Background
Alzheimer’s disease (AD) is the most common form of age-related dementia. It affects over 46 million individuals worldwide, and the incidence is estimated to double every 20 years according to a 2017 Alzheimer’s Association report [
1]. Despite the improved diagnostic tools available to clinicians, including cerebrospinal fluid (CSF) biomarkers [
2], a definite diagnosis of AD can still only be made on the basis of a neuropathological assessment [
3]. Importantly, more efficient biomarkers, preferably in plasma owing to the less invasive sampling procedure, are needed to assess the risk of developing AD and to identify patients who will convert from mild cognitive impairment (MCI) to AD. Ultimately, biomarkers per se should offer a window to the pathological processes coupled to the development and progression of AD [
4] and clearly distinguish between these processes and those related to normal aging. Aging alone is one of the major risk factors for AD [
5], but little is known about the molecular processes that distinguish healthy aging from pathological aging [
6]. In addition to accumulation of amyloid-β (Aβ) peptides and hyperphosphorylated tau, other age-related alterations such as processes linked to inflammation are believed to play an important role in AD pathogenesis [
7,
8]. Furthermore, the presence of the apolipoprotein E (
APOE) ε4 allele, the strongest genetic risk factor for sporadic AD [
9], may affect the course of aging in individuals carrying this AD risk allele [
10]. Owing to the heterogeneity in the age at onset of sporadic AD, future efforts need to address whether a different or additional set of biomarkers would be more appropriate for patients with AD developing disease at a younger or older age.
Kallikrein 6 (KLK6) is a protein highly expressed in the brain and spinal cord, with levels in body fluids increasing with age [
11‐
13]. This age association was proposed to be disrupted in AD [
14]. These initial findings led to the assessment of KLK6 as a candidate biomarker for AD for which plasma, CSF and tissue levels of human KLK6 were analysed. Diamandis and colleagues pioneered the field of KLK6 quantification and developed an immunofluorometric assay to assess human KLK6 levels [
15‐
17]. Menendez-Gonzalez and co-authors reported a decrease in plasma KLK6 in subjects with AD compared with age-matched control subjects [
13] but no significant alteration between patients with AD and patients with MCI [
18]. Researchers in another study reported significantly decreased CSF KLK6 levels in patients with synucleinopathies, but not in patients with AD [
19]. Hence, consensus remains to be reached on the potential usefulness of KLK6 fluid levels as a biomarker for AD.
In this study we analysed KLK6 levels in plasma and CSF from subjects included in two separate cohorts including patients with AD and control subjects using a previously developed and validated quantification method. We hypothesised that the discrepancy in reported KLK6 levels results from the use of different and not well-standardised quantification methods, as well as from possible age differences in previously examined cohorts. To address these issues, we used a previously developed and validated quantification method to quantify CSF and plasma levels of KLK6 in a cross-sectional study of older patients with AD with a more advanced disease stage and a longitudinally followed cohort including patients with amnestic MCI and patients with sporadic AD with disease onset before and after 65 years of age. In addition, we analysed potential associations between KLK6 levels; age; APOE genotype; total apoE level (assessed in the cross-sectional setting only); and the AD biomarkers amyloid-β 1–42 (Aβ42), amyloid-β 1–40 (Aβ40), total tau (t-tau) and phosphorylated tau (p-tau).
Discussion
In the present study we aimed to analyse the potential of the age-related protease KLK6 as a biomarker for AD. With a comprehensive study design we have used the same methodology to investigate two separate cohorts—a cross-sectional and a longitudinal study—for KLK6 levels in both plasma and CSF. As expected, we found that the levels of CSF KLK6 were positively related to age in control individuals, validating previous data suggesting CSF levels of KLK6 as a marker for healthy aging [
14]. The results derived from our analyses endorse two previous reports, one by Menendez-Gonzalez [
18] and one by Wennström [
19] and colleagues, showing that CSF levels of KLK6 are not useful as a biomarker for AD. We further show that CSF KLK6 levels were not useful for identifying patients with MCI transitioning into an AD diagnosis over the 24-month period investigated. Importantly, our study results propose increased levels of plasma KLK6 in patients with AD with a more advanced disease stage (cohort 1). Increased plasma levels of KLK6 in these individuals also correlated with lower MMSE scores, supporting the notion that an increase in plasma KLK6 levels might be associated with cognitive decline. This statement was further supported by the lack of an age-associated increase in plasma KLK6 in the control group, corroborating our reasoning that increased plasma levels of KLK6 may indicate a pathological aging process.
Importantly, under normal aging there was no correlation between plasma KLK6 levels and age in either cohort. In fact, the regression slopes between age and the plasma KLK6 data in control subjects versus patient groups were different. Hence, despite the significant age difference between the control subjects and patients with AD in group 1, we were unable to adjust the outcome of our analysis for age. Instead, enabled by the similar regression slopes between age and plasma KLK6 in the AD patient groups in both cohorts, which differed significantly in age (median age of 78 years in cohort 1 versus 64 years in cohort 2; p < 0.0001), we found that an age-corrected comparison of plasma KLK6 levels in the AD groups of the two cohorts yielded significantly higher plasma levels in the patients with AD included in cohort 1. The latter AD group differed significantly from the AD group of cohort 2 in regard to disease severity (p < 0.0001). These results support an assumption that plasma KLK6 levels are associated with AD disease severity.
The identification of significant correlations between CSF levels of KLK6 and levels of the conventional CSF AD biomarkers in both cohorts, although in different directions for Aβ42, underlines a potential involvement of KLK6 in the pathological cascade of events leading to AD. Levels of KLK6 in the CSF from subjects in both cohorts showed a strong and significant correlation with CSF levels of t-tau and p-tau. The conflicting results of either positive or negative associations between CSF KLK6 levels and Aβ42 levels between cohort 1 and cohort 2 further suggest a potential involvement of KLK6 in AD pathological processes at different stages of disease progression, with patients with AD in cohort 1 being both older and more advanced in their cognitive dysfunction than the patients with AD in cohort 2. Whether the levels of KLK6 per se, the composition of total KLK6 and the pool of active KLK6 are of importance for AD-related pathways needs to be determined in future studies.
KLK6 is generated as a pre-pro-enzyme that becomes active through cleavage of the pre- and pro-peptides by matrix metalloproteinases and proteases of the thrombostasis axis, both of which co-exist with KLK6 in brain tissue [
28,
29]. In CSF, KLK6 was earlier found to be mostly of the pro-form [
30], but the specific activity of this form needs to be elucidated. Our analyses in the present study provided only the total KLK6 levels in plasma and CSF without regard to activity, specific pro-forms or KLK6 isoforms. Hence, future studies need to address whether the balance between the inactive and active forms of KLK6 is altered during the development of AD. Three different KLK6 isoforms have been described [
31], generated as a result of alternative splicing. The classic KLK6 comprises the full-length pre-pro-enzyme and is predominant in nervous tissue, whereas the other two isoforms are truncated proteins. It is noteworthy that the alternate isoforms may constitute 10–20% of total KLK6. With our detection method, we used monoclonal antibodies generated against full-length recombinant human KLK6 protein, so we cannot rule out the simultaneous detection of alternative isoforms. Whether the truncated isoforms are functional proteins, and if so, whether they have activity similar to that of the full-length KLK6, are yet to be investigated.
We cannot properly explain the involvement of
APOE ε4 in our results. However, a recent study by Tamboli and colleagues showed that a secreted serine protease (other than thrombin and cathepsin G) can proteolytically yield fragmented apoE and that the activity of this protease can be inhibited by the astrocyte-secreted serine protease inhibitor α
1-antichymotrypsin (ACT) [
32]. The exact identification of this protease remains to be determined, although speculating that this protease may be KLK6 is not too far-fetched. It was previously found that approximately 5% of the active KLK6 in human milk and ascites fluid was stably bound to ACT, whereas KLK6 in serum and CSF was found to be free and in an uncomplexed form [
30]. More recently, however, it was found that the serine protease α
1-antitrypsin (AAT) is the main inhibitor of KLK6 in biological fluids [
33]. Both of these described findings are of relevance when considering the previously reported increase in plasma KLK6 and CSF levels of both ACT and AAT in patients with AD [
34]. Importantly, the same study further showed that patients with dementia with Lewy bodies (DLB) also exhibited increased levels of CSF but not plasma ACT and AAT. That patients with DLB in a later study were found to have significantly lower CSF KLK6 levels than both patients with AD and control subjects [
19] unfortunately complicates a conclusion. Importantly, the relevance of the activity of the different pools of KLK6 in the periphery versus the central nervous system is unknown, and it may be of importance if speculating that KLK6 is an apoE-cleaving enzyme. We found in both cohorts that the two pools of KLK6 were not correlated (data not shown), which could have resulted from the fact that plasma KLK6 and CSF KLK6 have different sources of production [
11]. To our knowledge, we are the first to report a correlation between
APOE ε4 and KLK6 levels, which was particularly evident in the CSF. The biological significance of these findings is currently being addressed in our laboratory.
Last, with the opportunity provided through the longitudinal nature of cohort 2, we were able to assess whether plasma and CSF KLK6 levels increase over time (24 months) in (1) patients with MCI that was stable over 2 years or in those converting to AD and (2) patients with AD. Although CSF levels of KLK6 were altered between time points, they followed a similar trend in the investigated patient groups and were not related to APOE ε4 status or cognition. Hence, alterations in CSF KLK6 levels within a 24-month follow-up period could not differentiate patients with AD from amnestic MCI and MCI converters in the age group 60–65 years.