Evaluation of two assays for CXCL13 analysis in
cerebrospinal fluid for laboratory diagnosis of Lyme
neuroborreliosis
ANNA J. HENNINGSSON,1PAULA GYLLEMARK,2MALIN LAGER,3 BARBRO HEDIN SKOGMAN4and IVAR TJERNBERG5,6
1Department of Clinical Microbiology, Division of Medical Services, Region J€onk€oping County, J€onk€oping; 2Infectious Diseases, Region J€onk€oping County, J€onk€oping;3Laboratory Medicine, Division of Medical Services, Region J€onk€oping County, J€onk€oping;4Center for Clinical Research Dalarna, Falun;5Department
of Clinical Chemistry and Transfusion Medicine, Kalmar County Council, Kalmar; and6Division of Microbiology and Molecular Medicine, Department of Clinical and Experimental Medicine, Link€oping
University, Link€oping, Sweden
Henningsson AJ, Gyllemark P, Lager M, Skogman BH, Tjernberg I. Evaluation of two assays for CXCL13 analysis in cerebrospinal fluid for laboratory diagnosis of Lyme neuroborreliosis. APMIS 2016; 124: 985–990 We evaluated the diagnostic performance of two assays, one bead-based assay and one enzyme-linked immunosorbent assay (ELISA), for the determination of CXCL13 levels in cerebrospinal fluid (CSF) from patients with suspected Lyme neuroborreliosis (LNB). Patients investigated for LNB were retrospectively included (n= 132): 35 with definite LNB, 8 with possible LNB with CSF pleocytosis but normal antibody index (AI), 6 with possible LNB with elevated AI but no CSF pleocytosis and 83 non-LNB patients. CSF samples had been drawn before antibiotic treatment and were analysed for CXCL13 by Quantikine ELISA (R&D Systems) and recomBead (Mikrogen). Receiver operating characteristic anal-yses based on the definite LNB and non-LNB groups revealed a best performance cut-off of 56 pg/mL for Quantikine and 158 pg/mL for recomBead (sensitivity and specificity 100% for both assays). When applying these cut-off levels on the study groups, the two assays performed equally well regarding sensitivity and specificity. In the group of patients with pleocytosis but negative AI, the majority of whom were children with short symptom duration, the CXCL13 anal-ysis supported the LNB diagnosis in half of the cases. We consider CSF-CXCL13 analanal-ysis a useful diagnostic tool, in addition to Borrelia-specific AI, in laboratory diagnostics of LNB.
Key words: Lyme neuroborreliosis; cerebrospinal fluid; CXCL13; sensitivity; specificity.
Anna J. Henningsson, Clinical Microbiology, Division of Medical Services, Region J€onk€oping County, S-551 85 J€onk€oping, Sweden. e-mail: anna.jonsson.henningsson@rjl.se
AJH and PG contributed equally to this work.
Lyme neuroborreliosis (LNB) is the most common form of disseminated Lyme borreliosis in Sweden as well as in other parts of Europe (1, 2). Accord-ing to European guidelines, the diagnosis of definite LNB requires neurological symptoms, pleocytosis in the cerebrospinal fluid (CSF) and intrathecal production of Borrelia-specific antibodies (positive antibody index, AI) (3). Patients fulfilling only two of the three criteria may be diagnosed with possible LNB. However, in very early disease, the sensitivity of antibody tests may be low, and a positive AI
may also persist for a long time after a passed LNB (4, 5).
The B-cell-attracting chemokine CXCL13 has in previous studies been shown to be reliably elevated in the CSF of patients with early LNB, and further-more, to decrease rapidly after antibiotic treatment (6–9). Thus, analysis of CXCL13 in the CSF may be helpful in AI-negative patients with possible early LNB, as well as a marker for active disease and in control of therapeutic response in AI-posi-tive patients. Furthermore, CXCL13 has been proven to be useful for discriminating acute LNB from other CNS disorders (10–12). However, the Received 7 April 2016. Accepted 26 July 2016
appropriate cut-off level has been debated and pre-vious studies based on the same enzyme-linked immunosorbent assay (ELISA) have used different cut-off values (8–15).
We have previously evaluated the performance of CSF CXCL13 in LNB diagnosis, and the results suggested an improved diagnostic sensitivity in addition to Borrelia-specific AI, especially in chil-dren with possible LNB (14). In this study, the aim was to evaluate and compare the diagnostic utility and performance of two different assays for the determination of CXCL13 concentrations in CSF in adult and paediatric patients with acute LNB.
MATERIALS AND METHODS
Patients investigated for suspected LNB during 2007 through 2009 were included retrospectively (n= 132): 35 with definite LNB according to the European guidelines (group 1) (3), 8 with possible LNB based on CSF pleocy-tosis and neurological symptoms strongly suggestive of LNB (all had meningitis and/or facial palsy) but normal AI (group 2) and 6 with possible LNB based on elevated Borrelia-specific AI but no CSF pleocytosis (group 3). As a non-LNB reference group, we selected, from the patients investigated for suspected LNB, 83 patients with normal Borrelia-specific AI and no CSF pleocytosis (group 4). The patients in group 4 had symptoms such as headache, vertigo, paraesthesia, numbness, fatigue and myalgia, and some received other specified diagnoses such as Bell’s palsy (n= 16), cerebral vascular disease (n = 6), autoim-mune disorders (n= 8), psychiatric diagnoses (n = 4), sci-atica (n= 3), trigeminal neuralgia (n = 2), Alzheimer’s disease (n= 2), Parkinson’s disease (n = 1), amyotrophic lateral sclerosis (n= 1), multiple sclerosis (n = 1), lym-phoma (n= 1), hypertension (n = 1) and deterioration of post-poliomyelitis syndrome (n= 1). Characteristics of the patient groups are presented in Table 1. The classification of patients was based on clinical evaluation according to medical records, CSF cell count (pleocytosis was defined as mononuclear cells >5/lL) and AI determined by the recomBead Borrelia IgM and IgG assays (Mikrogen Diag-nostik GmbH, Neuried, Germany). We have previously evaluated the performance of this CSF Borrelia anti-body assay and found that this bead-based multiple
antigen test achieved both high sensitivity and specificity in laboratory diagnostics of LNB (16).
CSF specimens were collected prior to antibiotic treat-ment and analysed for CXCL13 by the Quantikine ELISA Human CXCL13/BLC/BCA-1 (R&D Systems, Min-neapolis, MN, USA) and the bead-based assay recomBead CXCL13 (Mikrogen Diagnostik GmbH) according to instructions from the manufacturers. All samples were stored at 20°C pending analysis, and the samples were frozen and thawed once between the Quantikine ELISA CXCL3 and the recomBead CXCL13 analyses. The recomBead CXCL13 assay uses Luminex xMAP technol-ogy and, in this study, a Bio-Plex 200 System (Bio-Rad Laboratories, Inc., Hercules, CA, USA) was used together with the software Luminex Xponent version 3.1 (Luminex Corporation, Austin, TX, USA) and recomQuant version 4.93 (Mikrogen Diagnostik GmbH). The typical standard curve for the Quantikine ELISA CXCL13 allowed mea-surements between 0.04 and 500 pg/mL. Samples with higher CXCL13 concentrations were re-analysed after dilution 1:10 and 1:50, but this was not carried out for the recomBead CXCL13 assay as this test allowed measure-ments between 9 and 1000 pg/mL and the available sam-ple volumes were limited. The low-positive and the high-positive controls were within the defined target range for the recomBead CXCL13 assay in every run. Samples with CXCL13 concentrations below the detection limit were given half the value of the lowest point of the standard curve.
Intra- and inter-assay coefficient of variations (CVs) for the Quantikine ELISA CXCL13 were calculated previ-ously from one internal control at approximately 50 pg/ mL run at three separate locations per 96-well microplate and based on the results from 15 microplates (14). Intra-and inter-assay CVs for the recomBead CXCL13 assay were determined using a sample with a CXCL13 concen-tration around 250 pg/mL run at four different locations per 96-well microplate and based on the results from four microplates.
Statistical analyses were performed using IBM SPSS Statistics version 20.0 (SPSS Inc., Chicago, IL, USA). For comparison of continuous variables between multiple study groups, the Kruskal–Wallis ANOVA test was used, followed by the Mann–Whitney U-test as a post hoc anal-ysis. Two-tailed tests were used, and p-values <0.05 were considered to be significant. In order to determine and compare the ability of the two different CXCL13 assays to discriminate between diagnostic groups, cut-off values and area under curve (AUC) were calculated using group 1
Table 1. Characteristics of the patient groups CSF mononuclear cells>5/lL CSF anti-Borrelia AI Age (years): median; range
Males (%) Duration of symptoms (days): median; range Definite LNB (group 1)
n= 35 + +
38; 3–72 49 14; 2–730 Possible LNB pleo (group 2)
n= 8 + 21; 3–55 88 5; 1–28 Possible LNB AI (group 3) n= 6 + 64; 50–81 50 294; 21–730 Non-LNB (group 4) n= 83 39; 1–83 43 28; 1–36501 CSF, cerebrospinal fluid; AI, antibody index; LNB, Lyme neuroborreliosis; Pleo, pleocytosis.
1
(definite LNB) and group 4 (non-LNB) in receiver operat-ing characteristic (ROC) analyses usoperat-ing MedCalc version 16.1 (MedCalc Software, Ostend, Belgium).
The study was approved by the Regional Ethical Review Board in Link€oping, Sweden (2012/246-31). RESULTS
The CSF CXCL13 levels as measured by Quan-tikine ELISA CXCL13 and recomBead CXCL13, respectively, are presented for all patient groups in Fig. 1. The intra-assay CV for the Quantikine ELISA CXCL13 was 9.0%, and the inter-assay CV was 15.3% (14). For the recomBead CXCL13
assay, the intra-assay CV was 6.7% and the inter-assay CV was 10.2%.
ROC analyses based on group 1 (definite LNB) and group 4 (non-LNB) revealed a best perfor-mance cut-off value of 56 pg/mL for Quantikine ELISA CXCL13 (sensitivity and specificity 100%) and 158 pg/mL for recomBead CXCL13 (sensitivity and specificity 100%).
There are, however, well-known biological limita-tions to the AI used in the classification of patients, that is it may be negative in early LNB and it usu-ally remains elevated for years even after success-fully treated LNB (3). As CXCL13 levels are expected to be elevated in CSF before B lympho-cytes are recruited to the CNS, and thus, before intrathecal Borrelia-specific antibodies can be detected, we wanted to evaluate the diagnostic per-formance of the two CXCL13 assays in AI-negative patients with pleocytosis and short duration of symptoms strongly suggestive of LNB. Due to the lack of a laboratory gold standard, we performed a thorough evaluation of clinical symptoms and med-ical history by scrutinizing the patients’ medmed-ical records.
The patients in group 2, who are classified as pos-sible LNB cases according to the European guideli-nes (3), were assessed as early acute LNB after review of their medical records. They all had menin-gitis and/or facial palsy, CSF mononuclear pleocyto-sis and short duration of symptoms (all, except two patients, had had their symptoms less than a week). Three of them had Borrelia-specific IgM or IgG anti-bodies in serum. One patient had an erythema migrans and one patient had noticed a tick bite prior to the onset of neurological symptoms.
In contrast, the patients in group 3, who are also classified as possible LNB according to the current guidelines, had an elevated Borrelia-specific AI, no CSF pleocytosis and symptoms of significantly longer duration compared to the other LNB groups, groups 1 and 2 (p = 0.002 and 0.001 respectively). Review of the medical records revealed that one of the patients in group 3 had been diagnosed with, and treated for, LNB 2 years earlier. The other five patients had symptoms such as paraesthesia, dysar-thria, cognitive dysfunction, headache, fatigue, myal-gia, arthralgia, neck pain, hyperaesthesia and radiating pain. All of the patients had speci-fic IgG antibodies in serum and elevated Borrelia-specific IgG AI, whereas Borrelia-Borrelia-specific IgM could not be detected neither in serum nor in CSF (Borre-lia-specific antibodies in serum had been analysed previously by Enzygnost Lyme link VlsE/IgG and Enzygnost Borreliosis/IgM, Siemens, Marburg, Ger-many, as part of the routine investigation of the patients’ symptoms). Thus, the diagnosis of acute
Group 1 Group 2 Group 3 Group 4
Group 1 Group 2 Group 3 Group 4
0.01 0.1 1 10 100 1000 10 000 10 000 1000 100 10 1 100 000 Quantikine ELISA CXCL13 recomBead CXCL13 CXCL13 i n CSF ( p g/ m L ) CXCL13 i n CSF ( p g/ m L ) p<0.0001 p<0.0001 p<0.0001 p = 0.013 p = 0.002p<0.0001 p<0.0001 p = 0.01 p<0.0001 p = 0.04 p<0.0001 A B
Fig. 1. CXCL13 concentrations in cerebrospinal fluid (CSF) from patients in the different groups. (A) CXCL13 measured by Quantikine ELISA CXCL13. (B) CXCL13 measured by recomBead CXCL13. Group 1 (definite Lyme neuroborreliosis, LNB), group 2 (possible LNB with CSF pleocytosis but normal Borrelia-specific antibody index, AI), group 3 (possible LNB with elevated Borrelia-specific AI but no CSF pleocytosis) and group 4 (non-LNB patients). Black lines indicate the median level in each group.
LNB at the time of the lumbar puncture and CSF sampling used for this study is dubious, and there-fore, we chose to calculate the sensitivity and speci-ficity for the two CXCL13 assays without this group (Table 2).
When applying the cut-off level of 56 pg/mL for the Quantikine ELISA CXCL13, four of the eight patients in group 2 had CXCL13 levels above the cut-off, and when applying the cut-off level of 158 pg/mL for the recomBead CXCL13 assay, five of the eight patients had CSF-CXCL13 levels above the cut-off. In group 3, all of the six patients had CSF-CXCL13 levels below the suggested cut-off with the Quantikine ELISA CXCL13, whereas with the recomBead CXCL13 assay, two of the six patients had CSF-CXCL13 levels slightly above the suggested cut-off.
DISCUSSION
In this study, we wanted to evaluate and compare the diagnostic performance of two different assays for CXCL13 analysis in CSF, one ELISA and one bead-based assay, and to assess their possible con-tribution to laboratory diagnostics in acute LNB. We observed substantial differences in concentra-tions when measuring CXCL13 by the bead-based assay compared to the ELISA, with generally much higher levels for measurements carried out by the bead-based assay. Discrepancies between these two assay formats have been described before (17, 18), and the variations in absolute levels are probably due to different design of antibody pairs. Another, but less plausible, explanation for the higher con-centrations found with the bead-based assay is the interference of heterophilic antibodies (19).
The ROC analysis revealed that the two assays could equally well discriminate between group 1 (definite LNB) and group 4 (non-LNB), but at two completely different cut-off levels. This illustrates clearly that the best performance cut-off is not absolute but entirely dependent on the method and assay that is used. Moreover, previous studies, even if based on the same ELISA, have suggested differ-ent cut-off values (8–15). These difficulties to
establish general cut-off levels for CSF-CXCL13 analysis in LNB diagnostics render interpretation of results more complicated, especially for moder-ately elevated levels, and may constitute an impedi-ment for laboratories and clinicians in the implementation of this analysis in clinical practice.
The manufacturer of the bead-based assay recomBead CXCL13 suggests that CXCL13 levels of 190–300 pg/mL are considered as borderline results, and levels >300 pg/mL are interpreted as strongly supportive of the acute LNB diagnosis. Our results roughly corroborate these recommenda-tions, as 190 pg/mL is comparable to the best per-formance cut-off level of 158 pg/mL that we found in this study, and 86% of the acute LNB patients (groups 1 and 2) had CSF-CXCL13 levels >300 pg/ mL with the recomBead CXCL13 assay.
The Quantikine ELISA CXCL13 and the recom-Bead CXCL13 assays performed equally well when applying the suggested cut-off levels on patients assessed as acute LNB (groups 1 and 2). It is our opinion that group 3 (elevated Borrelia-specific AI but no CSF pleocytosis) does not represent acute LNB, in contrast to group 2 (CSF pleocytosis, short duration of symptoms strongly suggestive of LNB, but negative Borrelia-specific AI) which is very likely to consist of patients with early acute LNB. Indeed, we find that there are certain limita-tions in the classification of LNB patients in the current European guidelines (3), according to which both our groups 2 and 3 are classified as possible LNB, whereas we find important differences between them. We suggest that patients correspond-ing to group 3 could be considered as previous LNB, and patients corresponding to group 2 as probable acute LNB.
Both CXCL13 assays evaluated here could sup-port the acute LNB diagnosis in about 50% of the patients in group 2, mainly children, which we think is of value for the clinician who is offered additional information to be evaluated in relation to the medical history, clinical signs and symptoms, as well as other laboratory findings. In the group of patients with possible LNB and elevated Borrelia-specific AI but no CSF pleocytosis (group 3), the CSF-CXCL13 levels were below the suggested
cut-Table 2. Sensitivity and specificity of the two assays Quantikine ELISA CXCL13 and recomBead CXCL13 based on groups 1 and 2 taken together (definite LNB and possible LNB with pleocytosis) vs group 4 (non-LNB)
Assay Result LNB (n= 43) Non-LNB (n= 83) Sensitivity (%)
Specificity (%)
Quantikine ELISA (cut-off 56 pg/mL) Positive 39 0 91
100
Negative 4 83
recomBead CXCL13 (cut-off 158 pg/mL) Positive 40 0 93
100
Negative 3 83
off levels in all but two patients who had CXCL13 levels slightly above the cut-off for the recomBead CXCL13 assay, but the CXCL13 analysis did not give any conclusive additional information besides the cell count.
To be applicable in clinical practice, assays for CXCL13 measurements in CSF should not only be accurate, but they should also be fast, easy to use, relatively inexpensive and require the least possible sample volume. The recomBead CXCL13 assay met several of these criteria as it required little hands-on time, all reagents were ready to use after thaw-ing and data analysis was performed by the soft-ware. However, this assay is rather expensive and it also requires access to expensive equipment that may not be available in every laboratory. The two assays required the same sample volume (50lL) and had comparable incubation steps. The Quan-tikine ELISA CXCL13, however, involved standard dilution and the manual generation of a standard curve. The ELISA format does not require the use of a special instrument apart from an ordinary spectrophotometer, and another advantage may be the rather extensive amount of data and widespread experience from this assay, as it has been used in most of the previously published studies on CXCL13 in LNB.
Possible limitations in our study may be the ret-rospective study design and the lack of a fully reli-able gold standard for classification of patients. However, we have based our classification of the patient groups on the recomBead Borrelia IgM and IgG AI, a test that we previously have evaluated on a similar group of patients, the CSF cell count and review of the patients’ medical records. Another possible limitation may be that the non-LNB patients did not have CSF pleocytosis and therefore had lower levels of CSF-CXCL13. On the other hand, the use of CSF-CXCL13 in discriminating LNB from other CNS conditions with pleocytosis has been addressed in previous studies (6, 9, 10, 12, 13, 20) and was not the principal aim of this study.
In conclusion, the Quantikine ELISA CXCL13 and the recomBead CXCL13 assay performed equally well regarding sensitivity and specificity in patients evaluated for acute LNB. In the group of patients with pleocytosis but negative Borrelia-spe-cific AI, the majority of whom were children with short duration of symptoms, the CXCL13 analysis supported the LNB diagnosis in about half of the cases. Thus, we consider CXCL13 analysis in CSF, a useful diagnostic tool, in addition to Borrelia-spe-cific AI, in laboratory diagnostics of LNB. How-ever, a limitation of CSF-CXCL13 analysis is the difficulty to establish generally applicable cut-off levels.
The work was financially supported by the Division of Medical Services, Region J€onk€oping County, and by ScandTick Innovation, an Interreg V A project.
REFERENCES
1. Berglund J, Eitrem R, Ornstein K, Lindberg A, Ringer A, Elmrud H, et al. An epidemiologic study of Lyme disease in southern Sweden. N Engl J Med 1995;333:1319–27.
2. Cimmino MA. Relative frequency of Lyme borreliosis and of its clinical manifestations in Europe. European Community Concerted Action on Risk Assessment in Lyme Borreliosis. Infection 1998;26:298–300.
3. Mygland A, Ljøstad U, Fingerle V, Rupprecht T, Schmutzhard E, Steiner I. EFNS guidelines on the diagnosis and management of European Lyme neu-roborreliosis. Eur J Neurol 2010;17:8–16.
4. Hammers-Berggren S, Hansen K, Lebech AM, Karls-son M. Borrelia burgdorferi-specific intrathecal anti-body production in neuroborreliosis: a follow-up study. Neurology 1993;43:169–75.
5. Kruger H, Reuss K, Pulz M, Rohrbach E, Pflughaupt KW, Martin R, et al. Meningoradiculitis and encephalomyelitis due to Borrelia burgdorferi: a fol-low-up study of 72 patients over 27 years. J Neurol 1989;236:322–8.
6. Ljøstad U, Mygland A. CSF B-lymphocyte chemoat-tractant (CXCL13) in the early diagnosis of acute Lyme neuroborreliosis. J Neurol 2008;255:732–7. 7. Rupprecht TA, Pfister HW, Angele B, Kastenbauer S,
Wilske B, Koedel U. The chemokine CXCL13 (BLC): a putative diagnostic marker for neuroborreliosis. Neurology 2005;65:448–50.
8. Sillanp€a€a H, Skogman BH, Sarvas H, Sepp€al€a IJ, Lahdenne P. Cerebrospinal fluid chemokine CXCL13 in the diagnosis of neuroborreliosis in children. Scand J Infect Dis 2013;45:526–30.
9. Bremell D, Mattsson N, Edsbagge M, Blennow K, Andreasson U, Wikkels€o C, et al. Cerebrospinal fluid CXCL13 in Lyme neuroborreliosis and asymptomatic HIV infection. BMC Neurology 2013;13:2.
10. Schmidt C, Plate A, Angele B, Pfister HW, Wick M, Koedel U, et al. A prospective study on the role of CXCL13 in Lyme neuroborreliosis. Neurology 2011;76:1051–8.
11. van Burgel ND, Bakels F, Kroes AC, van Dam AP. Discriminating Lyme neuroborreliosis from other neu-roinflammatory diseases by levels of CXCL13 in cere-brospinal fluid. J Clin Microbiol 2011;49:2027–30. 12. Hyt€onen J, Kortela E, Waris M, Puustinen J, Salo J.
Oksi J (2014) CXCL13 and neopterin concentrations in cerebrospinal fluid of patients with Lyme neurobor-reliosis and other diseases that cause neuroinflamma-tion. J Neuroinflam 2014;11:103.
13. Senel M, Rupprecht TA, Tumani H, Pfister HW, Ludolph AC, Brettschneider J. The chemokine CXCL13 in acute neuroborreliosis. J Neurol Neuro-surg Psychiatry 2010;81:929–33.
14. Tjernberg I, Henningsson AJ, Eliasson I, Forsberg P, Ernerudh J. Diagnostic performance of cerebrospinal fluid chemokine CXCL13 and antibodies to the
C6-peptide in Lyme neuroborreliosis. J Infect 2011;62:149–58.
15. Cerar T, Ogrinc K, Lotric-Furlan S, Kobal J, Levic-nik-Stezinar S, Strle F, et al. Diagnostic value of cytokines and chemokines in Lyme neuroborreliosis. Clin Vaccine Immunol 2013;20:1578–84.
16. Henningsson AJ, Christiansson M, Tjernberg I, L€ofgren S, Matussek A. Laboratory diagnosis of Lyme neuroborreliosis: a comparison of three CSF anti-Borrelia antibody assays. Eur J Clin Microbiol Infect Dis 2014;33:797–803.
17. Lash GE, Scaife PJ, Innes BA, Otun HA, Robson SC, Searle RF, et al. Comparison of three multiplex cyto-kine analysis systems: Luminex, SearchLightTM
and FAST QuantÒ. J Immunol Methods 2006;309:205–8.
18. Dossus L, Becker S, Achaintre D, Kaaks R, Rinaldi S. Validity of multiplex-based assays for cytokine measurements in serum and plasma from “non-dis-eased” subjects: comparison with ELISA. J Immunol Methods 2009;350:125–32.
19. de Jager W, Prakken BJ, Bijlsma JW, Kuis W, Rijkers GT. Improved multiplex immunoassay performance in human plasma and synovial fluid following removal of interfering heterophilic antibodies. J Immunol Methods 2005;300:124–35.
20. Wutte N, Berghold A, L€offler S, Zenz W, Daghofer E, Krainberger I, et al. CXCL13 chemokine in pedi-atric and adult neuroborreliosis. Acta Neurol Scand 2011;124:321–8.
