Neuroinflammatory mechanisms contribute to the pathology of both PD and APD, which comprise microglia activation, astrocytosis and lymphocyte infiltration as observed in post mortem obtained brain tissue(10). Moreover, polymorphisms in genes associated with inflammation, such as LRRK2, S100B and NURR1, increase the risk for PD(23–26). Consequently, the expression levels of inflammatory proteins in CSF might translate into biomarkers for diagnosis and/or prognosis of PD and APD.
From a panel of 92 inflammatory proteins, we identified CCL28 as a biomarker of neuroinflammation that differentiated only PD from controls, which showed a different expression trend (lower expression levels in controls). We also identified 8 biomarkers that differentiated only MSA from controls and 5 biomarkers that differentiated both PD and MSA from controls. Interestingly, DNER and β-NGF could significantly differentiate MSA from PD and controls.
CCL28 (Mucosae-associated epithelial chemokine; MEC) is a chemokine constitutively expressed in mucosal tissue and moderately expressed in small intestine, kidney and brain –in neurons rather than glia cells. CCL28 bridges the innate and adaptative immune response; the C-terminus has antimicrobial activity and the N-terminus mediates lymphocyte migration. Signaling of CCL28 via CCR10 drives the homing of T and B lymphocytes, and via CCR3 the migration of eosinophils(27). CCL28 was the only biomarker that was up-regulated in PD patients compared to controls. In a mouse model of epilepsy, down-regulated expression of CCL28 in brain tissue was associated with neuronal loss(28). Another study performed in human PD brain tissue detected lower levels in PD patients compared to controls(29). This apparent inconsistency with our results might be because they analyzed brain tissue and thus, neuronal expression of CCL28, while we measured levels in CSF detecting expression of other cell types and also systemic inflammation. The elevated levels in CSF might be in line with the idea that viral and microbial infections, as well as altered gut-microbiota, increase the risk of PD or they may even be an early trigger of the disease(30–32). Another possible reason of elevated levels of CCL28 in CSF might be its release from degenerating neurons.
DNER was observed at higher levels in the CSF of PD patients as compared to MSA patients. DNER is highly expressed in the substantia nigra and, just as Parkin protein, is an activator of the NOTCH1 pathway, which has a role in neuronal and glial cell differentiation and neuroprotection(33, 34). A down-regulation of DNER protein in MSA might highlight the loss of neuroprotection and thus, the higher disease severity. Interestingly, DNER was also down-regulated in MSA compared to PD to a similar degree as it was described in a previous publication that used the same PEA inflammatory panel for the discovery of biomarkers for PD and APD in two independent cohorts(11).
β-NGF was also significantly down-regulated in MSA compared to PD patients. β-NGF is a trophic factor for sympathetic and sensory fibers found in the peripheral nervous system and in the central nervous system in cholinergic neurons projecting to the cerebral cortex and hippocampus. β-NGF has neuroprotective effects in cholinergic neurons(35–37). Therefore, the reduction in β-NGF levels may indicate a more advanced neuronal cell loss in MSA. β-NGF was not identified as a potential biomarker to differentiate PD from MSA in the above-mentioned double cohort study(11), which highlights the importance of independent biomarker discovery studies in different cohorts and laboratories.
In contrast to the previously published study using PEA for biomarker identification of PD and APD(11), we did not find significantly different levels of FGF–5, VEGF-A and FGF–19 in PD versus MSA, despite observing a similar trend, i.e. higher expression levels in PD than in MSA. These differences between our and this previous study could possibly be explained by our relatively small number of MSA patients in comparison to PD patients (3-fold difference). A more likely explanation, however, could be that our cohort had unique characteristics, namely all patients who were included in our study had a clear evidence of a form of parkinsonism, but with uncertainty of the specific diagnosis at baseline. Moreover, the diagnosis of the patients has been re-evaluated after three years of follow-up and updated again twelve years after inclusion according to the revised clinical criteria. This is unlike the previous study, in which, despite including a few uncertain cases in one of the two cohorts, most patients had a clear diagnosis at CSF withdrawal. Moreover, they followed their patients for no more than 5 years(11, 38, 39).
ROC analysis showed that the combination of DNER and β-NGF do not yield a higher AUC than each of them individually. However, they could be valuable in a bigger panel including biomarkers of different biological processes, such as NFL. In our cohort of study, NFL alone yielded the same diagnostic accuracy as NFL in combination with DNER and β-NGF. The poor added value of these two inflammatory proteins might be caused by their lower expression levels and the remodulation of the immune system at older age, losing the ability to fine-tune inflammation(40). Further studies need to determine the positive impact of adding these inflammatory proteins to a larger diagnostic panel to discriminate PD from MSA patients but our data suggests that such impact is likely to be minimal.
It can be hypothesized that PD (or MSA) patients with more pronounced neuroinflammation than others will have a more severe disease progression. For this reason, we correlated levels of CSF proteins at baseline with disease progression over a 3-year time-frame. Both MCP–1 and MMP–10 showed a significant positive correlation with parameters of PD progression. MCP–1 plays an important role in monocyte recruitment and propagation of inflammation. Previous studies showed that plasma levels of MCP–1 correlated with cognitive decline in patients with Alzheimer’s disease(41). Other studies in mouse models suggest that MCP–1 causes neuronal loss and that its downregulation is neuroprotective(42, 43). MMP–10 is a secreted metalloproteinase with a key role on modulation of macrophage activation and function. MMP–10 is not expressed in unchallenged tissues, but is increased in response to a variety of insults(44). Thus, a positive correlation of MMP–10 CSF levels with disease progression might indicate increased inflammation and neuronal loss.
The major strength of our study is the uniqueness of our patient cohort. Patients with diagnostic uncertainty were included in the study and their diagnosis was reevaluated after three and twelve years. Thus, our study exactly reflects the clinical situation when biomarkers are actually needed, i.e. biomarkers have diagnostic value when the diagnosis is not yet clear. Unlike our study, many biomarker studies have been performed with patients with a clear-cut diagnosis, but in such situations biomarkers will not add anything to the diagnostic work-up.
Our study also presents four limitations. First, our group of MSA patients was relatively small, which may have affected our analyses. However, the patients were very well-defined as a result of the long-term follow-up. Second, the final diagnosis was based on clinical evaluation according to international diagnostic criteria, but has not been confirmed yet by neuropathologic examination. This may have caused potential misdiagnoses, but we have reduced this risk by the very long follow-up of the patients. For most patients, a ‘silver standard’ diagnosis can be made after some three years of follow-up, when the rate of progression is known, new red flags may have appeared, and the levodopa responsiveness has been tested. Third, the disease progression is calculated based on the 3-year follow-up scores. A stronger correlation of biomarkers with prognosis might be observed with data from longer follow-up periods. Fourth, the study did not include patients with other forms of atypical parkinsonism, such as PSP, CBS and DLB, due to the small number of patients with these diagnoses included in the longitudinal study.