Transferrin Receptor (TFRC): A Potential Biomarker for The Diagnosis and Prognosis of Sepsis

Objective This study applies the data independent acquisition (DIA) technique combined with bioinformatics to identify differential proteins in sepsis patients and performed ELISA method to validate the candidate protein of clinical value, in an attempt to nd new biomarkers for the diagnosis and prognosis of sepsis. Blood samples from sepsis patients (Sepsis group, n = 50) and healthy individuals (NC group, n = 10) were collected from Aliated Hospital of Southwest Medical University. Mass spectrometry analysis was designed for 22 sepsis samples (randomly selected) and 10 healthy controls by DIA method, and the obtained differential proteins were subjected to GO annotation, meta-analysis and survival analysis to identify the candidate biomarker protein. ELISA was applied to validate the protein expression in original cohorts. ROC curves based on ELISA data were plotted to discuss the diagnostic and prognostic performance of the candidate protein and several clinical indexes, including C-reactive protein (CRP), procalcitonin (PCT) and lactate (Lac). potential biomarker for sepsis.


Introduction
Sepsis has now been a great di culty and challenge in the eld of medicine. Despite the intensive studies from all elds and the management of a series of active "rescues" in the world, sepsis still suffers from a high morbidity and fatality [1]. There was research covering adult sepsis in 27 developed countries and reporting that, the annual incidence of sepsis over the last 10 years was around 437/100,000, and approximately 17% of them were succumbed to death. Besides, 270/100,000 had a sever disease with 26% deaths [2]. In China, a study in 2020 which involved the multi-center intensive care unit (ICU) from 44 hospitals uncovered that, the incidence and fatality of ICU sepsis were approximately 20.6% and 35.5%, respectively, and the fatality of those with severe disease was up to 50% [1]. Sepsis is a life-threatening organ dysfunction caused by a dysregulated host response to infection [3]. During the onset process, there exist a series of physiological and pathological processes, including pathogen invasion, release of cytokines, microcirculation dysfunction, and the imbalance between the pathogen and the human immune system. Currently, diagnosis of sepsis adopts the Sepsis3.0, where sepsis is de ned when Infection + ΔSequential Organ Failure Assessment (SOFA) ≥ 2. In terms of the treatment for sepsis, the rescue and recovery of organ function are the focus. However, the prevention, treatment and prognosis of sepsis remain unsatisfactory, although great efforts have been made over the past two decades.
Biomarker is a sort of molecule of biological signi cance that can subjectively indicate the physiological, pathological and pharmacological conditions in patients [4], including DNA, RNA, proteins, and small metabolites, which are mainly applied in the diagnosis and staging of disease, as well as the prediction of therapeutic effect in certain population [5]. In clinic, patients having the same symptoms or vital signs but different biomarkers often have different prognoses and therapeutic outcomes [6]. In this context, biomarkers are increasingly important in disease diagnosis and prognosis, especially in precision medicine and individualized treatment [7]. Currently, C-reactive protein (CRP), serum procalcitonin (PCT) and lactate (Lac) are frequently used in infection determination and pathogen identi cation in sepsis, and PCT is the speci c biomarker of bacterial infection [8]. Since the issue of the international diagnostic criteria for sepsis (Sepsis3.0) in 2016, PCT, CRP and Lac have less value in the diagnosis and prognosis of sepsis. Hence, searching for new markers that can predict the development and outcome of sepsis can be a new direction for the prevention and treatment.
DIA (data independent acquisition) is a label-free quantitative technique that provides high repeatability, high detection sensitivity, high quantitative accuracy, and allows for data informatization [9]. It has been the most noteworthy technique in recent years. It allows to conduct protein identi cation and quantitative analysis in multiple samples or in samples of different batches at the same time, which is available for large-scale clinical studies into plasma/serum protein biomarkers. In 2015, international proteomics researchers Ruedi et al. [10] used DIA to uncover the in uence of genetic and environmental factors on individuals, which is helpful for the discovery and evaluation of clinical biomarkers. By now, DIA has been used in the research of biomarkers in multiple elds, such as tumor [11,12], lung disease [13], and obesity [14].
In this study, DIA method was applied to screen out differential proteins in serum samples from sepsis patients and normal controls, and a public database was consulted for veri cation. Besides, the candidate protein was further validated by ELISA test. This study attempts to nd new biomarkers for diagnosis and prognosis of sepsis. DIA mass spectrometry DIA was completed with the samples randomly selected from the sepsis cohort (n = 22) and the normal cohort (n=10). Liquid Chromatography-Mass Spectrometry (LC-MS) (Q-Exective HFX, Thermo Scienti c) was applied to conduct mass spectrometry for protein enzymatic peptides of the samples. A spectra database was constructed by traditional data dependent acquisition (DDA). The Ratio values and corresponding p-values that indicate the changes in protein expression abundance of the 32 samples were identi ed and quantitatively analyzed by using mProphet algorithm.

Materials And Methods
Functional enrichment and screening of target proteins All data were rstly subjected to log treatment, and then mapped into a box plot to identify the homogeneity of the samples. Principal component analysis (PCA) was implemented to exclude outliers from the two cohorts. Proteins with differential expression between the normal and sepsis cohorts were screened out assisted by the online tool iDEP91 (http://bioinformatics.sdstate.edu/idep/) [15,16]. Protein meeting fold change (FC) ≥2.0 and false discovery rate (FDR) <0.05 were selected, and then analyzed in enrichment analysis for main biological functional pathways.

Expression of target protein at transcription level
To know more about the transcription level of target protein in different groups, Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo/) was consulted to obtain seven sepsis-related data sets: GSE28750 [17], GSE54514[18], GSE63042 [19], GSE67652 [20], GSE69528 [21], GSE65682 [22] and GSE95233 [23]. All the datasets were documented with human peripheral blood samples with a sample size ≥20. The GSE65682 dataset contains 28-day prognostic data for 479 patients with sepsis, which were here used for survival analysis of the candidate target. Data from the 7 data sets were normalized (log2 treatment). Four groups were generated: sepsis group (Sepsis) and normal group (NC), sepsis survival group (Survival) and death group (Nonsurvival). A comprehensive meta-analysis was designed for single genes which were from different datasets but in the same group, based on R language.
Enzyme-linked immunosorbent assay (ELISA) All samples from the Sepsis group (n = 50) and the NC group (n = 10) were tested by ELISA. Double antibody sandwich method was applied to quantitatively analyze the proteins of each sample. Brie y, puri ed speci c antibodies were seeded into a micropore plate to immobilize, followed by an addition of tested protein samples and horseradish peroxidase (HRP) -conjugated testing antibodies in turn. An antibody-antigen-enzyme labeled antibody complex was formed. Substrate TMB (3,3,5',5'-Tetramethyl benzidine) was added for color development after washing. The darkness of the color is positively correlated with the amount of protein in the sample. The optical density (OD) at 450 nm was read on a microplate reader, and the protein content was calculated by a standard curve. The obtained data were loaded to GraphPad Prism8.0 software. Between-group differences were statistically analyzed by independent sample t-test, and considered signi cant when p < 0.05.

ROC
To assess the diagnostic and prognostic value of the candidate biomarker, receiver operating characteristic (ROC) curve was drawn by MedCalc 15.2 software, and the area under the ROC curve (AUC) was calculated. Favorable diagnostic performance was de ned when AUC >0.7 in the NC group versus Sepsis group, and excellent prognostic performance was de ned when AUC >0.7 in the Survival group versus Death group of sepsis. Besides, the diagnostic and prognostic performance of the candidate were compared with those of several clinical indexes, including CRP, PCT and Lac.

DIA data
Two hemolysis samples were excluded. In box plots and PCA analysis by iDEP91, there was good homogeneity in NC and Sepsis samples and the between-group discrimination is good as well, with no outliers (Fig. 1A-B). There were 142 differential proteins in the Sepsis group versus NC group, composed of 36 down-regulated proteins and 106 up-regulated proteins (Fig. 1C-D).

Screening of candidate biomarkers
Based on the differential proteins, Gene Ontology (GO) annotation was operated to nd the most enriched GO terms, which were immune response, response to stress, in ammatory response, cell activation, etc. (Fig. 2A). The top 10 proteins with the greatest differential expression were ltered according to corresponding p-values in DIA anlaysis, including FUCO2, MGAT1, OAF, AACT, TFRC, CCL14, EXTL2, KLKB1, TETN, CRP, and SAA1 (Fig. 2B). All the differential expressions in the NC versus Sepsis were statistically signi cant (p < 0.05).

Transcriptional expression of candidate biomarker
A meta-analysis was conducted based on the transcription data of candidate protein (TFRC) documented in GEO database. It demonstrated a signi cant increase of TFRC in the Sepsis group versus the NC group (Fig. 3A). Besides, TFRC expression was even higher in the Nonsurvival group versus the Survival group in sepsis patients, with statistically signi cant difference (Fig. 3B). Further survival analysis for TFRC was carried out using the GSE65682 data. It was found that patients poorly expressing TFRC had a higher survival rate (p = 0.00034) (Fig. 4A).

ELISA veri cation
There were 10 samples in the NC group, 48 in the Sepsis group (excluding 2 hemolysis samples), 14 in the Nonsurvival group and 34 in the Survival group. ELISA analysis identi ed a signi cant increase of TFRC expression in the Sepsis group versus the NC group (156.83 ± 84.71 nmol/L versus 87.99 ± 47.89 nmol/L), and the difference between the two groups was of statistical signi cance (p < 0.05) (Fig. 4B). In the sepsis cohort, the expression of TFRC in the Survival group was 130.97 ± 40.45 nmol/L, much lower than 219.63 ± 125.59 nmol/L in the Non-survival group, with a statistically signi cant difference (p < 0.05) (Fig. 4C).
Though the diagnostic performance was inferior to PCT, the prognostic performance was superior to PCT, CRP and Lac.

Discussion
Page 7/16 TFRC (Transferrin receptor/CD71) has two subtypes: TFR1 and TFR2. TFR1 commonly expresses on the surface of common cells, while TFR2 speci cally expresses in liver cells. TFRC has balancing act on iron metabolism, as a mediator of iron transport through endocytosis by binding to transferrin. Besides, it has biological activities in multiple biological processes, such as production of free radicals, hemoglobin synthesis, oxygen transport, DNA synthesis, production and release of neurotransmitters, and the regulation of steroid hormones [24]. The expression of TFRC is regulated by hypoxia-inducible factors and the iron regulatory protein-iron response element system. Presence of both hypoxia and iron de ciency will augment TFRC expression, which yet can be regulated by some microRNAs (miRNAs) [24]. Currently, there have been many studies into the role of TFRC in tumors. For instance, TFRC over expresses in tumor cells due to excessive proliferation and increased iron requirement. In addition, TFRC can facilitate tyrosine phosphorylation to inhibit apoptosis, and also can promote tumor cell growth by activating JNK signaling pathway in multiple cancers, such as breast cancer, liver cancer, ovarian cancer, etc. [25][26][27] It suggests that the high expression of TFRC in tumor may indicate adverse outcomes of patients.
Recent studies have found that TFRC can regulate immune function, while the speci c mechanism is not clear. TFRC missense mutation (Y20H) does not affect peripheral lymphocytes but leads to impairment of the function of peripheral T cells and a decrease of memory B cells. In the meantime, the ability to produce antibody and the function of immunoglobulin class conversion can also be impaired, manifested as serious immunode ciency [28]. In a mRNA sequencing study for preterm and term infants in Korean population, TFRC was found to be the core factor involved in T cell activation and closely related to the occurrence of preterm birth, indicating that the TFRC could be a predictor of preterm birth [29]. In addition, TFRC de ciency is regarded as a syndrome of immunode ciency, manifested as recurrent severe lung infections. In this case, there are TFRC mutations, and the proliferation, function and transformation of T cells and B lymphocytes are affected [30]. A study reported the clinical manifestations and immunological characteristics of 8 patients with TFRC mutations and found T cell function impairment in all patients [31].
Due to the certain morbidity and high fatality [32], sepsis has always been a hot topic in Emergency Medicine, Critical Care Medicine, Infectious Disease and even Surgery. However, there are certain limitations in the prevention and understanding of sepsis because of the complex pathological and physiological processes. It might be possible to reduce the morbidity and fatality of sepsis if more positive and effect managements are provided, in case of an early diagnosis via cytokines or biomarkers, awareness of disease severity and a prediction of possible outcomes. This study combined bioinformatics and DIA analysis to identify TFRC as a potential biomarker of sepsis. According to literature, we found that TFRC has the function of immunoregulation, yet the underlying mechanism remains to be elucidated, and its role in sepsis is rarely reported. We further performed ELISA and found that the expression of TFRC in sepsis cohort was signi cantly increased relative to that in healthy individuals, and it was much higher in patients who died. Besides, the ROC curves for the TFRC along with several clinical indexes (PCT, CRP and Lac) identi ed that, the TFRC had diagnostic and prognostic performance to a certain extent. It was superior to PCT in prognostic judgement with both higher speci city and sensitivity. A. Box plot shows the normalized DIA data of each sample. The proteins are distributed at the same level, which is comparable; B. PCA analysis shows good discrimination of the two cohorts, and there are no outliers; C. Volcano plot shows the differential proteins screened by t-test (up-regulated in red and downregulated in green). The X-axis is the log2 fold change, and the Y-axis is the -log10; D. There are 142 differential proteins in the Sepsis group versus the NC group, including 36 down-regulated and 106 upregulated. PCA, principal component analysis; NC, normal control.  A. Meta-analysis for TFRC based on GSE28750, GSE54514, GSE67652, GSE69528 and GSE95233 data.
TFRC shows a low expression in the NC group versus the Sepsis group in all datasets (heterogeneity, p < 0.01; random effect model; 95% CI, -0.93 -0.00); B. Meta-analysis for TFRC based on GSE54514, GSE63042 and GSE95233 data. TFRC shows a low expression in the Survival group versus the Nonsurvival group in all datasets (heterogeneity, p = 0.63; xed effect model; 95% CI, -0.70 --0.18). NC, normal control; 95% CI, 95% con dence interval.  A. Survival analysis for TFRC based on GSE65682 data. Patients with low TFRC expression had better prognosis than patients with high expression; B-C. ELISA analysis showed that the expression of TFRC was signi cantly higher in the Sepsis group (B, versus NC group, p < 0.05) and the Nonsurvival group (C, versus Survival group, p < 0.05). NC, normal control. Figure 5