A Panel of Urine Derived Biomarkers to Identify and Distinguish Sepsis From Systemic Inammatory Response Syndrome

Background Sepsis is a fatal condition caused by infection. It is frequently dicult to distinguish sepsis from systemic inammatory response syndrome (SIRS), often resulting in poor prognoses and the misuse of antibiotics. Hence, highly sensitive and specic biomarkers are needed to differentiate sepsis from SIRS. Methods Urine samples were collected and segregated by group (sepsis group, SIRS group, healthy control group). iTRAQ was used to identify the differentially expressed protein among the three groups. The identied proteins were measured by ELISA from urine samples. Finally, all the acquired data was analyzed by SPSS.


Introduction
Sepsis was rst de ned in 1992 [1] as a systemic in ammatory response syndrome (SIRS) caused by con rmed infection (Sepsis 1.0). The signs and symptoms of sepsis were expanded in Sepsis 2.0 [2].
Sepsis is now de ned as organ dysfunction caused by systemic in ammatory response to pathogenic microorganisms, which can be fatal (Sepsis 3.0) [3]. In the last decade, the incidence rate of sepsis was 437 per 100,000 person-years and the incidence of severe sepsis (de ned by Sepsis 2.0 [2]) was 270 per 100,000 person-years in developed countries [4]. Fleischmann et al. [4] also inferred that there were 31.5 million sepsis and 19.4 million severe sepsis cases, worldwide, with 5.3 million deaths annually. The case-fatality rate could reach up to 30% in sepsis, 50% in severe sepsis and 80% in cases of septic shock [5]. In addition, the prevalence of sepsis and contribution of sepsis to all-cause mortality rates have been increasing in the last several years [6]. Severe sepsis in elderly patients was an independent risk factor for substantial and persistent new cognitive impairment and functional disability among survivors [7].
Delays in treatment and inappropriate antibiotic therapy would dramatically reduce the survival rates in septic shock [8,9]. Hence, the early diagnosis of sepsis is particularly important. At present, the diagnosis of sepsis is di cult and complicated. Although there are clinical guidelines and many laboratory tests, e.g. C-reactive protein (CRP) test, procalcitonin (PCT) test, white blood cell (WBC) count, etc., to diagnose the sepsis, the speci city is unsatisfactory. Blood culture remains the conclusive evidence for the diagnosis of sepsis, but the sensitivity of blood culture is very low and blood culture is time-consuming, usually resulting in the delay of diagnosis. On the other hand, SIRS patients could be misdiagnosed with sepsis, which leads to antibiotic misuse and possible selection for drug resistance, because the symptoms and signs of SIRS are very similar to sepsis when the blood culture is negative. Biomarkers refer to measurable and quanti able biological parameters which could be molecules, genes, proteins or other characteristics. There are hundreds of biomarkers, but useful biomarkers against sepsis are only a small fraction [10], e.g. CRP, PCT, serum amyloid A (SAA), triggering receptors expressed on myeloid cells-1 (TREM-1) and others. Biomarkers are a promising way to diagnose sepsis which will facilitate early and accurate diagnosis, forecast organ dysfunction and assist in de ning the appropriate therapeutic plan [11]. Human urine contains thousands of proteins [12,13] and extracellular vesicles [14] which could be good resources for biomarkers. Isobaric tags for relative and absolute quanti cation (iTRAQ) is a mass spectrometry (MS) based relative proteomic quanti cation method, utilizing the derivatization of primary amino groups in intact proteins and isobaric tags for different peptide fragments [15,16]. iTRAQ can be used to screen the differentially expressed proteins among eight samples simultaneously and it is highly sensitive and speci c [17]. In this study, we sought to identify new diagnostic biomarkers of sepsis in urine, utilizing iTRAQ, and verify the biomarkers using enzyme linked immunosorbent assays (ELISA). In addition, we explored the diagnostic value of the combined biomarkers through logistic regression in order to enhance the diagnostic sensitivity and speci city to sepsis and help clinicians determine the appropriate therapeutic strategy.

Urine samples and data collection
This study was performed at the Second A liated Hospital of Chongqing Medical University. The procedures used in this study were in accordance with the 2008 Declaration of Helsinki and approved by the Ethics Committee of the Second A liated Hospital of Chongqing Medical University (Grant no. 201916). The urine samples and patient data came from the Adult Multidisciplinary Intensive Care Unit (ICU) and Infection department. All patients or their bailors were informed about the study procedures, risk and privacy policy and written consent was signed by each participant. Urine samples were collected from patients in the sepsis group, SIRS group and healthy control group. Consecutive patients with sepsis or SIRS were enrolled and the healthy control group were enrolled by recruiting. Sepsis inclusion criteria [3] were: (1) Suspected or proven infection, (2) Sequential [sepsis-related] Organ Failure Assessment (SOFA) score or 2 or more points. The speci c criteria of the SOFA score are listed in Table 1 [3,18] and the effectiveness of the SOFA was assessed [19]. SIRS inclusion criteria [1] were: (1)Temperature >38°C or <36°C; (2) heart rate >90/min; (3) respiratory rate >20/min or PaCO 2 <32mmHg; (4) white blood cell count >12´10 9 /L or <4´10 9 /L or >10% immature bands in peripheral blood. The patients who were diagnosed with SIRS must have presented with two or more of the four criteria. Healthy control inclusion criteria were that the healthy volunteer who did not suffer from infection disease or take antibiotics. Exclusion criteria were: (1) immune de ciency, (2) autoimmune diseases, (3) refusal to take part in the study or refusal to provide signed, written consent. The urine samples were collected as soon as the clinical diagnosis was made and the interval time was more than four hours from the patients' last micturition.
The urine was collected in 50 mL centrifuge tube directly from the bladder and no protease inhibitor was added to the samples. The collected samples were transported in an ice-water mixture. These samples were centrifuged at 2000g for 10min to separate the cellular or tissue debris. Finally, the samples were divided to ve 2.0 ml Eppendorf tubes and stored at -80℃ until analysis [20]. The patient data were collected through the electronic medical system of the Second A liated Hospital of Chongqing Medical University.
PaO2, partial pressure of oxygen; FIO2, fraction of inspired oxygen; MAP, mean arterial pressure; a Catecholamine doses are given as μg/(kg×min) for at least 1 hour.

Materials and measurement procedure
Protein extraction from urine and iTRAQ procedure We collected 10 samples from each of the groups to perform a preliminary analysis of the differentially expressed proteins among the three groups. The samples were pooled by group (sepsis, SIRS, and healthy control) and the proteins were precipitated using two-fold volumes of cooled acetone (2h) [13].
Finally, the three samples were centrifuged at 2,5000g at 4℃ for 15 min and the supernatants were discarded. This procedure was repeated twice.
The precipitates were air-dried and dissolved at room temperature in an appropriate amount of lysis buffer using an ultrasonic homogenizer. Finally, the solutions were centrifuged at 2,5000g at 4℃ for 15 min and the supernatants were collected. The Bradford method was used to measure the protein concentrations.
An iTRAQ 8 Plex Reagent Kit (Applied Biosystems, Foster City, CA, USA) was used to label the protein. The prepared protein was precipitated, re-dissolved, alkylated, cysteine-blocked and digested following the iTRAQ Kit instructions and the study by Wisniewski [22]. The Sepsis group was labelled with tag 113, the SIRS group was labelled with tag 119 and the healthy control group was labelled with tag 121. The labelled peptide solutions were pooled and freeze-dried under vacuum prior to further analysis.
The peptides were fractioned as previously described [23,24]. The prepared peptides were analyzed on the TripleTOF 5600 system coupled to an Eksigent NanoLC-2D system (AB Sciex, Framingham, MA, USA) and each of the samples was analyzed twice. The data was processed by ProteinPilot V2.0 (Applied  Table 2. The duplicate standards or samples were set and mean absorbance of duplicate wells was calculated. The concentration of urinary proteins (CRP, LRG1, SAA) was normalized to urinary creatinine (u-Cr) and expressed as protein/u-Cr in μg/mmol to adjust for individual differences and potential kidney injury [25,26].  ). There were no statistically signi cant differences in the distributions of age or gender among the three groups. The distribution of WBC count and neutrophil percentage were statistically similar and could not distinguish the sepsis and SIRS patients in this study. In addition, the sepsis patients presented higher urinary creatinine levels and pulse rates compared to the healthy control group. Patient data is summarized in Table 3. iTRAQ identi cation of differentially expressed proteins iTRAQ identi ed 11278 and 14904 proteins in duplicate analyses. There were 1970 and 1985 proteins with unused ProtScore > 1.3. Among these proteins, there were 114 and 118 with fold change> 2 between the sepsis group and healthy control group. There were 69 proteins which were detected in the both iTRAQ procedures. After eliminating the proteins with P values > 0.05, there were 45 proteins meeting the screening criteria. The ow chart is presented in Figure 1. CRP, LRG1, SAA were included in the 45 proteins. These three biomarkers were further analyzed by ELISA and their iTRAQ information are listed in Table 4. Quantitative and statistical analysis The concentrations of uCRP, uLRG1 and uSAA are shown in Table 3. Considering the variability of glomerular ltration rate (GFR), 24-hours urinary volume and glomerular ltration barrier among the different subjects, the concentrations normalized to urinary creatinine levels were also listed and expressed as uCRP/u-Cr, uLRG1/u-Cr and uSAA/u-Cr in μg/mmol and were used for further statistical analyses. The normality of acquired data was tested and we found that not all the data followed a Gaussian distribution. Although efforts were made to transform the data to a Gaussian distribution by converting initial data to the square root or logarithm, the data did not conform. So the data were presented as medians (25 percentile%, 75% percentile) and non-parametric tests were used to analyze the statistical characteristics. The Spearman's rank correlation coe cient was calculated ( Table 5). The serum CRP was strongly correlation with the urine CRP, which may indicate that the urinary CRP came from the blood. In addition, the three biomarkers (uCRP, uLRG1, uSAA) were pairwise correlated, which may indicate that they had similar status during the infection process. The median values of sepsis patients were highest, the SIRS patients were second and the healthy control group were lowest (P <0.001). When these biomarkers were adjusted to urine creatinine, the same conclusion could be inferred too and their values are presented in Figure 2. For the adjusted concentrations and sCRP, receiver operating characteristic curves (ROC) were drawn by SPSS 25 (Figure 3). The area under the curve values were 0.878 (uCRP/u-Cr), 0.874 (uLRG1/u-Cr), and 0.849 (uSAA/u-Cr), 0.891 (sCRP). The Youden index (sensitivity + speci city -1) was used to determine the best cut-off value. The sensitivity, speci city, cutoff value and area under the curve values are listed in Table 6. The sCRP had the highest diagnostic e ciency. The diagnostic difference was small among uCRP/u-Cr, uLRG1/u-Cr and uSAA/u-Cr. The adjusted concentrations (uCRP/u-Cr, uLRG1/u-Cr and uSAA/u-Cr) of the sepsis and SIRS groups were transformed to binary variables based on whether they were greater than the cut-off value and further analyzed by logistic regression. Forward stepwise (likelihood ratio) method was employed to process the data. The Omnibus test was used to verify the statistical signi cance (P<0.001) and Hosmer and Lemeshow test showed the model had taken full advantage of the acquired data (P=0.712). The regression coe cient and P values are listed in Table 7. According to the regression results, all three biomarkers were included in the regression model. If the patients' parameters exceeded the cut-off value, the corresponding risk of sepsis would increase 9.913-fold (uCRP/u-Cr), 15.936-fold (uLRG1/u-Cr) and 12.793-fold (uSAA/u-Cr) compared to SIRS patients. The risk score of sepsis is Logit (P) =2.294*X 1 +2.734*X 2 +2.549*X 3 -3.714 (X 1 =1, if uCRP/u-Cr >746.1 μg/mmol; X 1 =0, if uCRP/u-Cr>746.1 μg/mmol; X 2 =1, if uLRG1/u-Cr >1174.8μg/mmol; X 2 =0, if uLRG1/u-Cr <1174.8μg/mmol; X 3 =1, if uSAA/u-Cr >4.4μg/mmol; X 3 =0, if uSAA/u-Cr <4.4μg/mmol). According to the Logit(P), the prediction probabilities were calculated by P=e Logit(P) /(1+e Logit(P) ) and its ROC was also drawn (Figure 4). The area under the curve was 0.937. The cut-off of Logit(P) was 0.514. Its sensitivity and speci city were 0.906 and 0.896, respectively. This indicated that sensitivity and speci city would be greatly enhanced when diagnosis was made by combining the panel of biomarkers.

Discussion
Because the symptoms of sepsis and SIRS are quite similar, it is frequently di cult to distinguish a septic infection from SIRS. After Sepsis 3.0 was issued, the situation become more di cult, in part because patients suffering from infection-related SIRS, but without organ dysfunction, can be confused with patients having SIRS without sepsis. Biomarkers, and especially acute-phase proteins, could be key to distinguishing sepsis from SIRS, but their speci city has been unsatisfactory, making it necessary to combine multiple biomarkers in order to enhance overall speci city [27,28]. The three biomarkers identi ed in the present study were all acute phase reaction proteins [29][30][31][32]. CRP was identi ed in the 1930s and is synthesized in the liver and released in response to infection, trauma and immune disorders. Mean sCRP levels were 98mg/L in sepsis patients and 70mg/L in SIRS patients [33]. Most recently, bacterial infection is suspected when sCRP is greater than 100mg/L in the clinical context. Conversely, absence of severe bacterial infection is indicated when sCRP is less than 20mg/L [30]. In the present study, we rst veri ed the existence of CRP in urine and uCRP/u-Cr was strongly relation to sCRRP.
Although the diagnostic e ciency of uCRP/u-Cr was slightly lower than sCRP, uCRP/uCr could still be a good substitution for sCRP considering the convenience of specimen collection, strong relation and its noninvasiveness. Leucine rich α-2 glycoprotein 1 (LRG1), rst identi ed in 1977 [34], is synthesized by hepatocytes, stored in the neutrophils [35] or myeloid cells [36], and released to serum when the body is in acute phase stimulation by bacterial or viral infection [37]. LRG1 belongs to the leucine rich repeat (LRR) protein family and plays a role in protein interactions, innate immunity, platelet aggregation and angiogenesis. LRG1 is usually used as a biomarker of tumors, appendicitis, diabetes complications, and in ammatory disease [38][39][40][41]. According to our measurement results, uLRG1/u-Cr of sepsis patient was obviously higher in sepsis patients than in either SIRS patients or healthy volunteers (P<0.001), indicating that urinary LRG1 is a promising biomarker in diagnosing sepsis and distinguishing sepsis from SIRS. Considering the lower sensitivity (0.774) and higher speci city (0.854), LRG1 would be a good choice to combine with other biomarkers. Serum amyloid A protein is also a well-known acute phase protein produced by liver. Its concentration is very low (1 μg/ml) in healthy people, but can increase dramatically (100-fold or more) when human body is in acute phase [42]. Human SAA consists of 4 isotypes (SAA1, SAA2, SAA3,SAA4) [43], but sequence identity of SAA1 and SAA2 is more than 93% and SAA1 is the main isotype in serum [44]. The SAA ELISA Kits we employed in the current study were not sensitive to the SAA isotypes, and the concentration represented the concentration of total SAA. SAA is regulated by cytokines (Interleukin-1β, Interleukin-6, Tumor necrosis factor, etc.) and regulate cytokines (Interleukin-23, Interleukin-33, Interleukin-10, etc.) which is summarized as the cytokine-SAA-chemokine network [45].
Considering its interaction with cytokines, SAA was deemed to take part in in ammatory diseases, angiogenesis and tumor growth as do other acute phase proteins [46][47][48]. In our study, the sensitivity and speci city of uSAA/u-Cr were between the uCRP/u-Cr and uLRG1/u-Cr. It was also the rst time the value of uSAA to sepsis diagnosis and differential diagnosis was veri ed.
Although the three biomarkers were normalized to urinary creatinine in the statistical process, adjusting concentration had only a small in uence on the statistical results and no in uence on the nal conclusion. When the panel of biomarkers were combined, the sensitivity to sepsis was increased to 0.906 and speci city was increased to 0.896, which greatly exceeded any single biomarker. The area under ROC curve was 0.937, which indicated favorable diagnostic e ciency. Serum CRP data was acquired through the medical electronic system, but the serum LRG1 and SAA were not available by routine examination in the Second A liated Hospital of Chongqing Medical University. Hence serum LRG1 and SAA data was lacking in our study. The correlation between serum LRG1 and urinary LRG1 was not veri ed, and neither was the correlation of serum SAA and urinary LRG1. According to the correlation between serum CRP and urinary CRP, we inferred that urine CRP came from serum CRP. Considering the similar mechanism of protein excretion in kidney, we hypothesized urinary LRG1 and SAA also came from blood. There were only 101 patients included in our study, and it would be necessary to recruit more patients to further verify the e ciency of the combined biomarkers. In addition, the diagnostic value of LRG1 to sepsis was not fully investigated, and additional research is necessary to further understand the diagnostic signi cance of LRG1.
Conclusions C-reactive protein, serum amyloid A protein and leucine-rich alpha glycoprotein-1 exist in the urine. The urinary C-reactive protein comes from blood and may be a substitution of serum C-reactive protein in sepsis diagnosis. Urine SAA and leucine-rich alpha glycoprotein-1 were also favorable biomarkers in sepsis diagnosis. The sensitivity and speci city of differential diagnosis between sepsis and systemic in ammatory response syndrome could be greatly improved when the three biomarkers are combined. JH was responsible for statistical analysis. YT wrote the paper. HR and YXY edited the clinical theses. All the authors have approved the nal manuscript and declare that there exist no con icts among them.