Circulating Long Noncoding RNAs Positively Correlate with the Increased Risk, Elevated Severity and Unfavorable Prognosis in the Sepsis Patients

Objective This study aimed to evaluate the correlation of circulating long noncoding RNAs (lncRNAs) expression with disease risk, severity, inammatory cytokines levels and prognosis in patients with sepsis. Differential expression proles of lncRNA in the serum of sepsis rats were screened by high-throughput transcriptome sequencing. Homologous lncRNAs in the upregulation group were identied by homology analysis in rats and humans. The expression differences of these homologous lncRNAs in the serum of 176 sepsis patients and 176 healthy controls (HCs) were detected using reverse transcription quantitative polymerase chain reaction (RT-qPCR). And inammatory cytokines levels were detected by enzyme-linked immunosorbent assay (ELISA). A receiver operating characteristic (ROC) curve was used to verify the diagnostic and prognosis values. Spearman correlation coecient was used to analyze the correlation between the variables. Follow-up was performed to observe the 28-day mortality.

Introduction Sepsis is a life-threatening organ dysfunction resulting from a dysregulation of the host's response to infection, meanwhile, septic shock should be de ned as a subset of sepsis [1][2][3]. According to clinical epidemiology statistics, globally, there are an estimated 31.5 million cases of sepsis and 19.4 million cases of septic shock each year, with 5.3 million deaths [4]. Despite the signi cant advances in clinical management and use of devices, sepsis remains a signi cant public health problem worldwide due to high morbidity, mortality, and rising costs associated with the complex care of patients [5]. In addition, the new data from the current emerging COVID-19 pandemic indicate a close relationship between severe acute respiratory syndrome coronavirus-2 (SARS-COV-2) and sepsis; characterized by organ damage and systemic in ammation due to viral invasion [6][7][8]. Thus, it is necessary to search for new diagnostic and therapeutic approaches to improve clinical prognosis by elucidating the pathogenesis of sepsis comprehensively.
Long non-coding RNAs (lncRNAs) are a large class of 200-nucleotide long noncoding transcripts that lack the ability to encode proteins [9,10]. Previously, they were considered to be the "noise" produced by genome transcription and have no biological functions [10]. However, now more and more studies have found that lncRNAs can participate in a variety of biological processes, including in ammation, transcriptional activation, transcriptional interference, histone modi cation, chromatin remodeling, cell cycle regulation, epigenetics, and RNA splicing [11][12][13]. With the advancement of sequencing technology and chip detection technologies, a growing number of lncRNA expression pro les have been observed recently [14]. The lncRNAs have a regulatory function in human diseases, especially cancer, which have become diagnostic markers and therapeutic targets for related tumors [15,16]. The lncRNAs were detected not only in cancer tissue, but also in various body uids, including urine, blood and saliva [17]. Through detection of lncRNAs in body uids and further molecular mechanism studies, it was found that these lncRNAs play an important regulatory role in cardiovascular diseases [18] and in ammatory diseases [19]. But beyond that the corresponding regulatory function of lncRNAs in multiple organ failure caused by sepsis has been gradually reported in recent years [20]. Therefore, these reported studies on lncRNAs provide new possibilities for early rapid identi cation and clinical treatment of sepsis.
However, compared to the large lncRNAs families, current studies related to sepsis are very limited [21]. Therefore, through high-throughput transcriptome RNA sequencing, we found some up-regulated lncRNAs in peripheral blood of septic rat model, and two lncRNAs were identi ed by homology analysis of human and rat, respectively PKN2-AS1 and AC068888.1. In this paper, we further assessed their ability to distinguish between sepsis and healthy controls, as well as their association with disease severity, in ammation, and survival in patients with sepsis.

Animals and LPS model of sepsis
Nine Speci c Pathogen Free (SPF) male Sprague-Dawley rats (7-8 weeks old, weighing 250g to 300g) were used as an in vivo model species, which were provided by the Animal Experiment Center of Anhui Medical University(Anhui, China). The rats were placed at 3 rats per cage and labeled cage 1, 2, and 3, respectively. The food and water were freely provided, and the room temperature (22 ± 2 ℃), humidity (60%-80%), and 12-hours light/dark cycle were maintained. Before onset of the experiment, the rats were allowed to adapt to the environment for a week. All animal procedures were approved by the Ethics Committee of Anhui Medical University (Ethics Approval Number LLSC20190721). The lipopolysaccharide (LPS) (Sigma, St. Louis, MO, USA) was dissolved in sterile normal saline (0.9% NaCl) at a concentration of 1 mg/mL and injected intraperitoneally at a dose of 10 mg/kg. The rats of cage 2 and 3 were treated for 6 hours and 24 hours respectively. The control group (cage 1) was injected with the same amount of saline in the same way [22].

Treatment of model specimens and high-throughput transcriptome sequencing analysis
The rats of cage 2 and 3 were given pentobarbital sodium (Sigma, St. Louis, MO, USA) (100 mg/kg) intraperitoneally 6 h and 24 h after LPS treatment, and the control group (cage 1) was anesthetized using the same method. After satisfactory anesthesia, 7 mL of whole heart blood was collected under direct vision, and the serum was separated using a low temperature and high-speed centrifugation method (Step 1: 4 ℃, 3000 rpm, 10 min, Step 2: 4 ℃, 12000 rpm, 10 min). High-throughput transcriptome sequencing was performed on the collected sera to detect the information of abnormally expressed lncRNAs by BGI Company (Shenzhen, China). The sequencing item number was F19FTSCCKF1548_ou201909271103114221RA0305. Signi cantly DElncRNAs were studied using DESeq R software package (3.5.1). A threshold value of |log2 (fold change)|>1 with a P value <0.05 was determined. In order to obtain an overview of the expression pro les of lncRNAs, 'pheatmap' and 'ggplot2' R packages were used to draw the heat map and volcano plot, respectively. Through genomic analysis, human-rat homologous lncRNAs were screened out from all the highly expressed lncRNAs, and representative highly expressed lncRNAs were selected for clinical veri cation.  . All participants or their family members provided written informed consent before the registration for this study.

Inclusion and exclusion criteria
The screening criteria were as follows: (a) patients diagnosed with sepsis according to the third International Consensus De nitions for Sepsis and Septic Shock (1), (b) patients with the age of ≥ 18 years old and ≤75 years old, (c) patients who admitted to the intensive care unit (ICU) within the previous 24 hours, (d) patients who did not have other fatal diseases (e.g., hematologic malignancies, solid tumors, or acquired immune de ciency syndrome), (e) patients without immunosuppressive therapy within 3 months before the enrollment, (f) patients who were not in pregnancy or lactation.

Date collection
Sepsis patients' clinical characteristics were recorded after admission, which included demographic characteristics, complications, primary infection sites, organ dysfunction, glucocorticoid drugs intakes, biochemical indexes, and disease severity. The severity of the sepsis was assessed within 24 hours after admission using the APACHE II score and SOFA score. All patients were treated in accordance with the current guidelines for treatment of sepsis and followed up for 28 days (28-day survival was recorded as well).

Sample collection
The peripheral blood (PB) samples of the sepsis patients were collected within 24 hours after admission, and the PB samples of the HCs were obtained at the time of enrollment. After collection, the PB samples were centrifuged at 3000 rpm for 10 minutes at 4°C to separate plasma. Then, the plasma samples were centrifuged at 12000 rpm for 10 minutes at 4°C to separate serum, which was preserved at −80°C for the subsequent analysis. The absolute expression levels of lncRNAs in the serum samples were detected using reverse transcription-quantitative polymerase chain reaction (RT-qPCR).
Real-time quantitative polymerase chain reaction (RT-qPCR) The serum was separated from the blood samples by centrifugation, and the total RNA in serum was of lncRNA PKN2-AS1 and AC068888.1 transcripts per milliliter of serum were quanti ed by using RT-qPCR assay. In this assay, serially diluted RT-PCR products of lncRNAPKN2-AS1andAC068888.1 were used as templates to formulate standard curves, and then, the exact copies of lncRNAPKN2-AS1 and AC068888.1 per milliliter of serum were calculated accordingly. Using the data of three independent tests, the absolute expression of the target gene was calculated by the standard curve method [23].
Detections of in ammatory cytokines in the sepsis patients Serum in ammatory cytokines, IL-6 and TNF-α, were detected by ELISA kit (Shanghai Enzyme-linked Biotechnology Co., Ltd., Shanghai, China) in patients with sepsis.
Statistical analysis SPSS26.0 software (IBM Corporation, Armonk, NY, USA) and R language version 3.5.1 (R Foundation for Statistical Computing, Vienna, Austria) were used for statistical analysis and generating graphs, respectively. Continuous data were represented as the mean ± standard deviation (SD) or the median (IQR) of interquartile intervals. The categorical data were expressed in numbers (percentages). The T-test, Chi-square test or Wilcoxon rank-sum test were used to compare the differences of variables between the two groups. The Spearman rank correlation test was used to analyze the correlation between the two variables. The ROC curve was plotted to evaluate the abilities of different subjects using the AUC, sensitivity, and speci city of the optimal cut-off point. The survival probability was described by the Kaplan-Meier method, and the difference of survival probability between the two groups was described by logarithmic rank test. The univariate and likelihood ratio forward stepwise multivariate logistic regression model were used to predict the independent risk factors in patients with sepsis. P < 0.05 was considered statistically signi cant.

Results
High-throughput transcriptome sequencing RNA gene analysis Unsupervised hierarchical cluster analysis was used to analyze the expression pattern of lncRNAs in the serum from the sepsis rats at different periods (0-6-24 hour). The results showed that there was a signi cant difference in the lncRNA expression pro les between volcano map (Fig. 1a) and heat map (Fig. 1b). In this study, 63 lncRNAs (29 up-regulated and 34 down-regulated) were identi ed as signi cantly differentially expressed between the normal control group (NC) and sepsis group (Supplementary Table S1  converted multiple glucocorticoids into hydrocortisone equivalents, and the nal statistical median use of glucocorticoids was 80.0 (1000.0). Besides, the median APACHE II score was 20.0 (8.0) and the median SOFA score was 10.0 (4.0) in sepsis patients. The detailed information regarding other characteristics such as body mass index (BMI) and biochemical indexes were shown in Table 1.
Absolute expression of serum lncRNA PKN2-AS1 and AC068888.1 and their diagnostic value in the sepsis The absolute expression of serum lncRNA PKN2-AS1 was 329.3 (77.4) gc/mL in patients with sepsis and 203.1 (96.1) gc/mL in the HCs. Meanwhile, the absolute expression of lncRNA AC068888.1 was 312.1 (91.5) gc/mL in patients with sepsis and 175.0 (93.6) gc/mL in the HCs, and further comparison analysis showed that lncRNA PKN2-AS1 and AC068888.1 expression were increased in in patients with sepsis compared with the HCs (P < 0.001) (Fig. 2a, 2b). From the ROC curve, the lncRNA PKN2-AS1 and AC068888.1 were differentiated between the sepsis patients and the HCs with an AUC of 0.879 (95% CI: 0.843-0.915) and 0.842 (95% CI: 0.801-0.884) (Fig. 2c, 2d). The optimal cut-off point for the level of lncRNA PKN2-AS1 that distinguished patients with sepsis from HCs was >238.0, and the speci city and sensitivity were 93.8% and 67.0%, respectively. Correspondingly, the optimal cut-off point for lncRNA AC068888.1 level that distinguished patients with sepsis from HCs was >210.6, and the speci city and sensitivity were 88.6% and 71.0%, respectively. These data suggested that both lncRNA PKN2-AS1 and AC068888.1 might be biomarkers for the sepsis risk. However, the predictive effect of lncRNA PKN2-AS1 was more prominent in comparison ( Table 2).
Clinical characteristics of septic shock and no septic shock sepsis patients and risk factors for septic shock During the 28-day follow-up period, 100 (56.8%) sepsis patients occurred septic shock, and they were grouped as septic shock patients, 76 (43.2%) patients did not occur septic shock, and they were grouped as no septic shock patients. Univariate analysis revealed that the number of organ dysfunction, platelet (PLT), serum creatinine (Scr), PCT, Lac, prothrombin time (PT), activated partial thromboplastin time (APTT), APACHE II score, SOFA score (P < 0.001), other infections (P = 0.021), CRP (P = 0.022), brinogen (FIB) (P = 0.016), the absolute expression of lncRNA PKN2-AS1 (P < 0.001, Fig. 4a) and lncRNA AC068888.1 (P < 0.001, Fig. 4b) were signi cantly higher in patients with septic shock compared with in those without septic shock. The others were not signi cantly associated with the occurrence of septic shock (Table 3). In addition, likelihood ratio forward stepwise multivariate logistic regression analysis demonstrated that the highly absolute expression of lncRNA AC068888.1 (P < 0.001), Lac level (P = 0.045), SOFA score (P = 0.031), and PT (P = 0.026) were independent risk factors for septic shock, whereas the absolute expression of lncRNA PKN2-AS1 and others exhibited no association with the severity of septic shock (Table 4).
Diagnostic value of the lncRNA PKN2-AS1 and AC068888.1 level for septic shock The ROC curve analysis exhibited signi cant predictive value for lncRNA PKN2-AS1 (AUC = 0.704) and AC068888.1 (AUC = 0.812) in distinguishing patients with septic shock from those without septic shock (Fig. 4c). The predictive value of lncRNA AC068888.1 was parallel to Lac (AUC = 0.816), but apparently higher compared with those for APACHE II score (AUC = 0.668), SOFA score (AUC = 0.791) and lncRNA PKN2-AS1. At the optimal cut-off point of >302.3 for the serum absolute expression level of lncRNA AC068888.1, the speci city and sensitivity were 78.0% and 77.6%, respectively. At the optimal cut-off point of >297.6 for the serum absolute expression level of lncRNA PKN2-AS1, the speci city and sensitivity were 85.0% and 51.3%, respectively. At the optimal cut-off point of >3.6 for Lac levels, the speci city and sensitivity were 81.0% and 76.3%, respectively. At the optimal cut-off point of >9.5 for SOFA score, the speci city and sensitivity were 81.0% and 64.5%, respectively. At the optimal cut-off point of >20.5 for APACHE II score, the speci city and sensitivity were 56.0% and 72.4%, respectively. Those independent risk factors of septic shock were used to construct the predictive model for septic shock risk in sepsis patients (including lncRNA AC068888.1 score, SOFA score, Lac level), then the following ROC curve analysis manifested that the predictive model exhibited a good value for identifying septic shock risk in sepsis patients (AUC = 0.882) (Fig. 4c and Table 5).
Clinical characteristics of survival sepsis and non-survival sepsis patients, and risk factors for the unfavored prognosis of sepsis 96 (54.6%) sepsis patients survived during the 28-day follow-up period, and they were grouped as survival patients, 80 (45.4%) sepsis patients died, and they were grouped as non-survival patients. As presented in Table 6, univariate analysis revealed that number of organ dysfunction, the levels of PCT and Lac, APACHE II score, SOFA score (P < 0.001), glucocorticoid drugs (P = 0.007), septic shock (P = 0.021), PLT (P = 0.010), albumin (P = 0.030), Scr (P = 0.004), CRP (P = 0.013), PT (P = 0.011) and FIB (P = 0.008) were signi cantly different between survivors and non-survivors in patients with sepsis. Meanwhile, lncRNA PKN2-AS1 level (P < 0.001, Fig. 5a) and lncRNA AC068888.1 level (P<0.001, Fig. 5b) were signi cantly increased in non-survivors compared with survivors. The other variables were not signi cantly different between survivors and non-survivors. In addition, the likelihood ratio forward stepwise multivariate logistic regression analysis demonstrated that APACHE II score (P < 0.001), high PCT level (P = 0.032), number of organ dysfunction (P = 0.001), septic shock (P = 0.008), the highly absolute expression of lncRNA PKN2-AS1 (P = 0.019) and lncRNA AC068888.1 levels (P = 0.006) were independent risk factors for the poor prognosis of sepsis, whereas other variables exhibited no association with prognosis in patients with sepsis (Table 7).
Prognostic value of the lncRNA PKN2-AS1 and AC068888.1 levels for sepsis The ROC curve analysis exhibited signi cant predictive value for lncRNA PKN2-AS1 (AUC = 0.747) and AC068888.1 (AUC = 0.717) in distinguishing non-survivors from survivors (Fig. 5c), which were higher compared with its for Lac level (AUC = 0.694), but lower compared with those for APACHE II score (AUC = 0.806) and SOFA score (AUC = 0.778). At the optimal cut-off point of >342.13 for the serum absolute expression of lncRNA PKN2-AS1, the speci city and sensitivity were 60.0% and 81.2%, respectively. At the optimal cut-off point of >330.1 for the serum absolute expression of lncRNA AC068888.1, the speci city and sensitivity were 52.5% and 83.3%, respectively. At the optimal cut-off point of >5.7 for Lac level, the speci city and sensitivity were 45.5% and 88.5%, respectively. At the optimal cut-off point of >12.5 for SOFA score, the speci city and sensitivity were 46.3% and 94.8%, respectively. At the optimal cut-off point of >19.5 for APACHE II score, the speci city and sensitivity were 78.8% and 67.7%, respectively. The addition of either SOFA score alone (AUC = 0.778) or APACHE II score alone (AUC = 0.806) or both (AUC = 0.825) did not signi cantly improve the predictive ability to predict prognosis of septic patients. However, if SOFA score and APACHE II score were combined with lncRNA PKN2-AS1 and AC068888.1, the prognosis of sepsis patients was improved to a certain extent (AUC = 0.860) (Fig. 5c and Table 8).

Discussion
Sepsis is a medical emergency of serious organs dysfunction caused by the host's imbalanced and extreme in ammatory and immune response to an infection, which can lead to poor prognosis [1,24].
Common clinical symptoms are fever, leukocytosis, edema, and accumulation of in ammatory cells (neutrophils, macrophages, and monocytes) in various tissues and organs [25]. Septic shock, a more severe form of sepsis, is a subset of sepsis in which severe circulatory, cellular, and metabolic abnormalities are found with a greater risk of death than sepsis alone [26]. The clinical de nition of septic shock is that, on the basis of sepsis, vasopressors are required to maintain a mean arterial pressure of 65 mmHg or greater and serum lactic acid level greater than 2 mmol/L, despite adequate uid resuscitation [1][2][3]. Sepsis and its consequence of multiple organ dysfunction are still one of the dominating causes of morbidity and mortality in critically ill patients in ICU [27]. The latest guidelines for sepsis management indicated that early identi cation and diagnosis of sepsis is critical for early targetoriented treatment of sepsis and septic shock, helping to reduce mortality in patients with sepsis [28]. Therefore, the search for reliable biomarkers of sepsis is of great signi cance for the early diagnosis and prognosis of sepsis.
In this context, high-throughput transcriptome RNA sequencing has emerged, providing unprecedented insights into the study of the human genome. Sepsis is a multi-system affected condition, involving not only the early activation of in ammatory responses [29], but also major changes in non-immune pathways, such as cardiovascular, renal, autonomic nervous, hepatic, metabolic, and coagulation pathways, all of which are regulated by lncRNAs [1]. Studies showed that lncRNAs were involved in the pathological process of sepsis and sepsis-induced organ dysfunction [30], and identi ed them being as potential biomarkers and therapeutic targets. At present stage, numerous reports have proved that lncRNAs were involved in the occurrence and development of sepsis with crucial roles in modulating gene expression and signaling pathways in the serum of sepsis patients [31][32][33]. Meanwhile, Zheng et al. showed that lncRNAs regulated the in ammatory immune-related genes and may serve as potential diagnostic biomarkers for sepsis [34]. Furthermore, the study of Chen et al. demonstrated that the lncRNA MALAT1 could be used as a diagnostic marker and therapeutic target for sepsis by in uencing the p38 MAPK/NFκB signaling pathway through miR-125b related effects, and thereby exacerbating cardiac in ammation and dysfunction during sepsis [35].
In this study, human-rat homologous lncRNA PKN2-AS1 and AC068888.1 were selected as biomarker candidates of sepsis, which were screened from the rat sepsis model induced by LPS and veri ed clinically. Before this trial, lncRNA PKN2-AS1 and AC068888.1 were only associated with the prognosis of bladder cancer [36] and postoperative survival of patients with glioblastoma multiforme [37], respectively. Nevertheless, they were no reports related to other diseases, including sepsis. In homo sapiens, lncRNA PKN2-AS1 is located on chromosome 1 and its length was 3160-bp with the NONCODE transcript ID ENST00000645056.1. Likewise, lncRNA AC068888.1 is located in chromosome 12, and its length was 1849-bp with the NONCODE transcript ID ENST00000663863.1. To verify the sequencing results, the RT-qPCR was therefore performed in 176 expended pair samples. Gratifyingly, the absolute expression of serum lncRNA PKN2-AS1 and AC068888.1 in patients with sepsis were higher than those in the healthy controls, re ecting the involvement of those lncRNAs in sepsis. We speculate that the above-mentioned two lncRNAs may activate multiple in ammatory signaling pathways to regulate the expression of in ammation-related proteins, thus inducing systemic in ammatory response in patients with sepsis, resulting in higher expression in patients with sepsis. The ROC curve showed that the AUC of lncRNA PKN2-AS1 and AC068888.1 were 0.881 and 0.842, respectively. Compared with the conventional evaluation of sepsis score index or laboratory indicators, including PCT, CRP and Lac levels, as well as SOFA and APACHE II scores, the two lncRNAs have apparent advantages, which may be favorable candidates for early diagnosis of sepsis.
The CRP and PCT have been reported to be closely related to the diagnosis of infection and the severity of sepsis, but due to the low speci city, they can no longer be used as the only criteria for septic infection [38]. A multitude of in ammatory cytokines, including IL-6 and TNF-α, were closely related to the degree of tissue and organ damage during the occurrence and development of sepsis [39,40]. Current tools used to stratify the in ammation in sepsis included clinical severity scores, such as SOFA and APACHE II scores, as well as Lac level [41]. To verify the role of lncRNA PKN2-AS1 and AC068888.1 in the assessment of disease severity and in ammation, our clinical study found that those two lncRNAs were signi cantly correlated with SOFA, APACHE II scores and Lac level, but weakly correlated with PCT and CRP levels. The possible explanation is that, like other lncRNAs (such as MALAT1 and NEAT1), these two lncRNAs may promote the production of in ammatory factors through the NF-κB or TLRs pathway [42,43], aggravating the in ammation and organ dysfunction. In the diagnosis of sepsis, lncRNA PKN2-AS1 (AUC = 0.704) and AC068888.1 (AUC = 0.812) clearly distinguished patients with septic shock from those without septic shock. Furthermore, the lncRNA AC068888.1 showed a certain degree of superiority. Multivariate logistic regression analysis showed that high expression of AC068888.1 was an independent risk factor for shock in patients with sepsis as well as SOFA score and high Lac level, but the PKN2-AS1 was not associated with them. In the septic shock prediction model, the combination of the above independent risk factors well predicted the occurrence of septic shock (AUC = 0.882).
We further evaluated the prognostic values of lncRNA PKN2-AS1 and AC068888.1 in patients with sepsis, and found that the absolute expression levels of lncRNA PKN2-AS1 and AC068888.1 in non-survivors at 28 days were higher than those in survivors at 28 days. Multivariate logistic regression analysis showed that lncRNA PKN2-AS1 and AC068888.1 were independent risk factors for the prognosis of sepsis patients, and the patients with high expression of PKN2-AS1 and AC068888.1 had worse prognosis. In addition, lncRNA PKN2-AS1 and AC068888.1 well predicted the 28-day death risk of sepsis patients by the ROC curve. In the prognostic model, the combination of those two lncRNAs with the SOFA and APACHE II scores was able to better predict the poor prognosis of the patients with sepsis (AUC = 0.860). To verify the validity of the prognostic model, we used the K-M survival curve to represent the survival time of survival group and non-survival group. The P-values of K-M survival curves of those two lncRNAs were less than 0.001, indicating that our predicted model was strongly correlated with the survival outcome of patients with sepsis.
There were still some limitations in this study. First of all, the sample size was small, and it was a singlecenter study, so the sample size needs to be expanded. Secondly, the expression of those two lncRNAs was detected only once in patients with sepsis (within 24 hours after admission), so it is necessary to further elucidate the changes of those two lncRNAs in the course of disease and treatment. Third, only 28-day mortality in patients with sepsis was followed up, so it is necessary to extend the follow-up time and evaluate the long-term predictive values of those two lncRNAs. Finally, the speci c regulatory mechanism of those two lncRNAs in sepsis is not clear.