Combination of blood-based biomarkers and neuropsychological assessment enables reliable classification of tested subjects by controls, mild cognitive impairment and Alzheimer's disease Martin Kleinschmidt

Alzheimer’s Disease: The classical and the more detailed Amyloid hypothesis

Alzheimer’s Disease: Reducing plaques by lowering total Aβ Reducing total Aβ Approaches designed to reduce total Aβ Intercalation: Scyllo-inositol (ELND005) failed to show efficacy, Ph2, Aug 2010 Passive Immunization: Bapinuzumab & Solanezumab reduce Aβ, but do not improve cognition (18mts treatment 25% reduction) Aducanumab showed promising results in Ph1b - β-secretase inhibition: Specific β-secretase inhibitors block 15% Aβ deposition in tg2576, 2mts treatment, 25% in 7mts treatment, only minimal cognition improvement, no patient trial yet - γ-secretase inhibition: Semagacestat worsened cognition in patients, studies terminated

Alzheimer’s Disease: Reducing aggregation by lowering pGlu-Aβ QC pGlu-Aβ drives aggregation pGlu-Aβ pathogenic hypothesis Pathogenic species of AD may not be the monomeric Aβ40 or Aβ42 N-terminally truncated Aβ species develop EARLY in AD N-terminally truncated Aβ species make up >20% of deposits Most prevalent (>60%) N-terminally truncated Aβ are pyroglutamated Aβ (pGlu-Aβ) pGlu-Aβ species seed to full-length Aβ, have fast aggregating kinetics, are long-living, accumulate aggressively in AD brain pGlu-Aβ accumulation correlates well with AD progression and pathology Early intervention is decisive for treatment success – Early diagnosis is required

Alzheimer’s Disease: Strategy for suitable biomarkers MRI and PET pros: established methods, state of the art cons: resource intensive, not suitable for phase I, good in phase II / III CSF biomarkers (e.g. Aβ42/Aβ40, Tau, P-Tau) pros: reliable biomarker, state of the art, commercially available diagnostic tools cons: invasive, most volunteers decline LP, not suitable for area-wide screening Plasma (e.g. Aβ42/Aβ40, Tau, P-Tau) pros: blood withdrawal is generally accepted, suitable for initial screening cons: currently no reliable biomarker, 10-times lower Aβ conc. than in CSF, but 100-times higher overall protein content Development of plasma based biomarker assay systems

Aβ quantification: ELISA vs. extraction assay Spiking of human plasma and ELISA buffer with 20 ng/ml Aβ(1-42) Quantification of Aβ(1-42) in spiked plasma / ELISA buffer by specific ELISA Recovery rate was analyzed in two independent experiments spiked ELISA buffer spiked plasma Aβ peptides can be difficult to detect in conventional immunoassays Observing no signal or too much signal – both can be caused by plasma proteins Many diagnostic studies analyzing plasma Aβ have failed or gave inconsistent data Conclusion: Quantitative extraction of Aβ peptides and removing of interfering proteins should improve reliability of plasma Aβ determination

Separation of A peptides Aβ quantification: ELISA vs. extraction assay Plasma separation Capture Aβ from plasma Blood withdrawal Separation of A peptides Quantification of the different Aβ species Aβ(pE3-40/42)? 100 50 Specific pattern of Aβ peptides might characterize stage of Alzheimer Using pan-specific anti-Aβ antibody recovery (~13%) rate comparable with ELISA

Aβ quantification: ELISA vs. extraction assay SPR interaction analysis – monovalent vs. bivalent binding of Aβ Monovalent binding – KD= 7.43 nM Bivalent binding – KD= 0.061 nM Elevation of binding strength due to avidity effect Every anti-Aβ antibody is immobilized by many anti-mouse antibodies to the magnetic beads pan: epitope aa 17-24; C40: epitope aa 35-40; C42: epitope aa 37-42 Recovery rate from spiked plasma: 13 % Recovery rate from spiked plasma: 98 %

Aβ quantification: ELISA vs. extraction assay Quantification of pE3-Aβ in plasma and CSF pE3-Aβ is not quantifiable neither in human plasma nor in human CSF, although these species are the most abundant in brain deposits

Healthy controls or possible MCI (MMSE ≥ 21) Diagnostic study: Parameters Design: Cross-sectional Multicentric (three centers in Halle, one in Magdeburg, one in Kiel) Blinded Biomarker analysis: experimenter does not know classification age groups survey target Healthy controls or possible MCI (MMSE ≥ 21) 18-30 yrs 22 20 40-65 yrs 25 66-85 yrs 78 20 + 40 AD (MMSE < 21) 18 40 total 143 140

Diagnostic study – Neuropsychology and Biomarker neuropsychological analysis (independent variable) learning & memory: CERAD (word lists), WMS-R (narration), ROCF (recall) attention & executive func.: ROCF (copy), RWT (semantic/phonetic, alternating), TMT (psycho-motor activity/cog. Flexibility) cognitive decline: premorbid IQ (MWT-B, social formula), “non-hold” function (HAWIE-R Similarity) functioning: performance activities of daily living (FAST) Analysis of major depression episode (HADS-D & BDI) – subject exclusion Genetic risk factor – Apo E allele analysis plasma and serum biomarker (dependent variables) plasma Aβ(1-40) plasma Aβ(1-42) total hCCL2 / pE1-hCCL2 in serum auto antibodies against Aβ(1/pE3-x): IgG2, IgG3, IgM, IgA, IgG, Ig’s pro-inflammatory cytokines: IL-10, IL-1β, IFN-γ, IL-6 and TNF-α

Diagnostic study: Neuropsychological Tests - overview multivariate classification scheme to create the independent variable Kleinschmidt et al., 2015

Diagnostic study: Neuropsychological classification sample size of further analyses N = 94 Controls (18-30 yrs) Controls (40-65 yrs) Controls (66-85 yrs) naMCI aMCI MCI AD n 19 13 3 14 15 age (years) 24 (3.5) 54 (9.0) 74 (6.5) - 77 (5.7) 74 (5.7) 74 (4.8) female 53% 39% 58% 43% 50% 47% Kleinschmidt et al., 2015 excluded cases: classification impossible n = 19 possible Major depression episode n = 14 exclusion by CRF or premorbid IQ < 85 n = 12 naMCI n = 3 blood samples missed n = 1

Diagnostic study: ApoE allele status Kleinschmidt et al., 2015 ApoE allele frequencies is similar to distribution in German population ApoE ε4 allele is the major genetic risk factor for Alzheimer’s disease 76% of AD cases are carrying the ApoE ε4 allele

Diagnostic study: Plasma Aβ level Absolut level were normalized to an internal standard in every measurement One-way ANOVA, contrasts: t-statistics (α=0.05) Kleinschmidt et al., 2015 Aβ(1-42) is significant lower in MCI & AD compared to controls Age-dependent increase of Aβ(1-40) (maybe prevent Aβ deposition) Stepwise decrease of Aβ(1-42)/Aβ(1-40) ratio, similar to data of CSF analysis

Diagnostic study: Plasma Aβ level vs. cognitive test Plot CERAD delayed recall (z-score) vs. Aβ(1-42) / Aβ(1-40) ratio Pearson, r = 0.50, p<0.001 Kleinschmidt et al., 2015 Aβ(1-42)/Aβ(1-40) ratio of the entire cohort correlate with memory impairments in verbal delayed recall

Diagnostic study: Plasma Aβ level vs. Apo E *** Kleinschmidt et al., 2015 One-way ANOVA, contrasts: t-statistics (α=0.05) Reduced Aβ(1-42) is strongly associated with ApoE ε4, similar to CSF data Aβ(1-42) is more closely linked to ApoE allele than to clinical status Aβ(1-40) does not correlate with ApoE allele status

Diagnostic study: Inflammatory marker in serum Kleinschmidt et al., 2015 No differences in biomarker levels between controls, aMCI, MCI and AD Large inter-individual variations of CCL2 in controls, due to peripheral processes Levels of IL-1β, IL-6, IFN-γ were mainly below LLOQ

Diagnostic study: Autoantibodies against pE3-Aβ pE3-Aβ is not quantifiable in blood or CSF, but maybe leave an “imprint” Quantification of different (sub)classes of anti-pE3-Aβ autoantibodies Kleinschmidt et al., 2015 H test, contrasts: Mann-Whitney (α=0.05) One-way ANOVA, contrasts: t-statistics (α=0.05) Strong increase in oldest controls and decrease in AD exclusively of IgG2 Stepwise decrease of IgG2 / total IG ratio (comparable to plasma Aβ) Autoantibodies against Aβ(1-x) do not differ (data not shown)

Diagnostic study: Autoantibodies vs. plasma Aβ and ApoE Plot Aβ(1-42) / Aβ(1-40) ratio vs. anti-pE3-Aβ IgG2 / total Ig ratio Kleinschmidt et al., 2015 Spearman rank corr. r=-0.33, p<0.001 One-way ANOVA, p=0.529 anti-pE3-Aβ IgG2 / total IG ratio correlate inversely with plasma Aβ Autoantibodies do not correlate with Apo E allele status Indicating an independent influence of autoantibodies and ApoE isoforms on cognitive performance, most likely by modifying Aβ metabolism

Diagnostic study: Multivariate visualization of data Biomarker readouts were standardized to z-scores Spider plots contain group medians of seven different dimensions Total sum of all seven profile measures (area score) was calculated Stepwise decrease of area score In 93% aMCI, 100% MCI and 100% AD area score is below 1172 At a glance evaluating existence and progression of Alzheimer’s disease Kleinschmidt et al., 2015

Acknowledgements Prof. Hans-Ulrich Demuth Kathrin Gnoth Nadine Taudte Nadine Jänckel Mercedes Scharfe Claudia Göttlich Anja Weber Prof. Bernd Leplow Robby Schönfeld Medical facilities in Halle: Gallmed GmbH (Monitoring) Clinics for geriatrics of Diakonie Neurology Eger & Jakob Daniel Bittner Prof. Josef Aldenhoff