ARTICLE OPEN ACCESS

Brain myoinositol as a potential marker ofamyloid-related pathologyA longitudinal study

Olga Voevodskaya MSc Konstantinos Poulakis MSc Pia Sundgren MD PhD Danielle van Westen MD PhD

Sebastian Palmqvist MD PhD Lars-Olof Wahlund MD PhD Erik Stomrud MD PhD Oskar Hansson MD

PhD and Eric Westman PhD for the Swedish BioFINDER Study Group

Neurologyreg 201992e395-e405 doi101212WNL0000000000006852

Correspondence

Olga Voevodskaya

olgavoevodskayakise

AbstractObjectiveTo investigate the association between longitudinal changes in proton magnetic resonancespectroscopy (MRS) metabolites and amyloid pathology in individuals without dementia andto explore the relationship between MRS and cognitive decline

MethodsIn this longitudinal multiple time point study (a subset of the Swedish BioFINDER) weincluded cognitively healthy participants individuals with subjective cognitive decline andindividuals with mild cognitive impairment MRS was acquired serially in 294 participants (670individual spectra) from the posterior cingulateprecuneus Using mixed-effects models weassessed the association betweenMRS and baseline β-amyloid (Aβ) and betweenMRS and thelongitudinal Mini-Mental State Examination accounting for APOE age and sex

ResultsWhile baseline MRS metabolites were similar in Aβ positive (Aβ+) and negative (Aβminus) indi-viduals in the Aβ+ group the estimated rate of change was +19y for myo-inositol (mI)creatine (Cr) and minus20y for N-acetylaspartate (NAA)mI In the Aβminus group mICr andNAAmI yearly change was minus005 and +12 however this was not significant across timepoints The mild cognitive impairment Aβ+ group showed the steepest MRS changes with anestimated rate of +293y (p = 007) for mICr and minus355y (p lt 001) for NAAmIFurthermore in the entire cohort we found that Aβ+ individuals with low baseline NAAmI hada significantly higher rate of cognitive decline than Aβ+ individuals with high baseline NAAmI

ConclusionWe demonstrate that the longitudinal change in mICr and NAAmI is associated with un-derlying amyloid pathology MRS may be a useful noninvasive marker of Aβ-related processesover time In addition we show that in Aβ+ individuals baseline NAAmI may predict the rateof future cognitive decline

RELATED ARTICLE

EditorialDevelopment of 1H MRSbiomarkers for trackingearly predementiaAlzheimer disease

Page 209

These authors contributed equally as senior authors

From the Division of Clinical Geriatrics (OV KP L-OW EW) Department of Neurobiology Care Sciences and Society Karolinska Institute Stockholm Department of DiagnosticRadiology (PS DvW) Lund University Imaging and Function (DvW) Skaringne University Health Care Lund Clinical Memory Research Unit (SP ES OH) Department of ClinicalSciences Malmo Lund University Memory Clinic (ES OH) Skaringne University Hospital Malmo Sweden and Department of Neuroimaging (EW) Centre for Neuroimaging SciencesInstitute of Psychiatry Psychology and Neuroscience Kingrsquos College London UK

Go to NeurologyorgN for full disclosures Funding information and disclosures deemed relevant by the authors if any are provided at the end of the article

The Article Processing Charge was funded by the Swedish Research Council

This is an open access article distributed under the terms of the Creative Commons Attribution License 40 (CC BY) which permits unrestricted use distribution and reproduction in anymedium provided the original work is properly cited

Copyright copy 2019 The Author(s) Published by Wolters Kluwer Health Inc on behalf of the American Academy of Neurology e395

Accumulation of β-amyloid (Aβ) in the brain is the main hall-mark of Alzheimer disease (AD) and can be detected by ana-lyzing CSF Aβ42 levels In AD CSF Aβ42 might change beforeAβ fibrils become detectable with PET1 and is the earliestdisease biomarker to become abnormal2 Magnetic resonancespectroscopy (MRS) is a noninvasive technique that allowsquantifications of specific brain metabolites in vivo

Recent MRS studies provide evidence of myo-inositol (mI)being the most relevant MRSmetabolite in AD3ndash5 Specificallyelevated mI seems to be related to Aβ plaque pathologymdashhigher baseline levels of mI are associated with increasedAβ accumulation over time5 Another MRS metaboliteconsistently changed in dementia is the neuronal markerN-acetylaspartate (NAA)

In this study we investigate the association between longi-tudinal changes in the ratios mIcreatine (Cr) and NAAmIand amyloid pathology MRS measurements from a singlevoxel located in the posterior cingulate cortex (PCC)precuneus were acquired serially from 294 participants atbaseline and 2- and 4-year follow-up Using information onthe presence of amyloid pathology and the genetic risk factorAPOE e4 we aimed to investigate the extent to which the ratesof change in MRS metabolites mICr and NAAmI are asso-ciated with Aβ pathology Furthermore we explored whetherbaseline levels of NAAmI were able to predict future cognitivedecline

MethodsStandard protocol approvals registrationsand patient consentsAll participants gave written consent to participate in thestudy Ethical approval was given by the ethical committee ofLund University Sweden

ParticipantsThis study only contained data from individuals without de-mentia (characterized as cognitively normal [CTL]) indi-viduals with subjective cognitive decline (SCD) or those withmild cognitive impairment (MCI) The participants stemmedfrom the Swedish BioFINDER (Biomarkers for IdentifyingNeurodegenerative Disorders Early and Reliably) Study (de-tailed information available at biofinderse)

The CTL subset included participants based on the fol-lowing characteristics (1) baseline age 60 years or older

(2) 28ndash30 Mini-Mental State Examination (MMSE) pointsat the time of the screening (3) no subjective cognitive im-pairment and (4) fluent Swedish Participants were excluded ifthere was evidence of significant neurologic or psychiatricdisease MCI or dementia

Participants with SCD and MCI were enrolled from 3outpatient memory clinics Individuals were eligible if they(1) were referred to the memory clinic because of cognitiveimpairment (2) did not fulfill the criteria for any dementiadisorder (3) had an MMSE of 24ndash30 points at the time ofscreening (4) were 60 to 80 years old and (5) were fluentin Swedish Participants were excluded if (1) the causeof their cognitive impairment was clearly other than pro-dromal dementia (2) they had significant somatic diseaseor (3) they refused to undergo neuropsychological as-sessment or lumbar puncture The classification into SCDand MCI was based on a comprehensive battery of neu-ropsychological tests as well as an evaluation by a seniorneuropsychologist

In the current study only participants with baseline CSF andhigh-quality MRS data were eligible A total of 294 partic-ipants were investigated healthy controls (n = 137) from co-hort 1 and patients with SCD (n = 83) and patients with MCI(n = 74) from cohort 2

MRS acquisition and analysisMRS was performed on a 3-tesla Siemens TrioTim scanner(Siemens Medical Solutions Malvern PA) using the point-resolved spectroscopy single-voxel sequence at an echotime of 30 milliseconds (ms) and a repetition time of 2000ms The 2 times 2 times 2 cm3 voxel was situated in the midsagittalPCCprecuneus region This area has been recommendedas the appropriate region for MRS in AD6 and has fre-quently been the region of choice by previous large-scaleMRS studies578 Metabolite levels were quantified relativetotal creatine (creatine + phosphocreatine Cr) concen-tration due to the relative stability of the resonance peak ofCr9 Another marker routinely examined using MRS ischoline (Cho) Although the significance of Cho changesfor AD pathology is unclear we present these data briefly inthe report

MRS spectra were processed using the LCmodel1011 soft-ware and inspected for artifacts and quality A total of 214individual spectra were excluded from the analysis In-dividual spectra were excluded because of insufficient

GlossaryAβ = β-amyloid AD = Alzheimer disease BioFINDER = Biomarkers for Identifying Neurodegenerative Disorders Early andReliably Cho = choline Cr = creatine CTL = cognitively normal control MCI = mild cognitive impairment mI = myo-inositolMMSE =Mini-Mental State ExaminationMRS =magnetic resonance spectroscopyNAA =N-acetylaspartate PCC =posterior cingulate cortex SCD = subjective cognitive decline

e396 Neurology | Volume 92 Number 5 | January 29 2019 NeurologyorgN

quality individuals were excluded when no baseline spec-trum was available as well as when spectra from only onetime point existed (flowchart data available from DryadAdditional Methods doiorg105061dryadrd681sp) Themean FWHM (full width at half maximum) of the remainingspectra was 66 Hz (SD = 16 Hz) mean signal-to-noise ratiowas 24 (SD = 45) LCModel uses estimated SDs (Cramer-Rao lower bounds) as reliability indicators for metaboliteconcentrations the value of SD lt20 is often used asa limit for acceptable reliability However metabolites mINAA and Cho are readily detectable at echo time = 30 msfor mICr NAACr and ChoCr SD was le7 for allspectra

CSF acquisition and analysisThe procedure of CSF collection and analysis followed theAlzheimerrsquos Association flowchart for CSF biomarkers1213

Lumbar CSF samples were collected at 3 centers stored inpolypropylene tubes at minus80degC and analyzed at the same timeusing 2 ELISAs CSF Aβ42 was analyzed by INNOTESTELISAs (Fujirebio Europe Ghent Belgium)14 CSF Aβ42 con-centration le530 ngL was considered abnormal as previouslypublished4

Statistical proceduresCross-sectional between-group differences in MRS data wereassessed using analysis of variance accounting for age sex andAPOE e4 carriership and Tukey honestly significant differ-ence test for post hoc analyses

Longitudinal mICr and NAAmI changes were analyzedusing linear mixed-effects models with subject-specificintercepts Fixed effects were selected depending on theprimary research question In model 1 the fixed effectswere baseline CSF Aβ group (normalabnormal) visitnumber and the Aβ group-by-visit interaction In model 2the fixed effects were baseline clinical diagnosis visit num-ber and the diagnosis-by-visit interaction Models 1 and2 were created using data from all participants Model 3was analogous to model 1 while only using data frompatients with MCI Models 1ndash3 took into account baselineage sex and APOE e4 carriership Annualized percentualchanges in MRS measures were derived from the modelestimates by taking the average of the estimated rates ofmetabolite concentration change between visits 1 and 2 andvisits 1 and 3 The interaction term APOE times visit was notincluded in the final models since it was found to be in-significant in all models for both outcome variables mICrand NAAmI

A similar approach was used in the analysis of longitudinalcognitive data Here we modeled the change in MMSE scoreover time as a function of baseline NAAmI values The fixedeffects were baseline NAAmI (highlow) visit number andthe NAAmI-by-visit interaction When modeling cognitivedecline over time we took into account age sex years ofeducation and APOE e4 status

Since the APOE e2 allele may have a neuroprotective effect ofthe e2 allele 8 participants with genotype e2e4 were ex-cluded from the analyses

Statistical analyses were performed using R (R Foundation forStatistical Computing Vienna Austria r-projectorg)

Data availabilityAnonymized data will be shared by request from any qualifiedinvestigator for the sole purpose of replicating procedures andresults presented in the report provided that data transfer is inagreement with EU legislation on the general data protectionregulation

ResultsStudy sampleA total of 670 individual spectra were acquired averaging ap-proximately 22 spectra per individual for CTL 24 for SCDand 23 for MCI The timing and number of spectra are pre-sented in table 1 Demographic details of the diagnostic groupsare presented in table 2 As expected there were differences incognitive function across the 3 diagnostic groups both atbaseline and follow-up (table 2) Voxel placement and an ex-ample of serially acquired spectra are presented in figure 1 Aand B

Table 1 Study sample demographics

CTL(n = 137)

SCD(n = 83)

MCI(n = 74)

Sex MF 5780 3647 4133

Age y 728 (47) 702 (55) 699 (49)

CSF Aβ42 status Normalabnormal lt530 ngL

9833 5130 2548

APOE genotype laquo4noncarrierslaquo4 carriers

9339 4832 3338

Education y 123 (36) 131 (36)a 116 (34)ab

MMSE score

Baselinec 290 (10) 285 (15) 275 (18)

2-y follow-upd 289 (17) 284 (27) 253 (35)

4-y follow-upef 286 (17) 273 (43) 235 (39)

Abbreviations Aβ42 = β-amyloid 1ndash42 CTL = cognitively normal controlMCI = mild cognitive impairment MMSE = Mini-Mental State ExaminationSCD = subjective cognitive declineValues are reported as mean (SD)a There was a significant difference in the years of education between theSCD and the MCI groups (p = 004)b Education data were missing for 1 patient with MCIc BaselineMMSE data significantly different between all diagnostic groups (plt 001)d MMSE 2-year follow-up scores were significantly different between CTLand MCI groups (p lt 0001)e MMSE 4-year follow-up scores were significantly different between allgroups (p lt 005)f At 4-year follow-upMMSE data weremissing for 47 CTL 26 individuals withSCD and 24 with MCI

NeurologyorgN Neurology | Volume 92 Number 5 | January 29 2019 e397

Cross-sectional MRSMRS values at baseline visit 2 and visit 3 and cross-sectionalcomparisons across diagnostic groups are presented in figure 1C and D and data available from Dryad (Additional Resultsdoiorg105061dryadrd681sp) At baseline none of themetabolite ratios differed between diagnostic groups On visit2 MCI displayed higher mICr and ChoCr values and lowerNAAmI and NAACr than CTL and SCD (p lt 005) Onvisit 3 NAAmI remained significantly lower in MCI thanCTL and SCD (p lt 005) whereas mICr remained higher

for MCI compared to CTL (p lt 005) and NAACr remainedlower for MCI compared to SCD (p lt 005)

When we compared groups based on the individualsrsquo Aβstatus we found no differences at baseline At visits 2 and 3the differences in mICr NAACr and NAAmI were sig-nificant between Aβ-positive and -negative cases (p lt 005)MRS values and cross-sectional comparisons across groupsbased on Aβ status are shown in data available from Dryad(Additional Results doiorg105061dryadrd681sp)

Longitudinal MRSThe rate of change of MRS measures mICr and NAAmIwas estimated using mixed-effect models (table 3)

Model 1 was created using data from the entire cohort in orderto assess whether baseline Aβ status affects the rate of change inmICr and NAAmI accounting for baseline age sex andAPOE e4 carriership Thus the primary parameter of interestfor model 1 was the interaction between Aβ status and visitnumber We found that although higher Aβ was not associatedwith higher mICr levels at baseline the estimates of longitu-dinal change inmICr andNAAmIwere significantly different

Table 2 Visit time and number of acquired spectra

Time from baseliney mean (SD)

Allparticipants CTL SCD MCI

Visit 1 0 294 137 83 74

Visit 2 228 (038) 294 137 83 74

Visit 3 410 (019) 82 24 34 24

Abbreviations CTL = cognitively normal control MCI = mild cognitive im-pairment SCD = subjective cognitive decline

Figure 1 Voxel placement serial MRS acquisition and metabolite levels across diagnoses and biomarker groups

(A) Example of an MRS voxel placement and (B) acquired serial spectra at baseline visit 1 and visit 2 from a 73-year-old cognitively healthy woman (CndashF)Serial spectroscopic data across diagnostic groups (C) mICr ratio levels (D) NAACr ratio levels (E) NAAmI ratio levels (F) ChoCr ratio levels Significantbetween-group differences (reported at p lt 005) were obtained using analysis of variance accounting for age sex and APOE e4 carriership with Tukeyhonestly significant difference tests for post hoc comparisons Aβ+ = CSF Aβ42 le530 ngL Aβminus = CSF Aβ42 gt530 ngL Aβ = β-amyloid Cho = choline Cr =creatine CTL = cognitively normal control MCI = mild cognitive impairment mI = myo-inositol MRS = magnetic resonance spectroscopy NAA = N-acety-laspartate SCD = subjective cognitive decline

e398 Neurology | Volume 92 Number 5 | January 29 2019 NeurologyorgN

Table 3 Specifications of the mixed models estimating the rate of change of MRS measures mICr and NAAmI

Fixed effects

Outcome measure mICr Outcome measure NAAmI

Estimate (SE) p Value Estimate (SE) p Value

Model 1 Estimated rate of change in MRSmetabolites as a function ofbaseline Aβ status in the entire cohort

Intercept 096 (008) lt0001 143 (020) lt0001

Aβ 002 (002) 018 minus004 (004) 029

Visit 2 minus001 (001) 021 007 (037) 0001

Visit 3 002 (001) 024 005 (002) 015

Age minus000 (000) 009 000 (000) 020

APOE 002 (001) 016 minus007 (003) 0045

Sex minus003 (001) lt001 minus007 (033) 005

Aβ times visit 2 005 (001) lt0001 minus016 (003) lt0001

Aβ times visit 3 004 (002) lt005 minus016 (005) lt001

Model 2 Estimated rate of change in MRSmetabolites as a function ofbaseline diagnoses in the entire cohort

Intercept 088 (008) lt0001 161 (021) lt0001

Visit 2 minus001 (001) 052 004 (002) 009

Visit 3 minus001 (002) 069 008 (005) 011

SCD 001 (002) 070 minus002 (004) 061

MCI 002 (002) 035 minus007 (004) 013

Age minus000 (000) 045 000 (000) 068

APOE 004 (001) lt001 minus011 (003) lt0001

Sex minus003 (001) lt001 005 (003) 008

SCD times visit 2 001 (002) 071 minus000 (004) 091

SCD times visit 3 003 (003) 018 minus007 (007) 030

MCI times visit 2 005 (002) lt001 minus012 (004) lt001

MCI times visit 3 009 (003) lt001 minus020 (007) lt001

Model 3 Estimated rate of change in MRSmetabolites as a function ofbaseline Aβ status in the MCI group

Intercept 100 (018) lt0001 173 (040) lt0001

Aβ 005 (003) 019 minus004 (008) 056

Visit 2 002 (002) 030 001 (004) 072

Visit 3 003 (003) 033 009 (006) 016

Age minus000 (000) 031 minus000 (001) 082

APOE 001 (031) 068 minus005 (068) 049

Sex minus003 (003) 020 003 (006) 059

Aβ times visit 2 004 (002) 007 minus015 (005) lt001

Aβ times visit 3 007 (004) 006 minus031 (008) lt0001

Abbreviations Aβ = β-amyloid Cr = creatine MCI = mild cognitive impairment mI = myo-inositol MRS = magnetic resonance spectroscopy NAA = N-acetylaspartate SCD = subjective cognitive declineBaseline levels of factor variables in themodels Aβ = normal visit number = visit 1 sex =male APOE = e4 noncarrier The primary predictors of interest werethe interaction terms Aβ times visit for model 1 and model 3 and diagnosis times visit for model 2

NeurologyorgN Neurology | Volume 92 Number 5 | January 29 2019 e399

depending on the subjectsrsquo baseline Aβ status (figure 2 A andB table 3) In the Aβ+ group mICr increased significantlywith time at an estimated rate of 19y whereas in the Aβminus

group the increase was minimal at 005y The change in thecomposite ratio NAAmI also followed significantly differenttrajectories over time depending on the baseline Aβ status

Figure 2 Estimated rates of change in mI measures across different biomarker and diagnostic groups

Estimated rates of change inmICr andNAAmI in different biomarker and diagnostic groups Model 1 was created using data from the entire cohort to assess whetherbaseline Aβ status affects the rate of change in (A) mICr and (B) NAAmI Model 2 was built using data from the entire cohort in order to examine whether baselinediagnosisaffects therateofchange in (C)mICrand (D)NAAmIModel3onlyuseddata fromtheMCIgroupas inputexamining therateof change in (E)mICrand (F)NAAmI Based on these models we present estimated trajectories of change over time in (G) mICr and (H) NAAmI for a 71-year-old man who is an APOE e4 carrier Aβ =β-amyloidCr= creatineCTL= cognitivelynormal controlMCI=mild cognitive impairmentmI=myo-inositolNAA=N-acetylaspartate SCD= subjectivecognitivedecline

e400 Neurology | Volume 92 Number 5 | January 29 2019 NeurologyorgN

NAAmI declined with the rate of 20y in the Aβ+ groupIn the Aβminus group we detected a 12 yearly increase inNAAmI However the increase was only significant be-tween visits 1 and 2

Model 2 was built using data from the entire cohort in orderto examine whether baseline diagnosis affects the rate ofchange in mICr and NAAmI accounting for baseline agesex and APOE e4 carriership irrespective of baseline Aβstatus Therefore the primary predictor of interest for model2 was the interaction between baseline diagnosis and visitnumber The only significant change over time was found inthe MCI groupmdashhere mICr increased by an average of23y whereas NAAmI decreased by 20y (figure 2 Cand D table 3)

To further characterize these changes we created model 3using only data from theMCI group as input As expected therates of change were highest in the MCI Aβ+ group at 293y (p = 007) for mICr and minus355y (p lt 001) for NAAmI(figure 2 E and F table 3)

For an overview of the differences in the rates of change inmICr and NAAmI we present the estimated MRS trajec-tories for a 71-year-old male APOE e4 carrier for the differentAβ groups and diagnoses (figure 2 G and H)

MRS and cognitive declineTo investigate whether baseline MRS measurements may beuseful for predicting future cognitive decline we went on tocreate a mixed-effects model using data from the entire co-hort where serial MMSE data were predicted as a function of

baseline NAAmI Thus the main predictor of interest was theinteraction termbetweenNAAmI and visit numberWe foundthat in participants who were Aβminus at baseline the decline inMMSE over 4 years was negligible and effectively independentof the baseline NAAmI However for Aβ+ participants thetrajectory of cognitive decline differed significantly dependingon baseline NAAmI (p lt 001) (figure 3)

Individuals who were Aβ+ and had low NAAmI at baseline(lt15) declined by a total of 63 MMSE points (31 pointsbetween visits 1 and 2 and 32 points between visits 2 and 3)Individuals who were Aβ+ at baseline yet had higher baselineNAAmI (ge15) lost a total of 22 MMSE points (06 be-tween visits 1 and 2 and 16 between visits 2 and 3)

The low-NAAmI group comprised 25 healthy controls 22individuals with SCD and 22 individuals with MCI The high-NAAmI group comprised 6 controls 6 SCD and 19 MCIThe relationship between baseline MRS and longitudinalcognition was not significant when the diagnostic groups wereanalyzed separately

Furthermore we investigated whether the absolute change inNAAmI during the time between baseline and the last visitwas associated with the absolute change in MMSE during thistime We found a significant positive association betweenDNAAmI andDMMSE in the Aβ+ group but not in the Aβminusgroup In a linear regression model (built using data from theAβ+ group) where DNAAmI was used to estimate the out-come DMMSE accounting for sex age APOE and educationlevel the variable DNAAmI was found to be the only signif-icant predictor of the DMMSE (p lt 005) The association

Figure 3 Baseline NAAmI and cognitive decline over time

Cognitive decline as a function of baseline NAAmI SerialMMSE data were predicted as a function of baseline NAAmIvalues We found a negligible decline in MMSE score for par-ticipants who were Aβminus at baseline For Aβ+ participants thetrajectory of the decline in MMSE score differed significantlydepending on baseline NAAmI Aβ+ individuals with lowbaseline NAAmI (lt15) declined by a total of 63 MMSE pointsCSF Aβ+ individuals with high NAAmI (ge15) lost a total of 22MMSE points Aβ = β-amyloid mI = myo-inositol MMSE = Mini-Mental State Examination NAA = N-acetylaspartate

NeurologyorgN Neurology | Volume 92 Number 5 | January 29 2019 e401

between DNAAmI and DMMSE is presented in figure 4 inthe Aβ+ group a larger decline in MMSE is associated witha greater negative change (ie greater decline) in NAAmI

DiscussionThe current study highlights the link between longitudinalchanges in MRS metabolites and the main pathologic hall-mark of ADmdashamyloid accumulation A total of 670 individualhigh-quality spectra were analyzed for this studymdasha numberthat far surpasses previous 3-tesla MRS study sample sizes Ingeneral few studies have assessed serial MRS measurementsin the context of AD315ndash17 of which only 2 included patientswith MCI1617 and none have reported information regardingamyloid pathology Based on longitudinal MRS data froma diagnostically well-characterized cohort of 294 participantsincluding controls and individuals with SCD and MCI ourstudy provides evidence of mICr and NAAmI being po-tentially useful markers of Aβ-related pathology Longitudinalspectroscopic changes were evaluated in the PCCprecuneusThis region is highly involved throughout the progression ofAD pathology18ndash20 and has recently been demonstrated as theearliest site for amyloid accumulation21

The association between increased amyloid deposition andelevated levels of mI has previously been described cross-sectionally in vivo48 in human tissue22 and murine models23

A recent study also demonstrated that higher baseline levels ofmI are associated with an increased rate of Aβ accumulationover time in cognitively healthy individuals5 In the present

longitudinal MRS study we expand on these findings bydemonstrating that abnormal Aβ levels at baseline are pre-dictive of elevated mICr and NAAmI concentrations overtime and examine this relationship in different clinical di-agnoses One previous multiple time-point MRS study reportsthat only the change in the combined ratio NAAmI was sig-nificantly different between controls and patients with AD Theauthors reported a 37 yearly decline of NAAmI in patientswith AD and effectively no change in controls24 Although ourstudy did not include patients with a baseline AD diagnosis wefound a similar value for the annualized change in NAAmI inthe Aβ+ MCI group (36 yearly decline) which is coherentwith this group being most similar to the AD patient group A2 time-point study that included patients with MCI (but didnot report NAAmI values) found a 26 annual mICr in-crease for patients with MCI subsequently converting to ADand 0 for MCI with stable cognitive function We founda similar rate of change in theMCI group (23y) In the Aβ+MCI group the composite ratio NAAmI yielded a yearlydecline of 355 which is comparable to the reported rates ofhippocampal atrophy in MCI25 Therefore the ratio NAAmImay be useful for monitoring disease progression also becauseworking with this composite marker eliminates the need forinternal referencing

The signal provided by MRS cannot explicitly determine thetype of cell from which the detected resonance originatesHowever since NAA is known to be synthesized exclusivelyin the neuronal mitochondria there is some consensus withregard to NAA being a suitable marker of neuronal integrityand mitochondrial function2627 There is also evidence ofNAA recovery as cerebral energy balance is restored in trau-matic brain injury28 This sensitivity to transient neuronaldysfunction probably limits the usefulness of NAA alone orNAACr as a biomarker in AD The distribution of mI amongdifferent cell types in the brain is less well understood Gen-erally glial cells seem to contain higher concentrations ofmI2930 but some neuronal cell lines have also been shown tocontain high mI levels31 Although brain mI levels are elevatedin a number of conditions associated with gliosis mI can alsoincrease in absence of glial activation32 One of the most im-portant findings of the current study is the substantial differ-ence observed in the rate of mI increase between the Aβminus andAβ+ groups Several mechanisms (independent or interlinked)may be responsible for this phenomenon A known physiologicrole of mI is that of an organic osmolyte33 intracellular mIconcentrations increase as a response to extracellular hyper-tonicity34 and decrease in hypotonic conditions35 Thus it ispossible that the longitudinal accumulation of mI in the Aβ+detects osmoregulationmdashan effort to resist homeostatic aber-rations associated with pathology A recent study demonstratedthat mI accumulation in response to hypertonic stress is able todisturb electrical properties of excitable cells36 suggesting thatincrease in mI observed in our study may be linked to synapticdysfunction independently of Aβ However there is con-vincing evidence of a relationship between Aβ plaque pa-thology and mI Brain mI is elevated at predementia stages of

Figure 4 Absolute change in NAAmI and MMSE over 4years

A larger decline in MMSE score is associated with a greater negative change(decline) in NAAmI in Aβ+ participants but not in Aβminus participants Aβ =β-amyloidmI =myo-inositol MMSE =Mini-Mental State Examination NAA =N-acetylaspartate

e402 Neurology | Volume 92 Number 5 | January 29 2019 NeurologyorgN

Down syndrome37 and familial AD38 and antemortem mIlevels are associated with postmortem cored and diffuse am-yloid plaques23 Furthermore removal of Aβ plaques attenu-ates mI levels in APP-PS1 mice39 A recent retrospectivelongitudinal MRS study found that individuals who had pro-gressed to AD at 7-year follow-up already had higher mICrlevels at baseline than those who had remained stable40 It istherefore probable that the increase in mICr in the Aβ+group observed in this study is indicative of the ongoing ac-cumulation of cortical Aβ burden within this group A serialamyloid PET study performed in conjunction with serial MRSwould confirm this hypothesis

It has been shown that individuals with high baseline load ofAβ are at risk of a steeper cognitive decline41 Also there isrecent evidence suggesting that MRS measurements havethe potential to identify individuals at risk of a more rapidlyprogressing AD pathology5 This evidence of an interplaybetween MRS concentrations amyloid accumulation andcognition led us to explore the potential of baseline mINAAto predict the trajectory of future cognitive decline Ourresults demonstrate that in the entire sample Aβ+ individ-uals with low baseline NAAmI display a steeper cognitivedecline trajectory losing on average 4 more MMSE points intotal than those with higher baseline NAAmI levels Ofinterest this apparent ability of MRS metabolites to predictthe extent of future decline was only evident in the subgroupof individuals who were Aβ+ at baseline We demonstratethat additional differentiation of Aβ+ individuals based on anMRS threshold may reveal subgroups with markedly differentrates of cognitive decline over time The composite ratio NAAmI has previously been demonstrated to be a good predictor ofan AD diagnosis1640 Our finding expands on this knowledgeby demonstrating that NAAmI may be useful for predictingthe extent of cognitive changes (one aspect of disease severity)at an earlier disease stage The composite ratio of NAA andmI in AD is emerging as the most useful MRS marker in ADsince this measure may incorporate 2 of the main diseasehallmarksmdashamyloid pathology and neurodegeneration

Although our findings support the assumption of the existenceof an AD-specific MRS metabolic signature it is important torecognize that this study does not aim to evaluate the usefulnessof MRS as a marker of Aβ positivity nor does it assess theability of MRS alone to predict cognitive decline in the generalpopulation Rather we highlight the value in combining CSFand MRS data and demonstrate that MRS may provide newinformation regarding the rate of future cognitive decline onceamyloid positivity has been established The nature of our studypopulation is such that the existence of comorbidities cannotbe ruled out It is therefore possible that some of the alterationsin the MRS profile may be due to other causes Althoughabnormal CSF amyloid measures provide a good indication ofongoing AD pathology in an ideal setting information aboutthe timing and the number of individuals who convert to ADshould also be used Furthermore the lack of normativespectroscopic data or a clear understanding of the biological

processes causing MRS alterations corroborate the need forCSF measures in order to interpret any MRS changes

The design of our study was such that a cutoff value has beenused to stratify participants into Aβminus and Aβ+ based on base-line CSF Aβ42 levels Furthermore one of our results suggeststhat there may exist some cutoff value for NAAmI that isrelevant for predicting a worsening of cognitive symptomsThere are certain disadvantages to working with dichotomizedbiomarkers such as the possible masking of subthresholdeffects and the assessment of individuals close to the cut pointNevertheless applying a normalabnormal biomarker classifi-cation is an approach that apart from being practical is es-sential for determining eligibility for clinical trials42 Howeverwe recognize that a putative cutoff value for NAAmI whichmay have relevance for determining an individualrsquos cognitivedecline profile must be scrutinized by future studies and vali-dated in even larger longitudinal cohorts

The inclusion criteria of the BioFINDER cohort and the finalconstellation of our study sample may limit the generalizabilityof the results of this study The potential sources of selectionbias in the recruitment of the participants to the BioFINDERmay stem from the relatively high cognitive scores at entry andthe requirement to undergo a lumbar puncture and an MRIHowever although study participants might be healthier thanthe general population it is unclear in which direction thiswould affect the associations found in this study

In conclusion our findings demonstrate that the longitudinalchanges in mICr and NAAmI are largely governed by thepresence of underlying amyloid pathology warranting theirpotential usefulness as noninvasive dynamic disease biomarkersduring the predementia stages of AD Today amyloid andtau deposition can be imaged using PET allowing us to assessmolecular pathology in vivo However the need remains fora more widely available cost-effective technique that can beused for screening dementia and monitoring disease pro-gression and treatment effects in a clinical setting Large-scalemultimodal studies that include MRS will help further locatespectroscopic changes in the continuum of AD pathophysio-logic processes

AcknowledgmentThe authors thank the collaborators of this study and theSwedish BioFINDER Study Group

Study fundingWork at the authorsrsquo research centers was supported by theEuropean Research Council the Swedish Research Councilthe Marianne and Marcus Wallenberg Foundation the Stra-tegic Research Area MultiPark (Multidisciplinary Research inParkinsonrsquos Disease) at Lund University the Swedish BrainFoundation The Kockska Foundation the Bundy AcademyThe Swedish Alzheimer Foundation the Skaringne UniversityHospital Foundation the Swedish Alzheimer Association andthe Swedish federal government under the ALF agreement

NeurologyorgN Neurology | Volume 92 Number 5 | January 29 2019 e403

Swedish Brain Power the Strategic Research Programmein Neuroscience at Karolinska Institutet (StratNeuro) theSwedish Foundation for Strategic Research (SSF) the regionalagreement on medical training and clinical research (ALF)between Stockholm County Council and Karolinska Institutetthe Aringke Wiberg Foundation and the foundation Olle EngkvistByggmastare The authors also thank Birgitta och Sten West-erberg for additional financial support

DisclosureO Voevodskaya K Poulakis P Sundgren D van WestenS Palmqvist LWahlund E Stomrud andEWestman report nodisclosures relevant to the manuscript O Hansson has acquiredresearch support (for the institution) from Roche GE Health-care Biogen AVID Radiopharmaceuticals Fujirebio andEUROIMMUN In the past 2 years he has received consultancyspeaker fees (paid to the institution) from Lilly Roche andFujirebio Go to NeurologyorgN for full disclosures

Publication historyReceived by Neurology March 12 2018 Accepted in final formSeptember 18 2018

References1 Palmqvist S Mattsson N Hansson O Alzheimerrsquos Disease Neuroimaging Initiative

Cerebrospinal fluid analysis detects cerebral amyloid-β accumulation earlier thanpositron emission tomography Brain 20161391226ndash1236

2 Bateman RJ Xiong C Benzinger TLS et al Clinical and biomarker changes indominantly inherited Alzheimerrsquos disease N Engl J Med 2012367795ndash804

3 Schott J Bartlett J Fox N Barnes J Alzheimerrsquos Disease Neuroimaging InitiativeInvestigators Increased brain atrophy rates in cognitively normal older adults withlow cerebrospinal fluid Aβ1ndash42 Ann Neurol 201068825ndash834

4 Voevodskaya O Sundgren PC Strandberg O et al Myo-inositol changes precedeamyloid pathology and relate to APOE genotype in Alzheimer disease Neurology2016861754ndash1761

5 Nedelska Z Przybelski SA LesnickTG et al 1H-MRSmetabolites and rate of β-amyloidaccumulation on serial PET in clinically normal adults Neurology 2017891391ndash1399

6 Oz G Alger JR Barker PB et al Clinical proton MR spectroscopy in central nervoussystem disorders Radiology 2014270658ndash679

7 Gomar JJ Gordon ML Dickinson D et al APOE genotype modulates protonmagnetic resonance spectroscopy metabolites in the aging brain Biol Psychiatry201375686ndash692

8 Kantarci K Lowe V Przybelski SA et al Magnetic resonance spectroscopy β-amyloidload and cognition in a population-based sample of cognitively normal older adultsNeurology 201177951ndash958

9 Valenzuela M Sachdev P Magnetic resonance spectroscopy in AD Neurology 200156592ndash598

10 Provencher SW Estimation of metabolite concentrations from localized in vivoproton NMR spectra Magn Reson Med 199330672ndash679

11 Provencher SW Automatic quantitation of localized in vivo 1H spectra with LCModelNMR Biomed 200114260ndash264

12 Blennow K Hampel H Weiner M Zetterberg H Cerebrospinal fluid and plasmabiomarkers in Alzheimer disease Nat Rev Neurol 20106131ndash144

13 Palmqvist S Zetterberg H Blennow K et al Accuracy of brain amyloid detection inclinical practice using cerebrospinal fluid beta-amyloid 42 a cross-validation studyagainst amyloid positron emission tomography JAMA Neurol 2014711282ndash1289

14 Vandermeeren M Mercken M Vanmechelen E et al Detection of proteins in normaland Alzheimerrsquos disease cerebrospinal fluid with a sensitive sandwich enzyme-linkedimmunosorbent assay J Neurochem 1993611828ndash1834

15 Dixon RM Bradley KM Budge MM Styles P Smith AD Longitudinal quantitativeproton magnetic resonance spectroscopy of the hippocampus in Alzheimerrsquos diseaseBrain 20021252332ndash2341

16 Kantarci K Weigand SD Petersen RC et al Longitudinal 1H MRS changes in mildcognitive impairment and Alzheimerrsquos disease Neurobiol Aging 2007281330ndash1339

17 Modrego PJ Fayed N Sarasa M Magnetic resonance spectroscopy in the predictionof early conversion from amnestic mild cognitive impairment to dementia a pro-spective cohort study BMJ Open 20111e000007

18 Braak H Braak E Neuropathological stageing of Alzheimer-related changes ActaNeuropathol 199182239ndash259

19 Lehmann M Rohrer JD Clarkson MJ et al Reduced cortical thickness in the pos-terior cingulate gyrus is characteristic of both typical and atypical Alzheimerrsquos diseaseJ Alzheimers Dis 201020587ndash598

20 Minoshima S Giordani B Berent S Frey K Foster N Kuhl D Metabolic reductionin the posterior cingulate cortex in very early Alzheimerrsquos disease Ann Neurol19974285ndash94

21 Palmqvist S Scholl M Strandberg O et al Earliest accumulation of β-amyloid occurswithin the default-mode network and concurrently affects brain connectivity NatCommun 201781214

22 Kantarci K Knopman DS Dickson DW et al Alzheimer disease postmortem neu-ropathologic correlates of antemortem 1H MR spectroscopy metabolite measure-ments Radiology 2008248210ndash220

23 Murray ME Przybelski SA Lesnick TG et al Early Alzheimerrsquos disease neuropa-thology detected by proton MR spectroscopy J Neurosci 20143416247ndash16255

24 Schott JM Frost C MacManus DG Ibrahim F Waldman AD Fox NC Short echotime proton magnetic resonance spectroscopy in Alzheimerrsquos disease a longitudinalmultiple time point study Brain 20101333315ndash3322

25 Henneman WJ Sluimer JD Barnes J et al Hippocampal atrophy rates in Alzheimerdisease added value over whole brain volume measures Neurology 200972999ndash1007

26 Dautry C Franccediloise V Emmanuel B et al Early N-acetylaspartate depletion is a markerof neuronal dysfunction in rats and primates chronically treated with the mitochondrialtoxin 3-nitropropionic acid J Cereb Blood Flow Metab 200020789ndash799

27 Moffett JR Ross B Arun P Madhavarao CN Namboodiri AMA N-acetylaspartate inthe CNS from neurodiagnostics to neurobiology Prog Neurobiol 20078189ndash131

28 Signoretti S Marmarou A Aygok GA Fatouros PP Portella G Bullock RM As-sessment of mitochondrial impairment in traumatic brain injury using high-resolutionproton magnetic resonance spectroscopy J Neurosurg 200810842ndash52

29 Glanville NT Byers DM Cook HW Spence MW Palmer FB Differences in themetabolism of inositol and phosphoinositides by cultured cells of neuronal and glialorigin Biochim Biophys Acta 19891004169ndash179

30 Brand A Richter-Landsberg C Leibfritz D Multinuclear NMR studies on the energymetabolism of glial and neuronal cells Dev Neurosci 199315289ndash298

31 Fisher S Novak J Agranoff B Inositol and higher inositol phosphates in neuraltissues homeostasis metabolism and functional significance J Neurochem 200282736ndash754

32 Bartha R SmithM Rupsingh R Rylett J Wells JL Borrie MJ High field (1)HMRS ofthe hippocampus after donepezil treatment in Alzheimer disease Prog Neuro-psychopharmacol Biol Psychiatry 200832786ndash793

33 Kwon HM Yamauchi A Uchida S et al Cloning of the cDNa for a Na+myo-inositolcotransporter a hypertonicity stress protein J Biol Chem 19922676297ndash6301

34 Lee JH Arcinue E Ross BD Organic osmolytes in the brain of an infant withhypernatremia N Engl J Med 1994331439ndash442

35 Videen JS Michaelis T Pinto P Ross BD Human cerebral osmolytes during chronichyponatremia a proton magnetic resonance spectroscopy study J Clin Invest 199595788ndash793

36 Dai G Yu H Kruse M Traynor-Kaplan A Hille B Osmoregulatory inositol trans-porter SMIT1 modulates electrical activity by adjusting PI(45)P2 levels Proc NatlAcad Sci USA 2016113E3290ndashE3299

37 Huang W Alexander GE Daly EM et al High brain myo-inositol levels in thepredementia phase of Alzheimerrsquos disease in adults with Downrsquos syndrome a 1HMRSstudy Am J Psychiatry 19991561879ndash1886

38 Godbolt AK Waldman AD MacManus DG et al MRS shows abnormalities beforesymptoms in familial Alzheimer disease Neurology 200666718ndash722

39 Marjanska MWeigand SD Preboske G et al Treatment effects in a transgenic mousemodel of Alzheimerrsquos disease a magnetic resonance spectroscopy study after passiveimmunization Neuroscience 201425994ndash100

40 Waragai M Moriya M Nojo T Decreased N-acetyl aspartatemyo-inositol ratio inthe posterior cingulate cortex shown by magnetic resonance spectroscopy may be oneof the risk markers of preclinical Alzheimerrsquos disease a 7-year follow-up studyJ Alzheimers Dis 2017601411ndash1427

41 Donohue MC Sperling RA Petersen R et al Association between elevated brainamyloid and subsequent cognitive decline among cognitively normal persons JAMA20173172305ndash2316

42 Jack CR Jr Wiste HJ Weigand SD et al Defining imaging biomarker cut points forbrain aging and Alzheimerrsquos disease Alzheimers Dement 201713205ndash216

e404 Neurology | Volume 92 Number 5 | January 29 2019 NeurologyorgN

Appendix 1 Authors contribution

Name Location Role Contribution

OlgaVoevodskayaMSc

Division of Clinical Geriatrics Department of Neurobiology CareSciences and Society Karolinska Institute Stockholm Sweden

Author Study design and concept analysis andinterpretation of data statistical analysis draftingthe manuscript

KonstantinosPoulakis MSc

Division of Clinical Geriatrics Department of Neurobiology CareSciences and Society Karolinska Institute Stockholm Sweden

Author Statistical analysis and interpretation of datadrafting the manuscript

Pia SundgrenMD PhD

Department of Diagnostic Radiology Lund University Sweden Author Acquisition and interpretation of data criticalrevision of the manuscript for importantintellectual content

Danielle vanWesten MDPhD

Department of Diagnostic Radiology Lund University and Imaging andFunction Skaringne University Health Care Lund Sweden

Author Acquisition and interpretation of data criticalrevision of the manuscript for importantintellectual content

SebastianPalmqvistMD PhD

Clinical Memory Research Unit Department of Clinical Sciences MalmoLund University Sweden

Author Acquisition and interpretation of data criticalrevision of the manuscript for importantintellectual content

Lars-OlofWahlund MDPhD

Division of Clinical Geriatrics Department of Neurobiology CareSciences and Society Karolinska Institute Stockholm Sweden

Author Interpretation of data critical revision of themanuscript for important intellectual content

Erik StomrudMD PhD

Clinical Memory Research Unit Department of Clinical Sciences MalmoLund University and Memory Clinic Skaringne University Hospital MalmoSweden

Author Acquisition and interpretation of data criticalrevision of the manuscript for importantintellectual content

OskarHansson MDPhD

Clinical Memory Research Unit Department of Clinical Sciences MalmoLund University and Memory Clinic Skaringne University Hospital MalmoSweden

Author Study design and concept drafting themanuscript study supervision

Eric WestmanPhD

Division of Clinical Geriatrics Department of Neurobiology CareSciences and Society Karolinska Institute Stockholm Sweden andDepartment of Neuroimaging Centre for Neuroimaging SciencesInstitute of Psychiatry Psychology and Neuroscience Kingrsquos CollegeLondon UK

Author Study design and concept drafting themanuscript study supervision

Appendix 2 Coinvestigators

Name Location Role Contribution

Kaj Blennow MDPhD

Clinical Neurochemistry Laboratory University ofGothenburg Sweden

Coinvestigator BioFINDERcollaborator

Responsible for the analysis of CSFbiomarkers

Henrik ZetterbergMD PhD

Clinical Neurochemistry Laboratory University ofGothenburg Sweden

Coinvestigator BioFINDERcollaborator

Responsible for the analysis of CSFbiomarkers

NeurologyorgN Neurology | Volume 92 Number 5 | January 29 2019 e405

DOI 101212WNL0000000000006852201992e395-e405 Published Online before print January 4 2019Neurology

Olga Voevodskaya Konstantinos Poulakis Pia Sundgren et al study

Brain myoinositol as a potential marker of amyloid-related pathology A longitudinal

This information is current as of January 4 2019

ServicesUpdated Information amp

httpnneurologyorgcontent925e395fullincluding high resolution figures can be found at

References httpnneurologyorgcontent925e395fullref-list-1

This article cites 42 articles 10 of which you can access for free at

Citations httpnneurologyorgcontent925e395fullotherarticles

This article has been cited by 1 HighWire-hosted articles

Subspecialty Collections

httpnneurologyorgcgicollectionmrsMRS

httpnneurologyorgcgicollectionmci_mild_cognitive_impairmentMCI (mild cognitive impairment)

httpnneurologyorgcgicollectionalzheimers_diseaseAlzheimers diseasefollowing collection(s) This article along with others on similar topics appears in the

Permissions amp Licensing

httpwwwneurologyorgaboutabout_the_journalpermissionsits entirety can be found online atInformation about reproducing this article in parts (figurestables) or in

Reprints

httpnneurologyorgsubscribersadvertiseInformation about ordering reprints can be found online

ISSN 0028-3878 Online ISSN 1526-632XWolters Kluwer Health Inc on behalf of the American Academy of Neurology All rights reserved Print1951 it is now a weekly with 48 issues per year Copyright Copyright copy 2019 The Author(s) Published by

reg is the official journal of the American Academy of Neurology Published continuously sinceNeurology