Abnormalities of DYRK1A-Cytoskeleton Complexes in the Blood Cells as Potential Biomarkers of Alzheimer’s Disease

Brain tissue

Frozen brain tissue was obtained from the Brain and Tissue Bank for Developmental Disabilities and Aging of the Institute for Basic Research (IBR Brain Bank). Each group (Control, DS, and AD) consisted of six subjects representing both sexes and multiple ethnicities. Diagnosis of DS cases was genetically confirmed. All AD patients were clinically classified as severely demented and postmortem diagnosed as AD. Procedures involving human brain tissue were performed in accordance with the Declaration of Helsinki. Experimental protocols were approved by the Institutional Review Board of the NYS Institute for Basic Research in Developmental Disabilities.

Lymphoblastoid cell lines

EBV-immortalized LCLs of control, AD, LBD, and FTLD were obtained from the Coriell Cell Depositories (Camden, NJ). One control, all six DS, and four Fragile X cell lines were established in the Department of Human Genetics of our Institute. The full list of LCLs used in the study is given in Table 1.

In addition, frozen cell pellets of three FAD and three SAD LCLs, each with age-matched control, were prepared at the Laboratory of Preclinical Testing of Higher Standard of the Nencki Institute from individuals enrolled at the Alzheimer’s Disease Division, Neurology Clinic of the Central Clinical Hospital of the Ministry of Interior (MSWiA) in Warsaw, Poland. Diagnostics, blood collection, EBV-immortalization of lymphocytes and their culture were performed as described earlier [23]. AD diagnosis was based on the NIA-AA criteria [24]. Three subjects were diagnosed as FAD based on the presence of mutation in PS1 gene as demonstrated either by RT-PCR or genomic DNA analysis. All control subjects were non-demented with no clinical signs of neurological and/or psychiatric diseases, and with normal cognitive and neuropsychological test results. Characteristics of the donors of those cell lines are given in Table 2.

Lymphocyte cultures were maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum and a mixture of penicillin and streptomycin. All culture media and supplements were purchased from Mediatech, Inc. (Manassas, VA). Cells after harvest were washed three times in cold PBS and stored as dry pellet at –20°C until use.

Peripheral blood mononuclear cells

PBMC were isolated using a standard Ficoll-Paque protocol from blood collected from patients with SAD or with MCI due to AD at prodromal AD phase who were enrolled at the Alzheimer’s Disease Clinic at MSWiA in Warsaw, Poland. The diagnosis was performed according to NIA-AA criteria [24, 25] and involved detailed medical, neuropsychological, neurological, and psychiatric examinations. The AD diagnosis was supported by cerebrospinal fluid (CSF) ELISA assays (Innogenetics, Gent, Belgium) of standard markers: Aβ₄₂, total tau protein (t-tau), and phosphorylated form of tau (p-tau, INNOTEST^®PHOSPHO-TAU 181p). Cut-off values for CSF markers were as described previously [26]. Age- and sex-matched control individuals for the study were recruited from non-demented individuals presented at the Alzheimer’s Disease Clinic at MSWiA. The exclusion criteria for the control group included family history of AD, the potential AD risk factor, and other serious medical conditions that might influence cognition, such as depressive symptoms, previous diagnosed psychiatric disorders, and alcohol or drug abuse [26]. Postmortem blood sample of a 65-year-old DS patient was also included. White blood cells from this sample were isolated by first lysis of red blood cells in ACK lysis buffer (0.15 M NH₄Cl, 10 mM KHCO₃, and 0.1 mM EDTA) according to the protocol described by the manufacturer (ThermoFisher Scientific).

Ethic statement

All study protocols were approved by the Ethics Committee for studies on Human Subjects at the MSWiA Hospital in Warsaw (Agreement no. 89/2014) according to the National and European Union legislation, and the Code of Ethical Principles for Medical Research Involving Human Subjects of the World Medical Association. Peripheral blood and CSF samples were collected from all subjects after obtaining written informed consent for participation in the study.

Antibodies

Rabbit polyclonal anti-DYRK1A (H143, C-terminus) and goat polyclonal anti-HSP90 antibodies were obtained from Santa Cruz Biotechnology (Dallas, TX). The R420 rabbit polyclonal antibody targeting isoform 2 of DYRK1A was developed and produced at IBR [27]. For detection of DYRK1A on immunoblots, mouse monoclonal antibody 8D9 (Millipore-Sigma, Burlington, MA) was used [28, 29]. Mouse monoclonal antibodies (anti-β-actin clone AC-15, anti-α-tubulin clone B-5-1-2, and anti-neurofilament 200 clone N52) were from Sigma-Aldrich (St. Louis, MO). Anti-β-actin rabbit polyclonal antibody was purchased from Cytoskeleton, Inc (Denver, CO). Rabbit polyclonal anti-α-tubulin antibody was obtained from ThermoFisher Scientific.

Preparation of brain and lymphocyte lysates

Frozen dissected frontal cortex tissue was mortar ground in liquid nitrogen. Pools were made by mixing 0.1 g of powdered tissue from each subject. 100 mg of pooled tissue was lysed in 1 ml of ice-cold RIPA buffer (1×PBS, pH 7.4, 1% of NP-40, 0.1% of SDS, and 0.5% sodium deoxycholate) containing Complete protease inhibitors (Roche, Indianapolis, IN) and 200 nM of okadaic acid. After short-pulse sonication, lysates were clarified by microfuge centrifugation at 14,000 rpm for 10 min and protein concentrations were determined by bicinchoninic acid assay (BCA) (ThermoFisher Scientific). Preparation of lymphocytes lysates was essentially the same as brain lysates: 2×10⁷ frozen cells were thawed on ice and lysed in 1 ml of RIPA buffer supplemented with OA and Complete protease inhibitors, followed by short-pulse sonication. Lysates were clarified by centrifugation and protein concentration determined by BCA assay.

Immunoprecipitation (IP) and immunoblotting

To reduce the non-specific binding, lysates were pre-cleared by incubation with Invitrogen DYNAL Dynabeads Protein G (ThermoFisher Scientific). For IP of brain tissue and lymphocytes, total proteins of 1 mg and 500μg, respectively, were mixed with Dynabeads primary antibody conjugates, which were prepared by binding 2–4μg of the antibody to 50μl of beads suspension. Antibodies used for IP were either H-143 or R420 for DYRK1A, AC-15 for β-actin, and B-5-1-2 for α-tubulin. Samples were incubated at 4°C with rotation overnight. Using the magnet, the complexes were washed three times with 0.5 ml of RIPA buffer. After the final wash, samples were eluted by boiling for 5 min in 50μl of 1×tricine sample buffer (Bio-Rad, Hercules, CA). Samples were electrophoresed in 8% tricine SDS-polyacrylamide gels and semi-dry transferred onto PVDF membranes (Bio-Rad). The membranes were blocked with 5% nonfat milk or Pierce protein-free blocking buffer (ThermoFisher Scientific) and incubated overnight at 4°C with appropriate dilution of primary antibody. The immunoreactive bands were visualized using anti-IgG alkaline phosphatase-conjugated secondary antibody (Sigma-Aldrich) and CDP-Star chemiluminescence reagent (New England Biolabs, Ipswitch, MA). After chemiluminescence detection, blots were further developed by color reaction using BCIP/NBT (5-bromo-4-chloro-3-indolyl-phosphate/nitro blue tetrazolium) substrate (Sigma-Aldrich).

Fractionation of lymphocytes: G-/F-actin spin-down assay

Lysates prepared as described above were subjected to high-speed centrifugation: 150μg of total protein, usually between 100–150μl of the lysate, were spun at 100,000×g (50,000×rpm, TLC 100 rotor) for 1 h at 4°C in a Beckman Tabletop Ultra centrifuge. Supernatants (G-actin) and pellets (F-actin) were collected separately and placed on ice. Pellets were re-suspended in 100μl of ice-cold water with 10μM cytochalasin D, and incubated on ice for 1 h with occasional stirring by pipetting. Supernatants and pellet extracts were subsequently precipitated with 10% trichloroacetic acid (TCA) as follows: 10 min on ice, followed by centrifugation on a benchtop centrifuge for 5 min (14000 rpm), then by washing in cold acetone. Samples were left to dry overnight at 4°C. Precipitates of supernatants (S) and pellets (P) were re-suspended in 120μl and 100μl of 1×tricine sample buffer, respectively. To improve solubility of S samples short pulse sonication was applied. All samples were boiled for 10 min before loading on to 8% tricine SDS-polyacrylamide gel or NuPAGE™ 4-12% Bis-Tris gel (ThermoFisher Scientific). For better detection of differences in electrophoretic mobility of actin bands in PBMC samples (Fig. 6), samples were run on longer 8% tricine gels with extended duration (10×10 cm, 105 V, 150 min). Western blot analysis of DYRK1A, α-tubulin, and β-actin was carried out as described above. To visualize all protein bands, gels were stained with Coomassie blue. Flow chart in Fig. 1 summarizes all steps leading to obtaining G- and F-actin enriched fractions and samples for IP. Actin patterning conversion was performed by incubating RIPA lysate prepared from control subjects in the presence of 1 mM ATP and 2 mM MgCl₂ at 37°C for 15 min. Untreated mock lysate was similarly prepared except without the addition of ATP and MgCl₂. Incubated lysates were then subjected to the spin-down assay exactly as described above.

Fig.1

Fractionation of lymphocytes: Flow chart depicting steps leading to obtaining G- and F-actin enriched fractions and samples for immunoprecipitation.

Data analysis

The densitometric quantification of immunoreactive bands was carried out with 1D Scan EX 3.1 software program (Scanalytic Corp., Rockville, MD). The results were evaluated for statistical significance with the 2-tailed Student t-test for independent samples using GraphPad Prism version 4 software program (GraphPad Software, Inc., San Diego, CA). Difference with p < 0.05 was considered significant and marked.

DS and AD subjects display reduced abundance of cytoskeletal proteins co-immunoprecipitated (co-IP) with DYRK1A antibody

Brain tissue

DYRK1A is associated with cytoskeleton, as shown by co-IP of all major cytoskeletal proteins with anti-DYRK1A antibodies in the brain tissue (Fig. 2). Apparently, DYRK1A is a part of the large complex involving all major structural cytoskeletal components and other auxiliary proteins, like HSP90. DYRK1A co-IP assay of pooled frontal cortex tissue of control, DS, and AD subjects, revealed a marked reduction in the yield of this complex in pathological brains. Both DS and AD brains displayed similar patterns of reduction, which seems to be specific for DS/AD since they were not present in three unrelated mental retardation cases (data not shown). Among DS, AD, and controls, the levels of DYRK1A were highest in DS and lowest in AD, suggesting that the reduced formation of DYRK1A-cytoskeleton complexes may be linked to abnormal levels of this kinase. As demonstrated by our studies in DS [21], DYRK1A action on cytoskeleton is regulated by an enzyme-dosage sensitive mechanism. Of special attention is the electrophoretic patterning of actin bands in immunoprecipitates. In control, actin appears as a doublet of closely spaced bands while in DS/AD, only the upper band of the doublet is present. This actin patterning was another distinctive feature of DS/AD pathological brains, which emerged as a potential diagnostic marker of AD.

Fig.2

Co-IP of cytoskeletal proteins with DYRK1A. RIPA lysate, prepared from the pooled samples of frontal cortex of six individuals, was IP with anti-DYRK1A antibody (H143) and the immunoprecipitates were subsequently analyzed by immunoblotting for the cytoskeletal proteins indicated as described in the Materials and Methods.

Lymphoblastoid cell lines

Availability of LCLs established from DS and AD donors allowed us to investigate whether the diseased status confers similar cytoskeleton alterations in peripheral cell lines. All cases were carefully selected to have well-stratified representation of sporadic and familial AD and other neurodegenerative disorders. Accordingly, twelve controls and ten AD cases, of different age and gender, including SAD and FAD, were chosen. These included a pair containing an asymptomatic PS1 mutation carrier and her healthy cousin, and a pair of siblings from AD family, one of whom is affected, but at the time of sampling had no symptoms, and the second one probably an escapee from AD. LCLs from DS donors were represented by six cases. Three cases of non-AD neurodegenerative dementia: two LBD and one FTLD were also included. Neurological disorders other than neurodegenerative dementia were represented by four Fragile X cases. The full list of cases is presented in Table 1.

Analysis of LCLs started with measuring the levels of DYRK1A protein in crude lysates. As shown in Fig. 3A, which depicts results for three controls, one DS, and four AD cases, the levels of DYRK1A in all AD lysates were considerably lower compared to controls. SAD, FAD, and asymptomatic PS1 mutation carrier (PS1as), all displayed similar reduction. In DS lysate, as expected from triplication of the gene, the levels of DYRK1A were accordingly higher. We subsequently analyzed the changes of DYRK1A and cytoskeletal proteins in pooled samples of control, DS, and AD donors to eliminate an effect of single case outliers (Fig. 3E–H). The differences were also reproduced in lysates of pooled controls, DS, and AD, as shown in Fig. 3E, where DYRK1A levels compared to controls were enhanced by 39% in DS, while those in AD were reduced by 42%. Those results closely reproduce findings from the brain tissues demonstrating that abnormal levels of DYRK1A are maintained in peripheral tissue of subjects with AD-type dementia.

Fig.3

Co-IP analysis of DYRK1A-cytoskeleton complexes in lymphoblastoid cell lines (LCLs). A) The levels of DYRK1A and β-actin in crude lysates. RIPA lysates prepared from each test sample were probed for DYRK1A and β-actin by immunoblotting as described in the Materials and Methods. B) Analysis of β-actin co-IP with DYRK1A. RIPA lysates were IP with anti-DYRK1A antibody (R420) and the precipitates were then probed for DYRK1A and β-actin. Mouse anti-rabbit IgG antibody was included with anti-β-actin antibody in immunoblotting as the internal loading control for quantifying precipitates. IgG marks the band of rabbit IgG from antibody R420. C) Analysis of β-actin co-IP with DYRK1A in an individual with asymptomatic carrier PS1 mutation and her normal cousin. Experiments were performed similarly as described in (B). D) Reciprocal analysis of DYRK1A co-IP with β-actin. IP was conducted with anti-β actin antibody with lysates prepared from test samples and then analyzed for DYRK1A similarly as described above in (B). E) The levels of DYRK1A and β-actin in pooled lysates. Each pool was prepared from four distinct cases and analyzed as described in (A). The relative ratio of DYRK1A and β-actin in DS and AD to that of the control was plotted. Control (100%) was indicated by a dotted line. Difference to the control with statistical significance at the level of p < 0.05 was marked with asterisk. F) Quantification of β-actin co-IP with DYRK1A in pooled lysates. Co-IP experiments were conducted with pooled samples similarly as described in (B). The relative ratio of DYRK1A and β-actin in DS and AD was quantified similarly as described in (E). G) Quantification of β-actin and α-tubulin IP and co-IP in pooled samples. Pooled samples were first IP using either anti-β actin or anti-α-tubulin antibody and the precipitates were then analyzed for IP and co-IP with the congruent and the counter rabbit polyclonal antibody, respectively. The plots shown are the ratio of actin and tubulin IP and co-IP in AD relative to that of the control. H) Quantification of β-actin co-IP with α-tubulin in pooled lysates. Pooled samples were first IP using anti-α-tubulin and then analyzed for α-tubulin and β-actin as described in (F). The relative ratio of β-actin and α-tubulin in DS and AD was presented.

IP experiments were carried out on each cell lines individually and on the pools prepared from several controls, DS, and AD. Although DYRK1A was found to be part of a large complex containing three main cytoskeleton networks with preserved proportionality of all components (see Fig. 2), we focused on the yield and electrophoretic patterning of actin, which displayed easily detectable and consistent changes. Immunoblotting profiles of DYRK1A IP of seven AD, including five SAD, one FAD, one asymptomatic case (at the time of sampling), together with controls, are shown in Fig. 3B and 3C. In all AD samples, actin co-IP yield was significantly reduced. In some samples, there was no apparent actin at all. To rule out the possibility that the observed reduction of actin was caused by loading error, those samples were also probed with anti-rabbit IgG antibody to detect the rabbit anti-DYRK1A IgG (R420) used for IP. The resulting consistent signal of the IgG bands indicates that the loading of the IP complexes was quite even and thus the failure to detect actin cannot be attributed to this technical error. The marked reduction of actin in DYRK1A IP complexes was also seen in an asymptomatic PS1 mutation carrier indicating that AD-associated cytoskeleton abnormalities are present before clinical onset of the disease (Fig. 3C). Actin levels in controls were highly variable with subject 7045 having massively more actin than others. This variability does not seem to correlate with the amount of DYRK1A in those samples. Again, the levels of DYRK1A in AD samples, with some exceptions, were lower than that in controls. In this approach, both cytoskeleton-bound and unbound DYRK1A are detected, and therefore the amount of DYRK1A associated with cytoskeleton cannot be evaluated.

To assess the level of actin-associated DYRK1A, a reciprocal IP experiment with actin antibody was performed. As shown in Fig. 3D, there was indeed less DYRK1A associated with actin cytoskeleton in all AD samples, including the asymptomatic PS1 mutation carrier. This was in agreement with data obtained for crude lysates (Fig. 3A, DYRK1A panel). However, the most consistent finding distinguishing AD from healthy donors was the characteristic reduction in the yield of actin co-IP in AD. The changes were quantified in the pools of controls, DS, and AD donors (Fig. 3F). DYRK1A in DS was enhanced by up to 48% over control and that in AD was reduced by 22% versus control, while co-IP actin in both DS and AD was reduced by about 40% (Fig. 3F). Notably, both DS and AD displayed similar actin patterning, e.g., one band of actin as opposed to two bands in controls, which confirms the finding in the brain tissues (Fig. 2). The reduced abundance of other major cytoskeletal proteins could also be detected by targeting either actin or tubulin in co-IP experiments. This was revealed in actin and tubulin IPs of pools of controls and AD where the yield of actin and tubulin in immunoprecipitates was the measure of the abundance of associated cytoskeletons. As shown in Fig. 3G, both actin and tubulin in cross co-IP were markedly reduced in AD: actin by 80% in tubulin IP, and tubulin by 70% in actin IP. Tubulin was also reduced in tubulin IP by around 50% (including free tubulin). We repeated tubulin IP on pools of controls, DS, and another set of AD samples (Fig. 3H). Again, both DS and AD samples showed significantly lower yield of actin and tubulin. In DS, actin was decreased by 35% and in AD, by 68%. Reduction of tubulin in tubulin IP complexes was by 40% in both DS and AD. In summary, all immunoprecipitation data points to cytoskeletal impairments in DS and AD resulting in reduced interactions of different cytoskeletal networks. Importantly, the same reduction is observed in both DS and AD pathological brains, as well as in cultured lymphocytes from DS and AD donors.

G-/F-actin spin-down assay: Disease-specific electrophoretic profiles of actin bands in G-actin enriched fractions

Another feature that emerged from the co-IP experiments was electrophoretic patterning of actin. Co-IP actin in DS and AD samples was not only less abundant, but its electrophoretic mobility was slower. Upon closer inspection of the immunoblots, it appears that in controls actin is present as two closely spaced bands, but in DS/AD, only the upper band of the actin doublet is present. This was visible in the brain IP experiment (Fig. 2) and also in LCLs IP with DYRK1A and α-tubulin antibodies (Fig. 3). This distinctive feature of DS/AD led us to explore the possibility of using it for qualitative differentiation between healthy and diseased donors. Qualitative assay can be used to avoid the challenging task of achieving the accuracy required for quantitative co-IP assay. Accordingly, after testing many experimental conditions, we developed a protocol allowing unambiguous detection of this actin cytoskeleton abnormality. Our standard protocol is based on separation of G- and F-actin (spin-down assay) by high-speed centrifugation of RIPA lysates and analysis of G-actin (Supernatant) and F-actin (Pellet) enriched fractions by western blotting as depicted in Fig. 1 and in the Materials and Methods. Figure 4A presents results of analysis of G- (S) and F- (P)-actin enriched fractions of control, AD, and Fragile X LCLs. Both control and Fragile X supernatant (S) samples displayed a doublet of actin bands while in AD, there was a single band corresponding to the upper band of the doublet. All pellet samples (P), irrespective of the donor, produced only the upper band; therefore, these are not of diagnostic value. This electrophoretic patterning of actin can also be seen in Coomassie Blue (CB) stained gels (Fig. 4A). Besides visualization of actin band patterning, CB stained gels allow assessment of samples preservation and the quality of electrophoresis. As shown in Fig. 4A, bands in all samples were distinctly resolved with minimal smearing, ruling out the possibility that actin patterning was artefactually created by electrophoretic conditions.

Fig.4

G-/F-actin spin-down assay. A) Analysis of G- and F-actin enriched fractions from LCLs. RIPA lysates were subjected to spin-down fractionation as described in Fig. 1 and in the Materials and Methods. The resulting G-actin enriched supernatant (S) and F-actin enriched pellet (P) fractions were analyzed by immunoblotting for α-tubulin and β-actin. To visualize protein bands, a parallel gel was prepared and stained with Coomassie Blue (CB below). β-actin bands are marked. B) Effect of incubating with ATP and MgCl₂ on electrophoretic patterning of actin. Actin pattering conversion was performed as described in the Materials and Methods. +ATP: incubated ATP and MgCl₂; – ATP: untreated mock sample. Human platelets G-actin standard (AcSt) (Cytoskeleton Inc.) was included for comparison. C) Spin down assay of Lewy body disease (LBD) and frontotemporal lobe dementia (FTLD) LCLs. Details were as described in (A). D) Spin-down assay of pools LCLs lysates. Each pool was prepared by combining three distinct LCLs lysates and subjected to spin-down assay probing for DYRK1A, α-tubulin, and β-actin as described in (A).

The molecular identity of these two actin species is not known. Some hints regarding the nature of these two species may be gained from experiment where G- and F-actin enriched fractions of control LCL were analyzed before and after incubation with ATP and Mg^2 +. As shown in Fig. 4B, under those conditions, the actin doublet normally seen in G-actin fraction is converted into a single slower running actin band with electrophoretic mobility similar to that of G-actin seen in AD. Electrophoretic mobility of globular G-actin standard from human platelets considered to be G-actin monomer, was served as the reference. The lower band of the doublet in the untreated S sample appears to be lined-up with the actin standard; therefore, it likely represents G-actin monomer.

Next we examined whether those cytoskeleton changes are specific for AD-type dementia or shared by other age-associated neurodegenerative diseases. The changes were not present in four LCLs from Fragile X donors. Like controls, they all have a doublet of actin bands in G-actin fractions (Fig. 4A). We further examined two LCLs from patients diagnosed with LBD and one from FTLD by the G-/F-actin spin-down assay. As shown in Fig. 4C, all three displayed characteristic features of the controls, specifically, a doublet in the G-actin fractions. Taken together our findings support the view of cytoskeleton abnormalities in AD appears to be specifically associated with the disease process. Importantly, despite a significant overlap in clinical symptoms and pathology between AD and those two types of dementia [30, 31], they apparently are driven by different pathologic mechanisms. The changes are not only present in AD but also in DS subjects, as was seen earlier in the brain and in the following LCLs IP experiments. As shown in Fig. 4D, which presents results of spin-down analysis of pooled controls, DS, and AD LCLs, both DS and AD pools showed typical AD patterning, i.e., single band of actin in S fractions as opposed to a doublet as in the controls. Not surprisingly, DYRK1A has been found in F-actin enriched fractions indicating that it is an integral part of cytoskeleton assemblies. Therefore, single-banding pattern of β-actin is an apparent AD (and DS)-specific signature, which as a qualitative change, does not require calibration and statistical validation, as do measurements of actin levels for assessing AD/DS.

Among AD cases analyzed, six are SAD and four are from AD families (FAD) (Table 1). Irrespective of their origins, they all presented with cytoskeleton changes specific for AD suggesting a common mechanism of AD pathology. Of special interest is a pair of siblings from consanguineous marriage, one who is affected (AG10987) and the other (AG11009) is a probable AD escapee. Out of ten siblings, four were known to have AD (Fig. 5A) [32]. The affected subject AG10987 was diagnosed with AD by age 78 but, at the time of blood sampling for establishing LCL at age 74, was asymptomatic. The escapee AG11009 showed no symptoms at age 92. Having similar genetic make-up, these two cases offer an opportunity to investigate the effect of a still unidentified genetic change on conferring susceptibility to AD in this family. Accordingly, they were tested by DYRK1A co-IP (Fig. 5B) and the spin-down assays (Fig. 5C), which revealed the presence of the putative AD markers in the affected sibling only. This finding suggests that genetic susceptibility to AD may be driven by a mechanism involving impairment of the cytoskeleton.

Fig.5

Comparison of DYRK1A co-IP and G-/F-actin profiles of two siblings from AD family. A) Partial family tree showing parents and four out of ten siblings, of whom four (three are shown) were known to be affected by AD. Solid diamonds denote affected members while the open diamond represents the normal member of the family. Age of sampling is in parentheses. Donors AG11009 and AG10987, together with control 3657, were analyzed by (B) DYRK1A co-IP assay and by (C) G-/F-actin spin-down assay according to protocols described above.

Blind test of AD cytoskeleton biomarkers in LCLs

The accuracy of the assays was subsequently tested in blinded experiments involving three SAD, three FAD cases, and their age-matched controls. The details of the diagnostic procedures leading to selection of the donors and samples acquisition are given in Materials and Methods. To ensure unbiased reading, cells pellets were shipped from a collaborating laboratory in Poland (see affiliation) in encoded tubes. The list of cases, age of AD onset and PS1 mutations identified in FAD donors is presented in Table 2. Cells were processed for IP and spin-down assays as depicted in Fig. 1. The assays correctly identified 10 out of 12 cases, including all three FAD cases (Table 2). However, the clinical diagnosis of one of the misidentified cases, the donor of 129S LCL (SAD), may have been inaccurate due to overlap of clinical symptoms between AD and other types of dementia. Also, the unusually early onset of the disease at age of 48 years does not fit the standard picture of sporadic AD. Another misidentified sample, 145K(F-CTR), contained not enough cells preventing repeated testing. Irrespective of these two mismatches, the results of blind tests further validate cytoskeletal abnormalities in cultured lymphocytes as potential biomarkers of AD.

Blind test of AD cytoskeleton biomarkers in PBMC

The study conducted so far clearly establishes the usefulness of cultured LCLs in identifying AD but ideally if blood samples from patients can be examined directly. Therefore, we extended the analysis to ten blinded PBMC samples, received from the collaborative laboratory in Poland, of patients diagnosed with AD or MCI due to AD at prodromal AD phase (PDAD) and age-matched control individuals. Samples were processed for G-/F-actin spin-down assay according to the standard protocol used in this study. Characteristics of blood donors together with clinical diagnosis and their corresponding identification by G-/F-actin spin-down assay are presented in Table 3.

Like cultured LCLs, PMBC samples could be distinguished based on the electrophoretic mobility of actin in G-actin enriched fractions. As shown in Figure 6A, actin in some cases runs faster than in the others. However, unlike in LCLs of controls and non-AD dementia, the characteristic doublet of actin bands could not be detected. Upon comparing the actin patterning results obtained above, we reason that the single faster running actin band in the G-actin enriched fraction seen in cases #1, 2, 9, and 10 of Fig. 6A is likely to be the lower band of the doublet and can be used as the proxy for the doublet (Fig. 4A) to represent controls and non-AD dementia. On the other hand, the single slower running actin band seen in case #4, 5, and 7 of Fig. 6A, which has the same mobility as the actin in the F-actin enriched fraction, specifically marks AD. Based on this criteria, in the set of ten samples (Table 3), three (samples # 8, 9, and 10) were identified as normal controls, one (#7) as asymptomatic at risk of developing AD, three (#3, 4, and 5) as prodromal AD (PDAD), two (#2, and 6) as non-AD MCI, and one (#1) as non-AD dementia. Diagnosis based on clinical symptoms only is hampered by pathological overlap with other types of dementia. To improve accuracy of clinical diagnosis CSF biomarkers (a combination of the levels of Aβ₄₂, T-tau, and P-tau) were included [24, 26]. Unfortunately, the diagnostic performance of the CSF biomarker-trios was disappointing. Only two cases classified by MCI symptoms (#2 and 3) and one diagnosed as AD (#1) met some or all CSF AD biomarkers criteria. Since CSF Aβ_42/40 ratio was not assessed in this study, it likely became a limiting factor for reaching complete accordance between CSF biomarker-based, traditional diagnosis, and the G-/F-actin markers [33]. However, even CSF biomarker-negative MCI patients can progress to AD, as illustrated by a study population of patients with MCI where half of their participants who later progressed to AD were negative for CSF markers, although their progression was slower than in the biomarker positive group [34]. By clinical symptoms all cases in the present study met MCI criteria, the G-/F-actin marker diagnosis could further differentiate them allowing distinction between PDAD and non-AD MCI cases. This was evident in the identification of cases #2 and #6 as non-AD MCI, while case #1, deemed to represent AD in the set, was identified as non-AD dementia. Case #7, a normal control patient, displayed AD-specific patterning of actin marking this patient as being at risk of developing AD dementia. The effectiveness of classification of MCI cases by G/F actin markers awaits verification by studies conducted on larger groups of patients. The actin patterning displayed by PBMCs and the similarity to what has been observed in LCLs, speaks in favor of the diagnostic validity of G-/F-actin markers.

Fig.6

A) G-/F-actin spin-down assay of peripheral blood mononuclear cell (PBMC) from MCI donors. Spin-down assay was performed on PBMC obtained from donors listed in Table 3 as described in the Materials and Methods. B) Analysis of white blood cells (WBC) from postmortem blood sample of a 65-year-old DS donor. WBCs were isolated and processed for G-/F-actin spin-down assay similarly as described above. Samples of S fraction from Fragile X (FX) and control (CTR) are included for comparison.

Case #1, which by clinical criteria and CSF biomarkers, was diagnosed as AD dementia, tested negative in the G-/F-actin spin-down assay. Because of this uncertainty, our PBMC set of samples lacked an unambiguously identified case of AD required for validating the test results of MCI cases. Therefore, we reached out for a blood sample of a DS patient who would have clearly defined disease background in order to serve as the reference for identifying AD. Accordingly, white blood cells isolated from postmortem blood sample from a 65-year-old DS patient was processed for the spin-down assay (Materials and Methods). The results of G-/F-actin markers analysis of this blood sample together with S fraction samples of control and Fragile X of LCLs are presented in Fig. 6B. As with the LCLs (Fig. 4) and in prodromal AD MCI PMBC cases (Fig. 6A), S and P fractions from the blood of the DS patient displayed only the equivalent of the slower running band of the actin doublet with the patterning that specifically marks AD/DS.