM1 (HOXA transcript antisense RNA myeloid-specific 1). HOTAIRM1 upregulation 
by Notch/Hes1 was necessary for immunosuppression. HOTAIRM1 fatigued T cells by 
increasing PD-1+, regulatory, and PD-L1 T cells. Blocking Notch/Hes1 signaling or HOTAIRM1 
in late sepsis decreased T cell fatigue. Mechanistic investigations further revealed that 
HOTAIRM1 regulates PD-L1 expression in lung alveolar epithelial cells by targeting the 
transcription factor HOXA1 [33]. Hence, suggesting that the 
Notch/Hes1/HOTAIRM1/HOXA1/PD-L1 axis may treat sepsis-induced immunosuppression. 

In the case of T-reg cells, de Lima et al., indicated that at 15 days post-CLP mouse model, 
the Helios, neuropilin 1 and Tnfrsf18 and Pdcd1-expressing TIGIT+ Treg subgroup had 
expanded significantly due to sepsis. Increased vulnerability to a challenge with nosocomial 
bacteria coincided with the rise in TIGIT+ Tregs, showing that these cells are linked to 
immunosuppression in the aftermath of sepsis [34]. The authors found that the TIGIT+ Treg 
fraction did not expand mechanistically when ST2 was deleted during sepsis. In addition, 
recombinant IL-33 administration led to STAT6 and M2 macrophage-dependent expansion of 
TIGIT+ Tregs. These results showed that only TIGIT+ Tregs continue to develop steadily in the 
late stage of sepsis. TIGIT+ Tregs require the IL-33/STAT2/STAT6/M2 macrophage axis in order 
to grow. 

Apoptosis of lymphocytes, proliferation of myeloid-derived suppressor cells and regulatory 
T cells, and T-cell fatigue are all factors that contribute to immunosuppression. New secondary 
infections, typically caused by less virulent bacteria, arise in 30–40% of individuals with 
prolonged sepsis, demonstrating the effect of immunosuppression. [31,35].  The demographics 
of those who don't make it through sepsis are suggestive of the importance of host immunity 
in survival; these include the elderly, who are more likely to succumb to immunosenescence, as 
well as cancer patients, and those with serious coexisting conditions. Patients with chronic 
sepsis have severely suppressed immune systems, as evidenced by the reactivation of several 
latent viruses in 50% of cases.  

MAIT and γδ T cells, which line the bronchial mucosa, are innate-like lymphocytes that 
quickly respond to pathogen invasion by secreting interferon- and interleukin-17, cytokines that 
kill microbes. Amezcua Vesely et al. added CD4 TRM cells to this category. In mice infection 
models, these lung cells, which are generated from effector TH17 cells that release interleukin-
17, killed bacteria, including carbapenem-resistant Klebsiella pneumoniae [36]. Interleukin-7 
was needed to maintain CD4 TRM cells in the lungs. In a mouse model of latent viral infection, 
CD4 T cells that release interleukin-21 were shown to induce the formation of virus-specific CD8 
T cells [37]. 

More recently, Choi et al. used a clinically relevant mouse model and serial patient 
peripheral blood samples to assess mucosa-associated invariant T (MAIT) cell antibacterial 
activity in sepsis. After sublethal cecal ligation and puncture, B6-MAITCAST mice' hepatic and 
splenic MAIT cells expressed CD69 and produced IFNc [38]. Early MAIT cell enumeration in 
septic ICU patients hence may serve as a prognostic metric to use in routine assessment. 
Patients who died were shown to have lower peripheral blood MAIT cell frequencies. MAIT cells 
from sepsis survivors launched higher IFNc responses to many bacterial species upon 
admission to the ICU than those from those who died [38]. It is interesting that while low 
frequencies of human leukocyte antigen (HLA)-DR+ monocytes, a surrogate sign of sepsis-
induced immunosuppression, were restored, MAIT cell numerical insufficiency and CD69 



 
 

 

 

16 

expression continued to fall. Sepsis caused MAIT cells to lose their responses to bacteria, an 
MR1 ligand, IL-12, and IL-18, which did not return by ICU/hospital release. MAIT cell 
dysfunctions may cause immunosuppression after sepsis. 

Patients with sepsis frequently have immunosuppression due to decreased B and T 
lymphocyte counts (a clinical condition known as B and T lymphopenia) [39]. B and T 
lymphopenia in sepsis is caused by extensive lymphocyte apoptosis and blocking this 
apoptosis with caspase inhibitors greatly reduced mortality.  Anergy is a type of immune cell 
tolerance in which cells do not establish a typical innate immunological response (IR) to an 
antigen. There is a prolonged time in which the cells are hyporesponsive and inert [40]. 
Immunodepression is associated with a decrease in T cell activation via TCR signaling or Ca 
(2+) mobilization [41]. A drop in CD4+, CD8+, and total T cell counts has been associated to 
sepsis. A decrease in CD4+, CD8+, and total T cell counts has been linked to sepsis. Given the 
importance of lymphocytes in warding off infection, the correlation between lymphopenia and 
poor sepsis survival is likely more than coincidence, but also a critical pathophysiologic 
process. 

Patients in the intensive care unit were categorized by their changing Absolute 
lymphocyte count (ALC) using group-based trajectory modeling (GBTM). The independent 
correlation of trajectory endotypes with mortality and persistent inflammation, 
immunosuppression, and catabolism syndrome (PICS) was determined using a multivariate cox 
regression model. Based on dynamic ALC, the persistent lymphopenia endotype had the highest 
PICS (24.9%), hospital mortality (14.5%), and 28-day death (10.8%) of the four trajectory 
endotypes. In the multivariate cox regression model, chronic lymphopenia increased the risk of 
28-day death, hospital mortality, and PICS (HR: 1.54; 95% CI: 1.06-2.23, 1.66, 1.20-2.29) [42]. The 
ALC trajectory models of healthy patients and healthy older patients successfully classified 91% 
and 90% of critically unwell patients into the same endotypes, in the sensitivity analysis. The 
ALC trajectory model is a useful tool for categorizing patients in the intensive care unit, and pre-
existing lymphopenia is predictive of a worse prognosis. Notably, lymphopenia that lasts for an 
extended period of time may be an extremely strong indicator of immunosuppression in the 
critically ill [42-43].  

Although the exact mechanisms that cause B-cells to be drastically reduced in adult sepsis 
patients and in septic mice remain unclear, several hypotheses have been put forth. Reduced B-
cell counts may be caused by poor bone marrow production, however there were no changes 
between survivors and non-survivors in the amount of naive B cells or cytokines that inhibited 
or promoted B-cell proliferation [44]. These results rule out the possibility that a malfunction in 
bone marrow formation is to blame for low B-cell counts. Apoptosis, or the death of cells, is 
thought to have a role in the decline of B-cells. Evidence from previous studies [45-48] shows 
that extensive B-cell apoptosis occurs in the blood, intestines, and peripheral lymphoid organs, 
resulting in a progressive profound loss of B lymphocytes in sepsis patients and experimental 
animals. Indeed, memory B cells are more vulnerable to sepsis than naive B cells.  

Their numbers could be lower because of a defect in B-cell maturation. Septic patients' 
secondary lymphoid organs have been shown to have a lower cellular density compared with 
healthy controls [49], and this decrease coincides with a decrease in the number of circulating 
Tfh cells. These findings show that a decrease in B-cell count may arise from a loss of APCs 
and Tfh cells [50-51]. Animal studies have shown that B cells are more vulnerable to Fas-
mediated cell death after being exposed to B-cell activating factor (BAFF) and 
lipopolysaccharide (LPS) than B cells cultivated without BAFF [52]. The substantial elevation of 
BAFF serum levels following LPS challenge has been linked to the loss of memory B cells in 
human endotoxemia models [53]. Although BAFF enhances B-cell maturation under normal 
settings [53], these findings imply that it may be another component leading to B-cell shortage 
in sepsis. The decrease in B-cell numbers and the asymmetry of the B-cell compartment may 
also be attributable to the transfer of circulating B cells into peripheral organs under specific 
conditions. 

These mechanisms and modulation involved in sepsis-induced dysfunctions have distinct 
pathobiological and transcriptomic profiles in pediatric patients, including a decrease in TNF- 
production indicating a decreased innate immune response capacity, which has been identified 



 
 

 

 

17 

in children with nosocomial infection, prolonged MOF, and increased mortality rates [54]. Early 
adaptive immune suppression in children with septic shock has also been linked to decreased 
lymphocyte numbers and a reduced ability of lymphocytes to respond to stimulation with 
infectious complications on Day 0 [55-56]. The discovery of these endotypes prompted more 
recent research into changes in genomic profiles in immunoparalyzed and non-
immunoparalyzed children with sepsis. The researchers discovered 778 differentially expressed 
genes (DEGs) in immunocompromised children with sepsis. These youngsters had higher levels 
of expression of genes that suppress the immune system and lower levels of expression of 
genes involved in regulation and activation, with indications of protein level potential targets. 
Interestingly, the network analysis revealed that Zc3h12a may influence the level or activity of 
the protein products of other highly significant DEGs (CD86, CRTAP, LY86, TGFBI) via an 
interaction with TRAF6, determining its direct participation in the immunoparalysis state [56]. 
Many of the genes in the annotated clusters were found to be functionally connected to HLA-
DR [56]. This finding supports the use of ex vivo TNF- production in identifying a subset of 
children with sepsis who have innate immune suppression, as a decrease in monocyte HLA-DR 
could be an alternate method of defining immunoparalysis in these individuals. STAT3 has been 
demonstrated to downregulate TNF-synthesis by human monocytes during systemic 
inflammation in pediatric patients following cardiopulmonary bypass [56].       

 
 

1.3 Emergence of immunosuppressive Myeloid derived suppressor cells  
 

 
Sepsis causes a ‘reprogramming' of myeloid‐derived suppressor cells (MDSCs) along 

central metabolic pathways. These intricate metabolic shifts are crucial for the maintenance 
and expansion of MDSCs and for the promotion of their immunosuppressive properties. 

The inability of antigen-driven T cells to grow and release pro-inflammatory molecules like 
IL-4 and IFN- was demonstrated by co-culturing MDSCs obtained from patients with sepsis or 
septic shock with T cells [57]. Secondary infections and length of hospital stay are both strongly 
connected with the higher proportion of MDSCs in the peripheral blood of sepsis patients. 
Severe sepsis is associated with an uptick in the granulocytic subtype of MDSCs, which 
produces arginase-1 and is linked to T-cell dysfunction [58]. Similarly, G-MDSCs phenotypic 
changes gradually as their suppressive role increases over time following the onset of sepsis 
[59]. Finally, sepsis patients exhibit induction and accumulation of MDSCs, which promote the 
establishment of the immunosuppressed state of sepsis by suppressing the activity of the host 
immune system. 

In sepsis, researchers have experimented with a number of stimulant techniques in an 
effort to restore immunological homeostasis. It is possible that pathologically activated MDSCs 
are responsible for the lack of improvement in 28-day mortality seen in clinical studies where 
recombinant GM-CSF or G-CSF was used to treat patients with sepsis [60]. Similarly, clinical 
investigations have demonstrated that IL-7 is well-tolerated and has no serious side effects in 
increasing T cell numbers and function in murine sepsis [60-61]. However, therapy with IL-7 has 
been shown in one research to increase and prolong the sepsis-induced expansion of MDSCs 
[61]. In both murine and human sepsis, IFN- production is reduced, making IFN- a promising 
immunomodulatory treatment during sepsis [61]. Restoring IFN- increases survival in mouse 
sepsis [62], and a human study is now recruiting participants to test the effects of IFN- on sepsis 
patients (NCT01649921). 

Several strategies to reduce MDSC numbers and/or suppress their functions have been 
proposed on the basis of available evidence [63]. 

The primary purpose of a recent study Coudereau and colleagues, 2021 was to look at LOX-
1+ MDSC in bacterial and viral sepsis. The results showed that LOX-1+ MDSC were considerably 
growing in both patient categories. The peak number of LOX-1+ MDSC was 1-week delayed after 
ICU admission. COVID-19 levels were higher in patients with acute respiratory distress 
syndrome. Long-term immunosuppression caused by the persistence of these cells raises the 
possibility that infections will not be properly treated [64]. 



 
 

 

 

18 

Most research has been done in hosts with tumors, but this therapeutic approach will be 
effective in other pathological circumstances, such as sepsis, where the inhibition of MDSCs is 
an essential therapeutic aim. Similar immunological abnormalities characterize sepsis and 
malignancy. In late sepsis patients, a unique transcriptomic pattern of multiple immune cell 
subtypes was found. These included B- and CD4+, CD8+, activated CD4+, activated CD8+ T-
lymphocytes, natural killer (NK) cells, NKT cells, and plasmacytoid dendritic cells. Differential 
expression of cytotoxic genes among CD8+ T lymphocytes in late bacterial sepsis [60] is just 
one example of how the transcriptome of circulating lymphoid cells reflects persistent 
immunosuppression and low-grade inflammation. It follows that patients who exhibit a host 
endotype response caused by chronic inflammation may benefit from immunomodulatory 
medication if its administration is preceded by the identification of this transcriptome pattern 
in late sepsis in non-myeloid cell. Darden et al. want to bring attention to the transcriptional and 
clinical significance of myeloid derived suppressor cells in late sepsis. Sepsis was linked to a 
21-day rise in the relative expansion of granulocytic (G-) MDSCs, monocytic (M-) MDSCs, and 
early (E-) MDSCs [61-62]. However, the results showed that CCI depends in part on the initial 
septic insult, and the preliminary data suggests a unique immunosuppressive pathway in late 
sepsis that can be targeted for further research, particularly in patients with cancer-related 
sepsis considering previous research into the subsets and proliferation role of cancer-induced 
MDSCs. 

 
1.4 Diagnostic and prognostic value of biological markers of injury-induced 

immunosuppression  
 

Biochemical and immunological biomarkers have also been studied for their ability to 
distinguish infectious from non-infectious sepsis and their prognostic value. Over this course, 
more than 250 biomarkers have been found and tested, but none of them can reliably 
distinguish between sepsis and sepsis-like condition. Biomarkers have been shown to be useful 
in several research settings, including pathogen detection, clinical diagnosis, and antibiotic 
dosing optimization.  
 

 
1.4.1 Procalcitonin and C-reactive protein 

 
 
Early diagnosis of sepsis and illness determination can be greatly aided by measuring 

inflammatory factor levels, such as calcitoninogen and C-reactive protein (CRP), as has been 
shown in recent clinical investigations [65]. When an organism is infected with bacteria, levels 
of PCT and CRP, two important inflammatory cytokines involved in apoptosis and able to enable 
cell lysis, rise dramatically; however, the significance of changes in their levels has been studied 
less frequently in patients with sepsis due to bloodstream infection.  

The predictive usefulness of procalcitonin (PCT) in septic patients has been the subject 
of extensive research. However, different research has reached different conclusions. 
Heterogeneity in PCT testing duration was strongly indicated (P = 0.020). Patients with sepsis 
had poor initial PCT value predictive significance. PCT non-clearance was a prognostic factor 
of death in patients with sepsis. In a fixed-effects meta-analysis (I2 = 37.9%), the pooled RR was 
3.05 (95% CI, 2.35-3.95). Sensitivity was calculated to be 0.72 (95% CI, 0.58-0.82), specificity 
was calculated to be 0.77 (95% CI, 0.55-0.90), and the area under the SROC curve was 0.79 (95% 
CI, 0.75-0.83) [65].  More research is required to determine the best threshold for PCT non-
clearance and the best definition of risk. 

 Liang and Yu (2022) assessed the clinical added value of CRP, PCT, and NLR in 
predicting severity and correlating outcome of bloodstream infections and sepsis [66]. 
Correlation analysis revealed a positive relationship between CRP, PCT, and NLR and APACHE II 
scores (P 0.05), and the levels of NLR, CRP, PCT, total bilirubin (TBIL), glutamic oxaloacetic 
transaminase (AST), and serum creatinine (Scr) were all significantly higher in the critically ill 
group than in the non-critically ill group. Patients' prognoses were linked to their CRP, PCT, NLR, 



 
 

 

 

19 

and APACHE II levels, according to univariate logistic regression analysis (P 0.05). Patients' PCT, 
NLR, and APACHE II scores were all determined to be independent risk factors for mortality 
within 28 days (P 0.05) in a multi-factor logistic regression analysis. Both PCT and NLR had an 
AUC > 0.7 for predicting patient death within 28 days (P 0.05), as measured by the receiver 
operating characteristic curve [66]. Hence, inflammatory variables such as neutrophil lysis rate, 
C-reactive protein, and procalcitonin have significant therapeutic uses in the evaluation of the 
severity of disease and prognosis in patients with bloodstream infection and sepsis. It has been 
demonstrated that therapeutically using procalcitonin is an effective way to reduce antibiotic 
exposure and enhance clinical outcomes; nevertheless, this approach still has significant 
drawbacks in reference to low protocol adherence, improvements in regular care, and 
insufficient data in particular patient populations. 

Meanwhile, de Jong et al. [67] randomized 1575 critically ill patients to either a 
nonbinding PCT-based algorithm (with a suggestion to stop antimicrobials once the PCT level 
is 80% less than its original value or 0.5 ng/mL) or a standard-of-care approach. PCT instruction 
yielded a significant reduction in daily antibiotic usage and in total mortality [67], despite only 
44% compliance in the PCT arm. There was no increased mortality or bad outcomes linked with 
the adoption of PCT guidelines in patients with sepsis [66-67]. 

Increased MALAT1, STAT3, and PCT gene expression was seen in sepsis patient serum 
and LPS-induced U937 cells, but miR-125b expression was reduced. The MALAT1 transcript 
was found to be mostly nuclear by fluorescence in situ hybridization analysis.  Transfection of 
MALAT1 siRNA into LPS-stimulated U937 cells resulted in a decrease in STAT3 protein 
expression, phosphorylation level, and PCT expression. The results indicated that MALAT1 
might increase STAT3 and PCT expressions through targeted adsorption of miR-125b [68].  
 
 

1.4.2 HLA-DR  
 

Human leukocyte antigen-DR (HLA-DR) is a surface-expressed molecule present on 
immune system monocytes, macrophages, and dendritic cells. The immune system's ability to 
recognize and respond to external invaders such as bacteria and viruses is strongly reliant on 
HLA-DR. Reduced HLA-DR expression has been linked to sepsis and its complications [69], and 
this is thought to suggest immune suppression. Immunological suppression, also known as 
immune paralysis, results in the inability to fight off an infection, as well as the development of 
secondary infections and other issues [69-70]. HLA-DR expression may also be diminished in 
macrophage activation-like syndrome (MALS), a rare but potentially fatal disease characterized 
by an abnormal immune response. This imbalance may result in the overproduction of 
inflammatory cytokines and other immune chemicals [69], which can induce systemic tissue 
damage and organ malfunction. Monitoring HLA-DR expression may be useful in the diagnosis, 
treatment, and management of a range of diseases, including sepsis, MALS, and immunological 
paralysis, since it may provide information about the immune system's ability to respond to 
infections and other stressors. 

Previous research attempted to identify sepsis endotypes by examining how mHLA-DR 
(monocyte HLA-DR) expression varies in the first week after sepsis onset. Approximately two-
thirds of septic patients had either low or declining mHLA-DR expression, while the remaining 
patients had rising mHLA-DR expression. Bodinier et al. [71] discovered that assessing mHLA-
DR expression on days one and three following the onset of sepsis can achieve early risk 
assessment. Clinicians may be able to utilize this data to better identify which patients are most 
vulnerable to difficulties and hence require more intensive care. This study's findings support 
the use of mHLA-DR as a biomarker for early risk stratification and endotype identification in 
sepsis [71], emphasizing the importance of monitoring mHLA-DR expression in septic patients. 
More research is needed to validate these findings and determine the clinical significance of 
mHLA-DR monitoring in sepsis therapy. 

Grondman et al. discovered substantial differences between MALS and the 
immunological paralysis caused by sepsis. CD4 T cell counts were found to be considerably 
lower in sepsis patients who satisfied the immunological paralysis criteria [72]. This has 



 
 

 

 

20 

significant therapeutic implications because it shows that decreased monocytic HLA-DR 
expression is only one component of the pathophysiology of sepsis-induced immunological 
paralysis; it also shows that adaptive immunity is disturbed in these patients. Higher clinical 
scores from the Acute Physiology and Chronic Health Evaluation II (APACHE II), the Sequential 
Organ Failure Assessment (SOFA), and the Charlson Comorbidity Index (CCI) were also linked 
to the identification of a sepsis-specific monocyte cluster and higher mortality in sepsis patients 
[73]. An underlying immunological endotype's predictive value. 

Using an unsupervised clustering method, Leijte et al., analyzed the dynamics of mHLA-
DR expression in 241 patients with septic shock who had varying main sites of infection and 
bacteria. Researchers investigated the correlation between initial infection site and mHLA-DR 
expression levels and found no evidence of a correlation [74]. However, decreased mHLA-DR 
expression was linked to slower overall recovery times. The increased likelihood of unfavorable 
outcome in these patients further supports the idea that variations in mHLA-DR over time are 
clinically relevant in septic shock recovery and innate immune activation [74]. 

More recently, Layios et al., conducted a pilot study where leukocyte subsets were 
phenotyped using flow cytometry and cytokine measures from samples taken at admission (T1) 
and 48-72 hours later (T2) from patients admitted to a tertiary ICU for organ support after severe 
injury (elective heart surgery, trauma, prolonged breathing, or stroke) [75]. At T1, septic patients 
exhibited higher absolute monocyte counts (1030/l against 550/l, p = 0.013) and CRP (5.1 
mg/ml versus 2.5 mg/ml, p = 0.046). They had a higher SOFA score (p 0.0001), lower mHLA-DR 
at T2, and more CD62Lneg monocytes (p = 0.049) [75]. Stepwise logistic regression study 
connected monocyte markers and a high SOFA score (> 8) to nosocomial sepsis. In critically ill 
patients who have sustained an injury, elevated monocyte numbers and a phenotypic shift was 
shown to be linked to the development of secondary sepsis. 

In 1997, IFN- treatment was effective for eight out of nine patients with sepsis; two 
patients reverted after IFN- treatment was stopped. Results showed that IFN- restored TNF- 
production and HLA-DR expression in monocytes in a dose-dependent manner. Research into 
methods of immunological activation increased dramatically as a result of these findings. 
Radboud University Medical Center researchers are currently recruiting an expected 200 
participants for a new multi-center clinical trial with IFN- in patients with candidemia 
(NCT04979052). This study is a follow-up to a 2012 pilot study (NCT01649921) that sought to 
investigate the effects of IFN- therapy in sepsis [76]. 
 

1.4.3 ProADM  
 

The MR-proADM biomarker, or mid-region fragment of pro-adrenomedullin, is largely 
produced by vascular endothelial cells. Adrenomedullin, a strong vasodilator with metabolic and 
immune-modulating properties, has an immediate influence on MR-pro-ADM plasma levels. 
High plasma clearance on day 5 has been associated with improved outcomes [77]; MR-
proADM levels have been observed to grow in sepsis. Furthermore, this biomarker has been 
recognized for its potential application in the early diagnosis of patients at high risk of organ 
malfunction.  

MR-proADM surpassed lactate, PCT, C-reactive protein, and the SOFA score in 
predicting death in recent research [78]. Previous research has looked at the capacity of MR-
proADMs to predict critical care unit admission and the need for rapid therapy [79]. As a result, 
MR-pro-ADM is useful in directing therapeutic decisions for the allocation of ICU and hospital 
resources. 

In a multivariate investigation, MR-proADM was revealed to be a significant predictor of 
mortality from respiratory, coagulation, cardiovascular, neurological, and renal failures. In 
instance, PCT predicted only one additional system (renal), whereas lactate predicted three 
(coagulation, cardiovascular, and neurological). CRP could not predict any of SOFA's basic 
pieces. Detection of cardiovascular disease (AUROC, CI95%): MR-proADM (0.82 [0.76-0.88]), 
PCT (0.81 [0.75-0.87] (P.05), and renal failure (MR-proADM (0.87 [0.82-0.92] (P.05), PCT (0.81 
[0.75-0.86] (P.05)), respectively. With the exception of hepatic failure, MR-proADM was the 



 
 

 

 

21 

biomarker most capable of detecting all other elements of the SOFA score in infected patients 
[80]. 

 
1.4.4 Omentin-1 
 
Omentin-1, also known as intelectin-1, is a new anti-inflammatory adipokine 

associated with inflammation and sepsis. omentin-1 inhibits nuclear factor kappa B (NF-kB) 
and lowers the inflammatory response by activating AMP-activated protein kinase (AMPK), 
which inhibits insulin sensitization, atherosclerosis, and cardioprotection. These functions are 
linked to the role of omentin-1 in regulating energy balance and immunological responses.  

Experiments have shown that Omentin-1 suppresses the NF-kappa B (NF-kB) 
signaling pathway and the production of adhesion molecules including VCAM-1 and ICAM-1 
[81]. Furthermore, omentin-1 inhibits the expression of inflammatory mediators in 
macrophages exposed to lipopolysaccharide (LPS). Notably, omentin-1 plays a role in 
microbial surveillance and detection by acting as a ligand to lactoferrin and bacteria-specific 
carbohydrate residues [81-82]. 

 
An ELISA study which was performed by Kampela et al. revealed that sepsis patients 

and survivors had greater kinetics ((omentin-1) % 39.8 35.9% vs. 20.2 23.3%, p = 0.01 and 39.4 
34.3% vs. 13.3 18.1%, p 0.001) [82]. Greater omentin-1 levels at sepsis onset and 1 week later 
were linked to a higher risk of death (HR: 2.26, 95% CI 1.21-4.19, p = 0.01; HR: 2.15, 95% CI 
1.43-3.22, p 0.001). Finally, omentin-1 highly linked with severity scores, white blood cells, 
coagulation biomarkers, and CRP, but not with procalcitonin or other inflammatory biomarkers 
[82]. This implies that sepsis raises serum omentin-1 levels, and that higher levels and slower 
kinetics during the first week of sepsis increase the risk of death by day 28. In sepsis, 
Omentin-1 may predict early-stage onset functional immunity. Its involvement in sepsis merits 
more investigation. 
 

1.4.5 IL-7 
 

In the absence of interleukin-7 (IL-7), a growth factor produced by the body, 
lymphocytes cannot survive or proliferate. IL-7 increases cell proliferation and inhibits 
lymphocyte death via the JAK/STAT pathway by boosting the anti-apoptotic protein Bcl-2 and 
lowering the pro-apoptotic molecule Bim [83]. IL-7 becomes a modulation marker when both 
CD4 helper T cells and CD8 effector T cells increase. CD4 T cells, often known as "helper" cells, 
coordinate both innate and adaptive immune responses. With the exception of injection site 
responses, IL-7 has been found to be safe in clinical trials including over 500 individuals with 
cancer and infectious disorders [83]. Despite maximal antifungal treatment and debridement, 
an aggressive soft tissue infection caused by Trichosporon asahii, Fusarium, and Saksenaea 
persisted in a non-immunocompromised patient. A polymicrobial necrotizing soft tissue 
infection in the buttocks and perineum spread swiftly after surgery on day 7 [83-84]. 

The infection was not eradicated after forty days of antibiotic treatment and regular 
debridement in the operating room, as well as positive blood and tissue cultures from the 
operating room. Interleukin-7 immunotherapy [85] resulted in clinical response, fungus 
clearance, lymphopenia reversal, and improved T-cell activity. Immunoadjuvant treatment, 
which enhances the host's immune system, may be useful in cases of potentially lethal fungal 
infections. Both blood and tissue cultures were negative after 5 days of IL-7 therapy and 
remained so throughout the hospitalization [85]. 
Miller et al. described a reporter mouse with enhanced GFP expression using the endogenous 
Il7 gene. Both lymphatic endothelial cells (LECs) and fibroblastic reticular cells in the systemic 
lymphatic vasculature release IL-7, and we show that STAT5 phosphorylation is higher in 
lymphatics than in blood [86]. Il7 transcription is elevated in stromal fractions but not in myeloid 
fractions in lymph nodes depleted of lymphocytes. These findings reinforce prior findings that 
proximity to secondary lymphoid organs affects lymphocyte homeostasis and identify LECs as 



 
 

 

 

22 

an important in vivo source of IL-7, which is released by immune cells during both normal and 
lymphopenic stages when they are in transit. 

This, along with prior findings [86], shows that LECs are a significant source of in vivo 
IL-7 and that Il7 mRNA is expressed at high levels in a range of cell types and regions. This 
discovery begs the question of where T-cell homeostasis is most critical. This problem will not 
be solved until a conditional and tissue specific Il7/ mouse is produced. 

In sepsis, extreme lymphopenia is an independent predictor of poor clinical outcomes. 
Interleukin 7 (IL-7) is necessary for lymphocyte growth and maintenance. Intramuscular 
infusion of CYT107, a glycosylated recombinant human IL-7, enhanced lymphocyte activity and 
corrected sepsis-induced lymphopenia in a phase II trial [87]. The absolute number of 
lymphocytes (both CD4+ and CD8+ T cells; both p 0.05) rose two- to thrice after intravenous 
injection of CYT107 compared to placebo. This recovery was equivalent to that seen following 
intramuscular CYT107 administration, was sustained during the follow-up period, reversed 
severe lymphopenia, and was associated with an increase in organ support-free days (OSFD) 
[87]. 
 
1.5 Transcriptomic and Immunological Metrics Guiding Immunotherapy  

 
1.5.1 Nivolumab plus IFN-γ in the treatment of intractable mucormycosis  

 
In a recent clinical case report, an individual who received a diagnosis of fungal sepsis 

had the anticipated immunosuppressive reaction subsequent to a severe incident, as evidenced 
by compromised innate and adaptive immunological capabilities. The infusion of interferon- 
and the monoclonal antibody nivolumab, which specifically targets the programmed cell death 
protein 1 (PD-1) [88], resulted in the restoration of the deficiencies. Interferon- has been utilized 
as a therapeutic intervention for persons suffering from severe fungal infections that have 
demonstrated resistance to conventional treatment methods. The interaction between PD-1 
and its ligands, PD-L1 and PD-L2, is impeded by the binding of nivolumab to PD-1. Consequently, 
the inhibitory effects on T-cell proliferation and cytokine production, which are mediated by the 
PD-1 pathway, are alleviated. The effectiveness of Anti-PD-1 has been demonstrated in animal 
models of fungal peritonitis, as well as in patients afflicted with chronic hepatitis C virus 
infection. Subsequent immunological investigations revealed an increase in the total number of 
lymphocytes, accompanied by an enhanced expression of HLA-DR on monocytes, CD8 T-cells, 
and a decreased expression of PD-1 on T-cells [88]. The patient demonstrated a consistent 
amelioration in their health status, as indicated by successive CT scans revealing the absence 
of any signs of infection. The current evaluation of implementing combination immunotherapy 
as a potential advancement in the management of sepsis is ongoing.    
 
1.5.2 Soluble TREM-1 (sTREM-1), a potential mechanism-based biomarker, might facilitate 

enrichment of patient selection in clinical trials of nangibotide 
 

The phase 2b ASTONISH trial will be the first prospective effort to establish a cut-off 
value for a plasma biomarker of TREM-1 pathway activation in septic shock patients treated 
with nangibotide [89]. The inclusion of material from the scientific literature, observational data 
sets, and experimental clinical trials is required for the determination of both the cut-off value 
and the surrogate endpoint. Aside from its prognostic value in predicting outcome [89-90], 
sTREM-1 also represents the amount of TREM-1 circuit activation. This makes the use of 
sTREM-1 to enrich selection of a disease population getting nangibotide therapy appealing, as 
it identifies not only a high-risk group of patients, but also one that is most likely to respond to 
an anti-TREM-1 strategy [90]. TREM-1-inhibiting therapeutic medicines are still being developed 
and could be utilized to treat sepsis. 
 
 
1.5.3 The PROVIDE randomized clinical trial  

 



 
 

 

 

23 

The PROVIDE randomized clinical trial found four stages of immunoparalysis of sepsis 
using ferritin and HLA-DR. MALS is defined by ferritin levels greater than 4420 ng/mL, whereas 
immunoparalysis is defined by less than 5000 HLA-DR receptors on CD14 monocytes [91]. 
Anakinra treatment for MALS has been demonstrated to improve 7-day outcomes by lowering 
the SOFA score. Patients with MALS had a lower absolute platelet count and a longer activated 
partial thromboplastin time (aPTT), as well as higher levels of aspartate aminotransferase 
(AST), alanine aminotransferase (ALT), and total bilirubin [91]. These characteristics suggested 
a beneficial therapeutic impact in the early stages of sepsis insult, but they did not improve by 
day 28, therefore treatment was discontinued due to a pro-inflammatory effect. Because the 
PROVIDE classification of immunoparalysis immunotypes corresponds to the Mars 1 
endotypes, more studies are needed to optimize T cell function-targeting therapies. 

 
1.5.4 Meta subclassification in pediatric septic shock   

 
Wong et al. investigated the consequences of septic shock heterogeneity for clinical 

trial implementation and the incorporation of gene-based subclassifications in the instance of 
pediatric patients and the validation of gene expression-based subclassifications. Based only 
on a 100-gene expression signature model, the study determined three possible subgroups of 
children with septic shock. Notably, subclass A patients had a more severe disease than 
subclasses B and C [92]. Maximum MOF, fewer ICU stays, and higher PRISM and APACHE 
scores were used to assess illness severity. Downregulation of genes (CD24, CEACAM8, CKS2, 
CX3CR1, DDIT4, EMR3, G0S2, IL8, MAFF, RGS1, TGFBI) associated with adaptive immunity and 
glucocorticoid receptor signaling was also observed in patients in subclass A [92].  

In a similar study, Grunwell et al. looked at the function of transcription and expression 
factors in the development of metabolic anomalies that cause juvenile septic shock sequelae. 
The study discovered 123 Nrf2-linked genes that were upregulated in children with septic shock 
and were predominantly engaged in cellular lipid metabolism pathways [92]. Patients with 
endotype A (higher MOF and mortality) have more Nrf2 associated gene downregulation. The 
findings imply that Nrf2-linked genes may contribute to oxidative stress, which is also 
responsible for mitochondrial dysfunction and sepsis-associated MODS [92]. Targeting Nrf2 
may modify the degree of oxidative signaling and metabolic dysfunction in juvenile septic shock 
patients. 
  These immunologically enriched biomarkers and other values (CCL3, IL8, HSPA1B, 
GZMB, and MMP8) have clinically informed the recently redesigned tPERSEVERE model, which 
incorporates the classification of tree segregated patients into two broad endotypes of septic 
shock characterized by either excessive inflammation or immune suppression to predict 
patterns of inflammation and immunity [92-93]. More importantly, it will help to determine how 
pediatric patients with sepsis and septic shock progress along this range during the course of 
their disease.  

 
 
 
 
 
 
 
 
 



 
 

 

 

24 

2. Materials, Methods, and Techniques learned 
 

2.1 Sepsis immunosuppression in critically ill patients’ study  

 
An Institutional Research Board approval was granted by the SJH/TUH Joint Research 

Ethics Committee and The Health Research Consent Declaration Committee (HRCDC) under 
the register REC: 2020-05 List 17 on 2 March 2020 and Project ID 0428 for the study titled: 
Sepsis Immunosuppression in critically ill patients (Figure 3).  

 
 
Fig. 3 Ethical approval process and amendments  

 

 
 

2.1.1 Patient enrollment and Inclusion criteria  
 
 
We collected 10 mL of whole blood from (n=16) critically ill sepsis and septic shock, severe 
COVID-19 and mild COVID-19 patients from the ward were enrolled in the “Sepsis 
Immunosuppression in Critically Ill Patients” study. After obtaining written informed consent 
from patients or assent from their next-of-kin, the study included a prospective, observational 
cohort from May 2021 to January 2023. All patients admitted to the 26-bed mixed medical and 
surgical intensive care unit (ICU) in St James’s hospital (SJH) were assessed for inclusion in 
the study. 
 
Inclusion Criteria 

▪ Adult patient >18 years of age fulfilling Sepsis 3 criteria.  
▪ Septic shock adult patient > 18 year of age admitted to the ICU requiring a 

vasopressor to maintain a mean arterial pressure of ≥65 mmHg or greater and, 
despite sufficient fluid resuscitation, a serum lactate level of >2 mmol/L. A CV-SOFA 
score of 4 points.  

▪ Fluid responsive 
▪ Patients under mechanical ventilation  
▪ Severe COVID-19 with a positive SARS-CoV-2 nasopharyngeal swab obtained by RT-

PCR, and a status of respiratory failure requiring ventilation. Patients screened for 
ARDS.  

▪ Before patients with treatment at t (0)  
 

Exclusion  



 
 

 

 

25 

▪ The exclusion criteria were the following: a significant history of bone marrow 
disease, any immunosuppressive drugs, bone marrow or solid organ transplant 
recipients, and leucopenia (<1000mm−3).  

▪ 3 patients withdrew consent during the consent to continue process following next-of-
kin assent.  

 
 
Fig. 3 Pre-processing of samples and storage workflow 

 
 

 
2.1.2 Clinical data parameters collected in the study are as follows:   

 



 
 

 

 

26 

 
 
 

Experiment 1: Staining PBMCs with viability markers and counting on the BD Rhapsody single 
cell analysis system. 
 
Following protocol: Ulbrich J, Lopez-Salmeron V, Gerrard I. BD Rhapsody™ Single-Cell Analysis 
System Workflow: From Sample to Multimodal Single-Cell Sequencing Data. InSingle Cell 
Transcriptomics: Methods and Protocols 2022 Dec 11 (pp. 29-56). New York, NY: Springer US 
[94]. 
 

2.2 Analysis of single-cell RNA sequencing data  
 

We used single-cell RNA-sequencing (scRNA-SEQ) to profile peripheral blood 
mononuclear cells (PBMCs) from sixteen patients at St. James Hospital in Dublin, Ireland, to 
identify pathways in peripheral immune cells that may lead to dysregulated inflammatory 
responses or protective immunity in mild versus severe COVID-19 and bacterial induced sepsis. 
Patients with COVID-19, a form of viral sepsis, who were admitted to the intensive care unit 
(ICU) for treatment of acute respiratory distress syndrome (ARDS) were among those recruited, 
as were patients with a milder form of the COVID-19 syndrome who were not hospitalized, as 
well as people with non-severe bacterial sepsis and septic shock. 

The inclusion criteria for individuals diagnosed with severe COVID-19 consisted of 
being at least 18 years old, testing positive for SARS-CoV-2 using a nasopharyngeal swab using 
reverse transcription polymerase chain reaction (RT-PCR), and experiencing respiratory failure 
necessitating the use of mechanical ventilation. The identification of patients with SARS-CoV-2 
who underwent additional phenotyping for acute respiratory distress syndrome (ARDS) in the 
intensive care unit was conducted based on the Berlin criteria. These criteria include the 
presence of sudden hypoxemic respiratory failure, indicated by a PaO2/FIO2 ratio (the ratio of 
arterial oxygen partial pressure to the percentage of inspired oxygen) of less than 300 while 
receiving a minimum of 5 cm of positive end-expiratory pressure, as well as the observation of 
bilateral infiltrates on a chest X-ray. The user has provided a numerical reference without any 
accompanying text or context. Individuals exhibiting a minor manifestation of SARS-CoV-2 



 
 

 

 

27 

infection demonstrated a positive result on the SARS-CoV-2 quantitative reverse transcription 
polymerase chain reaction (qRT-PCR) test. These patients displayed indications of a little lower 
respiratory tract ailment, as evidenced by an oxygen saturation (SpO2) level ranging from 94% 
to 96%. Furthermore, these individuals did not necessitate admission to a hospital ward or the 
administration of vasopressor medications. 

The screening process for septic adult patients who meet the SEPSIS-3 criteria involved 
identifying patients who had received a diagnosis of sepsis within the past 36 hours, as outlined 
in the current SEPSIS-3 definition. This definition requires the presence of a suspected or proven 
infection, as well as an increase in the Sequential Organ Failure Assessment (SOFA) score by 
two points or more. The criteria for identifying patients in septic shock included admission to 
the intensive care unit (ICU) as a result of a bacterial infection, the need for a vasopressor to 
maintain a mean arterial pressure of 65 mmHg or higher, and a serum lactate level over 2 
mmol/L despite adequate fluid resuscitation. The Third International Consensus Definitions for 
Sepsis and Septic Shock [95] indicated that a CV-SOFA score of 4 points was observed.  

The exclusion criteria encompassed the following: individuals with hematological 
malignancy or a notable background of bone marrow disease, those on immunosuppressive 
medications, recipients of bone marrow or solid organ transplants, and individuals with 
leucopenia (<1000 mm−3). The clinical characteristics further elucidated a notable disparity in 
the total duration of hospitalization and an escalation in the duration of intensive care unit (ICU) 
stay among patients with septic shock, as compared to those in the severe COVID-19 and sepsis 
cohort. In general, there was a consistent relationship between in-hospital mortality rates 
among patients diagnosed with sepsis and those diagnosed with severe COVID-19. The study 
observed that the severe COVID-19 group exhibited a much greater proportion of patients 
receiving continuous renal replacement therapy (CRRT) in comparison to the group with 
bacterial-induced sepsis and septic shock. Please refer to Table 1 for further details. 
 

Table 1. Patient clinical characteristics and parameters  

DIAGNOSIS SEPSIS SEPTIC SHOCK SEVERE COVID-
19 

MILD COVID-19 

Age  57.25 65.00 65.25 76.33 

male Sex (%) 75 75 50 66 

Weight (KG) 86.25 100.00 82.35 76.27 

Height (cm) 183.63 159.00 170.50 171.00 

Surgical Admission  0.25 0.50 0.25 0.33 

TEMP (°C) 37.88 39.70 37.30 37.47 

RR (bpm) 29 36 30 22 

GCS 3 3 7 15 

PaO2/FiO2 (worst) 15 17 15 29 

LACTATE (mmol/L) 2.3 2.86 2.43 2.77 

Noradrenaline dose 
day of sample, min-

max (mean), 
mcg/kg/min 

0 − 0.667(0.149) 0.02 − 0.353(0.115) 0 − 0.169(0.031) 0.00 

CRRT (%) 25 25 50 0 

CREATININE 
(mmol/L) 

152.50 182.00 92.50 95.67 

egfr (ml/min/1.73m2) 19.67 33.00 59.50 52.00 

BILIRUBIN (µmol/L) 21 18 12 17 

HB (g/dL) 10.33 7.10 12.35 11.67 

PLT (109/L) 305.00 717.00 346.75 332.00 

WCC (109/L) 28.83 18.60 12.83 8.70 

APTT (seconds) 42.73 30.20 54.50 31.67 

CRP (mg/L) 229.70 341.89 74.35 96.80 

PCT (ng/mL) 0.32 0.48 0.21   

FERRITIN (µg/L) 15207.40   1469.00 763.93 

DDIMER (mg/L 
FEU) 

1378.00   3386.75 780.00 

NEUTS (109/L) 25.43 14.90 10.85 5.17 



 
 

 

 

28 

LYMPHOCYTES 
(109/L) 

0.38 0.80 0.98 0.77 

APACHE 37.00 41.00 21.00 9.00 

SOFA 9.75 12.75 7.00 1.33 

Hospital Length of 
Stay (days) 

124 149 17 31 

Survivors (%) 75 75 75 66 

ICU LOS (days) 24.25 38.5 24.5 0 

 
 
 

 
2.2.1 Details of the Step-by-Step isolation of PBMCs Method  

For all critically ill and non-severe patients with viral and bacterial induced sepsis, blood 
was collected into 10 mL EDTA tubes (Becton, Dickinson and Co., Franklin Lakes, NJ, USA) and 
loaded using SepMate 50 mL tubes (Stemcell Technologies, Saint-Egreve, France). Gradients 
were centrifuged at 1200× g for 10 min with the brake on at room temperature. The cell interface 
was carefully removed by pipetting and washed with PBS-EDTA by centrifugation at 400× g for 
7 min. PBMC pellets were suspended in an ammonium chloride solution (Stemcell 
Technologies, France) and incubated for 10 min at room temperature on a mixing platform to 
lyse contaminating red blood cells. The lysed pellet was resuspended in 10 mL of PBS (a small 
sample was taken for counting) and underwent spin suspension at 400× g for 7 min with the 
brake on. Isolated PBMCs were finally washed with PBS-EDTA and then resuspended for 
downstream analyses. The blood was processed within 4 h of collection for all samples. All 
samples obtained on the day were processed side-by-side to avoid variation from processing. 
They were analyzed fresh on the BD Rhapsody Single-Cell Analysis System platform to avoid 
clumpy cells and achieve a higher viability % of cells.  

2.2.2 Materials used:  

 

BD Rhapsody Cartridge Reagent Kit 

BD Rhapsody Cartridge Reagent Kit 

BD Rhapsody cDNA Kit 

BD Rhapsody Cartridge Kit 

BD Rhapsody Targeted Amplification Kit 

BD Rhapsody Immune Response Panel, Human

2 mM Calcein AM (in DMSO), Cat. No. 564061, BD Biosciences (stored at -20)

0.3 mM DRAQ7 Cat. No. 564904, BD Biosciences (stored at 4)

Falcon Tube with Cell Strainer Cap, Cat. No. 352235

INCYTO disposable hemocytometer, Cat. No. DHC-NO1-5

Sample Buffer Rhapsody Cartridge Reagent kit, Cat. No. 633731



 
 

 

 

29 

 
2.2.3 Step-by-step processing and BD RhapsodyTM targeted library workflow:  

 
2.2.3.1 Incubation steps 

1. Thaw Calcein AM at room temperature - protect from light 
2. Keep DRAQ7 on ice - protect from light 

3. Keep Sample buffer on ice 
4. Pellet cells at 400 X g for 5 mins 

5. Remove all of the supernatant using a P1000 but be careful not to disturb the pellet, can 
leave ~ 20 ul of supernatant 

6. Resuspend the cells in 620 ul of cold sample buffer 
7. Add 3.1 ul of 2 mM Calcein AM 

8. Add 3.1 ul of 0.3 mM DRAQ7 
9. Pipette gently to mix using a P1000 

10. Incubate at 37 C in the dark for 5 min 
11. Filter cells through Falcon Tube with Cell strainer Cap 

12. Gently pipette 10 ul into the chamber in the disposable hemocytometer 
13. Keep remaining cells on ice and protected from light 
14. Place hemocytometry in the hemocytometer adaptor 

15. The hemocytometer has two sides A and B so you need to know which sample is where. 
16. You will now be given the cell concentration and viability 

17. The viability ideally needs to be above 90% to really have a good chance of success but 
you can use lower viabilities 

18. The cell concentration needs be <1000/ul, if not then you need to dilute and count again - 
the idea concentration is between 200-800 cells/ul 

19. Select which samples you want to calculate from the desired volume is 650 ul 
20. Desired number of cells will be 30,000; aim to overestimate for now to try and get at least 

20,000 
21. Calculator will then tell you what volumes you will need to prepare a new cell stock. 

22. Prepare this exactly as described in Rhapsody Sample Buffer. 
23. Store on ice 

 
2.2.3.2 Lysing Cells 

1. This is the most important step, have a stopwatch out and no other distractions 
2. Add 75 ul of 1 M DTT to a 15 ml bottle of lysis buffer 

3. This is now stable for 24 hours 
4. Write on lysis buffer bottle and date it 

5. Vortex to mix 
6. Move left slider to Lysis 

7. Set pipette to Lysis mode 
8. Aspirate up 550 ul of lysis buffer 

9. Affix tip to cartridge and pipette in lysis buffer 
10. Incubate cartridge for 2 mins 

11. Place 5 ml LoBind tube in drawer 
12. Set p5000 to Retrieval 

13. Move the front slider to Beads 
14. Get your stopwatch ready 

15. Move left slider to Retrieval, 
16. Start stopwatch- Leave for 30 secs 

17. Aspirate up 5 ml of lysis buffer 
18. Affix tip to cartridge 
19. Move left slider to 0 

20. Pipette in the lysis buffer 
21. Remove tip from cartridge and discard 

22. Move front slider to open and remove the 5 ml tube and close 
23. Place 5 ml tube on magnet for 1 min 

24. Put cartridge in Rhapsody and perform the Retrieval scan 



 
 

 

 

30 

 
2.2.3.3 Washing capture beads 

1. Place 5 ml tube is on the magnet from before 
2. Remove 4 ml of the supernatant but be very careful not to make bubbles or to disturb the 

beads 
3. Remove tube from magnet 

4. Gently resuspend the beads 
5. Remove beads to a new 1.5 tube 

6. Wash out 5 ml tube with 0.5 ml of lysis buffer 
7. Place tube on magnet for 2 min 

8. Very carefully remove the supernatant - avoid bubbles as much as possible 
9. Remove tube from magnet 

10. Pipette in 1 ml of bead wash buffer and gently mix 
11.Place back on magnet for 2 min 

 
2.2.3.4 Loading Cells on Cartridge 
1. Aspirate up 700 ul of air 

2. Affix tip to cartridge and pipette in the air 
3. Remove tip from cartridge and dispose 

4. Set pipette to mode Cell Load 
5. Aspirate up 40 ul of air 

6. Place tip in suspension of cells and aspirate up 575 ul of cells 
7. Affix tip to cartridge and pipette in all the cells 

8. Incubate cells on cartridge for 20 mins - this allows the cells to fall into the wells 
 
 

2.2.3.5 Preparing Cell Capture Beads 
1. Place capture beads on magnet for 1 min 

2. Now using an ordinary pipette remove the supernatant without disturbing the beads 
3. Remove the tube from the magnet 
4. Pipette in 750 ul of sample buffer 

5. Store on ice until ready 
 

Performing Cell Load Scan 
1. Hit Scan on the BD Rhapsody 

2. Place cartridge on tray 
3. Note: can now set a delay timer to help with cell incubation 

 
Loading and washing Cell Capture Beads 

1. Set Pipette to Prime/Treat 
2. Place cartridge on the express 

3. Aspirate up 700 ul of air 
4. Affix tip to cartridge and pipette in the air 

5. Gently mix your bead with your finger 
6. Set the automatic pipette to Bead/load 

7. Aspirate up the beads 
8. Affix tip to cartridge 

9. Pipette in your beads 
10. Incubate for 3 minutes 

11. Place cartridge in Rhapsody 
12. Hit Scan and perform Bead Load 

13. Remove cartridge and place on Express 
14. Set pipette to Wash 

15. Aspirate up 700 ul of air 
16. Affix tip to cartridge and pipette in the air 

17. Remove tip from cartridge 
18.  Remove supernatant very carefully 

19. Remove tube from magnet 



 
 

 

 

31 

20. Pipette in 1 ml of Bead wash buffer, mix gently and place on ice 
21. Now have 30 mins to start the reverse transcription 

 

2.2.4 scRNA Sequencing by Seq-Well Targeted mRNA and AbSeq Amplification Kit 

Using the BD Rhapsody Single-Cell Analysis System platform (BD, Biosciences, Franklin 
Lakes, NJ, USA), we measured the expression of 399 transcripts related to the immune cell 
system in human PBMCs at the single-cell level. Reference (Figure 4) shows that FITC-
conjugated antibodies were used to stain human mononuclear cells for the markers CD19, 
CD14, CD16, CD3, CD4, CD8, CD304, CDTCRgd, CD56, and HLA-DR. 

Ten thousand cells pooled from each patient were placed into BD Rhapsody cartridges for 
single-cell isolation. There were 16 cartridges altogether. Following the manufacturer's 
instructions (BD Biosciences, Franklin Lakes, NJ, USA), single cells were separated by 
performing single-cell capture and cDNA synthesis using the BD Rhapsody Express Single-Cell 
Analysis System. Following the manufacturer's instructions, we used the BD Rhapsody targeted 
amplification kit (cat. 633,774, BD, Biosciences, Franklin Lakes, NJ, USA) to amplify the 399 
targeted transcripts that make up the BD Rhapsody immune response panel. A side cleanup 
using AMPure XP Beckman magnetic beads (catalog number A63880; Beckman Coulter; Brea, 
California, USA) removed any stray PCR products or other small compounds. Electrophoresis 
utilizing the Agilent 2200 TapeStation, cartridge (cat. #5067-5584, Agilent, Santa Clara, CA, 
USA) and the QubitTM dsDNA HS Assay Kit (cat. #Q32851, ThermoFisher Scientific, Waltham, 
MA, USA) were used to check the DNA's concentration and purity. 

2.2.5 scRNA-SEQ Computational Pipelines and Analysis 

Single-cell sequence data matrices were acquired from BD Rhapsody samples. The DBEC 
algorithm was used to acquire readings that had been filtered based on their unique molecular 
identification (UMI). Patients with similar diagnoses were clustered together. The merge 
function in base R (version 4.1.2) was used to collect the most common cell types across all 
conditions. The cell-expressing matrices of all patients with each disease were merged into 
one. After excluding low-coverage genes and cells, researchers were left with the ten most-
expressed genes among nine types of immune cells taken from patients with three different 
diagnoses. Principal component analysis (PCA) was used to eliminate objects based on their 
loading scores; this analysis was conducted with the R package stats, and only the top two 
components were used to account for volatility in the data. We constructed heatmaps depicting 
the common expression pattern of genes expressed by cell types in at least two diseases. R's 
scale function was used to adjust the transcript counts for size. Heatmaps were visualized with 
the geom_tile function of a ggplot2 object, and datasets were stretched with the melt function. 

To categorize cells based on their transcripts, a low dimensional analysis of immune cell 
expression was carried out. The RTsne package (version 0.16) in R was used to perform a t-
distributed stochastic neighbor embedding (T-SNE) study. Wattenberg et al. (2016) detailed 
experiments with perplexity levels of 5, 30, and 50 and these were tested in our data.  

 
2.2.6 Functional Evaluation 
The Protein Analysis Through Evolutionary Relationship (PANTHER) program (version 17, 

22 February 2022) was used to submit lists of the unique, most expressed genes by cell type 
across progressive disease scenarios. The outcomes for pathways, biological processes, and 
molecular activities were visualized using pie and bar charts. 
 

 



 
 

 

 

32 

 
Fig. 4 Single-cell transcriptional landscape of severe and non-severe patients with viral and bacterial 
sepsis and immune cell dysfunction. (A) A schematic outline depicting the experimental workflow for data 
collection from the published literature and the subsequent integrated analysis. Numbers indicate the 
number of samples of different cohorts (severe COVID-19, mild COVID-19, non-severe sepsis, and septic 
shock) and expression of 399 transcripts analyzed. (B) Specifically, human mononuclear cells were 
stained with FITC-conjugated antibodies (CD19, CD14, CD16, CD3, CD4, CD8, CD304, CDTCRgd, CD56, and 
HLA-DR). (C) Single cells isolated using single-cell capture and cDNA synthesis with the BD Rhapsody 
Express Single-Cell Analysis System, according to the manufacturer’s recommendations. (D) The BD 
AbSeq workflow integrated into the BD Rhapsody system. (E–G) Cells are colored by cell subtypes. Dashed 
circles indicate the major cell types; differential expression profiles were run among mild and severe 
manifestations of infection and have varied immune cell profiling. Adapted from 
https://doi.org/10.3390/biomedicines11030778 [96]. Created in Biorender.  

 

2.3 Experiment 2: Public reference in-silico scRNA-SEQ atlas  

2.3.1. Ethics statement  

All required ethical guidelines were followed, and ethics committee approvals were 
obtained by the original study groups.  

2.3.2 Public curated datasets  

Single-cell RNA sequencing (scRNA-SEQ) raw datasets in Seurat and scanpy h5ad. 
formats were obtained from the National Center for Biotechnology Information's gene 
expression omnibus (GEO) database, and additional metadata files relating to clinical 
outcomes, treatment, cell subtype annotations, and outcome measures were obtained directly 
from the author if additional permission was required, and further standardized for cross-study 
applications. Study ID: SCP548 [1], jonas_et_al_2020_10x_pbmc [2], 

https://doi.org/10.3390/biomedicines11030778%20%5b96


 
 

 

 

33 

EGAS00001004571_cohort4 [3], GSE175453 [4], GSE167363 [5], GSE175453 [6], 
PMID_35216673_TNK [7], PMID_35216673 [8], PMID_35216673_B [9], SCP1323 [10], and 
SCP1322 [11] datasets were used for the transcriptome analysis [97-101]. In brief, mild patients 
were those who were not admitted to an ICU for more than three days and did not require 
mechanical ventilation/intubation. Following a positive SARS-CoV-2 nasopharyngeal swab 
obtained by RT-PCR and the status of respiratory failure, patients admitted to an ICU for an 
extended period of time and/or requiring mechanical ventilation/intubation were categorized 
as serious patients. Sepsis patient samples from various illness states met sepsis-3 criteria, 
and additional severity scores (such as SOFA and APACHE II) were included. 

 

 

2.3.3 Single-cell data analysis  

BBrowserX, a GPU-accelerated platform for analyzing a scRNA-SEQ was used for the 
data quality control and analysis, as comprehensively outlined by the package developer. The 
software uses CUDA programming, cache-sensitive data structures, and other optimization 
strategies, we obtain hundreds (in some cases, thousands) times speedup in running time 
(while reducing memory) for essential algorithms, such as PCA, t-SNE, UMAP, Louvain 
clustering, Harmony batch effect removal, k-nearest neighbors, and AUCell (for geneset 
enrichment). The analysis was created with BBrowserX software (BioTuring Inc., San Diego, CA, 
USA). Briefly, the Seurat/scanpy objects were created from individual expression matrices. 
Unique molecular identifier (UMI) counts were scaled by library size and a natural log 
transformation. Gene counts for each cell were divided by the total UMI count of that cell, scaled 
by a factor of 10, 000, and then transformed via the natural log plus 1 function, “NormalizeData”. 
Filtering parameters are further applied to allow you to keep only high-quality cells. The 
parameters include the minimum and the maximum number of reads (10 – 1,000,000), the 
minimum and the maximum number of detected genes (10 – 100,000), and the maximum 
percentage of mitochondrial genes set at 25%. The normalized data were further scaled so that 
the mean expression across cells was 0 and the variance was 1 [102]. 



 
 

 

 

34 

Following cell filtration, data was normalized using the "LogNormalize" global-scaling 
normalization method, which divides the specific feature counts of each cell by the overall 
counts of that cell, multiplies it by a scaling factor (104) and then performs natural log-
transformation. In the context of scRNA-SEQ, a Z-scoring metric reveals how far one cell's 
frequency for a particular characteristic deviate from the mean of all cells' frequencies for that 
feature. The log-normalized counts are used to construct Z-scores. PCA was used in BBrowserX 
with the "RunPCA" function to minimize data dimensionality by clustering comparable cells from 
diverse datasets. PCA is typically used to map high-dimensional data to low-dimensional space 
using linear models. Following that, the system identified anchors across datasets by 
embedding cells in a k-nearest neighbor-based technique (KNN) to identify mutual neighbors 
from distinct datasets and scoring them based on their mutual nearest neighbors using the 
Venice binarizer.  

The differences in noise and batch effect among the datasets were taken into account 
by using reference cells from each dataset. Harmony [102] is a batch effect reduction technique 
that groups cells into a shared embedding based on cell kinds rather than dataset-specific 
variables. Harmony uses a cell-specific factor to correct each cell: a linear mixture of dataset 
correction factors weighted by the cell's soft cluster assignments made in step. The harmony 
approach is used in this system by removing matrix calculation and using the Woodbury matrix 
identity [102]. 

 
 

The utilization of graph-based hierarchical clustering enables the identification of 
linkages that exist both inside and between cellular populations on a manifold. In this analysis, 
we utilized the structural shared nearest neighbor-Louvain (SSNN-Louvain) method to analyze 
the relationship between a node and its shared nearest neighbors. The weight of an edge is 
determined by considering the ratio of shared nearest neighbors to nearest neighbors, thereby 
incorporating the structural characteristics of the graph. The utilization of the uniform manifold 
approximation and projection (UMAP) technique facilitated the visualization of cellular data 
with high dimensionality in a manner that is both accessible and complete. The identification 
of differentially expressed genes (DEGs) was conducted using the Venice technique [103], 
which involves the designation of marker genes for a certain cell population. These marker 
genes serve the purpose of distinguishing cells within the population from cells that are not 
part of the population. Based on this concept, the methodology develops an optimal classifier 
that aims to discern the cellular population based on the expression level of individual genes. 
In the conventional approach, genes that exhibit differential expression are often categorized 
as either up-regulated or down-regulated. In addition to log-fold change, the methodology 
incorporates a novel metric for categorizing genes into up-regulated, down-regulated, and 
transitional states [103]. 

2.3.4. Cell type annotation  

The analysis refers to both the author's original cell subtype annotations and a cell type 
prediction algorithm developed internally using the BBrowserX software. The method makes 
use of Talk2Data's publicly available curated scRNA-SEQ and spatial transcriptomics datasets. 
It creates huge training sets of > 120 million cells that span as many cell types as feasible. It 
employs the Venice 2 approach for grouping analogous genes (genes with similar expression 
patterns). The workflow searches for the most important combinations that can get the best 
score across all equivalent gene combinations (branch and bound), adding up to eight high-
scoring panels [104] while adjusting for specificity and sensitivity in projected marker panels. 
 
2.3.5 Analysis of dynamic changes in cell type composition compared with healthy donors 
 



 
 

 

 

35 

To investigate the dynamic changes in cell type composition, we calculated the 
proportion of cell types in each individual dataset (n = 10). As a control, we calculated the 
relative variation in each cell type composition between all pairs of healthy donors. Similarly, 
for each disease group, we calculated the relative variation in each cell type by dividing the 
fraction of the cell type in individual patient by that of individual healthy donor. After log2 
transformation, we conducted statistical analysis using the relative variation in composition 
between the control and disease groups using a data ranking system. For each cell in each 
single-cell dataset, we rank all its genes by the expression level. The underlying rationale is that: 
If gene A is expressed higher than gene B in one cell, gene A should be ranked higher than gene 
B, no matter which technology is used to measure their expression level. As there can be 
multiple genes with the same absolute expression value, their ranks are then  assigned by the 
average of their original ranks (SciPy, 2021). Given the assumption that a cell cannot express 
more than 5000 genes (in the current technology detection limit), we then convert the rank into 
5000 - rank [105]. In this setting, the gene with the highest expression value will have rank 5000.   

Although the ranking system helps to overcome the differences between different 
units, the perturbation is still too large. A slight change in the original expression values can 
lead to significant differences in rank values. Therefore, we further transform the data into 
zero and one - one means this gene is expressed in the cell and vice versa by applying 
Boolean transformation. A gene is considered as expressed in a cell if it has a positive 
expression value and ranks among the top 5000 genes of the cell. From now, when we say 
gene A is expressed in cell B, it means gene A has the rank value smaller or equal to 5000 in 
cell B. Cell ontology and meta data ontology is also standardized across all datasets using EBI 
ontology for cell type and tissue annotations https://www.ebi.ac.uk/ols/ontologies/cl, ncbi for 
curated disease https://www.ncbi.nlm.nih.gov/mesh/, and pubchem ncbi 
https://pubchem.ncbi.nlm.nih.gov/ for curated drug names.   

Enrichr https://maayanlab.cloud/Enrichr/ was applied as a source to examine each cluster-
specific genes by gene set enrichment analysis (GSEA) using cytokine-responsive gene sets 
originated from each cytokine-treated cells further immune checkpoint inhibitors/co-
stimulatory markers. AuCell enrichment is applied evaluates the activity of the gene set in each 
cell based on the ranking of all genes and returns a score ranging from 0 to 1. It allows to 
identify cells with active gene regulatory networks in single-cell RNA-SEQ data which the user 
can establish based on own gene sets for specific pathways. The input to AUCell is a gene set, 
and the output is the gene set in each cell. 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

https://www.google.com/url?q=https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.rankdata.html&sa=D&source=editors&ust=1681469833054218&usg=AOvVaw1F7ynUCFEJP3cbObPW1TKg
https://www.ebi.ac.uk/ols/ontologies/cl
https://www.ncbi.nlm.nih.gov/mesh/
https://pubchem.ncbi.nlm.nih.gov/
https://maayanlab.cloud/Enrichr/


 
 

 

 

36 

Title Study_ID Number 
of cells 

Disease 
status 

Number of 
samples 
included 

Plasma from patients with bacterial sepsis or severe 
COVID-19 induces suppressive myeloid cell 
production from hematopoietic progenitors in vitro 
(Sepsis patients) 

SCP548 122,557 Bacterial 
Sepsis, ICU 
non-sepsis, 
ICU sepsis, 
Control  

 N = 31 

Severe COVID-19 Is Marked by a Dysregulated 
Myeloid Cell Compartment (10X - PBMC) 

jonas_et_al_2020_10x_pbmc 33654 COVID-19  
Healthy 
controls 

 N = 20  

Single Cell RNA-seq of Human Myeloid Derived 
Suppressor Cells in Late Sepsis Reveals Multiple 
Subsets with Unique Transcriptional Responses: A 
Pilot Study 

Darden_et_al_2021 4,705 Sepsis  
Healthy 
controls 

 N = 4 

Early IFN-α signatures and persistent dysfunction are 
distinguishing features of NK cells in severe COVID-
19 

EGAS00001004571_cohort4 2316 Mild-COVID-
19 
Severe 
COVID-19  
Healthy 
controls 

 N = 76 

Dynamic changes in human single-cell transcriptional 
signatures during fatal sepsis 

GSE167363 57138 Survived 
septic patient  
Non-survived