Increasing the speed of protein digest analysis using nano- and micro- LC-MS strategies

(1)

Bachelor Thesis Chemistry

Increasing the speed of protein digest analysis using nano-

and micro- LC-MS strategies

by

Dewi Mooij

22 March 2018

Studentnumber

10752978

Research Institute

Supervisor

Van ’t Hoff Institute for Molecular Sciences

Prof. Dr. Garry L. Corthals

Research Group

Daily supervisor

(2)

Populair wetenschappelijke samenvatting

Figure 1: Experimental workflow1_.

Eiwitten worden gesynthetiseerd door het menselijk lichaam en zijn verantwoordelijk voor het functioneren van ons lichaam. Wanneer het lichaam niet goed functioneert en iemand ziek is, is dit te herleiden tot een defect of een tekort aan een bepaald eiwit. Al bijna 20 jaar kunnen de eiwitten die aanwezig zijn in een bepaald monster in kaart worden gebracht met behulp van de

analysetechniek nano liquid chromatography-mass spectrometry (nano-LC-MS). In de veelgebruikte strategie “bottom-up proteomics” worden de eiwitten in het monster eerst worden ontvouwen en afgebroken tot peptiden en vervolgens geanalyseerd met de nano-LC-MS. In de nano-LC-MS worden de peptiden van elkaar gescheiden op een scheidingskolom (LC) en daarna gefragmenteerd (MS). Er wordt een massaspectrum verkregen waarop de massa’s van de gedetecteerde fragmenten worden weergegeven. Aan de hand van de massa van een fragment kan de structuur van het fragment worden bepaald. De fragmenten kunnen alleen van een bepaald eiwit afkomstig zijn en door de gevonden fragmenten te vergelijken met databases kunnen de eiwitten uit een monster worden geïdentificeerd. Om eiwitten te kunnen kwantificeren worden ze gelabeld. Door verschillende monsters met elkaar te vergelijken is relatieve kwantificatie van eiwitten mogelijk. Problemen met de huidige methoden zijn dat het analyseren van een monster lang duurt (2 uur per monster) en dat de labels die gebruikt worden voor het kwantificeren erg duur zijn. Er zijn kortere analysemethoden nodig die eiwitten correct kunnen kwantificeren zodat er bij ziekenhuispatiënten sneller een diagnose kan worden gesteld en een behandeling kan worden gestart. In dit onderzoek worden nieuwe, sneller analysemethoden ontwikkeld en geëvalueerd. Om de analyse te versnellen wordt de flowrate, de snelheid waarmee het monster door de kolom gaat, 10 keer versneld van 300 nl/min naar 3 μl/min. Deze aanpassing naar de micro-flowrate zou het idealiter mogelijk moeten maken om

(3)

een monster in 30 minuten te kunnen analyseren. Ook worden verschillende

scheidingskolomlengten en kolomdiameters getest, omdat er grotere kolommen nodig zijn om de hogere flowrate te kunnen verwerken. Voor het kwantificeren van de eiwitten wordt gebruik gemaakt van de SWATH-MS-methode, met deze methode kunnen eiwitten worden gekwantificeerd zonder label. De volgorde waarin de eiwitanalyse wordt uitgevoerd wordt weergegeven in figuur 1. De resultaten van dit onderzoek tonen aan dat het mogelijk is om eiwitten te kwantificeren door gebruik te maken van de snellere micro-flow methode. Echter, doordat de flowrate 10 keer sneller is wordt er wel sensitiviteit verloren en kunnen er minder eiwitten worden gekwantificeerd dan met de huidige methoden. Verder onderzoek is noodzakelijk om de nieuwe

micro-LC-SWATH-MS-methode te optimaliseren zodat ziekenhuispatiënten uiteindelijk sneller en beter behandeld kunnen worden.

(4)

Abstract

Current nano-LC-MS methods for the quantitative analysis of proteins in clinical samples have enabled us to study and monitor diseases at the protein level. By quantitatively studying proteins, comparisons between sick and healthy patients can be made and based on these comparisons suitable treatments can be developed. However, the current analysis methods take too much time. Moreover, the labelling methods that are often used for protein quantitation, such as SILAC and iTRAQ are challenging and time-consuming. Recently, new data-independent acquisition strategies like SWATH-MS have been developed that enable label-free quantification. Shorter analysis methods that accurately quantify proteins in complex clinical samples are required for faster diagnosis and treatment of patients. For this purpose, in this work novel LC-SWATH-MS strategies are evaluated in terms of analysis time and proteome coverage. Yeast digest was studied as a model for method development. To speed up the peptide separation, different gradient lengths as well as nano- and micro flowrates were tested. Results of the yeast digest analysis showed that when a 10 cm long column is used in nano-LC-MS, halving the gradient length has little influence on the number of proteins quantified, indicating that the analysis time can be shortened without losing proteome coverage. After transitioning to micro-LC-MS, it was found that when using the same gradient and sample amount, 74% of the proteins identified in nano- could also be identified in micro-LC-MS. Half of the yeast proteins quantified in nano-LC-SWATH-MS could be quantified in micro-LC-SWATH-MS as well. For proof of concept, fresh frozen and FFPE kidney tissue samples were analysed in nano- and micro-LC-MS. For the fresh frozen tissue half of the proteins identified in nano- could also be identified in micro-LC-MS. For the FFPE tissue this percentage was found to be 20%. The findings of this work show a decrease in number of identified and accurately quantified proteins when transitioning from nano- to micro-flowrates. Further studies are necessary to improve methods for fast protein quantification using micro-LC-SWATH-MS.

(5)

1. Introduction

1.1 Proteomic approaches

The term proteomics was first used in 1995 and defined as the large-scale characterisation of the protein complement of a cell line, tissue or organism with the aim of understanding proteins and their role in a larger biological system2,3_{. Before proteomics emerged, genes were studied. However, a lot}

of information is missed when studying genes alone. Proteins, the products of genes, are of vital importance for many processes in cells like cell structure, communication between cells, energy production, cell division and may more. Only by studying proteins can one begin to fully comprehend cell processes in terms of protein modifications, protein interactions and identify potential targets for drugs in case of malfunctions3_{. Liquid chromatography coupled to tandem mass spectrometry is the}

most powerful approach for the identification and quantitation of proteins in complex biological samples4–7_{. Two main LC-MS based approaches are being used: bottom-up and top-down proteomics.}

In bottom up proteomics, the proteins are digested into peptides prior to LC-MS analysis. The protease trypsin is most often used for this purpose. Trypsin is a serine protease that cleaves the bonds at the carboxyl side of arginine and lysine. Trypsin cleaved peptides thus should have an arginine or lysine at the C-terminus, and this information can be used to identify peptides8_{. In top down proteomics, on the}

other hand, intact proteins are analysed. Top down proteomics is the preferred method when protein modifications are studied as intact proteins are analysed using this approach. However, this presents the disadvantage that intact proteins are more difficult to ionize and fragment compared to peptides8_.

Although both approaches have their advantages and disadvantages, bottom up proteomics is the more conventional approach.

1.2 Protein separation (chromatography-based methods)

While the advancements in the mass spectrometry technology have significantly boosted the speed and coverage of proteome analysis, little improvements have been made in the chromatographic peptide separation methods. Since 2004, the standard procedures involve the use of ultra-high pressure pumps to facilitate reversed-phase chromatography at nano-flowrates,9_{with fused silica}

capillary microcolumns packed with C18 microparticles10_{. Currently, these procedures are still being}

employed11–14_{. In this work, new methods are evaluated to test that maintain a high protein coverage,}

while decreasing the analysis time.

The obvious starting point when wanting to decrease the analysis time is switching to a higher flowrate. Most LC-MS/MS instruments operate at a nano-flowrate when peptide research is conducted. Typical nano-flowrates are in the order of 200-300 nl/min10_{. Under there flowrates, high}

(7)

7 sensitivity is reached resulting in increased protein identifications.10_{Additionally, the efficiency of the}

interface to the MS is improved, which contributes to increased sensitivity. At lower flowrates, the electrospray ionization is more efficient because the charged droplets are smaller. Smaller droplets means more efficient solvent evaporation and a higher number of charge per analyte, leading to increased sensitivity15_{. Warm, chemically inert nebulizer gas needs to be used to evaporate part of the}

droplets, making them smaller and hence the electrospray ionization more efficient. An additional benefit of operating the LC system at nano-flowrates is the reduction of solvent consumption16_{. At}

microflow, the droplets are larger, decreasing the ionisation efficiency. Research has shown that at micro-flowrates (1-10 μl/min) 85% of the proteins could be quantified as compared to nano-flowrates when injecting the same sample amount17_{. However, the gains in sample throughput are up to 400%}17_.

It appears a trade-off needs to be made regarding flowrate and sensitivity when using the same sample size. Loading more sample at micro-flow than at nano-flow does set off the sensitivity difference, making micro-flow systems more favourable when sample size can be increased17_{. Generally, around}

4 times the amount of sample used at nano-flowrates is required at micro-flowrates to give a similar sensitivity and number of identifications18_{. Besides, enhanced analysis throughput, operating at}

micro-flow rates also has other advantages. Also notable is the improvement in ease of operation as for instance, the leaks are easier to detect. Additionally, at micro-flowrates the robustness of the workflow is increased17,19_.

To improve and speed up the chromatographic separation multiple parameters can be altered. Factors that were examined in this research, besides testing nano- and microflows, include the length and internal diameter of the separation column, the particle size of the C18 stationary phase and the gradient length. In nano-LC-MS proteomics, capillary columns with an internal diameter of 75 μm are used most frequently20_{. Larger column internal diameters would in theory speed up the separation}

process, but cause a loss of sensitivity too21,22_{. As the flowrate in micro-LC-MS increases tenfold, wider}

columns are required to support this. When developing new methods using different flow rates the following formula (1) is used to calculate which target flow rate matches with which column inner diameter, to keep the speed at which the mobile phase travels through the column (linear velocity) constant23_.

𝑇𝑎𝑟𝑔𝑒𝑡 𝑓𝑙𝑜𝑤 𝑟𝑎𝑡𝑒 = 𝑂𝑟𝑖𝑔𝑖𝑛𝑎𝑙 𝑓𝑙𝑜𝑤 𝑟𝑎𝑡𝑒 𝑥 𝑑𝑡𝑎𝑟𝑔𝑒𝑡

2

𝑑_{𝑜𝑟𝑖𝑔𝑖𝑛𝑎𝑙}2 (1)

Therefore, if the target flow rate in micro-LC-MS is 3 μl/min, ten times higher than the original flowrate of 300 nl/min in nano-LC-MS using a 75 μm inner diameter column, the target diameter would be 237 μm. Since this exact diameter is not available, a 200 μm inner diameter column will be tested

(8)

8 at a flowrate of 3 μl/min. Commonly used in micro-LC-MS are columns with an inner diameter of 300 μm at flowrates of 5 μl/min, which is in compliance with formula (1)18,24,25_.

Previous research into nano-LC-MS has pointed out that chromatographic peak capacity increases when longer columns and gradients are used, resulting in more peptide identifications9,26,27_.

Longer gradients allow more time for the separation, enabling a better separation and higher peak capacity, defined as the maximum number of peaks that can be separated over a chromatographic column. When using the same MS parameters, longer gradients also result in more spectra being recorded, increasing the chance of identifying more unique peptides27_{. However, longer columns and}

gradients means a longer sample analysis time which is undesirable for this research. Another way reported to increase peak capacity is by reducing the particle size of the C18 stationary phase9,26,28_{. As}

mentioned before, sensitivity is lost when increasing the flowrate from nano-LC-MS to micro-LC-MS. To avoid losing sensitivity due to a higher flowrate and wider column, and improve the quality of the separation in micro-LC-MS, a column packed with 3 μm particle size C18 was tested rather than the conventional 5 μm particle size C18 used in nano-LC-MS. Furthermore, a method used in proteomics with the aim of improving peptide separation quality is two-dimensional LC-MS. As the maximum peak capacity results from multiplying the peak capacities of the first and the second dimension, the 2D set-up enables higher peak capacities and provides the possibility to combine different types of stationary phases to achieve higher proteome coverage. A combination commonly used is a ion exchange column followed by a reverse-phase column to achieve orthogonality between the first and second dimension29,30_{. Although multi-dimensional separation techniques have promising perspectives for the}

analysis of complex proteomic samples, it is beyond the scope of this project and therefore will not be researched in this work.

1.3 Protein identification (mass spectrometry)

The two most common strategies in bottom up proteomics are shotgun proteomics and selected reaction monitoring (SRM)5–7_{. Proteomic strategies can be targeted or untargeted, also known as}

discovery proteomics. In targeted proteomics such as SRM only a preselected group of peptides of interest is analysed, making the analysis more sensitive and accurate than when all peptides in a sample are analysed such as in untargeted proteomics31_{. In shotgun proteomics, the mass}

spectrometer operates in the data dependent acquisition mode (DDA)6,7,32_{. In DDA, a full mass}

spectrum is generated displaying all peptide intensity signals and from this spectrum the most abundant precursor ions are then selected for fragmentation producing the tandem MS spectra32_{. The}

fragment ion spectra are matched to their corresponding peptide sequences by database searching. Shotgun proteomics is a great tool for discovering proteins in a sample, but has its limitations due to

(9)

9 low reproducibility and its inability to identify low-abundant proteins5_{. SRM, on the other hand can}

reproducibly identify and quantify peptides from complex samples5,6_{. In this strategy, the peptides of}

interest are selected prior to analysis and then searched for their presence in a sample. SRM proceeds by the acquisition across the LC retention time window of pre-defined pairs of precursor and product ion masses, called transitions. The transitions make up an assay for the detection of a peptide. Data analysis computes the probability that the transition is derived from the targeted peptide6_{. Transitions}

should be validated to prove that they belong to the targeted peptide before the peptide can be identified and quantified from the transitions33_{. Though SRM can accurately identify and quantify the}

proteins of interest, it is unfit for large scale proteomics as the number of transitions that can be measured per run is limited by the MS instrument33_.

To combine the advantages of shotgun and targeted proteomics resulting in the reproducible and accurate identification of large numbers of peptides, new methods have been developed. These methods operate the MS in the data independent mode (DIA). In DIA, all peptides within a predetermined mass-to-charge window and retention-time range are fragmented in tandem MS5–7,32_.

Whereas in DDA only the most abundant peptides are fragmented in tandem MS in narrow isolation windows (2 m/z), hereby restricting the dynamic range to the peptides that ionize best34_{. Since in DIA}

the ions are not selected for tandem MS based on precursor scans, the dynamic range is no longer restricted. In DIA, all ions in the preselected m/z range can be isolated and fragmented, or the m/z range can be divided into smaller m/z isolation windows. Peptides within these windows of about 6-90 m/z are isolated and fragmented independently and consecutively34_.

(10)

10

Figure 2: Tandem MS analysis in DDA and DIA. DDA acquires MS/MS scans with narrow isolation windows of peptides selected in a precursor scan. DIA acquires MS/MS scans with wide isolation windows of all peptides within the m/z range34_.

The acquired DIA data is continuous both in time and fragment-ion intensity, which is beneficial as it increases the dimensionality but it also complicates the data analysis7_{. The DIA data used to be}

analysed by methods originally designed for the analysis of DDA data, though suffering from the high complexity of the spectra. To overcome this, a novel DIA method called sequential window acquisition of all theoretical mass spectra (SWATH)-MS has been developed. In SWATH-MS data independent acquisition is combined with targeted data extraction. SWATH-MS data is obtained on a quadrupole time-of-flight MS instrument by cycling repeatedly through predefined sequential isolation windows over the whole chromatographic elution range, called swaths35_{. The swath isolation windows are}

typically 25 Da wide, meaning 32 swath windows are typically needed to cover the 400-1200 m/z range. The precursor ions from a selected window accumulate and are collectively fragmented. A tandem MS spectrum is recorded and the window shifts to the next where the process repeats itself until all

(11)

11 windows have been covered34_{. The complex fragment ion spectra obtained by SWATH-MS are analysed}

using a targeted data extraction strategy. The targeted data extraction method uses information contained in spectral libraries. For the generation of spectral libraries DDA measurements are required5_{. The fragment ion maps obtained, are mined for signals corresponding to known}

coordinates of a targeted peptide, and in this manner the peptides can be identified in the map6_{. For}

automated targeted data extraction analysis software’s like OpenSWATH and Spectronaut have been developed7_{. These programmes generate the most intense transitions of a targeted peptide from all}

the corresponding tandem MS spectra, resulting in data similar to SRM data7_{. Because the transitions}

are specified after the data is acquired, the data can easily and flexibly be re-examined36_.

Figure 3: SWATH MS sequential acquisition of precursor isolation windows. In this figure 32 scans of 25 Da are used for the recording of fragment ion spectra. Once all the fragment ion spectra within an isolation window are aligned a MS2/SWATH map is obtained (right side of figure)6_.

1.4 Protein quantitation

In proteomics, relative quantification is achieved by comparing the differences between two or more physiological states of the system. Based on these comparisons, distinctions between sick and healthy patients can be made as well as decisions regarding the treatment of patients. Several methods are used for quantification, distinguishable are labelling and label-free methods. Methods using labels on peptides for quantification are particularly suitable for assessing small changes in protein levels or post-translational modifications due to their high accuracy. Labelling methods use stable isotopes to mass tag peptides. Peptides are quantified by tagging a known amount of peptides with a heavy isotope and comparing this amount to an unknown number of peptides tagged with a lighter isotope8,37_{. Labelling can be achieved by means of a chemical reaction or metabolic. In chemical}

labelling, the label is added to a reactive group on the peptide, an example of this method is isotope-coded affinity tag (ICAT). Metabolic labelling is achieved using a stable isotope-enriched medium, through which the isotopes are incorporated into the proteome during protein synthesis. Stable

(12)

12 isotope labelling with amino acids in cell culture (SILAC) is an example of a metabolic labelling method8_.

Disadvantages of using labels for peptide quantification include the fact that full labelling is hard to achieve and that peptides incorporate labels at different rates, complicating data analysis38_.

For label-free quantification two main strategies are used. The first is spectral counting, in this strategy the number of fragment spectra identifying peptides of a given protein are counted and compared between samples for relative quantification38_{. The second method measures and compares}

the MS signal intensity of peptide precursor ions belonging to a protein to quantify38_{. Influences from}

other signals and spectral background noise make label-free quantification the least accurate quantification technique. Quantification techniques often use data dependent acquisition. As mentioned above, in DDA a survey scan is made to determine which precursors to select for product scanning. The main disadvantage is that low intensity ions are often missed, failing the quantification of low-abundance proteins38_.

In the DIA mode, by using the SWATH-MS strategy all the precursor ions are fragmented, enabling even the quantification of low-abundance proteins in a sample. Peptide quantification in SWATH MS is linked to identification and proceeds by integration of the peak area of the fragment ion signals across the chromatographic elution of a validated peak group that identifies the peptide.6_However,

as SWATH-MS needs a spectral library for identification, the content of the spectral library strongly influences the number of peptides identified and quantified5,34_{. Spectral libraries are generated from}

fragment ion spectra that have been assigned to certain peptides using database searching. For the creation of a spectral library, DDA runs are required39_{. Though public libraries of multiple organisms}

are available, it is advised to generate sample specific libraries. To make sure that the library matches the sample DIA as good as possible, the DDA data must be acquired on the same instrument and column as the DIA data40_{. Many DDA runs of each sample are required to generate a high quality}

spectral library, making it a time-consuming process. Biognosys has studied the ratio between time investment and number of protein and precursor identifications and concluded that 6 DDA runs is the optimum number. Furthermore, it is important that the sample is both in DDA and DIA spiked with reference peptides to enable retention time alignment between the runs and hence, accurate quantitation40_{. iRT peptides are used to accurately predict the elution time of peptides. The iRT peptide}

mix that can be purchased from Biognosys contains 11 synthetic peptides, the retention times of the peptides present in a sample are defined relative to those 11 peptides. A disadvantage of this method is the relatively high costs of the iRT peptides, and since they are being added to the analysed samples it makes the method not entirely label-free.

(13)

13

1.5 Aim of the research

Using current day nano-LC-MS methods that take 2 hours to analyse one sample, a maximum of nine samples could be analysed per day, with 1-hour calibration methods every six hours and 2-hour quality control runs every 12 hours included. In micro-LC-MS a sample could be analysed in, for example, 30 minutes41_{, and this would mean that the sample throughput could be increased four times and even}

more when shorter calibration and quality control methods are developed. Shortening the sample analysis time, while maintaining a high proteome coverage and accurate protein quantitation is especially important in clinical settings. Faster protein analysis methods would enhance the sample throughput, leading to faster diagnosis and treatment for patients. Also, faster biomarker discovery will be of substantial influence in drug development and the understanding of diseases. To speed up protein analysis, improvements are necessary in the analytical workflow as well as the data analysis strategies. When trying to reduce analysis time, it is meaningful to keep in mind that a full blown proteomic analysis is often not needed. The focus should be on the development of synoptic assays that show only the information required to find the right cure for a patient. To achieve this, targeted data analysis strategies like SWATH-MS could be very useful.

The aim of this research is to evaluate new data independent LC-MS strategies, SWATH-MS, to increase the speed and coverage of protein digest analysis. For this purpose, various methods and instrumental set-ups were tested and the results are discussed herein. Reversed-phase columns of different inner diameters and length packed with C18 stationary phases were tested, as well as different gradient lengths. Nano-LC-MS and micro-LC-MS were compared in terms of number of proteins identified and quantified. Samples analysed in this research include yeast as a model for method development, and kidney tissues as these are exemplary of clinical samples that would ideally be analysed using faster methods.

(14)

14

2. Materials and method

2.1 Chemicals

Acetonitrile and LC/MS water of ULC/MS grade were purchased from Biosolve Chimie, Dieuze, France. Urea, dithiothreitol, iodoacetamide, ammonium acetate and trifluoroacetic acid were purchased from Sigma-Aldrich. Ammonium bicarbonate was purchased from Fluka analytical. Modified trypsin was purchased from Worthington Biochem. Extraction disk cartridges were purchased from Empore 3M. Fused silica capillaries were obtained from Polymicro technologies. Glass microfiber filters for the column frits were purchased from Whatman GE healthcare life sciences, Kasil 1 and formamide from Next Advance Frit Kit. The magic C18 stationary phase was purchased from Bulk Packing Materials. Ethanol absolute for analysis was purchased from Emsure.

2.2 Sample preparation

2.2.1 Yeast

Yeast lysate was precipitated in 5 volumes of cold acetone for 2 hours for sample clean-up. After centrifugation the supernatant was collected, and the sample dried at room temperature. For the digestion, the final volume in the sample vial was set to be 1 ml. The sample was reconstituted in 100 μl of 6 M urea. 1 mM of dithiothreitol was added for protein reduction, and the sample was reduced for 1 hour at 37 °C. For alkylation 4 mM of iodoactemide was added and the sample was stored in the dark for 1 hour at room temperature. 4 mM of dithiothreitol was added to consume the leftover alkylating agent. The urea concentration was diluted to 600 mM using ammonium bicarbonate (25 mM). Trypsin was added in 1:30 trypsin to protein ratio for digestion, and the sample was digested overnight at 37 °C. The next day, the sample was desalted by solid phase extraction using C18 Ziptips, freeze-dried and stored at -80 °C. For analysis the sample was dissolved in 50 μl of loading buffer (0.1% TFA, 2% ACN, 97.9% LC/MC water) and 5 μl of iRT peptides solution, to achieve of final concentration of 0.2 μg/μl of protein in the vial. 5 μl of sample, containing 1 μg of protein was injected for analysis.

2.2.2 Kidney tissue

Fresh frozen and formaline-fixed, paraffin-embedded (FFPE) kidney tissues were digested for peptide analysis by Marvin Dittrich. From the fresh frozen kidney tissue a sample area of 1 mm2_{with a thickness}

of 15 μm was analysed in 5 μl. Of the formaline-fixed, paraffin-embedded kidney tissues 1 mm2_of

(15)

15

2.3 Instrumentation

2.3.1 Chromatography

The peptide separation was performed using an Ekspert nano LC 452 system (Eksigent, AB Sciex). Under nano-flow rate conditions, an in-house packed trap column (200 μm I.D., 2 cm length) packed with 5 μm particle size Magic C18 was used. For the yeast samples two different 75 μm I.D. capillary columns (10 cm length and 5 cm length) packed in-house with 5 μm particle size Magic C18 stationary phase were tested. The kidney tissue samples were analysed using a 75 μm I.D. capillary column with a length of 10 cm packed in-house with 5 μm particle size Magic C18 stationary phase. All runs in the nanoflow set-up were ran at a flowrate of 300 nl. For the yeast samples linear gradients with mobile phase B (ACN + 0.1% FA) increasing from 5% to 40% in 15 min, 30 min and 60 min were tested on the 5 cm and 10 cm length columns. Additionally, a 90 min gradient was also tested on the 10 cm length column. The kidney tissues samples were measured using a linear gradient 11.25, 22.5 and 45 min during which mobile phase B increases from 5% to 40%.

For the experiments performed at micro-flow rate (yeast and kidney tissue), an in-house packed LC column (200 μm I.D., 15 cm length) packed with 3 μm particle size Magic C18 was used for peptide separation. Samples were injected directly onto the column. All runs in the microflow set-up were ran at a flowrate of 3 μl/min. For the yeast samples a linear gradient with mobile phase B (ACN + 0.1% FA) increasing from 10% to 40% in 60 min was tested. For the kidney samples linear gradients of 11.25 min, 22.50 min and 45 min were tested, each increasing mobile phase B from 5% to 40% during the gradient.

2.3.2 Mass spectrometry

All experiments were performed using a TripleTOF 5600+ system (AB SCIEX) mass spectrometer. For the nano-flow studies, a Nanospray III source (AB SCIEX) and a pulled quartz tip as the emitter were used for data acquisition. The MS source parameters are shown in table 1. MS parameters for the DDA runs of the yeast samples include a cycle time of 3301 ms in TOF MS and MS/MS. The accumulation time was set to 250 ms in TOF MS and to 100 ms in MS/MS. For protein quantification purposes 32 SWATH windows with a size of 25 m/z were generated from the 400-1250 m/z range. During the SWATH runs accumulation times of 50 ms in TOF MS and 100 ms in MS/MS were used. Gradients of 15, 30 and 60 minutes were tested using SWATH methods. MS parameters for the DDA runs of the kidney tissue samples were a cycle time of 3553 ms in both TOF MS and MS/MS, an accumulation time of 500 ms in TOF MS and an accumulation time of 100 ms in MS/MS.

In the micro flow LC-MS setup, the TripleTOF 5600+ system was fitted with a DuoSpray Ion Source with a 50 μm electrode (AB SCIEX). The MS source parameters are shown in table 1. For the

(16)

16 yeast DDA runs the cycle time in TOF MS and MS/MS was set to 1649 ms. In TOF MS the accumulation time was set to 100 ms and to 50 ms in MS/MS. For the SWATH runs for protein quantification in the yeast samples, the range 400-1250 m/z was covered by 48 SWATH windows with a size of 18 m/z. In both TOF MS and MS/MS the period cycle time was set to 2499 ms, and the accumulation time was set to be 50 ms. MS parameters for the DDA runs of the kidney tissue samples were a cycle time of 3553 ms in both TOF MS and MS/MS, an accumulation time of 500 ms in TOF MS and an accumulation time of 100 ms in MS/MS.

Table 1: MS Source parameters in nano- and micro-LC-MS.

Parameter nano-LC-MS micro-LC-MS

Gas 1 (psi) 15 10

Gas 2 (psi) 0 15

Nebulizer gas (psi) 30 30 Ionspray voltage (V) 2200 5500 Temperature (°C) 150 100

2.4 Data analysis

The DDA runs were processed with ProteinPilot (AB SCIEX), the FDR for identification was set to 1%. The ProteinPilot output files were used to generate sample specific spectral libraries in Spectronaut 11 (Biognosys), the default settings for spectral library generation were used. The data-independent SWATH runs were processed in Spectronaut with the previously generated spectral library. The SWATH runs were processed with the default settings for protein quantitation. Figure 4 shows the data analysis workflow. The protein group report was exported from Spectronaut to reveal how many and which protein groups had been identified in the SWATH runs. Venn diagrams were constructed using InteractiVenn (BMC Bioinformatics).

(17)

17

Figure 4: Data analysis workflow. The sample is spiked with iRT peptides and analysed. Raw data from the DDA runs is processed with ProteinPilot to confirm the number of identified proteins in the runs. The ProteinPilot output excel file is loaded into Spectronaut 11 to generate the spectral library. The SWATH raw files are assigned to the spectral library and processed in Spectronaut 11 for the quantitation of identified proteins using the iRT peptides for calibration.

Sample spiked with iRT peptides TripleTOF 5600+ MS Eksigent LC nano/micro SWATH runs DDA runs ProteinPilot: identification Spectral library generation Quantification of identified proteins

(18)

18

3.

Results and discussion

3.1 Yeast analysis

3.1.1 Effect of column and gradient length on protein identification and quantitation in

nano-LC-MS

To be able to develop new methods to decrease the analysis time of proteomic samples while maintaining a high proteome coverage, current methods were evaluated first and adapted to see if they could be improved. Literature reports that the method of choice for yeast digest analysis by nano-LC-MS, uses a 15 cm long, 75 μm inner diameter column packed with C18 stationary phase with a particle size of 3 and 5 μm6,14,21,42,43_{. Gradients are usually between 90-180 minutes long and show a}

linear increase in solvent B (acetonitrile with approximately 0.1% formic acid) from 5-35%6,14,21_{. For}

shortening the analysis time, shorter columns and gradients were tested. To resemble current methods, a conventional inner diameter and C18 particle size of 75 μm and 5 μm respectively, were used. For spectral library generation, the yeast was run in DDA using the 10 cm column and gradients of 60 and 90 minutes. A total number of six DDA runs were used to generate the spectral library, as this is the minimum number of runs required to make a high quality sample specific spectral library according to literature40_{. For the SWATH runs 32 windows of 25 Da were generated. Six different}

conditions were evaluated through combination of the two different column lengths of 5 and 10 cm with the different gradient lengths of 15, 30 and 60 minutes. Every condition was run in triplicate to rule out any biases or errors.

Figure 6 is a heatmap that shows the differences in protein groups quantified using different column and gradient lengths. The results clearly show that significantly more protein groups can be quantified on the 10 cm column as compared to the 5 cm column. The part of the heatmap marked in red that displays the data obtained on the 5 cm column with the gradient increasing from left to right, is clearly emptier than the right part of the figure (marked in black) that shows data obtained on the 10 cm column. The right part of the heatmap is much denser populated, meaning that many more protein groups have been quantified in the runs using the 10 cm column than on the 5 cm column. Further analysis of the chromatographic data also shows that the quality of the separation is better when a longer column is used. The average full width at half maximum (FWHM) is 0.30 min for the 10 cm column and 0.80 min for the 5 cm column, hence the peaks obtained on the 10 cm column are narrower. Using a longer column also gives a higher average peak capacity of 250 for the 10 cm column, compared to 80 for the 5 cm column. So, using a longer column improves the quality of the separation resulting in the successful quantitation of more protein groups.

(19)

19 When further examining just the part of the data obtained on the 5 cm column (marked in red), it becomes apparent that the number of identification does as expected increase with longer gradients. The Venn diagram in figure 5A also confirms this for the number of protein groups quantified per condition, more protein groups have been quantified in the longer runs. With this result it is once again shown, that the number of protein groups that can be quantified in short length columns strongly depends on the length of the gradient and as such increases when longer gradients are used due to higher peak capacities.

However, this difference does not seem to be so strong when comparing the quantified protein groups for the 10 cm column. Although the population of the heatmap does look to be increasing with longer gradients just like on the 5 cm column, the difference between the gradients is not as significant as it is for the 5 cm column. The Venn diagram below in figure 5B confirms this observation as the number of quantified protein groups only increases very little when going from a 15 min to a 30 minute gradient, and not at all when comparing the 30 minute gradient with the 60 minute gradient. So as it seems from the obtained results, when a longer column is being used the influence of the gradient on the number of quantified protein groups becomes less strong. This is a promising outcome that could be benefitted from when transitioning to micro flow rates because by simply using a longer column, the analysis time can be decreased from 60 minutes to 15 minutes without losing much proteome coverage. Also, the small difference in quantified protein groups between the different gradients could be overcome by optimization of MS parameters and the SWATH-MS windows.

Figure 5: Venn diagram of quantified protein groups using (a) 5cm column and (b) 10 cm column with gradients of 15, 30 and 60 minutes. The number of quantified proteins clearly increases with longer gradients.

(20)

20

Figure 6: Heatmap showing the protein group identifications per condition. The data marked in red was ran on the 5 cm column, with the gradient length increasing from 15 to 30 to 60 minutes from left to right. The black marked data was obtained on a 10 cm column, here the gradient also increases in length from left to right.

(21)

21 The quality of the data was evaluated by the coefficient of variation for protein quantitation for each of the conditions. The protein coefficient of variation (CV) is calculated based on the summed intensities of the peptides identifying the proteins. The standard deviation of the peptide intensities is divided by the mean and this is reported as the percentage CV. Figure 7 below shows the coefficient of variation per condition. The first striking observation is the fact that the coefficient of variation is lower for the conditions on the 10 cm column than on the 5 cm column. Not only is the median CV lower for the 10 cm column, also the minimum and maximum show less variance as compared to the 5 cm column conditions. Thus, the longer column causes there to be less variance, making the experiment more precise and repeatable. Based on the lower coefficient of variations it can be stated that the results obtained on the 10 cm column are of higher quality than those obtained on the 5 cm column. This is another reason for continueing with the 10 cm column in micro-LC-MS, as it not only gives more quantified protein groups but also the quality of the results is better as the coefficient of variation is lower.

Figure 7: Coefficient of variation per condition. From left to right: 10cm 15min (9.1%), 10cm 30min (7%), 10cm 60min (7.2%), 5cm 15min (8.7%), 5cm 30min (18.4%), 5cm 60min (16%).

3.1.2 Protein quantitation using SWATH-MS under nano and micro flow rate conditions

The previous section has shown that the number of quantified protein groups in nano-LC-MS is the same for the 30- and the 60-minute gradient. As conventional gradients in nano-LC-MS are often even longer than 60 minutes, it was decided that to evaluate the quality of measurements at a ten times higher micro-flowrate, the same relatively long 60-minute linear gradient was used in micro. Also, the length of the column was kept at 10 cm as the previous experiment indicated that using this column length a higher proteome coverage and higher quality of results were achieved. To support the higher flowrate the 75 μm ID column was replaced by a column with an inner diameter of 200 μm, keeping

(22)

22 the linear velocity constant. As wider columns and higher flowrates cause a loss of sensitivity which results in fewer protein identifications, the 5 μm C18 stationary phase was changed to 3 μm. This smaller particle size should increase the peak capacity and hence the quality of the chromatographic separation9,26,28_{. By improving the chromatographic separation an attempt was made to decrease the}

effects of sensitivity loss at micro-flowrates. Figure 8 shows the number of identified proteins in the 60-minute gradient DDA runs in both nano- and micro-LS-MS with a false discovery rate of 1%.

Figure 8: Number of protein identifications in a nano- and micro-LC MS DDA run with a linear gradient of 60 minutes. The DDA data has been processed with ProteinPilot.

73% of the proteins identified in nano-LC-MS can also be identified in micro-LC-MS when the same amount of sample was injected. The lower number of identified proteins in micro-LC-MS can be explained by multiple causes. One of the reasons could be that the average of three nano DDA runs is compared to the result of one micro DDA run, as for time management reasons there was no opportunity to record more micro runs in DDA and one DDA runs should in theory be sufficient to generate the spectral library needed for the SWATH runs later. Furthermore, multiple parameters were altered at the same time in the development of the microflow method. The linear velocity is slightly different with a flowrate of 3 μl/min using a 200 μm inner diameter column then with a flowrate of 300 nl/min and a 75 μm inner diameter column. To keep the linear velocity constant in both methods, ideally the flowrate in micro should be 2.1 μl/min or the column diameter should be 237 μm. Another inconsistency in the methods is the C18 particle size, being 5 μm in nano- and 3 μm in micro-LC-MS, although this should not lower the number of identifications in micro-LC-MS because it should improve the quality of the separation.

574 783 0 100 200 300 400 500 600 700 800 900

MICRO 60 min NANO 60 min

n u m b er o f id en tifi cat ion s DDA run

(23)

23 For quantification of proteins in micro-LC-MS 48 SWATH windows of 20 Da were generated. The number of SWATH windows in micro-LC-MS was set to be higher than the number of windows used in nano-LC-MS (32 windows of 25 Da), again as an attempt to obtain as many identified and quantified proteins as possible. Literature reports that cycling through more SWATH windows, results in a higher number of protein identifications and a decrease in the coefficient of variation44,45_{. For accurate}

quantitation using SWATH MS, at least 8 data points per peak are required25_{. In micro-LC-MS, sharper}

peaks are obtained due to the higher flowrate and this sharpness of peaks is even further enhanced by using the smaller particle size C18. To ensure enough data points per peak and sufficient signal intensity, the SWATH MS method was adapted, and the accumulation time shortened to 50 ms in both TOF MS and MS/MS. Using these settings an average of 7 data points per peak were obtained in micro-LC-MS, compared to 10 data points per peak in nano-LC-MS. These numbers should be sufficient for accurate quantitation of proteins. Figure 9 shows the number of protein groups quantified in both micro-LC-MS and nano-LC-MS as well as the protein groups that were quantified in both methods.

Figure 9: Venn diagram of quantified protein groups in nano- and micro-LC-MS.

311 protein groups could be quantified in both nano- and micro-LC-MS, an additional 310 protein groups were uniquely quantified in nano, and 28 protein groups were uniquely quantified in micro-LC-MS. 54% of the total number of protein groups identified in nano, could be identified in micro-LC-micro-LC-MS. It appears that in terms of quantitation the difference between the two methods is greater than in terms of identification. However, this is not a comparison that can be made, as the data for identification and quantification have been processed with different processing software’s. The difference in number of quantified protein groups between nano- and micro-LC-MS could partially arise from the fewer data points per peaks in micro. It would be interesting to ensure the same number of data points per peak in both nano and micro and compare the number of quantified protein groups then. Using the smaller stationary phase particle size in micro-LC-MS resulted in a peak capacity of 370. In nano-LC-MS a peak capacity of 217 was achieved using the larger 5 μm C18 particles. The quality of

(24)

24 the separation is better in micro-LC-MS due to the higher peak capacity and lower full width at half maximum (FWHM). Using micro flow rates, the FWHM is 0.17 min whereas the FWHM is 0.30 min in nano-LC-MS. This shows that the chromatographic peaks are sharper in micro-LC-MS. However, while narrow peaks are wanted in the sharpness of the peaks also decreases the number of data points per peak needed for quantitation, resulting in undersampling. Undersampling occurs when more peptides elute than the mass spectrometer can process41_{. Although one of the reasons for using the}

SWATH-MS method was to avoid this undersampling problem by fragmenting all precursors simultaneously in isolation windows, the SWATH-MS method used in micro-LC-MS needs to be adapted further. Further improvements to the SWATH-MS methods should lead be done to achieve more data points per peak and hopefully a higher number of quantified protein groups. Even though, the quality of the separation is better in nano- than micro-LC-MS, there is a significant loss in signal intensity that contributes to the lower number of quantified protein groups using micro-LC-MS. Signal intensity using nano flow is around 10 x 106_{, whereas the signal intensity using micro flow is only 2.5 x 10}6_{. The signal intensity is}

reduced fourfold by increasing the flow rate 10 times. As indicated by the results, the quality of the separation obtained in micro-LC-MS is better than the one in nano-LC-MS, however sensitivity is lost. To overcome this problem and obtain more proteome coverage in micro-LC-MS, more sample could be loaded on the column to offset the reduced sensitivity. Lastly, in SWATH MS quantitation is related to identification, hence the quality of the spectral library plays a crucial role in the number of protein groups that can be identified. Possibly the quality of the library could still be improved upon by measuring more sample in DDA in both the nano- and the micro-setup.

3.2 Kidney tissue analysis

3.2.1 Comparison of protein identifications between nano- and micro LC-MS

Having seen that for the model organism yeast, 73% of the proteins identified in DDA using nano-LC-MS could also be identified in micro-LC-nano-LC-MS, kidney tissues were analysed next using both nano- and microflow. Two different types of tissue specimen were analysed, fresh-frozen and formaline-fixed paraffin-embedded (FFPE). FFPE specimen are regarded as difficult to analyse using LC-MS due the fixation with formalin, which reacts with the protein aminogroups causing cross-links to form46_{. In this}

work, the two different specimens are compared as well as the number of identifications for each specimen in nano- and micro-LC-MS. To be able to compare between nanoflow and microflow, 3 different gradient lengths were created for both flowrates, 11.25 minutes, 22.5 minutes and 45 minutes. All three gradients show a linear increase in mobile phase B from 5-40% during the gradient length. Figures 10-13 show bar graphs displaying the number of identifications and the number of spectra obtained for each condition for the two specimens. When comparing the figures, it becomes

(25)

25 apparent that for the fresh frozen tissues more identifications and spectra were obtained in every condition compared to the FFPE tissues. This result is in line with the general idea that FFPE tissues are difficult to analyse using LC-MS and show a relatively low number of identifications. As stated earlier, in FFPE tissues protein crosslinks can form due to reactions of formaldehyde with the aminogroups of the tissue proteins. These crosslinks decrease the solubility of the proteins, making it hard to extract proteins from FFPE samples46,47_{. As fewer proteins are extracted from the sample, the number of}

proteins identified after LC-MS analysis will logically also be low. Furthermore, the reaction with formaldehyde could also cause unknown peptide modifications, because these modifications are unknown database searches will not be able to identify these peptides. Lastly, the crosslinks complicate the structure of the proteins making it difficult for trypsin to access the cleavage sites47_.

This disruption of trypsin digestion influences the number of peptides, which in turn influences the number of proteins that will ultimately be identified. The obtained results once again prove the difficulty of analysing FFPE samples using LC-MS. Further research is required into sample preparation protocols for FFPE tissue analysis by LC-MS. However, as the aim of this project solely was to investigate the number of identifications for FFPE tissues in both nano- and micro-LC-MS, improvement of sample preparation protocols has not been attempted in this work.

Figure 10: Number of protein identifications obtained in nano- and micro-LC-MS for fresh frozen kidney tissue samples.

0 100 200 300 400 500 600 700 800 900 NANO 45 min NANO 22 min NANO 11 min MICRO 45 min MICRO 22 min MICRO 11 min N u m b er o f p ro tein id en tifi cat ion s DDA run

Fresh frozen: number of protein identifications in

nano- and micro-LC-MS

(26)

26

Figure 11: Number of spectra recorded in nano- and micro-LC-MS for fresh frozen kidney tissue samples.

Figure 12: Number of protein identifications obtained in nano- and micro-LC-MS for FFPE kidney tissue samples.

Figure 13: Number of spectra recorded in nano- and micro-LC-MS for FFPE kidney tissue samples.

0 10000 20000 30000 40000 50000 60000 NANO 45 min NANO 22 min NANO 11 min MICRO 45 min MICRO 22 min MICRO 11 min N u m b er o f spectra DDA run

Fresh frozen: number of spectra in nano- and

micro-LC-MS

0 50 100 150 200 250 300 350 400 NANO 45 min NANO 22 min NANO 11 min MICRO 45 min MICRO 22 min MICRO 11 min N u m b er o f p ro tein id en tifi cat ion s DDA run

FFPE: protein identifications in nano- and

micro-LC-MS

0 5000 10000 15000 20000 25000 NANO 45 min NANO 22 min NANO 11 min MICRO 45 min MICRO 22 min MICRO 11 min N u m b er o f spectra DDA run

FFPE: number of spectra in nano- and

micro-LC-MS

(27)

27 When examining figure 10 and 11 it becomes apparent that for fresh frozen tissue samples in both nano- and micro-LC-MS the number of identifications and spectra decreases with the gradient length. This was expected as the same MS parameters have been used for all the experiments and therefore less spectra were recorded for the shorter gradient lengths. A lower number of spectra results in fewer identifications as the chances of not recording a specific peptide fragment that belongs to a unique protein are higher. This effect is especially strong on low abundance proteins, these proteins will most likely be missed when gradients are shortened. When comparing the number of protein identifications between nano- and micro-LC-MS for fresh frozen tissue samples, around 50% of the proteins identified in nano-LC-MS can also be identified in micro-LC-MS. This percentage for tissue samples is lower than the value of 73% for yeast obtained earlier in this work. However, this could be explained by the simple reason that contrary to the amount of yeast sample analysed (1 μg), the amount of tissue analysed in these runs is unknown. It is very well possible that less than 1 μg of tissue sample was analysed, which would explain the lower percentage found. Other possible explanations could be the fact that the tissue sample is more complex than yeast or that it contains more low abundance proteins that could not be identified in micro-LC-MS due to the decrease in sensitivity. Throughout the experiment the same MS parameters have been used for the nano- and micro flow rate experiments. The number of spectra recorded for the 22- and 11-minute gradient is similar in both nano- and micro-LC-MS. For the 11-minute gradient, more spectra were obtained in micro than in nano, but the number of identifications in nano remains higher than in micro. The same database and search settings were used to process all the data, so this result is unexpected as it was expected that more spectra would result in a higher number of identifications. However, this appears not to be the case in this experiment. A likely explanation would be that the spectra recorded in micro are of lesser quality than those obtained in micro. When only considering the micro conditions, the results also reveal that the difference in the number of identified proteins is small between the 45- and 22.5 -minute gradient. Halving the gradient length to 22.5 minutes does not affect the number of identifications significantly. However, the difference in identifications between the 22.5- and 11.25-minute gradient is significant as half of the proteome coverage is lost. So, when analysing fresh frozen samples using a micro-flow rate, the gradient can be shortened to 22 minutes without losing proteome coverage. Shortening the gradient even further does affect the proteome coverage.

For the FFPE tissues, only 20% of the proteins identified using the nano flow rate could also be identified using the micro flow rate. This is likely because FFPE samples are difficult to analyse in nano-LC-MS already. As mentioned earlier, sensitivity is lost when the flowrate is increased. It seems that for FFPE samples the loss of sensitivity has a big impact, possibly because the proteins in the sample are low in abundance making it hard to identify them in micro when the sensitivity and number of

(28)

28 recorded spectra are lower than in nano. When using the nano flowrate, the number of spectra and protein identifications clearly decreases with the gradient length for the same reason as for the fresh frozen samples before. In micro-LC-MS, although the number of identified proteins is much lower than in nano-LC-MS, the difference in number of spectra and identifications between the different gradient conditions is little. Striking is also that contrary to all the previous obtained results, the longest gradient leads to the lowest number of protein identifications. Apparently, for the analysis of FFPE samples in micro-LC-MS the gradient length does not influence the number of protein identifications. So, for FFPE tissue analysis the analysis time in micro-LC-MS can be shortened. However, this might not be a good idea as the proteome coverage in micro-LC-MS is very low in comparison to nano-LC-MS.

(29)

29

4.

Conclusion and further perspectives

In this work new, faster methods for protein digest analysis using nano- and micro-LC-MS have been developed and evaluated. By doing so, an attempt has been made to decrease sample analysis time, which is of great importance in clinical settings. Using nano flowrates, longer columns lead to an increased number of protein quantifications because the quality of the separation is improved. When a 10 cm long column is used in nano-LC-MS, the gradient length has little effect on the number of proteins quantified, indicating that the analysis time can be shortened without losing proteome coverage. When comparing nano- to micro-LC-MS for yeast samples, 73% of the proteins identified in nano- can also be identified in micro-LC-MS when using the same 60-minute gradient. For quantitation using SWATH-MS the recovery after switching from nano- to micro-LC-MS is 54%, this could be improved upon by optimization of the SWATH-MS parameters. When analysing kidney tissue samples, for fresh frozen samples 50% of the proteins identified in nano- were also identified in micro-LC-MS. For FFPE samples this percentage was found to be 20%. The results of this work indicate that analysis time can be shortened by using micro flow and shorter gradients, however proteome coverage is lost.

Improvements remain to be done to the micro-methods developed in this work. For instance, the SWATH-MS parameters should be optimized further and experimented with to see how they affect the number of quantified proteins. Also, the fresh frozen kidney tissue samples could be analysed using SWATH-MS to investigate how many proteins can accurately be quantified in nano- and micro-LC-MS. To try to improve the proteome coverage in micro-LC-MS, experiments injecting more sample could be done. Another way of improving proteome coverage could be using a 2D-LC-MS setup using micro flow. Moreover, it would be interesting to try 320 μm inner diameter columns at a 5 μl/min flowrate to decrease the analysis time further and see what the effects would be. Though a lot of work remains to be done, with this work an attempt at the implementation of faster quantitative methods using micro-LC-SWATH-MS for the analysis of clinical samples has been made.

(30)

30

5.

Acknowledgements

I would like to thank dr. Alina Astefanei for her daily supervision and dedication to helping me with this project. Many thanks to prof. dr. Garry Corthals for making this project possible and to the biomolecular systems analytics group for welcoming me. Furthermore, I would like to thank Marvin Dittrich for allowing me to analyse his samples and Peter Schoenmakers for being my second examiner.

(31)

31

6. References

(1) Mardamshina, M.; Geiger, T. Am. J. Pathol. 2017, 187 (10), 2175–2184.

(2) Karpievitch, Y. V; Polpitiya, A. D.; Anderson, G. A.; Smith, R. D.; Dabney, A. R. Ann Appl Stat

2011, 4 (4), 1797–1823.

(3) Graves, P. R.; Haystead, T. a J. Microbiol. Mol. Biol. Rev. Microbiol. Mol. Biol. Rev. 2002, 66 (1), 39–63.

(4) Aebersold, R.; Mann, M. Nature 2003, 422, 198–207.

(5) Bruderer, R.; Bernhardt, O. M.; Gandhi, T.; Miladinovic, M.; Cheng, L.; Messner, S.;

Ehrenberger, T.; Zanotelli, V.; Butscheid, Y.; Escher, C.; Vitek, O.; Rinner, O.; Reiter, L. Mol.

Cell. Proteomics 2015, 14 (5), 1400–1410.

(6) Gillet, L. C.; Navarro, P.; Tate, S.; Ro, H.; Selevsek, N.; Reiter, L.; Bonner, R.; Aebersold, R. Mol.

Cell. Proteomics 2012, 11 (6), 1–17.

(7) Röst, H. L.; Rosenberger, G.; Navarro, P.; Gillet, L.; Milandinovic, S. M.; Schubert, O. T.; Wolski, W.; Collins, B. C.; Malmström, L.; Aebersold, R. Nat. Biotechnol. 2014, 32 (3).

(8) Zhang, Y.; Fonslow, B. R.; Shan, B.; Baek, M.; Yates, J. R. Chem. Rev. 2013, 113, 2343–2394. (9) Shishkova, E.; Hebert, A. S.; Coon, J. J. Cell Syst. 2016, 3 (4), 321–324.

(10) Ishihama, Y. J Chromatogr A 2005, 1067, 73–83.

(11) Hebert, A. S.; Richards, A. L.; Bailey, D. J.; Ulbrich, A.; Coughlin, E. E.; Westphall, M. S.; Coon, J. J. Mol. Cell. Proteomics 2014, 13 (1), 339–347.

(12) Pirmoradian, M.; Budamgunta, H.; Chingin, K.; Zhang, B.; Astorga-wells, J.; Zubarev, R. A. Mol.

Cell. Proteomics 2013, 3330–3338.

(13) Olkowicz, M.; Rybakowska, I.; Chlopicki, S.; Smolenski, R. T. Talanta 2015, 131, 510–520. (14) Selevsek, N.; Chang, C.-Y.; Gillet, L. C.; Navarro, P.; Bernhardt, O. M.; Reiter, L.; Cheng, L.;

Vitek, O.; Aebersold, R. Mol. Cell. Proteomics 2015, 14 (3), 739–749. (15) Needham, S. R.; Valaskovic, G. A. Bioanalysis 2015, 7, 1061–1064. (16) Sneekes, E.; Rieux, L.; Swart, R. Thermo Fish. Sci. 2013, 2–7.

(32)

32 (17) Hunter, C.; Zhu, X.; Soest, R. Van. Biomarkers Omi. SCIEX 1–3.

(18) Motamedchaboki, K.; Soest, R. Van; Hunter, C. L. Biomarkers Omi. SCIEX 2016, 2–5. (19) Wenner, B. R.; Lynn, B. C. J. Am. Soc. Mass Spectrom. 2004, 15 (2), 150–157. (20) Camerini, S.; Mauri, P. J. Chromatogr. A 2015, 1381, 1–12.

(21) Xu, P.; Duong, D. M.; Peng, J. J. Proteome Res. 2009, 8 (8), 3944–3950.

(22) Meiring, H. D.; van der Heeft, E.; ten Hove, G. J.; de Jong, A. P. J. M. J. Sep. Sci. 2002, 25 (9), 557–568.

(23) Dolan, J. W. LCGC North Am. 2010, 28 (2), 114–120.

(24) Hunter, C. L.; Morrice, N. Biomarkers Omi. SCIEX 2016, 1–5.

(25) Vowinckel, J.; Zelezniak, A.; Kibler, A.; Bruderer, R.; Mülleder, M.; Reiter, L.; Ralser, M. bioRxiv

Prepr. 2016, 1–17.

(26) Köcher, T.; Pichler, P.; De Pra, M.; Rieux, L.; Swart, R. Proteomics 2014, 14 (17–18), 1999– 2007.

(27) Hsieh, E. J.; Bereman, M. S.; Durand, S.; Valaskovic, G. A.; Maccoss, M. J. J Am Soc Mass

Spectrom 2013, 24 (1), 148–153.

(28) Tolley, L.; Jorgenson, J. W.; Moseley, M. A. Anal. Chem. 2001, 73 (13), 2985–2991.

(29) Eeltink, S.; Dolman, S.; Swart, R.; Ursem, M.; Schoenmakers, P. J. J. Chromatogr. A 2009, 1216 (44), 7368–7374.

(30) Nägele, E.; Vollmer, M.; Hörth, P. J. Biomol. Tech. 2004, 15 (2), 134–143. (31) Marx, V. Nat. Methods 2013, 10 (1), 19–22.

(32) Doerr, A. Nat. Publ. Gr. 2010, 7 (1), 35.

(33) Picotti, P.; Aebersold, R. Nat. Methods 2015, 9 (6), 555–566.

(34) Chapman, J. D.; Goodlett, D. R.; Masselon, C. D. Mass Spectrom. Rev. 2014, 33, 452–470. (35) Liu, Y.; Hüttenhain, R.; Surinova, S.; Gillet, L.; Mouritsen, J.; Brunner, R.; Navarro, P.;

(33)

33 (36) Bourassa, S.; Fournier, F.; Nehmé, B.; Kelly, I.; Tremblay, A. PLoS One 2015, 1–16.

(37) Zhu, W.; Smith, J. W.; Huang, C. J. Biomed. Biotechnol. 2010, 2010, 1–6.

(38) Bantscheff, M.; Schirle, M.; Sweetman, G.; Rick, J.; Kuster, B. Anal. Bioanal. Chem. 2007, 389 (4), 1017–1031.

(39) Schubert, O. T.; Gillet, L. C.; Collins, B. C.; Navarro, P.; Rosenberger, G.; Wolski, W. E.; Lam, H.; Amodei, D.; Mallick, P.; Maclean, B.; Aebersold, R. Nat. Protoc. 2015, 10 (3), 426–441.

(40) DIA / SWATH Course 2017 2017, 1–13.

(41) Vowinckel, J.; Zelezniak, A.; Bruderer, R.; Mülleder, M.; Reiter, L.; Ralser, M. Nat. Sci. reports

2018, 8:4346 (February), 1–10.

(42) Ho, Y.; Gruhler, A.; Heilbut, A.; Bader, G. D.; Moore, L.; Adams, S.; Millar, A.; Taylor, P.; Bennett, K.; Boutilier, K.; Yang, L.; Wolting, C.; Donaldson, I.; Schandorff, S.; Shewnarane, J.; Vo, M.; Taggart, J.; Goudreault, M.; Muskat, B.; Alfarano, C.; Dewar, D.; Lin, Z.; Michalickova, K.; Willems, A. R.; Sassi, H.; Nielsen, P. A.; Rasmussen, K. J.; Andersen, J. R.; Johansen, L. E.; Matthiesen, J.; Hendrickson, R. C.; Gleeson, F.; Pawson, T.; Moran, M. F.; Durocher, D.; Mann, M.; Hogue, C. W. V; Figeys, D.; Tyers, M. Nature 2002, 415 (January), 2–5.

(43) Godoy, L. M. F. De; Olsen, J. V; Souza, G. A. De; Li, G.; Mortensen, P.; Mann, M. 2006, 1–15. (44) Navarro, P.; Kuharev, J.; Gillet, L. C.; Bernhardt, O. M.; Röst, H. L.; Tate, S. A.; Tsou, C.; Reiter,

L.; Rosenberger, G.; Perez-riverol, Y.; Nesvizhskii, A. I. 2017, 34 (11), 1130–1136.

(45) Bruderer, R.; Bernhardt, O. M.; Gandhi, T.; Xuan, Y.; Sondermann, J.; Schmidt, M. 2017, 41 (0). (46) Paulo, J. A.; Lee, L. S.; Banks, P. A.; Steen, H.; Conwell, D. L. Pancreas 2012, 41 (2), 175–185. (47) Magdeldin, S.; Yamamoto, T. Proteomics 2012, 12 (7), 1045–1058.

Increasing the speed of protein digest analysis using nano- and micro- LC-MS strategies

Bachelor Thesis Chemistry