This is the accepted version of the article: Papafilippou L., Claxton A., Dark P., Kostarelos K., Hadjidemetriou M.. Protein corona fingerprinting to differentiate sepsis from non-infectious systemic inflammation. Nanoscale, (2020). 12. : 10240 - . 10.1039/d0nr02788j. Available at: https://dx.doi.org/10.1039/d0nr02788j 1 Nanoparticle protein corona-enabled biomarker 2 discovery to differentiate bacterial sepsis from non- 3 infectious acute systemic inflammation 4 Lana Papafilippou,a Andrew Claxton,b Paul Dark,b,c Marilena Hadjidemetriou*a and 5 Kostas Kostarelos*a 6 7 a Nanomedicine Lab, Faculty of Biology, Medicine & Health, AV Hill Building, The University of Manchester, Manchester, 8 M13 9PT, UK 9 10 bCentre for Acute Care Trauma, Manchester Academic Health Science Centre, Health Innovation Manchester, Division 11 of Critical Care, Salford Royal NHS Foundation Trust, Greater Manchester, UK 12 13 cDivision of Infection, Immunity and Respiratory Medicine, Faculty of Biology, Medicine & Health, AV Hill Building, 14 The University of Manchester, Manchester, M13 9PT, UK. 15 16 17 18 19 _______________________________________ 20 * Correspondence should be addressed to: marilena.hadjidemetriou@manchester.ac.uk; 21 kostas.kostarelos@manchester.ac.uk 22 23 24 25 26 27 28 29 30 31 1 32 Abstract 33 34 Rapid and accurate diagnosis of sepsis remains clinically challenging. The lack of specific 35 biomarkers that can differentiate sepsis from non-infectious systemic inflammatory diseases 36 often leads to excessive antibiotic treatment. Novel diagnostic tests are urgently needed to 37 rapidly and accurately diagnose sepsis and enable effective treatment. Despite investment on 38 cutting-edge technologies available today, the discovery of disease-specific biomarkers in blood 39 remains extremely difficult. The highly dynamic environment of blood restricts access to vital 40 diagnostic information that can be obtained by proteomic analysis. Here, we employed clinically 41 used lipid-based nanoparticles (AmBisome®) as an enrichment platform to analyze the human 42 plasma proteome in the setting of sepsis. We exploited the spontaneous interaction of plasma 43 proteins with nanoparticles (NPs) once in contact, called the ‘protein corona’, to discover 44 previously unknown disease-specific biomarkers for sepsis diagnosis. Plasma samples obtained 45 from non-infectious acute systemic inflammation controls and sepsis patients were incubated ex 46 vivo with AmBisome® liposomes, and the resultant protein coronas were thoroughly 47 characterised and compared by mass spectrometry (MS)-based proteomics. Our results 48 demonstrate that the proposed nanoparticle enrichment technology enabled the discovery of 67 49 potential biomarker proteins that could reproducibly differentiate non-infectious acute systemic 50 inflammation from sepsis. This study provides proof-of-concept evidence that nanoscale-based 51 ‘omics’ enrichment technologies have the potential to substantially improve plasma proteomics 52 analysis and to uncover novel biomarkers in a challenging clinical setting. 53 54 Keywords: sepsis, infectious disease, inflammation, nanomedicine, biomarkers, liposomes 55 56 57 58 59 60 61 62 63 64 65 66 2 67 Introduction 68 69 Sepsis is defined as a life-threatening organ dysfunction triggered by a dysregulated 70 immune response to an infection.1 It is a leading cause of mortality, accounting for more than six 71 million deaths each year worldwide.2 Delays in diagnosing sepsis and initiating appropriate 72 antimicrobial treatments are associated with higher mortality.3 Hence, there is an unmet need for 73 the development of biomarkers to rapidly diagnose sepsis and monitor its progression. 74 A significant challenge in the clinical setting is the diagnostic uncertainty in differentiating 75 patients with sepsis in a highly heterogeneous, critically ill patient population. Microbiological 76 cultures are the current gold standard for identifying causative pathogen phenotypes and guiding 77 antimicrobial treatment,4 however they are limited by a lack of sensitivity and by long incubation 78 times (up to 72 hours).5 Given the high mortality rate associated with delayed treatments and the 79 lack of specific diagnostic tools, broad-spectrum antimicrobial treatment is recommended in all 80 patients suspected of developing sepsis.6 As a result, broad-spectrum antimicrobials are 81 frequently administered to patients with acute sterile inflammation or viral infections, contributing 82 to the emergence and propagation of antimicrobial resistant pathogens.7 Blood-circulating C- 83 reactive protein and procalcitonin biomarkers are increasingly used in routine clinical practice to 84 help identify unwell patients,8, 9 yet they lack diagnostic specificity as they are known to be 85 upregulated in other acute inflammatory disorders.10, 11 86 Considering the complexity of the sepsis syndrome and the molecular pathways 87 underlying immune responses that can also arise from non-infectious diseases,12 the use of a 88 single diagnostic biomarker is unlikely to offer the required specificity and sensitivity.13 To date, 89 the lack of protein biomarkers or biomarker combinations to identify sepsis is partially attributed 90 to the incapacity of currently available proteomics platforms to offer an in-depth analysis of the 91 plasma proteome.14 92 Although nanotechnology-based platforms have been developed to enable the 93 quantification of specific blood molecules, ongoing efforts are focusing on the discovery of 94 previously undetectable disease-specific biomarkers.15 We have previously exploited the 95 spontaneous interaction of intravenously administered nanoparticles with proteins, named the 96 ‘protein corona’,16, 17 to capture and amplify low molecular weight and low abundant proteins 97 from the blood circulation of tumour-bearing mice and ovarian carcinoma patients.18, 19 The 98 protein corona formed around intravenously injected clinically-used liposomes was found to be 99 enriched with disease-specific proteins which could not be detected by conventional plasma 100 proteomic analysis. The elimination of background, highly abundant proteins and the 3 101 identification of only the nanoparticle-bound proteins make this nano-platform technology very 102 promising for the discovery of novel diagnostic biomarkers and more studies are needed to 103 explore the prospective applications. 104 In the present study, we aimed at further exploring the potential use of the proposed 105 nano-scavenger tool in the highly complex clinical challenge of differentiating sepsis from non- 106 infectious acute systemic inflammation in humans. The occurrence of activated systemic 107 inflammation pathways in both conditions adds another level of complexity in comparison to our 108 previous work focusing on the discovery of cancer biomarkers.18 To prove our hypothesis, we 109 incubated the commercially available amphotericin B-containing liposomes (AmBisome®), 110 clinically used to treat serious, life-threatening fungal infections,20 with plasma samples obtained 111 from patients with confirmed sepsis and from phenotypically similar sterile tissue injury patients 112 with systemic inflammatory response syndrome (SIRS) (Figure 1A). We chose AmBisome® as 113 the preferred liposomal formulation because of its clinical use in high-risk patients with 114 suspected sepsis (most notably in the setting of immunosuppression in hematological oncology, 115 cell and tissue transplantation).20 Our results demonstrated that comprehensive comparison of 116 the resultant protein coronas led to the identification of 67 potential biomarker proteins that could 117 reproducibly distinguish non-infectious acute systemic inflammation from sepsis. 118 119 Results 120 121 Physicochemical characterisation of bare and corona-coated Amphotericin B-intercalated 122 liposomes (AmBisome®). Amphotericin B-intercalated liposomes (AmBisome®) were 123 physicochemically and structurally characterised prior to and after their ex vivo incubation with 124 human plasma samples obtained from SIRS control group (n=7) and sepsis patients (n=12). 125 Patient clinical characteristics and blood cell culture results are summarized in Tables S1 and 126 S2. Liposomes were also incubated with plasma obtained from healthy donors (n=12) as a 127 control. AmBisome® liposomes were allowed to interact with human plasma proteins for 1 hour 128 followed by a two-step purification protocol, for the separation of corona-coated liposomes from 129 unbound and weakly bound plasma proteins, as we have previously described.18, 19, 21, 22 130 Dynamic light scattering (DLS) and negative stain transmission electron microscopy 131 (TEM) were conducted before and after the ex vivo incubation of liposomes with human plasma 132 to assess their physicochemical properties prior to and after corona formation (Figures 1B, S1 133 and 1C). The physicochemical characteristics of the commercially available liposomal 4 134 formulation AmBisome® employed in this study are summarized in Figure 1B. Bare liposomes 135 displayed a mean hydrodynamic diameter of 102.6 nm, a negative surface charge of -54.3 mV 136 and low polydispersity index values (0.093) indicating a narrow size distribution (Figure 1B). 137 TEM imaging revealed a well-dispersed liposomal population with a homogenous size 138 distribution correlating that of DLS measurements (Figure 1C). 139 Dynamic light scattering measurements of corona-coated AmBisome® liposomes 140 demonstrated that their size was not significantly affected upon corona formation, while their 141 surface charge was shifted towards less negative values (Figures 1B and 1S, Table S3) 142 indicating their interaction with protein molecules. TEM confirmed the presence of protein 143 molecules onto the surface of AmBisome® liposomes and revealed that the recovered corona- 144 coated liposomes remained intact post-incubation with human plasma and purification, showing 145 no structural differences compared to the bare NP (Figure 1C). 146 147 Quantitative and qualitative comparison of the ex vivo protein coronas formed onto 148 AmBisome® liposomes. To quantitatively compare the total amount of protein adhered onto 149 AmBisome® liposomes in the three different conditions under investigation, we calculated the 150 protein binding value (Pb), expressed as the amount of proteins in μg per each μmole of lipid. As 151 shown in Figure 2A, the total amount of proteins adsorbed onto liposomes after their incubation 152 with plasma samples obtained from sepsis patients was significantly higher compared to the 153 amount of proteins adsorbed onto liposomes after their incubation with plasma samples 154 obtained from healthy volunteers (** indicates p<0.01 (p=0.0068) using the non-parametric 155 Kruskal-Wallis test). This clearly indicates that the formation of protein corona onto AmBisome® 156 liposomes is quantitatively influenced by human sepsis condition. This observation is in 157 agreement with our previous studies, showing that the total amount of protein molecules 158 adsorbed onto the NPs surface is determined by the presence or absence of tumorigenesis, 159 reflecting thus the ongoing pathophysiological alterations in blood proteome.18 160 Interestingly, the average Pb values observed post-incubation of AmBisome® liposomes 161 with plasma samples obtained from SIRS patients was similar to that of sepsis patients. This 162 reflects the clinical challenge in differentiating sepsis from other non-infectious events, such as 163 acute trauma and burns. As shown in Figure 2A, the SIRS group exhibited greater variations 164 among the individual patients Pb values compared with the healthy controls and the sepsis 165 patients group. This was expected considering that SIRS patients suffered from a wide range of 166 acute inflammation-related pathological conditions other than infection. It is worth mentioning 167 that the total amount of protein adsorbed onto the surface of AmBisome® liposomes was higher 5 168 compared to what we have previously observed for PEGylated doxorubicin-encapsulated 169 liposomes (Caelyx®) of similar size.19 This is in agreement with the literature suggesting that 170 non-PEGylated surfaces tend to adsorb a higher amount of proteins once in contact with 171 biological fluids.23 172 To investigate whether the higher amount of protein adsorbed onto AmBisome® 173 liposomes could offer a comprehensive coverage of the plasma proteome, corona proteins 174 associated with liposomes were separated by SDS-PAGE and visualized by Imperial Protein 175 stain, as illustrated in Figure 2B. The elimination of highly abundant proteins allowed the 176 enrichment of the low MW blood proteome, as we have previously reported (Figure 2B).18, 19, 22 177 Distinct protein bands were profiled in the case of corona samples, while the protein pattern of 178 plasma control verified the masking effect of albumin (Figure 2B). In agreement with the BCA 179 assay data, the amount of protein adsorbed onto AmBisome® post-incubation with plasma 180 obtained from sepsis patients was higher in comparison to the total amount adsorbed post- 181 incubation with plasma obtained from healthy volunteers. Similarly, the considerable variation in 182 the total amount of liposome-bound protein (Figure 2A) was reflected in the corona profiles of 183 SIRS patients (Figure 2B), indicating the high heterogeneity of this group. 184 185 The NP protein corona-enabled biomarker discovery to differentiate sepsis from non- 186 infectious acute systemic inflammation. To identify potential biomarker proteins that can 187 differentiate sepsis from non-infectious acute systemic inflammation, we comprehensively 188 characterised and compared the resultant protein coronas by label-free liquid chromatography– 189 mass spectrometry (LC-MS/MS). 190 Progenesis QI data analysis (version 3.0; Nonlinear Dynamics) enabled us to statistically 191 compare the protein corona profiles of healthy controls, SIRS patients and sepsis patients. Raw 192 data generated from LC-MS/MS analysis were processed and the mean normalized abundance 193 of each group, the relative protein expression (fold-change) and the reliability of measured 194 differences (ANOVA, p value) were calculated (Figure 3, Figure S2 and Tables 1, S4, S5 and 195 S6). As shown in Figure S2 and Table S5, common proteins between the three different groups 196 displayed quantitative differences and enhanced our hypothesis that proteomic analysis of the 197 liposomal protein coronas unravels differences between healthy and diseased states. In order to 198 identify differentially expressed proteins between healthy controls and the sepsis patients, we 199 compared the protein patterns of the ex vivo formed coronas. Results were filtered to present a p 200 value <0.05 and interestingly, n=135 proteins were found to be differentially expressed, of which 6 201 n=88 were upregulated and n=47 were downregulated in sepsis patients samples (Figure S2B 202 and Table S6). 203 Considering that the prevalent challenge in the clinic is to distinguish between sepsis and 204 non-infectious acute systemic inflammation, we performed further analysis focused on the 205 identification of differentially abundant proteins between SIRS controls and sepsis patients. As 206 shown by the normalized abundance values of Figure 3A and Table S4, quantitative differences 207 were observed between the two groups with n=67 proteins being differentially expressed (n=34 208 upregulated; n=33 downregulated), (Figure 3B, Figure S3 and Table 1). Interestingly, the 209 clinically-used CRP protein biomarker was not found to be differentially abundant between 210 sepsis patients and SIRS controls (p value=0.29, max fold-change=1.97, Table 1 and S4), 211 which is consistent with its acknowledged lack of diagnostic specificity.10 We also performed 212 Principal Component Analysis (PCA) to assess whether these 67 identified on the surface of 213 AmBisome® proteins were able to discriminate sepsis from SIRS. Strikingly, PCA analysis 214 unveiled the formation of two distinct clusters (Figure S3). 215 Ingenuity Pathway Analysis (IPA® QIAGEN Bioinformatics) was then performed to 216 investigate if the corona proteins found to be differentially expressed between sepsis and SIRS 217 groups (n=67) have been previously associated with bacterial sepsis and/or infection. Disease 218 and function IPA search revealed the association of n=7 corona proteins with bacterial infection 219 pathways (Figure 3C and Table S7): serum amyloid P component (APCS); B-cell 220 lymphoma/leukemia 10 (BCL10), CD5 antigen like (CD5L), gamma-aminobutyric acid type A 221 receptor alpha1 subunit (GABRA1), immunoglobulin heavy constant gamma 3 (IGHG3), 222 immunoglobulin kappa variable 2-24 (IGKV2-24), lipopolysaccharide binding protein (LBP). 223 However, none of the above differentially abundant proteins was previously described in the 224 literature as a potential biomarker for sepsis. 225 Noteworthy, serum amyloid P has been previously described to act as a pathogen 226 recognition receptor due to its binding affinity towards microbial surface components.24-26 227 Another study revealed the key role of APCS in complement-mediated immunity against 228 Streptococcus pneumoniae, which suggests that APCS is a significant component of the innate 229 immunity against this pathogen.27 In addition, B-cell lymphoma/leukemia 10 (BCL10) has been 230 reported to induce the activation of lymphocytes,28, 29 as well as to be elevated following 231 exposure of intestinal epithelial cells to LPS.30 Similar to APCS and BCL10, CD5 antigen like 232 (CD5L) has been shown to be implicated in monocyte inflammatory pathways in response to the 233 surface components of bacteria LPS and lipoteichoic acid (LTA).31 More interestingly, a recent 234 clinical study has shown that CD5L serum levels of 150 sepsis patients were significantly more 7 235 elevated on the day of admission to intensive care unit (ICU) than the levels of PCT and CRP 236 compared to ICU controls and healthy controls.32 As far as lipopolysaccharide binding protein 237 (LBP) concerns, the results has been so far controversial. Some studies have reported a 238 moderate to low diagnostic capacity for sepsis,33 while others demonstrated that LBP levels 239 were significantly higher in 97% of sepsis patients than healthy controls.34 The presence of these 240 proteins on the surface of liposomes suggests that the NP-protein corona mirrors the ongoing 241 pathophysiological alterations in plasma. 242 In order to provide some initial validation of the proteomic data obtained, we performed 243 enzyme-linked immunosorbent assay (ELISA) experiments using plasma samples obtained from 244 the same SIRS and sepsis patient groups. In agreement with our mass spectrometry data, 245 ELISA experiments demonstrated that the amount of the clinically used biomarker CRP, was not 246 significantly different between SIRS and sepsis groups (95% CI, AUC = 0.7619) (Figure S4). 247 Interestingly, serum amyloid P (APCS) and cathepsin G (CTSG), identified in the corona 248 samples obtained from SIRS and sepsis patients exhibited higher AUC values than CRP (95% 249 CI, AUC = 0.7937 for APCS and 95% CI, 0=9048 for CTSG). 250 Overall, we show that the use of a different from our previous studies18, 19 liposomal 251 formulation, AmBisome®, facilitates the enrichment of complex biofluids with low abundant 252 molecules, enabling thus the discovery of disease-specific biomarkers upon interaction with 253 plasma proteins. The above data suggest that proteomic analysis of the NP coronas uncovers 254 previously unknown potential biomarker proteins that can differentiate sepsis from non-infectious 255 acute systemic inflammation with higher specificity and sensitivity than the clinically used 256 biomarkers. 257 258 259 260 261 262 263 264 265 8 266 Discussion 267 268 There is an unmet need for the development of biomarkers to rapidly diagnose sepsis 269 and monitor its progression. The complexity of sepsis pathology and the lack of accurate 270 diagnostic tests impede early diagnosis which in turn impact on treatment decision-making.35 271 Currently, diagnosis of sepsis relies on the time-consuming process of blood culturing, often 272 associated with false negative results.4, 5 The diagnostic uncertainty in differentiating patients 273 with sepsis from those suffering from non-infectious acute systemic inflammation makes sepsis 274 the Achilles’ heel of health care.2 The clinically-used diagnostic and disease monitoring blood 275 marker, C reactive protein (CRP) is an indicator of acute-phase responses and thus is unable to 276 distinguish sepsis from other systemic inflammation-associated diseases.10, 11, 36 In addition, the 277 clinical signs and symptoms of sepsis in its early stages mirror those of non-infectious acute 278 systemic inflammation12 and this leads to antibiotics being administered to patients with acute 279 sterile inflammation or viral infection. To date, there is no biomarker or any diagnostic test with 280 the capacity to distinguish between sepsis and non-infectious acute systemic inflammation.37 281 Currently available proteomic techniques are capable of sampling a relatively small 282 fraction of the blood proteome which is mainly composed of highly abundant proteins. The 283 signal-to-noise issue, mainly caused by the albumin and immunoglobulins restricts access to the 284 vital diagnostic information that could be obtained.38-41 _ENREF_36Plasma immunodepletion 285 and fractionation methods are predominantly employed to tackle this issue, however their 286 extensive use leads to a significant loss of the low molecular plasma proteome along with the 287 highly abundant plasma proteins.39, 42 MS-based proteomics have significantly aided to the 288 discovery of several sepsis biomarkers and provided valuable information about the complex 289 molecular mechanisms underlining this condition.43 However, to the best of our knowledge, there 290 are only a few proteomic-based studies reporting the discovery of a single biomarker or a panel 291 of biomarkers that can differentiate sepsis from non-infectious acute systemic inflammation.37, 44 292 Even though there is an increasing evidence that nanoparticle-assisted discovery of 293 novel biomarkers by high-throughput proteomics can revolutionise the field of diagnostics,18, 19 294 nanoparticles have been so far utilised in the biomarker verification and validation phases to 295 boost immunoassays sensitivities and detect already known biomarker molecules.45, 46 296 Specifically in the case of sepsis, NPs have been successfully employed as contrast agents and 297 biosensors47 to facilitate the detection of either already known protein biomarkers (CRP, PCT),48 298 pathogenic DNA49-52 or bacterial cells53-56 by amplifying their signal and improving read-outs. 9 299 Among the emerging applications, the exploitation of the NP-protein corona for the discovery of 300 novel biomarkers in conjunction with high-throughput proteomics, has been recently proposed 301 by us17-19 and others.57, 58 In our previous work, we demonstrated that the spontaneous 302 interaction of intravenously injected clinically-used liposomes with plasma proteins allowed the 303 discovery of multiple proteins differentially abundant between healthy and tumour bearing 304 mice.18 It should be emphasised that the proposed liposome-based technology facilitates the 305 initial untargeted discovery step of the biomarker development pipeline and not the detection of 306 a specific biomarker molecule for clinical diagnosis. 307 In this study, we aimed to further explore and validate the use of the human ex vivo 308 nanoparticle-protein corona for the discovery of previously undetectable sepsis biomarkers. To 309 evaluate whether the human ex vivo protein corona could reveal differences in the complex case 310 of differentiating between two diseased states, we thoroughly characterised and compared the 311 protein corona patterns formed around the clinically-used Amphotericin B-intercalated liposomal 312 formulation (AmBisome®), post-incubation with plasma obtained from healthy donours, SIRS 313 patients and sepsis patients. 314 Our results here are in agreement with our previous findings in tumour-bearing mice and 315 humans,18, 19 showing that the protein corona changes both quantitatively (Figure 2A) and 316 qualitatively (Figure 2B and Figure S2) in the presence or absence of the disease. A 317 significantly higher total amount of protein was adsorbed onto the surface of liposomes after 318 their incubation with plasma from sepsis in comparison with healthy controls (Figure 2A). The 319 above differences in the protein corona patterns of healthy individuals compared with sepsis 320 patients reinforce our hypothesis that the NP-protein corona mirrors the ongoing 321 pathophysiological alterations in plasma proteome. Equally consistent with our previous 322 observations18, 19 were the gel electrophoresis results, demonstrating that the AmBisome® 323 protein corona is reproducibly enriched by low molecular weight proteins (Figure 2B), which 324 conventional proteomics are unable to sensitively detect. 325 Considering that sepsis pathology is a highly complex physiological process, there is a 326 prevalent concord that a panel of multiple biomarkers, rather than a single biomarker will be 327 needed for its accurate diagnosis.13, 59 The clinically challenging scenario of distinguishing non- 328 infectious acute systemic inflammation from sepsis,2 prompt us to investigate whether the NP- 329 protein corona platform can be utilised to identify biomarkers specific to sepsis. Comparative 330 analysis between the protein corona fingerprints of SIRS controls and sepsis patients unveiled 331 67 novel potential biomarkers, with n=34 and n=33 proteins being upregulated and 332 downregulated in sepsis, respectively (Figures 3A, 3B, S3 and Table 1). Especially notable was 10 333 that the discovered corona proteins presented max fold-change and p values much higher than 334 the clinically-used CRP protein biomarker. Indeed, CRP protein was not found to be differentially 335 expressed between SIRS controls and sepsis patients (p value>0.05). This observation is 336 consistent with the acknowledged lack of CRP diagnostic specificity10 and highlights the pressing 337 need for new diagnostic biomarkers. PCA analysis of the 67 identified protein corona biomarkers 338 revealed the formation of two distinct clusters, suggesting that these proteins can differentiate 339 between the two diseased states (Figure S3). 340 To date, there is not any clinically used biomarker that differentiates sepsis from non- 341 infectious acute inflammation. Even though, disease and function IPA search revealed the 342 association of n=7 corona proteins with bacterial infection-related pathways (Figure 3C), none of 343 the potential biomarker proteins discovered in this study (n=67) was found to be previously 344 proposed for sepsis diagnosis. This suggests that the NP-corona platform can potentially unravel 345 information about the human pathophysiology and the underlining mechanisms of human 346 bacterial sepsis. ELISA validation experiments demonstrated that the corona proteins APCS and 347 CTSG displayed greater sensitivity and specificity compared to the clinically-used biomarker 348 CRP (Figure S4). More work is needed to examine the role of the above-identified proteins in 349 bacterial sepsis and/or inflammation and their potential utility for the development of future 350 diagnostic tests. 351 In our previous work, we demonstrated that protein crorona analysis formed around 352 Doxil® liposomes enriches the identification of low abundant and low MW proteins allowing an in 353 depth analysis of the plasma proteome. [ref] Here, we explore the exploitation of protein corona 354 or biomarker discovery in a much more clinically challenging scenario. Our results reveal that 355 analysis of protein corona formed around AmBisome® liposomes allows mining of the blood 356 proteome and discovery of potential diagnostic biomarker panels in the presence of severe 357 blood inflammatory responses. AmBisome® liposomes were found to harvest disease-specific 358 molecules upon incubation with plasma, which paves the way towards further development of 359 this technology by using different types of NPs. It is worth mentioning that this study focuses on 360 the initial phase of biomarker discovery, which renders meticulous future exploration on the 361 verification and validation phases of the biomarker development pipeline important, however out 362 of the scope of this work. Clearly, further studies are needed using a higher number of human 363 clinical samples to determine the specificity and sensitivity of the discovered biomarkers that can 364 be used in the future as a meaningful and accurate panel in combination with clinical 365 assessment to diagnose sepsis. 366 11 367 368 Conclusion 369 370 In this study, we describe the formation of protein corona around the clinically-used liposomal 371 Amphotericin B formulation (AmBisome®). Our data reinforces our previous proposition to 372 exploit the nanoparticle-biomolecule corona in order to allow in depth analysis of the blood 373 proteome and to uncover potential biomarker proteins. To explore the adaptability of this nano- 374 scavenging tool to enrich disease-specific molecules with diagnostic potential in a more clinically 375 challenging scenario, we molecularly compared the ex vivo coronas formed upon incubation of 376 liposomes with plasma samples obtained from phenotypically identical non-infectious acute 377 systemic inflammation patients and sepsis patients. Despite the similar inflammatory responses 378 arisen in both conditions, proteomic comparison of the biomolecule coronas revealed 67 plasma 379 proteins that could reproducibly differentiate sepsis from non-infectious acute systemic 380 inflammation. The nanoparticle-corona platform has the potential to accelerate the development 381 of panels of biomarkers that can in combination with clinical evaluation rapidly and accurately 382 diagnose sepsis, limiting unnecessary antibiotic treatments. 12 383 Experimental 384 Ethical approvals of patient plasma samples. The patient samples were collected and stored at Salford 385 Royal NHS Foundation Trust following review and approval by the NHS National Research Ethics Service 386 Committee North West – Greater Manchester South (Ref#14/NW/1404). Sample collection from patients 387 was undertaken by designated members from Salford Royal NHS Foundation Trust’s research and 388 development team. All experiments were performed in compliance with the University of Manchester 389 policy on human plasma use and ethics. 390 391 Selection of patient samples. Suitable patient samples were identified via study case report forms. The 392 selection of patient samples was based on the clinical criteria for acute systemic inflammation and the 393 blood/urine culture results. For inclusion in the sepsis group, patients had to have blood culture evidence 394 of bacterial infection and a score of 2 or more out of 4 criteria for acute systemic inflammation.1 For 395 patients to be included in the non-infectious acute systemic inflammation group, they had to have 2 or 396 more clinical criteria for acute systemic inflammation, no clinical or microbiological laboratory evidence of 397 infection (negative blood/urine culture tests) and given no more than one dose of antibiotics. 398 399 Blood samples collection. Blood samples from patients had been collected after recognition of a score 400 of two or more (out of 4) clinical criteria for acute systemic inflammation, and immediately prior to antibiotic 401 administration. Samples were collected in commercially available anticoagulant-treated tubes (K2 EDTA 402 BD Vacutainer®). Plasma was then prepared by inverting the collection tubes to ensure mixing of blood 403 with EDTA and subsequent centrifugation for 15 minutes at 1500x g at 4 0C. Following centrifugation, 404 supernatant was immediately collected into labelled 1.5 mL screw-cap polypropylene tubes (Thermo 405 Fisher), and samples were maintained on dry ice while handling. Finally, samples were stored in a -80 0C 406 freezer and were thawed only before the incubations. 407 For the ex vivo protein binding study of healthy controls, 6 female and 6 male human samples were 408 supplied by Seralab, UK, Batches #BRH1401254 (H1), #BRH1401253 (H2), #BRH1401256 (H3), 409 #BRH1401259 (H4), #BRH1401252 (H5), #BRH1401257 (H6), #BRH1401243 (H7), #BRH1401247 (H8), 410 #BRH1401244 (H9), #BRH1401246 (H10), #BRH1401249 (H11) and #BRH1401248 (H12). Considering 411 the impact of the anticoagulant agent on the formation of the protein corona,60 healthy plasma samples 412 contained the same anticoagulant agent (K2 EDTA BD Vacutainer® tubes) as that described above for 413 the human clinical samples and were subjected to the same preparation protocol (centrifugation for 12 min 414 at 1300 rpm at 4 0C). Healthy human plasma samples were received on dry ice and were stored in a -80 415 0 C freezer upon arrival from Seralab, UK. Finally, samples were thawed only before the incubations. 416 417 Preparation of AmBisome® liposomes suspension. 2 vials of AmBisome® liposomal formulation 418 (HSPC:Cholesterol:DSPG:Amphotericin B, batch #006572D) were kindly donated by Salford Royal NHS 419 Foundation Trust’s research and development team. 12 mL of HEPES buffer solution (HBS, 150 mM NaCl 420 (Sigma Aldrich, S6191) and 20 mM HEPES (Sigma Aldrich, H3375), pH=7.4) were added to the vial and 421 the reconstituted suspension was gently and thoroughly mixed. The concentration of liposomes in the 422 reconstituted suspension was quantified by Stewart assay (see below) and was found to be ~41 mM. 423 424 Ex vivo incubation of AmBisome® liposomes with human plasma. The ex vivo protein corona was 425 allowed to form using AmBisome® liposomal formulation. AmBisome® was added into human plasma 426 obtained from healthy individuals (n=12), non-infectious acute systemic inflammation (SIRS) patients 427 (n=7) and sepsis patients (n=12). A concentration of 1 mM of liposomes (1 mM lipids/mL blood) was 428 chosen for incubation based upon a dose of 3 mg of Amphotericin B/kg body weight.61, 62 Thus, 24 μL of 429 AmBisome® were incubated with 820 μL of human plasma and 156 μL of HEPES buffer solution (HBS) to 430 reach 1 mL volume for 1 hour at 37 0C in orbital shaker (ThermoFisher, MaxQ™ 4450 Benchtop Orbital 431 Shaker) at 250 rpm, set to mimic in vivo conditions. 1 hour of ex vivo incubation was chosen because it is 432 the equivalent time for the infusion of AmBisome®. 433 434 Separation of corona-coated AmBisome® liposomes from unbound and weakly bound proteins. 435 Corona-coated liposomes were separated form excess plasma proteins following a two-step purification 436 protocol, including size exclusion chromatography and membrane ultrafiltration, as we have previously 437 described.18, 19, 21, 22 13 438 439 Size and zeta potential measurements using dynamic light scattering (DLS). The size and surface 440 charge of AmBisome® liposomes were measured by Intensity using a Zetasizer Nano ZS (Malvern, 441 Instruments, UK). For both size and ζ-potential measurements, samples were diluted with ultra-pure water 442 in 1 mL polystyrene disposable cuvettes and 1 mL disposable Zetasizer cuvettes, respectively. Size and ζ- 443 potential data was taken in three records. 444 445 Transmission Electron Microscopy (TEM). Liposomes were visualized with transmission electron 446 microscopy (FEI Tecnai 12 BioTwin) before and after their ex vivo interaction with human plasma proteins. 447 10 uL of each sample were diluted in 600 uL of ultra-pure water. A drop from each sample was placed 448 onto a Carbon Film Mesh Copper Grid (CF400-Cu, Electron Microscopy Science) and subsequently was 449 stained using aqueous uranyl acetate solution 1%. 450 451 SDS-PAGE electrophoresis. Proteins associated with 0.025 μM of liposomes were loaded in 4-20% 452 NovexTM Tris-Glycine Protein Gels (WedgeWellTM, ThermoFisher Scientific) along with a 10 – 180 kDa 453 Prestained Protein Ladder (PageRulerTM Prestained Protein Ladder 10 – 180 kDa, ThermoFisher 454 Scientific) and run through a 10 times diluted 10xNovex TM Tris-Glycine SDS Running Buffer 455 (ThermoFisher Scientific) in ultra-pure water for 25-40 minutes at 225 V and 125 mA until the proteins 456 reached the end of the gel. Gels were stained for 1 hour using the Imperial protein Gel Staining reagent 457 (Imperial protein stain, ThermoFisher Scientific) followed by a couple of washings with ultra-pure water for 458 1-2 days. 459 460 Quantification of recovered lipids and adsorbed proteins. Lipid concentration (mM) was quantified by 461 Stewart Assay using a Cary 50 Bio Spectrophotometer (Agilent Technologies), as we have previously 462 described.18, 19, 21, 22 The total amount of the adsorbed proteins onto the surface of AmBisome® was 463 quantified by BCA Protein assay kit (ThermoFisher Scientific), according to the manufacturer’s 464 instructions, using a plate reader (Fluostar Omega plate reader, BMG Labtech). Protein binding (Pb) 465 values, expressed as μg of protein/μmole of lipid were then calculated per patient and per group, 466 presented as the average ± standard deviation (healthy controls n=12, SIRS patients n=7 and sepsis 467 patients n=12). 468 469 Mass Spectrometry. Corona proteins (20 μg) were loaded in 10 % NOVEX Tris-Glycine Protein Gels 470 (WedgeWellTM, ThermoFisher Scientific) and run through a 10 times diluted 10xNovex® Tris-Glycine SDS 471 Running Buffer (ThermoFisher Scientific) in ultra-pure water for 2-5 minutes at 225 V and 125 mA until the 472 formation of a single band at the top of the gel. Gels were stained for 1 hour using the Imperial protein Gel 473 Staining reagent (Sigma Life Science) followed by a couple of washings with ultra-pure water for 1-2 days. 474 Gel bands were cut out in 1 mm small square pieces and placed into 96 perforated plate wells. Samples 475 were reduced with 10 mM dithiothreitol, alkylated with 55 mM iodoacetamide, washed with ammonium 476 bicarbonate and acetonitrile and digested by 12.5 ng/μL trypsin overnight at 37 0C. Following overnight 477 incubation, samples were extracted using 20 mM ammonium bicarbonate and 50 % acetonitrile, 5 % 478 formic acid, desalted and dried by vacuum centrifugation. Dried samples were reconstituted in 10 μL 5 % 479 acetonitrile 0.1 % formic acid and analyzed by LC-MS/MS using an UltiMate® 3000 Rapid Separation LC 480 (RSLC, Dionex Corporation, Sunnyvale, CA) coupled to a Q Exactive™ Hybrid Quadrupole-Orbitrap™ 481 (ThermoFisher Scientific, Waltham, MA) mass spectrometer. 482 483 Mass Spectrometry Data Analysis. RAW files were imported into Progenesis LC-MS software (version 484 3.0; Nonlinear Dynamics) with automatic feature detection enabled. A representative reference run was 485 selected automatically to which all other runs were aligned in a pair-wise manner. Automatic processing 486 was selected to run with applied filters for peaks charge state (maximum charge 5) and protein 487 quantitation method the relative quantitation using Hi-N with N=3 peptides to measure per protein. The 488 resulting MS/MS peak lists were exported as a single Mascot generic file and loaded onto a local Mascot 489 Server (Matrix Science, London, UK; version 2.5.1). The spectra were searched against the 490 SwissProt_2018_01 database (selected for Homo sapiens, 161629 entries) using the following 491 parameters: tryptic enzyme digestion with one missed cleavage allowed, peptide charge of +2 and +3, 492 precursor mass tolerance of 15 mmu, fragment mass tolerance of 8 ppm, oxidation of methionines as 493 variable modifications and carbamidomethyl as fixed modifications, with decoy database search disabled 14 494 and ESI-QUAD-TOF the selected instrument. Each search produced an XML file from Mascot and the 495 resulted peptides (XML files) were imported back into Progenesis LC-MS to assign peptides to features. 496 Data was filtered to present a score above 21 through the ‘refine identification’ tab of Progenesis QI 497 toolbox. A table of all identified features along with their normalized peptide and protein abundance in 498 each sample was generated. The max fold-change and p value (ANOVA) of each protein was then 499 calculated by Progenesis and data including the normalized abundance, p-value and max fold-change 500 were exported for further analysis. Finally, results were filtered to present a mean normalized abundance 501 of more than 50,000 in at least one of the three groups. 502 Mass Spectrometry data was further analyzed through the use of QIAGEN’s Ingenuity Pathway Analysis® 503 (IPA®, QIAGEN Redwood City, www.qiagen.com/ingenuity). Diseases and functions IPA tool was used to 504 identify proteins involved in inflammation and bacterial infection pathways. The biomarker overlay IPA tool 505 was then used to identify proteins described in the literature as potential systemic inflammatory response 506 and sepsis biomarkers for diagnosis, efficacy or any unspecified application. 507 508 Enzyme-linked Immunosorbent Assay (ELISA). Sandwich-ELISA kits for the human proteins CRP 509 (CRP, ab99995, Abcam, UK), serum amyloid P (APCS, ab137970, Abcam, UK) and cathepsin G (CTSG, 510 abx251221, Abbexa Ltd, UK) were purchased for the quantitative measurement of each corona protein in 511 plasma. Experiments were performed according to manufacturer’s instructions. All assays employed an 512 antibody specific for each human protein coated on a 96-well plate. Standards and samples were pipetted 513 into the wells and protein present in a sample was bound to the wells by the immobilized antibody. The 514 wells were then washed and biotinylated anti-human protein antibody was added. After washing away 515 unbound biotinylated antibody, HRP-conjugated Streptavidin was pipetted to the wells. Following a 516 second washing, a TMB substrate solution was added to each well to allow colour development in 517 proportion to the amount of each bound protein. Finally, stop solution was added to the wells, colour 518 changed from blue to yellow and the absorbance was measured at 450 nm using a plate reader (Fluostar 519 Omega plate reader, BMG Labtech). A standard curve was prepared based upon the intensity of the 520 colour of the standard solutions by GraphPad Prism 7 regression analysis and the concentration of the 521 bound protein per sample was calculated. 522 523 Statistical analysis. Statistical analysis of the data was performed using GraphPad Prism 7 software. 524 The non-parametric Kruskal-Wallis tests was used and p values<0.05 were considered significant. 525 Receiver operating curves (ROC) and the respective Area under the Curve (AUC) values were produced 526 and calculated based on the ELISA data of each selected for validation protein using GraphPad Prism 7 527 software (Figure S3). 528 529 Conflicts of interest 530 There are no conflicts to declare. 531 532 Author contributions 533 L.P., A.C. and M.H. designed and performed the experiments, analyzed all data and took responsibility for 534 planning and writing the manuscript. A.C. and P.D. contributed to the clinical design, provided oversight of 535 the ethical approval process and were responsible for access and storage of the samples in the Salford 536 Royal NHS Foundation Trust. P.D., M.H. and K.K. initiated, designed, directed, provided intellectual input 537 throughout the study and contributed to the writing of the manuscript. 538 539 Acknowledgements 540 This research was partially funded by the Manchester Molecular Pathology Innovation Centre 541 (MMPathIC). Authors wish to acknowledge Salford Royal NHS Foundation Trust’s research and 542 development team for their assistance in the collection of blood samples. The authors also wish to thank 543 the Faculty of Life Sciences EM Facility at the University of Manchester for their assistance in Electron 544 Microscopy imaging. In addition, Mass Spectrometry Facility staff at the University of Manchester for their 545 support. We would also like to thank the patients for their kind donation of blood samples to support this 546 research. 547 548 15 549 FIGURE LEGENDS 550 551 Figure 1: Physicochemical characterization of corona-coated Amphotericin B-intercalated liposomes 552 (AmBisome®). (A) Schematic description of the experimental design. Liposomes were incubated ex vivo (1 hour at 553 370C, 250 rpm) with human plasma samples obtained from a) healthy donors (n=12), b) SIRS patients (n=7) and c) 554 sepsis patients (n=12). Ex vivo corona-coated liposomes were recovered and purified from unbound proteins and the 555 formed protein coronas were qualitatively and quantitatively characterised and compared between the groups; (B) 556 Mean hydrodynamic diameter (nm) and ζ-potential (mV) distributions for the liposomal formulation AmBisome®, 557 before and after ex vivo incubation with human plasma. Table shows the average mean hydrodynamic diameter (nm), 558 polydispersity index (PDI) and ζ-potential (mV) values of bare and corona-coated liposomes; (C) Negative stain TEM 559 imaging of bare and corona-coated AmBisome® liposomes, recovered post-incubation with human plasma obtained 560 from healthy controls, SIRS patients and sepsis patients. All scale bars are 100 nm. 561 562 Figure 2: Quantitative and qualitative comparison of the ex vivo protein coronas formed onto AmBisome® 563 liposomes. (A) Comparison of the total amount of protein adsorbed onto liposomes after their ex vivo incubation with 564 plasma obtained from healthy donors (n=12), SIRS patients (n=7) and sepsis patients (n=12), (expressed as μg 565 protein/μmole of NP). Pb values represent the average and standard error. ** indicates p<0.01 (p=0.0068) between 566 healthy donors and sepsis patients using the non-parametric Kruskal-Wallis test; (B) Imperial stained SDS-PAGE gels 567 of plasma control and corona proteins associated with liposomes after incubation with plasma obtained from healthy 568 donors, SIRS patients and sepsis patients. 569 570 Figure 3: The NP protein corona-enabled biomarker discovery to differentiate sepsis from non-infectious 571 acute systemic inflammation. (A) Heatmap of normalized abundance values of corona proteins differentially 572 abundant between SIRS controls and sepsis patients, after identification by LC-MS/MS and data analysis by 573 Progenesis QI. Protein columns are sorted according to the abundance values (from highest to lowest) of the first 574 sample. The list of proteins shown in the heatmap and their respective accession numbers, p-values and max fold- 575 change are shown in Table S4; (B) Volcano plot represents the potential protein biomarkers differentially abundant 576 between sepsis group and SIRS control group (n=67) identified in corona samples. Upregulated biomarkers in sepsis 577 group are shown in red (n=34), whereas downregulated biomarkers in the presence of sepsis are shown in blue 578 colour (n=33). Only proteins with p value<0.05 are reported. The list of proteins shown in the volcano plot and their 579 respective accession numbers, p-values and max fold-change are shown in Table 1; (C) Ingenuity Pathway Analysis 580 (IPA) of potential biomarker corona proteins associated with bacterial infection. The name of proteins illustrated in the 581 diagram and their respective gene symbol are shown in Table S7. 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 16 598 FIGURE 1 599 600 601 602 603 17 604 FIGURE 2 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 18 627 FIGURE 3 628 629 630 631 632 633 634 635 636 637 638 639 640 19 641 TABLE 1 642 643 644 645 Table 1: Candidate corona protein biomarkers for differentiating between sepsis and non-infectious acute 646 systemic inflammation patients, as identified by proteomic analysis of the ex vivo liposome coronas. Full list of 647 proteins identified by Progenesis QI for proteomics to be upregulated or downregulated in sepsis patients in comparison 648 with SIRS controls classified from the highest max fold-change to the lowest. Only proteins with p<0.05 are shown. 649 Accession Anova Max fold Accession Anova Max fold Identified Proteins (n=67) Identified Proteins (n=67) Number (p) change Number (p) change UPREGULATED DOWNREGULATED OTU domain-containing protein 7A OTU7A_HUMAN 4.32E-03 13.23 N-terminal EF-hand calcium-binding protein 2 NECA2_HUMAN 1.02E-03 317.68 Immunoglobulin kappa variable 1D-13 KVD13_HUMAN 4.34E-03 10.14 Flotillin-2 FLOT2_HUMAN 2.27E-02 18.41 Dual specificity mitogen-activated protein Dematin DEMA_HUMAN 4.38E-02 11.90 MP2K7_HUMAN 4.60E-02 4.11 kinase kinase 7 B-cell lymphoma/leukemia 10 BCL10_HUMAN 1.86E-02 6.11 NEDD4-binding protein 2 N4BP2_HUMAN 4.11E-03 3.19 ATP-dependent 6-phosphofructokinase, liver PFKAL_HUMAN 7.88E-03 3.56 Complement component C8 gamma chain CO8G_HUMAN 1.68E-03 3.07 type Sentrin-specific protease 2 SENP2_HUMAN 7.28E-03 2.88 Cancer-associated gene 1 protein CAGE1_HUMAN 2.04E-02 3.51 Ficolin-2 FCN2_HUMAN 2.92E-05 2.79 Ras-related protein Ral-B RALB_HUMAN 3.18E-02 3.17 Fibulin-1 FBLN1_HUMAN 2.62E-03 2.64 Immunoglobulin lambda variable 1-44 LV144_HUMAN 1.26E-02 3.05 Endophilin-A3 SH3G3_HUMAN 4.97E-02 2.64 Guanine nucleotide-binding protein G(i) GNAI2_HUMAN 8.28E-04 2.96 Sorting nexin-32 SNX32_HUMAN 4.75E-02 2.64 subunit alpha-2 Immunoglobulin heavy constant gamma 2 IGHG2_HUMAN 3.75E-02 2.44 Cytoplasmic dynein 1 heavy chain 1 DYHC1_HUMAN 3.24E-02 2.86 Complement component C7 CO7_HUMAN 1.29E-03 2.36 Immunoglobulin kappa variable 1-12 KV112_HUMAN 2.64E-02 2.65 Immunoglobulin gamma-1 heavy chain IGG1_HUMAN 2.69E-02 2.31 Aminopeptidase N AMPN_HUMAN 4.48E-02 2.59 Tenascin-X TENX_HUMAN 7.46E-04 2.54 Immunoglobulin heavy constant gamma 3 IGHG3_HUMAN 4.21E-02 2.28 Immunoglobulin heavy variable 3-48 HV348_HUMAN 4.13E-02 2.51 Unconventional myosin-XV MYO15_HUMAN 4.77E-02 2.22 Immunoglobulin kappa variable 1-17 KV117_HUMAN 9.46E-03 2.40 Inter-alpha-trypsin inhibitor heavy chain H4 ITIH4_HUMAN 3.59E-03 2.14 Ubiquitin-40S ribosomal protein S27a RS27A_HUMAN 5.41E-05 2.17 Complement C5 CO5_HUMAN 2.49E-03 2.12 Gamma-aminobutyric acid receptor subunit Hemoglobin subunit alpha HBA_HUMAN 7.97E-03 2.04 GBRA1_HUMAN 1.98E-02 2.16 alpha-1 Complement component C6 CO6_HUMAN 1.72E-02 2.00 Serine/arginine-rich splicing factor 4 SRSF4_HUMAN 2.50E-02 2.11 Complement component C8 beta chain CO8B_HUMAN 3.47E-02 1.94 Immunoglobulin heavy variable 6-1 HV601_HUMAN 1.65E-02 2.08 Synapsin-3 SYN3_HUMAN 3.89E-02 1.91 Annexin A2 ANXA2_HUMAN 2.84E-02 2.04 Hemoglobin subunit delta HBD_HUMAN 1.36E-02 1.90 Sulfhydryl oxidase 2 QSOX2_HUMAN 2.70E-02 2.03 Inter-alpha-trypsin inhibitor heavy chain H3 ITIH3_HUMAN 3.51E-02 1.90 PI-PLC X domain-containing protein 2 PLCX2_HUMAN 1.23E-02 1.99 Complement factor H-related protein 1 FHR1_HUMAN 1.64E-02 1.89 Immunoglobulin kappa variable 1-5 KV105_HUMAN 3.97E-02 1.99 Lipopolysaccharide-binding protein LBP_HUMAN 4.48E-02 1.86 Immunoglobulin heavy variable 3-23 HV323_HUMAN 3.20E-02 1.96 FANCD2 opposite strand protein FACOS_HUMAN 3.76E-02 1.85 Immunoglobulin kappa variable 2-24 KV224_HUMAN 3.17E-02 1.90 Complement C3 CO3_HUMAN 1.77E-02 1.85 Immunoglobulin lambda variable 2-11 LV211_HUMAN 3.28E-02 1.90 Fibrinogen alpha chain FIBA_HUMAN 1.59E-02 1.78 Desmoglein-1 DSG1_HUMAN 2.90E-02 1.79 Histone H2B type 1-A H2B1A_HUMAN 1.70E-02 1.70 Immunoglobulin lambda variable 1-51 LV151_HUMAN 2.88E-02 1.73 Serum amyloid P-component SAMP_HUMAN 3.12E-02 1.68 CD5 antigen-like CD5L_HUMAN 4.63E-02 1.72 Carboxypeptidase N catalytic chain CBPN_HUMAN 3.98E-02 1.66 Immunoglobulin kappa variable 3-11 KV311_HUMAN 2.28E-02 1.69 Complement factor H CFAH_HUMAN 3.87E-02 1.53 Apolipoprotein L1 APOL1_HUMAN 4.15E-02 1.62 Hemoglobin subunit beta HBB_HUMAN 2.69E-02 1.48 Oxysterol-binding protein-related protein 11 OSB11_HUMAN 3.70E-02 1.60 Vitronectin VTNC_HUMAN 4.17E-02 1.38 Vitamin K-dependent protein S PROS_HUMAN 4.96E-02 1.34 20 References 1. M. Singer, C. S. Deutschman, C. W. Seymour, M. Shankar-Hari, D. Annane, M. Bauer, et al., JAMA, 2016, 315, 801-810. 2. K. Reinhart, R. Daniels, N. Kissoon, F. R. Machado, R. D. Schachter and S. Finfer, New England Journal of Medicine, 2017, 377, 414-417. 3. R. Ferrer, I. Martin-Loeches, G. Phillips, T. M. Osborn, S. Townsend, R. P. Dellinger, et al., Critical Care Medicine, 2014, 42, 1749-1755. 4. K. Chun, C. Syndergaard, C. Damas, R. Trubey, A. Mukindaraj, S. Qian, et al., J Lab Autom, 2015, 20, 539-561. 5. J. Phua, W. Ngerng, K. See, C. Tay, T. Kiong, H. Lim, et al., Crit Care, 2013, 17, R202. 6. A. Rhodes, L. E. Evans, W. Alhazzani, M. M. Levy, M. Antonelli, R. Ferrer, et al., Intensive Care Medicine, 2017, 43, 304-377. 7. C. A. Michael, D. Dominey-Howes and M. Labbate, Front Public Health, 2014, 2, 145. 8. J. D. Faix, Crit Rev Clin Lab Sci, 2013, 50, 23-36. 9. K. R. Walley, Curr Infect Dis Rep, 2013, 15, 413-420. 10. B. Clyne and J. S. Olshaker, J Emerg Med, 1999, 17, 1019-1025. 11. B. M. Tang, G. D. Eslick, J. C. Craig and A. S. McLean, Lancet Infect Dis, 2007, 7, 210- 217. 12. J. L. Vincent, PLoS Med, 2016, 13, e1002022. 13. V. Sankar and N. R. Webster, J Anesth, 2013, 27, 269-283. 14. N. L. Anderson and N. G. Anderson, Molecular & Cellular Proteomics, 2003, 2, 50-50. 15. K. Hamad-Schifferli, Nanomedicine (Lond), 2015, 10, 1663-1674. 16. T. Cedervall, I. Lynch, S. Lindman, T. Berggard, E. Thulin, H. Nilsson, et al., Proc Natl Acad Sci U S A, 2007, 104, 2050-2055. 17. M. Hadjidemetriou and K. Kostarelos, Nature Nanotechnology, 2017, 12, 288-290. 18. M. Hadjidemetriou, Z. Al-Ahmady, M. Buggio, J. Swift and K. Kostarelos, Biomaterials, 2018, 188, 118-129. 19. M. Hadjidemetriou, S. McAdam, G. Garner, C. Thackeray, D. Knight, D. Smith, et al., Adv Mater, 2018, DOI: 10.1002/adma.201803335, e1803335. 20. M. D. Moen, K. A. Lyseng-Williamson and L. J. Scott, Drugs, 2009, 69, 361-392. 21. M. Hadjidemetriou, Z. Al-Ahmady, M. Mazza, R. F. Collins, K. Dawson and K. Kostarelos, ACS Nano, 2015, 9, 8142-8156. 22. M. Hadjidemetriou, Z. Al-Ahmady and K. Kostarelos, Nanoscale, 2016, 8, 6948-6957. 23. Q. Dai, C. Walkey and W. C. Chan, Angew Chem Int Ed Engl, 2014, 53, 5093-5096. 24. C. J. C. de Haas, M. E. van der Tol, K. P. M. Van Kessel, J. Verhoef and J. A. G. Van Strijp, J Immunol, 1998, 161, 3607-3615. 25. C. J. C. de Haas, Fems Immunol Med Mic, 1999, 26, 197-202. 26. R. A. Schwalbe, B. Dahlback, J. E. Coe and G. L. Nelsestuen, Biochemistry-Us, 1992, 31, 4907-4915. 27. J. Yuste, M. Botto, S. E. Bottoms and J. S. Brown, Plos Pathog, 2007, 3, 1208-1219. 28. S. W. Li, X. Yang, J. Shao and Y. Q. Shen, Plos One, 2012, 7. 29. H. Y. Lu, B. M. Bauman, S. Arjunaraja, B. Dorjbal, J. D. Milner, A. L. Snow, et al., Front Immunol, 2018, 9. 30. S. Bhattacharyya, A. Borthakur, N. Pant, P. K. Dudeja and J. K. Tobacman, Am J Physiol Gastrointest Liver Physiol, 2007, 293, G429-437. 31. L. Sanjurjo, N. Amezaga, G. Aran, M. Naranjo-Gomez, L. Arias, C. Armengol, et al., Autophagy, 2015, 11, 487-502. 32. X. Gao, Y. Liu, F. Xu, S. Lin, Z. Song, J. Duan, et al., Chest, 2019, DOI: 10.1016/j.chest.2019.04.134. 21 33. K. F. Chen, C. H. Chaou, J. Y. Jiang, H. W. Yu, Y. H. Meng, W. C. Tang, et al., Plos One, 2016, 11, e0153188. 34. S. M. Opal, P. J. Scannon, J. L. Vincent, M. White, S. F. Carroll, J. E. Palardy, et al., J Infect Dis, 1999, 180, 1584-1589. 35. R. S. Hotchkiss, L. L. Moldawer, S. M. Opal, K. Reinhart, I. R. Turnbull and J. L. Vincent, Nat Rev Dis Primers, 2016, 2, 16045. 36. C. Pierrakos and J.-L. Vincent, Critical Care 2010, 14. 37. T. C. Hall, D. K. Bilku, D. Al-Leswas, C. Horst and A. R. Dennison, Ir J Med Sci, 2011, 180, 793-798. 38. K. Merrell, K. Southwick, S. W. Graves, M. S. Esplin, N. E. Lewis and C. D. Thulin, J Biomol Tech, 2004, 15, 238-248. 39. C. E. Parker and C. H. Borchers, Mol Oncol, 2014, 8, 840-858. 40. L. A. Liotta, M. Ferrari and E. Petricoin, Nature, 2003, 425, 905. 41. R. Schiess, B. Wollscheid and R. Aebersold, Mol Oncol, 2009, 3, 33-44. 42. N. Rifai, M. A. Gillette and S. A. Carr, Nature Biotechnology, 2006, 24, 971-983. 43. Z. Cao and R. A. Robinson, Proteomics Clin Appl, 2014, 8, 35-52. 44. M. E. Lissauer, S. B. Johnson, G. Siuzdak, G. Bochicchio, C. Whiteford, B. Nussbaumer, et al., J Trauma, 2007, 62, 1082-1092; discussion 1092-1084. 45. J. Kim, M. A. Abdou Mohamed, K. Zagorovsky and W. C. W. Chan, Biomaterials, 2017, 146, 97-114. 46. X. Zhu, A. F. Radovic-Moreno, J. Wu, R. Langer and J. Shi, Nano Today, 2014, 478-498. 47. P. D. Howes, R. Chandrawati and M. M. Stevens, Science, 2014, 346, 1247390. 48. A. H. Nguyen, Y. Shin and S. J. Sim, Biosensors and Bioelectronics, 2016, 85, 522-528. 49. H. J. Chung, C. M. Castro, H. Im, H. Lee and R. Weissleder, Nature Nanotechnology, 2013, 8, 369-375. 50. M. Liong, A. N. Hoang, J. Chung, N. Gural, C. B. Ford, C. Min, et al., Nat Commun, 2013, 4, 1752. 51. J. J. Storhoff, A. D. Lucas, V. Garimella, P. Y. Bao and U. R. Mueller, Nature Biotechnology, 2004, 22, 883-887. 52. H. D. Hill and C. A. Mirkin, Nature Protocols, 2006, 1, 324-336. 53. B. A. Borsa, B. G. Tuna, F. J. Hernandez, L. I. Hernandez, G. Bayramoglu, M. Y. Arica, et al., Biosensors and Bioelectronics, 2016, 86, 27-32. 54. D. Issadore, H. J. Chung, J. Chung, C. Budin, R. Weissleder and H. Lee, Advanced Healthcare Materials, 2013, 1224-1228. 55. C.-I. Weng, H.-T. Chang, C.-H. Lin, Y.-W. Shen, B. Unnikrishnan, Y.-J. Li, et al., Biosensors and Bioelectronics, 2015, 68, 1-6. 56. X. Zhao, L. R. Hilliard, S. J. Mechery, Y. Wang, R. P. Bagwe, S. Jin, et al., PNAS, 2004, 101, 15027-15032. 57. M. Magro, M. Zaccarin, G. Miotto, L. Da Dalt, D. Baratella, P. Fariselli, et al., Analytical and Bioanalytical Chemistry, 2018, 410, 2949-2959. 58. G. Miotto, M. Magro, M. Terzo, M. Zaccarin, L. Da Dalt, E. Bonaiuto, et al., Colloids and Surfaces B-Biointerfaces, 2016, 140, 40-49. 59. R. S. Samraj, B. Zingarelli and H. R. Wong, Shock, 2013, 40, 358-365. 60. S. Schottler, K. Klein, K. Landfester and V. Mailander, Nanoscale, 2016, 8, 5526-5536. 61. N. R. Stone, T. Bicanic, R. Salim and W. Hope, Drugs, 2016, 76, 485-500. 62. R. Hamill, Drugs, 2013, 73, 919-934. 22