Search tips
Search criteria 


Logo of jbtJBT IndexAssociation Homepage
J Biomol Tech. 2005 December; 16(4): 414–422.
PMCID: PMC2291749

Reproducibility of Retention Time using a Splitless nanoLC Coupled to an ESI-FTICR Mass Spectrometer


Replicate injections of a myoglobin tryptic digest, ultrafiltrates of human serum, and ultrafiltrates of human plasma made on a splitless nanoscale liquid chromatography system coupled to a Fourier-transform ion cyclotron resonance mass spectrometer were utilized to assess analytical reproducibility. The mean (across 19 tryptic fragments detected in at least 3 of 24 replicate injections) of the 95% CIM of retention time is ±6.3 sec and the maximum is ±11.6 sec; when only those tryptic fragments that were found in 24 of 24 replicates are considered, the maximum 95% CIM of retention time drops to ±6.7 sec. This represents a deviation of at most seven spectra. Similarly, in the serum (and plasma) filtrates, 95% of the 393 (312) species observed in 3 replicate injections had a 95% CIM of retention time of ±22.0 (±18.5) sec or less. Ion abundance was similarly reproducible, with an average across those tryptic fragments observed in all 24 replicates of the coefficient of variation of ion abundance equal to 37.0%. This reproducibility represents a significant improvement over prior work, which required flow splitting in order to achieve nanoliter-per-minute flow rates. These improvements in retention time reproducibility will also be observed with mass spectrometers employing mass analyzers other than FT-ICR.

Keywords: Biomarkers, liquid chromatography, mass spectrometry, splitless, nLC

The search for biomarkers1 using mass spectrometry, while extremely promising, faces significant challenges.2 Initial efforts have focused on surface-enhanced laser desorption/ionization (SELDI) time of flight (TOF) methodologies, and, while some recent studies with large sample sizes and well-designed validation may have borne fruit,3 significant questions remain to be answered concerning reproducibility and bias.4 Where SELDI-TOF focuses on speed and throughput with minimal separation before ionization, alternative approaches utilize one or more fractionation methods to reduce the complexity of the resulting spectra and provide a direct correspondence between a discriminatory signal/peak and a biomolecule.

A number of separation or fractionation techniques have been applied to the search for protein biomarkers. A common and promising method is to couple one or more dimensions of liquid chromatography (LC) with soft ionization techniques such as electrospray ionization (ESI), resulting in the observation of both a mass and a retention time for each tryptic fragment57 or intact protein.8 A recent study found that out of 1175 non-redundant proteins identified by four different methods (2D electrophoresis, 2D LC with and without ultrafiltration, literature search), only 195 were observed by more than one of the four techniques,9 suggesting that multiple analytical approaches may be required to achieve substantial proteome coverage.

Our laboratory has chosen to focus on pre-fractionating serum via molecular-weight cutoff (MWCO) filters, coupling nanoflow LC (nLC) with high-resolution FT-ICR mass spectrometry, using accurate mass and retention time as metrics to pinpoint potential bio-markers, quantifying their abundances in cohorts of diseased and healthy patients, applying statistical classification methodology, and subsequently identifying discriminatory markers by MS/MS.10 An early study applied low-resolution TOF mass spectrometry and preliminary bioinformatics,11 while later work utilized Fourier-transform ion cyclotron resonance (FT-ICR) and algorithms more appropriate for this high-resolution instrumentation.12 Both of these studies used chromatographic equipment that required flow splitting to achieve nanoliter-per-minute flow rates.

In these experiments, precisely determining the chromatographic retention time of a molecular component can allow it to be distinguished from other components that have the same or very similar mass. Silva et al.7 report that with 10-ppm mass tolerance and ±1-min retention time tolerance, 36 species out of 2582 detected in human serum were found to overlap, whereas at ±5-min retention time tolerance, 293 overlapped. As we probe further into the complexities of the human proteome, the likelihood will increase of molecules being found that have similar masses, even at the ppm level, but that are separated chromatographically. Therefore, reproducible retention time measurements impact the number of components that can be resolved in a biomarker discovery study.

However, flow splitting is prone to introducing variability in retention time, especially for the analysis of complex mixtures. Because splitters equalize back pressure between their two flow paths, any change in flow resistance of the nanocolumn (caused, for example, by sample aggregation or particulates on either a pre-column or the nLC column, or partial blockage of nano-ESI emitters) will immediately be compensated by diverting flow from the nanocolumn to the balance flow. An earlier study of nLC retention time reproducibility by our group12 showed a 95% confidence interval of the mean (95% CIM) (the uncertainty interval around the mean value within which 95% of replicate measurements could be expected to fall) of ±54 sec or less for 95% of 497 species detected in at least four of five replicate injections of human serum. Linear increases, or drift, in retention time were observed for each successive injection. Similarly, Wang et al.6 used a ±5-min “loose” tolerance before applying a complex and computationally intensive chromatographic alignment technique. A number of computational methods6,13,14 are known for alignment of chromatograms, generally based on a dynamic programming technique known as “time warping.”15 These algorithms could be used to correct for drift due to splitting.

While there are multiple variables that affect retention time reproducibility (such as reproducibility of gradient mixing, temperature, varying component complexity from one sample to another, and deterioration of the column with increasing number of injections), a significant amount of retention time variability in nLC can be attributed to variations of mobile-phase flow rates caused by the splitting process.

A number of vendors, employing a variety of flow-control strategies, have recently introduced HPLC instruments that provide direct flow control at sub-microliter-per-minute flow rates to eliminate the component of retention time variability that arises from variations in splitter-based nanoLC pumping systems. With the Eksigent system used for this work, flows for each mobile phase are generated and controlled by pneumatic amplifiers, which control flow by continually adjusting the head pressure applied to each mobile phase. The system incorporates three pressure transducers—one in each of the mobile-phase lines, and a third that monitors the column pressure. Solvent flow for each mobile phase is regulated by reading the pressure differential between the mobile-phase transducer and the column transducer, and feeding those values into the pneumatic amplifiers. The relationship between pressure and flow is established via a recommended monthly calibration for each mobile phase.

To demonstrate the viability of this splitless nanoLC system for biomarker discovery, we performed replicate nanoLC-FTICR-MS injections of a myoglobin digest reference sample interdigitated with injections of plasma and serum samples that had been filtered through 50-kDa MWCO filters. In this paper, we describe the reproducibility of retention time and ion abundance achieved using this system.



HPLC-grade water and acetonitrile were purchased from Burdick and Jackson (Muskegon, MI). Formic acid (ACS Reagent, 98–100%) was purchased from Fluka Chemical (Buchs, Switzerland). Zwittergent 3-16 was purchased from Calbiochem (La Jolla, CA). Hepta-fluorobutyric acid (HFBA) was purchased from Pierce Chemical (Milwaukee, WI).

Sample Preparation

Aliquots of serum and plasma from four patients with advanced-stage ovarian cancer and four age-matched, post-menopausal, healthy control patients were diluted and passed through 50-kDa ultrafiltration devices (Microcon, YM50, Millipore, Bedford, MA). Filter devices were first rinsed with 400 μL of 25 mM NH4CO3 (pH 7.6) containing 20% acetonitrile, and the rinsate was discarded. Three hundred twenty micro-liters of 25 mM NH4CO3/20% acetonitrile was applied to the rinsed filters, and 80 μL of serum or plasma was added. Additionally, reference pools of serum and plasma were created by combining 100-μL aliquots from all 10 patients. Five ultrafiltrates were prepared from each of the two reference pools. The samples were centrifuged for 60 min at 3000g. The filtrates were then briefly frozen at −80°C and lyophilized in a vacuum centrifuge. Samples were stored at −80°C until analysis. Each sample was reconstituted in 500 μL of solvent (water/acetonitrile/n-propanol, 98.5/1/0.5 by volume) that contained overall concentrations of 0.2% formic acid, 0.0005% HFBA, and 0.001% Zwittergent 3-16. Samples were vortexed 5 min, sonicated 1 min, and centrifuged 5 min at 12,000g. Twenty-microliter aliquots were then diluted fivefold into autosampler vials for nLC-FTICR-MS. The reference pool samples were initially reconstituted in 200 μL of solvent, followed by a 10-fold dilution into autosampler vials.

Nanoscale Liquid Chromatography (nLC)

nLC was performed using an Eksigent NanoLC-1D (Eksigent Technologies, Livermore, CA), a binary microfluidic LC system that was used to create gradients at a total flow rate of 400 nL/min without the use of flow splitting. The nLC column was a 15 cm long by 100 μm i.d. IntegraFrit column (NewObjective, Woburn, MA) packed in-house with Targa C18, 5-μm, 300-Å particles (Higgins Analytical, Mountain View, CA). Mobile phase A was water/acetonitrile/n-propanol/formic acid (98/1/1/0.2 by volume), while mobile phase B was acetonitrile/n-propanol/water/ formic acid (80/10/10/0.2 by volume). A multistep gradient was used, starting at 0% B and holding for 3 min, with linear increases to 40% B at 40 min and 95% B at 50 min. The column was held at 95% B for 8 min before being re-equilibrated at 0% B. A 0.25-μL reversed-phase pre-column (Optimize Technologies, Oregon City, OR), packed with Michrom Magic C8, 5μm, 300 Å, was used to pre-concentrate and desalt samples within the sampling loop of a 10-port valve (VICI, Houston, TX). Twenty-five-microliter sample aliquots were loaded into a 35-μL loop in the autosampler. Contents of the loop were loaded and rinsed on the pre-column for 5 min using a 15-μL-per-min flow of water/acetonitrile/n-propanol/formic acid (98.5/1/0.5/0.2 by volume) containing 0.0005% HFBA and 0.5 mM ammonium acetate. After 5 min, the precolumn was switched in line with the nLC column, and ICR measurements were started.

nLC-ESI-FT-ICR-MS Measurements

nLC-FT-ICR-MS data were obtained on an IonSpec FT-ICR mass spectrometer (Irvine, CA) with a 9.4 Tesla actively shielded magnet, equipped with a Waters Z-spray nano-electrospray interface. Ions were accumulated for 1 sec in the hexapole region of the electrospray interface and then pulsed into the ICR cell. The ICR transient was measured at 1 MHz with 1024k data points (~1 sec transient). Total FT-ICR script time was 2.05 sec. Dual digitizers in the IonSpec interface electronics allowed a mean transient-to-transient time of 2.11 sec.

Computational Data Processing

Software for transient processing, mass calibration, THRASH, and grouping was written in C++ on Linux and Mac OS X. Each 1024k word transient was processed with a Blackman windowing function, zero-filled twice, and Fourier transformed to the frequency domain. The THRASH algorithm16 for isotope cluster detection (implemented in house largely as described by Horn et al. but with minor modifications12) was applied to each spectrum of the replicate chromatograms. THRASHing all 71 chromatograms (a total of 118,980 spectra, or 235 GB of data) took nearly 28 CPU hours per chromatogram, or ~5 d on a 16-processor Opteron 248 cluster (Western Scientific, San Diego, CA). The THRASH signal-to-noise threshold used was 4; the FOM threshold was 8. Charge states between 1 and 20 were accepted. Some filtering was performed to remove isotope clusters that were clearly artifactual (generally resulting from polymer, detergent, or mobile-phase contaminants). Specifically, masses that appeared in at least 10 scans from the first 4 min of a chromatogram (before the nLC gradient reached the column) were removed from that entire chromatogram with a 500-ppm tolerance; isotope clusters with retention time <5 min or >40 min were also removed, as were isotope clusters with charge state (z) of 1 and neutral mass (Mr) <900 Da.

Those isotope clusters that passed the filters were subjected to grouping in mass (30-ppm tolerance) and retention time (120-sec tolerance) both within and between chromatograms. The grouping algorithm is simple and is described elsewhere.12 In order to reduce false positives, at least three isotope clusters (whether at multiple charge states or within multiple spectra, or a combination) were required for detection of a given molecular species. A centroid retention time was determined for each molecular species by calculating an abundance-weighted average retention time across the chromatographic peak from each spectrum containing that molecular species. Retention-time reproducibility is expressed as the 95% confidence interval of the s mean (An external file that holds a picture, illustration, etc.
Object name is 414eqn1.jpg), where t is the critical value from

a t-distribution with (n – 1) degrees of freedom associated with an α = .05 significance level, and s is the sample standard deviation across the n replicates.

The in-silico myoglobin digest allowed up to two missed cleavages.


To study the reproducibility of chromatography in our biomarker discovery platform, replicate injections of myoglobin tryptic fragments, ultrafiltrates of human serum, and ultrafiltrates of human plasma were made on a splitless nanoLC system coupled to an FT-ICR mass spectrometer, as shown in Figure 11.. In order to understand how the presence of complex serum and plasma samples affected the reproducibility of the chromatography and ESI response, myoglobin tryptic digest reference samples were interdigitated with both individual and pooled 50-kDa molecular-weight cutoff (MWCO) filtrates of serum and plasma samples. Filtrates from four patients with advanced-stage ovarian cancer and four age-matched, post-menopausal, healthy control patients were randomized; serum and plasma from the same patient were injected sequentially. The individual plasma and serum samples were also pooled for use as “complex” quality control (QC) standards to examine variability independent of biological variations; both disease and control samples were pooled together. The known myoglobin tryptic fragments provided a set of consistently identifiable species whose retention times and abundances could be confidently tracked; they were used as “simple” QC standards. A total of 71 injections were made, 20 of which were used to identify optimal chromatography conditions; the remaining 51 injections were used to analyze reproducibility. Twenty-four of these were myoglobin “simple” QC standards, 16 were patient samples (plasma and serum from four cancer patients and four age-matched, healthy control patients), and 6 were pooled “complex” QC samples (3 plasma, 3 serum); there were 5 blanks.

Experimental design: Serum and plasma samples from four patients with advanced-stage ovarian cancer and from four age-matched, post-menopausal controls were filtered through 50-kDa MWCO filters. The individual samples were also pooled to be used as “complex” ...

The THRASH algorithm for isotope cluster detection was applied to each scan of the replicate injections, giving a monoisotopic neutral mass, retention time, and abundance for each isotope cluster; the clusters were then filtered and grouped; the details of the algorithmic data processing are given above.

In previous work,12 the splitting necessary to obtain nanoliter flow rates caused changes in column back-pressure to change column flow rate across the LC runs, introducing shifts in retention time that increased with increasing retention time. In this work, splitless LC pumps were used, which automatically regulated flow rate independent of changes in column back pressure.

Figure 2A2A shows the column pressure and total flow rate recorded during the analysis of a representative myoglobin reference sample. The pressure excursion at 5 min occurred when the pre-column was switched inline with the nLC column. The flow trace shown was obtained by summing the flows for mobile phases A and B, as acquired by the LC data system. The total flow trace demonstrates the ability of the two pressure amplifiers to achieve their requested outputs, which change through the course of the LC gradient. The LC post-analysis summary reported RMS flow precision of 2.6 nL/min for mobile phase A and 1.6 nL/min for mobile phase B relative to their set points during the gradient.

A: Column pressure (top curve) and LC total flow (bottom curve) graphs from a representative replicate nanoLC-FT-ICR MS injection of a myoglobin digest. Modulating pressure allowed flow to remain nearly constant at ~400 nL/min. The sharp spike at ~5 min ...

Retention time reproducibility is demonstrated in Figure 2B2B,, where the 95% confidence interval of the mean (95% CIM) of retention time for each tryptic fragment is plotted against retention time. Across all tryptic fragments detected, the average of the 95% CIM of retention time is ±6.3 sec, and the maximum is ±11.6 sec; when only those tryptic fragments that were found in 24 of 24 replicates are considered, the maximum 95% CIM of retention time drops to ±6.7 sec, as shown in Table 11.. A representative TIC is also shown.

Figure 3A3A depicts the pressure and total flow rate from a representative injection of pooled human serum filtrate, and shows similarly consistent flow rates. RMS flow-rate precision for this analysis was reported as 2.8 nL/min and 2.0 nL/min for mobile phases A and B, respectively. Figure 3B3B plots the 95% CIM of retention time for those species found in all three replicate injections of human serum filtrate. While the complexity of serum indeed increases the variability of the retention time as compared to the myoglobin data above, the average of the 95% CIM of retention time across all species detected in the serum is still ±8.8 sec. Ninety-five percent of 393 species detected in all three serum replicates have a 95% CIM of retention time of ±22.0 sec or less. This corresponds to a maximum variability of 22 spectra. It is important to note that the variability does not seem to depend on the total ion current, indicating that the number or abundance of species eluting does not affect the reproducibility of chromatography, at least within the limits observed here.

A: Column pressure (top curve) and LC total flow (bottom curve) graphs for a representative replicate injection of human serum. Flow, again, remains constant. B: Plot of 95% CIM of retention time (circles, right hand axis; in sec) vs retention time (in ...

Figure 4A4A shows the pressure and total flow rate from a representative injection of pooled human plasma filtrate; again, pressure monitoring resulted in uniform flow. RMS flow-rate precision for this analysis was reported as 2.8 nL/min and 1.8 nL/min for mobile phases A and B, respectively. Figure 4B4B shows a plot of the 95% CIM of retention time for those species found in all three replicate injections of human plasma; the average of the 95% CIM of retention time across all species is ±10.3 sec. Ninety-five percent of the 312 species in all three of the plasma replicates have a 95% CIM of retention time of ±18.5 sec or less, equivalent to 18 spectra.

A: Column pressure (top curve) and LC total flow (bottom curve) graphs for a representative replicate injection of human plasma. B : Plot of 95% CIM of retention time (circles , right hand axis; in sec) vs retention time (in min) for 312 species found ...

Next, we examined the effect that the complex serum and plasma had on the reproducibility of abundances; we wanted to ensure that column performance was unaffected by sample composition. Also, earlier work had shown a systematic decline in abundance across the course of 50 or more LC runs (data not shown), representing several days of instrument time, also probably due to splitting. The interdigitated myoglobin reference samples allowed us to assess these characteristics.

Figure 55 shows the abundances for 22 myoglobin fragments detected in 24 replicate injections across more than 50 total LC-MS runs. Little change is observed across the injections and over nearly 2.5 orders of magnitude of fragment abundance. For those fragments observed in 24 of 24 replicate injections, the minimum, average, and maximum coefficients of variation (CV) of abundance were 13.6%, 37.0%, and 120.5%, respectively. Sequence coverage of myoglobin, also shown in Figure 55,, is quite consistent as well, with a minimum of 70.6%, average of 88.4%, and maximum of 100%.

Abundance of 22 tryptic myoglobin fragments (lines, left hand axis; arbitrary units) identified in 24 replicate injections interdigitated in over 50 total LC-MS runs; 20 injections performed using the same trap and column in order to optimize chromatography ...


Replicate injections of tryptic myoglobin, ultrafiltrates of human serum, and ultrafiltrates of human plasma made on a splitless nanoLC system coupled to an FT-ICR mass spectrometer were used to assess analytical reproducibility, as summarized in Table 11.. The consistency in retention time (maximum 95% CIM ±6.7 sec, 7 spectra), in abundance (37% average CV), and in sequence coverage (average 88.4%) of the 19 tryptic myoglobin fragments detected in at least 3 of 24 replicate injections indicate that reproducibility is preserved in spite of the complexity of the samples that were interdigitated. The serum and plasma samples themselves showed somewhat greater variability in retention time (±22.0 sec or less for serum and ±18.5 sec or less for plasma).

The increase in reproducibility as compared to previous studies is in large part due to the elimination of flow splitting through the use of newer LC instrumentation that controls flows independent of column backpressure . While this work was done with an FT-ICR mass analyzer, similar or larger improvements in retention-time reproducibility would also be expected with other types of mass spectrometers that utilize higher sampling rates. It is also important to note that these experiments have utilized a pre-column for concentration and desalting of samples prior to injection. The presence of the pre-column could account for some of the variability observed. However, the overall analytical variability is certainly an improvement over previous work.


This work was supported by the National Institutes of Health (CA105295-1), the W.M. Keck Foundation, and the Mayo Clinic College of Medicine.


1. Vlahou A, Fountoulakis M. Proteomic approaches in the search for disease biomarkers. J Chromatogr B 2005;814:11–19.
2. Diamandis EP. Mass spectrometry as a diagnostic and a cancer biomarker discovery tool: Opportunities and potential limitations. Mol Cell Proteomics 2004;3:367–378. [PubMed]
3. Zhang Z, Bast RC Jr, Yu Y, Li J, Sokoll LJ, Rai AJ, et al. Three biomarkers identified from serum proteomic analysis for the detection of early stage ovarian cancer. Cancer Res 2004;64:5882–5890. [PubMed]
4. Ransohoff DF. Lessons from controversy: Ovarian cancer screening and serum proteomics. J NCI 2005;97(4):315–319.
5. Pasa-Tolic L, Masseon C, Barry RC, Shen Y, Smith RD. Proteomic analyses using an accurate mass and time tag strategy. Biotechniques 2004;37:621–639. [PubMed]
6. Wang W, Zhou H, Lin H, Roy S, Shaler TA, Hill LR, Norton S, Kumar P, Anderle M, Becker CH. Quantification of proteins and metabolites by mass spectrometry without isotopic labeling or spiked standards. Anal Chem 2003;75:4818–4826. [PubMed]
7. Silva JC, Denny R, Dorschel CA, Gorenstein M, Kass IJ, Li GZ, et al. Quantitative proteomic analysis by accurate mass retention time pairs. Anal Chem 2005;77(7):2187–2200. [PubMed]
8. Wall DB, Kachman MT, Gong SS, Parus SJ, Long MW, Lubman DM. Isoelectric focusing nonporous silica reversed-phase high-performance liquid chromatography/electrospray ionization time-of-flight mass spectrometry: A three-dimensional liquid-phase protein separation method as applied to the human erythroleukemia cell-line. Rapid Commun Mass Spectrom 2001;15:1649–1661. [PubMed]
9. Anderson NL, Polanski M, Pieper R, Gatlin T, Tirumalai RS, Conrads TP, et al. The human plasma proteome: A non-redundant list developed by combination of four separate sources. Mol Cell Proteomics 2004;3:311–326. [PubMed]
10. Muddiman DC. Cliby W, Bergen HR III. Cancer bio-markers: How proteomics is leading to the discovery of new markers. Clin Lab News 2003;29:12–16.
11. Bergen HR III, Vasmatzis G, Cliby WA, Johnson KL, Oberg AL, Muddiman DC. Discovery of ovarian cancer biomarkers in serum using NanoLC electrospray ionization TOF and FT-ICR mass spectrometry. Disease Markers 2004;19(4/5):239–250.
12. Johnson KL, Mason CJ, Muddiman DC, Eckel JE. Analysis of the low molecular weight fraction of serum by LC-dual ESI-FT-ICR mass spectrometry: Precision of retention time, mass, and ion abundance. Anal Chem 2004;76:5097–5103. [PubMed]
13. Wang CP, Isenhour TL. Time-warping algorithm applied to chromatographic peak matching gas chromatography/Fourier transform infrared/mass spectrometry. Anal Chem 1987;59:649–654.
14. Nielsen NP, Carstensen JM, Smedsgaard J. Aligning of single and multiple wavelength chromatographic profiles for chemometric data analysis using correlation optimized warping. J Chromatogr A 1998;805:17–35.
15. Sakoe H, Chiba S. Dynamic programming algorithm optimization for spoken word recognition. IEEE Trans Acous Speech Sig Proc 1978;26(1):43–49.
16. Horn DM, Zubarev RA, McLafferty FW. Automated reduction and interpretation of high resolution electrospray mass spectra of large molecules. J Am Soc Mass Spectrom 2000;11:320–332. [PubMed]

Articles from Journal of Biomolecular Techniques : JBT are provided here courtesy of The Association of Biomolecular Resource Facilities