This is the accepted version of the following article: Hadjidemetriou M., Papafilippou L., Unwin R.D., Rogan J., Clamp A., Kostarelos K.. Nano-scavengers for blood biomarker discovery in ovarian carcinoma. Nano Today, (2020). 34. 100901: - . 10.1016/j.nantod.2020.100901, which has been published in final form at https://dx.doi.org/10.1016/j.nantod.2020.100901 © https://dx.doi.org/10.1016/j.nantod.2020.100901. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ 1 2 3 Nano-scavengers for Blood Biomarker Discovery in Ovarian 4 Carcinoma 5 6 7 Marilena Hadjidemetriou1*, Lana Papafilippou1, Richard D. Unwin2, Jane Rogan3, Andrew Clamp4 and Kostas 8 Kostarelos1,5* 9 10 11 1Nanomedicine Lab, Faculty of Biology, Medicine & Health, AV Hill Building, The University of Manchester, Manchester, , 12 United Kingdom 13 14 2Divisionof Cardiovascular Sciences, Faculty of Biology, Medicine & Health, The University of Manchester, Manchester, 15 United Kingdom 16 17 3Manchester Cancer Research Centre Biobank, The Christie NHS Foundation Trust, CRUK Manchester Institute, 18 Manchester, United Kingdom 19 20 4Institute of Cancer Sciences and The Christie NHS Foundation Trust, Manchester Cancer Research Centre (MCRC), 21 The University of Manchester, Manchester, United Kingdom 22 23 5Catalan Institute of Nanoscience and Nanotechnology (ICN2), Campus UAB, Bellaterra, 08193 Barcelona, Spain 24 25 26 _______________________________________ 27 * Correspondence should be addressed to: marilena.hadjidemetriou@manchester.ac.uk 28 kostas.kostarelos@manchester.ac.uk 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 1 46 Abstract 47 48 49 The development and implementation of biomarker-based screening tools for ovarian cancer require 50 novel analytical platforms to enable the discovery of biomarker panels that will overcome the 51 limitations associated with the clinically used CA-125.The systematic discovery of protein 52 biomarkers directly from human plasma using proteomics remains extremely challenging, due to the 53 wide concentration range of plasma proteins. Here, we describe the use of lipid-based nanoparticles 54 (NPs) as an ‘omics’ enrichment tool to amplify cancer signals in the blood and to uncover disease 55 specific signatures. We aimed to exploit the spontaneous interaction of clinically-used liposomes 56 (Caelyx®) with plasma proteins, also known as ’protein corona’ formation, in order to facilitate the 57 discovery of previously unreported differentially abundant molecules. Caelyx® liposomes were 58 incubated with plasma samples obtained from advanced ovarian carcinoma patients and healthy 59 donors and corona-coated liposomes were subsequently recovered. Comprehensive comparison 60 between ‘healthy’ and ‘diseased’ corona samples by label-free proteomics resulted in the 61 identification of multiple differentially abundant proteins. Moreover, immunoassay-based validation 62 of selected proteins demonstrated the potential of nanoparticle-platform proposed to discover novel 63 molecules with great diagnostic potential. This study proposes a nanoparticle-enabled workflow for 64 plasma proteomic analysis in healthy and diseased states and paves the way for further work 65 needed to discover and validate panels of novel biomarkers for disease diagnosis and monitoring. 66 67 68 Keywords: protein corona, biomarkers, ovarian cancer, liposomes, nanomedicine 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 2 92 Introduction 93 94 Much effort is currently focused on the development of robust and high-throughput ‘omics’ 95 platforms for the discovery of minimally invasive molecular biomarkers to aid early and accurate 96 cancer diagnosis, monitor tumour growth and response to therapies. Despite significant investment 97 by major stakeholders, few protein cancer biomarkers have been validated and received FDA 98 approval, raising concerns regarding the efficiency of the biomarker-development pipeline. It is 99 noteworthy that of the FDA-approved biomarkers, the majority are used to monitor the progression 100 of cancer, rather than enabling its early diagnosis.1 101 Proteins are the biological endpoints that govern most pathophysiological processes and 102 they have therefore attracted most interest so far as biomarkers for cancer diagnostics.2 Blood is 103 frequently the biosample of choice for biomarker identification; however the discovery of tumour- 104 derived protein signatures directly from blood is hindered by the wide concentration range of blood 105 proteins, in addition to the preponderance of highly abundant proteins.3 Despite significant 106 improvement in the sensitivity of mass spectrometry-based proteomics, the issue of the high 107 dynamic range of plasma protein abundances still remains unresolved and the diagnostic 108 information blood can offer is partially inaccessible.4 109 Nanotechnology-based platforms hold great promise in addressing the above issues 110 associated with biomarker discovery.5 It should be emphasised however, that the vast majority of 111 nanoparticle-based technologies developed so far have been designed to capture and quantify 112 already known cancer-specific analytes,6-9 enabling the verification and validation phases of 113 biomarker development. The NP-enabled discovery of new plasma buried biomarkers has only 114 been recently attempted.10 115 The fact that the surface of NPs is instantly covered by a wide range of adsorbed proteins 116 and other biomolecules once in contact with blood, a self-assembly phenomenon known as 117 ‘protein’ or ‘biomolecule’ corona formation, 11,12 makes NPs ideal biomarker discovery platforms. 118 Biomolecule corona formation has become a popular line of research in the last decade and 119 ongoing research is mainly focused on the proteomic analysis of corona profiles after the ex vivo 120 and more recently the in vivo interaction of NPs with biofluids (mainly plasma).13-17 Nanoparticle- 121 protein interactions at the bio-nano interface not only can shed new light on the development of 122 nanotechnologies but are now gradually being exploited as an engineering tool with therapeutic 123 and diagnostic capabilities. 10,11,18-20 124 The surface-capture of a complex blood proteome by NPs as well as the recently proposed 125 concept of ‘personalized corona’ has sparked interest for utilizing the biomolecule corona 3 126 fingerprinting as a proteomic discovery platform.10,18,21,22 We have recently demonstrated that the 127 NP protein corona formed in the blood circulation of humans has the potential to be exploited as an 128 enrichment and pre-fractionation tool that allows in depth coverage of the plasma proteome.18 In a 129 subsequent study, we employed two different tumour mouse models (a subcutaneous melanoma 130 model and human lung carcinoma xenograft model) to demonstrate that intravenously injected 131 lipid-based NP-scavengers (liposomes) surface-capture low MW, low abundant and disease- 132 specific plasma proteins which cannot be detected by conventional plasma proteomic analysis.10 133 Moreover, this study demonstrated that protein coronas, formed around intravenously injected 134 NPs, differ both quantitatively and qualitatively in the presence and absence of a disease, allowing 135 the uncovering of differentially abundant potential biomarker proteins.10 136 When animal models are employed for biomarkers discovery the exploitation of the 137 molecularly richer in vivo protein corona is advantageous as opposed to its counterpart ex vivo 138 corona.16 However, hypothesis-free discovery proteomics often require the use of human clinical 139 samples and therefore, in this study we aimed to explore the use of the ex vivo protein corona 140 formed around the clinically used PEGylated liposomal doxorubicin formulation (Caelyx®), to 141 identify disease-specific proteins directly from plasma samples, obtained from patients with 142 recurrent ovarian carcinoma. 143 The work flow of this study is summarized in Figure 1A and involved the incubation of 144 Caelyx® liposomes with plasma samples from patients with recurrent ovarian cancer and from 145 healthy donors and the comprehensive comparison of the resultant protein coronas by label-free 146 mass spectrometry. The above analysis led to in the discovery of 413 differentially abundant 147 proteins between ‘healthy’ and ‘diseased’ corona samples, of which nine were quantified by 148 immunoassays to further validate the potential use of the nanoparticle-protein corona technology 149 for plasma proteomic analysis and biomarkers discovery. 150 151 . 152 153 154 155 156 157 158 159 160 161 162 4 163 Results 164 165 Recovery and purification of corona-coated liposomes from plasma samples obtained from 166 ovarian carcinoma patients and healthy donors. To investigate the exploitation of the ex vivo 167 formed NP protein corona for biomarker discovery, Caelyx® liposomes (20 µl of 1.5mM) were 168 incubated with plasma samples (980 µl) obtained from patients with recurrent ovarian carcinoma 169 about to commence the first cycle of Caelyx® as part of standard-of-care treatment (n=19) and 170 age- matched female healthy donors (n=10). Patient clinical and basic blood analysis 171 characteristics are summarized in Tables S1 and S2. The physicochemical characteristics of the 172 Caelyx® liposomes employed are summarized in Table S3. It should be noted that Caelyx® was 173 employed because of its clinical use for the treatment of advanced ovarian cancer. The presence 174 of the encapsulated doxorubicin has been shown not to affect the surface properties of liposomes 175 and therefore corona formation.15 176 The ex vivo protein corona was allowed to form upon incubation of Caelyx® liposomes with 177 plasma samples for 90 min at 37°C (Figure 1A). A purification protocol dependent on size 178 exclusion chromatography was immediately performed to separate corona-coated liposomes from 179 unbound plasma proteins. Membrane ultrafiltration was then used to concentrate the corona 180 samples and to remove any large unbound or softly attached proteins, as previously optimised and 181 described.10,13-16,18 The above two-step purification process results in the complete elimination of 182 unbound plasma proteins as demonstrated by plasma control experiments (Figure S1). 183 To confirm corona formation and to assess the morphology of Caelyx® liposomes before 184 and after corona formation, Transmission Electron Microscopy (TEM) was performed. A well- 185 dispersed liposome suspension was observed before and after incubation with plasma samples 186 and purification. Corona-coated Caelyx® liposomes retained their size and spherical structure, 187 while the occurrence of the proteins attached onto their surface revealed protein corona formation 188 (Figure 1B). 189 To quantitatively compare ‘healthy’ and ‘diseased’ protein coronas, we calculated the total 190 amount of protein associated with each µmole of lipid (Protein binding value; Pb). As shown in 191 Figure1C, the average Pb value for ovarian carcinoma patients was 4 times higher than the 192 average Pb value observed for healthy controls. These results are in agreement with our previous 193 investigations in preclinical mouse models showing that protein corona fingerprints quantitatively 194 differ in the absence and presence of tumorigenesis.10 195 Proteins associated with Caelyx® liposomes were separated by SDS-PAGE and visualized 196 by Imperial Protein stain, as illustrated in Figure 1D. Despite the higher total amount of protein 197 observed in the ‘diseased’ coronas, well distinct protein bands even at the low MW region were 5 198 observed, demonstrating the ability of the NP enrichment platform technology to minimise the noise 199 of highly abundant proteins, such as albumin, and allow the interaction with low abundant proteins. 200 The extensive purification processes applied to retrieve the corona-coated liposomes and purify 201 them from the unbound proteins, worked as fractionation tool allowing the uncovering of the low 202 MW plasma proteome (Figure 1D). 203 Proteomic comparison of ‘healthy’ and ‘diseased’ protein coronas. The goal of the 204 proteomic discovery experiment was to comprehensively compare the ‘healthy’ and ‘diseased’ 205 coronas in order to identify differentially abundant proteins. 206 In order to assess the reproducibility of the analysis of protein corona by LC-MS/MS, we 207 isolated corona-coated liposomes from 6 aliquots of the same plasma sample (obtained from one 208 healthy donor). The results demonstrated that the purification and quantification of the 209 nanoparticle-bound protein fraction was reproducible, with 73% of proteins being measured with 210 <30% CV (Figure S2A). To further validate the experimental reproducibility, two of the above 211 replicated samples were analyzed in triplicates to evaluate the repeatability of the LC:MS/MS 212 platform and the results demonstrated high analytical precision with approximately 95% of 213 identified proteins with <30% CV (Figure S2B & S3). To assess the linearity of protein adsorption 214 we incubated the same concentration of Caelyx® liposomes in full and diluted plasma. 215 Interestingly, the total amount of corona proteins adsorbed onto the surface of liposomes was 216 directly proportional to the total protein concentration in the incubation medium (Figure S4). 217 To compare ‘healthy’ and ‘diseased’ coronas, equal amounts of total protein from each 218 corona sample were digested and subsequently analysed by LC-MS/MS. It should be emphasized 219 that albumin and immunoglobulins were not depleted from corona samples prior to proteomic 220 analysis. Processing of the raw data generated by LC-MS/MS analysis with Progenesis QI (version 221 3.0; Nonlinear Dynamics) software tool was carried out to statistically compare the abundance of 222 proteins present in the ‘healthy’ and ‘diseased’ liposomal coronas. The Relative Protein Expression 223 (fold change) and the reliability of measured differences (ANOVA, p value) were calculated. 224 Figures 2A & 2B highlight the subset of differentially abundant proteins that met our confirmation 225 criteria (see Experimental Section for further details). Out of 1187 identified proteins, 413 were 226 found to be differentially abundant between the two groups with a p value <0.05, of which 171 were 227 upregulated and 242 downregulated (Figure 2B &Table S4). Considering the importance of 228 achieving high confidence in the discovered proteins, we applied even more stringent criteria and 229 interestingly out of the above 413 differentially abundant proteins 303 had a p value<0.01 (FDR of 230 2%) and a fold change> 2, which represents 25.5% of all proteins identified (Figure 2B). 6 231 The majority of highly specific cancer biomarkers are low MW intracellular proteins released 232 from the tumour microenvironment into the blood circulation by leakage or secretion, however, their 233 detection by plasma proteomic analysis remains challenging due to their extremely low 234 concentration in the ng/ml to pg/ml range.23 Interestingly, classification of the differentially abundant 235 proteins identified in this study according to their cellular localization, demonstrated the enrichment 236 of 189 intracellular proteins (present in the cytoplasm or nucleus) onto the surface of liposomes 237 (Figure S5). In addition, ~50% of the differentially abundant proteins discovered had a MW<60 238 kDa (Figure S6). 239 The above observation prompted us to investigate whether the differentially abundant 240 corona proteins have been previously associated with ovarian carcinoma pathways. Disease and 241 function IPA search revealed the association of 335 and 60 corona proteins with solid tumour 242 pathways and metastasis processes, respectively. Interestingly, 72 proteins have been previously 243 associated with ovarian cancer pathways, of which 15 have been described in the literature as 244 potential biomarkers for ovarian cancer (n=8 for diagnosis; n=5 for unspecified applicability, n=1 for 245 safety; and n=2 for efficacy), (Figure 3). 246 The plasma-incubated liposomes also surface-captured the clinically used blood biomarkers 247 CA 125 (MUC 16), Transthyretin (TTR) and Apolipoprotein A1 (APOA1), all included in the FDA 248 approved OVA1 diagnostic test which is used to evaluate ovarian masses for cancer prior to 249 planned surgery.24 In agreement with OVA1 test, CA125 and B2M were found to be upregulated in 250 the ‘diseased’ corona whereas APOA1 was found to be downregulated. This suggests that the 251 abundance of corona proteins, as calculated by LC-MS/MS analysis, directly reflects their 252 concentration in blood. 253 Overall, the above data suggest that analysis of protein coronas formed after the ex vivo 254 incubation of liposomes with plasma samples obtained from cancer patients and healthy controls 255 can be used to uncover differentially abundant proteins, otherwise buried under the overwhelming 256 signal of albumin. 257 258 Validation of the nanoparticle protein corona technology. To verify that the level of fold change 259 observed by proteomic analysis of the ‘healthy’ and ‘diseased’ coronas is representative of the 260 plasma proteome in healthy and diseased states, we performed ELISA experiments using plasma 261 samples obtained from the same ovarian carcinoma patients and healthy controls. Distribution of 262 the differentially abundant proteins identified by statistical significance and magnitude of change 263 revealed that the majority exhibited fold change values much higher than the clinically used 264 biomarkers CA125, TTR, and APOA1 (Figure 2B & 4A). 7 265 Nine corona proteins were selected to be validated and were divided in three groups: a) 266 clinically used biomarkers for ovarian cancer (CA125, APOA1 and TTR; shown in orange); b) 267 proteins mapped by IPA software to be associated with ovarian carcinoma pathways (THBS1, 268 ENO1 and TGF-b1; shown in green) and c) proteins that have not been previously associated with 269 ovarian carcinoma pathways but exhibited very promising fold change and p values (NME1, PDIA4 270 and PRKCSH; shown in blue). The fold change and p values of the nine selected proteins (as 271 calculated by LC-MS/MS analysis) are illustrated in (Figures 4A & S7). 272 The plasma concentration profiles of selected proteins and their respective ROC curves, as 273 calculated by ELISA experiments are shown in Figure 4B. In agreement with LC-MS/MS data 274 (Figures 4A & S7), APOA1 and TTR proteins were found to be downregulated while CA125, 275 THBS1,ENO1, TGF-b1, NME1, PDIA4 and PRKCSH proteins were found to be upregulated in 276 ovarian carcinoma patients (Figure 4B). This indicates that changes in the plasma proteome are 277 directly reflected in the protein corona composition. 278 As illustrated in Figure 4B, the second group of proteins showed greater specificity and 279 sensitivity than the clinically used biomarkers with AUC values ranging between 97.6% and 99.4%. 280 Strikingly, an AUC value of 100% was observed for PDIA4 and PRKCSH proteins. The above 281 ELISA validation data provide the first experimental evidence that ex vivo corona proteomic 282 profiling provides allows increased penetration into the plasma proteome and has the potential to 283 allow the discovery of candidate biomarkers. 284 285 Discussion 286 287 In the UK, 55-58% of ovarian carcinoma patients are diagnosed at stage III or IV and 42- 288 45% are diagnosed at stage I or II. Survival for ovarian cancer is strongly related to the stage of the 289 disease at diagnosis (99% of patients diagnosed at stage I survive their disease for at least one 290 year, versus 51% of patients diagnosed at stage IV).25 The lack of disease-specific symptoms, in 291 addition to the limited performance of the clinically used CA-125 serum biomarker, indicates the 292 need for new biomarker-based tools to accurately detect ovarian cancer and to monitor disease 293 progression.26 294 Label-free proteomics profiling of blood is a powerful tool to detect molecular biomarkers that 295 are differentially expressed between healthy and disease states.27 The concentration of a complex 296 network of proteins that regulate tumorigenic pathways is often altered in the blood circulation of 297 cancer patients; however their identification is hampered by the wide dynamic range of plasma 298 proteins. Due to the limited analytical sensitivity, currently available mass spectrometry-based 8 299 approaches predominantly detect highly abundant proteins of limited diagnostic use and fail to detect 300 low MW tumour-tissue derived proteins of lower abundance. To overcome the issue of albumin 301 masking, plasma immunodepletion columns are extensively employed, however their use is limited 302 by a significant loss of low MW proteins along with the highly abundant carrier plasma proteins. 303 Therefore, the multifaceted process of tumorigenesis necessitates the development of ‘omics- 304 enrichment’ platforms that will enable the discovery of biomarker panels with sufficient specificity and 305 sensitivity. 306 We have recently proposed the use of nanoparticles (NPs) as protein-scavenging 307 enrichment platforms to address the above fundamental issues associated with biomarker 308 discovery in plasma. Our results demonstrated that intravenously administered lipid-based NPs, in 309 mice and humans, were able to surface-capture and amplify low abundant tumour-released 310 molecules that could not be detected by conventional plasma proteomics analysis.10,18 This idea of 311 using nanoparticles to allow an in depth-analysis of the blood proteome is based on the 312 spontaneous and non-targeted adsorption of hundreds of proteins onto the nanoparticles surface 313 once in contact with biological fluids, a phenomenon known as ‘protein corona’ formation.11 314 Herein, we aimed to further explore and validate the potential use of the nanoparticle-protein 315 corona to discover novel disease-specific proteins from plasma samples obtained from ovarian 316 carcinoma patients (Figure 1A and Tables S1 & S2). Even though analysis of the in vivo formed 317 protein corona (after intravenous administration of NPs) has been shown to result in a richer sampling 318 of the blood proteome,16 exploitation of the ex vivo protein corona (after incubation of NPs with 319 plasma samples) could be more easily incorporated in the discovery phase of the biomarker pipeline 320 and deserves further investigation. It should be emphasized that unlike other nanoparticle-based 321 technologies aiming to increase the sensitivity of detection of already known molecules, 8,28 the 322 technology platform proposed here aims to identify previously unseen potential biomarker proteins 323 that could potentially offer higher specificity and sensitivity than the clinically used biomarkers. 324 To assess the potential of the NP protein corona technology as a tool for biomarker 325 discovery, we comprehensively compared protein coronas formed around a clinically used 326 liposomal formulation, Caelyx®, upon incubation with plasma samples obtained from healthy 327 donors (n=10) and from patients with recurrent ovarian cancer (n=19). Immediately after plasma 328 incubations, corona-coated liposomes were purified from any unbound plasma proteins and only 329 liposome-bound proteins were analysed by LC-MS/MS. It should be emphasized that albumin and 330 immunoglobulins were not depleted from corona samples prior to LC-MS/MS analysis. Even 331 though a fraction of highly abundant proteins interacts with the surface of liposomes, any unbound 332 highly abundant proteins are removed by the purification process which addresses the ‘signal-to- 9 333 noise’ issue that biomarker discovery suffers from. We therefore propose that analysis of the 334 nanoparticle protein corona can substitute plasma fractionation and immunodepletion 335 methodologies. 336 In agreement with our previous studies in tumour-bearing mice,10 we observed a 337 significantly higher total amount of protein adhered onto the NPs surface in ovarian carcinoma 338 patients in comparison to healthy controls (Figure 1C). This observation, not only confirms our 339 hypothesis that protein corona is greatly affected by the ongoing tumorigenesis but it also paves 340 the way for the development of diagnostic tests. More studies are needed to assess if the 341 fluctuation in the amount of protein adsorbed onto the NPs surface can be used to indicate the 342 onset of a disease or to monitor disease progression and response to the treatment. 343 Despite the higher amount of protein adhered onto the NPs surface after incubation with 344 plasma samples obtained from ovarian carcinoma patients, gel electrophoresis experiments 345 indicated that the analysis of the NP protein corona eliminates the issue of albumin masking and 346 enables the analysis of a broad range of the plasma proteins including low MW proteins (Figure 347 1D & Figure S5). Subsequent analysis of the NPs protein coronas by LC-MS/MS revealed 413 348 proteins that were differentially abundant between ovarian carcinoma patients and healthy controls 349 (with a p value < 0.05), of which 171 were under-expressed and 242 over-expressed (Figures 2A 350 & B). Recent ongoing biomarker development efforts indicate that multiple markers, used 351 individually or as part of a panel, are required to provide sufficient sensitivity and specificity. 352 Noteworthy, although the majority of proposed cancer biomarkers are proteins found to be 353 upregulated in the blood circulation of cancer patients, downregulated biomarkers are currently 354 clinically used and should be taken into consideration.26 355 According to Ingenuity Pathway Analysis, out of 413 differentially abundant proteins 356 discovered in this study 57 have been previously associated with ovarian carcinoma pathways and 357 only 15 have been previously proposed as potential biomarkers for ovarian cancer (Figure 3). 358 Moreover, distribution of the differentially abundant proteins identified by statistical significance and 359 magnitude of change (Figure 2B) revealed that n=303 proteins had a p value<0.01 and a fold 360 change> 2. It should be noted that even though the clinically used biomarkers CA125, TTR and 361 APOA1 were found to interact with the surface of liposomes, the majority of differentially abundant 362 proteins identified, exhibited higher fold change values with higher statistical significance in 363 comparison to the clinically used biomarkers (Figure 3B). These results suggest that analysis of 364 the NP protein corona can unveil previously unseen disease-specific molecules, otherwise buried 365 under the overwhelming signal of albumin and immunoglobulins. More studies are needed to prove 366 the ability of this nanoplatform to enrich disease-specific molecules at the early asymptomatic 10 367 stages of cancer. Moreover, pathway analysis of the ‘diseased’ nanoparticle corona could provide 368 valuable information about the ongoing pathological pathways and the mechanism of cancer 369 initiation and progression and could potentially lead to discovery of novel therapeutic strategies. 370 The identification of previously unknown disease-specific proteins prompted us to verify the 371 differences observed between ‘healthy’ and ‘diseased’ coronas by commercially available ELISA 372 kits, using plasma samples obtained from the same ovarian carcinoma patients and healthy 373 controls. We chose to validate and compare 3 groups of proteins: a) clinically used biomarkers for 374 ovarian cancer (CA125, APOA1 and TTR), b) proteins mapped by IPA software to be associated 375 with ovarian carcinoma pathways (THBS1, ENO1 and TGF-b1) and c) proteins that have not been 376 previously associated with ovarian carcinoma pathways (NME1, PDIA4 and PRKCSH). ELISA data 377 were found to directly reflect the changes observed by LC-MS/MS analysis of ‘healthy’ and 378 ‘diseased’ corona samples (Figure 4), indicating that protein corona composition mirrors the 379 concentration fluctuations of the plasma proteome in the presence of a disease. Interestingly, 380 ELISA quantification of the last two groups of proteins revealed higher performance than the 381 clinically used biomarkers with AUC values between 97.6% and 99.4%, while an AUC value of 382 100% was shown for PDIA4 and PRKCSH proteins. Given the same plasma source was used for 383 the corona formation and ELISA experiments, the above data represent solely an orthogonal 384 validation of the LC-MS/MS data. Clearly, more validation studies will be required to prove the 385 ability of the proposed proteins to discriminate between ovarian carcinoma patients and healthy 386 controls. 387 Although the samples employed in this study were obtained from advanced ovarian cancer 388 patients and have limited value for the discovery of screening biomarkers, the above validation 389 data provide the first experimental evidence of the exploitation of the NP protein corona, formed ex 390 vivo in human clinical samples, for the discovery of potential biomarker proteins. Considering the 391 low number of samples used in this study, it should be emphasized that the clinical utility of the 392 differentially abundant proteins identified as screening or monitoring biomarkers will require 393 validation in much larger and well-defined patient cohorts and with the appropriate control groups 394 (i.e. benign gynaecological conditions). 395 Collectively, our results suggest that NPs dispersed in biological fluids have the potential to 396 be used as an enrichment ‘omics’ platform for biomarker discovery. It is now well established that 397 the physicochemical properties of NPs directly affect the composition of protein corona11 and more 398 work is needed to investigate whether the use of other types of NPs and/or combinations of 399 different NPs will further increase the range of plasma proteome detected. 400 11 401 Conclusion 402 403 In this study, we propose the use of lipid-based nanoparticles as ‘omics’ enrichment 404 platforms to reveal disease specific signatures in the blood of ovarian carcinoma patients. We 405 demonstrate that the molecular composition of protein corona, spontaneously formed around NPs 406 upon incubation with plasma samples, reflects the concentration fluctuations of the blood proteome 407 in the presence of tumorigenesis. Comprehensive comparison between the ex vivo formed ‘healthy’ 408 and ‘diseased’ protein coronas by label-free proteomics (LC-MS/MS), revealed the discovery of 413 409 differentially abundant proteins. Subsequent immunoassay- based validation demonstrated the 410 potential of the nanoparticle-platform proposed to identify novel potential biomarker proteins. This 411 work is thought to pave the way for many more studies needed to allow the clinical exploitation of 412 protein corona fingerprinting as a novel tool to track tumours over time and discover panels of novel 413 biomarkers for early and accurate disease diagnosis. 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 12 430 Experimental 431 Ethical Approvals. This project was reviewed and approved by the Manchester Cancer Research Centre 432 Biobank Sample Access Committee and all sample collection was conducted under the MCRC Biobank 433 Research Tissue Bank Ethics (ref: 07/H1003/161+5). 434 435 Blood sample collection. Eligible cases for this study included women with recurrent ovarian cancer 436 commencing Caelyx® chemotherapy as part of standard chemotherapeutic management for disease 437 progression. Caelyx® contains 2mg/ml doxorubicin hydrochloride encapsulated in a PEGylated liposomal 438 formulation (16 mg lipid content) and is indicated for the treatment of advanced ovarian cancer in women 439 who have failed a first-line platinum-based chemotherapy. Plasma samples (before cycle 1 Caelyx® infusion) 440 were collected into commercially available anticoagulant-treated tubes (K2 EDTA BD Vacutainer®). Plasma 441 was then prepared by inverting the collection tubes 10 times to ensure mixing of blood with EDTA and 442 subsequent centrifugation for 12 minutes at 1300 RCF at 4 °C. Following centrifugation supernatant was 443 immediately collected into labelled Protein LoBind Eppendorf tubes and stored in -800C. Age-matched 444 plasma samples from healthy female donors (n=2 Caucasian; n=4 Black; n=4 Hispanic) were purchased from 445 Seralab UK (LOT BRH1221742-BRH1221751). Considering the impact of the anticoagulant agent on the 446 formation of the protein corona,29 healthy plasma samples contained the same anticoagulant agent (K2 447 EDTA BD Vacutainer® tubes) as that described above for the human clinical samples and were subjected to 448 the same preparation protocol (centrifugation for 12 min at 1300 rpm at 4 0C). Healthy human plasma 449 samples were received on dry ice and were stored in a -800C. Finally, samples were thawed only before the 450 incubations. 451 452 Ex vivo protein corona formation. To investigate the ex vivo protein corona, Caelyx® liposomes were 453 incubated with plasma samples obtained from recurrent ovarian cancer patients and from healthy donors. 454 Caelyx® liposomes (20 ul of 0.15 mM) were incubated with 980 ul of plasma for 90 min at 37°C in orbital 455 shaker at 250 rpm. The ex vivo protein corona was allowed to form using the same liposome concentration 456 (0.15 mM) as that extracted in 1 mL of plasma from intravenously injected patients.18 This liposome 457 concentration results in a final sample protein concentration (upon purification of corona-coated liposomes) 458 that allows in gel digestion of 20ug of protein/sample and subsequent LC-MS/MS analysis. Our previous time 459 evolution studies of the nanoparticle protein corona demonstrated that a complex protein corona forms as 460 early as 10 min post-incubation and does not quantitatively change over time.15 In the present study, we 461 chose to incubate liposomes for 90 min, which reflects the time of Caelyx® infusion in ovarian carcinoma 462 patients. 463 464 Separation of corona-coated liposomes from unbound and weakly bound proteins. Corona-coated 465 liposomes were separated as we have previously described.15,16 Briefly, ex vivo incubated liposomes were 466 separated form unbound plasma proteins by size exclusion chromatography followed by membrane 467 ultrafiltration. Immediately after incubation, samples (1ml) were loaded onto a Sepharose CL-4B (Sigma- 468 Aldrich) column (15cm) equilibrated with HEPES buffer. Fractions containing liposomes (4,5,6) were then 469 pooled together and concentrated to 500 μL using a Vivaspin 6 column (10 000 MWCO, Sartorious, Fisher 470 Scientific) at 9000 rpm. Vivaspin 500 centrifugal concentrator (1 000k MWCO, Sartorious, Fisher Scientific) 471 was then used at 9000 rpm, to further concentrate the samples to 100 μL and to ensure separation of 472 protein-coated liposomes from the remaining large unbound proteins. Corona-coated liposomes were then 473 washed 3 times with 100 μL HEPES buffer to remove weekly bound proteins. To validate the separation of 474 corona-coated liposomes from unbound proteins, the same procedure was performed with controls of plasma 475 samples (without prior incubation with liposomes (Figure S1). 476 477 Transmission Electron Microscopy (TEM). Bare and corona-coated liposomes were stained by uranyl 478 acetate solution 1% and visualized with transmission electron microscopy (FEI Tecnai 12 BioTwin) before 479 and after their in vivo interaction with plasma proteins. Samples were diluted to 0.5 mM lipid concentration 480 and carbon Film Mesh Copper Grids (CF400-Cu, Electron Microscopy Science) were used. 481 482 SDS-PAGE electrophoresis. Proteins associated with 0.025 μM of liposomes were loaded onto a 4-20% 483 NOVEX Tris-Glycine Protein Gel (ThermoFisher Scientific). The gel was run until the proteins neared the end 13 484 of the gel (25-40 minutes at 225V). Staining was performed with Imperial Gel Staining reagent (Sigma Life 485 Science). 486 487 Quantification of adsorbed proteins. Proteins associated with recovered liposomes were quantified by 488 BCA Protein assay kit according to manufacturer's instructions. To make sure that liposomes in solution do 489 not interfere with the absorbance at 562 nm we measured the absorbance of corona-coated liposomes in 490 HEPES buffer and subtracted it from the total absorbance, measured when corona-coated liposomes were 491 mixed with the BCA reagent. Lipid concentration was quantified by Stewart assay and Protein binding ability; 492 Pb values (μg of protein/µM lipid) were then calculated. 493 494 Mass Spectrometry. In-gel digestion of corona proteins (20ug/sample) was performed prior to LC-MS/MS 495 analysis, as we have previously described.15,16,30 Before SDS-PAGE samples were boiled for 5 min at 90°C 496 in the presence of Tris-Glycine SDS Sample Buffer and NuPAGE reducing agent (ThermoFisher). Digested 497 samples were analysed by LC-MS/MS using an UltiMate® 3000 Rapid Separation LC (RSLC, Dionex 498 Corporation, Sunnyvale, CA) coupled to a Q Exactive™ Hybrid Quadrupole-Orbitrap™ (Thermo Fisher 499 Scientific, Waltham, MA) mass spectrometer. 500 501 To assess the repeatability of the sample processing workflow we incubated Caelyx® liposomes with plasma 502 obtained from a healthy donor and repeated the same protocol for 6 times/replicates (Figure S2A). To 503 assess the analytical variation/repeatability of the LC:MS/MS platform two of the samples used above to 504 determine the repeatability of the sample processing workflow were run in triplicates (Figure S2B & S3). To 505 assess the linearity of the method we incubated the same concentration of Caelyx® liposomes (0.15 mM) in 506 full and diluted plasma (Figure S4). 507 508 Mass Spectrometry data analysis. To statistically compare the abundance of proteins identified in the 509 ‘healthy’ and ‘diseased’ coronas MS peak intensities were analyzed using Progenesis LC-MS software 510 (version 3.0; Nonlinear Dynamics). RAW files were imported into Progenesis LC-MS software (version 3.0; 511 Nonlinear Dynamics) with automatic feature detection enabled. The Progenesis QI default method of 512 normalisation was applied (‘Normalise to all proteins’) to compensate for experimental variations. A 513 representative reference run was selected automatically, to which all other runs were aligned in a pair-wise 514 manner. Automatic processing was selected to run with applied filters for peaks charge state (maximum 515 charge 5). Protein quantitation method was selected to be the relative quantitation using Hi-N with N=3 516 peptides to measure per protein. The resulting MS/MS peak lists were exported as a single Mascot generic 517 file and upload onto a local Mascot Server (version 2.3.0; Matrix Science, UK). The spectra were searched 518 against the UniProt database using the following parameters: tryptic enzyme digestion with one missed 519 cleavage allowed, peptide charge of +2 and +3, precursor mass tolerance of 15 mmu, fragment mass 520 tolerance of 8 ppm, oxidation of methionines as variable modifications and carbamidomethyl as fixed 521 modifications, with decoy database search disabled and ESI-QUAD-TOF the selected instrument. Each 522 search produced an XML file from Mascot and the resulted peptides (XML files) were imported back into 523 Progenesis LC-MS to assign peptides to features. The peptide intensities were compared between groups by 524 one way analysis of variance. Subsequently data were exported in Excel format. Finally, results were filtered 525 to present a mean normalized abundance of more than 50,000 in at least one of the two groups. 526 527 Mass Spectrometry data were also analysed with QIAGEN’s Ingenuity® Pathway Analysis (IPA®, QIAGEN 528 Redwood City, www.qiagen.com/ingenuity). Diseases and functions IPA tool was used to identify proteins 529 involved in ovarian carcinoma pathways. The biomarker overlay IPA tool was then used to identify proteins 530 described in the literature as potential biomarkers for ovarian cancer. 531 532 Enzyme-linked Immunosorbent Assay (ELISA). ELISA kits for human CA125 (MUC16, ab195213, Abcam, 533 UK), apolipoprotein AI (APOAI, ab189576, Abcam, UK), prealbumin (Transthyretin TTR, ab108895, Abcam, 534 UK), thrombospondin 1 (THBS1, ab193716, Abcam, UK), alpha-enolase (ENO1, ab181417, Abcam, UK), 535 transforming growth factor beta 1 (TGF-β1, DB100B, R&D Systems Europe, LTD.), nucleoside diphosphate 536 kinase A (NME1, orb406403, Biorbyt Ltd., UK), glucosidase 2 subunit beta (PRKCSH, EH2259, Wuhan Fine 537 Biotech Co., Ltd.) and disulphide-isomerase A4 (PDIA4, abx250438, Abbexa Ltd., UK) were purchased for the 538 quantitative measurement of each human protein in plasma. Experiments were performed according to 539 manufacturer’s instructions 14 540 541 Statistical analysis. Statistical analysis of the data was performed using GraphPad Prism software. Mann- 542 Whitney t-test was used for the quantification of the total amount of protein adsorbed (Pb values of Figure 543 1C) and for ELISA experiments (Figure 4B). 544 545 Author contributions 546 M.H. initiated, designed and performed the experiments, analyzed all data and took responsibility for 547 planning and writing the manuscript. L.P. contributed to the analysis of the mass spectrometry data. R.U. 548 provided guidance on the proteomic data analysis. J.R. and A.C. contributed to the clinical design, provided 549 oversight of the ethical approval process and were responsible for access and storage of the samples in the 550 MCRC Biobank. K. K. provided intellectual input throughout the study and contributed to the writing of the 551 manuscript. 552 553 Data availability 554 The raw/processed data required to reproduce these findings are all included in the supporting information. 555 556 Acknowledgements 557 This research was partially funded by the CRUK Pioneer Award (C54921/A25435). We would like to 558 acknowledge the CRUK/NIHR Manchester Experimental Cancer Medicine Centre for providing research 559 nurse funding and their clinical infrastructure used for the blood sampling. This work was partially funded by 560 the CRUK Centre (C147/A18083) and Major Centre Award (C147/A25254) awarded to the Manchester 561 Cancer Research Centre. We would like to acknowledge the CRUK/NIHR Manchester Experimental Cancer 562 Medicine Centre for providing research nurse funding and their clinical infrastructure used for the blood 563 sampling. Authors wish to acknowledge Manchester Cancer Research (MCRC) Biobank team for their 564 assistance in the collection of blood samples. The authors also would like to thank the Faculty of Life 565 Sciences EM Facility at the University of Manchester for their assistance in Electron Microscopy imaging. In 566 addition, Biological Mass Spectrometry (Bio-MS) Facility staff at the University of Manchester for their 567 support. We would also like to thank the patients for their kind donation of blood samples to support this 568 research. 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 15 591 592 References 593 594 1 Fuzery, A. K., Levin, J., Chan, M. M. & Chan, D. W. Translation of proteomic biomarkers into FDA 595 approved cancer diagnostics: issues and challenges. Clin Proteom 10, doi:Artn 13 596 10.1186/1559-0275-10-13 (2013). 597 2 Anderson, N. L. & Anderson, N. G. The human plasma proteome: history, character, and diagnostic 598 prospects. Molecular & cellular proteomics : MCP 1, 845-867 (2002). 599 3 Hanash, S. M., Pitteri, S. J. & Faca, V. M. Mining the plasma proteome for cancer biomarkers. 600 Nature 452, 571-579, doi:10.1038/nature06916 (2008). 601 4 Rifai, N., Gillette, M. A. & Carr, S. A. Protein biomarker discovery and validation: the long and 602 uncertain path to clinical utility. Nat biotechnol 24, 971-983, doi:10.1038/nbt1235 (2006). 603 5 Liotta, L. A., Ferrari, M. & Petricoin, E. Clinical proteomics: written in blood. Nature 425, 905, 604 doi:10.1038/425905a (2003). 605 6 Tam, J. O. et al. A comparison of nanoparticle-antibody conjugation strategies in sandwich 606 immunoassays. J Immunoassay Immunochem 38, 355-377, doi:10.1080/15321819.2016.1269338 607 (2017). 608 7 Huo, Q. et al. A facile nanoparticle immunoassay for cancer biomarker discovery. J 609 Nanobiotechnology 9, 20, doi:10.1186/1477-3155-9-20 (2011). 610 8 Luchini, A. et al. Smart hydrogel particles: biomarker harvesting: one-step affinity purification, size 611 exclusion, and protection against degradation. Nano letters 8, 350-361, doi:10.1021/nl072174l 612 (2008). 613 9 Moro, L., Turemis, M., Marini, B., Ippodrino, R. & Giardi, M. T. Better together: Strategies based on 614 magnetic particles and quantum dots for improved biosensing. Biotechnol Adv 35, 51-63, 615 doi:10.1016/j.biotechadv.2016.11.007 (2017). 616 10 Hadjidemetriou, M., Al-Ahmady, Z., Buggio, M., Swift, J. & Kostarelos, K. A novel scavenging tool for 617 cancer biomarker discovery based on the blood-circulating nanoparticle protein corona. Biomaterials 618 188, 118-129, doi:10.1016/j.biomaterials.2018.10.011 (2019). 619 11 Hadjidemetriou, M. & Kostarelos, K. Nanomedicine: Evolution of the nanoparticle corona. Nat 620 Nanotechnol 12, 288-290, doi:10.1038/nnano.2017.61 (2017). 621 12 Cedervall, T. et al. Understanding the nanoparticle-protein corona using methods to quantify 622 exchange rates and affinities of proteins for nanoparticles. P Natl Acad Sci USA 104, 2050-2055, 623 doi:DOI 10.1073/pnas.0608582104 (2007). 624 13 Al-Ahmady, Z. S., Hadjidemetriou, M., Gubbins, J. & Kostarelos, K. Formation of protein corona in 625 vivo affects drug release from temperature-sensitive liposomes. J Control Release 276, 157-167, 626 doi:10.1016/j.jconrel.2018.02.038 (2018). 627 14 Garcia-Alvarez, R., Hadjidemetriou, M., Sanchez-Iglesias, A., Liz-Marzan, L. M. & Kostarelos, K. In 628 vivo formation of protein corona on gold nanoparticles. The effect of their size and shape. Nanoscale 629 10, 1256-1264, doi:10.1039/c7nr08322j (2018). 630 15 Hadjidemetriou, M., Al-Ahmady, Z. & Kostarelos, K. Time-evolution of in vivo protein corona onto 631 blood-circulating PEGylated liposomal doxorubicin (DOXIL) nanoparticles. Nanoscale 8, 6948-6957, 632 doi:10.1039/c5nr09158f (2016). 633 16 Hadjidemetriou, M. et al. In Vivo Biomolecule Corona around Blood-Circulating, Clinically Used and 634 Antibody-Targeted Lipid Bilayer Nanoscale Vesicles. ACS nano 9, 8142-8156, 635 doi:10.1021/acsnano.5b03300 (2015). 636 17 Chen, F. et al. Complement proteins bind to nanoparticle protein corona and undergo dynamic 637 exchange in vivo. Nat Nanotechnol 12, 387-393, doi:10.1038/nnano.2016.269 (2017). 638 18 Hadjidemetriou, M. et al. The Human In Vivo Biomolecule Corona onto PEGylated Liposomes: A 639 Proof-of-Concept Clinical Study. Adv Mater, e1803335, doi:10.1002/adma.201803335 (2018). 640 19 Ke, P. C., Lin, S., Parak, W. J., Davis, T. P. & Caruso, F. A Decade of the Protein Corona. ACS nano 641 11, 11773-11776, doi:10.1021/acsnano.7b08008 (2017). 642 20 Westmeier, D. et al. Nanomaterial-microbe cross-talk: physicochemical principles and 643 (patho)biological consequences. Chem Soc Rev 47, 5312-5337, doi:10.1039/c6cs00691d (2018). 644 21 Miotto, G. et al. Protein corona as a proteome fingerprint: The example of hidden biomarkers for cow 645 mastitis. Colloids Surf B Biointerfaces 140, 40-49, doi:10.1016/j.colsurfb.2015.11.043 (2016). 16 646 22 Magro, M. et al. Analysis of hard protein corona composition on selective iron oxide nanoparticles by 647 MALDI-TOF mass spectrometry: identification and amplification of a hidden mastitis biomarker in 648 milk proteome. Anal Bioanal Chem 410, 2949-2959, doi:10.1007/s00216-018-0976-z (2018). 649 23 Schiess, R., Wollscheid, B. & Aebersold, R. Targeted proteomic strategy for clinical biomarker 650 discovery. Mol Oncol 3, 33-44, doi:10.1016/j.molonc.2008.12.001 (2009). 651 24 Nolen, B. M. & Lokshin, A. E. Biomarker testing for ovarian cancer: clinical utility of multiplex assays. 652 Mol Diagn Ther 17, 139-146, doi:10.1007/s40291-013-0027-6 (2013). 653 25 Office for National Statistics. Cancer survival by stage at diagnosis for England (experimental 654 statistics) , 2016). 655 26 Nolen, B. M. & Lokshin, A. E. Protein biomarkers of ovarian cancer: the forest and the trees. Future 656 Oncol 8, 55-71, doi:10.2217/fon.11.135 (2012). 657 27 Sandin, M., Chawade, A. & Levander, F. Is label-free LC-MS/MS ready for biomarker discovery? 658 Proteomics Clin Appl 9, 289-294, doi:10.1002/prca.201400202 (2015). 659 28 Longo, C. et al. Core-shell hydrogel particles harvest, concentrate and preserve labile low 660 abundance biomarkers. PloS one 4, e4763, doi:10.1371/journal.pone.0004763 (2009). 661 29 Schottler, S., Klein, K., Landfester, K. & Mailander, V. Protein source and choice of anticoagulant 662 decisively affect nanoparticle protein corona and cellular uptake. Nanoscale 8, 5526-5536, 663 doi:10.1039/c5nr08196c (2016). 664 30 Al-Ahmady, Z. S., Hadjidemetriou, M., Gubbins, J. & Kostarelos, K. Formation of protein corona in 665 vivo affects drug release from temperature-sensitive liposomes. Journal of Controlled Release 276, 666 157-167, doi:10.1016/j.jconrel.2018.02.038 (2018). 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 17 FIGURE LEGENDS Figure 1: Protein corona formation after the ex vivo incubation of PEGylated, doxorubicin- ® encapsulated liposomes (Caelyx ) with plasma samples obtained from healthy controls (n=10) and ovarian carcinoma patients (n=19). (A) Schematic description of the experimental design. Caelyx® liposomes were incubated ex vivo with plasma samples obtained from patients with recurrent ovarian carcinoma (n=19) and from healthy donors (n=10) for 90 min at 37°C. Corona-coated liposomes were isolated and purified from unbound proteins by size exclusion chromatography and membrane ultrafiltration. ‘Healthy’ and ‘diseased’ protein coronas were comprehensively characterized and compared by label-free mass spectrometry to identify differentially abundant potential biomarker proteins. Selected potential biomarker proteins were further validated by commercially available ELISA kits. (B) Negative stain TEM of liposomes before and after corona formation. All scale bars are 100nm. (C) The total amount of protein adsorbed onto the surface of liposomes recovered from plasma samples obtained from healthy and ovarian carcinoma patients expressed as Pb values (μg of protein/μM lipid). Individual biological replicates are shown. Error bars indicate mean +/− SEM. (Mann-Whitney t-test; ****indicates p<0.0001). (D) Imperial stained SDS-PAGE gel of representative ‘healthy’ and ‘diseased’ corona samples. Figure 2: Proteomic comparison of the liposomal protein coronas formed in plasma samples obtained from ovarian carcinoma patients (n=19) and healthy controls (n=10). MS peak intensities were analyzed using Progenesis LC-MS software (version 3.0; Nonlinear Dynamics). Results were filtered to present a mean normalized abundance of more than 50,000 in at least one of the two groups. The peptide intensities were compared between groups by one way analysis of variance (ANOVA). (A) Heatmap of Normalized Abundance (NA) values of proteins found to be differentially expressed between ‘healthy’ and ‘diseased’ coronas. Only proteins with p value < 0.05 are shown (n=413). Proteins are classified from highest to the lowest max fold change. Average abundance of each protein for each group is also shown. The full list of differentially abundant corona proteins and their respective mean normalized abundance, p value and max fold change are shown in Table S4. (B) Volcano plot displays the relationship between fold change and significance between the two groups. The y-axis depicts the negative log10 of p- values and the x-axis is the difference in expression between the two groups as log10 fold changes. Only proteins with at least 2-fold change and a p value<0.01 value are highlighted (n=303; downregulated proteins are shown in red and upregulated proteins in blue). Figure 3: Ingenuity Pathway Analysis (IPA) of potential biomarker corona proteins. Out of 413 potential biomarker proteins n=15 (shown in orange) were previously reported as potential biomarkers for ovarian cancer and n=57(shown in green) have been previously associated with ovarian carcinoma pathways. The name of proteins illustrated in the diagram and their respective gene symbols are shown in Table S5. Figure 4: ELISA validation of the nanoparticle protein corona technology. (A) Scatter plot displays the relationship between fold change and significance of the nine potential biomarker proteins selected to be further validated. The y-axis depicts the negative log10 of p-values and the x-axis is the difference in expression between the two groups as log10 fold changes. Clinically used biomarkers are shown in orange, proteins previously associated with ovarian carcinoma pathways (according to IPA) are shown in green and proteins that have not been previously associated with ovarian carcinoma pathways are shown in blue. (B) Plasma concentration profiles of selected potential biomarker proteins in healthy controls (n=10) and ovarian carcinoma patients (n=15-17) and their respective ROC curves based on ELISA assays. AUC values are also shown; Mann-Whitney t-test; * indicates p<0.05, *** indicates p < 0.001, **** indicated p<0.0001. 18 (CA125=mucin 16, APOA1=apolipoprotein A1, TTR=transthyretin, THBS1=thrombospondin 1, ENO1=enolase 1, TGF-b1= transforming growth factor b1, NME1= nucleoside diphosphate kinase A, PRKCSH=glucosidase 2 subunit beta and PDIA4= protein disulfide-isomerase A4). 19 FIGURE 1 20 FIGURE 2 21 FIGURE 3 22 FIGURE 4 23