Study cohort

The study cohort consisted of archival samples of paired sera and CSF from patients tested for LNB using the intrathecal anti-Borrelia (IgM/IgG) antibody index (AI) test. One-hundred and four paired samples were selected from AI positive patients, indicating certain prior exposure to I. ricinus ticks, and 110 paired samples were selected from AI negative patients. As some of the archival samples only had small remaining volumes, insufficient for further testing, some of them had to be discarded. A total of 99 sera and 87 CSF samples from AI positive patients, and 101 sera and 103 CSF samples from AI negative patients, remained for inclusion in this study after duplicates and samples containing insufficient volumes had been removed (Fig. 1).

All samples were collected in the months of June to November during the years 2011–2015 and have been stored at the State Serum Institute (SSI) at − 20 °C. The serological analysis for LNB was during these years mostly centralized at SSI and the samples thus represent patients from across the whole country.

Intrathecal anti-Borrelia antibody index test

The intrathecal anti-Borrelia antibody index test had been performed using an in-house capture enzyme-linked immunosorbent assay (ELISA) that uses native purified flagellum from B. afzelii (strain DK1) as antigen [21]. The AI had been calculated using the formula: (ODcsf/ODserum)*(ODcsf − ODserum), with ODcsf and ODserum representing the optical density in CSF and sera, respectively. An AI ≥0.3 is regarded as positive.

Rickettsia serology

The sera were examined for antibodies against SFG rickettsiae using a commercially available indirect immunofluorescense assay (IFA) (IF0100G, Rickettsia IFA IgG; Focus Diagnostics, Inc., Cypress, CA, USA). Inactivated R. rickettsii (RR) was used as antigen to detect SFG rickettsiae. The samples were screened at dilutions of 1:64 (IgM) and 1:128 (IgG) and were titrated to end-point fluorescence in two-fold dilutions. A titer of 1:64 or above was considered positive for IgM. The cut-off for IgG had previously, for diagnostic purposes, been raised to 1:512, following testing of healthy Danish blood donors, to achieve a specificity of 95% and reduce cross-reactivity with other bacteria [22]. However, as this elevated cut-off likely results in a loss of sensitivity, a lower cut-off of 1:128 was used in this study. One hundred and eleven serum samples from healthy Danish blood donors served as controls for IgG and 60 blood donor samples served as controls for IgM. These donor samples provided an estimate of the background seroprevalence of anti-rickettsia antibodies in the Danish general population.

Rickettsia PCR

Statistical analysis

Statistical significance was calculated with Fisher’s exact test (two-tailed) and the two-sample Wilcoxon rank sum test using the statistical software R Studio (v1.0.153) [24]. A p-value ≤0.05 was considered to be statistically significant.

Serology

When using the routine, diagnostic cut-off of 1:512 [22], 12 patients in the AI positive group and five of the healthy blood donors exhibited anti-rickettsia antibodies. This difference was, however, also non-significant (p = 0.07).

PCR

At the genus-level, Rickettsia DNA was identified in eight of 190 (4%) archival CSF samples, with cycle threshold (Ct) values ranging between 39 and 44. Compared with the 103 negative controls, this is a non-significant difference (p = 0.054). However, five of these patients were AI positive, reflecting a significantly increased probability of detecting Rickettsia DNA in this subgroup compared with the controls (p = 0.02). Sequencing was attempted but failed on all eight samples. None of these patients displayed anti-Rickettsia antibodies in the corresponding sera. No CSF in the negative control group, regardless of biochemical findings, showed evidence of Rickettsia DNA and no Rickettsia DNA could be detected in the sera. All internal negative controls remained negative and all positive controls displayed positive reactions.

Cases

One of the patients (Case 5) had, in the months prior to hospital admission, experienced a tick bite in Sweden. No other patients had any mention of international travel, leading up to the hospital admission. As R. helvetica is also the most common tick-borne rickettsia found in Sweden, all patients had thus most likely been infected with this species or possibly, R. monacensis [25].

Serology

The high seroprevalence of IgG antibodies against SFG rickettsiae among healthy blood donors is indicative of considerable non-specific background reactivity, revealed by the lowering of our titer cut-off for this study. Nevertheless, the absence of a significant difference in the seroprevalence of antibodies between healthy blood donors and patients examined for LNB, regardless of titer cut-off and Borrelia AI, is indicative of low rates of exposure in this group of patients.

Important caveats to be emphasized are that the precise time elapsed from being bitten to being tested is not known and that the precise kinetics of the human antibody response to R. helvetica and R. monacensis have not been studied, factors that could potentially impact the current results. However, considering that adult Danish patients with LNB are often diagnosed late in the course of their disease with a median delay of 20 days from the start of neurological symptoms until first hospital contact, one can assume that enough time had passed to allow seroconversion in most of the AI positive patients exposed to SFG rickettsiae [26].

Lastly, R. rickettsii, and not R. helvetica, was used as IFA antigen in the current study. Rickettsia conorii is often used as antigen to screen for SFG rickettsiae and cross-reactivity between this species and R. helvetica has previously been demonstrated [7, 14]. As R. rickettsii is phylogenetically closely related to R. conorii and due to the generally considerable homology between SFG rickettsiae resulting in high levels of serological intra-group cross-reactivity, we expect R. helvetica to be detectable using this current assay, as has been suggested by a previous study [27]. This has been corroborated by preliminary tests on Norwegian sera from patients seropositive for R. helvetica, indicating that these antibodies are detectable using R. rickettsii as antigen and the current cut-offs (unpublished data).

PCR

Whether the detected Rickettsia DNA represents true neuroinfection or not can be discussed. All internal negative controls remained negative, making contamination less likely. Furthermore, Rickettsia DNA was not detected in the negative control group, consisting of patients tested for other neuroinfections than LNB and representing a range of varying biochemical values, reducing the chances of it being an artifact. Sequencing was attempted but failed on all eight samples, most likely due to the small amounts of bacterial DNA in the samples, as reflected in the high Ct values. However, the general lack of discerning clinical features in these cases and the apparent lack of pleocytosis and clinical signs of infection in two of the subjects necessitates cautious interpretation of these PCR findings. It is important to note that none of the patients received treatment with doxycycline (Table 3). Despite Rickettsia spp. being inherently resistant to beta-lactam antibiotics [28] and these patients thus never receiving any effective treatment against their putative Rickettsia infection, all patients with spinal fluid pleocytosis and available follow-up data recovered from the acute phase of their infections, with two of them experiencing a full recovery 1 month post treatment. One of the previously described cases of putative R. helvetica neuroinfection had concomitant PCR-verified herpes simplex meningitis and it is, likewise, unknown what role R. helvetica played in this infection and how the patient would have fared had he not received treatment with doxycycline [12]. Furthermore, it is important to emphasize that detection of bacterial DNA is not equivalent to detecting the organism itself, and the present study does not establish any causal relationship between these findings and the clinical features of the patients.

Interestingly, none of the patients exhibiting DNA in their CSF had detectable antibodies against SFG rickettsiae. Possible explanations could be that the patients were tested prior to the development of anti-rickettsia antibodies, a lack of sensitivity in the IFA assay or sample contamination with rickettsial DNA. It is possible that the lack of antibodies facilitated the detection of Rickettsia DNA by allowing the bacteria to survive and spread, as suggested by a recent study [29].

Strengths and limitations

The study is limited by its retrospective nature and its use of frozen archival specimens of variable volume, ill-suited for culture and lacking convalescent samples for serology. However, the use of archival material also allowed the inclusion of a greater number of patients with a confirmed anti-Borrelia AI, ascertaining prior tick exposure, than would have been possible in a reasonable time frame following a prospective study design. It is important to remember that all samples had initially been sent to a serological laboratory and have thus, not been handled in a clean working environment appropriate for PCR analysis, why potential sample contamination cannot be ruled out. This also applies to the negative CSF controls, many of which had previously been tested for HSV/VZV using an intrathecal antibody index test. However, contamination seems unlikely as we have never amplified or cultured rickettsiae in our serological laboratory and have only used commercial IFA kits with preformed slides, on which the antigen is already fixed, for serological analysis.

The effect of prolonged storage in − 20 °C freezers and multiple cycles of thawing-freezing on the quality and yield of potential bacterial DNA can also be discussed. The former was not considered a major problem as R. helvetica has previously been detected in CSF (using both culture and PCR) after having been stored for over a year in a − 20 °C freezers [11]. Regarding the latter, prior studies on Treponema pallidum have shown only negligible effects of multiple freezing-thawing cycles on the quality and yield of spirochetal DNA in CSF, and this is likely the case for other forms of bacterial DNA as well [30].