Tideglusib reduces progression of brain atrophy in progressive supranuclear palsy in a randomized trial

Tideglusib Reduces Progression of Brain Atrophy in ProgressiveSupranuclear Palsy in a Randomized Trial

G€unter U. H€oglinger, MD,1,2,3* Hans-J€urgen Huppertz, MD,4 Stefan Wagenpfeil, PhD,5 Mar�ıa V Andr�es, PhD,6

Vincente Belloch, MD,7 Teresa Le�on, MD, PhD,6 and Teodoro del Ser, MD8 for the TAUROS MRI Investigators

1Department of Neurology, Philipps Universit€at, Marburg, Germany2Department of Neurology, Technische Universit€at M€unchen, Munich, Germany3German Center for Neurodegenerative Diseases (DZNE), M€unchen, Germany

4Swiss Epilepsy Centre, Z€urich, Switzerland5Institute for Medical Biometry, Epidemiology and Medical Informatics (IMBEI), Universit€atsklinikum des Saarlandes, Homburg/Saar, Germany

6Clinical Operations Department, Noscira SA, Madrid, Spain7Scientific Department, Exploraciones Radiol�ogicas Especiales (ERESA), Valencia, Spain

8Medical Department, Noscira SA, Madrid, Spain

ABSTRACT: It is believed that glycogen synthasekinase-3 hyperphosphorylates tau protein in progressivesupranuclear palsy (PSP). The Tau Restoration on PSP(TAUROS) trial assessed the glycogen synthase kinase-3inhibitor tideglusib as potential treatment. For the mag-netic resonance imaging (MRI) substudy reported here,we assessed the progression of brain atrophy. TAUROSwas a multinational, phase 2, double-blind, placebo-con-trolled trial in patients with mild-to-moderate PSP whowere treated with oral tideglusib (600 mg or 800 mgdaily) or with placebo for 1 year. A subset of patientsunderwent baseline and 52-week MRI. Automated,observer-independent, atlas-based, and mask-basedvolumetry was done on high-resolution, T1-weighted,three-dimensional data. For primary outcomes, progres-sion of atrophy was compared both globally (brain, cere-brum) and regionally (third ventricle, midbrain, pons)between the active and placebo groups (Bonferroni cor-rection). For secondary outcomes, 15 additional brainstructures were explored (Benjamini & Yekutieli correc-tion). In total, MRIs from 37 patient were studied (pla-

cebo group, N 5 9; tideglusib 600 mg group, N 5 19;tideglusib 800 mg group, N 5 9). The groups comparedwell in their demographic characteristics. Clinical resultsshowed no effect of tideglusib over placebo. Progressionof atrophy was significantly lower in the active groupthan in the placebo group for the brain (mean 6 stand-ard error of the mean: 21.3% 6 1.4% vs. 23.1% 6

2.3%, respectively), cerebrum (21.3% 6 1.5% vs.23.2% 6 2.1%, respectively), parietal lobe (21.6% 6

1.9% vs. 24.1% 6 3.0%, respectively), and occipitallobe (20.3% 6 1.8% vs. 22.7% 6 3.2%, respectively).A trend toward reduced atrophy also was observed inthe frontal lobe, hippocampus, caudate nucleus, mid-brain, and brainstem. In patients with PSP, tideglusibreduced the progression of atrophy in the whole brain,particularly in the parietal and occipital lobes. VC 2014International Parkinson and Movement Disorder Society

Key Words: progressive supranuclear palsy;randomized controlled clinical trial; volumetric MRI;glycogen synthase kinase-3; tideglusib

------------------------------------------------------------------------------------------------------------------------------Additional Supporting Information may be found in the online version of this article.

*Correspondence to: Dr. G€unter U. H€oglinger, Department of Translational Neurodegeneration, German Center for Neurodegenerative Diseases (DZNE),M€unchen, Max Lebsche Platz 30, D-81677 Munich, Germany; [email protected]

The first two authors contributed equally to this work.

Funding agencies: This study was funded by Noscira SA, Madrid, Spain. Dr. H€oglinger is supported by the Deutsche Forschungsgemeinschaft (DFG,HO2402/6-1).

Relevant conflicts of interest/financial disclosures: Three authors were employees of Noscira SA, Madrid, Spain, the company that funded this study.Full financial disclosures and author roles may be found in the online version of this article.

Members of the TAUROS MRI Investigators are listed in the Appendix.

Received: 29 April 2013; Revised: 1 October 2013; Accepted: 20 December 2013

Published online 31 January 2014 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/mds.25815

R E S E A R C H A R T I C L E

Movement Disorders, Vol. 29, No. 4, 2014 479

Progressive supranuclear palsy (PSP) is a rapidlyprogressive neurodegenerative disease leading to fron-tal lobe dysfunction, supranuclear gaze palsy, pseudo-bulbar palsy, and postural instability. Itsneuropathological hallmark is the somatodendriticaccumulation and aggregation of hyperophosphory-lated tau protein. Glycogen synthase kinase-3 (GSK-3)phosphorylates tau at multiple serine and threonineresidues1 and is thought to be up-regulated2,3 as anupstream event in the cascade of events leading toneurodegeneration in PSP. Hyperphosphorylation oftau reduces its affinity to microtubules, presumablyleading to toxic tau aggregates and to destabilizationof microtubules with ensuing deficiencies in axonaltransport.4 Neuronal and glial deposits of hyperphos-phorylated tau in patients with PSP are immunoreac-tive for GSK-3.5 GSK-3 inhibition reduces tauphosphorylation both in vitro and in vivo.6 Therefore,GSK-3 may be a promising drug target for PSP.

Tideglusib is a thiadiazolidinone7 that irreversiblyinhibits GSK-3 and reduces tau phosphorylation inhuman neuroblastoma cells, murine primary neurons,8

and brains of transgenic mice.9 Tideglusib has beenrecently tested as potential treatment for patients withPSP in the Tau Restoration on PSP (TAUROS) trial.The detailed methods and results of that trial will bereported elsewhere. The main findings showed (1)safety with the exception of some asymptomatic andreversible transaminases elevations, (2) good tolerabil-ity except for rare diarrhea, and (3) no evidence ofclinical efficacy of tideglusib in PSP. In the currentarticle, we report the objective of a nested substudyusing magnetic resonance imaging (MRI) to explorewhether tideglusib versus placebo would reduce theprogression of global or regional brain atrophy during1 year.

Patients and Methods

TAUROS was a double-blind, placebo-controlled,randomized, three parallel-group, multicenter, multi-national, phase 2 trial performed in the EuropeanUnion (Germany, Spain, UK) and the United States toevaluate clinical efficacy, safety, and biomarkers oftwo oral doses (600 mg or 800 mg per day) of tideglu-sib (NP031112) versus placebo in patients with mild-to-moderate PSP. The study protocol (Study CodeNP031112-08B02, EudraCT no. 2009-013097-40,ClinicalTrials.gov no. NCT01049399) was approvedby the corresponding national authorities and ethicscommittees. The Declaration of Helsinki was fol-lowed. Written informed consent was obtained fromeach patient and caregiver. Eligible participants wererandomly allocated to the active or placebo armsaccording to a 2:2:1 ratio, stratified by country. Ran-domization was done in blocks of 10 patients, follow-

ing a sequential list that was generated at random bya contract research organization. Tideglusib (ormatching placebo) was administered orally once dailyat doses of 600 mg or 800 mg for 52 weeks. Medica-tion was dispensed in uni-dose sachets as dry powderfor suspension in water and was administered daily inthe morning after overnight fasting.

Participants

Patients ages 40 to 85 years with possible or proba-ble PSP10 who scored � 4 on the Golbe Staging Sys-tem11; were without other relevant neurologic,psychiatric, or medical illnesses; and were living athome with a caregiver were recruited. A brain MRIwithin 24 months consistent with PSP and ruling outrelevant vascular pathology (Wahlund scale12 � 2)was required. Patients with drug or alcohol abuse orwith sensory or language disability were excluded.Chronic daily intake of neuroleptics (except atypical),anticonvulsants, systemic anticholinergics, acetylcho-linesterase inhibitors, nootropics, centrally active anti-hypertensive drugs, immunosuppressants, steroid ornon-steroid anti-inflammatory drugs, anticoagulants,digitoxin, memantine, or lithium was not allowed.

The MRI substudy reported here was conducted at17 European sites (Spain, 9 sites; Germany, 5 sites;UK, 3 sites) selected for capabilities and willingness tofollow the standard protocol. Participation wasoptional for patients. Requirement of sedation forimage acquisition or the presence of a cardiac pace-maker, defibrillator, cochlear implant, aneurism clip,insulin pump, or surgical staples precluded participa-tion. These conditions limited the sample of patientswho were eligible to participate in this MRI substudy.

Demographic and Clinical Data

Gender, age, disease duration, PSP stage, baselinescores, and changes in scores after 1 year on the PSPRating Scale,13 the Dementia Rating Scale,14 the Fron-tal Assessment Battery,15 the Lexical Verbal Fluency,the Clinical Global Impression Scale,16 and the modi-fied Schwab & England scale17 were recorded. Thesample size of the TAUROS trial was determined bypower analysis (power, 80%; 2-sided significancelevel, 5%) to determine a 41.4% (4-point) differencein progression on the PSP Rating Scale based on pub-lished natural history data.13

Image Acquisition

Images were acquired on 1.5-T or 3-T scanners. Ahead phantom was scanned before baseline and beforepost-treatment imaging. Participants were imagedwithin 4 weeks prior to the baseline visit and againwithin approximately 4 weeks of the last drug intake(actual range, 223 to 39 days). Images were centrally

H €O G L I N G E R E T A L

480 Movement Disorders, Vol. 29, No. 4, 2014

revised by a neuroradiologist (V.B.) for verification ofprotocol adherence and image quality and wereuploaded onto a secure server in pseudonymized Digi-tal Imaging and Communication in Medicine(DICOM) format. For volumetric analysis, a three-dimensional Magnetization Prepared Rapid Acquisi-tion Gradient Echo (MPRAGE) sequence with 1.0 3

1.0 3 1.0 mm isotropic voxels was acquired accordingto the Alzheimer’s Disease Neuroimaging Initiative(ADNI) recommendations (University of California-Los Angeles, Los Angeles, CA, USA; www.loni.ucla.edu/ADNI). Straight axial gradient echo and fast spinecho T2 sequences were obtained to detect white mat-ter or vascular lesions.

Image Analysis

The volumes of 17 structures (total brain, gray/white matter, cerebrum, cerebral cortex, frontal/parie-tal/temporal/occipital lobes, hippocampus, caudatenucleus, cerebellum, brainstem, midbrain, pons,medulla oblongata, third ventricle) and the areas ofmidsagittal planes across 3 structures (midbrain, mid-brain tegmentum, pons) were determined using a fullyautomated, observer-independent method of atlas-based and mask-based volumetry, as described in theonline Supporting Materials. Each baseline or 12-month follow-up data set was processed individuallywith the same protocol. The method had been vali-dated18,19 and applied to neurodegenerative diseasesin cross-sectional and longitudinal studies, demonstrat-ing its sensitivity to detect volume changes within 6months.19 In contrast to other possible approaches (egthe boundary shift integral20), this method providesabsolute volumetric measures instead of only relativechanges. It can therefore be used in cross-sectionaland longitudinal studies without co-registration offollow-up scans to baseline scans.

Statistical Analysis

Demographic and quantitative baseline clinical datawere compared between groups using either one-wayanalysis of variance (ANOVA) with treatment as afactor for continuous variables or the v2 test for non-continuous variables. The score changes after 52weeks of treatment were compared between groupsusing one-way analysis of covariance (ANCOVA) withtreatment as a factor and baseline score as a covariate.For neuroimaging data, a primary analysis was per-formed in areas predefined in the study protocol,including whole (brain, cerebrum) and regional (thirdventricle, midbrain pons) brain atrophy. A secondary,extensive post-hoc analysis examined additional 15cerebral substructures that were not listed in the pro-tocol. The atrophy rates ([baseline volume 2 final vol-ume] / baseline volume 3 100%) of these structureswere compared by using two-sided t tests followed by

Bonferroni correction. This correction, however, isconsidered too conservative when simultaneous meas-urements are interrelated. It does not take intoaccount the expected dependency between atrophy ofdifferent brain regions due to coherent anatomical pat-terns of neurodegenerative changes. Therefore, analy-ses of the additional 15 cerebral substructures weredone using the false discovery rate in multiple testingunder the assumption of mutual dependency of theanalyzed structures, by comparison with two-sidedt tests and correction according to Benjamini andYekutieli21 using the matrix laboratory (MATLAB)script (The MathWorks Inc., Natick, MA, USA; avail-able online at: www.mathworks.com/matlabcentral/fileexchange/27418-benjamini-hochbergyekutieli-procedure-for-controlling-false-discovery-rate). On an explora-tory basis, we also showed the uncorrected P valuesfrom two-sided t tests for all measured structures. Tocontrol for a potential bias of the relative atrophyrate as main dependent variable, the absolute volumechange in every structure was analyzed by ANCOVAusing baseline volume as a covariate. To control forany noise derived from diagnostic uncertainty, themost relevant analyses were repeated after removingfive cases diagnosed as possible PSP, includingpatients with only probable PSP. To control for aputative bias by unbalanced disease duration orstage, atrophy rates as well as volume changes wereanalyzed by ANCOVA using treatment as a factorand disease duration and baseline values as covari-ates, by two-way ANOVA using treatment and dis-ease stage as factors, and by two-way ANCOVAusing treatment and disease stage as factors and dis-ease duration and baseline values as covariates. Thebivariate Pearson correlation (r) was calculated forlinear correlation analysis, and the Fisher r to z testwas used to compare those correlations betweengroups. P values � 0.05 were considered significant.

Results

The Consolidated Standards of Reporting Trials(CONSORT) flow diagram of the TAUROS MRI sub-study is shown in Figure 1. In total, 146 patients wererecruited between December 2009 and August 2011.From patients enrolled in the TAUROS trial whoagreed to participate in the MRI substudy, 41 hadMRI data in sufficient quality. Before disclosing theirgroup assignment, four patients were excluded fromanalyses: participant 02-007 was excluded for adoubtful PSP diagnosis (he was rather young [aged 56years], had a long disease duration [8 years], and hadno MRI abnormality suggestive of PSP), and threeother participants were excluded for technical reasons(participant 05-006 had midbrain over-segmentation,with the pons being partially included; participant

T I D E G L U S I B I N P S P


http://www.loni.ucla.edu/ADNI


http://www.mathworks.com/matlabcentral/fileexchange/27418-benjamini-hochbergyekutieli-procedure-for-controlling-false-discovery-rate



14-010 had movement artifacts in one MRI, leadingto blurred borders between structures and midbrainover-segmentation; in participant 23-001 normaliza-tion and segmentation failed due to extensive atrophyand white matter lesions). Therefore, in total 37,patients were analyzed.

The demographic and clinical parameters at baseline(Table 1) and the field strength of the MRI scannersused did not significantly differ between the groups;only disease duration was longer for the low-dosegroup compared with the other groups. The clinicalscores of the patients at baseline and their changes atthe 52-week follow-up visit (Table 2) showed no sig-nificant differences between the high-dose, low-dose,and either-dose groups versus placebo among partici-pants in this MRI substudy.

The global and regional atrophy rates over the 52-weekfollow-up period are shown in Table 3. The primary anal-ysis demonstrated lower progression of whole-brain atro-phy in the active treatment group (either dose) than in theplacebo group (brain: 21.3% 6 1.4% vs. 23.1% 6

2.3%; cerebrum: 21.3% 6 1.5% vs. 23.2% 6 2.1%).The secondary analysis demonstrated additional signifi-

cant effects for the parietal lobes (21.6% 6 1.9% vs.24.1% 6 3.0%) and the occipital lobes (20.3% 6

1.8% vs. 22.7% 6 3.2%). The exploratory analysis sug-gested further effects of tideglusib versus placebo on theatrophy rates of gray matter (21.0% 6 0.4% vs. 23.0%6 1.0%), cerebral cortex (21.2% 6 0.5% vs. 23.4% 6

0.4%), and frontal lobe (22.0% 6 0.4% vs. 24.0% 6

0.2%), which did not reach significance after correctionfor multiple comparisons in the primary or secondaryanalyses. Other volumes of interest that exhibited strongatrophy in the placebo group over the follow-up periodwere the midbrain (24.4% 6 0.9%), hippocampus(24.3% 6 1.5%), caudate nucleus (24.1% 6 1.3%),and brainstem (23.6% 6 1.0%); the largest percentagechange of all substructures investigated in the placebogroup were observed in the midbrain plane (27.5% 6

1.9%) and the midbrain tegmentum plane (26.7% 6

2.1%); in all of these structures, there was a nonsignifi-cant trend toward reduced atrophy in the tideglusibgroup. When the absolute volume change was analyzedfor every structure controlling for baseline value as acovariate, similar results were obtained. When fivepatients diagnosed with possible PSP were removed to

FIG. 1. This is a Consolidated Standards of Reporting Trials (CONSORT) flow diagram of the Tau Restoration in Progressive Supranuclear PalsyMagnetic Resonance Imaging (TAUROS MRI) substudy.



analyze only those with probable PSP, similar significantdifferences were observed, albeit with lower P values dueto the loss of statistical power (data not shown).

Figure 2 shows average MRIs from patients whoreceived placebo or high-dose tideglusib. In the high-dose tideglusib group, the rate of atrophy was lowerin all investigated substructures (except for white mat-ter) than in the low-dose group, but ANOVA showedno significant difference.

After correcting for disease duration or stage, theoverall (brain, cerebrum) and regional (parietal andoccipital lobes) rate of atrophy (relative to baseline) aswell as the absolute volume change were still signifi-cantly lower in the active group. In the correspondingANCOVA and two-way ANOVA and ANCOVAmodels, treatment always had a significant effect,whereas the covariates disease duration and diseasestage did not. Using the baseline brain, cerebrum, and

parietal lobe volumes as covariates also had a signifi-cant effect in the model, with more atrophic casesshowing higher progression.

The 28 tideglusib-treated patients had a 52-weekfollow-up MRI scan within 23 days before to 39 daysafter the last drug intake: 11 of those patients werescanned during the treatment period (washout, 0days), and 17 were scanned after the termination oftideglusib treatment (washout, 1-39 days). The meanwashout period for all 28 treated patients was 5.8 6

1.6 days. Post-hoc analyses did not suggest that theeffect of active treatment on brain atrophy would bereversible. First, Pearson’s correlation between theduration of the washout phase with the atrophy ratewas not significant in any of the analyzed regions. Sec-ond, the mean rate of brain, cerebrum, parietal, andoccipital atrophy was not significantly different inpatients who were still on treatment at the second

TABLE 1. Baseline characteristics of the participants in the TAUROS MRI substudy

Tideglusib Group

Characteristic Placebo Group 600 mg 800 mg Either Dose

Total no. of patients 9 19 9 28Gender: Men:women 4:5 8:11 7:2 15:13Age at baseline: Mean 6 SEM (range), y 69.9 6 2.3 (56–76) 69.3 6 1.5 (59–85) 67.0 6 3.1 (52–81) 68.6 6 1.4 (52–85)Disease duration: Mean 6 SEM (range), y 1.7 6 0.6 (0.1–5.1) 4.1 6 0.6 (0.2–8.2)a 2.1 6 0.6 (0.0–4.8) 3.4 6 0.5 (0.0–8.2)PSP stage: No. of patients1 0 2 2 42 3 4 73 6 3 124 0 0 5

PSP diagnosis: No. of patientsProbable 6 18 8 26Possible 3 1 1 2

MRI field strength: No. of patient1.5 T 7 13 203.0 T 2 6 8

aP 5 0.038 vs. placebo (analysis of variance followed by post-hoc least-squares difference test).Abbreviations: SEM, standard error of the mean; PSP, progressive supranuclear palsy; MRI, magnetic resonance imaging; T, Tesla.

TABLE 2. Clinical scales at baseline and change at the 52-week follow-up in the TAUROS MRI substudya

Mean 6 SEM

Tideglusib Group, N 5 28

Placebo Group, N 5 9 600 mg, N 5 19 800 mg, N 5 9 Either Dose, N 5 28

Scale Baseline Change Baseline Change Baseline Change Baseline Change

PSP Rating Scale 33.1 6 4 10.9 6 1.5 36.4 6 2.9 10.0 6 1.6 29.2 6 3.7 12.1 6 1.9 34.1 6 2.3 10.6 6 1.3Dementia Rating Scale 121.8 6 3.9 20.6 6 3.6 117.2 6 4.3 24.9 6 3.8 121.6 6 4.3 25.1 6 5.3 118.6 6 3.2 25.0 6 3.1Frontal Assessment Battery 12.3 6 1.4 1.1 6 1.0 11.6 6 0.9 20.3 6 0.8 12.1 6 1.3 0.9 11.8 6 0.8 0.2 6 0.6Lexical Verbal Fluency 12.7 6 2.2 20.6 6 2.4 10.8 6 1.8 20.1 6 0.8 11.1 6 3.3 21.0 6 1.7 10.9 6 1.6 0.4 6 0.8Clinical Global Impressionb 4.0 6 0.3 5.7 6 0.2 4.4 6 0.3 5.3 6 0.2 3.7 6 0.3 5.1 6 0.2 4.1 6 0.2 5.3 6 0.1Modified Schwab & England 63.3 6 6.2 225.6 6 4.4 57.9 6 5.0 214.7 6 3.5 71.1 6 7.4 217.8 6 3.6 62.1 6 4.2 215.7 6 2.6

aNo differences between groups in an analysis of variance test for baseline scores or in an analysis of covariance test with baseline as a covariate for scorechange after 1 year.bThe Clinical Global Impression assesses the severity at baseline and the change of severity after treatment with the same 1 to 7 scale.Abbreviations: SEM, standard error of the mean; PSP, progressive supranuclear palsy.



scan compared with those who were not (data notshown).

In placebo-treated patients, Pearson’s correlationswere significant between the atrophy rates in all mainvolumes of interest (brain, cerebrum, cerebral cortex,frontal/parietal/occipital lobe, brainstem, midbrain,pons; N 5 9; r 5 0.986-0.670; P < 0.0001 to P <0.05), suggesting a coherent pattern of neurodegenera-tion throughout the analyzed structures, with onlytwo exceptions (midbrain with cerebral cortex andfrontal lobe). In contrast, in tideglusib-treated patients,there were two separate clusters: the atrophy rates inbrain, cerebrum, and cerebral lobes correlated witheach other (N 5 28; r 5 0.977-0.336; P < 0.0001 toP 5 0.08); the atrophy rates in brainstem, midbrain,and pons also correlated with each other (N 5 28;r 5 0.935-0.739; P < 0.0001). However, the atrophyrates of the brainstem for most patients did not corre-late significantly with the cerebral structures (N 5 28;r 5 20.017 to 0.343; P 5 0.93 to P 5 0.07), andsome of them were significantly lower (Fisher’s r to ztest), suggesting a differential effect of tideglusib inthe brainstem and cerebrum.

Discussion

This study suggests that a 52-week treatment withthe GSK-3 inhibitor tideglusib reduces the progression

of brain atrophy, particularly in the parietal and occi-pital lobes, in patients with PSP, although withoutdetectable clinical correlation. The main limitation ofthis study is the limited number of analyzed cases andthe lack of specific randomization for this exploratoryMRI substudy, which was nested in the TAUROStrial. However, the main demographic and clinicalcharacteristics were balanced across the treatmentarms. Only disease duration was significantly longer inthe low-dose treatment group. Still, these findings donot appear to have a major impact on the conclusionsof this study, as indicated by the ANCOVA orANOVA models with disease duration or stage as acovariate or factor, respectively. There were also nosignificant differences in the clinical scales at baselineand at 52-week follow-up between the groups. Thus,the subsample of participants analyzed here, as typi-cally observed in this type of substudy,22 is in con-cordance in this regard with the entire TAUROS studypopulation, in which no effect of tideglusib has beenobserved on the rate of progression on the clinicalscales.

A further limitation is heterogeneity in the MRI fieldstrength (1.5 T vs. 3.0 T) and MRI equipment. How-ever, the distribution of field strength was not signifi-cantly different between the treatment arms; and, withrespect to the variability of volumetric results, it is farmore important that the two MRIs in each individualpatient were done with the same scanner and sequence.

TABLE 3. Atrophy rates during the 52-week follow-up period in the TAUROS MRI substudy

Atrophy Rate, %a

Tideglusib Group, N 5 28 Statistical Analysis: P<b

Structure Placebo Group, N 5 9 600 mg 800 mg Either Dose Primary Secondary Exploratory

Brain 23.1 21.4 21.0 21.3 0.035 0.048 0.007Gray matter 23.0 21.2 20.6 21.0 0.111 0.035White matter 23.2 21.6 21.7 21.6 0.139 0.066Cerebrum 23.2 21.4 21.2 21.3 0.044 0.048 0.009Cerebral cortex 23.4 21.4 20.8 21.2 0.117 0.043Frontal lobe 24.0 22.1 21.9 22.0 0.084 0.022Parietal lobe 24.1 21.6 21.4 21.6 0.048 0.005Temporal lobe 22.9 21.3 20.6 21.1 0.178 0.103Occipital lobe 22.7 20.4 20.1 20.3 0.048 0.010Hippocampus 24.3 22.6 20.7 22.0 0.178 0.101Caudate nucleus 24.1 22.8 21.5 22.4 0.198 0.125

Cerebellum 22.6 20.3 20.2 20.3 0.126 0.053Brainstem 23.6 22.8 21.7 22.5 0.277 0.220Midbrain 24.4 23.5 22.3 23.1 0.896 0.262 0.179Midbrain plane 27.5 26.0 24.5 25.5 0.356 0.337Midbrain tegmentum plane 26.8 24.7 24.4 24.6 0.277 0.223Pons 23.4 22.5 21.8 22.2 1.00 0.277 0.233Pons plane 23.2 22.3 21.3 22.0 0.277 0.248Medulla oblongata 22.9 22.7 20.1 21.9 0.495 0.495

Third ventricle 7.7 3.8 6.6 4.7 0.307 — 0.061

aData are presented as mean atrophy rates ([baseline volume 2 final volume] /baseline volume 3 100%). For statistical analyses,the placebo group was com-pared with either tideglusib dose group.The primary analysis was Bonferroni-corrected,the secondary analysis was Benjamini & Yekutieli-corrected (false dis-covery rate), and uncorrected t tests were used for the exploratory analysis.bValues in boldface are for comparisons that indicate statistically significant differences.



FIG. 2. Magnetic resonance imaging (MRI) changes are compared between (A) the placebo group and (B) the 800 mg tideglusib group. The left col-umn shows an overlay of all patients’ brains in the 2 groups as segmented from the baseline MRI, the middle column shows the correspondingoverlay for the follow-up MRI after 12 months, and the right column again shows the overlay from the follow-up MRI with a colored indication ofparenchymal loss relative to baseline. Each red-colored voxel represents a location in which at least one of the patients has lost brain parenchymabetween the baseline and follow-up investigations. As recognizable especially in the enlarged subfigures of the brainstem, the loss is larger in theplacebo group than in the treatment group. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]



http://wileyonlinelibrary.com

For most cerebral structures, this allows recognizingsignificant volume changes in the range from 1% to 3%on the individual patient level and even smaller volumechanges for groups.18 Another limitation with this studyis that we did not take into consideration white matterlesions, which may secondarily affect our measures ofgrey matter tissue volume.

In both the placebo group and the treatment group,the highest atrophy rates of all structures investigatedwere observed in the midsagittal midbrain planes. Thisfits with the expected pattern of atrophy in patients withPSP.19,23-26 Furthermore, this congruence indicates thatthe method of automated atlas-based and mask-basedvolumetry applied in this study is able to detect volumedifferences in a single-digit percentage range for inter-vals as short as 12 months. This is in accordance withformer studies in which similar accuracy was demon-strated even for at the single-patient level.18,26

In contrast to the clinical data, we observed a cleareffect of tideglusib on brain atrophy after the 52-weektreatment period compared with placebo-treatedpatients. The primary per protocol analysis confirmedthe hypothesis that global brain and cerebral atrophywas reduced significantly by about two-thirds. The sec-ondary analysis indicated that atrophy also was reducedin the parietal and occipital lobes. The exploratoryanalysis furthermore suggested also that other regions,such as the frontal lobe, hippocampus, caudate nucleus,midbrain, and brainstem, had a trend toward reducedatrophy. PSP neuropathology affects all of these struc-tures, yet with varying severity.27 There is not a definiteexplanation for the more pronounced effect of tideglu-sib in cerebral structures compared with brainstemstructures, but it is consistent with a stronger expressionof GSK-3 protein in the former (Allen Institute for BrainScience, Seattle, WA: available at: http://human.brain-map.org). Another potential explanation is that volu-metric measures of larger structures are less prone toregistration errors between probabilistic brain atlas andsingle patient brain and, thus, are more able to showsignificant differences.

Although the reduction of atrophy versus placebofor all regions analyzed was more pronounced in the800-mg group compared with the 600-mg group on anumerical basis, the difference was small, and therewas not enough statistical power to prove a dose-response effect in this sample. The discrepancybetween the clear evidence of reduced brain atrophyin the active group versus the placebo group and theabsence of parallel beneficial clinical effects remainselusive. Similar clinico-radiological discrepancies havebeen reported in other clinical trials. The possibilitythat the treatment would protect neuronal perikaryawithout preserving distal axonal connectivity will haveto be addressed in future studies. We cannot rule outwith certainty other unfavorable possibilities, eg that

the treatment would reduce brain volume loss byincreasing inflammatory processes or cerebral watercontent rather than by protecting neurons from dying.However, some secondary analyses that were not con-sidered in the initial study design have been performedto address this issue. Indirect evidence argues againstsuch mechanisms: first, the absence of significant cor-relations between the washout duration and theobserved atrophy rates argues against a rapidly revers-ible effect; and, second, the different pattern ofregional atrophy inter-correlations in the active andplacebo groups points to a stronger effect of the com-pound on cerebral regions than on the brainstem. Inthis regard, it is important to note that the tendencyfor reduced atrophy of tideglusib-treated patients inthe midbrain (ie, the brain region with the highestrelevance for the clinical dysfunction in PSP) did notreach statistical significance.

Finally, we cannot exclude in this limited sample ofpatients the possibility of a nonbalanced, random dis-tribution by treatment; however, the potential effectsof disease duration or diagnosis have been controlledin the statistical analyses. Therefore, the most plausi-ble explanation for the clinico-radiological discrepancyis that the differential regional effect of tideglusib maybe lower in the brain areas, which are most damagedand most relevant for clinical symptoms in PSP.

In summary, this is the first drug trial to evaluateGSK-3 as therapeutic target in human patients with aprimary tauopathy and the first indicative of disease-modifying effects in PSP. Although the cerebral areasthat appear to have responded best to tideglusib, theparietal and occipital lobes, are those least affected byPSP, this does not cast suspicion on the validity of thefindings. Rather, it appears logical that the areaswhere the disease process has progressed least arethose most amenable to neuroprotective intervention.This only supports the possibility that tideglusib isactually slowing progression of the disease process,albeit to a degree not noticeable by clinical historyand examination. The effect of GSK-3 inhibition war-rants further investigation. An intervention at an ear-lier stage of the disease or for longer treatmentduration in well selected patients might lead to meas-urable clinical effects with relevant impact on thepatient’s lives.

Appendix

Tau Restoration on PSP (TAUROS) MRI Investiga-tors: J.C. G�mez, MD, B. Tijero, MD; and R. Villoria,MD (Hospital de Cruces, Barakaldo, Spain; Site Inves-tigator); J. Garc�ıa de Yebenes, MD, and J.L. LopezSend�on, MD (Hospital Ram�on y Cajal, Madrid, Spain;Site Investigators); E. Tolosa, MD, M.T. Buongiorno,MD, and N. Bargall�o, MD (Hospital Clinic,



http://human.brain-map.org

http://human.brain-map.org

Barcelona, Spain; Site Investigators); J.A. Burguera,MD, and I. Martinez, MD (Hospital La Fe, Valencia,Spain; Site Investigators); J. Ruiz-Mart�ınez, MD, andJ. Villanua, MD (Hospital Donostia, San Sebasti�an,Spain; Site Investigator); F. Vivancos, MD, and I.Ybot, MD (Hospital La Paz, Madrid, Spain; SiteInvestigators); M. Aguilar, MD, and J.L. Dolz, MD(Hospital Mutua Terrassa, Terrassa, Spain; Site Inves-tigator); M. Boada, MD, A. Lafuente, MD, and M.A.Tejero, MD (Fundaci�on ACE, Barcelona, Spain; SiteInvestigators); J.J. L�opez-Lozano, MD, and M. Mata,MD (Hospital Puerta de Hierro, Madrid, Spain; SiteInvestigators); A. Kupsch, MD, and A. Lipp, MD(Virchow-Klinikum, Berlin, Germany; Site Investiga-tors); M. H€ollerhage, MD, W.H. Oertel, MD, G.Respondek, MD, M. Stamelou, MD, and S. Knake,MD (Universit€atsklinikum, Marburg, Germany; SiteInvestigators); D. Berg, MD, W. Maetzler, MD, K.K.Srulijes, MD, and A. Gr€oger, MD (Universit€atsklini-kum, T€ubingen, Germany; Site Investigators); A.Ludolph, MD, and J. Kassubek, MD (Universit€atskli-nikum, Ulm, Germany; Site Investigators); M. Steiger,MD, and K. Tyler, MD (Walton Center, Liverpool,UK; Site Investigators); D.J. Burn, MD, and L. Morris,MD (Clinical Ageing Research Unit, Newcastle uponTyne, UK; Site Investigator); A. Lees, MD, H. Ling,MD, and L. Strycharczuk, MD (Reta Lila WestonInstitute, London, UK; Site Investigators).

Acknowledgments: We thank the patients and families for partici-pating and the entire investigator team for recruiting patients for thisstudy. MRI data obtained from the Alzheimer’s Disease NeuroimagingInitiative (ADNI) database (www.loni.ucla.edu/ADNI) have been used toimprove and test the methods of volumetric MRI analysis applied in thisstudy.

References1. Sutherland C. What are the bona fide GSK3 substrates [serial

online]? Int J Alzheimers Dis 2011;2011:505607.

2. Ferrer I, Gomez-Isla T, Puig B, Freixes M, Ribe E, Dalfo E, AvilaJ. Current advances on different kinases involved in tau phospho-rylation, and implications in Alzheimer’s disease and tauopathies.Curr Alzheimer Res 2005;2:3-18.

3. Wang JZ, Grundke-Iqbal I, Iqbal K. Kinases and phosphatases andtau sites involved in Alzheimer neurofibrillary degeneration. Eur JNeurosci 2007;25:59-68.

4. Stamelou M, de Silva R, Arias-Carrion O, et al. Rational therapeu-tic approaches to progressive supranuclear palsy. Brain 2010;133:1578-1590.

5. Ferrer I, Barrachina M, Puig B. Glycogen synthase kinase-3 isassociated with neuronal and glial hyperphosphorylated tau depos-its in Alzheimer’s disease, Pick’s disease, progressive supranuclearpalsy and corticobasal degeneration. Acta Neuropathol 2002;104:583-591.

6. Medina M, Castro A. Glycogen synthase kinase-3 (GSK-3) inhibi-tors reach the clinic. Curr Opin Drug Discov Devel 2008;11:533-543.

7. Martinez A, Alonso M, Castro A, Perez C, Moreno FJ. Firstnon-ATP competitive glycogen synthase kinase 3 beta (GSK-3beta)

inhibitors: thiadiazolidinones (TDZD) as potential drugs for thetreatment of Alzheimer’s disease. J Med Chem 2002;45:1292-1299.

8. Dominguez JM, Fuertes A, Orozco L, del Monte-Millan M,Delgado E, Medina M. Evidence for the irreversible inhibition ofglycogen synthase kinase-3beta by tideglusib. J Biol Chem 2000;287:893-904.

9. Sereno L, Coma M, Rodriguez M, et al. A novel GSK-3b inhibitorreduces Alzheimer’s pathology and rescues neuronal loss in vivo.Neurobiol Dis 2009;35:359-367.

10. Litvan I, Agid Y, Calne D, et al. Clinical research criteria for thediagnosis of progressive supranuclear palsy (Steele-Richardson-Ols-zewski syndrome): report of the NINDS-SPSP International Work-shop. Neurology 1996;47:1-9.

11. Golbe LI. The Medical Advisory Board of the Society for ProgressiveSupranuclear Palsy. A clinical rating scale and staging system forprogressive supranuclear palsy. Neurology 1997;48(suppl):A326.

12. Wahlund LO, Barkhof F, Fazekas F, et al. A new rating scale forage-related white matter changes applicable to MRI and CT.Stroke 2001;32:1318-1322.

13. Golbe LI, Ohman-Strickland PA. A clinical rating scale for pro-gressive supranuclear palsy. Brain 2007;130:1552-1565.

14. Jurica PJ, Leitten CL, Mattis S. Dementia Rating Scale-2 Professio-nal Manual. Lutz, FL: Psychological Assessment Resources; 2001.

15. Dubois B, Slachevsky A, Litvan I, Pillon B. The FAB: a FrontalAssessment Battery at bedside. Neurology 2000;55:1621-1626.

16. Guy W. Clinical Global Impressions: In: Guy W, editor. ECDEUAssessment Manual for Psychopharmacology. Revised DHEW Pub-lication (ADM). Rockville, MD: National Institute for MentalHealth; 1976; 218-222.

17. Schwab RS, England AC. Projection technique for evaluating sur-gery in Parkinson’s disease. In: Gillingahm FJ, Donaldson IML,editors. Third Symposium on Parkinson’s Disease. Edinburgh, UK:Livingston; 1969:152-157.

18. Huppertz HJ, Kroll-Seger J, Kloppel S, Ganz RE, Kassubek J.Intra- and interscanner variability of automated voxel-based volu-metry based on a 3D probabilistic atlas of human cerebral struc-tures. Neuroimage 2010;49:2216-2224.

19. Frings L, Mader I, Landwehrmeyer BG, Weiller C, Hull M,Huppertz HJ. Quantifying change in individual subjects affectedby frontotemporal lobar degeneration using automated longitudi-nal MRI volumetry. Hum Brain Mapp 2011;33:1526-1535.

20. Freeborough PA, Fox NC. The boundary shift integral: an accurateand robust measure of cerebral volume changes from registeredrepeat MRI. Trans Med Imaging 1997;16:623-629.

21. Benjamini Y, Yekutieli D. The control of the false discovery rate inmultiple testing under dependency. Ann Stat 2000;29:1165-1188.

22. Raman R, Thomas RG, Weiner MW, et al. MRI substudy partici-pation in Alzheimer disease (AD) clinical trials: baseline compara-bility of a substudy sample to entire study population. AlzheimerDis Assoc Disord 2009;23:333-336.

23. Oba H, Yagishita A, Terada H, et al. New and reliable MRI diag-nosis for progressive supranuclear palsy. Neurology 2005;64:2050-2055.

24. Quattrone A, Nicoletti G, Messina D, et al. MR imaging index fordifferentiation of progressive supranuclear palsy from Parkinsondisease and the Parkinson variant of multiple system atrophy.Radiology 2008;246:214-221.

25. Morelli M, Arabia G, Novellino F, et al. MRI measurements pre-dict PSP in unclassifiable parkinsonisms: a cohort study. Neurology2011;77:1042-1047.

26. Whitwell JL, Xu J, Mandrekar JN, Gunter JL, Jack CR Jr, JosephsKA. Rates of brain atrophy and clinical decline over 6 and 12-month intervals in PSP: determining sample size for treatment tri-als. Parkinsonism Relat Disord 2012;18:252-256.

27. Williams DR, Holton JL, Strand C, Pittman A, de Silva R, LeesAJ, Revesz T. Pathological tau burden and distribution distin-guishes progressive supranuclear palsy-parkinsonism from Richard-son’s syndrome. Brain 2007;130:1566-1576.




Documents

Tideglusib reduces progression of brain atrophy in progressive supranuclear palsy in a randomized trial