•  Comprehensive profiling of extracellular miRNAs in biofluids depends upon the availability of highly sensitive and reproducible methods for RNA quantification

•  Small RNA-Seq is a powerful method for miRNA profiling but the RNA isolation, adapter ligation, and reverse transcription steps can introduce both systematic variation (bias) and non-systematic variation (noise)

•  EdgeSeq (HTG Molecular) is an alternative method for miRNA profiling •  miRNA-containing solutions are hybridized to a large set of nucleic

acid probes followed by nuclease protection, amplification, and massively parallel sequencing

•  Suitable for analysis of biofluids (without RNA isolation) or isolated RNA

•  Study goals •  Analyze the reproducibility of the EdgeSeq assay for miRNAs in

human plasma using male and female plasma pools developed for ERCC U01 comparative analyses

•  Analyze sequence-specific biases of the EdgeSeq assay using a large set of synthetic human miRNAs

•  Compare EdgeSeq results with results of RNA-Seq studies of the same samples

Introduction

Reproducibility and Bias of miRNA Profiling with the HTG EdgeSeq Targeted Capture and Sequencing Assay Andrea J. Barczak1*, Paula Godoy1*, Rebecca Barbeau1, Joshua Pollack1, Alton Etheridge2, David Galas2, Ionita Ghiran3, Prescott G. Woodruff1, David J. Erle1

1University of California, San Francisco (UCSF); 2Pacific Northwest Diabetes Research Institute (PNDRI); 3Beth Israel Deaconess Medical Center (BIDMC)

EdgeSeq for Relative Quantification of miRNAs Between Samples

Figure 4: EdgeSeq comparisons of reference and test synthetic human miRNA pools. Reference pools had equimolar amounts of all synthetic miRNAs. In (A), total miRNA mass was 5 x 10-17 mol (170 zmol for each miRNA). In (B), total miRNA mass was 1.25 x 10-17 mol (43 zmol for each miRNA). For test samples, miRNAs were divided into 4 sub-pools at the indicated concentrations by serial 10-fold dilutions. Read count ratios (reference/test) for each miRNA were median normalized using the sub-pool of miRNAs present at equal amounts in reference and test samples (1:1). The green dashed line shows expected results (line of unity). EdgeSeq made accurate measurements of relative concentrations of most miRNAs in the range of 1:1 to 1:1/100 and with amounts of miRNA as low as ~1 zeptomole (<1000 molecules). At ratios of 1:1/1000 relative read counts systematically underestimated true differences.

Female Plasma

miRNAs with >1

read/104

(RNA-Seq)

miRNAs with ≤1

read/104

(RNA-Seq)

Total

miRNAs with >1

read/104

(EdgeSeq) 76 428 504

miRNAs with ≤1

read/104

(EdgeSeq) 23 1729 1752

Total miRNAs 99 2157 2256

Male plasma pool

EdgeSeq and RNA-Seq Bias Analysis Using Synthetic miRNAs

Figure 3: EdgeSeq and RNA-Seq analyses of an equimolar reference samples containing a pool of 284 synthetic human miRNAs. (A) Distributions of EdgeSeq and RNA-Seq read counts for synthetic miRNA pools. EdgeSeq, RNA-Seq, and randomized adapter RNA-Seq (4N-RNA-Seq) were performed using indicated amounts of synthetic miRNA pools. RNA-Seq was performed at UCSF and PNDRI, as indicated. Contour plots are overlaid on box plots to show distribution of read counts for each method. EdgeSeq was less biased than standard RNA-Seq despite use of much lower amounts of input miRNA. Use of randomized end adapters (4N-RNA-Seq) reduced bias to levels similar to those found with EdgeSeq. (B) Reproducibility of EdgeSeq. (C) Comparison of EdgeSeq and standard RNA-Seq. (D) Comparison of RNA-Seq results at two different sites using different amounts of input miRNA. (E) Comparison of EdgeSeq RNA-Seq and 4N-RNA-Seq. Each point represents one of the 284 miRNAs. EdgeSeq was highly reproducible. Standard RNA-Seq bias was similar at the two sites using different amounts of input miRNA. EdgeSeq and 4N-RNA-Seq had relatively modest (and somewhat different) biases.

Figure 2: EdgeSeq and RNA-Seq analyses of female and male plasma pools. (A) EdgeSeq replicates for samples from female plasma pool. 25 µl of a 1:1 mixture of plasma and hybridization buffer were used for each hybridization. Each point represents one miRNA. (B) Comparison of EdgeSeq and RNA-Seq read counts for female pools. RNA-Seq analyses were performed using RNA isolated from 500 µl of plasma as input for the Illumina TruSeq small RNA protocol. (C) Comparison of miRNAs with >1 read/104 from EdgeSeq and RNA-Seq assays. (D-F) Same analyses with male plasma pool. "Reads/104" indicates reads per 104 total mapped miRNA reads. miRNAs with read counts of <100/104 total reads were set to 100. >5 times more miRNAs yielded read counts above the arbitrary threshold of 1 read/104 with EdgeSeq. This could be due to decreased sensitivity of RNA-Seq (due to sequence-specific bias) and/or false positives in EdgeSeq (due to cross-hybridization).

Plasma miRNA Profiling

100

101

102

103

104

Edg

eSeq

Fem

ale

Pla

sma

#1R

eads

/104

EdgeSeq Female Plasma #2Reads/104

100 101 102 103 104

100

101

102

103

104

Edg

eSeq

Mal

e P

lasm

a #1

Rea

ds/1

04

RNA-Seq Male PlasmaReads/104

100 101 102 103 104

A

E

r = 0.98

100

101

102

103

104

Edg

eSeq

Mal

e P

lasm

a #1

Rea

ds/1

04

EdgeSeq Male Plasma #2Reads/104

100 101 102 103 104

Dr = 0.99

B

100

101

102

103

104

Edg

eSeq

Fem

ale

Pla

sma

#1R

eads

/104

RNA-Seq Female PlasmaReads/104

100 101 102 103 104

r = 0.33

r = 0.28

C

FMale Plasma

miRNAs with >1

read/104

(RNA-Seq)

miRNAs with ≤1

read/104

(RNA-Seq)

Total

miRNAs with >1

read/104

(EdgeSeq) 66 507 573

miRNAs with ≤1

read/104

(EdgeSeq) 19 1664 1683

Total miRNAs 85 2171 2256

Conclusions

•  EdgeSeq provided reproducible measurements of plasma and synthetic miRNAs and bias was substantially lower than observed with a standard RNA-Seq protocol.

•  EdgeSeq detected substantially more miRNAs than standard RNA-Seq even though much lower amounts of plasma and synthetic mIRNA were used with EdgeSeq.

•  Compared with standard RNA-Seq, 4N-RNA-Seq reduced bias to a level comparable to that seen with EdgeSeq.

•  EdgeSeq was accurate at quantifying relative levels of synthetic miRNAs over two orders of magnitude.

•  EdgeSeq advantages •  No RNA isolation required •  Small volumes of biofluid required •  Sensitive and reproducible •  Most reads map to miRNA probes (no adapter dimers or non-miRNA sequences)

•  EdgeSeq limitations •  Only analyzes miRNAs that match those in probe set •  Potential for cross-hybridization with closely related sequences (miRNAs and

non-miRNAs) References 1.  Raabe CA, Tang TH, Brosius J, Timofev, F. Biases in small RNA deep sequencing data.

Nucleic Acids Res. 2014 Feb; 42(3):1414-26. 2.  Jayaprakash AD, Jabado I, Brown BD, Sachidanandam R. Identification and remediation of

biases in the activity of RNA ligases in small-RNA deep sequencing. Nucleic Acids Res. 2011 Nov; 39(21):e141.

Acknowledgements Supported by NIH grants 1U01HL126493, 1U01HL126496, and 1U01HL126497. Thanks to Mark Schwartz, HTG Molecular Diagnostics, Inc.

Materials & Methods

miRNA Biomarker Identification Using a Sequencing-Based Nuclease Protection Assay �����!*����%�������������%$���$�������!���*���"����� ��#%"����$

Bruce Seligmann1, Thomas Beach2, Debrah Thompson1, Ihab Botros1; 1HTG Molecular Diagnostics, Inc., Tucson, AZ; 2Banner Sun Health Res. Inst., Sun City, AZ

Presented at ABRF Annual Meeting | Palm Springs, CA | March 2013

INTRODUCTION

Support and Acknowledgements Research Funded by the National Institute on Aging, NIH, HHS 5 R43 AG 040004-02 Phase I SBIR grant. We are grateful to the Banner Sun Health Research Institute Brain and Body Donation Program of Sun City, Arizona for the provision of the frozen, fixed, brain tissue and serum. The Brain and Body Donation Program is supported by the National Institute of Neurological Disorders and Stroke (U24 NS072026 National Brain �+"��'002#��#0,2/!#�$,/���/('+0,+90��'0#�0#��+"��#)�1#"��'0,/"#/0���1&#���1',+�)��+01'121#�,+� %'+%����� ������ /'7,+�� )7&#'*#/90�Disease Core Center), the Arizona Department of Health Services �!,+1/�!1��������� /'7,+�� )7&#'*#/90��#0#�/!&��#+1#/���1&#� /'7,+��Biomedical Research Commission (contracts 4001, 0011, 05-901 and 1001 to the Arizona Parkinson's Disease Consortium) and the �'!&�#)�����,5��,2+"�1',+�$,/���/('+0,+90��#0#�/!&

�2')"'+%�,+�1&#�!,**#/!'�)�.�� 8��00�6�4&'!&�&�0� ##+��21,*�1#"�,+�1&#�������601#*�-)�1$,/*�*�/(#1#"� 6������4#�"#3#),-#"���+ovel +2!)#�0#�-/,1#!1',+�1�/%#1#"��� �0#.2#+!'+%��00�6�����#.8��1&�1�20#0��+�#51/�!1',+�$/##�)60'0�-/,!#00�$,)),4#"� 6���+2!)#�0#�protection assay (NPA) to prepare a stoichiometric library of nuclease protection probes (NPP) for measurement. The NPSeq probes are amplified and barcoded by PCR and then sequenced. By barcoding and pooling many samples into a single sequencing run, the sequencing cost/sample can be reduced. Since all the NPP sequences are known, the data analysis is simple and fast. FASTQ files from the sequencer are simply compared to a look-up table of the probes, using the Bowtie short read aligner, and exact matches are counted. NPSeq can be used for biomarker identification in assays of 2,000+ genes. NPSeq retains the advantages qNPA has over other methods for measuring RNA accurately from formalin fixed paraffin embedded (FFPE) tissue to develop and perform assays providing useful additional information1. The EDGE System will automate the preparation of samples for NPSeq through the preparation the PCR plate. We have used NPSeq to measure the levels of 1,844 miRNA from *'���0#����'+������ /�'+�1'002#�$/,*�"#!#�0#"�+,/*�)��� )7&#'*#/90����/('+0,+90���+"���0!2)�/��#*#+1'��-�1'#+10�1,�'"#+1'$6� iomarkers and demonstrate feasibility. Future studies can be performed on the EDGE ArrayPlate assay once biomarkers are narrowed down to a selected few of high interest. 1Rimsza et al, Blood, 2008 Oct 15, 112 (8): 3425-3433.

PERFORMANCE NPP Cocktail QC

� 1,877 miRNA from miRBase 19 plus 19 mRNA housekeepers and negative control passed QC by sequencing

Inter-Sequencing Run Reproducibility � 6% average CV between triplicate NPA reactions o Genes with expression levels >100 reads,250 ng/sample

reference prostate RNA

Linearity and Sensitivity � Titration of reference RNA

� Titration of clinical FFPE

CROSS-PLATFORM PERFORMANCE Compare NPSeq to qPCR

� Test reference RNA o Human prostate total RNA (AM7988), human liver total RNA

(AM7960) o Compare fold change NPSeq results to qPCR performed on

select miRNA � With Removal of one outlier: R2 = 0.905

RESULTS: BIOMARKER DISCOVERY Comparison of 298 miRNA Expressed at >100 Reads

� �� )7&#'*#/90�� ���&,/'7,+1�)��5'0��vs Control (Ctr), full scale, all miRNA o B: AD vs Ctr, expanded scale, lower expressed miRNA

� ��� )7&#'*#/90�� ���&,/'7,+1�)��5'0��vs ��/('+0,+90�������$2))�0!�)#��all miRNA o D: AD vs PD, expanded scale, lower expressed miRNA

� ��� )7&#'*#/90�� ���&,/'7,+1�)��5'0��vs Vascular Dementia (VAD), full scale, all miRNA o F: AD vs VAD, expanded scale, lower expressed miRNA

CONCULSIONS AND NEXT STEPS

RESULTS: SELECTION OF BRAIN REGION � Compared the middle temporal gyrus of the cerebral cortex (MTG),

and subcortical regions of the hippocampus (HIP) and amigdala (AMG)

'�� ��"���$���� �&#� �����"����$"�������������)�����"+#�disease (B) tissue

� Identified miRNA that were differentially expressed between regions o No large fold change differences observed

� Selected the MTG region for biomarker discovery o Most accessible, largest number of Control (Ctr��� )7&#'*#/90��'0#�0#�� �����+"���/('+0,+90��'0#�0#�������+"���0!2)�/�Dementia (VAD) matched FFPE, frozen and serum samples

NPSeq miRBase 19 Assay � Assay established, measuring o 1,877 of the miRNA in miRBase 19 o 19 housekeeper mRNA o Internal negative control

� High sensitivity: requires low amounts of sample whether FFPE or RNA

� High reproducibility: average 6% CV for separately processed samples

� Low sequencing cost/sample afforded by multiplexing barcoded samples

� Performance of the assay has been characterized � �1')'16�"#*,+01/�1#"�4'1&� )7&#'*#/90� ',*�/(#/�012"6

��)�����"+#��$%�( � �',*�/(#/0�$,/� )7&#'*#/90��'0#�0#�&�3#� ##+�'"#+1'$'#"�20'+%�1&#�

NPSeq miRBase 19 assay o Several identified that differentiate AD from Ctr, PD, and VAD o Additional identified that differentiate AD from either Ctr, PD, or

VAD

Next Steps � Confirm biomarkers and extend study o Confirm on a larger independent cohort of patient samples o Extend study to measurements of miRNA biomarkers in serum

� Simplify protocol (e.g. eliminate library gel purification step) � Establish a mRNA biomarker identification assay o Surrogate whole transcriptome assay measuring 2600 genes in

development

����!*��������������������� Sample Preparation Kit and Protocol for Fixed Tissue

� Select samples o FFPE, Paraffin embedded FFPE

� Add lysis buffer & denaturation oil from sample prep kit o Incubate at 95oC for 10 min

� Add proteinase K from sample prep kit o Incubate at 50oC for 60 min

� Transfer lysates to 96 well sample plate

NPA Sample Processing � Add cocktail of nuclease protection probes (NPP) and incubate � Add S1 nuclease enzyme and incubate � Add Stop buffer and incubate

PCR to Prepare Sequencing Library from the NPP � Add aliquot of each NPA sample to PCR Master Mix � Add PCR forward and reverse primers with barcodes and adaptor

sequences � Carry out 15 cycles of PCR � Pool, Gel purify, Sequence

Sensitivity to <0.5mm2 area, 5 �m thick

section

Sensitivity to <6 ng/sample

!���*��������!*�������������� Xx

EDGE Reader

EDGE Processor

Nuclease Protection Assay (NPA)

Nuclease Protection Assay (NPA)

!���*����$��� EDGE System

SIMPLE. ACCURATE. RNA

Unlike any other molecular technology for RNA analysis

SIMPLE Walk-Away Automation Extraction-Free. Amplification-Free. Synthesis-Free. ACCURATE High Frozen-to-FFPE Correlation Multiplex up to 4,512 data points in a single run RNA 24-Hour Sample-to-Answer Single 5µm Section Input

Primer 9��"�-1#/�!,*-)#*#+1 i5 barcode complement

3430 E . G lobal Loop | Tucson, A Z 85706 | (877) 289-2615 | htgMolecular.com Learn&more&about&the&EDGE&System&&http://www.htgmolecular.com/products/htg-edg-system-edgeseq

Micro-RNA samples: •  Two pools of plasma were collected from human subjects (1 female pool, 1 male pool) at

BIDMC and distributed to ERCC U01 groups. •  284 human miRNAs were synthesized (PNDRI) and pooled at equimolar concentrations

(reference samples) or divided into 4 sub-pools that were combined in a ratio of 1000:100:10:1 (test samples).

Methods: •  Three methods were used: HTG EdgeSeq, Illumina TruSeq small RNA-Seq (RNA-Seq), and a

modified RNA-Seq using randomized adapters with 4N added to the 3’ ends1 (4N-RNA-Seq).

Nor

mal

ized

read

cou

nt ra

tio(r

efer

ence

sam

ple

/ tes

t sam

ple)

1:1

1:1/10

1:1/100

1:1/1000

1:1 1:1/10 1:1/100 1:1/1000Input miRNA ratio

(reference sample / test sample)

Nor

mal

ized

read

cou

nt ra

tio(r

efer

ence

sam

ple

/ tes

t sam

ple)

1:1 1:1/10 1:1/100 1:1/1000Input miRNA ratio

(reference sample / test sample)

Individual miRNA masses in zmol ( = 10-21 mol) Individual miRNA masses in zmol ( = 10-21 mol)

Test sample 170 17 1.7 0.17 43 4.3 0.43 0.043

Reference sample 170 170 170 170 43 43 43 43

A

Test sampleReference sample

B

1:1

1:1/10

1:1/100

1:1/1000