Influence of Next-Generation Sequencing and Storage Conditions on

Jul 24, 2015 - The PAXgene system is a commonly used blood storage system for miRNA analysis. ... The Supporting Information is available free of char...
0 downloads 0 Views 2MB Size
Page 1 of 20

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

The influence of Next-Generation Sequencing and storage conditions on miRNA patterns generated from PAXgene blood Christina Backes1, Petra Leidinger2, Gabriela Altmann3, Maximilian Wuerstle3, Benjamin Meder4, 1 1 3 3 2 Valentina Galata , Sabine C. Mueller , Daniel Sickert , Cord Stähler , Eckart Meese , Andreas 1,* Keller

1 2 3 2

Chair for Clinical Bioinformatics, Saarland University, Saarbrücken, Germany Institute of Human Genetics, Saarland University, Homburg, Germany Siemens AG, Germany Internal Medicine II, University Hospital Heidelberg, Heidelberg, Germany

* corresponding author: Andreas Keller, [email protected], Chair for Clinical Bioinformatics, Saarland University, 66123 Saarbrücken, Phone: +49 681 302 68611

17

ACS Paragon Plus Environment

1

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 20

Abstract

18 19 20

Whole blood derived miRNA signatures determined by Next-Generation Sequencing (NGS)

21

offer themselves as future minimally invasive biomarkers for various human diseases. The

22

PAXgene system is a commonly used blood storage system for miRNA analysis. Central to

23

all miRNA analyses that aim to identify disease specific miRNA signatures, is the question of

24

stability and variability of the miRNA profiles that are generated by NGS.

25

We characterized the influence of five different conditions on the genome wide miRNA

26

expression pattern of human blood isolated in PAXgene RNA tubes. In detail, we analyzed

27

15 miRNomes from three individuals. The blood was subjected to different numbers of

28

freeze / thaw cycles and analyzed for the influence of storage at -80 °C or 8 °C. We also

29

determined the influence of blood collection and NGS preparations on the miRNA pattern

30

isolated from a single individual, which has been sequenced 10 times. Here, five PAXGene

31

tubes were consecutively collected that have been split in two replicates, representing two

32

experimental batches. All samples were analyzed by Illumina NGS. For each sample,

33

approx. 20 million NGS reads have been generated. Hierarchical clustering and Principal

34

Component Analysis (PCA) showed an influence of the different conditions on the miRNA

35

patterns. The effects of the different conditions on miRNA abundance are, however, smaller

36

than the differences that are due to inter-individual variability. We also found evidence for an

37

influence of the NGS measurement on the miRNA pattern. Specifically, hsa-miR-1271-5p

38

and hsa-miR-182-5p showed coefficients of variation above 100% indicating a strong

39

influence of the NGS protocol on the abundance of these miRNAs.

40 41 42 43

Keywords: miRNA; PAXgene, Next-Generation Sequencing, technical evaluation

ACS Paragon Plus Environment

2

Page 3 of 20

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

44

Analytical Chemistry

Introduction

45 46

For the identification of biomarkers and even more for the translation from basic research to

47

clinical routine, it is crucial to understand how markers vary depending on different storage

48

conditions and technical analysis. Especially for complex marker profiles like miRNA

49

signatures, a systematic bias will compromise their diagnostic and prognostic values. While

50

tissue based miRNA profiles have first been in the focus of research, there are increasing

51

efforts to identify miRNA signatures as non- or minimally invasive markers in body fluids

52

such as blood, serum or urine. Besides Heparin and EDTA blood tubes, PAXgene blood

53

RNA tubes have frequently been used to collect patients’ blood. Examples of PAXgene

54

blood RNA pattern include biomarkers for myocardial infarction1, lung cancer2,3, multiple

55

sclerosis4,5, melanoma6, ovarian cancer7, chronic obstructive pulmonary disease8,

56

glioblastoma9 and Alzheimer´s disease10. More recently, miRNA profiles of single blood cell

57

types or exosomes have been accomplished11,12.

58

To obtain profiles of miRNAs, different technologies have been applied. In the early stages

59

of miRNA profiling, microarrays have been widely used to generate miRNA patterns. High-

60

throughput qRT-PCR platforms also enable the parallel measurement of hundreds of

61

miRNAs. As the most recent technology, Next-Generation Sequencing (NGS) generates

62

millions of short reads that can be aligned to known miRNAs annotated in the miRBase13.

63

Likewise, new miRNA candidates can be predicted by aligning the fragments to the target

64

genome. To facilitate clinical applications, other methods such as immunoassays are

65

currently developed14. Technical stability of the profiles for reliable biomarker discovery is of

66

high impact, independent of the applied screening technique. Since NGS is increasingly

67

applied to generate patient based miRNA signatures we investigated NGS-related variability

68

and stability of miRNA pattern that are derived from PAXgene samples. In detail, we

69

investigate three main question: First, markers are frequently discovered in retroperspective

70

studies. Often, samples that are stored in biobanks are thawed and a part of the sample is

71

used for measurement. We asked whether additional freeze / thaw cylces have a significant

72

influence of miRNAs. Second, we asked how storage at 8 degree Celsius relates to the

73

samples stored at -80 degree Celsius. The time was thereby restricted to four days. Third,

74

we also investigated the influence of NGS on the variability / stability of miRNA patterns by

75

performing two NGS batches echa containing technical replicates of five PAXgene blood

76

tubes, which were all taken from the same individual.

77 78 79

ACS Paragon Plus Environment

3

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 20

80

Materials and Methods

81 82

Study set-up and miRNA profiling

83

In this study we focused on the influence of freeze / thaw cycles and short time storage of

84

PAXGene blood samples. We performed 25 miRNA measurements from 4 individuals. To

85

minimize the influence of pathogenic processes healthy individuals without known diseases

86

affection were investigated. All blood donors participating in this study gave their informed

87

consent. For each of the first three individuals we collected 5 PAXgene Blood RNA tubes.

88

The first tube has been stored at -80 °C for 4 days, while the second tube has been frozen at

89

-80 °C and was subsequently subjected to one additional freeze / thaw cycle on the first day

90

and finally stored again at -80 °C for the remaining days. The third tube has been subjected

91

to an additional freeze / thaw cycle on the first day, frozen again at -80 °C, subjected to a

92

second additional freeze /thaw cycle on the second day and stored at -80 °C for the

93

remaining days. The fourth tube has been subjected to one freeze / thaw cycle on the first

94

day, a second freeze / thaw cycle on the second day, a third additional freeze / thaw cycle

95

on the third day and stored at -80 °C for the fourth day. The fifth tube has been stored at 8

96

°C for four days. The study set-up is sketched in Figure 1A. For all samples independent

97

miRNA isolation using the PAXgene Blood miRNA Kit and individual library preps using the

98

Illumina TruSeq small RNA Library Prep Kit have been generated according to

99

manufacturer’s instruction.

100

To determine the influence of NGS on the miRNA pattern, NGS was performed on miRNAs

101

isolated from a single individual. We have taken five PAXgene tubes from this donor

102

(storage at -80 °C), isolated the RNA and performed two batches of library preps and NGS

103

runs on these five RNA eluates (10 miRNomes). An overview of the study set-up is

104

illustrated in Figure 1B.

105

For each of the 25 libraries, Illumina HiSeq2500 runs have been carried out according to

106

manufacturer’s instruction. For all samples around 20 Million raw sequencing reads have

107

been generated. All miRNA extractions and sequencing runs have been carried out by

108

CeGaT GmbH (Tübingen, Germany).

109 110 111

Bioinformatics analysis

112

We preprocessed the raw sequencing data as described previously10. In brief, the reads

113

were mapped against the current miRBase v21 sequences by using the miRDeep2

114

pipeline15. The raw miRBase counts for all samples were summarized in an expression

115

matrix.

ACS Paragon Plus Environment

4

Page 5 of 20

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

116

In order to carry out hierarchical clustering and calculate heat maps source code from the

117

heat map.2 function, provided as part of the “gplots” CRAN package (version 2.12.1) has

118

been used. In more detail, hierarchical clustering relying on the Euclidian distance has been

119

carried out on quantile normalized data (normalization has been done by the

120

“preprocessCore” package using standard parameters). As alpha level, 0.05 has been used

121

through the manuscript. If not stated explicitly, p-values have been adjusted for multiple

122

testing using the Benjamini-Hochberg approach16. Analysis of variance has been calculated

123

by using the “anova” package.

124 125 126 127 128

ACS Paragon Plus Environment

5

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

129

Page 6 of 20

Results & Discussion

130 131 132

In total, we generated 25 miRNomes from four individuals by NGS. 15 miRNomes have

133

been derived from three individuals to determine the influence of different storage conditions

134

and freeze / thaw cycles (see Figure 1A) and from the fourth individual 10 NGS runs have

135

been performed (5 consecutive blood drawings, two technical replicates that have been

136

processed in batches, see Figure 1B).

137 138

The expression values of blood miRNAs showed a high dynamic range of approximately 7

139

orders of magnitude. Out of 2,588 human miRNAs annotated in miRBase 21, we found

140

1,252 different miRNAs in at least one of the 25 samples analyzed. 1,060 miRNAs remained

141

after markers covered by just a single read were removed. Regarding the distribution of NGS

142

reads to miRNAs we observed significant variations. 90.3% of all reads matched to hsa-miR-

143

486-5p and 5% to hsa-miR-92a-3p. The remaining 4.7% reads (approx. 20 million reads),

144

matched to 1,250 miRNAs. To minimize a potential bias introduced by very low abundant

145

miRNAs, we empirically determined a threshold of 50 counts and continued the analysis with

146

a remaining set of 455 miRNAs. All miRNAs with absolute read count and percentage of all

147

reads mapping to this miRNA are summarized in Supplemental Table 1.

148 149

Influence of storage conditions and additional freeze / thaw cycles on miRNA patterns

150

from human blood

151

In order to determine the effect of different storage conditions and additional freeze / thaw

152

cycles on the miRNA pattern, we first employed cluster analysis and Principal Component

153

Analysis (PCA) as two commonly applied statistical approaches, on the set of 455 miRNAs.

154

In detail, we performed a complete linkage hierarchical clustering using the Euclidian

155

distance as distance measure. The results for all miRNAs measured under five different

156

conditions for three individuals, are summarized in Figure 2 as heat map with dendrograms

157

on top for the storage conditions and on the left side for the miRNAs. The heat map indicates

158

clustering of samples that have been stored throughout the experiment at -80 °C without

159

additional freeze / thaw cycles (indicated by an orange color in the heat map). Likewise, the

160

samples that have been stored at 8 °C for four days without changing the storage condition

161

(indicated with green color in the heat map), cluster together as well. In addition, samples

162

that have been stored at -80 °C without thawing cluster together with samples that have

163

been stored at -80 °C with only one additional freeze / thaw cycle (indicated by an light blue

ACS Paragon Plus Environment

6

Page 7 of 20

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

164

color in the heat map). We also observed a clustering of samples that are stored at -80 °C

165

with three additional freeze / thaw cycles (indicated by a dark blue color in the heat map)

166

with samples that have been stored solely at 8 °C. The results may be indicative of an

167

overall influence of the time period during which a sample is stored either at -80 °C or at 8

168

°C. The PCA, which has been used to generate a 2-dimensional mapping of the high-

169

dimensional miRNA profiles, largely confirmed the clustering results.

170

After having investigated the systematic effects on the miRNA patterns, we analyzed

171

whether single miRNAs show differences under the tested storage conditions. Therefor an

172

analysis of variance (ANOVA) as well as the coefficient of variation (CV) have been applied.

173

ANOVA identified 41 markers that were significant according to raw ANOVA p-values and an

174

alpha level of 0.05. Since multiple markers were measured, the p-values had to be adjusted

175

for multiple testing, resulting in 5 miRNAs being still significant, including hsa-miR-320b

176

(p=0.0002), hsa-miR-320a (p=0.001), hsa-miR-16-5p (0.018), hsa-miR-18b-5p (0.037) and

177

hsa-miR-375 (0.0375). As one example of a significant miRNA, has-miR-375 is presented as

178

boxplot in Figure 3A. For miRNA-375 we found a significant difference between samples that

179

have been stored at 8 °C for four days and samples that have been stored at -80 °C. We did

180

not find a significant influence of the freeze / thaw cycles. All miRNAs along with the raw and

181

adjusted p-values are presented in Supplemental Table 2.

182

Finally, we addressed the question of the importance of the miRNA changes observed under

183

the different storage conditions. To this end, we compared the differences between the three

184

donors to the differences between the different storage conditions. An analysis of the

185

coefficient of variation highlighted 37 miRNAs with the standard deviations exceeding the

186

mean value (CV > 1). The mean value, standard deviation and CV for all miRNAs are

187

presented in Supplemental Table 3. Largest CV was calculated for miR-1291 (mean of 24.6,

188

standard deviation of 40.7, CV of 1.7). One example is provided in Figure 3B, where the log

189

of normalized read counts of the five different conditions for the three individuals is

190

presented for miR-99a-5p. For this miRNA, the mean value is 508, the standard deviation

191

624 and the CV 1.23. The variation of the five measurements for each of the individuals is

192

substantially smaller than the deviation between individuals 1 and 2 compared to individual

193

3, which had high read counts for this miRNA. In general, we observed that the variability of

194

miRNA abundance between different donors was higher than the miRNA variability under

195

different storage conditions.

196

We conclude that there is a general systematic influence of the storage conditions on miRNA

197

patterns. Largest variability was observed between storage at 8 degrees Celsius and

198

samples stored at -80 degree Celsius without additional freeze / thaw cycles. The specific

ACS Paragon Plus Environment

7

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 20

199

effect of the storage conditions has to be verified for each single miRNAs separately.

200

Importantly, the effects of storage conditions on miRNA abundance are generally smaller

201

than the differences due to inter donor variability.

202

While we investigated a short time period, the long time storage of samples has also been

203

investigated. Seelenfreund and co-workers reported that miRNA from PAXGene tubes can

204

be recovered even after periods of up to 4 years, if samples are frozen at -80 degrees

205

Celsius17. In this study, a subset of all known miRNAs analyzed by qRT-PCR has been

206

included. Viprey et al. considered even longer time periods of up to 5 years18. In detail they

207

evaluated the reliability of expression for a subset of 377 miRNAs by qRT-PCR. The authors

208

did not observe a correlation of miRNA abundance with storage time. As most stable

209

reference miRNAs, miR-26a, miR-28-5p, and miR-24 were identified. These miRNAs were

210

also not affected significantly in our study with respect to freeze / thaw cycles, indeed

211

rendering them as reasonable and stable reference markers.

212 213 214

Influence of NGS preparation on miRNA patterns from human blood

215

As mentioned above, two NGS runs with different library prep were performed with each on

216

miRNAs isolated from five PAXgene tubes from a single individual and thus resulting in 10

217

miRNomes. The five samples of the first NGS run have been processed together, and

218

likewise the samples of the second NGS run with a slightly changed preprocessing step as

219

mentioned in material & methods. The results of the analysis are summarized as heat map

220

in Figure 4. The dendrogram on top of the heat map show the clustering of different blood

221

tubes and sequencing procedures, while the dendrogram on the left side for the miRNAs.

222

The heat map indicates no clustering between the five different blood sample tubes

223

(indicated by five shadings of red). However, a strong clustering between the two different

224

NGS preparations is shown (indicated by an orange color for the first NGS run and a blue

225

color for the second NGS run). These results show a significant influence of the NGS

226

preparation on the miRNA pattern identified in human blood. The findings from this cluster

227

analysis were confirmed by a PCA.

228

We next addressed the question of importance miRNA changes observed for the NGS

229

preparations by analyzing the coefficient of variation. The analysis showed lower CV values

230

for the 10 miRNomes measured by two NGS preparations than for the miRNome obtained

231

under different storage conditions. The standard deviation exceeded the mean for only 21

ACS Paragon Plus Environment

8

Page 9 of 20

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

232

miRNAs the majority of which were low abundant. Only two miRNAs with expression levels

233

(normalized read counts >5) showed average CV values >1 including hsa-miR-182-5p with

234

average value of 2,824 and standard deviation of 2,960 resulting in a CV of 1.05 and hsa-

235

miR-1271-5p with average read count of 5.7 and standard deviation of 6.2 resulting in a CV

236

of 1.07. As indicated by the barplot of logarithmized normalized read counts in Figure 5A, the

237

second batch of NGS library preparation (indicated in blue) showed for hsa-miR-182-5p

238

significantly higher expression level as the first NGS library preparation (indicated in orange).

239

Notably, an analysis of raw read counts revealed the same behavior indicating that the

240

differences between the two NGS batches are not due to the normalization process (Figure

241

5B). The same values are presented in Figure 5C and 5D for miR-1271-5p. This miRNAs

242

was almost not present in the second NGS batch while substantially expressed in the first

243

batch. The CV values, mean and standard deviations for all miRNAs are presented in

244

Supplemental Table 4.

245

In sum, we found evidence for an influence of the NGS measurement on specific miRNAs’

246

profiles including the library preparation.

247

ACS Paragon Plus Environment

9

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 20

248

Conclusion

249 250

In this study we systematically explored the influence of different conditions and freeze /

251

thaw cycles on miRNA profiles generated by using NGS. Furthermore, we investigated the

252

stability and reproducibility of the respective miRNA patterns by carrying out 10 replicated

253

measurements of the same individual.

254

For selected miRNAs, we found an influence with respect to up to three additional freeze

255

thaw cycles. Directly processed samples showed overall closest proximity to samples

256

undergoing one freeze / thaw cycle. Interpreting the replicated measurements of the same

257

donor also revealed a certain degree of variability. Specifically, we observed that the

258

influence of the NGS procedure of miRNAs seems to be partially exceed the variability of the

259

the blood drawing and miRNA extraction. These results were determined by considering all

260

miRNAs for the analysis, with techniques such as clustering and PCA as well as considering

261

the coefficient of variation. However, the variability was only observed for a subset of all

262

miRNAs. It is therefore essential to be aware of the potential pitfalls of sample storage and

263

NGS preparation that can contribute to the variability of miRNAs. If specific case control

264

studies present these miRNAs as candidates to detect a disease, further in-depth validation

265

is required.

266

ACS Paragon Plus Environment

10

Page 11 of 20

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

267 268 269 270 271 272 273

Competing interest The study has been funded in parts by Siemens Healthcare.

274 275

Author’s contribution

276

CB, PL, VG, SM contributed in data analysis, GA contributed in study set up, coordinated

277

experiments, evaluated data, MW contributed in manuscript preparation and writing, DS

278

contributed in study design, CS contributed in study design and set up, contributed in

279

manuscript preparation, EM wrote the manuscript, AK wrote the manuscript and contributed

280

in data analysis.

281 282 283

Acknowledgement

284 285

This work has been funded in parts by internal funds of Saarland University, by Siemens

286

Healthcare and by the EU FP7 Project BestAgeing.

287 288 289

ACS Paragon Plus Environment

11

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327

Page 12 of 20

References (1) Meder, B.; Keller, A.; Vogel, B.; Haas, J.; Sedaghat-Hamedani, F.; Kayvanpour, E.; Just, S.; Borries, A.; Rudloff, J.; Leidinger, P.; Meese, E.; Katus, H. A.; Rottbauer, W. Basic Res. Cardiol. 2011, 106, 1323. (2) Keller, A.; Leidinger, P.; Borries, A.; Wendschlag, A.; Wucherpfennig, F.; Scheffler, M.; Huwer, H.; Lenhof, H. P.; Meese, E. BMC cancer 2009, 9, 353. (3) Leidinger, P.; Backes, C.; Blatt, M.; Keller, A.; Huwer, H.; Lepper, P.; Bals, R.; Meese, E. Mol. Cancer 2014, 13, 202. (4) Keller, A.; Leidinger, P.; Lange, J.; Borries, A.; Schroers, H.; Scheffler, M.; Lenhof, H. P.; Ruprecht, K.; Meese, E. PloS one 2009, 4, e7440. (5) Keller, A.; Leidinger, P.; Steinmeyer, F.; Stahler, C.; Franke, A.; Hemmrich-Stanisak, G.; Kappel, A.; Wright, I.; Dorr, J.; Paul, F.; Diem, R.; Tocariu-Krick, B.; Meder, B.; Backes, C.; Meese, E.; Ruprecht, K. Mult Scler 2013. (6) Leidinger, P.; Keller, A.; Borries, A.; Reichrath, J.; Rass, K.; Jager, S. U.; Lenhof, H. P.; Meese, E. BMC cancer 2010, 10, 262. (7) Hausler, S. F.; Keller, A.; Chandran, P. A.; Ziegler, K.; Zipp, K.; Heuer, S.; Krockenberger, M.; Engel, J. B.; Honig, A.; Scheffler, M.; Dietl, J.; Wischhusen, J. Br. J. Cancer 2010, 103, 693-700. (8) Leidinger, P.; Keller, A.; Borries, A.; Huwer, H.; Rohling, M.; Huebers, J.; Lenhof, H. P.; Meese, E. Lung cancer 2011, 74, 41-47. (9) Roth, P.; Wischhusen, J.; Happold, C.; Chandran, P. A.; Hofer, S.; Eisele, G.; Weller, M.; Keller, A. Journal of neurochemistry 2011, 118, 449-457. (10) Leidinger, P.; Backes, C.; Deutscher, S.; Schmitt, K.; Muller, S. C.; Frese, K.; Haas, J.; Ruprecht, K.; Paul, F.; Stahler, C.; Lang, C. J.; Meder, B.; Bartfai, T.; Meese, E.; Keller, A. Genome biology 2013, 14, R78. (11) Leidinger, P.; Backes, C.; Dahmke, I. N.; Galata, V.; Huwer, H.; Stehle, I.; Bals, R.; Keller, A.; Meese, E. What makes a blood cell based miRNA expression pattern disease specific? - A miRNome analysis of blood cell subsets in lung cancer patients and healthy controls, 2014. (12) Leidinger, P.; Backes, C.; Meder, B.; Meese, E.; Keller, A. BMC Genomics 2014, 15, 474. (13) Kozomara, A.; Griffiths-Jones, S. Nucleic Acids Res. 2011, 39, D152-157. (14) Kappel, A.; Backes, C.; Huang, Y.; Zafari, S.; Leidinger, P.; Meder, B.; Schwarz, H.; Gumbrecht, W.; Meese, E.; Staehler, C. F.; Keller, A. Clin. Chem. 2015, 61, 600-607. (15) Friedländer, M. R.; Mackowiak, S. D.; Li, N.; Chen, W.; Rajewsky, N. Nucleic Acids Res. 2011. (16) Benjamini, Y.; Hochberg, Y. Journal of the Royal Statistical Society. Series B. Methodological 1995, 57, 289-300. (17) Seelenfreund, E.; Robinson, S. E.; Amato, C. M.; Bemis, L. T.; Robinson, W. A. Biopreserv Biobank 2011, 9, 29-33. (18) Viprey, V. F.; Corrias, M. V.; Burchill, S. A. Anal. Biochem. 2012, 421, 566-572.

328 329

ACS Paragon Plus Environment

12

Page 13 of 20

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370

Analytical Chemistry

Figure Legends Figure 1: Study set up: 4 donors without known disease affection were included. For the first three donors 5 Blood Tubes were extracted and processed over the next four days (Panel A). For the fourth donor 5 blood tubes were extracted and handled in duplicates (Panel B). All patients are labeled by X-Y, where X is the patient number and Y the condition. Figure 2: Cluster heat map with dendrogram on top and left. Color scale for the clustering heat map is provided in the upper left corner. The samples are colored as follows: The three different cycles are shown in three different blue shadings above the heat map, orange samples have been directly handled and the four-day refrigerator samples are colored in green. The three different cycles that are shown in three different blue shadings above the heat map mix between the orange samples that have been directly handled (clustering on the left side) and the four-day fridge control samples that are colored in green (clustering on the right side). This left cluster also contains the samples that have been frozen and thawed once while the right cluster contains the samples with three freeze / thaw cycles. PCA, which has been used to generate a 2-dimensional mapping of the high-dimensional miRNA profiles, confirmed these clustering results

Figure 3A: Boxplot resulting from the ANOVA for conditions 1-5 and donors 1-3. For miR375 signals increase for samples stored for four days at the refrigerator. The color scheme corresponds to Figure 2.

Figure 3B: log of normalized read counts for conditions 1-5 and donors 1-3 for miR-99a-5p. The overall high variability is due to the overall higher expression of that miRNA in donor 3 as compared to the first two donors. Figure 4: Cluster heat map with dendrogram on top. The samples are colored with respect to the blood drawing (first row on top of the heat map) and according to the analytical NGS batch (second row on top of the heat map).

Figure 5: Normalized (Panel A) and raw (Panel B) read counts for hsa-miR-182-5p. Shown are logarithmized reads for five replicated blood drawings (denoted 4 1, 4 2, 4 3, 4 4 and 4 5) in two NGS batches. The first batch is highlighted in orange the second in blue. Panels C and D present the same analysis for the second variable miRNA, has-miR-1271-5p (please note the absolute scale in contrast to the logarithmic for Panel A and B).

ACS Paragon Plus Environment

13

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 14 of 20

371 372

ACS Paragon Plus Environment

14

Page 15 of 20

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

373 374

ACS Paragon Plus Environment

15

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 16 of 20

375 376 377

ACS Paragon Plus Environment

16

Page 17 of 20

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

378 379 380 381 382

ACS Paragon Plus Environment

17

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 18 of 20

383 384

ACS Paragon Plus Environment

18

Page 19 of 20

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

385

ACS Paragon Plus Environment

19

Analytical ChemistryPage 20 of 20

50000   40000   30000   20000  

10000  ACS 0  

stored  at  8°C  

60000  

3  freeze-­‐thaw  

70000  

2  freeze-­‐thaw  

80000  

1  freeze-­‐thaw  

NGS  read  count  

1 2 3 4 5 6 7 8 9 10

stored  at  -­‐80°C  

90000  

Paragon Plus Environment