FUNDAMANT: an interventional 72-week phase 1 follow-up study of AADvac1, an active immunotherapy against tau protein pathology in Alzheimer’s disease

Despite numerous setbacks, compounds based on the amyloid hypothesis (whether immunotherapies or small molecules) continue to be investigated vigorously, with focus shifting towards evaluating their efficacy in early and preclinical stages of AD (e.g., A4: Anti-Amyloid Treatment in Asymptomatic Alzheimer’s Disease [21] or the API: Alzheimer’s Prevention Initiative [22]).

Several other anti-tau immunotherapeutics are being tested in clinical trials. Three mAbs—C2N 8E12 (C2N; AbbVie, North Chicago, IL, USA) [23], RO 7105705 (AC Immune, Lausanne, Switzerland; Genentech, South San Francisco, CA, USA; Hoffmann-La Roche, Basel, Switzerland) and BMS-986168 (Biogen, Cambridge, MA, USA) [24]—recognise various epitopes on the N-terminus of tau protein. They target mostly extracellular N-terminal fragments and/or all six tau isoforms. The active liposome-based vaccine ACI-35 (AC Immune; Janssen, Beerse, Belgium) [25] is being tested in a phase 1 clinical trial. The vaccine generates antibodies recognising tau protein phosphorylated at pS396/pS404. The discontinued RG7345 passive vaccine similarly aimed for the C-terminus, specifically the phospho-epitope pS422 [26].

In contrast to the aforementioned tau therapeutic strategies, AADvac1 induces generation of antibodies targeting a part of the tau molecule that is preserved in all aggregating truncated tau species: the microtubule-binding repeat domain (MTBR). Some newly introduced tau-targeted immunotherapies seem to follow this rationale (LY3303560, UCB1017) (see also www.​alzforum.​org). Little has been reported to date about other recently introduced compounds (JNJ-63733657, BIIB076) [27].

The present study (protocol AC-AD-002) was a 72-week, open-label, single-arm interventional follow-up trial for the 24-week first-in-human study of AADvac1 (protocol AxonCO18700) [17]. The study was conducted in four centres in Austria (University Hospital Graz; Medical University Wien; Social and Medical Centre East, Danube Hospital Wien; and Christian-Doppler Clinic Salzburg).

The study enrolled only patients who had completed the preceding phase 1 study of AADvac1. Thus, the inclusion/exclusion criteria were minimal, requiring informed consent capability; the availability of a caregiver; and the absence of severe co-morbidities, immunosuppressive treatment, or contraindications for magnetic resonance imaging (MRI).

The preceding study enrolled patients aged 50–85 years with a diagnosis of Alzheimer’s dementia based on the National Institute of Neurological and Communicative Disorders and Stroke-Alzheimer’s Disease and Related Disorders Association criteria, a screening MRI study compatible with the diagnosis of AD, and a Mini Mental State Examination (MMSE) score of 15–26 (see Table 1 for demographic characteristics of the study population). See Additional file 1 for verbatim wording of the inclusion criteria of this study and the preceding first-in-human trial.

Each dose of AADvac1 consisted of 40 μg of Axon Peptide 108 (N-terminally cysteinylated tau 294–305/4R, amino acid sequence CKDNIKHVPGGGS) coupled to keyhole limpet haemocyanin (KLH) via a maleimide linker, with aluminium hydroxide adjuvant (containing 0.5 mg Al³⁺) in a phosphate buffer volume of 0.3 ml. The design of the vaccine has been described in detail previously [19].

The treatment regimen of the preceding study consisted of six doses of AADvac1 in 4-week intervals (three doses in the double-blind phase and three doses in the open-label phase). Patients randomised to placebo have instead received three doses of placebo in the double-blind phase, followed by three doses of AADvac1 in the open-label phase [17].

Patients assigned to AADvac1 in the double-blind phase of the preceding study have received two booster doses in 24-week intervals in this trial. The first booster dose was optional, to be administered if the titres of antibodies against Axon Peptide 108 declined below 75% of titres achieved following the initial six-dose vaccination regimen. The second booster dose was obligatory.

Patients assigned to placebo in the double-blind phase of the preceding study have first received three doses of AADvac1 in 4-week intervals (“catch-up visits”) to bring their vaccination regimen in line with patients who had received AADvac1 in the double-blind phase, and subsequently entered the above-described booster regimen.

Thus, in the two studies combined, both groups have received an identical AADvac1 treatment regimen: six doses administered in 4-week intervals, followed by two boosters in 24-week intervals. Both groups have a total of 96 weeks of on-treatment data and are pooled for analyses that measure on-treatment change. See Additional file 1 for an illustration.

Reports of adverse events were obtained by non-directive questioning from patients and their caregivers, as well as by reviewing the patient diary. A structured neurological and physical examination was done by the investigators at 12-week intervals. Vital signs were assessed at every visit and 1 h after AADvac1 administration. Standard laboratory panels (biochemistry, coagulation, haematology, and inflammation markers) and dipstick urinalysis were assessed at 12-week intervals; a laboratory urinalysis was done if indicated by pathological dipstick results. Investigators reviewed laboratory results and reported clinically significant abnormalities as adverse events. A standard electrocardiogram was done every 24 weeks. Safety MRI was done every 24 weeks, and each scan was assessed in parallel by the investigator and the sponsor’s radiologist. Accruing safety data were reviewed every 6 months by an independent Data and Safety Monitoring Board (DSMB).

An indirect enzyme-linked immunosorbent assay (ELISA) was used to measure the titres of vaccine-induced antibodies. Axon Peptide 108, tau151–391/4R, and KLH were used as solid phase, separately immobilised on microtitre plates (High Binding strip plates; Greiner Bio-One, Frickenhausen, Germany), and incubated overnight with serially diluted patient serum samples. After extensive washing, bound antibodies were detected by anti-human immunoglobulins conjugated to horseradish peroxidase (anti-human subclass-specific secondary antibodies for the detection of IgM and IgG, IgG subclasses IgG1, IgG2, IgG3 and IgG4, and anti-human IgA + IgG + IgM for anti-KLH antibodies; all provided by Pierce Biotechnology, Rockford, IL, USA). We measured the amounts of bound secondary antibodies through the activity of horseradish peroxidase with the ready-to-use chromogenic substrate TMB One (KEM-EN-TEC Diagnostic, Taastrup, Denmark) and the absorbance at 450 nm. The resulting signal was compared with that obtained for the patient’s serum collected at baseline.

We defined the titre of the antibodies in the serum as the highest dilution at which the absorbance at 450 nm was at least twice the absorbance of equally diluted pre-immunisation serum samples from the same patient. To obtain the titres, we first fitted the absorbance values of serially diluted serum samples into curves using a four-parameter logistic non-linear regression model in the R programming environment (R Development Core Team 2017, Vienna, Austria). We measured the titre as the dilution at which the curve of the post-immunisation serum crossed the twice multiplied curve of the corresponding pre-immunisation serum. For the purpose of assay consistency and quality, we used quality control samples with two concentrations of the humanised version of mAb DC8E8 (AX004) that was used in the design of AADvac1 [19, 20] as positive controls. AX004 is the passive vaccine counterpart to AADvac1.

Tissue samples from AD, Pick’s disease, corticobasal degeneration (CBD) and progressive supranuclear palsy (PSP) were obtained from the Netherlands Brain Bank. Sections were pre-treated with formic acid (98% for 1 min at 4 °C) and heat (autoclave, 121 °C, 20 min), followed by incubation with serum antibodies for 72 h. All sections were incubated with goat anti-human biotinylated secondary antibodies at room temperature for 2 h, and with avidin-biotin-peroxidase complex for 2 h. The immunoreaction was visualised with VIP substrate (VECTASTAIN Elite ABC Kit; Vector Laboratories, Burlingame, CA, USA) and counterstained with methyl green (Vector Laboratories).

We dissolved sarkosyl-insoluble tau [28] isolated from human brain samples (AD: Kuopio Brain Bank, Braak V–VI, gyrus temporalis; PSP and CBD: London Brain Bank, nucleus caudatus and globus pallidus; four AD brains and three different brains for each tauopathy), in 1× SDS sample loading buffer in 1:50 volume of the soluble fraction. We denatured sarkosyl-insoluble tau by heating at 95 °C for 5 min. Samples were loaded onto 5–20% gradient SDS-PAGE gels and electrophoresed in a Tris-glycine-SDS buffer system for 40 min at 25 mA. We transferred the proteins to polyvinylidene fluoride transfer membranes (1 h at 150 mA in 10 mmol/L 3[cyclohexylamino]propanesulphonic acid, pH 12). After the transfer, we blocked the membranes in 5% non-fat dry milk in PBS for 1 h at room temperature and then incubated them for 16 h at 4 °C with serum antibodies diluted in PBS supplemented with 5% bovine serum albumin, followed by three washes with a large volume of PBS supplemented with 0.1% Tween 20 (PBS-T; Sigma-Aldrich, St. Louis, MO, USA). After the washes, we used horseradish peroxidase-conjugated goat anti-mouse Ig (DAKO, Glostrup, Denmark) diluted 1:4000 with PBS-T as the secondary antibody. Incubation (1 h at room temperature) was followed by washing with PBS-T. We developed the blots with SuperSignal West Pico Chemiluminescent Substrate (Pierce Biotechnology) and detected the signal using an LAS3000 imaging system (FUJI Photo Film, Tokyo, Japan).

Flow cytometry was assessed from the peripheral blood collected into ethylenediaminetetraacetic acid-treated test tubes (VACUETTE 454036; Greiner Bio-One). Samples were analysed within 24 h of blood collection and transported in a temperature-controlled environment (25 °C). For T-cell identification, 50 μl of whole blood was labelled for 20 min in the dark with the following commercial mAbs: CD3 (peridinin chlorophyll protein complex-conjugated, clone UCHT1), CD4 (allophycocyanin [APC]-conjugated, clone MEM-241), CD8 (APC-conjugated, clone MEM-31), CD28 (phycoerythrin-conjugated, clone CD28.2), CD45RA (fluorescein isothiocyanate [FITC]-conjugated, clone MEM-56), CD45RO (FITC-conjugated, clone UCHL1). All antibodies were purchased from Exbio (Prague, Czech Republic). Red blood cells were lysed for 10 min in 100 μl of EXCELLYSE I lysing solution (Exbio). The remaining cells were washed in 2 ml of distilled water, centrifuged for 5 min at 300 g and resuspended in 500 μl of PBS for analysis. Finally, samples were acquired on a FACSCalibur flow cytometer (BD Biosciences, Heidelberg, Germany) and analysed with Cell-Quest software (BD Biosciences). Cells were electronically gated according to size (forward scatter) and granularity (side scatter) for the detection of 50,000 leucocytes. For each antibody, negative staining levels were set by comparison to unstained and fluorescence minus one controls. The following lymphocyte populations were analysed: CD3⁺ (T cells), CD3⁺/CD4⁺ (CD4 helper T cells), CD3⁺/CD8⁺ (CD8 cytotoxic T cells), CD3⁺/CD8⁺/CD28⁺ (co-stimulatory receptor CD8 cytotoxic T cells), CD3⁺/CD8⁺/CD28⁺/CD45RA⁺ (naïve co-stimulatory receptor cytotoxic CD8 T cells), CD3⁺/CD8⁺/CD28⁺/CD45RO⁺ (memory co-stimulatory receptor cytotoxic CD8 T cells). Results were expressed as number of the cells with the given phenotype counted per 50,000 leucocytes. Standard haematological assessments were done on a Siemens Advia 2120i haematology system (Siemens Healthcare, Erlangen, Germany) with peroxidase staining technology.

Cognition was assessed using a standard 11-item Alzheimer’s Disease Assessment Scale cognitive assessment (ADAS-Cog), the Controlled Oral Word Association Test (COWAT) and the Category Fluency Test (CFT). The ADAS-Cog11 is a weighted, multi-domain battery for rating the severity of the patient’s impairment from 0 to 70, with 0 indicating flawless performance and 70 representing pronounced dementia. The MMSE is a short cognitive battery rating the severity of the patient’s impairment from 30 to 0, with 30 indicating flawless performance and 0 representing pronounced dementia. The CFT is an executive function and language test, asking the patient to produce as many words fitting a specific category as possible within 60 s. The category “animals” was used in this study. The COWAT is also a test of language, fluency and executive function. The patient performs three trials, each time being asked to produce as many words as possible starting with a given letter within the allotted timeframe of 60 s per trial. The letters used in the German COWAT version were B, L and S.

We used the following scanners: Siemens Magnetom Essenza (1.5 Tesla), Siemens Magnetom Prisma (3 Tesla) and Siemens Magnetom Tim Trio (3 Tesla). MRI was done for safety assessment and volumetric analysis. Scans obtained at screening and in weeks 24, 48, 72 and 96 of treatment were used for the volumetric analysis. The following sequences were recorded: axial T2-weighted (fast spin echo), axial fluid-attenuated inversion recovery (IRfast spin echo), axial T2-star (spoiled gradient echo), and sagittal 3DT1 (IR-prepped fast three-dimensional gradient echo [spoiled]). Detailed scanner settings are listed in Additional file 1. The following ROIs were subjected to volumetric analysis: total brain, hippocampus (sum of left and right) and lateral ventricles.

Whole-brain volume change between follow-up visits and baseline was estimated with SIENA [29], part of FSL [30]. To extract volume estimates of the hippocampus and the lateral ventricles, images were automatically processed with the longitudinal stream [31] in FreeSurfer (version 6.0). Specifically, an unbiased within-subject template space and image was created using robust, inverse consistent registration [32]. Several processing steps, such as skull stripping, Talairach transforms, atlas registration as well as spherical surface maps and parcellations were then initialised with common information from the within-subject template, significantly increasing reliability and statistical power [31].

No sample size calculation was performed, because the study enrolled only patients who completed the preceding study (AxonCO18700). We performed the statistical analyses using Prism version 6.07 software (GraphPad software, La Jolla, CA, USA) and the R programming environment (R Development Core Team 2017). Patients whose signal was at least double compared with baseline in the ELISA for IgG against Axon Peptide 108 at any point of treatment were defined as responders. For correlations between antibody response and other outcomes, all completer patients in whom the dependent variable was recorded at the end-of-study visit were used. For descriptive statistics of antibody titres, responders who completed the study were analysed.

Because the titres of AADvac1-induced IgG antibodies fluctuated over the course of the study depending on whether the patients were recently (re)vaccinated, AUCs were calculated using IgG titre values measured over 96 weeks since the initiation of treatment as a measure of cumulative exposure to IgG antibodies against Axon Peptide 108 (and by proxy against pathological tau protein).

Because patients did not attend the end-of-study visit precisely on the last day of week 96, cubic spline smoothing with generalised cross-validation was used to estimate (if week 96 was between last two consecutive visits) or predict (if week 96 was after last two consecutive visits) values at week 96.

Correlations were analysed using Spearman’s correlation coefficient because multiple variables did not follow a normal distribution. To analyse the linear time trend for the variables with lower variability (CD markers), a mixed effects linear regression model was used. For demographic variables that were not re-recorded for this follow-up study (e.g., ApoE4 genotype), the results obtained in the preceding phase 1 trial are imputed. Descriptive statistics are listed as mean (±SD), except for antibody titres, which are listed as geometric means (95% CI), and counts are listed as number of events and percentage of the study population.

Transgenic R3/m4 mice (n = 10, female, aged ~ 3 months) expressing human truncated tau 151–391/4R under the Thy1 promoter [33] were housed under standard laboratory conditions, with a 12-h/12-h light/dark cycle, and access to food and water ad libitum. Animals were treated with five doses of AADvac1 (40 μg of Axon Peptide 108 coupled to KLH) in 21-day intervals.

One animal died spontaneously prior to evaluation. This animal was removed from analysis. Animals were perfused with PBS either at the terminal stage or at the age of 6.5–7 months. Transgenic mice were deeply anaesthetised with Zoletil (Virbac, Carros, France)/xylazine and perfused intracardially using a peristaltic pump for 2 min with PBS. The brain was post-fixed overnight in 4% paraformaldehyde in PBS, pH 7.2, cryoprotected with 15% and 25% sucrose solutions (subsequently overnight), frozen in 2-methylbutane (30 s at − 42 °C) and transferred to dry ice and finally stored at − 70 °C.

Blood samples were obtained at the time mice were killed, and serum was stored at − 70 °C. Antibody half-maximal effective concentration (EC₅₀) was measured by ELISA, using tau 151–391/4R as solid phase, and a horseradish peroxidase-conjugated anti-mouse-Ig secondary antibody. We measured the amounts of bound secondary antibodies through the activity of horseradish peroxidase with the ready-to-use chromogenic substrate TMB One (KEM-EN-TEC Diagnostic) and the absorbance at 450 nm. EC₅₀ was defined as the dilution at which the sample displays half-maximum absorbance at 450 nm.

Sagittal brainstem sections (40 μm thick) were cut on a Leica CM1850 cryomicrotome (Leica Biosystems, Buffalo Grove, IL, USA). For semi-quantitative analysis, neurofibrillary tangles (NFTs) were quantified in two sections from each brain. Tissue sections were incubated with primary antibodies AT8 (Pierce Endogen), pT212 and pS214 (Invitrogen, Carlsbad, CA, USA) overnight at 4 °C. Sections were immunostained using the standard avidin-biotin-peroxidase method (VECTASTAIN Elite ABC Kit) with VIP as the chromogen.

Antibody titres increased following each of the six doses in the initial vaccination regimen. Over the 6-month vaccination-free period thereafter, antibody titres declined in a majority of patients: IgG to a median value of 15.8%, IgM to 23.5%, and anti-KLH antibodies to 19.8% of values obtained 4 weeks after the initial vaccination regimen. Booster doses restored all types of antibody titres to previously achieved levels (see Fig. 1 and Table 2).

Similarly to the antibody response observed in the first-in-human study, the 96-week AUC of the IgG antibody response was varied, with a median AUC of 1.60 × 10⁶ (IQR, 0.96 × 10⁶–3.93 × 10⁶; minimum, 0; maximum, 7.4 × 10⁶). Patients who initially responded to AADvac1 treatment with high antibody titres continued to display high levels of antibody response; their AUC of IgG titre was highly correlated with the IgG titre at the end of the primary vaccination regimen (Spearman r = 0.889, p < 0.001). Over the entire course of treatment, IgG antibody titre against the peptide was correlated to anti-KLH titres (Spearman r = 0.595, p < 0.001).

A portion of the immunological predictors identified in the first-in-human study also predicted the cumulative amount of antibody production over 96 weeks. The IgG AUC values correlated with the following immunological variables measured prior to vaccination: absolute (r = 0.618, p = 0.004) and relative lymphocyte counts (r = 0.755, p = 0.001), CD3⁺/CD4⁺ lymphocyte counts (r = 0.585, p = 0.007), as well as absolute (r = − 0.492, p = 0.028) and relative segmented neutrophil granulocyte counts (r = − 0.786, p < 0.001). A negative correlation between age and IgG AUC was observed (r = − 0.480, p = 0.032). Relative lymphocyte and segmented granulocyte counts show a far stronger association with age, though (r = − 0.667, p = 0.001 and r = 0.642, p = 0.002, respectively). Similarly, there was a negative correlation between age and CD3⁺/CD4⁺ lymphocyte counts prior to treatment (r = − 0.485, p = 0.009).

No significant impact of sex on IgG AUC was observed; as expected, an impact of sex on several haematological variables was noticeable, though (p < 0.05 for CD3⁺, CD3⁺/CD4⁺, CD3⁺/CD8⁺/CD28⁺, CD3⁺/CD8⁺/CD28⁺/CD45RA⁺, and relative lymphocyte counts, as well as relative and absolute segmented neutrophil granulocyte counts).

No changes in lymphocyte populations were observed over the course of treatment.

The IgG subclass assessment was performed cross-sectionally at the end of the initial vaccination regimen. In all patients who mounted an IgG response, the response was IgG1-dominated: the patients’ IgG response consisted of 69.54% (61.85, 78.20) IgG1, 1.33% (0.50, 3.50) IgG2, 12.58% (6.70, 23.63) IgG3, and 0.17% (0.08, 0.39) IgG4 (geometric mean, 95% CI) (see Fig. 2).

Western blot analysis was used to assess whether the antibodies induced by vaccination with AADvac1 could label pathological tau protein in brains with AD and non-AD tauopathies (CBD, PSP). AADvac1-induced antibodies are reactive with sarkosyl-insoluble tau extracts in AD as well as non-AD tauopathies (three PSP and three CBD brains) (Fig. 3).

Antibodies elicited by vaccination were able to label all aspects of tau pathology in histological preparations from AD brains: NFTs, neuropil threads, and the neuritic corona of plaques (Fig. 4a–c). Similarly, the same patients’ sera were able to stain the Pick bodies in Pick’s disease (Fig. 4g–i), oligodendroglial coiled bodies in PSP (Fig. 5a–c) and astrocytic plaques in CBD (Fig. 5g–i).

Antibodies against Axon Peptide 108 were detected in the CSF of all patients who donated it. The titres in the CSF were 0.22 (±0.13)% of those measured in the periphery (mean ± SD) and strongly correlated (Pearson r = 0.884, p = 0.0003).

The only adverse events clearly linked to AADvac1 treatment were injection site reactions (erythema, swelling, warmth, pruritus, pain, nodule); these are usual for aluminium-containing vaccines. One or more of these were observed in 50% of patients on AADvac1 treatment. Injection site reactions were reversible and predominantly mild in presentation.

Six serious adverse events were observed (abdominal strangulated hernia, dehydration, acute psychosis, behavioural and psychiatric symptoms of dementia, second-degree atrioventricular block, and sinus bradycardia). None were judged by the investigators to be related to AADvac1 treatment.

No allergic or anaphylactic reactions were observed. No safety signals emerged in laboratory assessment (coagulation, blood biochemistry, haematology, and urinalysis), vital sign assessment, or neurological and physical examination. No safety signals were detected by MRI assessment. No oedematous changes occurred; no meningeal changes and no meningoencephalitis were observed. New micro-haemorrhages were observed in one ApoE4 homozygote, and superficial haemosiderin was detected in one ApoE4 heterozygote; both events were clinically silent. This is consistent with the background incidence of such lesions in the AD patient population [34].

All adverse events occurring in at least 10% of patients are listed in Table 3.

Over 96 weeks of observation, on average, patients declined on the ADAS-Cog11, the COWAT, the CFT, and the MMSE. The mean change (±SD) on the ADAS-Cog11 was 11.7 (±13.0) points, on the COWAT -4.7 (±7.3), on the CFT –2.2 (±4.6), and on the MMSE –5.3 (±6.6). Split by disease stage (mild AD defined as MMSE 20–26 and moderate AD as MMSE 15–19 at screening), the change in the mild AD population on the ADAS-Cog11 was 7.6 (±11.2), on the COWAT –2.9 (±8.5), on the CFT –1.4 (±5.6), and on the MMSE –3.2 (±7.2); the change in the moderate AD population on the ADAS-Cog11 was 18.1 (±13.9), on the COWAT –7.3 (±4.2), on the CFT –3.4 (±2.6), and on the MMSE –8.3 (±4.8).

Cognitive change measured by the COWAT correlated with IgG titre AUC, with patients who had higher IgG titres experiencing less cognitive decline (Spearman r = 0.488, p = 0.049) (Fig. 6c).

As is usual with AD populations recruited solely on the basis of clinical criteria, some patients were negative for AD biomarkers (hippocampal atrophy on MRI, and/or CSF AD biomarkers). Therefore, we analysed the correlation of cognitive change with IgG AUC also in the subgroup of patients with positive AD biomarkers according to McKhann [35]. A clearer association between cognitive benefit and IgG AUC was seen for COWAT (Spearman r = 0.673, p = 0.039); the association for ADAS-Cog11 likewise was stronger, but did not reach statistical significance (Spearman r = − 0.576, p = 0.088) (Fig. 6b, d).

Finally, we tested the hypothesis whether patients with more advanced disease, and thus presumably more tau pathology, require higher antibody titres for cognitive benefit. The IgG AUCs were divided by the respective patient’s baseline ADAS-Cog11 value, and analysed for association with cognitive change. Correcting thus for baseline disease severity, the association between IgG AUC and cognitive change was significant both for the ADAS-Cog11 (Spearman r = − 0.758, p = 0.015) and the COWAT (Spearman r = 0.806, p = 0.007) (Fig. 6f, h).

The change in CFT was correlated with IgG AUCs only in the general population; the association was not found in the subgroup analysis or when the IgG titre AUCs were corrected for disease severity.

The changes in brain volume observed over 96 weeks of treatment, shown as mean (SD), were as follows: TBV –5.433% (2.448); LVV + 16.84% (8.946); HCV –9.324% (3.488). These atrophy rates do not exceed those expected of mild to moderate AD populations [36, 37].

Analogously to cognitive assessment, correlations between AUC of IgG titres and brain atrophy were assessed for all completers and for the biomarker-positive subgroup. In the overall population, high IgG titre AUCs (both uncorrected and corrected for disease severity) were associated with less hippocampal atrophy (r = 0.544, p = 0.0196 and r = 0.476, p = 0.0460); no association was found for whole-brain volume and ventricular volume. In the biomarker-positive subgroup, a similar relationship between high IgG antibody titres and low hippocampal atrophy was observed (r = 0.683, p = 0.050 for uncorrected, and r = 0.750, p = 0.025 for corrected titres); no significant correlation was seen for total brain and ventricular volume (Fig. 7).

To evaluate whether the biological response in the central nervous system (CNS) is proportional to the strength of the AADvac1-induced antibody response, we administered AADvac1 to R3/m4 transgenic mice. These animals develop progressive and ultimately lethal neurofibrillary degeneration. Their antibody response was evaluated (mean EC₅₀, 8088; minimum, 4639; maximum, 12,704), and compared with the number of NFTs in the midbrain. Animals with stronger antibody response had consistently lower tangle counts than those with weak antibody response. The correlation was significant for all three phospho-antibodies (AT8, anti-pS214, anti-pT212) used for evaluation of neurofibrillary pathology (AT8, r = − 0.860, p = 0.006; pS214, r = − 0.747, p = 0.033; pT212, r = − 0.921, p = 0.003). See Additional file 1 for figures.