Neuroinflammation in Lyme neuroborreliosis affects amyloid metabolism

The trans-membranous protein amyloid precursor protein (APP) has been intensely studied in Alzheimer's disease (AD), since it is the source of β-amyloid (Aβ) peptides, recognized as key-components in AD pathophysiology [1]. Although ubiquitously expressed, the physiological role of APP in the adult organism remains largely unknown. APP may undergo non-amyloidogenic cleavage at the α-site, which inhibits formation of Aβ and releases an extracellular soluble α-sAPP fragment. Alternatively, APP is processed by combined cleavages by β-secretase and γ-secretase, releasing Aβ and β-sAPP. Aβ peptides vary in length due to variability in the γ-secretase cleavage site. Although CSF levels of α-sAPP and β-sAPP generally correlate tightly [2], it is not known how these pathways are orchestrated in vivo. CSF levels of α-sAPP and β-sAPP are reduced in MS and cerebral systemic lupus erythematosus SLE [3], and even lower levels are seen in HIV patients with cerebral engagement [4].

Lyme neuroborreliosis (LNB) is caused by a central nervous system (CNS) infection by the tick-borne spirochete Borrelia burgdorferi. LNB is often manifested by cranial nerve engagement, and common clinical findings are facial nerve palsy and radiculitic pain [5, 6]. Common laboratory findings are increased albumin ratio, indicating impaired blood-brain barrier function, and CSF monocytosis. In this study, we investigated CSF markers of amyloid metabolism and neural cell damage in LNB, to elucidate the influence of neuroinflammation on amyloid metabolism.

The project contained two clinical studies. The first was a cross-sectional study in patients with acute facial palsy caused by either LNB or idiopathic Bell's palsy. The second was a longitudinal prospective cohort-study, where LNB patients were followed with successive lumbar punctures to investigate time-dependent biomarker dynamics.

In the cross-sectional study all groups were comparable in age. The only exception was that Bell's palsy patients were slightly younger than the controls (P = .023, Table 1). LNB patients had longer history of neurological symptoms before the time of lumbar puncture than Bell's palsy patients (Table 1).

In the longitudinal study, LNB patients were younger than the controls (P = .031, Table 1). The median duration of neurologic symptoms was 7 days longer than in the cross-sectional study (21 days compared with 28 days). At follow-up, all LNB patients had improved in their clinical symptoms and their inflammatory reactions had diminished, with decreased CSF monocytic cell counts (Table 1).

Amyloid metabolism

In the cross-sectional study, LNB patients had lower α-sAPP and β-sAPP than the other groups (Figure 1), but there were no differences in Aβ38, Aβ40 or Aβ42 (Table 2). α-sAPP and β-sAPP correlated to most Aβ peptides in Bell's palsy patients (R = .47-.60, P ≤ .002) and controls (R = .42-.55, P < .05; the only exception was α-sAPP and Aβ38 in controls, where there was a trend towards significance, R = .40, P = .065), but not in LNB patients (P > .05).

Group	Aβ38 (ng/L)	Aβ40 (ng/L)	Aβ42 (ng/L)	T-tau (ng/L)	P-tau (ng/L)	NFLb
Controls	779 (323-1767)	6040 (3372-9885)	531 (240-992)	158 (74-553)	36 (20-99)	2/22
LNB	570 (270-2106)	5436 (3456-9465)	434 (198-978)	129 (75-335)	31 (22-71)	10/19d
Bell's palsy	694 (108-1615)	5819 (1316-9780)	440 (71-1010)	155 (75-420)	44c (15-80)	6/42e

CSF, cerebrospinal fluid; LNB, Lyme neuroborreliosis; ^adata presented as median (range); ^bnumber of patients with CSF concentrations above the detection limit 125 ng/L; ^cP = .02 vs LNB; ^dP = .002 vs controls; ^eP = .002 vs LNB

In the longitudinal study, LNB patients had lower baseline levels of α-sAPP and β-sAPP (Figure 2) and Aβ peptides than controls (Table 3). α-sAPP and β-sAPP increased after treatment (Figure 2), while Aβ levels were unaffected (Table 3). Conversely to what was seen in the cross-sectional study, α-sAPP and β-sAPP correlated to all Aβ peptides in LNB in the longitudinal study (R = .71-.98, P < .001), but not in controls (P > .05).

Group	Aβ38 (ng/L)	Aβ40 (ng/L)	Aβ42 (ng/L)	T-tau (ng/L)	P-tau (ng/L)
Controls	891 (279-1715)	6676 (3450-9672)	672 (293-1193)	349 (171-552)	47 (23-69)
LNB baseline	561b (254-1609)	4840c (2726-8857)	378d (211-997)	229e (171-526)	30c (16-60)
LNB follow-up	510f (234-1562)	4310g (2449-8690)	310h (171-894)	235i (171-532)	33j,k (16-74)

CSF, cerebrospinal fluid; LNB, Lyme neuroborreliosis; ^adata presented as median (range); ^bP = .015 vs controls; ^cP = .006 vs controls; ^dP = .013 vs controls; ^eP = .01 vs controls; ^fP = .021 vs controls; ^gP = .005 vs controls; ^hP = .004 vs controls; ⁱP = .031 vs controls; ^jP = .041 vs controls; ^kP = .004 vs LNB baseline.

Aβ38 correlated to Aβ40 (R = .91-.97, P < .001) and Aβ42 (R = .74-.92, P < .001), and Aβ40 correlated to Aβ42 (R = .69-.93, P ≤ .001), in all groups. These correlations were expected, since none of the subjects had any known amyloid plaque pathology, which could have entrapped Aβ40 and Aβ42 in the brain parenchyma. In the absence of plaques, these Aβ peptides are likely to diffuse at similar rates to the CSF. α-sAPP and β-sAPP correlated in all groups (R = .88-.95, P < .001), indicating that the pathways producing these peptides are tightly synchronized in vivo.

We found signs of altered brain amyloid metabolism in LNB. CSF levels of the soluble APP-fragments α-sAPP and β-sAPP were decreased in the acute stage of the disease, and increased following treatment with antibiotics. Patients with Bell's palsy had no inflammatory activity in CSF and higher CSF α-sAPP and β-sAPP concentrations than LNB patients. These findings add to previous evidence of altered amyloid metabolism in primary inflammatory CNS diseases such as MS, cerebral SLE and HIV with cerebral engagement. In MS, α-sAPP and β-sAPP levels were significantly lower in patients with ongoing or recent disease exacerbation compared to patients in stable remission [3]. In light of these findings, we suggest that CSF levels of α-sAPP and β-sAPP may be generally reduced in diseases with pronounced neuroinflammation. This should be compared to what is seen in AD. Although AD is primarily a neurodegenerative condition, it also has strong inflammatory components [12]. The dense core amyloid plaques in AD brain tissue are surrounded by activated microglia [13]. However, although these cells may phagocytose Aβ it is yet unknown precisely how the inflammatory activity affects amyloid metabolism in AD. Net effects on sAPP proteins differ from what is seen in MS, cerebral SLE and CNS infections, since AD patients have unaltered or even elevated CSF levels of α-sAPP and β-sAPP [2, 14, 15]. Regarding Bell's palsy, it is an important differential diagnosis to LNB. Although the primary aim of our study was not to assess the diagnostic performance of the amyloid markers, we think it is interesting to note the striking differences between facial palsy caused by LNB and Bell's palsy.

Although low α-sAPP and β-sAPP levels appear to be consistent findings in diseases with pronounced neuroinflammation, the picture becomes more complex when we turn to Aβ peptides. All subjects in the cross-sectional study in the current project had similar Aβ, but in the longitudinal study LNB patients had lower Aβ levels than controls. Note that the LNB patients differed in neurological symptoms between these two parts of the study. Other studies have found decreased Aβ42 in MS, bacterial meningitis and HIV with cerebral engagement [16] but not in viral meningitis [3, 17]. Thus, the etiology, duration or severity of the neuroinflammation might affect its influence on Aβ metabolism. In AD, CSF levels of Aβ42 are decreased, which is often explained by accumulation of Aβ42 in plaques [18]. However, low Aβ42 levels may also be seen in Creutzfeld-Jakob's disease, where the abundance of plaques is low [19]. Thus, other mechanisms than plaque accumulation could contribute to low CSF Aβ42 levels in neurodegenerative diseases.

In principle, reduced levels of CSF sAPP could be caused by 1) decreased expression of APP, 2) decreased processing of APP into sAPP, 3) decreased clearance of sAPP into CSF, or 4) increased degradation of sAPP. Presently, it is not possible to draw definite conclusions on which of these mechanisms that operate in neuroinflammation. In cases where no clear reductions in CSF Aβ peptides are seen, such as facial palsy LNB, a decreased APP expression is unlikely. Likewise, altered APP processing with lower β-secretase activity would cause lower levels of both β-sAPP and Aβ peptides. Regarding the third possibility, decreased clearance of sAPP into CSF, it is known that Aβ42 under certain conditions might be trapped in interstitial drainage pathways [20]. It is not known if this is also the case for sAPP. Finally, increased degradation of sAPP in neuroinflammation remains a vital possibility. Little definite data exist on the physiological roles of sAPP, although α-sAPP has been attributed with neuroprotective properties [21]. Particularly interesting in the context of neural damage is a recently described neurotoxic N-terminal APP-fragment, located within the sAPP-sequence, capable of inducing axonal degeneration through interaction with axonal DR6 receptors [22]. Further research should be undertaken to explore whether α-sAPP and β-sAPP share these axonal degenerative properties, or if they are metabolised into DR6-activating peptides in neuroinflammation.

In the follow-up part of the present study LNB patients improved clinically after oral doxycycline, with dramatic effects on the radiculitic pain and a slower recovery of peripheral paresis. CSF analysis with decreased CSF inflammation measured as number of monocytic cells and normalization of the blood-brain barrier measured as albumin ratio support the excellent therapeutic effect. The LNB patients were younger than the controls, which might affect the statistical analysis of some parameters. However, in the study on MS referred to earlier, CSF levels of α-sAPP and β-sAPP decreased in both patients and controls during a 10 year follow-up [3]. Thus, it is unlikely that the age difference explains the low baseline levels of sAPP in young LNB patients compared to older controls.

Regarding markers of neural damage and tau pathology, we confirm earlier findings of increased CSF NFL in some LNB patients, which indicates acute damage to large myelinated axons [23]. White matter lesions have also been described previously in LNB [24]. However, the CSF NFL levels were low compared to other CNS diseases such as herpes encephalitis [25]. NFL correlated to albumin ratio, likely since both axons and the blood-brain barrier are more affected in severe cases. For CSF T-tau, considered a marker of damage to cortical axons [26], we found no difference between LNB and Bell's palsy. This suggests that the neural damage in LNB is not located to cortical structures. T-tau levels were even lower in LNB patients than in controls. Since CSF T-tau increases with age, this could partly be explained by the age-difference between LNB patients and controls [27]. CSF P-tau was decreased in LNB patients, and increased after treatment in the follow-up study. Thus, the acute neuroinflammation may have affected tau phosphorylation.

We report an association of cerebral amyloid metabolism with LNB, with decreased CSF levels of soluble APP-fragments in the acute disease stage. Besides possible implications for diagnosis of LNB, these findings strengthen the case for a general involvement of amyloid in neuroinflammation and suggest that a physiological role of APP could be to participate in inflammatory pathways. We also confirm previous findings of elevated CSF NFL in LNB, suggesting damage to myelinated axons. Further research on amyloid metabolism in neuroinflammation will likely shed more light on the functions of APP in the adult human nervous system.