ARTICLE OPEN ACCESS

MicroRNAs in CSF as prodromal biomarkers forHuntington disease in the PREDICT-HD studyEric R Reed MS Jeanne C Latourelle DSc Jeremy H Bockholt BS Joli Bregu MS Justin Smock MD

Jane S Paulsen PhD and Richard H Myers PhD PREDICT-HD CSF ancillary study investigators

Neurologyreg 201890e264-e272 doi101212WNL0000000000004844

Correspondence

Dr Myers

rmyersbuedu

or Dr Paulsen

jane-paulsenuiowaedu

AbstractObjectiveTo investigate the feasibility of microRNA (miRNA) levels in CSF as biomarkers for prodromalHuntington disease (HD)

MethodsmiRNA levels were measured in CSF from 60 PREDICT-HD study participants using the HTGprotocol Using a CAGndashAge Product score 30 prodromal HD participants were selected basedon estimated probability of imminent clinical diagnosis of HD (ie low medium high n = 10group) For comparison participants already diagnosed (n = 15) and healthy controls (n = 15)were also selected

ResultsA total of 2081 miRNAs were detected and 6 were significantly increased in the prodromal HDgene expansion carriers vs controls at false discovery rate q lt 005 (miR-520f-3p miR-135b-3pmiR-4317 miR-3928-5p miR-8082 miR-140-5p) Evaluating the miRNA levels in each of theHD risk categories all 6 revealed a pattern of increasing abundance from control to low risk andfrom low risk to medium risk which then leveled off from the medium to high risk and HDdiagnosed groups

ConclusionsThis study reports miRNAs as CSF biomarkers of prodromal and diagnosed HD ImportantlymiRNAs were detected in the prodromal HD groups furthest from diagnosis where treatmentsare likely to be most consequential and meaningful The identification of potential biomarkersin the disease prodrome may prove useful in evaluating treatments that may postpone diseaseonset

Clinicaltrialsgov identifierNCT00051324

RELATED ARTICLE

EditorialCSF microRNA in patientswith Huntington disease

Page 151

From the Bioinformatics Program (ERR RHM) Boston University Department of Neurology (JCL JB RHM) Boston University School of Medicine MA and Departments ofNeurology and Psychiatry (JHB JSP) and Internal Medicine (JS) Carver College of Medicine University of Iowa Iowa City

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 ProcessingCharge was funded by the Jerry McDonald Huntingtonrsquos Disease Research Fund

Coinvestigators are listed at linkslwwcomWNLA152

This is an open access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License 40 (CC BY-NC-ND) which permits downloadingand sharing the work provided it is properly cited The work cannot be changed in any way or used commercially without permission from the journal

e264 Copyright copy 2017 The Author(s) Published by Wolters Kluwer Health Inc on behalf of the American Academy of Neurology

Huntington disease (HD) is an inherited neurodegenerativedisease typically diagnosed in midlife1 although symptomsmay appear as early as age 3 and as late as age 85 The mu-tation responsible for HD is an expanded cytosine adenineguanine (CAG) trinucleotide repeat in the first exon of thehuntingtin gene2 Neuropathologic changes involving theaccumulation of the huntingtin protein3 and the degenerationof neurons precede motor diagnosis with up to half of striatalneurons lost before diagnosis4 Volumetric changes in thestriatum are evident as early as 2 decades prior to predicteddiagnosis5 indicating that neuropathologic changes occurmany years prior to clinical motor manifestation and thateffective therapeutics to prevent neurodegeneration wouldneed to be administered long before clinical onset The lack ofvalidated biomarkers for onset and progression of neuro-degeneration prior to clinical manifestation impedes theevaluation of preventive therapies

MicroRNAs (miRNAs) are small noncoding ribose moleculeswith a bonded nucleotide base that negatively regulate mRNAlevels in a sequence-specific manner binding to the 39-un-translated region to initiate cleavage or translational repression67

miRNAs are abundant in the CNS and assist in various neuronalprocesses such as synaptic development maturation andplasticity89 Because they are encapsulated in small vesicles (ei-ther exosomes or microvesicles)10 and are associated withargonaute-2 (AGO2) proteins of the RNA-induced silencingcomplex miRNAs resist degradation by ribonuclease Mountingevidence suggests that disease-specific miRNA profiles can bedetected in CSF in Parkinson disease (PD) and Alzheimer dis-ease (AD)11ndash13

Studies of human HD prefrontal cortex identified 75 signifi-cantly altered miRNAs14 including several that were associ-ated with age at HD motor onset or the level ofneuropathologic involvement in the striatum1516 SomemiRNA levels were altered in brain samples of prodromal HDmHTT carriers16 Although there is evidence of alteredmiRNA levels in plasma samples of prodromal HD thechanges were subtle and not sufficiently sensitive for an ef-fective biomarker17 We therefore sought to assess the pres-ence of miRNAs in CSF from HD prodromal individuals asa biomarker of neurodegeneration prior to diagnosis

MethodsStudy design and participantsThe PREDICT-HD study is a prospective observationalstudy with 32 international sites conducted from

September 2002 to July 2014 All PREDICT-HD partic-ipants had genetic testing prior to study enrollment A totalof 1078 CAG-expanded (CAG gt35 64 female) indi-viduals prior to motor diagnosis of HD were enrolled in thisstudy As healthy controls 304 non-CAG-expanded sib-lings were also included (65 female) Annual assessmentsin the domains of motor cognitive psychiatric function-ing and brain imaging were obtained with collection ofDNA blood saliva and urine The goal of PREDICT-HDwas to find predictive markers for motor manifestation(clinical diagnosis) of HD

Standard protocol approvals registrationsand patient consentsAll participants gave informed written consent prior to studyparticipation and all study procedures were approved by eachsitersquos respective institutional review board

CSF sample acquisitionCSF acquisition was added to the PREDICT-HD protocolat the end of the study at 6 sites All participants underwentscreening for the lumbar puncture (LP) the day prior tosample acquisition so that biospecimens would be collectedafter fasting and that screening blood sample laboratoriescould be conducted Exclusion criteria for LP were (1) useof anticoagulant medication (ie warfarin heparin) orantiplatelets (aspirin) within 14 days (2) unable to fast for8 hours (3) any acute or chronic infection (4) history ofany chronic inflammatory disorder (5) unstable medical orpsychiatric disorder disease or illness and (6) abnor-malities in blood-based screening (eg abnormalities inprothrombin time partial thromboplastin time or lowplatelets) In a sterile environment a Sprotte 24-G atrau-matic spinal needle was used after adequate local anesthesiawas administered The site coordinator recorded the timefor each component of the protocol The first 1ndash2 mL ofCSF from the first syringe was immediately sent at roomtemperature for basic CSF analyses conducted locallywithin 4 hours (ie for cell count erythrocytes total pro-tein and glucose) Remaining CSF was transferred to 15mL conical polypropylene tubes at room temperaturemixed gently by inverting 3ndash4 times and then centrifugedat 2000 g for 10 minutes Aliquots of 15 mL of supernatantwere transferred to precooled 2-mL microcentrifuge tubesand stored at minus80degC until shipment on dry ice to the Na-tional Institute of Neurological Disorders and Stroke bio-repository Over 77 of the samples in this study werecollected by an internist at the University of Iowa (JS) andthe remaining samples were collected from 5 sites in thePREDICT-HD CSF ancillary study

GlossaryAD = Alzheimer disease CAG = cytosine adenine guanine CAP = CAGndashAge Product FDR = false discovery rate HD =Huntington disease LP = lumbar puncture miRNA = microRNA PD = Parkinson disease

NeurologyorgN Neurology | Volume 90 Number 4 | January 23 2018 e265

SamplesCSF samples for 60 participants were chosen by thePREDICT-HD Data Management Team1819 All sampleswere blinded by a unique code specific for this substudy Thesamples included 15 participants clinically diagnosed with HDaccording to traditional criteria with diagnostic confidencelevel of 4 on the Unified Huntingtonrsquos Disease Rating Scale20

30 participants determined to be prodromal gene expansioncarriers for HD and 15 healthy controls Disease burden inthe prodromal participants was determined by calculation ofthe CAGndashAge Product (CAP = age times [CAG minus 3366])21

developed to reflect age-adjusted cumulative exposure to theeffects of mutant huntingtin

miRNA preprocessing and quantificationFifteen microliters of CSF was processed for miRNA levelsusing the HTG Molecular Diagnostics miRNA whole tran-scriptome protocol HTG EdgeSeq system (htgmolecularcomproductshtg-edg-system-edgeseq) This processincludes specific probes for 2083 miRNAs producing bothraw small-RNA sequencing files and prequantified data Amaximum of 24 samples can be processed in a single run andsamples were randomly assigned to each of 3 batches Rawsequencing files were processed and eventually used for dif-ferential analyses Initial checks for sample quality as well asadapter sequence identification was performed using FastQC(version 0113 bioinformaticsbabrahamacukprojectsfastqc) For each sample low-quality reads were removedusing FastX (version 0014 hannonlabcshledufastx_tool-kit) FASTQ Quality Filter using a quality score of 80TruSeq Adapter Index 2 adapter sequence 9 (59-GATCG-GAAGAGCACACGTCTGAACTCCAGTCACCGATGTATCTCGTATGCCGTCTTCTGCTTG-39) was removedfrom each read using Cutadapt (version 171) removingreads with fewer than 15 remaining nucleotides Reads withthe same sequence were combined using FastX (version0014) Collapser reporting the number of duplicated readsper sequence22 Reads were aligned to human genome versionhg19 using Bowtie (version 111) allowing for 0 mis-matches23 Bam files were converted to bed files using bed-tools (version 2250) bamToBed24 miRNAs were defined asreads aligning within plusmn4 bases from the start coordinate ofannotated miRNAs frommirBase (version 20) filtered for the2083 probes25 miRNA reads were counted using Genomi-cRanges (version 1224) R package removing reads greaterthan 27 bases26 Of the 2083 probes we were able to count atleast one read across all samples for 2082 miRNAs OnemiRNA was removed due to low expression (mean rawcounts lt2 across all samples) Therefore differential analysisincludes 2081 individual miRNAs

Statistical analysisAll analysis was carried out using R (version 322) Counts werenormalized using the DESeq2variance stabilization trans-formation in DESeq2 (version 1101)27 These values were thenadjusted for batch effects from their sequencing run usingComBat (version 3180)28 Unless otherwise stated expression

values reported in this article are count values after trans-formation on a log2 scale Sample-level quality control was con-ducted across all samples All differential expression analyses werecarried out with linear models using miRNA expression as theoutcome variable

False discovery rate (FDR) q values were calculated fromnominal p values using the Benjamini-Hochberg procedureperformed by first ordering the p values where the smallest pvalue has a rank of 1 Each p value is then transformed bytaking the product of the p value and the total number of testsand then dividing by the p value ranking Finally the FDR qvalues are assigned as the cumulative minimum of this newset ordered by the reverse ranking of the original p value29

Inadequate power precluded analyses comparing expressionof individual miRNAs to CAG repeat size in the 45 partic-ipants with HD

Sample-level quality controlOutlier samples were detected via qualitative assessment ofplots of the first 2 principal components of expression valuesacross all samples After initial outlier samples were removedthe first 2 principal components of the remaining sampleswere replotted and the remaining samples were reevaluatedfor outliers After 2 iterations of this process no additionalsamples were removed

Diagnosed HD vs controlsDifferential expression analysis between diagnosed HD andcontrols was performed using both the complete set of miRNAsas well as a subset of 16 miRNAs previously reported16 as dif-ferentially expressed between postmortem HD and controlparticipants In each model age was included as a covariate

Ordinal scales of prodromal HD progressionIn order to explore the relationship of miRNA expression withestimated risk of clinical HD diagnosis we assigned ordinalvalues to each clinical group The following values wereassigned 0 to control 1 to low risk 2 to medium risk 3 tohigh risk and 4 to diagnosed manifest HD participants Agewas not included as a covariate in these models because it isa factor in assigning HD prodromal staging

Hierarchical clustering of diagnosed HDand controlsHierarchical clustering was carried out on diagnosed HD andcontrols using a subset of miRNAs determined to be signif-icantly differentially expressed (FDR q value lt01) betweenthe 2 groups Euclidean distance with theWard agglomerativemethod was used to cluster both the samples and miRNAs

ResultsDifferential analysis of miRNA expression inCSF between diagnosed HD and controlsIn order to evaluate altered miRNA expression in HD CSFwe performed differential expression in 2081 miRNA probes

e266 Neurology | Volume 90 Number 4 | January 23 2018 NeurologyorgN

that passed expression filtering quantified from small-RNAsequencing using the HTG EdgeSeq system Of the 60 sam-ples processed 56 passed quality control filtering including14 controls 10 low risk 8 medium risk 10 high risk and 14diagnosed HD (table 1)

The initial analysis compared diagnosed HD to controls Afternormalization and batch correction miRNAs were testedindependently using multivariate linear modeling adjustingfor age Of the 2081 miRNAs 25 reached FDR significance qvalue lt01 and 6 reached FDR significance q value lt005 Inall 25 of these miRNAs expression was upregulated in HDand 14 miRNAs had greater than 2-fold changes in expression(log2FC gt1) in HD compared to control participants (table2) The extent to which these 25 miRNAs separated HD casesfrom controls was further explored via hierarchical clusteringwhich revealed a clear partition between cases and controls

with all but 3 HD samples and 3 control samples clusteringwithin their group (figure 1)

None of the 16 miRNAs previously identified16 to be differ-entially expressed between postmortem HD and controlbrains reached statistical significance when performing FDRcorrections for either the full set 2081 miRNA or the candi-date set of 16 miRNAs though 4 miRNAs reached nominalsignificance (p value lt005 table e-1 linkslwwcomWNLA52)

Analysis of miRNA expression and estimatedrisk of HD diagnosisIn order to evaluate the association between miRNA ex-pression and progression in prodromal to diagnosed HDwe assigned each group an ordinal variable 0 to 4 where0 was assigned to controls 4 to diagnosed HD

Table 1 Sample information before and after sample-level quality control

Control

Prodromal HD

Diagnosed HDLow risk Medium risk High risk

Before sample quality control

n 15 10 10 10 15

Age y mean (SD) 4591 (1398) 3121 (989) 3893 (933) 5122 (1589) 5594 (869)

CAG mean (SD) 2053 (41) 416 (178) 424 (184) 43 (408) 42 (146)

Sex n ()

Male 7 (4667) 5 (5000) 5 (5000) 5 (5000) 5 (3333)

Female 8 (5333) 5 (5000) 5 (5000) 5 (5000) 10 (6667)

Batch n ()

1 6 (4000) 4 (4000) 4 (4000) 5 (5000) 5 (3333)

2 6 (4000) 4 (4000) 4 (4000) 4 (4000) 6 (4000)

3 3 (2000) 2 (2000) 2 (2000) 1 (1000) 4 (2667)

After sample quality control

n 14 10 8 10 14

Age y mean (SD) 4536 (1433) 3121 (989) 3985 (1013) 5122 (1589) 5551 (885)

CAG mean (SD) 2071 (42) 416 (178) 4238 (207) 43 (408) 4214 (141)

Sex

Male 6 (4286) 5 (5000) 4 (5000) 5 (5000) 5 (3571)

Female 8 (5714) 5 (5000) 4 (5000) 5 (5000) 9 (6429)

Batch n ()

1 6 (4286) 4 (4000) 4 (5000) 5 (5000) 4 (2857)

2 5 (3571) 4 (4000) 4 (5000) 4 (4000) 6 (4286)

3 3 (2143) 2 (2000) 0 (000) 1 (1000) 4 (2857)

Abbreviations CAG = cytosine adenine guanine HD = Huntington diseaseFour samples 1 control 1 diagnosed HD and 2 prodromal (both medium risk) did not meet quality control standards and were removed from the study

NeurologyorgN Neurology | Volume 90 Number 4 | January 23 2018 e267

participants and 1ndash3 to each of the prodromal groupsLinear modeling of the 2081 expressed miRNAs acrossthe 56 samples revealed no miRNAs that reached FDRsignificance although 16 had nominal p values lt0005(table 3) These 16 miRNAs included the top 5 signifi-cantly differentially expressed (q lt 005) in the HD vs

control analysis miR-520f-3p miR-135b-3p miR-4317miR-3928-5p miR-8082 Boxplots of the distribution ofexpression across each group for all 6 miRNAs differen-tially expressed between HD vs control participants areshown in figure 2 For each of these miRNAs the directionof the log2FC between adjacent nominal groups is con-sistent with the direction of altered expression seen be-tween HD vs controls

None of the candidate miRNAs reported previously16 as dif-ferentially expressed in HD vs control prefrontal cortexreached FDR q lt 01 and only 2 reached nominal significance(miR-132-3p p lt 0017 miR-5695 p lt 005 table e-2 linkslwwcomWNLA53)

DiscussionThis analysis reports the assessment of miRNAs in HDCSFas a biomarker for HD We evaluated the differential levelsof miRNAs for individuals diagnosed with HD vs controlsas well as the relationship of miRNA levels among geneexpansionndashpositive prodromal individuals with varyingestimated risk of diagnosis (table 1) We first sought todistinguish miRNAs that characterize diagnosed HD usinga discovery set of 2081 miRNAs Six miRNAs were dif-ferentially found in diagnosed HD vs control CSF (FDR qlt 005) and an additional 19 at FDR q lt 01 (table 2) All ofthe miRNAs were upregulated in HD CSF However noneof the miRNAs that we had previously identified with dif-ferential levels in diagnosed HD vs control prefrontalcortex brain samples16 were found to be different in theseearly diagnosed HD CSF samples

When examining the association of miRNA expression toan ordinal scale of diagnosis risk or time to diagnosiswhere 0 was assigned to controls 4 to diagnosed HDparticipants and 1ndash3 to each prodromal group with de-creasing proximity to (or risk of) diagnosis 16 miRNAs hadnominal p lt 0005 (FDR lt0326) including the top 5differentially expressed in diagnosed HD vs controls FDR qlt 005 (table 3) When we plotted the 6 FDR significantmiRNAs we observed a consistent pattern of associationbetween miRNA expression across prodromal groupsSpecifically miRNA increases from control to low risk andincreases again from low risk to medium risk but thenappears to remain elevated across the medium risk to highrisk and HD diagnosed groups (figure 2)

While this study shows altered miRNA expression in HDCSF similar studies have been performed for AD and PD Astudy11 using small-RNA sequencing to quantify 2228miRNAs in 69 AD 67 PD and 78 control participantsreported differential expression of 41 miRNAs in AD vscontrols and 17 miRNAs in PD vs controls A similarstudy13 using TaqMan low-density array human miRNApanels to quantify 746 exosomal miRNAs in CSF across 28

Table 2 Differentially expressed microRNAs (miRNAs)betweendiagnosedHuntington disease (HD) andcontrols

miRNAMeanexpression logFC p Value

FDRq value

miR-520f-3p 447 124 000005 0040

miR-135b-3p 353 116 000007 0040

miR-4317 603 120 000008 0040

miR-3928-5p 637 098 000008 0040

miR-8082 330 142 000013 0049

miR-140-5p 619 065 000014 0049

miR-509-3-5p 504 136 000020 0055

miR-6516-5p 406 150 000021 0055

miR-455-3p 376 095 000030 0059

miR-6838-3p 431 105 000030 0059

miR-552-5p 364 121 000033 0059

miR-761 347 095 000037 0059

miR-4659a-5p 487 118 000037 0059

miR-4781-5p 615 092 000041 0061

miR-4462 473 105 000053 0074

miR-132-5p 534 090 000058 0074

miR-6818-5p 381 103 000060 0074

miR-34c-3p 305 086 000072 0083

miR-4724-3p 687 108 000076 0083

miR-4307 597 095 000089 0090

miR-6874-5p 398 110 000091 0090

miR-5581-3p 376 095 000101 0094

miR-6807-5p 509 090 000104 0094

miR-922 313 128 000112 0094

miR-1322 373 133 000113 0094

Abbreviation FDR = false discovery rateResults of differential expression of miRNAs between 14 diagnosed HDand 14 control participants Shown are the 6 miRNAs with FDR q valueslt005 and an additional 19 with q lt 01 ordered by nominal p value Thesep values reflect the coefficient for HD status adjusted for participant age ina multivariate linear model FDR q values are calculated using the Benja-mini-Hochberg procedure for the set of 2081 miRNAs tested The meanexpression values are calculated from the DESeq2variance stabilized andbatch-corrected values across all 28 participants The values for logFCspecify the log2 transformation of the fold change of miRNA expression ofparticipants with HD vs controls

e268 Neurology | Volume 90 Number 4 | January 23 2018 NeurologyorgN

AD 47 PD and 27 control participants reported differen-tial expression of 27 miRNAs in AD vs controls and 6miRNAs in PD vs controls These studies show concor-dance with our results most notably miR-132-5p identifiedin the PD analysis11 as well as in both the AD and PDanalyses13 In our analysis miR-132-5p was differentiallyexpressed in diagnosed HD vs controls (table 2) as well asnominally associated with ordinal categorization of pro-dromal HD progression (p = 0035 FDR = 033) miR-132-3p was included in the set of miRNAs that were differen-tially regulated in HD brain16 Of these 16 miRNAs miR-132-3p had the second lowest nominal p value whencomparing diagnosed HD vs control CSF (p = 0025 FDR= 015 table e-1 linkslwwcomWNLA52) as well as thelowest nominal p value for the ordinal relationship (p =0020 FDR = 027) Also identified in the AD analysis11

miR-760 was one of the top 16 miRNAs in our ordinalanalysis (p = 00038 FDR = 036 table 3)

Several points can be made from these studies First we didnot see a strong relationship between miRNA levels thatdistinguish HD from control in brain with the miRNA levelsthat distinguish HD from control in CSF The process bywhich miRNAs are released into CSF is still not well-understood and it may be that miRNAs released into CSF arederived from the degeneration of neurons as the integrity ofthe neuronal cell membrane is lost while the predominantdifferential miRNA levels seen inHD brainmay instead reflectmiRNAs found in non-neuronal cell types (microglia astro-cytes and oligodendrocytes)

Second the pattern for miRNA increase present for theearliest prodromal stages of HD may be important for fu-ture clinical trials as those miRNAs may reflect changesoccurring in the brain that echo effects of the initial neu-rodegeneration seen in HD long before clinical diagnosisA panel of miRNAs may provide insight into whether

Figure 1 Hierarchal clustering of differentially expressed microRNAs (miRNAs)

Hierarchal clustering of 14 diagnosed Huntington disease cases and 14 controls presented on the X axis defined by the color at the top of the figure using thetop 25 most differentially expressed miRNAs presented on the Y axis (table 2) Samples and miRNAs have been clustered based on their normalizedexpression Colors in this heatmap reflect miRNA-wise z score transformation of normalized expression where darker shades of red represent increasedlevels and darker shades of blue represent decreased levels

NeurologyorgN Neurology | Volume 90 Number 4 | January 23 2018 e269

treatments are preventing the initiation of the degenerativeprocess in HD clinical trials that seek to prevent earlydamaging effects of the HTT gene on the integrity of thebrain These findings show particular promise since veryfew baselinecross-sectional measures have detected dif-ferences between the low-riskfar from diagnosis pro-dromal group and controls To our knowledge onlyemotion recognition30 and striatal volumes31 from MRI arereportedly different between controls and prodromal par-ticipants who are furthest from HD diagnosis Biomarkersto detect and track the earliest measures of disease willbecome important in future clinical trials of preventivetherapies

Finally we recognize that the sample size of 60 studied heremay not have sufficient power to detect all of the miRNAsthat are altered in diagnosed or prodromal HD relative tocontrols Additional studies of larger cohorts throughoutthe continuum of the disease spectrum and studies ofsamples taken longitudinally will almost certainly revealadditional important insights into the utility of miRNAmeasures in CSF as biomarkers for prodromal HD We also

recognize the imprecision of the prodromal staging variableCAP Although research has validated the utility of thisvariable19 biomarker studies are likely to improve as theprodromal stages of HD are further characterized andsubjected to clinimetrics Generalizability of our findingswith regards to sex ethnicity race and potential environ-mental factors is unknown since the PREDICT studyrecruited all participants with regards to HD risk Confi-dence in the findings might be strengthened since thecontrol group represented siblings whose genetic testrevealed a normal CAG length offering control for somevariation in individual differences An additional limitationof our study is that the prodromal groups are partially de-fined by age with those further from diagnosis beingyounger than those nearer to diagnosis risk Consequentlyadjusting for age across these groups is problematic Effectsof age on the levels of miRNAs may be a source of bias thatwe are not able to consider in a study of this size Finally werecognize that 3 of the HD cases did not cluster with theothers and cluster as controls in figure 1 These may bea consequence of assay failure or other unknown factorsthat alter the levels of miRNAs in CSF Additional work to

Table 3 MicroRNA (miRNA) expression association with ordinal categories of control prodromal and diagnosedHuntington disease (HD)

Mean expression logFC p Value FDR q value

miR-18b-5p 495 023 000052 0326

miR-135b-3pa 434 020 000086 0326

miR-875-3p 628 021 000091 0326

miR-3928-5pa 640 016 000095 0326

miR-520f-3pa 414 018 000146 0326

miR-4317a 630 020 000229 0326

miR-4252 548 014 000317 0326

miR-4499 475 022 000336 0326

miR-6838-3p 451 020 000337 0326

miR-8082a 486 022 000341 0326

miR-760 448 012 000379 0326

miR-4723-3p 425 minus009 000409 0326

miR-4491 541 022 000433 0326

miR-4327 633 017 000452 0326

miR-335-3p 520 014 000488 0326

miR-7705 569 023 000497 0326

Abbreviation FDR = false discovery rateResults of univariate linear modeling of miRNA expression vs ordinal categories of risk of diagnosis Shown are the 16 miRNAs with the lowest nominal pvalues These p values reflect the coefficient for ordinal groupmembership FDR q values are calculated using the Benjamini-Hochberg procedure for the setof 2081 miRNAs tested The mean expression values are calculated from the DESeq2variance stabilized and batch-corrected values across all 56 partic-ipants The logFC values represent the estimated change inmiRNAexpression between 2 adjacent ordinal groups calculated as the log2 transformation of thefold change between 2 adjacent study groups on the ordinal scalea Significantly differentially expressed between HD and controls (table 2)

e270 Neurology | Volume 90 Number 4 | January 23 2018 NeurologyorgN

Figure 2 Plots of microRNAs (miRNAs) across categories of control prodromal and diagnosed Huntington disease (HD)

Boxplots of the distribution of DESeq2variance stabilized and batch-corrected expression among the 5 ordinal groups (risk of diagnosis of HD) for each of the6 miRNAs differentially expressed between HD and control participants (table 2 A 50f-3p B 135b-3p C 4317 D 3928-5p E 8082 F 140-5p) p Values andlogFC values are the same as in table 3 The low-risk medium-risk high-risk and diagnosed HD groups are synonymous with the far from onset middle fromonset near onset and symptomatic HD groups

NeurologyorgN Neurology | Volume 90 Number 4 | January 23 2018 e271

determine the sources of miRNA variation is likely to im-prove the efficacy of these measures

Author contributionsRHM and JSP conceived and designed the study ERR JCL JHB JB JS JSP and RHM acquired analyzed orinterpreted the data ERR RHM and JSP drafted themanuscript ERR JCL JHB JB JS JSP and RHMprovided critical revision of the manuscript for importantintellectual content ERR JCL and JHB conducted dataand statistical analysis RHM and JSP obtained fundingand provided study supervision

AcknowledgmentThe authors thank the PREDICT-HD sites the studyparticipants the National Research Roster for HuntingtonDisease Patients and Families the Huntingtonrsquos DiseaseSociety of America the Huntington Study Group and theEuropean Huntingtonrsquos Disease Network

Study fundingSupported by the Jerry McDonald HD Research Fund and bythe NIH National Institute of Neurologic Disorders and Strokegrant (3R01-NS073947) awarded to Richard H Myers and bythe NIH National Institute of Neurologic Disorders and Strokegrants (5R01NS040068 5R01NS054893 5U01NS082089) andthe CHDI Foundation Inc (A6266 A2015) awarded to Jane SPaulsen This publication was supported by the National Centerfor Advancing Translational Sciences and the NIH The contentis solely the responsibility of the authors and does not necessarilyrepresent the official views of the NIH

DisclosureThe authors report no disclosures relevant to the manuscriptGo to NeurologyorgN for full disclosures

Received November 8 2016 Accepted in final form September 292017

References1 Myers RH Huntingtonrsquos disease genetics NeuroRx 20041255ndash2622 MacDonald ME Ambrose CM Duyao MP et al A novel gene containing a tri-

nucleotide repeat that is expanded and unstable on Huntingtonrsquos disease chromo-somes Cell 201672971ndash983

3 Gomez-Tortosa E Macdonald ME Friend JC et al Quantitative neuropathologicalchanges in presymptomatic Huntingtonrsquos disease Ann Neurol 20014929ndash34

4 Vonsattel JP Myers RH Stevens TJ Ferrante RJ Bird ED Richardson EP Jr Neu-ropathological classification of Huntingtonrsquos disease J Neuropathol Exp Neurol 198544559ndash577

5 Aylward EH Sparks BF Field KM et al Onset and rate of striatal atrophy in pre-clinical Huntington disease Neurology 20046366ndash72

6 Bartel DP MicroRNAs genomics biogenesis mechanism and function Cell 2004116281ndash297

7 Bartel DP MicroRNAs target recognition and regulatory functions Cell 2009136215ndash233

8 Schratt GM Tuebing F Nigh EA et al A brain-specific microRNA regulates dendriticspine development Nature 2006439283ndash289

9 Cao X Yeo G Muotri AR Kuwabara T Gage FH Noncoding RNAs in the mam-malian central nervous system Annu Rev Neurosci 20062977ndash103

10 Arroyo JD Chevillet JR Kroh EM et al Argonaute2 complexes carry a population ofcirculating microRNAs independent of vesicles in human plasma Proc Natl Acad SciUSA 20111085003ndash5008

11 Burgos K Malenica I Metpally R et al Profiles of extracellular miRNA in cerebro-spinal fluid and serum from patients with Alzheimerrsquos and Parkinsonrsquos diseases cor-relate with disease status and features of pathology PLoS One 20149e94839

12 Kumar S Reddy PH Are circulating microRNAs peripheral biomarkers for Alz-heimerrsquos disease Biochim Biophys Acta 201618621617ndash1627

13 Gui Y Liu H Zhang L Lv W Hu X Altered microRNA profiles in cerebrospinal fluidexosome in Parkinson disease and Alzheimer disease Oncotarget 2015637043ndash37053

14 Hoss AG Kartha VK Dong X et al MicroRNAs located in the hox gene clusters areimplicated in Huntingtonrsquos disease pathogenesis PLoS Genet 201410e1004188

15 Hadzi TC Hendricks AE Latourelle JC et al Assessment of cortical and striatalinvolvement in 523 Huntington disease brains Neurology 2012791708ndash1715

16 Hoss AG Labadorf A Latourelle JC et al miR-10b-5p expression in Huntingtonrsquosdisease brain relates to age of onset and the extent of striatal involvement BMCMedGenomics 2015810

17 Hoss AG Lagomarsino VN Frank S Hadzi TC Myers RH Latourelle JC Study ofplasma-derivedmiRNAsmimic differences in Huntingtonrsquos disease brain Mov Disord2015301961ndash1964

18 Paulsen JS Hayden M Stout JC et al Preparing for preventive clinical trials thePREDICT-HD study Arch Neurol 200663883ndash890

19 Paulsen JS Long JD Ross CA et al Prediction of manifest Huntington disease withclinical and imaging measures a 12-year prospective observational study LancetNeurol 2014131193ndash1201

20 Huntington Study Group Unified Huntingtonrsquos Disease Rating Scale reliability andconsistencyMov Disord 199611136ndash142

21 Zhang Y Long JD Mills JA Warner JH Lu W Paulsen JS PREDICT-HD Inves-tigators and Coordinators of the Huntington Study Group Indexing disease pro-gression at study entry with individuals at risk for Huntington disease Am J MedGenet B Neuropsychiatr Genet 2011156751ndash763

22 Martin M Cutadapt removes adapter sequences from high-throughput sequencingreads EMBnetjournal 20111710

23 Langmead B Trapnell C PopM Salzberg SL Ultrafast and memory-efficient alignment ofshort DNA sequences to the human genome Genome Biol 200910R25

24 Quinlan AR Hall IM BEDTools a flexible suite of utilities for comparing genomicfeatures Bioinformatics 201026841ndash842

25 Kozomara A Griffiths-Jones S MiRBase annotating high confidence microRNAsusing deep sequencing data Nucleic Acids Res 20144268ndash73

26 Lawrence M Huber W Pages H et al Software for computing and annotatinggenomic ranges PLoS Comput Biol 20139e1003118

27 LoveMI Huber W Anders S Moderated estimation of fold change and dispersion forRNA-seq data with DESeq2 Genome Biol 201415550

28 Johnson WE Li C Rabinovic A Adjusting batch effects in microarray expression datausing empirical Bayes methods Biostatistics 20078118ndash127

29 Benjamini Y Hochberg Y Controlling the false discovery rate a practical and pow-erful approach to multiple testing J R Stat Soc 199557289ndash300

30 Stout JC Paulsen JS Queller S et al Neurocognitive signs in prodromal Huntingtondisease Neuropsychology 2011251ndash14

31 Paulsen JS Nopoulos PC Aylward E et al Striatal and white matter predictors ofestimated diagnosis for Huntington disease Brain Res Bull 201082201ndash207

e272 Neurology | Volume 90 Number 4 | January 23 2018 NeurologyorgN

SOURCE ARTICLE NPuborgyadi3d

MicroRNAs in CSF as prodromal biomarkers forHuntington disease in the PREDICT-HD studyEric R Reed MS Jeanne C Latourelle DSc Jeremy H Bockholt BS Joli Bregu MS Justin Smock MD

Jane S Paulsen PhD and Richard H Myers PhD The PREDICT-HD CSF ancillary study investigators

Cite as Neurologyreg 201890e264-e272 doi101212WNL0000000000004844

Correspondence

Dr Myers

rmyersbuedu

or Dr Paulsen

jane-paulsenuiowaedu

Study questionCan microRNA (miRNA) levels in CSF serve as biomarkers ofneurodegeneration in Huntington disease (HD) prodromalindividuals

Summary answermiRNAs are effective CSF biomarkers for prodromal HD longbefore diagnosis

What is known and what this article addsVolumetric changes (in the striatum) are often the earliestindicators of HD onset and progression However there isa distinct lack of validated biomarkers This study reveals CSFbiomarkers that could help detect and prevent HD

Participants and settingCSF miRNA levels from 60 PREDICT-HD participants weremeasured 30 participants with prodromal HD were selectedThe rest were part of a comparison study where 15 individualswere healthy and 15 had HD

Design size and durationThis was a prospective observational study comprising 32 in-ternational sites from September 2002 to July 2014 In total 1078CAG-expanded individuals were enrolled prior to HD diagnosis aswell as 305 non-CAG-expanded siblings as healthy controls CSFmiRNA levels were measured via the HTG molecular diagnosticsmiRNA whole transcriptome protocol which includes specificprobes for 2083 miRNAs Patients with prodromal HD were se-lected based on estimation of imminent clinical HD diagnosisbased on a CAGndashAge Product (CAP) score

Primary outcomes risks and exposuresDifferential miRNA expression levels were determined for bothindividuals with diagnosedHD and controls and the relationshipof miRNA levels among prodromal individuals with different riskdiagnoses

Main results and the role of chanceOf the 2081 detectedmiRNAs differential expression of 6 miRNAs(miR-520f-3p miR-135b-3p miR-4317 miR-3928-5p miR-8082and miR-140-5p) was significantly higher in the prodromal HDgene expansion carriers than in the controls (q lt 005) This increase

in expression was significant in the low-HD risk group compared tothe control and in themedium-HD risk group compared to the low-HD risk group However there were no such observations betweenthe medium to high-HD risk and HD-diagnosed groups

Bias confounding and other reasons for cautionThe sample size may not have had sufficient detection power toidentify all differentially expressed miRNAs in diagnosed orprodromal HD compared to controls CAP scores may not beaccurate as a prodromal HD readout Age adjustment acrossgroups was problematic as prodromal groups were partially de-fined by age as those further from diagnosis were younger thanthose closer to diagnosis risk Furthermore effects of age onmiRNA levels may serve as a source of bias Three HD cases didnot cluster possibly owing to assay failure

Generalizability to other populationsGiven that the PREDICT study recruited all participants withregard toHD risk generalizability of the findings with regard to sexethnicity race and potential environmental factors may be limited

Study fundingpotential competing interestsThe study was funded by a group of foundation and governmentgrants Go to NeurologyorgN for full disclosures

Plots of microRNAs (miRNAs) across categories of controlprodromal and diagnosed Huntington disease (HD)

A draft of the short-form article was written by E Feric a writer with Editage a division of Cactus Communications The authors of thefull-length article and the journal editors edited and approved the final version

Copyright copy 2018 American Academy of Neurology 157

SHORT-FORM ARTICLE

DOI 101212WNL0000000000004844201890e264-e272 Published Online before print December 27 2017Neurology

Eric R Reed Jeanne C Latourelle Jeremy H Bockholt et al PREDICT-HD study

MicroRNAs in CSF as prodromal biomarkers for Huntington disease in the

This information is current as of December 27 2017

ServicesUpdated Information amp

httpnneurologyorgcontent904e264fullincluding high resolution figures can be found at

References httpnneurologyorgcontent904e264fullref-list-1

This article cites 31 articles 4 of which you can access for free at

Citations httpnneurologyorgcontent904e264fullotherarticles

This article has been cited by 1 HighWire-hosted articles

Subspecialty Collections

httpnneurologyorgcgicollectionhuntingtons_diseaseHuntingtons disease

httpnneurologyorgcgicollectioncerebrospinal_fluidCerebrospinal Fluid

httpnneurologyorgcgicollectioncase_control_studiesCase control studiesfollowing 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 2017 The Author(s) Published by

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