Exploiting the dynamic relationship between ... - ACS Publications

Subscriber access provided by UNIV OF LOUISIANA

Article

Exploiting the dynamic relationship between peptide separation quality and peptide co-isolation in a mPSM approach offers a new strategy to optimize bottom-up proteomics throughput and depth Manuel I. Villalobos Solis, Richard J. Giannone, Robert L. Hettich, and Paul E. Abraham Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.9b00819 • Publication Date (Web): 10 May 2019 Downloaded from http://pubs.acs.org on May 11, 2019

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 26 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

1 2 3 4 5 6

Analytical Chemistry

Exploiting the dynamic relationship between peptide separation quality and peptide coisolation in a mPSM approach offers a new strategy to optimize bottom-up proteomics throughput and depth Manuel I. Villalobos Solis1,2, Richard J. Giannone1, Robert L. Hettich1 and Paul E. Abraham1

7 8 9 10 11

1

12

*Corresponding author:

Chemical Sciences Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, United States. 2 Department of Genome Science and Technology, University of Tennessee, Knoxville, Tennessee 37996, United States.

13 14 15 16 17 18 19

Paul E. Abraham Oak Ridge National Lab, Oak Ridge, TN 37831 Email: [email protected] Phone: (865) 574-4968 Fax: (865) 241-1555

20 21 22 23

Keywords: tandem mass spectrometry, chimeric spectra, isolation interference, liquid chromatography

24 25 26 27 28 29 30 31 32 33 34 35 36

Notice: This manuscript has been authored by UT-Battelle, LLC under Contract No. DE-AC0500OR22725 with the U.S. Department of Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/downloads/doe-public-access-plan).

0 ACS Paragon Plus Environment

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 26

37

ABSTRACT

38

Peptide co-fragmentation leads to chimeric MS/MS spectra that negatively impact traditional

39

single-peptide match-per-spectrum (sPSM) search strategies in proteomics. The collection of

40

chimeric spectra is influenced by peptide co-elution and the width of precursor isolation windows.

41

Although peptide co-fragmentation can be reduced by advanced chromatography, such as UHPLC

42

and 2D-HPLC separation schemes, and narrower isolation windows, chimeric spectra can still be

43

as high as 30-50% of the total MS/MS spectra collected. Alternatively, co-fragmented peptides in

44

chimeric spectra and the use of wider isolation windows benefit multiple-peptide matches-per-

45

spectrum (mPSM) algorithms, such as CharmeRT, which facilitate the identification of several co-

46

fragmented peptides. Considering recent advancements in LC and mPSM methodologies, we

47

present a comprehensive examination of the levels of chimeric spectra collected in the analysis of

48

a HeLa digest measured using different LC modes of separation and isolation windows and

49

compare the depth of identifications obtained when these data are annotated using a sPSM or a

50

mPSM approach. Our results demonstrate that MS/MS data derived from 1D-HPLC strategies

51

under different gradient schemes and searched with CharmeRT yielded higher average numbers

52

of PSMs (11%-49%), peptide analytes (10%-16%), and peptide sequences (3%-10%) compared to

53

data derived from 1D-UHPLC runs but searched with a sPSM strategy. Interestingly, data from a

54

2D-HPLC separation strategy benefits more from the application of CharmeRT results when

55

compared to a 50 cm 1D-UHPLC column employing a 500 min gradient. Overall, these results

56

provide new insights into how to better configure LC-MS/MS measurements for improved

57

throughput and peptide identification in complex proteomes.


Page 3 of 26 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

58


INTRODUCTION

59

Despite advances in mass accuracy, resolving power, and scan speeds in mass spectrometry

60

instrumentation, one of the remaining challenges of any high-throughput bottom-up proteomics

61

experiment is that only a fraction of the collected tandem mass spectra (MS/MS) can be assigned

62

to peptide sequences with high confidence (usually ≤ 60%).1 Although several reasons contribute

63

to this effect,2 one that has been under scrutiny by the proteomics community is the occurrence of

64

chimeric spectra (also known as mixture or co-fragmented spectra).

65

Chimeric spectra are the result of the co-isolation and co-fragmentation of two or more

66

peptide precursor ions with similar m/z and retention time. The complex nature of the samples

67

commonly analyzed in bottom-up proteomics (>100,000 detectable peptide species) and the m/z

68

isolation widths typically used in data-dependent acquisition (DDA) experiments (2-4 m/z)1, 2 can

69

result in chimeric spectra representing 50% of the total MS/MS data collected.2 Recent

70

investigations employing tryptic digests of HeLa cells demonstrated that even in experiments with

71

a narrow isolation width of 2 m/z, 39% of the total MS/MS spectra collected is chimeric.3

72

The negative effects of chimeric spectra in proteomics studies have been well-studied. For

73

example, in database-driven peptide and protein identifications, the presence of chimeras

74

deteriorates the search scores of true peptide assignments given by algorithms such as MASCOT

75

and SEQUEST.2 In addition, chimeric spectra reduces the accuracy of quantitative isobaric

76

tagging-based quantification methods such as iTRAQ or TMT, in which the contribution of

77

reporter ion intensities from co-fragmented peptides, causes the under-estimation of

78

protein/peptide abundance differences (a phenomenon termed as “ratio compression”).4-6 Due to

79

these issues, experimental and computational approaches that minimize the negative impact of

80

chimeric spectra in bottom up experiments have been developed.



Page 4 of 26

81

Reducing the complexity of samples prior to MS analysis has the advantage of minimizing

82

the chance of co-eluting peptides.7, 8 By exploiting independent physicochemical properties of

83

peptides, protocols coupling orthogonal chromatographic separations before MS detection have

84

shown improved separation resolutions and increased peptide and protein identifications, albeit at

85

the cost of increased analysis times.9-11 With the introduction of ultra-high-pressure

86

chromatography (UHPLC), a new era of high-quality one-dimensional separations has resurfaced.

87

The use of chromatographic pumps that can tolerate up to 10,000 psi and stationary phases with

88

particles sizes of < 2 µM diameter,12, 13 not only has afforded narrower peptide elution profiles and

89

increased ion sensitivities, but the improved column capacities also reduces co-elution of peptides

90

and hence, the occurrence of chimeric spectra.1 This mode of operation has enabled “single-shot”

91

proteomics studies, in which complex proteomes are analyzed in-depth with the aid of 1D reverse

92

phase chromatographic columns of 50 cm or more in length and effective LC gradients times

93

ranging from 300-500 mins.12, 14-16 Although 1D-UHPLC (including “one-shot” separations) and

94

2D-HPLC based separations reduce the occurrence of chimeric spectra, even under the best

95

chromatographic separations, the co-elution of thousands of peptides is still unavoidable.1

96

Computationally, database searching algorithms aiming to deconvolute chimeric spectra

97

collected in DDA experiments have been developed.17-21 These algorithms make use of the

98

multiple peptides-per-spectrum-match approach (mPSM) that, in comparison to most commonly

99

employed search strategies using a single-peptide match-per-spectrum approach (sPSM), try to

100

assign more than one peptide per MS/MS spectrum. A newly developed computational workflow

101

called CharmeRT demonstrated substantial improvement in peptide identifications when

102

compared to other methods.3 This was achieved by implementing a second search strategy coupled

103

with a highly accurate retention time prediction algorithm method. The second search option of


Page 5 of 26 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


104

CharmeRT is integrated into the database search engine MSAmanda, with validation of multiple

105

peptide assignments to a given MS/MS spectrum performed by Elutator, a new tool built upon the

106

foundations of Percolator that incorporates retention time (RT) prediction in FDR calculations.

107

The impact of having RT prediction in FDR evaluations of first and second searches translated

108

into 25%-62% more peptide identifications in runs of HeLa tryptic digests compared to more

109

frequently used database search workflows.

110

Further advancements in LC and MS technology will continue to improve the limit of

111

detection for peptide sequencing in complex mixtures; however, these alone are unlikely to solve

112

the problem with chimeric spectra. We contend that a concomitant evaluation of LC configurations

113

and mPSM search algorithms is required to further increase the number of detectable peptides.

114

Therefore, our study was designed to systematically evaluate several LC peptide separation

115

techniques, ranging from short HPLC gradients to UHPLC and orthogonal separations, as well as

116

precursor isolation m/z windows, to better understand the effects of each on the subsequent

117

assignment of MS/MS spectra by a traditional sPSM search strategy compared to a mPSM one,

118

specifically CharmeRT.

119 120

EXPERIMENTAL PROCEDURES

121

Standard sample for LC-MS/MS analyses

122

Commercial Pierce HeLa Protein Digest Standards were purchased from Thermo Fisher

123

Scientific (20 µg total amounts). The lyophilized peptides were resuspended in 40 µL of water

124

with 0.1% formic acid as per the manufacturer instructions. Stock solutions of ~0.5 µg/µL were

125

frozen at -80°C and used as needed. Volumes equivalent to 2 µg of HeLa protein digest were

126

analyzed each time using different LC setups and gradients as described below. 4 ACS Paragon Plus Environment


127 128

Page 6 of 26

1D-LC-MS/MS runs of HeLa tryptic digest standards

129

1D-LC-MS/MS runs were carried out using a Proxeon EASY-nLC 1200TM system (Thermo

130

Fisher Scientific) interfaced with a Q Exactive Plus (Thermo Fisher Scientific) mass spectrometer

131

equipped with a nano-electrospray source. For HPLC measurements, a single 2 µg injection of

132

HeLa peptides were loaded in mobile phase A (0.1% formic acid, 2% acetonitrile) each time onto

133

a trap column (150 mm x 100 μm ID) packed in-house with ~10 cm of 5 μm Kinetex C18 resin

134

(Phenomenex). Peptide separation was conducted on an analytical column (250 mm x 75 μm ID)

135

packed in-house with 5 μm particle size Kinetex C18 resin using a linear gradient from 2 to 22% of

136

mobile phase B (0.1% formic acid, 80% acetonitrile) at a flow rate of 300 nL/min over 90, 210, or

137

240 min depending on the experiment. Each gradient was followed by an increase to 35% B within

138

5 minutes, a 5 mins hold in 35% B, and afterwards a decrease to 2% B in 5 minutes.

139

To test the effects of smaller particle diameter size in number of identifications, the same

140

LC gradients were used in an UHPLC setup, where the trap and analytical front columns were

141

packed with 1.7 μm particle size Kinetex C18 resin. In addition, a single 2 µg injection of HeLa

142

peptides was separated with an analytical 500 mm x 75 μm ID column packed with 1.7 μm Kinetex

143

C18 which was placed in a column heater (Sonation GmbH) at a temperature of 60°C. The linear

144

gradient for this configuration was from 2 to 22% solvent B over 500 mins at a flow rate of 250

145

nL/min.

146

Optimization of relevant MS parameters in the Q Exactive instrument were performed and

147

were found to agree well with the Q Exactive benchmarking study.22 In brief, mass spectra were

148

acquired with the Q Exactive Plus instrument in a top 10 data-dependent acquisition setup. Peptide

149

precursor MS spectra was collected within 300 to 1500 m/z with automatic gain control (AGC)

150

target value of 3 × 106 at a resolution of 70,000 with a maximum injection time (IT) of 25 ms. 5 ACS Paragon Plus Environment

Page 7 of 26 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


151

Precursor ions with charge states ≥2 and ≤ 5 and intensity threshold of 1.6 × 105 were selected for

152

higher-energy C-trap collision dissociation (HCD) with a normalized collision energy of 27.

153

Peptide precursor ions collected from the 1D HPLC and UHPLC gradient runs were isolated using

154

a 1.6 m/z isolation width; whereas for the best gradients under HPLC or UHPLC conditions (see

155

Results & Discussions), precursor m/z isolation widths of 0.8 and 3.0 m/z were additionally

156

employed in order to test the effects of co-isolation interference. Fragment ion spectra were always

157

acquired at a resolution of 17,500 at m/z 200 with an AGC target value of 1 × 105 and maximum

158

IT of 50 ms. Dynamic exclusion was set to 20 s to avoid repeated sequencing of peptides. All runs

159

were conducted in triplicate.

160

2D-LC-MS/MS runs of HeLa tryptic digest standards

161

2D-LC-MS/MS runs were performed using a Vanquish UHPLC interfaced with a Q

162

Exactive Plus mass spectrometer (Thermo Fisher Scientific) outfitted with a 100 µM ID triphasic

163

precolumn (RP-SCX-RP) coupled to a 250 mm x 75 µM ID nanospray emitter packed with 250

164

mm of 5 µm Kinetex C18 RP resin.23 For each sample, a single 2 µg injection of HeLa peptides

165

was loaded onto the precolumn with mobile phase A by direct flow (2 µL/min) then separated and

166

analyzed across two successive salt cuts of ammonium acetate (35 mM and 500 mM), with each

167

cut followed by a 210 min, split-flow (300 nL/min) organic gradient, wash, and re-equilibration:

168

0% to 2% solvent B over 2 min; 2% to 22% solvent B over 208 min; 22% to 50% solvent B over

169

10 min; 50% to 0% solvent B over 10 min, hold at 0% solvent B for 15 min. Mass spectra from

170

the eluting peptides was collected using the same MS and MS/MS parameter settings on the Q

171

Exactive Plus instrument as in the 1D-LC-MS/MS runs.

172

PSMs, peptide, and proteins identifications by database search



Page 8 of 26

173

All MS/MS spectra collected were processed in Proteome Discoverer v.2.2. (PD) with

174

MSAmanda v.2.2 and Percolator. Spectral data were searched against the most up-to-date human

175

reference proteome database from UniProt (ID. UP000005640) to which common laboratory

176

contaminants were appended. The following parameters were set up in MSAmanda to derive fully-

177

tryptic peptides: MS1 tolerance = 5 ppm; MS2 tolerance = 0.02 Da; missed cleavages = 2;

178

Carbamidomethyl (C, +57.021 Da) as static modification; oxidation (M, +15.995 Da) and

179

carbamylation (n-terminus, +43.006 Da) as dynamic modifications. The percolator FDR threshold

180

was set to 1% at the PSM and peptide level. In addition, MS/MS spectral data were searched with

181

MSAmanda in which a second search option was enabled and Elutator v2.2 (the CharmeRT

182

workflow). Parameters applied for MSAmanda second search were as described in the original

183

CharmeRT publication, with the exception that a maximum of 3 additional precursors per PSMs

184

were searched (referred as second searches).3 The Elutator FDR threshold was set to 1% at the

185

PSM and peptide level.

186 187

Data analysis

188

The following identification parameters were considered to assess the performance of each

189

1D and 2D LC-MS/MS runs: number of protein groups, number of modification-specific peptides

190

with charge (referred to as peptide analytes), number of peptides without modification and charge

191

(referred to as peptide sequences) and number of peptide-spectrum matches (PSMs). In addition,

192

we considered the precursor isolation interference percentage calculated by Proteome Discoverer,

193

as a measure of chimerism in the spectra collected:

194

[ (

Eq.1 % 𝑖𝑠𝑜𝑙𝑎𝑡𝑖𝑜𝑛 𝑖𝑛𝑡𝑒𝑟𝑓𝑒𝑟𝑒𝑛𝑐𝑒 = 100 × 1 ―

𝑝𝑟𝑒𝑐𝑢𝑟𝑠𝑜𝑟 𝑖𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦 𝑖𝑛 𝑖𝑠𝑜𝑙𝑎𝑡𝑖𝑜𝑛 𝑤𝑖𝑛𝑑𝑜𝑤 𝑡𝑜𝑡𝑎𝑙 𝑖𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦 𝑖𝑛 𝑖𝑠𝑜𝑙𝑎𝑡𝑖𝑜𝑛 𝑤𝑖𝑛𝑑𝑜𝑤

)]


Page 9 of 26 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


195

The full width at half-maximum (FWHM) region was calculated for each LC configuration

196

using the FeatureFinderMetabo node24 of OpenMS25 using default parameters except for the

197

expected chromatographic peak width (in seconds) setting, which was optimized for each

198

configuration. All spectral data collected in this study was deposited at the ProteomeXchange

199

Consortium via the MASSIVE repository. The project accession is PXD012635 and reviewers can

200

access the data under the username [email protected] and password qd5l9bhm.

201 202

Results & Discussion

203

Performance metrics across a range of nanoLC peptide separation techniques

204

Two common LC peptide separation techniques are HPLC and UHPLC. The former

205

employs analytical columns packed with stationary phases having particle sizes > 2 µM and flow

206

rates that operate below 450 bar,26 while the latter uses stationary phases with particle sizes ≤ 2

207

µM that drive the backpressure of the LC system to 600-1300 bar.13 Our approach started with

208

investigation of the performance of 5 µM C18 packed HPLC and 1.7 µM C18 packed UHPLC

209

columns (250 mm x 75 μm ID) employing three different linear gradient lengths from 2 to 22%

210

solvent B over 90 min, 210 min and 240 min at a constant flow rate of 300 nL/min.

211

Not surprisingly, the 1D-UHPLC setups (250 mm x 75 μm ID column, 1.7 µM C18)

212

outperformed the 1D-HPLC setups (250 mm x 75 μm ID column, 5 µM C18) in average number

213

of identifications across the different gradients tested, with the best performance achieved in the

214

longer analysis time (Figures S-1A-D). These data are undoubtedly explained by UHPLC

215

providing narrower FWHM and boosting the sensitivity of the analysis with increased ion

216

intensities (Figure S-2).14 In addition, the overall improved chromatographic resolution provided

217

by UHPLC setups identified >90% of all the identified HPLC peptides and further yielded between



Page 10 of 26

218

38-44% new peptide sequences and 25-32% new protein groups not found by their HPLC

219

counterparts (Figure S-1E).

220

The comparison across gradient time lengths on the same LC setups demonstrated the peak

221

dilution phenomena in HPLC.27, 28 For example, when the gradient lengths were adjusted from 210

222

min to 240 min, minor returns in the number of identifications were observed for the UHPLC

223

setup, with a maximum gain of 12% in the number of PSMs and less than 5% for the remaining

224

identifications. However, for HPLC, only an increase of 6% PSMs was observed at 240 min, but

225

the numbers of other identifications were reduced between 2%-4%, thereby suggesting longer

226

gradients for the HPLC column results would likely bring in diminishing returns due to a decrease

227

in peak heights which subsequently leads to loss of sensitivity and resolution.

228

One way to combat peak dilution is to use extra-long 1D-UHPLC columns (> 300 mm) or

229

to fractionate the sample via 2D-HPLC. While both these different modes of enhanced LC have

230

been shown to offer in-depth proteome analyses with comparable identification metrics, we

231

hypothesized that they would likely lead to differing degrees of co-isolation. Therefore, both 2D-

232

HPLC (250 mm x 75 μm ID column, 5 µM C18) and long 1D-UHPLC (500 mm x 75 μm ID

233

column, 1.7 µM C18) configurations were employed to assess their overall influence on chimeric

234

MS/MS spectra. As observed in Figure S-3, the percentage increases in the average number of

235

identifications were significant for both the 2D-HPLC and the 1D-UHPLC 500 min runs when

236

compared to the results from the 250 mm HPLC and UHPLC configurations tested.

237 238 239

Comparison of precursor isolation interferences across a range of nanoLC peptide separation techniques

240

An obvious benefit provided by enhanced LC separation is a reduction in the number of

241

co-eluting peptides. As fewer peptides with similar m/z ratios co-elute, the amount and degree of 9 ACS Paragon Plus Environment

Page 11 of 26 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


242

interfering precursor ions during isolation is expected to decrease. To evaluate this premise across

243

the tested LC configurations, the degree of isolation interference imparted by co-eluting peptides

244

was computed.

245

Given the already established relationship between precursor ion abundance and the degree

246

of isolation interference in MS/MS spectra,2 the median isolation interference percentages of

247

identified PSMs in each LC-MS/MS run were plotted against different ranges of precursor ion

248

abundance (Figure 1). As expected, higher precursor abundances have lower median isolation

249

interferences in each LC configuration. In general, the majority of PSM precursor abundances

250

ranged from 1e07-1e10. In this range, higher median isolation interference was observed in the

251

1D-HPLC setups relative to the 1D-UHPLC setups, while the lowest isolation interferences were

252

found in the 500 min 1D-HPLC runs and 2D-HPLC. Between these more specialized LC setups,

253

precursors identified in the 2D-HPLC runs had a slightly lower median isolation interference

254

percentage. Plotting the 2D-HPLC values for the independent fractions of peptides collected at

255

each salt pulse revealed similar median isolation interference percentages across both fractions,

256

suggesting they have similar peptide complexity (Figure S-4).

257 258 259

Evaluation of the CharmeRT mPSM search algorithm across different LC peptide separation techniques

260

Dorfer et al. (2018) benchmarked CharmeRT against other widely-used search algorithms,

261

including MSAmanda, and demonstrated improved measurement depth. However, an in-depth

262

evaluation of the performance gains across a broad range of peptide separation techniques has not

263

been reported. Therefore, our objective was to compare the CharmeRT workflow (mPSM) against

264

its foundational algorithms, MS Amanda and Percolator (sPSM), across the various LC

265

configurations implemented. As expected, the application of CharmeRT increased the average

266

numbers of identified PSMs between 31%-63% for the HPLC runs, 56%-58% for the UHPLC 10 ACS Paragon Plus Environment


Page 12 of 26

267

runs, and up to 64% and 26% for the more specialized 2D-HPLC and 500 min 1D-UHPLC runs,

268

respectively. These numbers translated to increases in the number of peptide sequences identified

269

overall with gains of 32%-55% per HPLC run, 45%-41% per UHPLC, 51% per 2D-HPLC and

270

22% per 500 min 1D-UHPLC (Figure 2). The percentage gains in the average number of peptide

271

sequences did not translate into comparable percentage of increases in the average number of

272

protein groups. For example, we observed a slight decrease (~2-3%) in the number of protein

273

groups identified in the 500 min 1D-UHPLC measurements, which we suspect is a reflection of

274

saturation in protein measurement depth; this observation is also expected to vary across peptide

275

mixtures of different complexities. Nearly all peptides and protein groups identified with

276

MSAmanda-Percolator were also found with CharmeRT (Figure S-5). Moreover, most peptides

277

identified from the second search mapped to protein groups identified in the first search (Figure

278

S-6). Overall, these observations agreed well with results reported in the original CharmeRT

279

publication.

280

A major goal of this study was to better understand how varying the quality of peptide

281

separation influences the performance of the CharmeRT search strategy and to determine whether

282

one could achieve similar depth using either a UHPLC with MSAmanda-Percolator or an HPLC

283

configuration with CharmeRT. A comparison of the average number of PSMs, peptide analytes,

284

and peptide sequences between these two approaches revealed that HPLC combined with

285

CharmeRT provided substantially more identifications than the conventional UHPLC Amanda-

286

Percolator search (Figure 2). More importantly, comparable performance was observed between

287

the HPLC 210 min CharmeRT and UHPLC 240 min Amanda-Percolator using the shorter

288

analytical column (250 mm x 75 μm ID). Intriguingly, this observation suggests that peptide

289

separation of a complex mixture on a HPLC column with shorter gradient times in combination


Page 13 of 26 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

with a mPSM search workflow can achieve identification metrics comparable to results given by

291

a UHPLC column and a sPSM search strategy.

292

Next, the degree of overlap between the newly identified peptides provided by either

293

UHPLC or the application of CharmeRT was evaluated. More specifically, the newly identified

294

peptide sequences and protein groups identified from HPLC runs searched with CharmeRT were

295

compared to the results obtained from their UHPLC counterparts searched with MSAmanda-

296

Percolator (Figure S-7). These comparisons demonstrate that the newly identified peptides for

297

both approaches were mostly complementary, with only a small overlap (~10%).

298

When comparing the CharmeRT performance between 2D-HPLC and the 1D-UHPLC 500

299

min gradient runs (500 mm x 75 μm ID), the 2D-HPLC measurements had the highest percentage

300

gains in terms of average number of PSMs, peptide analytes, peptide sequences and protein groups

301

(64%, 63%, 51%, 13%, respectively) (Figure 2), despite 2D-HPLC having the least interference

302

(Figure 1). To explore this further, the total number of PSMs derived from the 2D-HPLC and 1D-

303

UHPLC 500 min setups at different ranges of isolation interference were binned and quantified

304

based on their CharmeRT search depth (Figure 3, isolation width 1.6 m/z). Interestingly, across

305

every isolation interference bin, the 2D-HPLC configuration afforded 2-4x more CharmeRT gains

306

relative to the 1D-UHPLC 500 min configuration. Additionally, a greater search depth (i.e., a

307

higher percentage of scans having 2-3 additional precursors derived from second searches) was

308

observed at higher isolation interference ranges in 2D-HPLC. After exploring accompanying data

309

related to data quality and PSM scoring, it’s not immediately obvious what metrics explain this

310

phenomenon. However, a likely explanation is that the 2D separation scheme, which reduces the

311

complexity of the loaded peptide mixture for each salt pulse, retains the CharmeRT performance



Page 14 of 26

312

gains like the 1D-HPLC 210 min separation scheme, whereas the 1D-UHPLC 500 min CharmeRT

313

performance gains suffers from dilution of lower abundant secondary precursors.

314 315 316

Evaluation of spectra collected from LC peptide separation techniques with different isolation windows and searched with CharmeRT

317

Widely-acknowledged, precursor isolation widths can impact the performance in LC-

318

MS/MS measurements: narrower isolation widths lead to loss in signal and wider isolation widths

319

results in more precursor co-isolation/co-fragmentation. As previously demonstrated by Dorfer et

320

al., (2018) applying wider isolation widths during data acquisition improves the performance gains

321

of the CharmeRT workflow. Therefore, our approach was to evaluate different isolation widths

322

across the different LC configurations applied in this study.

323

To this end, three isolation widths (0.8, 1.6 and 3.0 m/z) that encompass the range often

324

employed in LC-MS/MS measurements were applied and performance assessed across the 1D-

325

HPLC 210 min and 1D-UHPLC 240 min setups, which gave the highest numbers of average

326

identifications across each configuration either when the results were searched with Amanda-

327

Percolator or CharmeRT, as well as the 1D-UHPLC 500 min and 2D-HPLC setups.

328

At each isolation width, the 1D-HPLC 210 min runs had the highest precursor isolation

329

interference medians, followed by the 1D-UHPLC 240 min runs. The 1D-UHPLC 500 min and

330

2D-HPLC setups presented similar isolation interference values (Figure 4A). Given the lower

331

signal and reduced potential for co-fragmentation, the narrower isolation width of 0.8 m/z had the

332

fewest identifications and lowest gains by the CharmeRT workflow (Figures 4B and S8). Similar

333

to the results observed above, CharmeRT gains in peptide and protein group identifications were

334

higher in the HPLC configurations relative the UHPLC setups, particularly the 2D-HPLC which

335

experienced the widest range in peptide identifications across the three different isolation widths.


Page 15 of 26 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


336

In agreement with the original CharmeRT publication, our results revealed CharmeRT

337

improvements as isolation widths increased, except for the 1D-UHPLC 500 min configuration.

338

This is counterintuitive, as the increase in precursor isolation interference becomes seemingly

339

more favorable for the CharmeRT workflow (Figure 3, isolation width 3 m/z and Figure S-9).

340

This novel insight into mPSM expectations implies that long 1D-UHPLC gradient times do not

341

experience the same benefit in the CharmeRT workflow as observed for the 2D-HPLC and other

342

more common HPLC configurations tested in this study.

343

Conclusions

344

By using state-of-the-art LC-MS platforms, proteomes of simple organisms can now be

345

sequenced to near completion in just over an hour.29 However, comparable coverage is still not

346

readily achieved in more complex peptide mixtures, like those derived from higher-order

347

eukaryotes or microbial communities. While steady advancements in MS technology and LC will

348

continue to transform the achievable throughput and depth of proteomic analyses, the results

349

presented herein suggest that a better understanding of the dynamic relationship between peptide

350

separation and co-fragmentation can be leveraged by mPSM approaches, such as CharmeRT, to

351

enhance saturation in peptide identifications with minimal measurement duration or without

352

advanced LC configurations, i.e. 500 mm long, heated, UHPLC-driven peptide separations.

353

Moreover, this approach coupled to more specialized LC configurations, like 2D-HPLC, offers

354

even greater depth when compared to more traditional single-PSM approaches.

355

While our results revealed an improvement in peptide separation leads to the expected

356

increases in the number of peptides identified, similar gains can be achieved using shorter gradients

357

coupled with a mPSM approach. This performance differential became less as the peptide

358

separation improved. This observation has substantial implications for LC-MS/MS approaches



Page 16 of 26

359

that use HPLC columns, as it suggests that a new approach towards improving peptide

360

identification rates is to reduce peptide separation while employing a mPSM approach. Moreover,

361

for the 250 mm x 75 μm ID HPLC and UHPLC columns, a similar depth can be achieved when

362

HPLC measurements employ the CharmeRT wofklow. Importantly, this suggests that research

363

laboratories separating complex mixtures on a HPLC column with shorter gradient times can

364

achieve comparable identification metrics as if the sample were separated on a UHPLC column.

365

To maximize peptide identification and dynamic range of a proteome measurement, there

366

are many approaches that have been introduced in recent years. Many LC-MS proteome specialists

367

view the use of longer columns, packed with sub-2 μm separation particles, as the most direct

368

configuration toward improving peptide identifications.14 Yet, our results offer an alternative

369

perspective, in which the application of CharmeRT performed substantially better for a 2D-HPLC

370

strategy when compared to a long 1D-UHPLC column. Therefore, the 2D-HPLC strategy

371

combined with mPSM represents the best path toward improving the number of detectable

372

peptides, enabling a more rapid and comprehensive proteome analysis when compared to the

373

longer UHPLC columns. Moreover, because the demands of UHPLC configurations require LC

374

platforms that withstand very high pressures, which are often unavailable to non-specialists,

375

MS/MS spectral data derived from HPLC separations searched with a mPSM approach could be

376

become broadly adopted by the proteomics community.

377 378

Acknowledgments

379

This work was supported by the Plant-Microbe Interfaces Science Focus Area supported by the

380

U.S. Department of Energy (DOE) and the Office of Biological and Environmental Research

381

(OBER). The manuscript has been authored by UT-Battelle, LLC, under contract no. DE-AC05-

382

00OR22725 with the U.S. Department of Energy. 15 ACS Paragon Plus Environment

Page 17 of 26 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


383

Supporting Information

384

A total of nine supporting figures (S-1 to S-9) with additional identification results and analyses

385

are presented in the Supplementary Results section.

386

Conflict of Interest Disclosure

387

The authors declare no financial conflicts of interest.

388



Page 18 of 26

389

Figure Legends:

390

Figure 1. Boxplots showing the median isolation interference of all identified PSMs precursors

391

across a range of abundances in all initial LC configurations tested. A distribution of the total

392

number of PSMs precursors per abundance ranges is also shown in blue bars. All spectral data

393

were searched with Amanda-Percolator.

394

Figure 2. Results from the LC-MS/MS analyses of 2µg of HeLa digest using different LC

395

configurations and MS/MS spectra search algorithms. (A-D) Bar charts depicting the average

396

numbers of identifications obtained from the spectral data collected for each LC setup tested (x-

397

axis) and searched with Amanda-Percolator or the CharmeRT workflow. Percentage increase or

398

decrease of the CharmeRT results compared to the ones obtained with Amanda-Percolator are

399

shown above or close to each pair of bars per LC setup. Error bars are the standard error of the

400

mean (n= 3 technical replicates).

401

Figure 3. Search depth achieved by CharmeRT in the spectral data collected from (A) 2D-HPLC

402

and (B) 1D-UHPLC 500 mins runs at different isolation windows. Histograms of the number of

403

PSMs per isolation interference range and identified in the first search of CharmeRT are shown.

404

The colored stacked bars below each histogram represent the percentage (above each stacked bar)

405

of PSMs identified in the first search of CharmeRT from which no additional PSMs derived from

406

second search were identified (green color, Depth 1); from which just one additional PSM was

407

identified (orange color, Depth 2, i); from which two additional PSMs were identified (green color,

408

Depth 2, ii); from which three additional PSMs were identified (pink color, Depth 2, iii).

409

Figure 4. Results from the spectral data collected with a range of isolation windows from the best

410

LC configurations. (A) Boxplots showing the median isolation interference of all identified PSMs

411

precursors from the spectral data of the optimized LC gradients collected under a range of isolation 17 ACS Paragon Plus Environment

Page 19 of 26 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


412

windows. (B) Bar charts depicting the average numbers of peptide sequences and protein groups

413

obtained from the spectral data collected for each LC setup (x-axis) under a range of isolation

414

windows. Percentage increase or decrease afforded from the CharmeRT results compared to those

415

from Amanda-Percolator are shown above or close to each pair of bars per setup and isolation

416

window. Error bars are the standard error of the mean (n= 3 technical replicates). The average

417

numbers of PSMs and peptides analytes are presented in Figure S-8.

418



Page 20 of 26

419

References

420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465

1. Michalski, A.; Cox, J.; Mann, M., More than 100,000 Detectable Peptide Species Elute in Single Shotgun Proteomics Runs but the Majority is Inaccessible to Data-Dependent LC-MS/MS. Journal of proteome research 2011, 10 (4), 1785-1793. 2. Houel, S.; Abernathy, R.; Renganathan, K.; Meyer-Arendt, K.; Ahn, N. G.; Old, W. M., Quantifying the Impact of Chimera MS/MS Spectra on Peptide Identification in Large-Scale Proteomics Studies. Journal of proteome research 2010, 9 (8), 4152-4160. 3. Dorfer, V.; Maltsev, S.; Winkler, S.; Mechtler, K., CharmeRT: Boosting Peptide Identifications by Chimeric Spectra Identification and Retention Time Prediction. Journal of proteome research 2018, 17 (8), 2581-2589. 4. Karp, N. A.; Huber, W.; Sadowski, P. G.; Charles, P. D.; Hester, S. V.; Lilley, K. S., Addressing Accuracy and Precision Issues in iTRAQ Quantitation. Molecular & Cellular Proteomics 2010, 9 (9), 18851897. 5. Savitski, M. M.; Mathieson, T.; Zinn, N.; Sweetman, G.; Doce, C.; Becher, I.; Pachl, F.; Kuster, B.; Bantscheff, M., Measuring and Managing Ratio Compression for Accurate iTRAQ/TMT Quantification. Journal of proteome research 2013, 12 (8), 3586-3598. 6. Evans, C.; Noirel, J.; Ow, S. Y.; Salim, M.; Pereira-Medrano, A. G.; Couto, N.; Pandhal, J.; Smith, D.; Pham, T. K.; Karunakaran, E.; Zou, X.; Biggs, C. A.; Wright, P. C., An insight into iTRAQ: where do we stand now? Analytical and Bioanalytical Chemistry 2012, 404 (4), 1011-1027. 7. Ow, S. Y.; Salim, M.; Noirel, J.; Evans, C.; Wright, P. C., Minimising iTRAQ ratio compression through understanding LC-MS elution dependence and high-resolution HILIC fractionation. Proteomics 2011, 11 (11), 2341-2346. 8. Yang, Y.; Qiang, X.; Owsiany, K.; Zhang, S.; Thannhauser, T. W.; Li, L., Evaluation of different multidimensional LC–MS/MS pipelines for isobaric tags for relative and absolute quantitation (iTRAQ)based proteomic analysis of potato tubers in response to cold storage. Journal of proteome research 2011, 10 (10), 4647-4660. 9. Toll, H.; Oberacher, H.; Swart, R.; Huber, C. G., Separation, detection, and identification of peptides by ion-pair reversed-phase high-performance liquid chromatography-electrospray ionization mass spectrometry at high and low pH. J Chromatogr A 2005, 1079 (1-2), 274-86. 10. Manadas, B.; English, J. A.; Wynne, K. J.; Cotter, D. R.; Dunn, M. J., Comparative analysis of OFFGel, strong cation exchange with pH gradient, and RP at high pH for first-dimensional separation of peptides from a membrane-enriched protein fraction. Proteomics 2009, 9 (22), 5194-8. 11. Gilar, M.; Olivova, P.; Daly, A. E.; Gebler, J. C., Orthogonality of separation in two-dimensional liquid chromatography. Anal Chem 2005, 77 (19), 6426-34. 12. Pirmoradian, M.; Budamgunta, H.; Chingin, K.; Zhang, B.; Astorga-Wells, J.; Zubarev, R. A., Rapid and Deep Human Proteome Analysis by Single-dimension Shotgun Proteomics. Molecular & Cellular Proteomics 2013, 12 (11), 3330-3338. 13. Lesur, A.; Domon, B., Advances in high-resolution accurate mass spectrometry application to targeted proteomics. Proteomics 2015, 15 (5-6), 880-890. 14. Shishkova, E.; Hebert, A. S.; Coon, J. J., Now, More Than Ever, Proteomics Needs Better Chromatography. Cell Syst 2016, 3 (4), 321-324. 15. Cristobal, A.; Hennrich, M. L.; Giansanti, P.; Goerdayal, S. S.; Heck, A. J.; Mohammed, S., Inhouse construction of a UHPLC system enabling the identification of over 4000 protein groups in a single analysis. Analyst 2012, 137 (15), 3541-8. 16. Thakur, S. S.; Geiger, T.; Chatterjee, B.; Bandilla, P.; Frohlich, F.; Cox, J.; Mann, M., Deep and highly sensitive proteome coverage by LC-MS/MS without prefractionation. Mol Cell Proteomics 2011, 10 (8), M110 003699. 19 ACS Paragon Plus Environment

Page 21 of 26 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

466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505


17. Wang, J.; Bourne, P. E.; Bandeira, N., Peptide identification by database search of mixture tandem mass spectra. Mol Cell Proteomics 2011, 10 (12), M111 010017. 18. Cox, J.; Neuhauser, N.; Michalski, A.; Scheltema, R. A.; Olsen, J. V.; Mann, M., Andromeda: a peptide search engine integrated into the MaxQuant environment. Journal of proteome research 2011, 10 (4), 1794-805. 19. Zhang, N.; Li, X. J.; Ye, M.; Pan, S.; Schwikowski, B.; Aebersold, R., ProbIDtree: an automated software program capable of identifying multiple peptides from a single collision-induced dissociation spectrum collected by a tandem mass spectrometer. Proteomics 2005, 5 (16), 4096-106. 20. Shteynberg, D.; Mendoza, L.; Hoopmann, M. R.; Sun, Z.; Schmidt, F.; Deutsch, E. W.; Moritz, R. L., reSpect: software for identification of high and low abundance ion species in chimeric tandem mass spectra. J Am Soc Mass Spectrom 2015, 26 (11), 1837-47. 21. Niu, M.; Mao, X.; Ying, W.; Qin, W.; Zhang, Y.; Qian, X., Determination of monoisotopic masses of chimera spectra from high-resolution mass spectrometric data by use of isotopic peak intensity ratio modeling. Rapid Commun Mass Spectrom 2012, 26 (16), 1875-86. 22. Scheltema, R. A.; Hauschild, J. P.; Lange, O.; Hornburg, D.; Denisov, E.; Damoc, E.; Kuehn, A.; Makarov, A.; Mann, M., The Q Exactive HF, a Benchtop mass spectrometer with a pre-filter, highperformance quadrupole and an ultra-high-field Orbitrap analyzer. Mol Cell Proteomics 2014, 13 (12), 3698-708. 23. Clarkson, S. M.; Giannone, R. J.; Kridelbaugh, D. M.; Elkins, J. G.; Guss, A. M.; Michener, J. K., Construction and Optimization of a Heterologous Pathway for Protocatechuate Catabolism in Escherichia coli Enables Bioconversion of Model Aromatic Compounds. Appl Environ Microbiol 2017, 83 (18). 24. Kenar, E.; Franken, H.; Forcisi, S.; Wörmann, K.; Häring, H.-U.; Lehmann, R.; Schmitt-Kopplin, P.; Zell, A.; Kohlbacher, O., Automated label-free quantification of metabolites from liquid chromatography–mass spectrometry data. Molecular & cellular proteomics 2014, 13 (1), 348-359. 25. Rost, H. L.; Sachsenberg, T.; Aiche, S.; Bielow, C.; Weisser, H.; Aicheler, F.; Andreotti, S.; Ehrlich, H. C.; Gutenbrunner, P.; Kenar, E.; Liang, X.; Nahnsen, S.; Nilse, L.; Pfeuffer, J.; Rosenberger, G.; Rurik, M.; Schmitt, U.; Veit, J.; Walzer, M.; Wojnar, D.; Wolski, W. E.; Schilling, O.; Choudhary, J. S.; Malmstrom, L.; Aebersold, R.; Reinert, K.; Kohlbacher, O., OpenMS: a flexible open-source software platform for mass spectrometry data analysis. Nat Methods 2016, 13 (9), 741-8. 26. Minakuchi, H.; Nakanishi, K.; Soga, N.; Ishizuka, N.; Tanaka, N., Octadecylsilylated porous silica rods as separation media for reversed-phase liquid chromatography. Analytical chemistry 1996, 68 (19), 3498-3501. 27. Liu, H.; Finch, J. W.; Lavallee, M. J.; Collamati, R. A.; Benevides, C. C.; Gebler, J. C., Effects of column length, particle size, gradient length and flow rate on peak capacity of nano-scale liquid chromatography for peptide separations. J Chromatogr A 2007, 1147 (1), 30-6. 28. Sandra, K.; Moshir, M.; D’hondt, F.; Verleysen, K.; Kas, K.; Sandra, P., Highly efficient peptide separations in proteomics: Part 1. Unidimensional high performance liquid chromatography. Journal of Chromatography B 2008, 866 (1-2), 48-63. 29. Hebert, A. S.; Richards, A. L.; Bailey, D. J.; Ulbrich, A.; Coughlin, E. E.; Westphall, M. S.; Coon, J. J., The one hour yeast proteome. Mol Cell Proteomics 2014, 13 (1), 339-47.

506 507



508

Page 22 of 26

Figure 1:

509 510


Page 23 of 26 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

511


Figure 2:

512



513

Page 24 of 26

Figure 3:

514

515 516


Page 25 of 26 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

517


Figure 4:

518 519



520

Page 26 of 26

TOC IMAGE:

521 522


Exploiting the dynamic relationship between ... - ACS Publications

Recommend Documents