Integrated Microfluidic Device for Mass Spectrometry-Based Proteomics and Its Application to...

Integrated Microfluidic Device for MassSpectrometry-Based Proteomics and ItsApplication to Biomarker Discovery ProgramsMarie-Helene Fortier,† Eric Bonneil,‡,§ Paul Goodley,| and Pierre Thibault*,†,‡,§

Department of Chemistry, Universite de Montreal, Montreal, Canada, Caprion Pharmaceuticals, Montreal, Canada,Institute for Research in Immunology and Cancer, Universite de Montreal, Montreal, Canada, and Agilent Technologies,5301 Stevens Creek Boulevard, Santa Clara, California 95051

The present investigation describes the analytical perfor-mances of a microfluidic device comprising an enrichmentcolumn, a reversed-phase separation channel, and ananoelectrospray emitter embedded altogether in poly-imide layers. This configuration minimizes transfer linesand connections and reduces postcolumn peak broaden-ing and dead volumes. This compact and versatile modu-lar nanoLC-chip system was interfaced to both ion trapand time-of-flight mass spectrometers, and its analyticalpotentials were evaluated in the context of proteomicsapplications. The figures of merit of this system in termsof peak capacity, reproducibility, sensitivity, and lineardynamic range of peptide detection were determinedusing tryptic digests of complex protein extracts includingalbumin- and immunoglobulin-depleted rat plasma sam-ples. The analysis of peak profiles for more than 600peptide ions reproducibly detected across replicatenanoLC-chip-MS runs (n ) 10) indicated that this systemprovided good reproducibility of retention time and peakintensity with RSD values of less than 0.5 and 9.1%,respectively. Variation in peptide abundance as low as2-fold changes was identified for spiked tryptic digestspresent at levels of 2-5 fmol in plasma samples. Sensitiv-ity measurements were performed on dilution series ofprotein digests spiked into rat plasma samples andprovided a detection limit of 1-5 fmol. The modularconcept of the microfluidic systems also facilitated theintegration of two-dimensional chromatography (strongcation exchange/C18) thereby increasing the sample load-ing and selectivity of the nanoLC-chip-MS system. Theapplication of this integrated device was evaluated forcomplex rat plasma samples to compare the number ofprotein identifications obtained using one- and two-dimensional nanoLC-chip-MS/MS.

The purpose of proteomics goes beyond the identification ofproteins in an organelle or in biological fluids. The need to

understand the biological mechanisms involved in cancer andinfectious and inflammatory diseases at the clinical levels impliesnot only comprehensive protein identification but also expressionprofiling of proteins across healthy and disease patient samples.In this context, the challenge of proteomics lies in the complexityof protein mixtures, the number of samples, and the reproduc-ibility of analysis. Most popular mass spectrometry-based pro-teomic approaches1,2 rely on two major strategies: (i) thetraditional protein gel-based separation followed by capillary liquidchromatography-mass spectrometry (LC-MS) analyses of trypticdigests of 1D bands or 2D spots or (ii) the shotgun proteomicapproach (2D-LC-MS) combining strong cation exchange (SCX)separation of tryptic digest followed by capillary LC-MS analysesof the salt fractions. Both techniques have shown to be veryeffective for protein identification3-6 and expression profiling.7-9

Capillary LC-MS systems used in those two strategies consistof an enrichment precolumn, an actuated rotor valve, and ananoscale C18 reversed-phase analytical column terminated by atransfer line to the nanoelectrospray emitter. In systems operatingat submicroliter per minute flow rate, transfer lines and deadvolume must be minimized significantly in order to preserveseparation efficiencies. Physical constraints with instrumentationand the requirements for routine operation and avoidance ofcapillary leakages and clogging often force the user to tradeseparation performance for ease of operation and maintenance.In addition, the serial nature of current capillary LC-MS systemslimits the sample throughput and potential application for multi-plex analyses although recent reports have described the suc-cessful implementation of a multiplex LC-MS system for multi-

* Corresponding author. E-mail: [email protected]. Phone: (514)343-6910. Fax: (514) 343-7586.

† Universite de Montreal.‡ Caprion Pharmaceuticals.§ Present address: Institute for Research in Immunology and Cancer,

Universite de Montreal, Montreal, Canada.| Agilent Technologies.

(1) Domon, B.; Broder, S. J. Proteome Res. 2004, 3, 253-260.(2) Marko-Varga, M.; Fehniger, T. E. J. Proteome Res. 2004, 3, 167-178.(3) Fountoulakis, M.; Tsangaris, G.; Oh, J.; Maris, A.; Lubec, G. J. Chromatogr.,

A 2004, 1038, 247-265.(4) Scirle, M.; Heurtier, M. A.; Kuster, B. Mol. Cell. Proteomics 2003, 2 (12),

1297-1305.(5) Jacobs, J. M.; Mottaz, H. M.; Yu, L.; Anderson, D. J.; Moore, R. J.; Chen,

W. U.; Auberry, K. J.; Strittmatter, E. F.; Monroe, M. E.; Thrall, B. D.; Camp,D. G., II; Smith, R. D. J. Proteome Res. 2004, 1, 68-75.

(6) Whenzong, Z.; Reich, C. I.; Olsen, G. J.; Giometti, C. S.; Yates, R. J., III. J.Proteome Res. 2004, 3, 538-548.

(7) Righetti, P. G.; Campostrini, N.; Pascali, J.; Hamdan, H.; Astner, H. Eur. J.Mass Spectrom. 2004, 10 (3), 335-348.

(8) Wang, W.; Zhou, H.; Lin, H.; Roy, S.; Shaler, T. A.; Hill, L. R.; Norton, S.;Kumar, P.; Anderle, M.; Becker, C. H. Anal. Chem. 2003, 75, 4818-4826.

(9) Radulovic, D.; Jelveh, S.; Ryu, S.; Hamilton, T. G.; Foss, E.; Mao, Y.; Emili,A. Mol. Cell. Proteomics, in press.

Anal. Chem. 2005, 77, 1631-1640

10.1021/ac048506d CCC: $30.25 © 2005 American Chemical Society Analytical Chemistry, Vol. 77, No. 6, March 15, 2005 1631Published on Web 02/08/2005

dimensional liquid chromatography separations.10 In this regard,a fully integrated system comprising desalting and separationcolumns terminated by a short nanoelectrospray emitter wouldoffer significant advantages over present nanoLC-MS configura-tions. Furthermore, miniaturization of the separation channelincreases sensitivity due to greater concentrations of analyte bandseluting from smaller inner diameter columns.11,12 Higher demandsfor increased throughput, sensitivity, and robustness from thepharmaceutical industry has led to the integration and miniaturiza-tion of sample separation systems as well as coupling of microde-vices with electrospray ionization mass spectrometry.13,14

For practical reasons, separation of peptides on microfluidicsdevices has been historically performed using systems based oncapillary electrophoresis (CE)15 and more recently using capillaryelectrochromatography.16 A number of reports have presented thecoupling of CE-based microfluidics devices to nanoelectrospraymass spectrometry.17-21 On-line stacking or adsorption precon-centration prior to electrophoretic separation was also developedto enhance sensitivity and to decrease concentration detectionlimits to low-nanomolar levels.22,23 Significant advances in automa-tion are still needed to achieve the required throughput andruggedness sought for the rapid and reproducible proteomicsanalyses. To this end, efforts are also devoted to microfabricatedchip designs enabling sample multiplexing.24,25

On the other hand, microchip LC-MS separations have beenfirst implemented using either microchip wall derivatization witha stationary phase26,27 or UV light-induced polymerization ofmonolithic materials into the microchannel.28,29 Although thosetwo approaches do not suffer from clogging and back pressure,

the reduced adsorption area often results in lower loading capacitythan traditional bead stationary phase. In the present work, weinvestigated the analytical performances of a newly designednanoLC-chip-MS device for proteomic analyses.30 This compactpolyimide microfluidic device comprises laser-ablated channelsembedding C18-packed desalting and analytical separation channelstogether with a nanoelectrospray tip. The chip is mounted on ahigh-pressure manifold that consists of a face-sealed rotary valvethat allows switching between sample desalting and separationwhile minimizing time delays and dead volumes. Application ofthis LC-chip-MS device using one- and two-dimensional separa-tions is demonstrated for expression profiling and protein iden-tification of rat plasma samples using both ion trap and time-of-flight (TOF) mass spectrometers.

EXPERIMENTAL SECTIONChemicals and Materials. Fused-silica capillaries were pur-

chased from Polymicro Technologies (Phoenix, AZ). Teflon andPEEK tubing were obtained from Supelco (Bellefonte, PA). HPLCgrade water and acetonitrile (ACN) were purchased from FisherScientific (Whitby, ON, Canada). Formic acid (FA) and ammoniumacetate were obtained from EM Science (Mississauga, ON,Canada). Protein standard digests used in the analyses werepurchased from Michrom Bioresources (Auburn, CA). Rat plasmasample was obtained from Bioreclamation (Hempstead, NY). TheJupiter C18 5-µm particle material used for packing the microfluidicdevices was purchased from Phenomenex (Torrance, CA). ThenanoLC-chip and interface system were provided by AgilentTechnologies (Santa Clara, CA). Strong cation exchange materialwas purchased from PolyLC (Columbia, MD).

Sample Preparation. A solution of eight protein standards(40 fmol/µL each) was prepared from individual digests of bovineserum albumin, rabbit phosphorylase b, yeast alcohol dehydro-genase, bovine deoxy-ribonuclease, horseradish peroxidase C1A,bovine glyceraldehyde 3P dehydrogenase, Escherichia coli â-ga-lactosidase, and bovine carboxypeptidase A (Michrom Biore-sources, Auburn, CA) diluted in a H2O/5% ACN/0.2% FA solution.Complex protein extracts from rat plasma (Bioreclamation, Hemp-stead, NY) were depleted with IgG (protein G, Amersham) andalbumin (MP Biomedicals) columns to remove these high-abundance proteins. The rat plasma sample was digested in 0.1M urea, 50 mM ammonium bicarbonate with Promega Trypsin(Fisher Scientific, Whitby, ON, Canada) overnight and dissolvedin a H2O/5% ACN/0.2% FA solution after being evaporated todryness.

Microchip Device Fabrication. The microchip device fabri-cation is described elsewhere.30 A polyimide film (6.5 × 2.5 cm)is laser-ablated to form microfluidic channels, ports, chambers,and columns. The ablation residues are removed, and the film islaminated with heat and pressure application. The chip is thentrimmed using laser ablation to its final shape and to form theelectrospray tip. The final 8-µm-i.d. tip shape is conical with acircular end with an emitter of 100-µm o.d. and 2 mm in length(Figure 1). A gold layer is electrodeposited near the electrospraytip for contacts to the fluid flow channel. The layer surface bearingthe electrical contacts is directly laminated such that the metal is

(10) Tessier, S.; Bonneil, E.; Carrier, A.; Thibault, P., Proc. of the 52nd Am. Soc.Mass Spectrom. Conf., Nasville, TN, 2004.

(11) Shen, Y.; Zhao, R.; Berger, S. J.; Anderson, G. A.; Rodriguez, N.; Smith, R.D. Anal. Chem. 2002, 74, 4235-4249.

(12) Shen, Y.; Moore, R. J.; Zhao, R.; Blonder, J.; Auberry, D. L.; Masselon, C.;Pasa-Tolic, L.; Hixson, K.; Auberry, K.; Smith, R. D. Anal. Chem. 2003, 75,3596-3605.

(13) Lion, N.; Rohner, T. C.; Dayon, L.; Arnaud, I. L.; Damoc, E.; Youhnovski,N.; Wu, Z.; Roussel, C.; Josserand, J.; Jensenn, H.; Rossier, J. S.; Prybylski,M.; Girault, H. H. Electrophoresis 2003, 24, 3533-3562.

(14) Marko-Varga, G.; Nilsson, J.; Laurell, T. Electrophoresis 2003, 24, 3521-3532.

(15) Belder, D.; Ludwig, M. Electrophoresis 2003, 24 (21), 3595-3606.(16) Bandilla, D.; Skinner C. S. J. Chromatogr., A 2004, 1044 (1-2), 113-129.(17) Xue, Q.; Foret, F.; Dunayevskyi, Y. N.; Zavracky, P. M.; McGruer, N. E.;

Karger, B. L. Anal. Chem. 1997, 69, 426-430.(18) Figeys, D.; Aebersold, R., Anal. Chem. 1998, 70, 3721-3127.(19) Li, J.; Thibault, P.; Bings, N. H.; Skinner, C. D.; Wang, C.; Colyer, C. L.;

Harrison, D. J. Anal. Chem. 1999, 71, 3036-3045.(20) Lazar, I. M.; Ramsey, R. S.; Jacobson, S. C.; Foote, R. S.; Ramsey, J. M. J.

Chromatogr., A 2000, 892, 195-201.(21) Deng, Y. Z.; Zhang, N. W.; Henion, J. Anal. Chem. 2001, 73, 1432-1449.(22) Li, J.; Wang, C.; Kelly, J. F.; Harrison, D. J.; Thibault, P. Electrophoresis 2000,

21, 198-210.(23) Li, J.; LeRiche, T.; Tremblay, T.-L.; Wang, C.; Bonneil, E.; Harrison, D. J.;

Thibault, P. Mol. Cell. Proteomics 2002, 1, 157-168.(24) Xue, Q.; Dunayevskiy, Y. M.; Foret, F.; Karger, B. L. Rapid Commun. Mass

Spectrom. 1997, 11 (12), 1253-1256.(25) Xue, Q.; Foret, F.; Dunayevskiy, Y. N.; Zavracky, P. M.; McGruer, N. E.;

Karger, B. L. Anal. Chem. 1997, 69, 426-430.(26) Kutter, J. P.; Jacobson, S. C.; Ramsey, J. M. J. Microcolumn Sep. 2000, 12,

93-97.(27) Kutter, J. P.; Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 1998, 70, 3291-

3297.(28) Yu, C.; Xu, M. C.; Svec, F.; Frechet, J. M. J. J. Polym. Sci. Polym. Chem.

2002, 40, 755-769.(29) Yu, C.; Davey, M. H.; Svec, F.; Frechet, J. M. J Anal. Chem. 2001, 73,

5088-5096.(30) Yin, H.; Killeen, K.; Brennen, R.; Sobek, D.; Werlich, M.; van de Goor, T.,

Anal. Chem., in press.

1632 Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

embedded inside the device. The enrichment precolumn and theLC column were packed with 5-µm Jupiter C18 material. Theinternal volume of the enrichment precolumn is 40 nL whereasthe LC column has a 75 by 50 µm cross section and a length of4.5 cm. The analytical column is connected to the tip via a 15 µmi.d. × 0.4 mm channel inside the chip.

Chip Valve Manifold. The chip valve manifold is describedin more detail elsewhere.30 Briefly, the chip is positioned betweenthe stator and the rotor of a two-position Rheodyne HPLC valve.A clamping arm mechanism ensures that the stator ports arealigned to the ports on top of the chip, and the rotor channels toports are aligned via registration pins. When the clamping arm isclosed, the valve stator, the chip, and the valve rotor are sealedtogether. The arm is rotated to position the chip orthogonally tothe sampling cone of the mass spectrometer.

NanoLC-Chip-MS and Nano-LC-Chip-MS/MS Analyses.An HP1100 capLC pump (Agilent Technologies) was used todeliver solvent H2O/3% ACN/0.2% FA during sample loading, andan Agilent nanoLC 1100 Series was used for gradient delivery.During sample loading the flow rate was set to 4 µL/min for 6min. LC gradient (mobile phase A: H2O, 0.2% FA; mobile phaseB: ACN, 0.2% FA) was delivered by an Agilent 1100 nanoPumpwith flow rate set at 300 nL/min. Peptides were eluted from thereversed-phase column into the mass spectrometer using a lineargradient elution of 10-60% B over 56 min. The nanoelectrosprayvoltage was set to 2400 V. All nanoLC-chip-MS experiments wereperformed on a TOF and a ion Trap XCT (Agilent Technologies)mass spectrometer.

Two-Dimensional LC Separations with On-Line LC-ChipMS/MS Analyses. The SCX column (0.5 mm i.d. × 6.3 cm) waspacked with PolyLC stationary phase (5-µm diameter, 300-Å poresize). The SCX column was connected directly to the switchingvalve, and was on-line with the chip C18 precolumn during sample

loading, and toggled off-line during reversed-phase peptide separa-tion on the analytical column. Peptides were sequentially elutedfrom the SCX column onto the C18 precolumn with eight saltfractions of ammonium acetate, pH 3.5. Each 10-µL salt fractionwas loaded from the autosampler at 4 µL/min for 15 min anddelivered to the SCX cartridge using an auxiliary pump. Peptideswere eluted from the SCX column using sequential fractions of 0,55, 65, 80, 100, 150, 400, and 1000 mM ammonium acetate, pH3.5. Peptides were eluted from the reversed-phase column intothe mass spectrometer using a gradient from 10 to 60% B over 56min. The nanoelectrospray voltage was set to 2400 V. Tandemmass spectra were acquired with the Agilent MSD Trap XCT massspectrometer using He as a collision gas. Multiply charged ionswith an intensity above a threshold of 40 000 counts were selectedfor MS/MS sequencing. The MS/MS fragmentation amplifiervoltage was set to 1.3V.

Peptide Detection and Clustering. A script was developedto convert the raw data files from the ion trap or TOF instrumentsinto text files representing all m/z, intensity, and time values abovea user-defined intensity threshold, typically 100 counts for TOFand 700 000 counts for ion trap data. These text files were thenprocessed using in-house software enabling data reduction, peptidedetection, and alignment of peptide ion maps (m/z, retention time,intensity) across all samples.31 Segmentation analyses wereperformed using hierarchical clustering with criteria based on m/zand time tolerance (typically (0.1 m/z and (1 min) to generatelists of nonredundant peptide clusters for all replicates or samplesto be compared.

Protein Identification. Database searches were performedagainst a nonredundant NCBI database using Mascot (MatrixScience, London, U.K.) selecting human and rodent species.

(31) Kearney, P.; Thibault, P. J. Bioinf. Comput. Biol. 2003, 1 (1), 183-200.

Figure 1. NanoLC-chip-MS system interfaced to an Agilent ion trap XCT mass spectrometer. The chip rests on an articulated manifold comprisinga valve rotor with a clamping mechanism ensuring proper port alignment and sealing. A two-way translation stage provides easy positioning ofthe device in front of the sampling orifice when the manifold is inserted in the MS inlet. The inset shows a closeup view of the chip device withthe precolumn (40 nL), separation channel (4.5 cm length × 75 µm × 50 µm cross section channel), and nanospray tip (8 µm i.d.).

Analytical Chemistry, Vol. 77, No. 6, March 15, 2005 1633

Parent ion and fragment ion mass tolerances were both set at(0.4 Da.

RESULTS AND DISCUSSIONThis work focuses on the evaluation of the analytical perfor-

mances of a nanoLC-chip system for the analysis of highly complexpeptide mixtures. A photograph of this microfluidic deviceinterfaced to an Agilent ion trap XCT is shown in Figure 1. Theinset shows a closeup of the device integrating together a 40-nLvolume precolumn and a 4.5-cm length analytical column with a75 × 50 µm cross section channel and a nanoelectrospray tip.The chip is placed on a support plate housing a valve rotor onone side and aligned inlet ports on the other side. The valve statoris mounted in a clamping mechanism enabling proper alignmentand sealing of the chip within the manifold. The chip, supportplate, and rotor valve are mounted on an adjustable two-waytranslation stage via a rotating bracket allowing easy positioningof the manifold in front of the mass spectrometer sampling cone.When in position, the nanoelectrospray tip is perpendicular andapproximately 2.5 mm axially and 1 mm off-axis from to the inletcone. Precise positioning of the spray emitter is facilitated througha camera and lighting system.

Analytical Performances of the LC-Chip-MS System. Theapplication of this microchip LC device coupled to the time-of-flight mass spectrometer (nanoLC-chip-TOF) for proteomicsanalyses was first tested using a tryptic digest of an eight-protein

mixture representing a diverse population of peptides withdifferent molecular masses and hydrophobicities. The base peakchromatogram corresponding to the injection of 80 ng of thecombined digest (200 fmol of each protein digest) is presentedin Figure 2a along with the contour profile of m/z versus timeversus intensity in Figure 2b. The logarithmic intensity scaleshows an increase in abundance for ions transitioning betweenblack and bright yellow. Each dot shown on this contour profilerepresents an individual peptide ion defined by specific m/z andtime coordinates. Peptide detection was facilitated using an in-house proprietary software program that analyze the peak profilesof all eluting isotopic components to return 12C monoisotopicpeptide ions based upon their corresponding m/z value, time,charge state, and intensity.31 The charge distribution of the 1578detected peptides is shown in the table of Figure 2a. Comparisonof peptide maps across replicates or sample sets is thus facilitatedby digitizing the nanoLC-chip-MS analysis to produce lists ofdetected peptide ions.

To determine the reproducibility of elution profiles in termsof time, m/z measurements, and intensity, we compared thedigitized peptide maps of 10 replicate analyses of the 8-proteindigest mixture. The identification of individual peptide clusteracross replicates (defined as unique ions reproducibly found within(0.1 m/z and (1.0 min) was achieved using segmentationanalyses with hierarchical clustering. These unique peptide

Figure 2. NanoLC-chip-LC-MS analysis of an 80-ng injection of a eight-proteins tryptic digest. (a) Base peak chromatogram. (b) Contourprofile of m/z vs retention time vs intensity. Peptide ions are displayed on a logarithmic scale of increasing intensity ranging from black to brightyellow. The number of detected peptide ions and their corresponding charge state are shown in the table.


clusters were then compared to determine their frequency ofobservation across all 10 replicates. Figure 3a represents thedistribution of cluster size defined as the number of times a givenpeptide cluster was detected across the 10 replicates. The numberof peptides was normalized to include all individual peptides foundin a specific cluster size. From a total number of 11 456 detectedpeptide ions, 55% of this peptide population was observed in all

10 replicates whereas 83% was detected in at least half of the entirepeptide population (cluster size g6). It is noteworthy that thiscomparison was performed across all peptide ions irrespective oftheir individual intensities and does not account for the highervariability in peptide detection typically observed for lowerintensity ions. Indeed, examination of the peptide ion populationrepresented by cluster size 1 indicates a higher proportion of lower

Figure 3. Reproducibility of peptide detection across 10 replicate nanoLC-chip-MS analyses of a 8-protein tryptic digest. (a) Number of peptideions found across replicates or (a) cluster size and (b) log of ion intensity.

Figure 4. Reproducibility of retention time and ion intensity for peptides reproducibly found in all 10 replicate injections of a 8-protein trypticdigest. (a) Distribution of RSD values for all of the 626 individual peptide clusters. (b) Log-log plot of individual ion intensity vs the averageintensity of the peptide cluster. Inset shows the frequency of peptide detection across the fold change in ion intensity.


abundance ions when compared to higher cluster sizes (Figure3b). For example, the median value of ion intensity for clustersize 1 was 594 counts (log intensity, 2.77) compared to 707 counts(log intensity, 2.85) and 1480 (log intensity, 3.17) for cluster sizes6 and 10, respectively. Obviously, low-abundance peptides aremore difficult to detect reproducibly across numerous replicatesdue to factors such as retention competition on the precolumnand analytical column, background variation, and ionizationsuppression.

To determine the reproducibility of observed retention times,m/z values, and intensities we compared the distribution of relativestandard deviations (RSD) across peptide clusters. Figure 4a givesthe relative standard deviation of retention times for 626 peptideclusters found in all 10 replicates. As indicated, RSD values variedtypically between 0.08 and 1.05% with a median value of 0.27%.Furthermore, 95% of peptide ions reproducibly detected acrossall replicates showed a variation in retention time below 0.15 min,corresponding to an RSD of less than 0.5%. Similarly, a comparisonof m/z measurements indicated that individual peptide clusterswere represented by RSD values ranging from 0.001 to 0.003%,and 95% of all reproducibly detected ions were within (25 ppm.It is noteworthy that these m/z measurements were performedwith external calibration and improvement in mass accuracy toless than 5 ppm could be achieved by co-infusion of an internalstandard.

The reproducibility in ion intensity observed across peptideclusters of size 10 is shown in Figure 4b, which compares thelog-log plot of individual versus average intensities. Most of theions are tightly distributed along a 45° angled line for all replicates,indicating a narrow dispersion across 2.5 orders of magnitude inintensity. As observed from Figure 4b, intensity variations ap-peared more significant for lower abundance ions with an averageRSD value of 9.12% for ions of less than 2000 counts compared toan average RSD value of 6.19% for ions above this intensity level.An expression plot of fold change variation in intensity for allpeptide clusters of size 10 (inset Figure 4b) highlights that 95%of peptide clusters showed a variation of less than (20% ((1.2 infold change) across replicate injections, attesting of the goodreproducibility of the present system. Reproducibility of retentiontime and peak intensity of the nanoLC-chip-TOF system iscomparable to that observed previously on a nanoscale capillaryLC system coupled to a Q-Star mass spectrometer32

The evaluation of peak capacity and peak asymmetry wasperformed on peptide clusters reproducibly observed across all

10 replicates. Peak width measurements were performed at 10and 50% height for detected ions. An example of a peak profilefor the doubly charge peptide ion at m/z 501.72 is shown in Figure5a. The distribution of peptide clusters as a function of peak widthsat 50 and 10% is presented in Figure 5b, and data obtained aresummarized in Table 1. The wider distribution observed for peakwidth at 10% is due to larger variability in intensity values formeasurements taken at levels approaching background noise. Themean values for peak widths measured at 50 and 10% height were0.34 and 0.76 min, respectively (Table 1). In comparison, packedcapillary columns of 12 cm × 75 µm i.d. gave average peak widthsof 0.27 and 0.65 min for measurements taken at 50 and 10% height(data not shown).

Peak asymmetry was also measured for a set of 240 repro-ducibly observed peptides with intensity above 10 000 counts tominimize variability attributed to poor peak definition at levelsapproaching background noise. The ratio of half-width along thecenter peak line, b/a (Figure 5b), gave median peak asymmetryvalues of 1.13 and 1.22 for measurements taken at 50 and 10%height, respectively (Table 1). In contrast, peak asymmetry valuesof 1.49 and 1.93 were obtained for packed capillary columns of 12cm × 75 µm i.d. for measurements taken at 50 and 10% height(data not shown). The low asymmetry values obtained for themicrofluidic device was attributed to low dead volumes andreduced postcolumn peak-broadening contributions. Correspond-ingly, the peak capacity values, nc, were calculated at 50% heightand 10% height from the peak width measurements (Table 1) andprovide an estimate of the number of components that can beseparated to a specified resolution within a given separationspace.33 An average nc value was calculated from the ratio of the

(32) Shen, Y., Zhao, R., Berger, S. J., Anderson, G. A., Rodriguez, N., Smith, R.D. Anal. Chem. 2002, 74, 4235-4249.

Figure 5. Peak asymmetry and peak width measurements on the nanoLC-chip-MS system. (a) Extracted ion chromatogram on a doublyprotonated peptide at m/z 501.7 eluting in the most dense region of the contour profile shown in Figure 2b. (b) Distribution of peptide ionsaccording to peak widths at 50 and 10% height.

Table 1. Average Peak Width and AsymmetryMeasurements Obtained from NanoLC-chip-MSAnalyses of Tryptic Peptides from an Eight-ProteinDigest

%height

peak width(min)

peak capacity(nc)

asymmetry(b/a)a

50 0.34 141 1.1310 0.76 63 1.22

a Statistics across 240 peptides (n ) 5) with intensity higher than10 000 counts.


elution time window and the peak width as described previ-ously.32,34 The modest average nc value of 141 (50% height)compared to other reports using the 70 cm × 50 µm i.d. column32

reflects the use of an on-line trapping column, the narrower elutionwindow, and the difference in particle size used. Indeed, increasedpeak capacity could be obtained by using shallower gradientelution for a given column configuration. However, more ap-preciable gains in separation performance are expected byreducing the precolumn volume, increasing the column length,and using smaller particle size.

Another important consideration is the ruggedness and easeof operation of the present system. Despite the reduced dimen-sions of the precolumn and analytical columns compared to moreconventional capillary LC systems, the integrated format of thisdevice enables a significant reduction in dead volumes whilesimultaneously minimizing leakage through tubing connections.These advantages were translated by reduced peak asymmetrycompared to capillary columns of comparable size. Although therobustness of this microfluidic system was not necessarily thesubject of the present investigation, the nanoLC-chip-MS systemwas operated in an uninterrupted manner for more than 100injections of complex proteins digests on a single chip withoutany noticeable degradation of performance. Moreover, the articu-lated manifold facilitates chip replacement and system recondition-ing within less than 5 min.

Detection and Identification of Spiked Proteins in a BloodBiomarker Sample. Replicate analyses described in the previoussection highlighted the good reproducibility of the present systemwhere RSD values of less than 0.5, 0.003, and 9.1% were obtainedon retention time, m/z value, and intensity measurements,respectively. These reproducibility measurements provided com-fortable operating margins for peak detection across comparablesamples. To evaluate the analytical potentials of the present micro-fluidic device in the context of expression profiling measurements,we performed a control experiment on complex rat plasma sam-ples. In this case, each sample injected on the chip correspondedto different levels ranging from 0 to 250 fmol of the eight-proteindigest spiked into a 100-ng digest of plasma sample. Spike serieswere injected in triplicate interleaving three blank samples beforethe next series of injections. Peptide maps and clustering analyseswere performed as described in the previous section.

The average number of reproducibly detected peptide ions inall three nonspiked replicate samples was 992 ions. The distribu-tion of peptide ions unique to the plasma sample and reproduciblyfound in the spike series was relatively uniform and ranged from931 to 976 ions. A comparison of peptide ions reproduciblydetected and specific to the spiked samples progressively in-creased from 4 to 604 ions for spike levels of 1.2 to 120 fmol. Asexpected, changes in peptide intensity observed at the low-femtomole level were observed for tryptic peptides showing ahigher signal response. To visualize the intensity changes oftryptic peptide ions from spiked proteins, we compared the log-log plots of the 24-, 48-, and 120-fmol spike with that of the 12-fmol spike (Figure 6a). At the lower spike level, we identified atotal of 61 unique peptide ions reproducibly detected in all threereplicate spike samples of 12 fmol and above. The intensity

progression of these ions with their corresponding spike level(vertical lines) can be visualized in Figure 6a for the secondreplicate of the spike series. Confirmation of the identity of thesuspected spiked ions was obtained by coupling the nanoLC-chipdevice to an ion trap instrument. Examples of MS-MS spectraacquired using this targeted approach are presented in Figure6b and c for m/z 480.6 and 435.2 (circled in Figure 6a), respective-ly. MS-MS spectra for the doubly protonated ion m/z 480.6 (Figure6b) and 435.2 (Figure 6c) enabled the unambiguous identificationof peptides DTIVNELR and FAAYLER corresponding to spikedproteins horseradish peroxidase and phosphorylase b, respectively.It is noteworthy that differential expression for spiked amount aslow as 4.8 fmol was reproducibly detected. Table 2 shows the lin-earity of response and limit of detection for 10 randomly selectedspiked peptides. The limits of detection (S/N > 3) obtained forthese ions ranged between 2.4 and 11.9 fmol. Excellent linearityof response was obtained across 2 orders of magnitude in intensityas indicated by the correlation coefficients ranging between 0.9928and 0.9998. In the present situation, the dynamic range of peptidedetection is ∼14 parts per million given a loading capacity of 200ng and a lower detection boundary of 2 fmol for an average trypticpeptide molecular mass of 1350 Da. It is noteworthy that thepresent chip configuration provides a loading capacity of ∼700ng, which can be further extended by 1 order of magnitude usingon-line two-dimensional chromatography.

Application of the NanoLC-Chip Device with On-Line Two-Dimensional Chromatography. The modular design of themicrofluidic LC system enables the integration of on-line two-dimensional chromatography (2D-LC) by inserting a SCX trappingcolumn directly into the sample inlet port of the switching valve.In addition to providing enhanced selectivity, the coupling of 2D-LC also increases the overall number of detected peptides throughhigher sample loading. In the present investigation, we comparedthe overall improvements in sequence coverage and in the numberof proteins assigned for the nanoLC-chip system with and withoutan SCX trapping column. This comparison was performed on abiomarker sample obtained from rat plasma.

The analysis of serum samples typifies one the most significantanalytical challenge in proteomics research given the diversity ofproteins, their corresponding dynamic range in abundance (>10orders of magnitude), and the high proportion of proteins suchas albumin and immunoglobulins.35,36 Indeed, 22 proteins alonerepresent more than 99% of the plasma protein content.37 Althoughcommon methods remove major proteins such as albumin, IgG,and transferrins, the proportion of these proteins still remainsignificant even after depletion. The inherent advantages of 2D-LC in terms of its high resolving power and increased loadingcapacity compared to one-dimensional reversed-phase chroma-tography thus offers a means to address the overwhelmingcomplexity and diversity of the plasma proteome.36,38

The number and distribution of major blood proteins obtainedfor the injection of 200 ng of an albumin and IgG-depleted rat

(33) Giddings, J. C. Anal. Chem. 1967, 39, 9, 1027-1028.(34) Lan, K.; Jorgenson, J. W. Anal. Chem. 1999, 71, 709-714.

(35) Anderson, N. L.; Anderson, N. G. Mol. Cell. Proteomics 2002, 1, 845-867.(36) Adkins, J. H.; Varnum, S. M.; Auberrys K. J.; Moore, R. J.; Angell, R. H.;

Smith, R. D.; Springer, D. L.; Pounds, G. J. Mol. Cell. Proteomics 2002, 1,947-955.

(37) Tirumalai, R. S.; Chan, K. C.; Prieto, D. A.; Isaaq, H. J.; Conrads, T. P.;Veenstra, T. D. Mol. Cell. Proteomics 2003, 2, 1096-1103.

(38) Shen, Y.; Jacobs, J. M.; Camp, D. J.; Fang, R.; Moore, R. J.; Smith, R. D.;Xiao, W.; Davis, R. W.; Tompkins, R. G. Anal. Chem. 2004, 76, 1134-1144.


plasma sample is shown in Figure 7a for data acquired withnanoLC-chip system coupled to an ion trap instrument. A total of270 MS-MS spectra of good quality were acquired from a singlenanoLC-chip-MS/MS run of which 206 were assigned to 47nonredundant protein clusters. The largest proportion of theassigned spectra was matched to classic plasma proteins (36%)such as albumin, transferrin, complement components, andapolipoproteins. Cellular leakage proteins represented 17% whereassecreted or extracellular proteins corresponded to 18% of allproteins detected.

The same sample was subsequently analyzed using an on-line2D-nanoLC-chip system. In this case, 1.5 µg of rat plasma digestwas loaded on the SCX trap and eight fractions of different salt con-centrations were injected to obtain a relatively uniform distributionof peptide ions with minimal carryover across fractions. A total of2189 MS/MS spectra of good quality were obtained of which 1098were assigned to 111 distinct protein clusters. Assigned proteinswere distributed across secreted or extracellular (34%), classicplasma (29%), and cell leakage (18%) proteins (Figure 7b). Inter-estingly, classic proteins and immunoglobulins represented re-

Figure 6. Detection of peptide ions from a digest of rat plasma proteins spiked with increasing amounts of a eight-protein tryptic digest. (a)Log-log plot of peptide intensity for protein spike levels of 24, 48, and 120 fmol compared to 12 fmol. Tandem mass spectra of (b) m/z 480.6(horseradish peroxidase, DTIVNELR) and (c) m/z 435.2 (phosphorylase b, FAAYLER), both precursor ions are indicated by the arrows in (a).

Table 2. Linearity and Limits of Detection for a Sampleof Spiked Tryptic Peptides Reproducibly Detected inRat Plasma Protein Digest

m/zLOD

(fmol) S/N r2assignedproteinsa

440.21 2+ 4.8 4.0 0.9983 unassigned453.25 3+ 11.9 6.6 0.9971 phospho b479.28 1+ 4.8 4.2 0.9960 unassigned501.76 2+ 11.9 3.5 0.9993 BSA548.94 3+ 4.8 3.6 0.9983 G3PDH566.21 2+ 4.8 3.7 0.9986 G3PDH577.28 2+ 4.8 4.1 0.9990 BSA593.28 2+ 2.4 3.4 0.9928 HSP679.26 2+ 4.8 4.1 0.9998 ADH721.84 2+ 4.8 4.46 0.9992 phospho b

a Phospho b, phosphorylase b; BSA, bovine serum albumin; ADH,alcohol dehydrogenase; G3PDH, glyceraldehyde 3P dehydrogenase;HSP, horseradish peroxidase.

Table 3. Concentration and Sequence Coverage ofProteins Identified from Rat Plasma Sample UsingNanoLC-chip-MS with and without Two DimensionalChromatography

sequencecoverage (%)

proteinsmass(kDa)

abundancein plasma

(Log concnpg/mL)a LC-MS 2D-LC-MS

hemopexin 51.3 8.7-9 [40] 17.0 40.9apolipoprotein A1 30.1 7-7.3 [35] 16.2 53.2complement C3 186.5 6.8-7.0 [35] 2.5 4.7haptoglobin 42.5 6.2-7 [35] 16.0 29.7transthyretin 15.8 5.9-6.2 [35] 0 10.2plasminogen 90.5 5.6-6.1 [35] 0 5.6retinol binding protein 23.2 5.5-5.8 [35] 0 5.0insulin-like growthfactor binding protein

67.0 4.7-5.0 [41] 0 5.1

a References are given in brackets.


spectively less than 36 and 7% irrespective of the method used.Furthermore, 2D-LC provided at least two to three times more se-quence coverage when compared to reversed-phase LC alone (Table3). In the case of proteins such as plasminogen, insulin-like growthfactor binding protein, retinol binding protein, and transthyretin,their identifications could only be obtained using 2D-LC. An ex-ample of this is shown in Figure 8 for the identification of insulin-like growth factor binding protein 1, a protein present at 80 ng/mLin plasma.41 Insulin-like growth factor binding protein prolongs thehalf-life of the insulin growth factor (IGF) and can modulate growth

factor-like activities on cell culture. This protein can also alter theinteraction of IGFs with their cell surface receptors.39

CONCLUSIONThe compact footprint of the present microfluidic system

combined with its modularity and ease of operation provides

(39) Anderson, N. L.; Polanski, M.; Pieper, R.; Gatlin, T.; Tirumalai, R. S.; Conrad,T. P.; Veenstra, T. D.; Adkins, J. N.; Pounds, J. G.; Fagan, R.; Loblei, A.Mol. Cell. Proteomics 2004, 3, 311-326.

(40) Anderson, N. L.; Anderson, N. G.; Haines, L. R.; Hardie, D. B.; Olafson, R.W., Pearson, T. W. J. Proteome Res. 2004, 3, 235-244

Figure 7. Distribution of the plasma proteins identified for the nanoLC-chip-MS/MS using (a) C18 reversed-phase and (b) two-dimensional(SCX/C18) chromatography. A total of 200 ng and 1.5 µg was injected on the nanoLC-chip-MS/MS system for (a) and (b), respectively.

Figure 8. Identification of low-abundance plasma proteins from nanoLC-chip-MS/MS with two-dimensional (SCX/C18) chromatography. (a)Narrow region of the contour profile of the 55 mM ammonium acetate fraction showing the doubly protonated peptide ion at m/z 657.0. (b)Tandem mass spectrum of m/z 657.0 corresponding to the tryptic peptide DFALQNPGVVPR from insulin-like growth factor binding protein 1.


significant advantages for the analysis of complex protein digestsas part of proteomics discovery programs. Although endurancetests were not performed on this system, the chip device wasoperated in an uninterrupted manner for more than 100 injectionsof complex tryptic digests with no apparent loss of performance.The figures of merit of this system in terms of reproducibility inretention time, m/z, and intensity measurements were determinedusing tryptic digests comprising more than 1000 peptide ions. Peakmeasurements taken on more than 626 different peptide clustersfrom replicate injections (n ) 10) of an 8-protein digest providedRSD values of less than 0.5, 0.003, and 0.2% for retention time,m/z, and intensity, respectively. Despite its small dimensions, thismicrofluidic device provided peak capacity of 141 (50% height)for peptide elution taking place over 45 min. The integration ofthe precolumn, analytical column, and nanoelectrospray emitterin a compact format also minimized dead volumes as reflectedby the small peak width (0.3 min at 50% height) and asymmetry(<1.2 at 10% height) measurements.

The excellent reproducibility of this microfluidic systemfacilitated the monitoring of changes in peptide abundance acrosscomplex and dense tryptic digests. In control experiments, 2-foldchanges in differential abundance were detected for as low as 2.4fmol of protein digests spiked in 100-ng plasma biomarker

samples. The integration of SCX trapping column provides anincrease in sample loading and selectivity, thus facilitating thedetection of low-abundance proteins such as that of insulin-likegrowth factor binding protein 1 present at less than 80 ng/mL inplasma biomarker samples. The ability to mine complex data setswith high sensitivity and reproducibility using microfluidic deviceswill be of great significance for protein expression and identifica-tion platforms given the expanding role of mass spectrometry-based proteomics in biomarker and drug discovery programs.

ACKNOWLEDGMENTWe thank Agilent Technologies for providing early access to

the prototype LC-chip-MS system, MSD TOF, and MSD trap XCTmass spectrometers. The authors also thank Agilent staff membersDr. John Michnowicz and Georges Gauthier for their invaluablehelp and assistance, and Linda Cote and Sebastien Marchand fortechnical support. Finally, we acknowledge Caprion colleaguesEric Thibodeau for providing the depleted plasma sample, AlainCarrier and Anik Forest for providing the SCX column, andNavdeep Jaitly for invaluable assistance with the clusteringalgorithms.

Received for review October 8, 2004. Accepted December30, 2004.

AC048506D(41) Heald, A. H.; Siddals, K. W.; Fraser, W.; Taylor, W.; Kaushal, K.; Morris, J.;

Young, R. J.; White, A.; Gibson, J. M. Diabetes 2002, 51 (8), 2629-2636.


Integrated Microfluidic Device for Mass Spectrometry-Based Proteomics and Its Application to...

Documents

Transcript of Integrated Microfluidic Device for Mass Spectrometry-Based Proteomics and Its Application to...