Mass-spectrometry-based proteomics enables rapid and accurate diagnosis of Lyme neuroborreliosis in adults

Mass-spectrometry-based proteomics enables rapid and accurate diagnosis of Lyme neuroborreliosis in adults | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Mass-spectrometry-based proteomics enables rapid and accurate diagnosis of Lyme neuroborreliosis in adults Nicolai Wewer Albrechtsen, Annelaura Nielsen, Lasse Fjordside, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6048306/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 27 Oct, 2025 Read the published version in Nature Communications → Version 1 posted You are reading this latest preprint version Abstract Lyme neuroborreliosis (LNB), a severe nervous system infection caused by tick-borne spirochetes of the Borrelia burgdorferi sensu lato complex, represents one of the most frequent bacterial infections of the nervous system in Europe. Early diagnosis remains challenging due to limited sensitivity of current methods and requires invasive lumbar punctures, underscoring the need for improved, less invasive diagnostic tools. Here, we applied mass spectrometry-based proteomics to analyse 308 cerebrospinal fluid (CSF) samples and 207 plasma samples from patients with LNB, viral meningitis, controls and other manifestations of Lyme borreliosis. Diagnostic panels of regulated proteins were identified and evaluated through machine learning-assisted proteome analyses. In CSF, the classifier distinguished LNB from viral meningitis and controls with AUCs of 0.92 and 0.90, respectively. In plasma, LNB was distinguished from controls with an AUC of 0.80. Our findings highlight the diagnostic potential of machine learning-assisted proteomics for LNB in CSF and plasma. Health sciences/Health care/Diagnosis/Laboratory techniques and procedures Health sciences/Diseases/Infectious diseases/Bacterial infection Tick-Borne Diseases Borrelia Lyme Borreliosis Lyme Disease Lyme Neuroborreliosis Central Nervous System Infections Meningitis Viral Delayed Diagnosis Diagnostic Techniques and Procedures Proteomics Machine Learning Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 INTRODUCTION Lyme neuroborreliosis (LNB) is a bacterial infection of the nervous system caused by spirochetes of the Borrelia burgdorferi sensu lato ( B. burgdorferi s.l.) complex transmitted through bites from hard-shelled ticks of the Ixodes genus 1 . With an approximate incidence in endemic countries between 3.2 and 6.3 per 100 000 persons/year 2–4 LNB is among the most frequent bacterial infections of the nervous system in Europe 5,6 . LNB can cause a wide range of clinical neurological conditions, but most frequently presents as a subacute painful meningo-radiculitis with radiating pain from the spine to neck, extremities, thorax or abdomen, lymphocytic meningitis and/or cranial neuropathies e.g., facial nerve palsy 7 . If relevant antibiotic therapy is administered at an early stage of disease, LNB has a favourable long-term prognosis 8–10 . However, delayed treatment is associated with an increased risk of residual symptoms and long-term sequelae 9 . In countries where LNB is endemic, the average time from onset of neurological symptoms to diagnosis is typically around 3 weeks and has remained unchanged for the last four decades 6 . The cause of the diagnostic delay is multifactorial. Overlapping symptomatology with other more common diseases and the fact that only around 40% of patients with LNB report a tick bite and only 25% report a history of the classic skin rash erythema migrans makes it less likely that physicians and patients consider LNB as a differential diagnosis, especially in the absence of facial nerve palsy 6,8 . However, even when LNB is clinically suspected at an early stage, the diagnosis requires detection of B. burgdorferi s.l.- specific antibodies in the cerebrospinal fluid (CSF). Lumbar puncture is an uncomfortable and expensive procedure that may require hospitalization and generel anesthesia for children. Direct pathogen identification of B. burgdorferi s.l. with polymerase chain reaction (PCR), cultivation or a combination has retained an exceedingly poor level of sensitivity both in blood and CSF despite several attempts to improve tools and techniques over the years 11–14 . Further, the serological response in LNB is not detectable early in the course of disease and antibodies may remain elevated for months to several years after full recovery of well-treated infections making it impossible to discriminate between past and current infection. Furthermore, the B. burgdorferi s.l.-specific IgG and IgM in blood, poorly predict nervous system involvement 15 . Thus, in order to reduce the diagnostic delay in LNB, the need for novel diagnostic tools is evident. Especially improved blood-based diagnostic tests would be extremely valuable as they would both potentially help reduce the diagnostic delay and offer a less invasive diagnostic tool. Untargeted proteomic analysis enables accurate measurement of proteins in a sample and can be used to compare patterns of protein regulation in patients with and without disease to identify disease-specific protein response signatures 16 . When combined with machine-learning (ML) for data-evaluation and analysis, proteomic test results can be made available with turnaround times of just 2 hours making it highly attractive in real-life clinical settings. The aim of this study was to explore the potential of ML-assisted MS-based proteomics as a novel diagnostic tool for LNB in CSF and plasma of adults. RESULTS Study population A total of 483 CSF and plasma samples from adult individuals were eligible for proteomics analysis: 155 samples from patients with LNB, 127 samples from patients with viral meningitis, 169 controls and 32 samples from patients with erythema migrans, acrodermatitis chronica atrophicans and post-treatment-Lyme-disease-syndrome (Table 1). Table 1 Baseline characteristics of patients with Lyme neuroborreliosis, viral meningitis, controls, and other Lyme borreliosis manifestations. Cohort Diagnosis n Age Mean [IQR] Females n (%) CSF Development Lyme neuroborreliosis 49 51[40-66] 26 (53) N = 145 Viral meningitis 44 38[27-42] 21 (47) Control 52 36[25-42] 17 (32) CSF Validation Lyme neuroborreliosis 69 53[46-67] 28 (40) N = 163 Viral meningitis 45 31[25-37] 23 (51) Control 49 44[33-53] 24 (48) Plasma Development Lyme neuroborreliosis 27 64[61-69] 10 (37) N = 95 Viral meningitis 20 35[27-38] 13 (65) Control 48 46[33-60] 28 (58) Plasma Validation Lyme neuroborreliosis 10 59[56-63] 4 (40) N = 80 Viral meningitis 18 38[24-43] 10 (55) Control 20 36[28-42] 6 (30) Acrodermatitis chronica atrophicans 6 66[59-74] 6 (100) Erythema migrans 9 47[36-54] 5 (55) Post-treatment-Lyme-disease-syndrome 17 58[48-64] 10 (58) Abbreviations: CSF = cerebrospinal fluid, IQR = Inner Quartile Range CSF proteomics CSF development cohort A total of 1,865 proteins were identified in CSF, with a median of 773 quantified pr. sample (Supplementary Figure 1). Fourteen samples were excluded due to low protein numbers. After filtering for ­missingness, a total of 654 proteins were included for further analysis. A total of 176 proteins were significantly different between LNB and viral meningitis, of these 10 had an absolute log2 fold change larger than 1 (Figure 1A, Supplementary Table1). A total of 464 proteins were significantly different between LNB and controls, of these 41 had an absolute log2 fold change larger than 1 (Figure 1B, Supplementary Table1). Of the significant proteins with high log2 fold changes in the two comparisons, 7 proteins overlapped (Figure 1C). CSF validation cohort In the validation cohort a total of 147 proteins were significantly different in LNB compared to viral meningitis and controls with high log2 fold changes, where 43/46 proteins from the development cohort overlapped (Figure 1B and Supplementary Figure 2,3). Protein signatures of LNB in CSF Protein signatures were dominated by immunoglobulins (e.g., IGLV3-25, IGLV2-18, FCGBP, IGHM), proteins involved in innate immune responses (e.g. enolase 1, S100A9), neuroendocrine signalling (e.g. ECRG4, CHGA), cell-migration and cell-damage (e.g. PFN1, APCS, YWHAZ, H4C1, ACTA2) (Figure 1C). Immunoglobulins were generally upregulated, while proteins involved in innate immune responses and neuroendocrine signalling were relatively downregulated in LNB compared to viral meningitis and controls. Diagnostic classifier for LNB based on machine learning and CSF proteomics The protein profiles of the CSF development cohort were used to develop a ML model to evaluate the potential of MS-based proteomics coupled with ML as diagnostic support in LNB. Of the twelve ML models tested, the Logistic Regression (LR) model obtained the highest performance in the classification of viral meningitis vs. LNB, whereas the Support Vector Classifier (SVC) model produced the best result in the control vs. LNB classification (Supplementary Figure 4 and 5 respectively). The performance of the diagnostic classification model of viral meningitis vs. LNB on the test set had an area under the curve (AUC) of 0.91 (std. = 0.11) and a Matthews Correlation Coefficient (MCC) of 0.81 (std.=0.08) (Figure 2A). When applied on the validation cohort, the model obtained an AUC and MCC of 0.92 (std. = 0.02) and 0.7 (std.=0.06). According to their SHAP (SHapley Additive exPlanations) values, the most important proteins in the discrimination between viral meningitis and LNB are visualized in Figure 2B. The performance of the diagnostic classification model of control vs. LNB on the test set had an AUC of 0.93 (std. = 0.07) and a MCC of 0.76 (std.=0.14) (Figure 2C). When applied on the validation cohort, the model obtained an AUC and MCC of 0.9 (std. = 0.01) and 0.63 (std.=0.06). The most important proteins in the model according to their SHAP values (Figure 2D) were similarly related to innate and humoral immune responses, neuroendocrine signalling, and cell-damage . Plasma proteomics Plasma development cohort A total of 379 proteins were identified in plasma, with a median of 268 proteins quantified per sample (Supplementary Figure 6). One sample was excluded due to low protein number. After filtering for missingness, a total of 232 proteins were included for further analysis. A total of 61 proteins were significantly different between LNB and viral meningitis (Figure 3A, Supplementary Table2) and 63 for LNB versus controls (Figure 3B). The overlap between these two comparisons of significant proteins were 36 (Figure 3C, Supplementary Table2). Plasma validation cohort In the validation cohort a total of 25 proteins were significantly different in LNB compared to viral meningitis and controls. Ten overlapped with the significant proteins from the development cohort (Figure 3B and Supplementary Figure 7,8). Protein signatures of LNB in plasma The differences in protein profiles between the three patient groups are dominated by proteins involved in innate and humoral immune responses (Figure 3C). When comparing protein profiles of plasma samples from LNB and viral meningitis, patients with LNB generally seem to have a relative upregulation of proteins associated with innate immunity and complement activation (e.g., FCN3, SERPING1, SERPINA5), lipid metabolism (e.g., APOE, APOC1, APOM), and coagulation regulation (e.g., F13A1, PROC). In contrast, patients with viral meningitis have a plasma protein profile dominated by acute-phase and pro-inflammatory markers (e.g., CRP, S100A8, S100A9) and immunoglobulin production. In the comparison between LNB and controls, upregulated plasma-proteins in LNB are similarly involved in complement activation (e.g., C3, C5), coagulation and inflammation regulation (e.g., SERPINA3, SERPIND1), and immune system modulators (e.g., HP, ITIH4). In contrast to CSF, most immunoglobulins had a higher protein level in the plasma of control individuals compared to LNB patients. Diagnostic classifier for LNB based on machine learning and plasma proteomics Similar to the CSF analysis, plasma proteome data were subjected to ML algorithm- and hyperparameter selection in a 3-fold cross validation scheme. It was not deemed feasible to develop a model for the discrimination of viral meningitis and LNB due to the low number of viral meningitis cases, hence, the only model tested was the discrimination of controls and LNB. Of the twelve models tested, the SVC model obtained the best performance in the test set (Supplementary Figure 9) with an AUC of 0.96 (std.=0.03) and a MCC of 0.83 (std.=0.008) (Figure 4A). When applied on the validation cohort the model obtained an AUC of 0.80 (std.=0.02) and a MCC of 0.48 (std.=0.04) (Figure 4A). According to the SHAP feature importance values, the most predictive proteins for the diagnostic classification of controls vs. LNB (Figure 4B) were associated with innate immunity, humoral immune defence, coagulation and cellular metabolism. When the LNB classification model was applied to other diagnostic groups, it identified groups with active Borrelia burgdorferi s.l. infections; acrodermatitis chronica atrophicans and erythema migrans, as having significantly higher similarities to the LNB profile than groups with absence of active infection; Post-treatment-Lyme-disease-syndrome or unrelated aetiology; viral meningitis. Mann-Whitney U Test (acrodermatitis chronica atrophicans+erythema migrans vs Post-treatment-Lyme-disease-syndrome +Viral meningitis): U statistic = 5475.0, p-value = 2.65e-07. (Figure 4c). Comparison of protein profiles in CSF and plasma We found an overlap of 72 proteins that were significantly different between LNB and controls or viral meningitis in both CSF and plasma (Figure 5A, Supplementary Table 3). The majority of overlapping proteins (62/72) were relatively upregulated in CSF from patients with LNB compared to controls and viral meningitis, and 30 of these were inversely regulated in plasma. This group primarily included immunoglobulins (18/30) (Figure 5B). DISCUSSION Our study shows that MS-based proteomics can accurately differentiate patients with LNB from patients with viral meningitis and non-LNB controls. Notably, we observed a high level of diagnostic accuracy in plasma, highlighting the potential of blood-based proteomics as a clinically valuable tool. Given the current need for novel diagnostic approaches to reduce delays and reliance on lumbar punctures, these findings could have significant implications for both initial diagnosis and monitoring of treatment response in LNB. To our knowledge, this is the first study reporting the use of untargeted proteomics as a diagnostic tool in a large and well-defined cohort of adult LNB patients, investigating both CSF and plasma samples. Additionally, the inclusion of viral meningitis cases strengthens our findings by assessing the ability of proteomics to distinguish between two conditions with overlapping symptomatology and CSF findings. This approach enhances the translational potential of our results. Previous studies investigating the diagnostic potential of proteomics in LNB have been limited in scope and methodology. A small US-based study assessing the diagnostic utility of proteomics in Lyme disease including a small sub-group with CNS involvement; Angel et al. (2013) found a panel of 13 proteins discriminated between cases and controls reaching an AUC of 0.8 17 . However, the focus on other Lyme disease manifestations than LNB likely reflects the, somewhat debated, geographical difference in prevalence of neurological involvement with a B. burgdorferi s.l.strain-dependent lower incidence in North America compared to Europe 18,19 . More recent European studies have investigated the diagnostic use of protein panels in LNB, though with different methodologies and sample types: Fredriksson et al. (Sweden, 2024) analysed serum from 119 paediatric patients with LNB (n=61) and non-LNB (n=58) using a multiplex proximity extension assay that included a panel of 92 predefined inflammatory markers and found that a 5-protein-panel identified LNB with an AUC of 0.88 20 . Though the diagnostic groups and the use of blood-samples were comparable to our study, it is unclear to what extent results in children can be extrapolated to adults and the multiplex assay used is not comparable to the untargeted MS-based proteomics used in our study. Gęgotek et al. (Poland, 2024) performed LC-MS based proteomics on serum from patients with LNB (n=10) and controls (n=27) and found significant differences, but did not report measures of diagnostic accuracy 21 . Other investigations have primarily focused on characterizing patients with post-treatment-Lyme-disease-syndrome rather than acute LNB 22,23 . Despite differences in study design, a consistent pattern of immune activation has emerged across studies, emphasizing the role of innate immune responses, complement system and humoral immune defence in the human host response to B. burgdorferi s.l. infection. Our proteomic findings in CSF, indicate a distinct immune response characterized by elevated levels of immunoglobulin chains, complement-related proteins, and proteins involved in immune cell migration and cytoskeletal dynamics. These findings suggest a highly targeted immune response against B. burgdorferi s.l. antigens, involving adaptive immunity with significant antibody production and complement activation. This aligns with established knowledge of host immune response during infections with B. burgdorferi s.l. 24 . Interestingly, the antibody response exhibited opposite regulation in CSF and plasma, likely illustrating the significant compartmentalization of the humoral immune response to the CNS in LNB and reflects the fact that all patients with LNB in this study had a positive B. burgdorferi s.l.- specific intrathecal antibody index. Moreover, we observed overlapping regulation of key immune-related proteins in both plasma and CSF, suggesting a coordinated immune response across compartments. Plasma and CSF both showed changes in immunoglobulin chains, complement factors, and innate immune modulators in LNB compared to controls, indicating a systemic and localized immune activation. The CSF findings highlight immunoglobulin diversity and a direct antibody response, reflecting the local immune reaction in the CNS. Meanwhile, plasma proteomics revealed a strong systemic immune engagement, with upregulation of innate immune components such as ficolins and coagulation proteins, underscoring systemic inflammation. Notably, viral meningitis exhibited a different pattern, with both plasma and CSF showing elevated acute-phase reactants (e.g., CRP, S100A9) and immunoglobulins, consistent with a strong inflammatory and humoral response. Opposing regulation of certain proteins in CSF and plasma in LNB likely reflects a compartmentalized immune strategy, where the CNS immune response is tightly regulated, while plasma changes mirror systemic inflammation. Additionally, factors such as blood-brain barrier permeability, local versus systemic protein production, and clearance mechanisms may further influence these patterns. Additionally, structural proteins involved in immune cell motility were upregulated in CSF, indicating active immune cell recruitment within the CNS, specifically in LNB. Viral meningitis patients present a pattern more focused on inflammatory markers and metabolic activity, consistent with an acute viral immune response. A small cluster of cytoskeleton and extracellular matrix proteins (VIM, ACTA2, PFN1, FN1 and ACTB) emerged among the top discriminatory markers, suggesting increased cell-turnover in LNB. While expected in LNB vs. controls, this is more surprising when compared to viral meningitis. Notably, B. burgdorferi s.l. spirochetes are hypothesized to extravasate early upon transmission and travel through the extracellular matrix to peripheral nerves eventually reaching the borders of the CNS and creating the inflammatory basis of radiculoneuritis 18,25 . Vimentin (VIM), a class III intermediate filament that regulates myelination in axons of peripheral nerves 26 , is particularly relevant given our limited understanding of CNS entry by B. burgdorferi s.l. spirochetes, the frequent cranial nerve involvement in LNB with painful radiculoneuritis 9,27–29 and because proteins released from damaged nerves to the peripheral circulation could be potential biomarker candidates for early LNB, as previously shown with neurofilament light chain (NfL) 30 . Additionally, fibronectin (FN1) which is also found in this cluster, is considered essential for B. burgdorferi s.l. cell adhesion and migration through binding to B. burgdorferi s.l. surface proteins BBK32 31 and RevA 32 . The clinical implications of our findings are considerable. If validated in independent cohorts, CSF proteomics could aid in diagnosing LNB among adult patients with CNS infections with unknown aetiology. This may be particularly relevant for PCR-negative suspected viral meningitis cases, where CSF proteomics suggestive of LNB could prompt further testing or the direct initiation of relevant antibacterial treatment e.g. doxycycline. However, the most significant clinical implication would be the ability to diagnose LNB using a blood sample alone - a true gamechanger in LNB diagnostics. It would spare patients the discomfort and risk for complications of lumbar puncture and reduce healthcare-costs associated with referral to specialized healthcare facilities. A diagnostic blood test for LNB would likely reduce the diagnostic delay and thereby reduce the risk of residual symptoms. Even a modest improvement over current serology-based approaches, could lead to fewer invasive procedures and significantly ease treatment-effect monitoring. This would be particularly beneficial in children, who account for around 30% of all LNB patients, and often require general anaesthesia to have a lumbar puncture performed 33 34,35 . Moreover, a blood-based test could address the unmet need for a less invasive method to monitor treatment response in patients with persistent symptoms after LNB treatment. While these findings demonstrate the diagnostic potential of proteomics in LNB, it is important to consider the strengths and limitations of our study. With 483 samples from patients with LNB, viral meningitis and non-LNB controls, our study represents the largest investigation to date. While we could not stratify patients in full accordance with the European Federation of Neurological Societies (EFNS) diagnostic criteria 7 due to limited clinical data, our use of uniform case definition and the specificity of a first time positive B. burgdorferi sl. intrathecal antibody index minimizes the risk of false-positive LNB diagnoses 36 . Our selection of clinically relevant comparison groups, including viral meningitis and patients who were clinically suspected for LNB, enhances the real-world applicability of our findings by testing the robustness of proteomics in distinguishing LNB from other CNS infections. A key strength of our study is its focus on diagnostic feasibility, employing a proteomic workflow with a short turnaround time suitable for clinical implementation. However, the limited detection range inherent to this approach means that certain low-abundance but highly disease-specific proteins, such as CXCL13, may not be captured. Given that previous proteomic studies in LNB have also failed to detect CXCL13, this highlights a general challenge in proteomic biomarker discovery rather than a limitation of our study specifically 37 . As with all retrospective analyses, our study is constrained by sample availability and cohort composition. Differences in cohort sizes necessitated varying cross-validation strategies, and some observed differences in protein signatures may be influenced by factors such as disease stage, age, and sex 38–42 . However, our study population reflects real-world clinical variation, as patients were primarily recruited from referral hospitals. Furthermore, prioritizing clinically relevant diagnostic groups over strictly age- and sex-matched controls enhances the translational value of our findings. In conclusion, our findings demonstrate that MS-based proteomics can accurately distinguish LNB from controls in both CSF and plasma. These results warrant further prospective validation, particularly with longitudinal sampling, to assess the diagnostic utility of plasma proteomics and its potential role in monitoring treatment response in LNB patients. Declarations ACKNOWEDGEMENTS We express our gratitude to Christine Rasmussen from the Department of Clinical Biochemistry at Copenhagen University Hospital, Bispebjerg, for her dedicated efforts in planning, preparing, and conducting the proteomic analysis. Additionally, we extend our appreciation to the Clinical Proteomic Group at the NNF Center for Protein Research, University of Copenhagen, for sharing their expertise on Mass Spectrometry. We would also like to thank lab technicians Stine Østergaard and Dorthe Hass from the clinical research unit at the Department of Infectious Diseases, Rigshospitalet, for their great help with handling and sending samples for analysis. Conflict of interest disclosures . AML reports speakers’ honorarium/travel grants/advisory board activity and unrestricted grant from Gilead, speakers honorarium/travel grants from GSK, speaker’s honorarium/advisory board activity from Pfizer outside this work. NJWA has received funding, served on scientific advisory panels, and/or speakers bureaus for Boehringer Ingelheim, MSD/Merck, Novo Nordisk, EvoSep, ROCHE, Janssen, and Mercodia. AJH reports a research collaboration agreement with Pfizer unrelated to this work. M.M. is an indirect shareholder in Evosep Biosystems. None of the other authors report any conflict of interests. Funding NJWA is supported by European Foundation for the Study of Diabetes Future Leader Award (NNF21SA0072746), Independent Research Fund Denmark, Sapere Aude (1052-00003B) and Novo Nordic Foundation (NNF23OC0084970, NNF19OC0055001 and NNF24OC0088402). Novo Nordisk Foundation Center for Protein Research is supported financially by the Novo Nordisk Foundation (Grant agreement NNF14CC0001). LF was supported by a research grant from the Research Fund of Copenhagen University Hospital - Rigshospitalet. AML was supported by a research grant from the Lundbeck foundation, Research Fund of Copenhagen University Hospital - Rigshospitalet, Independent Research Fund Denmark and Svend Andersen’s foundation. Role of the funder The Novo Nordisk Foundation, the Research Fund of Rigshospitalet, and the Lundbeck Foundation had no role in the design and conduct of the study; collection, management, analyses, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. Data and code availability The raw data and processed data have been deposited to the ProteomeXchange Consortium via the Proteomics Identifications Database (PRIDE) partner repository, the dataset identifier will be added upon publication 54 . The Jupyter notebooks are available at github upon request. References Hansen K, Crone C, Kristoferitsch W. Lyme Neuroborreliosis . Vol 115. 1st ed. Elsevier B.V.; 2013. doi:10.1016/B978-0-444-52902-2.00032-1 Dahl V, Wisell KT, Giske CG, Tegnell A, Wallensten A. Lyme neuroborreliosis epidemiology in Sweden 2010 to 2014: Clinical microbiology laboratories are a better data source than the hospital discharge diagnosis register. Eurosurveillance . 2019;24(20):1800453. doi:10.2807/1560-7917.ES.2019.24.20.1800453/CITE/PLAINTEXT Dessau RB, Espenhain L, Mølbak K, Krause TG, Voldstedlund M. Improving national surveillance of Lyme neuroborreliosis in Denmark through electronic reporting of specific antibody index testing from 2010 to 2012. Eurosurveillance . 2015;20(28):1-11. doi:10.2807/1560-7917.ES2015.20.28.21184 Tetens MM, Haahr R, Dessau RB, et al. Changes in Lyme neuroborreliosis incidence in Denmark, 1996 to 2015. Ticks Tick Borne Dis . 2020;11(6). doi:10.1016/j.ttbdis.2020.101549 Koelman DLH, Van Kassel MN, Bijlsma MW, Brouwer MC, Van De Beek D, Van Der Ende A. Changing Epidemiology of Bacterial Meningitis Since Introduction of Conjugate Vaccines: 3 Decades of National Meningitis Surveillance in The Netherlands. Clin Infect Dis An Off Publ Infect Dis Soc Am . 2021;73(5):e1099. doi:10.1093/CID/CIAA1774 Nordberg CL, Bodilsen J, Knudtzen FC, et al. Lyme neuroborreliosis in adults: A nationwide prospective cohort study. Ticks Tick Borne Dis . 2020;11(4):101411. doi:10.1016/j.ttbdis.2020.101411 Mygland Å, Ljøstad U, Fingerle V, Rupprecht T, Schmutzhard E, Steiner I. EFNS guidelines on the diagnosis and management of European lyme neuroborreliosis. Eur J Neurol . 2010;17(1):8-e4. doi:10.1111/j.1468-1331.2009.02862.x Obel N, Dessau RB, Krogfelt KA, et al. Long term survival, health, social functioning, and education in patients with European Lyme neuroborreliosis: nationwide population based cohort study. BMJ . 2018;361. doi:10.1136/bmj.k1998 Knudtzen FC, Andersen NS, Jensen TG, Skarphédinsson S. Characteristics and Clinical Outcome of Lyme Neuroborreliosis in a High Endemic Area, 1995-2014: A Retrospective Cohort Study in Denmark. Clin Infect Dis . 2017;65(9):1489-1495. doi:10.1093/cid/cix568 Haahr R, Tetens MM, Dessau RB, et al. Risk of Neurological Disorders in Patients With European Lyme Neuroborreliosis: A Nationwide, Population-Based Cohort Study. Clin Infect Dis . 2019;71(6):1511-1516. doi:10.1093/cid/ciz997 Nocton JJ, Bloom BJ, Rutledge BJ, et al. Detection of Borrelia burgdorferi DNA by polymerase chain reaction in cerebrospinal fluid in Lyme neuroborreliosis. J Infect Dis . 1996;174(3):623-627. doi:10.1093/infdis/174.3.623 Lebech AM, Hansen K, Brandrup F, Clemmensen O, Halkier-Sørensen L. Diagnostic value of PCR for detection of Borrelia burgdorferi DNA in clinical specimens from patients with erythema migrans and Lyme neuroborreliosis. Mol diagnosis a J devoted to Underst Hum Dis through Clin Appl Mol Biol . 2000;5(2):139-150. doi:10.1007/BF03262032 Skogman BH, Wilhelmsson P, Atallah S, Petersson AC, Ornstein K, Lindgren PE. Lyme neuroborreliosis in Swedish children—PCR as a complementary diagnostic method for detection of Borrelia burgdorferi sensu lato in cerebrospinal fluid. Eur J Clin Microbiol Infect Dis . 2021;40(5):1003-1012. doi:10.1007/s10096-020-04129-7 Leth TA, Nymark A, Knudtzen FC, et al. Detection of Borrelia burgdorferi sensu lato DNA in cerebrospinal fluid samples following pre-enrichment culture. Ticks Tick Borne Dis . 2023;14(3):102138. doi:https://doi.org/10.1016/j.ttbdis.2023.102138 Tetens MM, Dessau R, Ellermann-Eriksen S, et al. The diagnostic value of serum Borrelia burgdorferi antibodies and seroconversion after Lyme neuroborreliosis, a nationwide observational study. Clin Microbiol Infect . 2022;28(11):1500.e1-1500.e6. doi:10.1016/J.CMI.2022.06.001 Niu L, Thiele M, Geyer PE, et al. Noninvasive proteomic biomarkers for alcohol-related liver disease. Nat Med . 2022;28(6):1277. doi:10.1038/S41591-022-01850-Y Angel TE, Jacobs JM, Smith RP, et al. Cerebrospinal fluid proteome of patients with acute Lyme disease. J Proteome Res . 2012;11(10):4814-4822. doi:10.1021/PR300577P Halperin JJ, Eikeland R, Branda JA, Dersch R. Lyme neuroborreliosis: known knowns, known unknowns. Brain . 2022;145(8):2635-2647. doi:10.1093/brain/awac206 Koedel U, Fingerle V, Pfister HW. Lyme neuroborreliosis - Epidemiology, diagnosis and management. Nat Rev Neurol . 2015;11(8):446-456. doi:10.1038/nrneurol.2015.121 Fredriksson T, Brudin L, Henningsson AJ, Skogman BH, Tjernberg I. Diagnostic patterns of serum inflammatory protein markers in children with Lyme neuroborreliosis. Ticks Tick Borne Dis . 2024;15(4):102349. doi:10.1016/j.ttbdis.2024.102349 Gęgotek A, Skrzydlewska E, Groth M, Czupryna P, Moniuszko-Malinowska A. Changes in the serum proteome profile of patients with neuroborreliosis, foresters, and patients treated according to ILADS method. Microb Pathog . 2024;197(July). doi:10.1016/j.micpath.2024.107094 Nilsson K, Skoog E, Edvinsson M, Mårtensson A, Olsen B. Protein biomarker profiles in serum and CSF in 158 patients with PTLDS or persistent symptoms after presumed tick-bite exposure compared to those in patients with confirmed acute neuroborreliosis. PLoS One . 2022;17(11 November):1-17. doi:10.1371/journal.pone.0276407 Schutzer SE, Angel TE, Liu T, et al. Distinct cerebrospinal fluid proteomes differentiate post-treatment Lyme disease from Chronic fatigue syndrome. PLoS One . 2011;6(2):1-8. doi:10.1371/journal.pone.0017287 Bockenstedt LK, Wooten RM, Baumgarth N. Immune response to borrelia: Lessons from lyme disease spirochetes. Curr Issues Mol Biol . 2020;42:145-190. doi:10.21775/cimb.042.145 Rupprecht TA, Koedel U, Fingerle V, Pfister HW. The pathogenesis of lyme neuroborreliosis: From infection to inflammation. Mol Med . 2008;14(3-4):205-212. doi:10.2119/2007-00091.Rupprecht Triolo D, Dina G, Taveggia C, et al. Vimentin regulates peripheral nerve myelination. Development . 2012;139(7):1359-1367. doi:10.1242/dev.072371 Radzišauskienė D, Urbonienė J, Jasionis A, et al. Clinical and epidemiological features of Lyme neuroborreliosis in adults and factors associated with polyradiculitis, facial palsy and encephalitis or myelitis. Sci Rep . 2023;13(1):1-11. doi:10.1038/s41598-023-47312-4 Solheim AM, Skarstein I, Quarsten H, et al. Clinical and laboratory characteristics during a 1-year follow-up in European Lyme neuroborreliosis: A prospective cohort study. Eur J Neurol . 2024;(August):1-10. doi:10.1111/ene.16487 van Samkar A, Bruinsma RA, Vermeeren YM, et al. Clinical characteristics of Lyme neuroborreliosis in Dutch children and adults. Eur J Pediatr . 2023;182(3):1183-1189. doi:10.1007/s00431-022-04749-5 Mens H, Fjordside L, Gynthersen R, et al. Plasma neurofilament light significantly decreases following treatment in Lyme neuroborreliosis and not associated with persistent symptoms. doi:10.1111/ene.15707 Fischer JR, LeBlanc KT, Leong JM. Fibronectin binding protein BBK32 of the Lyme disease spirochete promotes bacterial attachment to glycosaminoglycans. Infect Immun . 2006;74(1):435-441. doi:10.1128/IAI.74.1.435-441.2006 Brissette CA, Bykowski T, Cooley AE, Bowman A, Stevenson B. Borrelia burgdorferi RevA antigen binds host fibronectin. Infect Immun . 2009;77(7):2802-2812. doi:10.1128/IAI.00227-09 Bruinsma RA, Zomer TP, Skogman BH, van Hensbroek MB, Hovius JW. Clinical manifestations of Lyme neuroborreliosis in children: a review. Eur J Pediatr . 2023;182(5):1965-1976. doi:10.1007/s00431-023-04811-w Garro A, Avery RA, Cohn KA, et al. Validation of the Rule of 7’s for Identifying Children at Low-risk for Lyme Meningitis. Pediatr Infect Dis J . 2021;40(4):306-309. doi:10.1097/INF.0000000000003003 Skogman BH, Sjöwall J, Lindgren PE. The NeBoP score - a clinical prediction test for evaluation of children with Lyme Neuroborreliosis in Europe. BMC Pediatr . 2015;15(1):1-9. doi:10.1186/s12887-015-0537-y Blanc F, Jaulhac B, Fleury M, et al. Relevance of the antibody index to diagnose Lyme neuroborreliosis among seropositive patients. Neurology . 2007;69(10):953-958. doi:10.1212/01.wnl.0000269672.17807.e0 Strle F, Henningsson AJ, Strle K. Diagnostic Utility of CXCL13 in Lyme Neuroborreliosis. Clin Infect Dis . 2021;72(10):1727-1729. doi:10.1093/cid/ciaa337 Baird GS, Nelson SK, Keeney TR, et al. Age-dependent changes in the cerebrospinal fluid proteome by slow off-rate modified aptamer array. Am J Pathol . 2012;180(2):446-456. doi:10.1016/j.ajpath.2011.10.024 Zhang J, Goodlett DR, Peskind ER, et al. Quantitative proteomic analysis of age-related changes in human cerebrospinal fluid. Neurobiol Aging . 2005;26(2):207-227. doi:10.1016/j.neurobiolaging.2004.03.012 Lehallier B, Gate D, Schaum N, et al. the lifespan. 2020;25(12):1843-1850. doi:10.1038/s41591-019-0673-2.Undulating Held F, Makarov C, Gasperi C, et al. Proteomics Reveals Age as Major Modifier of Inflammatory CSF Signatures in Multiple Sclerosis. Neurol Neuroimmunol neuroinflammation . 2025;12(1):e200322. doi:10.1212/NXI.0000000000200322 Wesenhagen KEJ, Gobom J, Bos I, et al. Effects of age, amyloid, sex, and APOE ε4 on the CSF proteome in normal cognition. Alzheimer’s Dement Diagnosis, Assess Dis Monit . 2022;14(1):1-12. doi:10.1002/dad2.12286 Laugesen K, Mengel-From J, Christensen K, et al. A Review of Major Danish Biobanks: Advantages and Possibilities of Health Research in Denmark. Clin Epidemiol . 2023;15(February):213-239. doi:10.2147/CLEP.S392416 Bader JM, Geyer PE, Müller JB, et al. Proteome profiling in cerebrospinal fluid reveals novel biomarkers of Alzheimer’s disease. Mol Syst Biol . 2020;16(6):9356. doi:10.15252/MSB.20199356 Geyer PE, Wewer Albrechtsen NJ, Tyanova S, et al. Proteomics reveals the effects of sustained weight loss on the human plasma proteome. Mol Syst Biol . Published online 2016. doi:10.15252/msb.20167357 Kulak NA, Pichler G, Paron I, Nagaraj N, Mann M. Minimal, encapsulated proteomic-sample processing applied to copy-number estimation in eukaryotic cells. Nat Methods . Published online 2014. doi:10.1038/nmeth.2834 Demichev V, Messner CB, Vernardis SI, Lilley KS, Ralser M. DIA-NN: neural networks and interference correction enable deep proteome coverage in high throughput. Nat Methods . 2020;17(1):41-44. doi:10.1038/s41592-019-0638-x Krismer E, Bludau I, Strauss MT, Mann M. AlphaPeptStats: An open-source Python package for automated and scalable statistical analysis of mass spectrometry-based proteomics. Bioinformatics . 2023;39(8):1-4. doi:10.1093/bioinformatics/btad461 Santos A, Colaço AR, Nielsen AB, et al. A knowledge graph to interpret clinical proteomics data. Nat Biotechnol . 2022;40(5):692-702. doi:10.1038/s41587-021-01145-6 Webel H, Niu L, Nielsen AB, et al. Imputation of label-free quantitative mass spectrometry-based proteomics data using self-supervised deep learning. Nat Commun . 2024;15(1):5405. doi:10.1038/s41467-024-48711-5 Behdenna A, Colange M, Haziza J, et al. pyComBat, a Python tool for batch effects correction in high-throughput molecular data using empirical Bayes methods. BMC Bioinformatics . 2023;24(1):1-9. doi:10.1186/s12859-023-05578-5 Ashburner M, Ball CA, Blake JA, et al. Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat Genet . 2000;25(1):25-29. doi:10.1038/75556 Lundberg SM, Lee SI. A unified approach to interpreting model predictions. In: Proceedings of the 31st International Conference on Neural Information Processing Systems . NIPS’17. Curran Associates Inc.; 2017:4768–4777. Perez-Riverol Y, Bai J, Bandla C, et al. The PRIDE database resources in 2022: A hub for mass spectrometry-based proteomics evidences. Nucleic Acids Res . 2022;50(D1):D543-D552. doi:10.1093/nar/gkab1038 MATERIALS AND METHODS Ethics Samples were obtained during diagnostic investigation and all patients had given their informed consent to the storage of biological material and its future use in research. The study was approved by Knowledge Center for Data Reviews (P-2019-707) and the local ethics committee (H-17024315). The biobanks were approved by the Danish Data Protection Agency (Rigshospitalet: j.nr.: 2012-41-0036). Study design This study was an observational retrospective cohort study. Study population and setting We identified a total of 483 CSF (n = 308) and plasma (n = 175) samples from adults (>18 years) with LNB, erythema migrans, acrodermatitis chronica atrophicans, post-treatment-Lyme-disease-syndrome, viral meningitis, and individuals who were investigated for suspected LNB, but had normal CSF (controls) from; i) The Danish National Biobank (samples collected from 2001-2011) 43 and ii) the Biobank of the Department of Infectious Diseases at Copenhagen University Hospital, Rigshospitalet (samples collected from 2016-2022). After informed patient consent for storage and future use in research was provided, CSF samples were labelled and stored at -80°C and blood samples in EDTA tubes were spun and the supernatant was transferred to Eppendorf tubes, labelled, and stored at -80°C. Samples from the two sites of origin were kept separate in the subsequent analyses to preserve independent cohorts for model development and validation. Cohorts Samples were divided in four cohorts for diagnostic model development and validation in CSF and plasma respectively. The CSF development cohort consisted of 145 CSF samples from patients diagnosed with LNB (n=49), viral meningitis (n=44) and controls (n=52). The CSF validation cohort included a total of 163 CSF samples from patients diagnosed with LNB (n=69), viral meningitis (n=45), and controls (n=49). The plasma development cohort included a total of 95 plasma samples from patients with LNB (n=27), viral meningitis (n=20) and controls (n=48) whereas the plasma validation cohort consisted of 80 plasma samples from patients with LNB (n=10), viral meningitis (n=18), controls (n=20) and an additional cohort of plasma samples from patients with manifestations of Lyme borreliosis without CNS involvement including post-treatment-Lyme-disease-syndrome (n=17), erythema migrans (n=9), and acrodermatitis chronica atrophicans (n=6). Definitions of diagnostic groups LNB: Patients diagnosed with LNB based on a first-time positive B. burgdorferi -specific intrathecal antibody test and the International Classification of Diseases 10 th revision (ICD-10) diagnosis code Borreliosis: A69. Viral Meningitis: Patients with a final clinical diagnosis of viral meningitis determined by an infectious disease specialist at a tertiary referral university hospital. The diagnosis was based on clinical presentation, CSF pleocytosis and exclusion of other diagnoses. In 76 cases a PCR-verified viral aetiology was determined (Herpes simplex virus 2 (n=32), Enterovirus (n=30), Varicella zoster (n=9), Influenza A virus (n=3), Epstein-Barr virus (n=1), Toscana virus (n=1)). Controls: Individuals who had a lumbar puncture performed because LNB was suspected, but where white-blood-cell (WBC) count in the CSF was within normal reference ranges (CSF-WBC: 0-5 x 10 6 /L) and the B. burgdorferi s.l. specific intrathecal antibody index was negative. Post-treatment - Lyme-disease-syndrome: Patients with persisting symptoms of Lyme borreliosis >6 months after diagnosis and treatment. Erythema migrans : Patients with a skin rash clinically diagnosed with erythema migrans at the Unit for Tick- Borne Infections, Copenhagen University Hospital, Rigshospitalet. Acrodermatitis chronica atrophicans : Patients clinically diagnosed with acrodermatitis chronica atrophicans skin rash and positive B. burgdorferi s.l.IgG in blood at the Unit for Tick Borne Infections, Copenhagen University Hospital, Rigshospitalet. Proteomic analysis CSF and plasma samples were thawed and 100 µl were transferred to 96-well plates for analysis. The sample preparation was optimized based on the previously published methods described in 44,45 . Briefly, 20 µl CSF was were first denatured with 30 µl PreOmics Lysis buffer, while 5 µl plasma was denatured with 45 µl PreOmics Lysis buffer 46 . Both sample types were subsequently digested using LysC/trypsin enzyme mix. The resulting peptides were purified using two-gauge SDB-RPS StageTips, and the eluate analysed on Evosep One (Evosep Biosystem, Denmark) liquid chromatography system, coupled online to an Orbitrap Exploris 480 mass spectrometer. Data acquisition was performed in data-independent analysis (DIA), using 60 samples per day (SPD) gradient and 8 cm Pepsep column. Data processing Initial data processing of the mass spectrometry raw files were performed with DIA-NN version 1.9 in a data independent search 47 . The DIA-NN data underwent further processing using the Clinical Knowledge Graph (CKG), alphapeptstats and Jupyter Notebook 48,49 . Initially, a stringent filter for missing data was applied: 1) samples with low protein count, defined by a value below 1.5IQR from the 25 th quantile of the combined distribution, were excluded, and 2) proteins with a missingness of more than 40% across samples were excluded. Data was log2 transformed. The remaining missing values were imputed with a variational autoencoder using the PIMMS software 50 . Assessment of sample quality was conducted as previously described 50 . Batch correction was executed using combat to overcome potential plate-specific bias on subsequent analyses 51 . Data analysis Proteins exhibiting significantly different levels between the cohorts were identified by unpaired t-tests. Multiple hypothesis correction was applied with the Benjamini-Hochberg method, with adjusted P-values < 0.05 deemed statistically significant. P-values and protein abundances were visualized in Volcano plots, with -log10(corrected p-value) and log2 of protein fold change between groups. Venn diagrams were used to visualize overlaps in significant proteins between statistical comparisons. Heatmaps in combination with Sankey plots of gene ontology terms were used to highlight protein changes of the significant proteins together with their biological processes 52 . In the heatmaps, protein abundances were z-scored. Gene ontology terms were retrieved for each protein through UNIPROT. Gene ontology terms with frequency of less than 5% across proteins were not visualized. Model development Z-scored data from the development cohorts were analysed using supervised machine learning to explore the potential for a diagnostic signature for LNB. Significant proteins identified by t-test analysis on the development cohorts were used as input features in the training data. Our data sets, comprising both samples from patients diagnosed with LNB and viral meningitis or control samples, was partitioned into a training set for model development and a test set for model validation using a 5-fold or 3-fold cross-validation (CV) approach for CSF and plasma respectively. The number of CV-folds was decided based on sample size with a minimum of 30 samples in each set. Both sets maintained an equal ratio of positive and negative cases (stratified k-fold cross-validation). The classification target used for analysis was the diagnosis of LNB and viral meningitis or LNB and control (yes/no). An appropriate ML model was determined by testing the performance of twelve different algorithms based on the area under the curve (AUC) and Matthews Correlation Coefficient (MCC) in the development test set cross-validations. For each algorithm, optimal hyperparameters and features were selected during cross-validation. Feature importance were highlighted for the top 10 most predictive features with SHAP (SHapley Additive exPlanations) values 53 . Model validation The models trained on the development cohorts were subsequently applied to the validation cohorts and the performance was assessed by AUC and MCC. ROC curves and confusion matrices were used to visualize the performance of the classifiers. Additional Declarations Yes there is potential Competing Interest. AML reports speakers’ honorarium/travel grants/advisory board activity and unrestricted grant from Gilead, speakers honorarium/travel grants from GSK, speaker’s honorarium/advisory board activity from Pfizer outside this work. NJWA has received funding, served on scientific advisory panels, and/or speakers bureaus for Boehringer Ingelheim, MSD/Merck, Novo Nordisk, EvoSep, ROCHE, Janssen, and Mercodia. AJH reports a research collaboration agreement with Pfizer unrelated to this work. M.M. is an indirect shareholder in Evosep Biosystems. None of the other authors report any conflict of interests. Supplementary Files SupplementaryMaterial.docx Mass-spectrometry-based proteomics enables rapid and accurate diagnosis of Lyme neuroborreliosis in adults Cite Share Download PDF Status: Published Journal Publication published 27 Oct, 2025 Read the published version in Nature Communications → Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6048306","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":420244485,"identity":"173b8ec6-6744-4ef4-b290-7bde3dab12fc","order_by":0,"name":"Nicolai Wewer Albrechtsen","email":"data:image/png;base64,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","orcid":"https://orcid.org/0000-0003-4230-5753","institution":"Copenhagen University Hospital - Bispebjerg and Frederiksberg","correspondingAuthor":true,"prefix":"","firstName":"Nicolai","middleName":"Wewer","lastName":"Albrechtsen","suffix":""},{"id":420244486,"identity":"15959211-12c8-4fb3-b1c4-a91f8c81922d","order_by":1,"name":"Annelaura Nielsen","email":"","orcid":"","institution":"Copenhagen University Hospital, Bispebjerg and Frederiksberg Hospital","correspondingAuthor":false,"prefix":"","firstName":"Annelaura","middleName":"","lastName":"Nielsen","suffix":""},{"id":420244487,"identity":"f088bad8-fc77-4fd2-a628-1ad923a5d17f","order_by":2,"name":"Lasse Fjordside","email":"","orcid":"","institution":"Copenhagen University Hospital, Rigshospitalet","correspondingAuthor":false,"prefix":"","firstName":"Lasse","middleName":"","lastName":"Fjordside","suffix":""},{"id":420244488,"identity":"65f587d1-1b46-4646-be5f-bf8538483926","order_by":3,"name":"Lylia Drici","email":"","orcid":"https://orcid.org/0000-0002-6633-1721","institution":"University of Copenhagen","correspondingAuthor":false,"prefix":"","firstName":"Lylia","middleName":"","lastName":"Drici","suffix":""},{"id":420244489,"identity":"656012fe-fea2-447d-89c2-3c47e62a3892","order_by":4,"name":"Maud Ottenheijm","email":"","orcid":"","institution":"Copenhagen University Hospital, Bispebjerg and Frederiksberg Hospital","correspondingAuthor":false,"prefix":"","firstName":"Maud","middleName":"","lastName":"Ottenheijm","suffix":""},{"id":420244490,"identity":"25f88702-1999-480d-8526-b979f5812eff","order_by":5,"name":"Christine Rasmussen","email":"","orcid":"","institution":"Copenhagen University Hospital, Bispebjerg and Frederiksberg Hospital","correspondingAuthor":false,"prefix":"","firstName":"Christine","middleName":"","lastName":"Rasmussen","suffix":""},{"id":420244491,"identity":"9a2c2f62-b2a8-444d-914a-6297696a7bf5","order_by":6,"name":"Anna Henningsson","email":"","orcid":"https://orcid.org/0000-0002-9315-8901","institution":"Linköping University","correspondingAuthor":false,"prefix":"","firstName":"Anna","middleName":"","lastName":"Henningsson","suffix":""},{"id":420244492,"identity":"f8e4a42d-57d0-48a4-9da7-18f66a294e49","order_by":7,"name":"Lene H. Harritshøj","email":"","orcid":"","institution":"Department of Clinical Immunology, Copenhagen University Hospital","correspondingAuthor":false,"prefix":"","firstName":"Lene","middleName":"H.","lastName":"Harritshøj","suffix":""},{"id":420244493,"identity":"7d5f2e0b-39b2-4c19-9191-759f733e5c7f","order_by":8,"name":"Matthias Mann","email":"","orcid":"https://orcid.org/0000-0003-1292-4799","institution":"Max Planck Institute of Biochemistry","correspondingAuthor":false,"prefix":"","firstName":"Matthias","middleName":"","lastName":"Mann","suffix":""},{"id":420244494,"identity":"ddc4bdc3-2d4c-42e8-a8ea-9cd3a4fbbc4c","order_by":9,"name":"Anne-Mette Lebech","email":"","orcid":"","institution":"Department of Infectious Diseases, Copenhagen University Hospital, Rigshospitalet","correspondingAuthor":false,"prefix":"","firstName":"Anne-Mette","middleName":"","lastName":"Lebech","suffix":""}],"badges":[],"createdAt":"2025-02-17 13:20:29","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6048306/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6048306/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41467-025-64903-z","type":"published","date":"2025-10-27T04:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":77401589,"identity":"fb40bbf2-61c3-44ef-a771-e4c64ade4247","added_by":"auto","created_at":"2025-02-28 08:40:02","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":392052,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA. \u003c/strong\u003e\u0026nbsp;Volcano plots highlighting protein differences between Lyme neuroborreliosis (LNB) vs. viral meningitis (VM) and LNB vs. Controls in the cerebrospinal fluid development cohort. Each point represents a protein, the colour purple represents upregulated proteins with a log2 fold change (FC) larger than 1, the colour blue represents downregulated proteins with a log2 fold smaller larger than -1 and the colour grey represents proteins with a log2 fold change between -1 and 1 or non-significant proteins. X-axis depicts log2 fold change and y-axis depicts the log 10 adjusted p-value. Horizontal dashed lines represent the significance threshold after multiple hypothesis correction at p=0.05. Vertical dashed lines represent large log2 fold changes: above 1 or below -1. \u003cstrong\u003eB\u003c/strong\u003e Overlap in significant proteins from the two volcano plots with log2 fold changes lower than -1 or larger than 1. \u003cstrong\u003eC\u003c/strong\u003e Heatmap of the significant proteins for each sample in the development cohort and the corresponding mean from the validation cohort. The Gene Ontology-term associated with the protein is depicted on the right.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6048306/v1/518cf9435a1fba97ffb5efc7.png"},{"id":77400641,"identity":"cf886dd4-17f7-4701-89a1-2e3066c82248","added_by":"auto","created_at":"2025-02-28 08:32:00","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":169135,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA\u003c/strong\u003e AUC for the final machine-learning model (Logistic Regression) developed to discriminate between LNB and viral meningitis (VM) cases in cerebrospinal fluid samples. Prediction performance is presented with Area Under ROC Curve (AUC) and Matthews Correlation Coefficient (MCC) for each cohort. Standard deviation (std.) from the 5-fold cross validation (CV) is shown as semi-transparent error intervals for the test set. \u003cstrong\u003eB\u003c/strong\u003e Feature importance of the LR model for the top 10 most predictive features depicted with SHAP values. \u003cstrong\u003eC\u003c/strong\u003e AUC for the final ML-model (Support Vector Classifier (SVC)) developed to discriminate between LNB and controls. Prediction performance is presented with AUC and MCC for each cohort. Standard deviation from the 5-fold cross validation is shown as semi-transparent error intervals for the test set. \u003cstrong\u003eD\u003c/strong\u003e Feature importance of the SVC model for the top 10 most predictive features depicted with SHAP values. Abbreviations: TPR; True Positive Rate, FPR; False Positive Rate.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6048306/v1/f4995e17551742bdf352edb6.png"},{"id":77400407,"identity":"e6dd4c37-8a99-4289-9ccc-2bdddccd506b","added_by":"auto","created_at":"2025-02-28 08:24:00","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":550340,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA\u003c/strong\u003e Volcano plots highlighting protein differences between Lyme neuroborreliosis (LNB) vs. viral meningitis (VM) and LNB vs. Controls in the development cohort in plasma samples. Each point represents a protein, the colour purple represents upregulated proteins with a log2 fold change larger than 1, the colour blue represents downregulated proteins with a log2 fold smaller larger than -1 and the colour grey represents proteins with a log2 fold change between -1 and 1 or non-significant proteins. X-axis depicts log2 fold change and y-axis depicts the log10 adjusted p-value. Horizontal dashed lines represent the significance threshold after multiple hypothesis correction at p=0.05. Vertical dashed lines represent large log2 fold changes: above 1 or below -1. \u003cstrong\u003eB\u003c/strong\u003e Overlap in significant proteins from the two volcano plots with log2 fold changes lower than -1 or larger than 1. \u003cstrong\u003eC\u003c/strong\u003e Heatmap of the significant proteins for each plasma sample in the development cohort and the corresponding mean from the validation cohort. The Gene Ontology-term associated with the protein is depicted on the right.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6048306/v1/13734fa9056a3f5bf6731a71.png"},{"id":77400364,"identity":"8de6c95c-3a0f-43bd-b433-d3bb37d90e85","added_by":"auto","created_at":"2025-02-28 08:23:59","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":110531,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA\u003c/strong\u003e AUROC for the final machine-learning (ML) model, Support Vector Classifier (SVC), developed to discriminate between Lyme neuroborreliosis (LNB) and controls in plasma samples. Prediction performance is presented with Area Under ROC Curve (AUC) and Matthews Correlation Coefficient (MCC) for each cohort. Standard deviation (Std.) from the 3-fold cross validation (CV) is shown as semi-transparent error intervals for the test set. \u003cstrong\u003eB \u003c/strong\u003eFeature importance of the SVC model for the top 10 most predictive proteins depicted with SHAP values. \u003cstrong\u003eC\u003c/strong\u003e Distributions of predicted probabilities from the LNB classification model on other diagnostic groups. Abbreviations: TPR; True Positive Rate, FPR; False Positive Rate, ACA; Acrodermatitis chronica atrophicans, EM; Erythema migrans, PTLDS; post-treatment-Lyme-disease-syndrome, VM; Viral meningitis.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6048306/v1/c3471dc52586e70bacef62c5.png"},{"id":77400417,"identity":"23aa9491-b4c7-4977-85ed-0d12d86248a7","added_by":"auto","created_at":"2025-02-28 08:24:00","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":295841,"visible":true,"origin":"","legend":"\u003cp\u003eComparison of cerebrospinal fluid (CSF) and plasma proteins. \u003cstrong\u003eA\u003c/strong\u003eVenn diagram highlighting overlap between significant proteins in the comparisons in the different tissues. \u003cstrong\u003eB\u003c/strong\u003e Heatmap of the log2 fold changes between LNB and VM or controls of the 72 overlapping proteins. Empty fields represent non-significant findings for that comparison. Gene ontology terms for the proteins are summarized in the Sankey plot on the right.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-6048306/v1/89cc3a5b2f86d7cad89b2c00.png"},{"id":94583230,"identity":"658db336-327c-45a5-9121-3cce1e2516a9","added_by":"auto","created_at":"2025-10-28 18:13:55","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2307075,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6048306/v1/818f7c14-b586-4586-b5d9-8f47c4476696.pdf"},{"id":77400456,"identity":"08c2c5fb-4a3d-403b-9a3b-8e9c16a00904","added_by":"auto","created_at":"2025-02-28 08:24:02","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":3562002,"visible":true,"origin":"","legend":"Mass-spectrometry-based proteomics enables rapid and accurate diagnosis of Lyme neuroborreliosis in adults","description":"","filename":"SupplementaryMaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-6048306/v1/22f12199d7786c0b4860e11f.docx"}],"financialInterests":"\u003cb\u003eYes\u003c/b\u003e there is potential Competing Interest.\nAML reports speakers’ honorarium/travel grants/advisory board activity and unrestricted grant from Gilead, speakers honorarium/travel grants from GSK, speaker’s honorarium/advisory board activity from Pfizer outside this work. NJWA has received funding, served on scientific advisory panels, and/or speakers bureaus for Boehringer Ingelheim, MSD/Merck, Novo Nordisk, EvoSep, ROCHE, Janssen, and Mercodia. AJH reports a research collaboration agreement with Pfizer unrelated to this work. M.M. is an indirect shareholder in Evosep Biosystems. None of the other authors report any conflict of interests.","formattedTitle":"Mass-spectrometry-based proteomics enables rapid and accurate diagnosis of Lyme neuroborreliosis in adults","fulltext":[{"header":"INTRODUCTION","content":"\u003cp\u003eLyme neuroborreliosis (LNB) is a bacterial infection of the nervous system caused by spirochetes of the \u003cem\u003eBorrelia burgdorferi\u003c/em\u003e sensu lato (\u003cem\u003eB. burgdorferi\u003c/em\u003e s.l.) complex transmitted through bites from hard-shelled ticks of the \u003cem\u003eIxodes\u003c/em\u003e genus\u003csup\u003e1\u003c/sup\u003e. With an approximate incidence in endemic countries between 3.2 and 6.3 per 100 000 persons/year\u003csup\u003e2\u0026ndash;4\u003c/sup\u003e LNB is among the most frequent bacterial infections of the nervous system in Europe\u003csup\u003e5,6\u003c/sup\u003e. LNB can cause a wide range of clinical neurological conditions, but most frequently presents as a subacute painful meningo-radiculitis with radiating pain from the spine to neck, extremities, thorax or abdomen, lymphocytic meningitis and/or cranial neuropathies e.g., facial nerve palsy\u003csup\u003e7\u003c/sup\u003e. If relevant antibiotic therapy is administered at an early stage of disease, LNB has a favourable long-term prognosis\u003csup\u003e8\u0026ndash;10\u003c/sup\u003e. However, delayed treatment is associated with an increased risk of residual symptoms and long-term sequelae\u003csup\u003e9\u003c/sup\u003e. In countries where LNB is endemic, the average time from onset of neurological symptoms to diagnosis is typically around 3 weeks and has remained unchanged for the last four decades\u003csup\u003e6\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe cause of the diagnostic delay is multifactorial. Overlapping symptomatology with other more common diseases and the fact that only around 40% of patients with LNB report a tick bite and only 25% report a history of the classic skin rash erythema migrans makes it less likely that physicians and patients consider LNB as a differential diagnosis, especially in the absence of facial nerve palsy\u003csup\u003e6,8\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eHowever, even when LNB is clinically suspected at an early stage, the diagnosis requires detection of \u003cem\u003eB. burgdorferi\u003c/em\u003e s.l.- specific antibodies in the cerebrospinal fluid (CSF). Lumbar puncture is an uncomfortable and expensive procedure that may require hospitalization and generel anesthesia for children. Direct pathogen identification of \u003cem\u003eB. burgdorferi\u003c/em\u003e s.l. with polymerase chain reaction (PCR), cultivation or a combination has retained an exceedingly poor level of sensitivity both in blood and CSF despite several attempts to improve tools and techniques over the years\u003csup\u003e11\u0026ndash;14\u003c/sup\u003e. Further, the serological response in LNB is not detectable early in the course of disease and antibodies may remain elevated for months to several years after full recovery of well-treated infections making it impossible to discriminate between past and current infection. Furthermore, the \u003cem\u003eB. burgdorferi\u003c/em\u003e s.l.-specific IgG and IgM in blood, poorly predict nervous system involvement\u003csup\u003e15\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThus, in order to reduce the diagnostic delay in LNB, the need for novel diagnostic tools is evident. Especially improved blood-based diagnostic tests would be extremely valuable as they would both potentially help reduce the diagnostic delay and offer a less invasive diagnostic tool.\u003c/p\u003e \u003cp\u003eUntargeted proteomic analysis enables accurate measurement of proteins in a sample and can be used to compare patterns of protein regulation in patients with and without disease to identify disease-specific protein response signatures\u003csup\u003e16\u003c/sup\u003e. When combined with machine-learning (ML) for data-evaluation and analysis, proteomic test results can be made available with turnaround times of just 2 hours making it highly attractive in real-life clinical settings.\u003c/p\u003e \u003cp\u003eThe aim of this study was to explore the potential of ML-assisted MS-based proteomics as a novel diagnostic tool for LNB in CSF and plasma of adults.\u003c/p\u003e"},{"header":"RESULTS","content":"\u003cp\u003eStudy population\u003c/p\u003e\n\u003cp\u003eA total of 483 CSF and plasma samples from adult individuals were eligible for proteomics analysis: \u0026nbsp;155 samples from patients with LNB, 127 samples from patients with viral meningitis, 169 controls and 32 samples from patients with erythema migrans, acrodermatitis chronica atrophicans\u003cem\u003e\u0026nbsp;\u003c/em\u003eand post-treatment-Lyme-disease-syndrome (Table 1).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e1\u003c/strong\u003e Baseline characteristics of patients with Lyme neuroborreliosis, viral meningitis, controls, and other Lyme borreliosis manifestations.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"588\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22.449%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCohort\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35.3741%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDiagnosis\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 4.93197%;\"\u003e\n \u003cp\u003e\u003cstrong\u003en\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.8367%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAge\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eMean [IQR]\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.4082%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eFemales\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003en (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22.449%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCSF Development\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35.3741%;\"\u003e\n \u003cp\u003eLyme neuroborreliosis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 4.93197%;\"\u003e\n \u003cp\u003e49\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.8367%;\"\u003e\n \u003cp\u003e51[40-66]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.4082%;\"\u003e\n \u003cp\u003e26 (53)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22.449%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eN = 145\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35.3741%;\"\u003e\n \u003cp\u003eViral meningitis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 4.93197%;\"\u003e\n \u003cp\u003e44\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.8367%;\"\u003e\n \u003cp\u003e38[27-42]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.4082%;\"\u003e\n \u003cp\u003e21 (47)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22.449%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35.3741%;\"\u003e\n \u003cp\u003eControl\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 4.93197%;\"\u003e\n \u003cp\u003e52\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.8367%;\"\u003e\n \u003cp\u003e36[25-42]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.4082%;\"\u003e\n \u003cp\u003e17 (32)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22.449%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCSF Validation\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35.3741%;\"\u003e\n \u003cp\u003eLyme neuroborreliosis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 4.93197%;\"\u003e\n \u003cp\u003e69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.8367%;\"\u003e\n \u003cp\u003e53[46-67]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.4082%;\"\u003e\n \u003cp\u003e28 (40)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22.449%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eN = 163\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35.3741%;\"\u003e\n \u003cp\u003eViral meningitis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 4.93197%;\"\u003e\n \u003cp\u003e45\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.8367%;\"\u003e\n \u003cp\u003e31[25-37]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.4082%;\"\u003e\n \u003cp\u003e23 (51)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22.449%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35.3741%;\"\u003e\n \u003cp\u003eControl\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 4.93197%;\"\u003e\n \u003cp\u003e49\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.8367%;\"\u003e\n \u003cp\u003e44[33-53]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.4082%;\"\u003e\n \u003cp\u003e24 (48)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22.449%;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePlasma Development\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35.3741%;\"\u003e\n \u003cp\u003eLyme neuroborreliosis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 4.93197%;\"\u003e\n \u003cp\u003e27\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.8367%;\"\u003e\n \u003cp\u003e64[61-69]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.4082%;\"\u003e\n \u003cp\u003e10 (37)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22.449%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eN = 95\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35.3741%;\"\u003e\n \u003cp\u003eViral meningitis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 4.93197%;\"\u003e\n \u003cp\u003e20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.8367%;\"\u003e\n \u003cp\u003e35[27-38]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.4082%;\"\u003e\n \u003cp\u003e13 (65)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22.449%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35.3741%;\"\u003e\n \u003cp\u003eControl\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 4.93197%;\"\u003e\n \u003cp\u003e48\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.8367%;\"\u003e\n \u003cp\u003e46[33-60]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.4082%;\"\u003e\n \u003cp\u003e28 (58)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22.449%;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePlasma Validation\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35.3741%;\"\u003e\n \u003cp\u003eLyme neuroborreliosis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 4.93197%;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.8367%;\"\u003e\n \u003cp\u003e59[56-63]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.4082%;\"\u003e\n \u003cp\u003e4 (40)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22.449%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eN = 80\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35.3741%;\"\u003e\n \u003cp\u003eViral meningitis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 4.93197%;\"\u003e\n \u003cp\u003e18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.8367%;\"\u003e\n \u003cp\u003e38[24-43]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.4082%;\"\u003e\n \u003cp\u003e10 (55)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22.449%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35.3741%;\"\u003e\n \u003cp\u003eControl\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 4.93197%;\"\u003e\n \u003cp\u003e20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.8367%;\"\u003e\n \u003cp\u003e36[28-42]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.4082%;\"\u003e\n \u003cp\u003e6 (30)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22.449%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35.3741%;\"\u003e\n \u003cp\u003eAcrodermatitis chronica atrophicans\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 4.93197%;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.8367%;\"\u003e\n \u003cp\u003e66[59-74]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.4082%;\"\u003e\n \u003cp\u003e6 (100)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22.449%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35.3741%;\"\u003e\n \u003cp\u003eErythema migrans\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 4.93197%;\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.8367%;\"\u003e\n \u003cp\u003e47[36-54]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.4082%;\"\u003e\n \u003cp\u003e5 (55)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 22.449%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 35.3741%;\"\u003e\n \u003cp\u003ePost-treatment-Lyme-disease-syndrome\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 4.93197%;\"\u003e\n \u003cp\u003e17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.8367%;\"\u003e\n \u003cp\u003e58[48-64]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.4082%;\"\u003e\n \u003cp\u003e10 (58)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eAbbreviations: CSF = cerebrospinal fluid, IQR = Inner Quartile Range\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCSF proteomics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eCSF development cohort\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eA total of 1,865 proteins were identified in CSF, with a median of 773 quantified pr. sample (Supplementary Figure 1). Fourteen samples were excluded due to low protein numbers. After filtering for \u0026shy;missingness, a total of 654 proteins were included for further analysis.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eA total of 176 proteins were significantly different between LNB and viral meningitis, of these 10 had an absolute log2 fold change larger than 1 (Figure 1A, Supplementary Table1). A total of 464 proteins were significantly different between LNB and controls, of these 41 had an absolute log2 fold change larger than 1 (Figure 1B, Supplementary Table1).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOf the significant proteins with high log2 fold changes in the two comparisons, 7 proteins overlapped (Figure 1C).\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eCSF validation cohort\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eIn the validation cohort a total of 147 proteins were significantly different in LNB compared to viral meningitis and controls with high log2 fold changes, where 43/46 proteins from the development cohort overlapped (Figure 1B and Supplementary Figure 2,3).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eProtein signatures of LNB in CSF\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eProtein signatures were dominated by immunoglobulins (e.g., IGLV3-25, IGLV2-18, FCGBP, IGHM), proteins involved in innate immune responses (e.g. enolase 1, S100A9), neuroendocrine signalling (e.g. ECRG4, CHGA), cell-migration and cell-damage (e.g. PFN1, APCS, YWHAZ, H4C1, ACTA2) (Figure 1C). Immunoglobulins were generally upregulated, while proteins involved in innate immune responses and neuroendocrine signalling were relatively downregulated in LNB compared to viral meningitis and controls.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDiagnostic classifier for LNB based on machine learning and CSF proteomics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe protein profiles of the CSF development cohort were used to develop a ML model to evaluate the potential of MS-based proteomics coupled with ML as diagnostic support in LNB. Of the twelve ML models tested, the Logistic Regression (LR) model obtained the highest performance in the classification of viral meningitis vs. LNB, whereas the Support Vector Classifier (SVC) model produced the best result in the control vs. LNB classification (Supplementary Figure 4 and 5 respectively). The performance of the diagnostic classification model of viral meningitis vs. LNB on the test set had an area under the curve (AUC) of 0.91 (std. = 0.11) and a Matthews Correlation Coefficient (MCC) of 0.81 (std.=0.08) (Figure 2A). When applied on the validation cohort, the model obtained an AUC and MCC of 0.92 (std. = 0.02) and 0.7 (std.=0.06). According to their SHAP (SHapley Additive exPlanations)\u0026nbsp;values, the most important proteins in the discrimination between viral meningitis and LNB are visualized in Figure 2B. The performance of the diagnostic classification model of control vs. LNB on the test set had an AUC of 0.93 (std. = 0.07) and a MCC of 0.76 (std.=0.14) (Figure 2C). When applied on the validation cohort, the model obtained an AUC and MCC of 0.9 (std. = 0.01) and 0.63 (std.=0.06). The most important proteins in the model according to their SHAP values (Figure 2D) were similarly related to innate and humoral immune responses, neuroendocrine signalling, and cell-damage\u003cem\u003e.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePlasma proteomics\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003ePlasma development cohort\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eA total of 379 proteins were identified in plasma, with a median of 268 proteins quantified per sample (Supplementary Figure 6). One sample was excluded due to low protein number. After filtering for missingness, a total of 232 proteins were included for further analysis.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eA total of 61 proteins were significantly different between LNB and viral meningitis (Figure 3A, Supplementary Table2) and 63 for LNB versus controls (Figure 3B). The overlap between these two comparisons of significant proteins were 36 (Figure 3C, Supplementary Table2).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003ePlasma validation cohort\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eIn the validation cohort a total of 25 proteins were significantly different in LNB compared to viral meningitis and controls. Ten overlapped with the significant proteins from the development cohort (Figure 3B and Supplementary Figure 7,8).\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eProtein signatures of LNB in plasma\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe differences in protein profiles between the three patient groups are dominated by proteins involved in innate and humoral immune responses (Figure 3C). \u0026nbsp;When comparing protein profiles of plasma samples from LNB and viral meningitis, patients with LNB generally seem to have a relative upregulation of proteins associated with innate immunity and complement activation (e.g., FCN3, SERPING1, SERPINA5), lipid metabolism (e.g., APOE, APOC1, APOM), and coagulation regulation (e.g., F13A1, PROC). \u0026nbsp;In contrast, patients with viral meningitis have a plasma protein profile dominated by acute-phase and pro-inflammatory markers (e.g., CRP, S100A8, S100A9) and immunoglobulin production. In the comparison between LNB and controls, upregulated plasma-proteins in LNB are similarly involved in complement activation (e.g., C3, C5), coagulation and inflammation regulation (e.g., SERPINA3, SERPIND1), and immune system modulators (e.g., HP, ITIH4). In contrast to CSF, most immunoglobulins had a higher protein level in the plasma of control individuals compared to LNB patients.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDiagnostic classifier for LNB based on machine learning and plasma proteomics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSimilar to the CSF analysis, plasma proteome data were subjected to ML algorithm- and hyperparameter selection in a 3-fold cross validation scheme. It was not deemed feasible to develop a model for the discrimination of viral meningitis and LNB due to the low number of viral meningitis cases, hence, the only model tested was the discrimination of controls and LNB. Of the twelve models tested, the SVC model obtained the best performance in the test set (Supplementary Figure 9) with an AUC of 0.96 (std.=0.03) and a MCC of 0.83 (std.=0.008) (Figure 4A). When applied on the validation cohort the model obtained an AUC of 0.80 (std.=0.02) and a MCC of 0.48 (std.=0.04) (Figure 4A). According to the SHAP feature importance values, the most predictive proteins for the diagnostic classification of controls vs. LNB (Figure 4B) were associated with innate immunity, humoral immune defence, coagulation and cellular metabolism. When the LNB classification model was applied to other diagnostic groups, it identified groups with active \u003cem\u003eBorrelia burgdorferi\u003c/em\u003e s.l. infections; acrodermatitis chronica atrophicans and erythema migrans, as having significantly higher similarities to the LNB profile than groups with absence of active infection; Post-treatment-Lyme-disease-syndrome or unrelated aetiology; viral meningitis. Mann-Whitney U Test (acrodermatitis chronica atrophicans+erythema migrans vs Post-treatment-Lyme-disease-syndrome +Viral meningitis): U statistic = 5475.0, p-value = 2.65e-07. (Figure 4c).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eComparison of protein profiles in CSF and plasma\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe found an overlap of 72 proteins that were significantly different between LNB and controls or viral meningitis in both CSF and plasma (Figure 5A, Supplementary Table 3). The majority of overlapping proteins (62/72) were relatively upregulated in CSF from patients with LNB compared to controls and viral meningitis, and 30 of these were inversely regulated in plasma. This group primarily included immunoglobulins (18/30) (Figure 5B).\u003c/p\u003e"},{"header":"DISCUSSION","content":"\u003cp\u003eOur study shows that MS-based proteomics can accurately differentiate patients with LNB from patients with viral meningitis and non-LNB controls. Notably, we observed a high level of diagnostic accuracy in plasma, highlighting the potential of blood-based proteomics as a clinically valuable tool. Given the current need for novel diagnostic approaches to reduce delays and reliance on lumbar punctures, these findings could have significant implications for both initial diagnosis and monitoring of treatment response in LNB. \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo our knowledge, this is the first study reporting the use of untargeted proteomics as a diagnostic tool in a large and well-defined cohort of adult LNB patients, investigating both CSF and plasma samples. Additionally, the inclusion of viral meningitis cases strengthens our findings by assessing the ability of proteomics to distinguish between two conditions with overlapping symptomatology and CSF findings. This approach enhances the translational potential of our results.\u003c/p\u003e\n\u003cp\u003ePrevious studies investigating the diagnostic potential of proteomics in LNB have been limited in scope and methodology. A small US-based study assessing the diagnostic utility of proteomics in Lyme disease including a small sub-group with CNS involvement; \u003cem\u003eAngel et al.\u003c/em\u003e (2013) found a panel of 13 proteins discriminated between cases and controls reaching an AUC of 0.8\u003csup\u003e17\u003c/sup\u003e. However, the focus on other Lyme disease manifestations than LNB likely reflects the, somewhat debated, geographical difference in prevalence of neurological involvement with a \u003cem\u003eB. burgdorferi\u0026nbsp;\u003c/em\u003es.l.strain-dependent lower incidence in North America compared to Europe\u003csup\u003e18,19\u003c/sup\u003e. \u0026nbsp;More recent European studies have investigated the diagnostic use of protein panels in LNB, though with different methodologies and sample types: \u003cem\u003eFredriksson et al.\u003c/em\u003e (Sweden, 2024) analysed serum from 119 paediatric patients with LNB (n=61) and non-LNB (n=58) using a multiplex proximity extension assay that included a panel of 92 predefined inflammatory markers and found that a 5-protein-panel identified LNB with an AUC of 0.88\u003csup\u003e20\u003c/sup\u003e. Though the diagnostic groups and the use of blood-samples were comparable to our study, it is unclear to what extent results in children can be extrapolated to adults and the multiplex assay used is not comparable to the untargeted MS-based proteomics used in our study.\u003cem\u003e\u0026nbsp;Gęgotek et al.\u003c/em\u003e (Poland, 2024) performed LC-MS based proteomics on serum from patients with LNB (n=10) and controls (n=27) and found significant differences, but did not report measures of diagnostic accuracy\u003csup\u003e21\u003c/sup\u003e. Other investigations have primarily focused on characterizing patients with post-treatment-Lyme-disease-syndrome rather than acute LNB\u003csup\u003e22,23\u003c/sup\u003e. Despite differences in study design, a consistent pattern of immune activation has emerged across studies, emphasizing the role of innate immune responses, complement system and humoral immune defence in the human host response to \u003cem\u003eB. burgdorferi\u0026nbsp;\u003c/em\u003es.l. infection.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOur proteomic findings in CSF, indicate a distinct immune response characterized by elevated levels of immunoglobulin chains, complement-related proteins, and proteins involved in immune cell migration and cytoskeletal dynamics. These findings suggest a highly targeted immune response against \u003cem\u003eB. burgdorferi\u003c/em\u003e s.l. antigens, involving adaptive immunity with significant antibody production and complement activation. This aligns with established knowledge of host immune response during infections with \u003cem\u003eB. burgdorferi\u0026nbsp;\u003c/em\u003es.l.\u003csup\u003e24\u003c/sup\u003e. Interestingly, the antibody response exhibited opposite regulation in CSF and plasma, likely illustrating the significant compartmentalization of the humoral immune response to the CNS in LNB and reflects the fact that all patients with LNB in this study had a positive \u003cem\u003eB. burgdorferi\u0026nbsp;\u003c/em\u003es.l.- specific intrathecal antibody index. \u0026nbsp;Moreover, we observed overlapping regulation of key immune-related proteins in both plasma and CSF, suggesting a coordinated immune response across compartments. Plasma and CSF both showed changes in immunoglobulin chains, complement factors, and innate immune modulators in LNB compared to controls, indicating a systemic and localized immune activation. The CSF findings highlight immunoglobulin diversity and a direct antibody response, reflecting the local immune reaction in the CNS. Meanwhile, plasma proteomics revealed a strong systemic immune engagement, with upregulation of innate immune components such as ficolins and coagulation proteins, underscoring systemic inflammation. Notably, viral meningitis exhibited a different pattern, with both plasma and CSF showing elevated acute-phase reactants (e.g., CRP, S100A9) and immunoglobulins, consistent with a strong inflammatory and humoral response. Opposing regulation of certain proteins in CSF and plasma in LNB likely reflects a compartmentalized immune strategy, where the CNS immune response is tightly regulated, while plasma changes mirror systemic inflammation. Additionally, factors such as blood-brain barrier permeability, local versus systemic protein production, and clearance mechanisms may further influence these patterns.\u003c/p\u003e\n\u003cp\u003eAdditionally, structural proteins involved in immune cell motility were upregulated in CSF, indicating active immune cell recruitment within the CNS, specifically in LNB. Viral meningitis patients present a pattern more focused on inflammatory markers and metabolic activity, consistent with an acute viral immune response.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eA small cluster of cytoskeleton and extracellular matrix proteins (VIM, ACTA2, PFN1, FN1 and ACTB) emerged among the top discriminatory markers, suggesting increased cell-turnover in LNB. While expected in LNB vs. controls, this is more surprising when compared to viral meningitis. Notably, \u003cem\u003eB. burgdorferi\u003c/em\u003e s.l. spirochetes are hypothesized to extravasate early upon transmission and travel through the extracellular matrix to peripheral nerves eventually reaching the borders of the CNS and creating the inflammatory basis of radiculoneuritis\u003csup\u003e18,25\u003c/sup\u003e. Vimentin (VIM), a class III intermediate filament that regulates myelination in axons of peripheral nerves\u003csup\u003e26\u003c/sup\u003e, \u0026nbsp;is particularly relevant given our limited understanding of CNS entry by \u003cem\u003eB. burgdorferi\u0026nbsp;\u003c/em\u003es.l. spirochetes, the frequent cranial nerve involvement in LNB with painful radiculoneuritis\u003csup\u003e9,27\u0026ndash;29\u003c/sup\u003e and because proteins released from damaged nerves to the peripheral circulation could be potential biomarker candidates for early LNB, as previously shown with neurofilament light chain (NfL)\u003csup\u003e30\u003c/sup\u003e. Additionally, fibronectin (FN1) which is also found in this cluster, is considered essential for \u003cem\u003eB. burgdorferi\u003c/em\u003e s.l. cell adhesion and migration through binding to \u003cem\u003eB. burgdorferi\u0026nbsp;\u003c/em\u003es.l. surface proteins BBK32\u003csup\u003e31\u003c/sup\u003e and RevA\u003csup\u003e32\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eThe clinical implications of our findings are considerable. If validated in independent cohorts, CSF proteomics could aid in diagnosing LNB among adult patients with CNS infections with unknown aetiology. This may be particularly relevant for PCR-negative suspected viral meningitis cases, where CSF proteomics suggestive of LNB could prompt further testing or the direct initiation of relevant antibacterial treatment e.g. doxycycline. However, the most significant clinical implication would be the ability to diagnose LNB using a blood sample alone - a true gamechanger in LNB diagnostics. It would spare patients the discomfort and risk for complications of lumbar puncture and reduce healthcare-costs associated with referral to specialized healthcare facilities. \u0026nbsp;A diagnostic blood test for LNB would likely reduce the diagnostic delay and thereby reduce the risk of residual symptoms. Even a modest improvement over current serology-based approaches, could lead to fewer invasive procedures and significantly ease treatment-effect monitoring. This would be particularly beneficial in children, who account for around 30% of all LNB patients, \u0026nbsp; and often require general anaesthesia to have a lumbar puncture performed\u003csup\u003e33\u003c/sup\u003e \u003csup\u003e34,35\u003c/sup\u003e. Moreover, a blood-based test could address the unmet need for a less invasive method to monitor treatment response in patients with persistent symptoms after LNB treatment.\u003c/p\u003e\n\u003cp\u003eWhile these findings demonstrate the diagnostic potential of proteomics in LNB, it is important to consider the strengths and limitations of our study. With 483 samples from patients with LNB, viral meningitis and non-LNB controls, our study represents the largest investigation to date. While we could not stratify patients in full accordance with the European Federation of Neurological Societies (EFNS) diagnostic criteria\u003csup\u003e7\u003c/sup\u003e due to limited clinical data, our use of uniform case definition and the specificity of a first time positive \u003cem\u003eB. burgdorferi\u0026nbsp;\u003c/em\u003esl. intrathecal antibody index minimizes the risk of false-positive LNB diagnoses\u003csup\u003e36\u003c/sup\u003e. \u0026nbsp;Our selection of clinically relevant comparison groups, including viral meningitis and patients who were clinically suspected for LNB, enhances the real-world applicability of our findings by testing the robustness of proteomics in distinguishing LNB from other CNS infections. A key strength of our study is its focus on diagnostic feasibility, employing a proteomic workflow with a short turnaround time suitable for clinical implementation. However, the limited detection range inherent to this approach means that certain low-abundance but highly disease-specific proteins, such as CXCL13, may not be captured. Given that previous proteomic studies in LNB have also failed to detect CXCL13, this highlights a general challenge in proteomic biomarker discovery rather than a limitation of our study specifically\u003csup\u003e37\u003c/sup\u003e. As with all retrospective analyses, our study is constrained by sample availability and cohort composition. Differences in cohort sizes necessitated varying cross-validation strategies, and some observed differences in protein signatures may be influenced by factors such as disease stage, age, and sex\u003csup\u003e38\u0026ndash;42\u003c/sup\u003e. However, our study population reflects real-world clinical variation, as patients were primarily recruited from referral hospitals. Furthermore, prioritizing clinically relevant diagnostic groups over strictly age- and sex-matched controls enhances the translational value of our findings.\u003c/p\u003e\n\u003cp\u003eIn conclusion, our findings demonstrate that MS-based proteomics can accurately distinguish LNB from controls in both CSF and plasma. These results warrant further prospective validation, particularly with longitudinal sampling, to assess the diagnostic utility of plasma proteomics and its potential role in monitoring treatment response in LNB patients.\u003cbr\u003e\u0026nbsp;\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eACKNOWEDGEMENTS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe express our gratitude to Christine Rasmussen from the Department of Clinical Biochemistry at Copenhagen University Hospital, Bispebjerg, for her dedicated efforts in planning, preparing, and conducting the proteomic analysis. Additionally, we extend our appreciation to the Clinical Proteomic Group at the NNF Center for Protein Research, University of Copenhagen, for sharing their expertise on Mass Spectrometry.\u0026nbsp;We would also like to thank lab technicians Stine Østergaard and Dorthe Hass from the clinical research unit at the Department of Infectious Diseases, Rigshospitalet, for their great help with handling and sending samples for analysis.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of interest disclosures\u003c/strong\u003e. AML reports speakers’ honorarium/travel grants/advisory board activity and unrestricted grant from Gilead, speakers honorarium/travel grants from GSK, speaker’s honorarium/advisory board activity from Pfizer outside this work. NJWA has received funding, served on scientific advisory panels, and/or speakers bureaus for Boehringer Ingelheim, MSD/Merck, Novo Nordisk, EvoSep, ROCHE, Janssen, and Mercodia. AJH reports a research collaboration agreement with Pfizer unrelated to this work. M.M. is an indirect shareholder in Evosep Biosystems. None of the other authors report any conflict of interests.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNJWA is supported by European Foundation for the Study of Diabetes Future Leader Award (NNF21SA0072746), Independent Research Fund Denmark, Sapere Aude (1052-00003B) and Novo Nordic Foundation (NNF23OC0084970, NNF19OC0055001 and NNF24OC0088402). Novo Nordisk Foundation Center for Protein Research is supported financially by the Novo Nordisk Foundation (Grant agreement NNF14CC0001).\u003c/p\u003e\n\u003cp\u003eLF was supported by a research grant from the Research Fund of Copenhagen University Hospital - Rigshospitalet.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAML was supported by a research grant from the Lundbeck foundation, Research Fund of Copenhagen University Hospital - Rigshospitalet, Independent Research Fund Denmark and Svend Andersen’s foundation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRole of the funder\u003cbr\u003e\u003c/strong\u003eThe Novo Nordisk Foundation, the Research Fund of Rigshospitalet, and the Lundbeck Foundation had no role in the design and conduct of the study; collection, management, analyses, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData and code availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe raw data and processed data have been deposited to the ProteomeXchange Consortium via the Proteomics Identifications Database (PRIDE) partner repository, the dataset identifier will be added upon publication\u003csup\u003e54\u003c/sup\u003e. The Jupyter notebooks are available at github upon request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eHansen K, Crone C, Kristoferitsch W. \u003cem\u003eLyme Neuroborreliosis\u003c/em\u003e. Vol 115. 1st ed. Elsevier B.V.; 2013. doi:10.1016/B978-0-444-52902-2.00032-1\u003c/li\u003e\n \u003cli\u003eDahl V, Wisell KT, Giske CG, Tegnell A, Wallensten A. Lyme neuroborreliosis epidemiology in Sweden 2010 to 2014: Clinical microbiology laboratories are a better data source than the hospital discharge diagnosis register. \u003cem\u003eEurosurveillance\u003c/em\u003e. 2019;24(20):1800453. doi:10.2807/1560-7917.ES.2019.24.20.1800453/CITE/PLAINTEXT\u003c/li\u003e\n \u003cli\u003eDessau RB, Espenhain L, M\u0026oslash;lbak K, Krause TG, Voldstedlund M. Improving national surveillance of Lyme neuroborreliosis in Denmark through electronic reporting of specific antibody index testing from 2010 to 2012. \u003cem\u003eEurosurveillance\u003c/em\u003e. 2015;20(28):1-11. doi:10.2807/1560-7917.ES2015.20.28.21184\u003c/li\u003e\n \u003cli\u003eTetens MM, Haahr R, Dessau RB, et al. Changes in Lyme neuroborreliosis incidence in Denmark, 1996 to 2015. \u003cem\u003eTicks Tick Borne Dis\u003c/em\u003e. 2020;11(6). doi:10.1016/j.ttbdis.2020.101549\u003c/li\u003e\n \u003cli\u003eKoelman DLH, Van Kassel MN, Bijlsma MW, Brouwer MC, Van De Beek D, Van Der Ende A. Changing Epidemiology of Bacterial Meningitis Since Introduction of Conjugate Vaccines: 3 Decades of National Meningitis Surveillance in The Netherlands. \u003cem\u003eClin Infect Dis An Off Publ Infect Dis Soc Am\u003c/em\u003e. 2021;73(5):e1099. doi:10.1093/CID/CIAA1774\u003c/li\u003e\n \u003cli\u003eNordberg CL, Bodilsen J, Knudtzen FC, et al. Lyme neuroborreliosis in adults: A nationwide prospective cohort study. \u003cem\u003eTicks Tick Borne Dis\u003c/em\u003e. 2020;11(4):101411. doi:10.1016/j.ttbdis.2020.101411\u003c/li\u003e\n \u003cli\u003eMygland \u0026Aring;, Lj\u0026oslash;stad U, Fingerle V, Rupprecht T, Schmutzhard E, Steiner I. EFNS guidelines on the diagnosis and management of European lyme neuroborreliosis. \u003cem\u003eEur J Neurol\u003c/em\u003e. 2010;17(1):8-e4. doi:10.1111/j.1468-1331.2009.02862.x\u003c/li\u003e\n \u003cli\u003eObel N, Dessau RB, Krogfelt KA, et al. Long term survival, health, social functioning, and education in patients with European Lyme neuroborreliosis: nationwide population based cohort study. \u003cem\u003eBMJ\u003c/em\u003e. 2018;361. doi:10.1136/bmj.k1998\u003c/li\u003e\n \u003cli\u003eKnudtzen FC, Andersen NS, Jensen TG, Skarph\u0026eacute;dinsson S. Characteristics and Clinical Outcome of Lyme Neuroborreliosis in a High Endemic Area, 1995-2014: A Retrospective Cohort Study in Denmark. \u003cem\u003eClin Infect Dis\u003c/em\u003e. 2017;65(9):1489-1495. doi:10.1093/cid/cix568\u003c/li\u003e\n \u003cli\u003eHaahr R, Tetens MM, Dessau RB, et al. Risk of Neurological Disorders in Patients With European Lyme Neuroborreliosis: A Nationwide, Population-Based Cohort Study. \u003cem\u003eClin Infect Dis\u003c/em\u003e. 2019;71(6):1511-1516. doi:10.1093/cid/ciz997\u003c/li\u003e\n \u003cli\u003eNocton JJ, Bloom BJ, Rutledge BJ, et al. Detection of Borrelia burgdorferi DNA by polymerase chain reaction in cerebrospinal fluid in Lyme neuroborreliosis. \u003cem\u003eJ Infect Dis\u003c/em\u003e. 1996;174(3):623-627. doi:10.1093/infdis/174.3.623\u003c/li\u003e\n \u003cli\u003eLebech AM, Hansen K, Brandrup F, Clemmensen O, Halkier-S\u0026oslash;rensen L. Diagnostic value of PCR for detection of Borrelia burgdorferi DNA in clinical specimens from patients with erythema migrans and Lyme neuroborreliosis. \u003cem\u003eMol diagnosis a J devoted to Underst Hum Dis through Clin Appl Mol Biol\u003c/em\u003e. 2000;5(2):139-150. doi:10.1007/BF03262032\u003c/li\u003e\n \u003cli\u003eSkogman BH, Wilhelmsson P, Atallah S, Petersson AC, Ornstein K, Lindgren PE. Lyme neuroborreliosis in Swedish children\u0026mdash;PCR as a complementary diagnostic method for detection of Borrelia burgdorferi sensu lato in cerebrospinal fluid. \u003cem\u003eEur J Clin Microbiol Infect Dis\u003c/em\u003e. 2021;40(5):1003-1012. doi:10.1007/s10096-020-04129-7\u003c/li\u003e\n \u003cli\u003eLeth TA, Nymark A, Knudtzen FC, et al. Detection of Borrelia burgdorferi sensu lato DNA in cerebrospinal fluid samples following pre-enrichment culture. \u003cem\u003eTicks Tick Borne Dis\u003c/em\u003e. 2023;14(3):102138. doi:https://doi.org/10.1016/j.ttbdis.2023.102138\u003c/li\u003e\n \u003cli\u003eTetens MM, Dessau R, Ellermann-Eriksen S, et al. The diagnostic value of serum Borrelia burgdorferi antibodies and seroconversion after Lyme neuroborreliosis, a nationwide observational study. \u003cem\u003eClin Microbiol Infect\u003c/em\u003e. 2022;28(11):1500.e1-1500.e6. doi:10.1016/J.CMI.2022.06.001\u003c/li\u003e\n \u003cli\u003eNiu L, Thiele M, Geyer PE, et al. Noninvasive proteomic biomarkers for alcohol-related liver disease. \u003cem\u003eNat Med\u003c/em\u003e. 2022;28(6):1277. doi:10.1038/S41591-022-01850-Y\u003c/li\u003e\n \u003cli\u003eAngel TE, Jacobs JM, Smith RP, et al. Cerebrospinal fluid proteome of patients with acute Lyme disease. \u003cem\u003eJ Proteome Res\u003c/em\u003e. 2012;11(10):4814-4822. doi:10.1021/PR300577P\u003c/li\u003e\n \u003cli\u003eHalperin JJ, Eikeland R, Branda JA, Dersch R. Lyme neuroborreliosis: known knowns, known unknowns. \u003cem\u003eBrain\u003c/em\u003e. 2022;145(8):2635-2647. doi:10.1093/brain/awac206\u003c/li\u003e\n \u003cli\u003eKoedel U, Fingerle V, Pfister HW. Lyme neuroborreliosis - Epidemiology, diagnosis and management. \u003cem\u003eNat Rev Neurol\u003c/em\u003e. 2015;11(8):446-456. doi:10.1038/nrneurol.2015.121\u003c/li\u003e\n \u003cli\u003eFredriksson T, Brudin L, Henningsson AJ, Skogman BH, Tjernberg I. Diagnostic patterns of serum inflammatory protein markers in children with Lyme neuroborreliosis. \u003cem\u003eTicks Tick Borne Dis\u003c/em\u003e. 2024;15(4):102349. doi:10.1016/j.ttbdis.2024.102349\u003c/li\u003e\n \u003cli\u003eGęgotek A, Skrzydlewska E, Groth M, Czupryna P, Moniuszko-Malinowska A. Changes in the serum proteome profile of patients with neuroborreliosis, foresters, and patients treated according to ILADS method. \u003cem\u003eMicrob Pathog\u003c/em\u003e. 2024;197(July). doi:10.1016/j.micpath.2024.107094\u003c/li\u003e\n \u003cli\u003eNilsson K, Skoog E, Edvinsson M, M\u0026aring;rtensson A, Olsen B. Protein biomarker profiles in serum and CSF in 158 patients with PTLDS or persistent symptoms after presumed tick-bite exposure compared to those in patients with confirmed acute neuroborreliosis. \u003cem\u003ePLoS One\u003c/em\u003e. 2022;17(11 November):1-17. doi:10.1371/journal.pone.0276407\u003c/li\u003e\n \u003cli\u003eSchutzer SE, Angel TE, Liu T, et al. Distinct cerebrospinal fluid proteomes differentiate post-treatment Lyme disease from Chronic fatigue syndrome. \u003cem\u003ePLoS One\u003c/em\u003e. 2011;6(2):1-8. doi:10.1371/journal.pone.0017287\u003c/li\u003e\n \u003cli\u003eBockenstedt LK, Wooten RM, Baumgarth N. Immune response to borrelia: Lessons from lyme disease spirochetes. \u003cem\u003eCurr Issues Mol Biol\u003c/em\u003e. 2020;42:145-190. doi:10.21775/cimb.042.145\u003c/li\u003e\n \u003cli\u003eRupprecht TA, Koedel U, Fingerle V, Pfister HW. The pathogenesis of lyme neuroborreliosis: From infection to inflammation. \u003cem\u003eMol Med\u003c/em\u003e. 2008;14(3-4):205-212. doi:10.2119/2007-00091.Rupprecht\u003c/li\u003e\n \u003cli\u003eTriolo D, Dina G, Taveggia C, et al. Vimentin regulates peripheral nerve myelination. \u003cem\u003eDevelopment\u003c/em\u003e. 2012;139(7):1359-1367. doi:10.1242/dev.072371\u003c/li\u003e\n \u003cli\u003eRadzi\u0026scaron;auskienė D, Urbonienė J, Jasionis A, et al. Clinical and epidemiological features of Lyme neuroborreliosis in adults and factors associated with polyradiculitis, facial palsy and encephalitis or myelitis. \u003cem\u003eSci Rep\u003c/em\u003e. 2023;13(1):1-11. doi:10.1038/s41598-023-47312-4\u003c/li\u003e\n \u003cli\u003eSolheim AM, Skarstein I, Quarsten H, et al. Clinical and laboratory characteristics during a 1-year follow-up in European Lyme neuroborreliosis: A prospective cohort study. \u003cem\u003eEur J Neurol\u003c/em\u003e. 2024;(August):1-10. doi:10.1111/ene.16487\u003c/li\u003e\n \u003cli\u003evan Samkar A, Bruinsma RA, Vermeeren YM, et al. Clinical characteristics of Lyme neuroborreliosis in Dutch children and adults. \u003cem\u003eEur J Pediatr\u003c/em\u003e. 2023;182(3):1183-1189. doi:10.1007/s00431-022-04749-5\u003c/li\u003e\n \u003cli\u003eMens H, Fjordside L, Gynthersen R, et al. Plasma neurofilament light significantly decreases following treatment in Lyme neuroborreliosis and not associated with persistent symptoms. doi:10.1111/ene.15707\u003c/li\u003e\n \u003cli\u003eFischer JR, LeBlanc KT, Leong JM. Fibronectin binding protein BBK32 of the Lyme disease spirochete promotes bacterial attachment to glycosaminoglycans. \u003cem\u003eInfect Immun\u003c/em\u003e. 2006;74(1):435-441. doi:10.1128/IAI.74.1.435-441.2006\u003c/li\u003e\n \u003cli\u003eBrissette CA, Bykowski T, Cooley AE, Bowman A, Stevenson B. Borrelia burgdorferi RevA antigen binds host fibronectin. \u003cem\u003eInfect Immun\u003c/em\u003e. 2009;77(7):2802-2812. doi:10.1128/IAI.00227-09\u003c/li\u003e\n \u003cli\u003eBruinsma RA, Zomer TP, Skogman BH, van Hensbroek MB, Hovius JW. Clinical manifestations of Lyme neuroborreliosis in children: a review. \u003cem\u003eEur J Pediatr\u003c/em\u003e. 2023;182(5):1965-1976. doi:10.1007/s00431-023-04811-w\u003c/li\u003e\n \u003cli\u003eGarro A, Avery RA, Cohn KA, et al. Validation of the Rule of 7\u0026rsquo;s for Identifying Children at Low-risk for Lyme Meningitis. \u003cem\u003ePediatr Infect Dis J\u003c/em\u003e. 2021;40(4):306-309. doi:10.1097/INF.0000000000003003\u003c/li\u003e\n \u003cli\u003eSkogman BH, Sj\u0026ouml;wall J, Lindgren PE. The NeBoP score - a clinical prediction test for evaluation of children with Lyme Neuroborreliosis in Europe. \u003cem\u003eBMC Pediatr\u003c/em\u003e. 2015;15(1):1-9. doi:10.1186/s12887-015-0537-y\u003c/li\u003e\n \u003cli\u003eBlanc F, Jaulhac B, Fleury M, et al. Relevance of the antibody index to diagnose Lyme neuroborreliosis among seropositive patients. \u003cem\u003eNeurology\u003c/em\u003e. 2007;69(10):953-958. doi:10.1212/01.wnl.0000269672.17807.e0\u003c/li\u003e\n \u003cli\u003eStrle F, Henningsson AJ, Strle K. Diagnostic Utility of CXCL13 in Lyme Neuroborreliosis. \u003cem\u003eClin Infect Dis\u003c/em\u003e. 2021;72(10):1727-1729. doi:10.1093/cid/ciaa337\u003c/li\u003e\n \u003cli\u003eBaird GS, Nelson SK, Keeney TR, et al. Age-dependent changes in the cerebrospinal fluid proteome by slow off-rate modified aptamer array. \u003cem\u003eAm J Pathol\u003c/em\u003e. 2012;180(2):446-456. doi:10.1016/j.ajpath.2011.10.024\u003c/li\u003e\n \u003cli\u003eZhang J, Goodlett DR, Peskind ER, et al. Quantitative proteomic analysis of age-related changes in human cerebrospinal fluid. \u003cem\u003eNeurobiol Aging\u003c/em\u003e. 2005;26(2):207-227. doi:10.1016/j.neurobiolaging.2004.03.012\u003c/li\u003e\n \u003cli\u003eLehallier B, Gate D, Schaum N, et al. the lifespan. 2020;25(12):1843-1850. doi:10.1038/s41591-019-0673-2.Undulating\u003c/li\u003e\n \u003cli\u003eHeld F, Makarov C, Gasperi C, et al. Proteomics Reveals Age as Major Modifier of Inflammatory CSF Signatures in Multiple Sclerosis. \u003cem\u003eNeurol Neuroimmunol neuroinflammation\u003c/em\u003e. 2025;12(1):e200322. doi:10.1212/NXI.0000000000200322\u003c/li\u003e\n \u003cli\u003eWesenhagen KEJ, Gobom J, Bos I, et al. Effects of age, amyloid, sex, and APOE \u0026epsilon;4 on the CSF proteome in normal cognition. \u003cem\u003eAlzheimer\u0026rsquo;s Dement Diagnosis, Assess Dis Monit\u003c/em\u003e. 2022;14(1):1-12. doi:10.1002/dad2.12286\u003c/li\u003e\n \u003cli\u003eLaugesen K, Mengel-From J, Christensen K, et al. A Review of Major Danish Biobanks: Advantages and Possibilities of Health Research in Denmark. \u003cem\u003eClin Epidemiol\u003c/em\u003e. 2023;15(February):213-239. doi:10.2147/CLEP.S392416\u003c/li\u003e\n \u003cli\u003eBader JM, Geyer PE, M\u0026uuml;ller JB, et al. Proteome profiling in cerebrospinal fluid reveals novel biomarkers of Alzheimer\u0026rsquo;s disease. \u003cem\u003eMol Syst Biol\u003c/em\u003e. 2020;16(6):9356. doi:10.15252/MSB.20199356\u003c/li\u003e\n \u003cli\u003eGeyer PE, Wewer Albrechtsen NJ, Tyanova S, et al. Proteomics reveals the effects of sustained weight loss on the human plasma proteome. \u003cem\u003eMol Syst Biol\u003c/em\u003e. Published online 2016. doi:10.15252/msb.20167357\u003c/li\u003e\n \u003cli\u003eKulak NA, Pichler G, Paron I, Nagaraj N, Mann M. Minimal, encapsulated proteomic-sample processing applied to copy-number estimation in eukaryotic cells. \u003cem\u003eNat Methods\u003c/em\u003e. Published online 2014. doi:10.1038/nmeth.2834\u003c/li\u003e\n \u003cli\u003eDemichev V, Messner CB, Vernardis SI, Lilley KS, Ralser M. DIA-NN: neural networks and interference correction enable deep proteome coverage in high throughput. \u003cem\u003eNat Methods\u003c/em\u003e. 2020;17(1):41-44. doi:10.1038/s41592-019-0638-x\u003c/li\u003e\n \u003cli\u003eKrismer E, Bludau I, Strauss MT, Mann M. AlphaPeptStats: An open-source Python package for automated and scalable statistical analysis of mass spectrometry-based proteomics. \u003cem\u003eBioinformatics\u003c/em\u003e. 2023;39(8):1-4. doi:10.1093/bioinformatics/btad461\u003c/li\u003e\n \u003cli\u003eSantos A, Cola\u0026ccedil;o AR, Nielsen AB, et al. A knowledge graph to interpret clinical proteomics data. \u003cem\u003eNat Biotechnol\u003c/em\u003e. 2022;40(5):692-702. doi:10.1038/s41587-021-01145-6\u003c/li\u003e\n \u003cli\u003eWebel H, Niu L, Nielsen AB, et al. Imputation of label-free quantitative mass spectrometry-based proteomics data using self-supervised deep learning. \u003cem\u003eNat Commun\u003c/em\u003e. 2024;15(1):5405. doi:10.1038/s41467-024-48711-5\u003c/li\u003e\n \u003cli\u003eBehdenna A, Colange M, Haziza J, et al. pyComBat, a Python tool for batch effects correction in high-throughput molecular data using empirical Bayes methods. \u003cem\u003eBMC Bioinformatics\u003c/em\u003e. 2023;24(1):1-9. doi:10.1186/s12859-023-05578-5\u003c/li\u003e\n \u003cli\u003eAshburner M, Ball CA, Blake JA, et al. Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. \u003cem\u003eNat Genet\u003c/em\u003e. 2000;25(1):25-29. doi:10.1038/75556\u003c/li\u003e\n \u003cli\u003eLundberg SM, Lee SI. A unified approach to interpreting model predictions. In: \u003cem\u003eProceedings of the 31st International Conference on Neural Information Processing Systems\u003c/em\u003e. NIPS\u0026rsquo;17. Curran Associates Inc.; 2017:4768\u0026ndash;4777.\u003c/li\u003e\n \u003cli\u003ePerez-Riverol Y, Bai J, Bandla C, et al. The PRIDE database resources in 2022: A hub for mass spectrometry-based proteomics evidences. \u003cem\u003eNucleic Acids Res\u003c/em\u003e. 2022;50(D1):D543-D552. doi:10.1093/nar/gkab1038\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"MATERIALS AND METHODS","content":"\u003cp\u003e\u003cstrong\u003eEthics\u003cbr\u003e\u003c/strong\u003eSamples were obtained during diagnostic investigation and all patients had given their informed consent to the storage of biological material and its future use in research. The study was approved by Knowledge Center for Data Reviews (P-2019-707) and the local ethics committee (H-17024315). The biobanks were approved by the Danish Data Protection Agency (Rigshospitalet: j.nr.: 2012-41-0036).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStudy design\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was an observational retrospective cohort study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStudy population and setting\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe identified a total of 483 CSF (n = 308) and plasma (n = 175) samples from adults (\u0026gt;18 years) with LNB, erythema migrans, acrodermatitis chronica atrophicans, post-treatment-Lyme-disease-syndrome, viral meningitis, and individuals who were investigated for suspected LNB, but had normal CSF (controls) from; i) The Danish National Biobank (samples collected from 2001-2011)\u003csup\u003e43\u003c/sup\u003e and ii) the Biobank of the Department of Infectious Diseases at Copenhagen University Hospital, Rigshospitalet (samples collected from 2016-2022). After informed patient consent for storage and future use in research was provided, CSF samples were labelled and stored at -80°C and blood samples in EDTA tubes were spun and the supernatant was transferred to Eppendorf tubes, labelled, and stored at -80°C. Samples from the two sites of origin were kept separate in the subsequent analyses to preserve independent cohorts for model development and validation. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCohorts\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSamples were divided in four cohorts for diagnostic model development and validation in CSF and plasma respectively. The CSF development cohort consisted of\u0026nbsp;145 CSF samples from patients diagnosed with LNB (n=49), viral meningitis (n=44) and controls\u0026nbsp;(n=52). The CSF validation cohort included a total of 163 CSF samples from patients diagnosed with LNB (n=69), viral meningitis (n=45), and controls (n=49). The plasma development cohort included a total of 95 plasma samples from patients with LNB (n=27), viral meningitis (n=20) and controls (n=48) whereas the plasma validation cohort consisted of 80 plasma samples from patients with LNB (n=10), viral meningitis (n=18), controls (n=20) and an additional cohort of plasma samples from patients with manifestations of Lyme borreliosis without CNS involvement including post-treatment-Lyme-disease-syndrome (n=17),\u0026nbsp;erythema migrans\u0026nbsp;(n=9), and\u0026nbsp;acrodermatitis chronica atrophicans\u0026nbsp;(n=6).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDefinitions of diagnostic groups\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLNB:\u003c/strong\u003e Patients diagnosed with LNB based on\u0026nbsp;a first-time positive \u003cem\u003eB. burgdorferi\u003c/em\u003e-specific intrathecal antibody test and the International Classification of Diseases 10\u003csup\u003eth\u003c/sup\u003e revision (ICD-10) diagnosis code Borreliosis: A69.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eViral Meningitis:\u003c/strong\u003e Patients with a final clinical diagnosis of viral meningitis determined by an infectious disease specialist at a tertiary referral university hospital. The diagnosis was based on clinical presentation, CSF pleocytosis and exclusion of other diagnoses.\u003c/p\u003e\n\u003cp\u003eIn 76 cases a PCR-verified viral aetiology was determined (Herpes simplex virus 2 (n=32), Enterovirus (n=30), Varicella zoster (n=9), Influenza A virus (n=3), Epstein-Barr virus (n=1), Toscana virus (n=1)).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eControls:\u003c/strong\u003e Individuals\u0026nbsp;who had a lumbar puncture performed because LNB was suspected, but where white-blood-cell (WBC) count in the CSF was within normal reference ranges (CSF-WBC: \u0026nbsp;0-5 x 10\u003csup\u003e6\u003c/sup\u003e/L) and the \u003cem\u003eB. burgdorferi\u0026nbsp;\u003c/em\u003es.l. specific intrathecal antibody index was negative.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePost-treatment\u003cem\u003e-\u003c/em\u003eLyme-disease-syndrome:\u003c/strong\u003e Patients with persisting symptoms of Lyme borreliosis \u0026gt;6 months after diagnosis and treatment.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eErythema migrans\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e:\u003c/strong\u003e Patients with a skin rash clinically diagnosed with \u003cem\u003eerythema migrans\u003c/em\u003e at the Unit for Tick- Borne Infections, Copenhagen University Hospital, Rigshospitalet.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eAcrodermatitis chronica atrophicans\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e:\u003c/strong\u003e Patients clinically diagnosed with \u003cem\u003eacrodermatitis chronica atrophicans\u003c/em\u003e skin rash and positive \u003cem\u003eB. burgdorferi\u0026nbsp;\u003c/em\u003es.l.IgG in blood at the Unit for Tick Borne Infections, Copenhagen University Hospital, Rigshospitalet. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eProteomic analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCSF and plasma samples were thawed and 100 µl were transferred to 96-well plates for analysis. The sample preparation was optimized based on the previously published methods described in\u003csup\u003e44,45\u003c/sup\u003e. Briefly, 20 µl CSF was were first denatured with 30 µl PreOmics Lysis buffer, while 5 µl plasma was denatured with 45 µl PreOmics Lysis buffer\u003csup\u003e46\u003c/sup\u003e. Both sample types were subsequently digested using LysC/trypsin enzyme mix. The resulting peptides were purified using two-gauge SDB-RPS StageTips, and the eluate analysed on Evosep One (Evosep Biosystem, Denmark) liquid chromatography system, coupled online to an Orbitrap Exploris 480 mass spectrometer. Data acquisition was performed in data-independent analysis (DIA), using 60 samples per day (SPD) gradient and 8 cm Pepsep column.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData processing\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eInitial data processing of the mass spectrometry raw files were performed with DIA-NN version 1.9 in a data independent search\u003csup\u003e47\u003c/sup\u003e. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe DIA-NN data underwent further processing using the Clinical Knowledge Graph (CKG), alphapeptstats and Jupyter Notebook\u003csup\u003e48,49\u003c/sup\u003e. Initially, a stringent filter for missing data was applied: 1) samples with low protein count, defined by a value below 1.5IQR from the 25\u003csup\u003eth\u003c/sup\u003e quantile of the combined distribution, were excluded, and 2) proteins with a missingness of more than 40% across samples were excluded. Data was log2 transformed. The remaining missing values were imputed with a variational autoencoder using the PIMMS software\u003csup\u003e50\u003c/sup\u003e. Assessment of sample quality was conducted as previously described\u003csup\u003e50\u003c/sup\u003e. Batch correction was executed using combat to overcome potential plate-specific bias on subsequent analyses\u003csup\u003e51\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eProteins exhibiting significantly different levels between the cohorts were identified by unpaired t-tests. Multiple hypothesis correction was applied with the Benjamini-Hochberg method, with adjusted P-values \u0026lt; 0.05 deemed statistically significant. P-values and protein abundances were visualized in Volcano plots, with -log10(corrected p-value) and log2 of protein fold change between groups. Venn diagrams were used to visualize overlaps in significant proteins between statistical comparisons. Heatmaps in combination with Sankey plots of gene ontology terms were used to highlight protein changes of the significant proteins together with their biological processes\u003csup\u003e52\u003c/sup\u003e. In the heatmaps, protein abundances were z-scored. Gene ontology terms were retrieved for each protein through UNIPROT. Gene ontology terms with frequency of less than 5% across proteins were not visualized.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eModel development\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eZ-scored data from the development cohorts were analysed using supervised machine learning to explore the potential for a diagnostic signature for LNB. Significant proteins identified by t-test analysis on the development cohorts were used as input features in the training data. Our data sets, comprising both samples from patients diagnosed with LNB and viral meningitis or control samples, was partitioned into a training set for model development and a test set for model validation using a 5-fold or 3-fold cross-validation (CV) approach for CSF and plasma respectively. The number of CV-folds was decided based on sample size with a minimum of 30 samples in each set. Both sets maintained an equal ratio of positive and negative cases (stratified k-fold cross-validation). The classification target used for analysis was the diagnosis of LNB and viral meningitis or LNB and control (yes/no). An appropriate ML model was determined by testing the performance of twelve different algorithms based on the\u0026nbsp;area under the curve (AUC) and Matthews Correlation Coefficient (MCC) in the development test set cross-validations. For each algorithm, optimal hyperparameters and features were selected during cross-validation. Feature importance were highlighted for the top 10 most predictive features with SHAP (SHapley Additive exPlanations) values\u003csup\u003e53\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eModel validation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe models trained on the development cohorts were subsequently applied to the validation cohorts and the performance was assessed by AUC and MCC. ROC curves and confusion matrices were used to visualize the performance of the classifiers.\u0026nbsp;\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Tick-Borne Diseases, Borrelia, Lyme Borreliosis, Lyme Disease, Lyme Neuroborreliosis, Central Nervous System Infections, Meningitis, Viral, Delayed Diagnosis, Diagnostic Techniques and Procedures, Proteomics, Machine Learning","lastPublishedDoi":"10.21203/rs.3.rs-6048306/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6048306/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eLyme neuroborreliosis (LNB), a severe nervous system infection caused by tick-borne spirochetes of the \u003cem\u003eBorrelia burgdorferi \u003c/em\u003esensu lato complex, represents one of the most frequent bacterial infections of the nervous system in Europe. Early diagnosis remains challenging due to limited sensitivity of current methods and requires invasive lumbar punctures, underscoring the need for improved, less invasive diagnostic tools.\u003c/p\u003e\n\u003cp\u003eHere, we applied mass spectrometry-based proteomics to analyse 308 cerebrospinal fluid (CSF) samples and 207 plasma samples from patients with LNB, viral meningitis, controls and other manifestations of Lyme borreliosis. Diagnostic panels of regulated proteins were identified and evaluated through machine learning-assisted proteome analyses.\u003c/p\u003e\n\u003cp\u003eIn CSF, the classifier distinguished LNB from viral meningitis and controls with AUCs of 0.92 and 0.90, respectively. In plasma, LNB was distinguished from controls with an AUC of 0.80.\u003c/p\u003e\n\u003cp\u003eOur findings highlight the diagnostic potential of machine learning-assisted proteomics for LNB in CSF and plasma.\u003c/p\u003e","manuscriptTitle":"Mass-spectrometry-based proteomics enables rapid and accurate diagnosis of Lyme neuroborreliosis in adults","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-02-28 08:23:17","doi":"10.21203/rs.3.rs-6048306/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"c17da92d-d0c4-4cf7-9714-c6799ae9fddf","owner":[],"postedDate":"February 28th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":44797931,"name":"Health sciences/Health care/Diagnosis/Laboratory techniques and procedures"},{"id":44797932,"name":"Health sciences/Diseases/Infectious diseases/Bacterial infection"}],"tags":[],"updatedAt":"2025-10-28T17:30:44+00:00","versionOfRecord":{"articleIdentity":"rs-6048306","link":"https://doi.org/10.1038/s41467-025-64903-z","journal":{"identity":"nature-communications","isVorOnly":false,"title":"Nature Communications"},"publishedOn":"2025-10-27 04:00:00","publishedOnDateReadable":"October 27th, 2025"},"versionCreatedAt":"2025-02-28 08:23:17","video":"","vorDoi":"10.1038/s41467-025-64903-z","vorDoiUrl":"https://doi.org/10.1038/s41467-025-64903-z","workflowStages":[]},"version":"v1","identity":"rs-6048306","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6048306","identity":"rs-6048306","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}