Determination of Flow Rates in Capillary Liquid Chromatography

Jun 28, 2016 - ... to a Nanoelectrospray Source using Droplet Image Analysis Software ... Faculty of Computer Science, Dalhousie University, 6050 Univ...
0 downloads 0 Views 3MB Size
Download PDF
Technical Note pubs.acs.org/ac

Determination of Flow Rates in Capillary Liquid Chromatography Coupled to a Nanoelectrospray Source using Droplet Image Analysis Software Alejandro M. Cohen,*,† Axel J. Soto,‡ and James P. Fawcett†,§ †

Faculty of Medicine, Dalhousie University, 5849 University Avenue, Halifax, NS B3H 4R2 Nova Scotia, Canada Faculty of Computer Science, Dalhousie University, 6050 University Avenue, Halifax, NS B3H 4R2 Nova Scotia, Canada § Departments of Pharmacology and Surgery, Faculty of Medicine, Dalhousie University, 5849 University Avenue, Halifax, NS B3H 4R2 Nova Scotia, Canada ‡

S Supporting Information *

ABSTRACT: Liquid chromatography coupled to electrospray tandem mass spectrometry (LC−ESI-MS/MS) is widely used in proteomic and metabolomic workflows. Considerable analytical improvements have been observed when the components of LC systems are scaled down. Currently, nano-ESI is typically done at capillary LC flow rates ranging from 200 to 300 nL/min. At these flow rates, trouble shooting and leak detection of LC systems has become increasingly challenging. In this paper we present a novel proof-of-concept approach to measure flow rates at the tip of electrospray emitters when the ionization voltage is turned off. This was achieved by estimating the changes in the droplet volume over time using digital image analysis. The results are comparable with the traditional methods of measuring flow rates, with the potential advantages of being fully automatable and nondisruptive. very difficult to detect at flow rates ranging from ∼200 to 300 nL/min. At these flow rates, leaks become almost invisible; the effluents are typically hidden inside dead volumes of connectors, and evaporation rates often match those of the leak. Monitoring the pressure at the pump heads becomes one of the best indicators to determine the presence of a leak or clogged line. However, postcolumn leaks are extremely hard to detect as these do not typically affect the pressure read backs. This type of leak is more common in setups where a connector is placed between the column and the nano-ESI emitter. A leak caused by a complete blockage in the emitter produces a very clear loss of MS signal. However, partial leaks only produce a reduced MS signal, and thus are harder to diagnose. The result is a loss of analytical sensitivity that commonly goes unnoticed. Measuring the flow rate at the tip of the emitter is an effective way to determine the presence of a leak along the chromatographic line. This has traditionally been done manually by collecting the eluent at the emitter using a graduated capillary, as explained in this technical note.6 However, this process is disruptive (requires interrupting data acquisition), cumbersome, and time-intensive. To overcome these issues, we present a novel approach to measure flow rates at the tip of nano-ESI emitters by calculating the change in the droplet volume over time. The droplet volumes were calculated through the analysis of video images

L

iquid chromatography coupled to electrospray tandem mass spectrometry (LC−ESI-MS/MS) has become one of the most widely used analytical platforms in proteomics and metabolomics workflows.1,2 The advancements in both chromatographic and mass spectrometric instrumentation have allowed a reduction in the scale of the analytical components. The chromatographic parameters such as flow rates, injection volumes, media (particle and pore size), and run times have all decreased their dimensions considerably in the past few decades. Simultaneously, new ionization sources have been designed to adapt to the miniaturization of the chromatographic conditions. Capillary LC coupled to nanoESI-MS/MS has shown to have some significant advantages (e.g., sensitivity) over the traditional higher flow rate setups.3,4 For many of these reasons, capillary LC nano-ESI-MS/MS has become the analytical platform of choice in many laboratories. The shift toward nano-ESI-MS/MS has nevertheless been gradual, mostly due to the challenges and technical expertise required to operate these sophisticated systems. A detailed practical guide published by Noga et al.5 clearly describes many of the issues encountered by laboratories adopting nano-LC− MS/MS technologies. Initially, most laboratories implemented restriction valves to produce microliter per minute outputs from chromatography pumps. The advent of capillary LC systems capable of delivering nanoliter per minute flow rates simplified this process, however, at a considerable cost. Despite of the advances in instrument hardware and software, maintenance and troubleshooting of capillary LC systems remains challenging. Leaks and clogged lines become © 2016 American Chemical Society

Received: April 18, 2016 Accepted: June 28, 2016 Published: June 28, 2016 7476

DOI: 10.1021/acs.analchem.6b01523 Anal. Chem. 2016, 88, 7476−7480

Technical Note

Analytical Chemistry

rates works as follows: frames are extracted at a rate of one 640 × 480 resolution grayscale image per second using the FFMpeg framework.8 We developed a script in Python to analyze the images extracted by FFMpeg. The algorithm scans the image from right to left aiming to detect the upper and bottom boundaries of the drop. This is achieved by identifying a clear contrast in the pixel intensities of the background and those of the capillary and the drop. In order to distinguish the capillary from the drop, a “roundness” constraint in the boundary identification is imposed; i.e., if the boundary aligns to a straight line, the algorithm identifies this as part of the capillary. The flow can be computed by calculating volume difference of the drop over consecutive frames. The diameter of the capillary is known and can be used as a reference to compute the scale of millimeters per pixel. The volume can be measured by identifying the lengths of the major and minor axes assumed in an oblate ellipsoid, where the occluded axis is estimated to be equal to the minor axis. The code of this software can be conveniently downloaded from: https://github.com/axelsoto/Droplet-Image-Processing.

captured by the camera installed on the nano-ESI source. This approach produced results comparable to those obtained by the traditional manual approach using graduated capillaries. Unlike the traditional manual approach, the method we developed is automatable, nondisruptive, and produces results within seconds. Furthermore, it could also be implemented into existing LC−MS/MS acquisition programs to enhance troubleshooting capabilities, by measuring flow rate in real time at the tip of the nano-ESI emitters.

■

MATERIALS AND METHODS LC−MS/MS. A capillary LC system (Ultimate 3000rslc, Thermo Fisher Scientific, U.S.A.) was connected to a monolithic C18 column (Onyx C18, Phenomenex, Torrance, CA, U.S.A.) using fused-silica capillary tubing. The column was plumbed to a pulled silica emitter (PicoTip Emitter SilicaTip, FS360-20-10-N20-10.5CT, New Objective, Woburn, MA, U.S.A.) via a micro tee liquid junction. A nanospray ion source (NSI-2, Thermo Fisher Scientific, U.S.A.) equipped with a charge-coupled device (CCD) camera (CV-A55 IR, JAI, Japan) was used to acquire video images (Osprey video capture card). The camera uses the ICX429ALL 1/2 in. sensor (Sony, JP) to achieve a high signal-to-noise ratio and records monochrome video at a maximum of 25 frames/s. The camera is located perpendicular to and directly above the capillary. It is integrated to the nanospray ion source to visualize the spray at the tip of the emitter when a high voltage is applied. The nanospray source is factory-equipped with a red LED light to enhance the spray image. A white LED light source was used to improve the image contrast of the droplets formed during video acquisition. The LC was set to deliver 3% acetonitrile with 0.1% formic acid at 100, 200, 300, and 400 nL/min nominal flow rates. For experimental purposes, we refer to “nominal flow rate” as the input values introduced in the software that controls the LC system. These flow rates are tightly controlled through pressure transducers that deliver the preset flow rate values. In contrast, we refer to measured (or empirical) flow rates, as those measured at the tip of the nano-ESI emitter. The flow rate accuracy and precision are not disclosed in the users’ manuals. The ionization voltage was set to 0 V throughout the experiment to allow the formation of the droplets. Manual Flow Rate Measurements. Manual flow rates were measured following the technical note from New Objective.6 Briefly, the LC system was sequentially set to deliver nominal flow rates at 100, 200, 300, and 400 nL/min. At each flow rate, the volume of the droplet was measured using a graduated capillary (microcaps disposable pipets, Drummond, PA, U.S.A.) at 2 and 4 min intervals. Each calibrated capillary holds 2 μL and is 64 mm long (31.25 nL/min). The total volume collected is calculated based on the length of liquid contained in the capillary. Digital Image Analysis with ImageJ. Video recordings of the droplets formed at each flow rate were exported as .avi files to ImageJ software for image analysis.7 The video footage was split into still images (frames) at 1 s intervals. The size of each droplet was manually estimated by fitting the perimeter of each droplet using the elliptical selection tool. The lengths in pixels of the major and minor axes were measured at each frame. Droplet Image Analysis Software (DIAS). The same video files described above were used to develop a computer program for automated postacquisition image analysis. We named this approach DIAS, for “droplet image analysis software”. The computational process for measuring the flow

■

RESULTS AND DISCUSSION The size and shape of droplets formed in contact with fibers has been studied extensively.9,10 Droplets formed on the end of a capillary tube adopt the shape that resembles a prolate ellipsoid. The video images captured at the tip of nanoelectrospray emitters (when the ionization voltage is turned off) could hypothetically be used to calculate the size of the droplets. The volume of the droplet can be estimated based on the following equation (eq 1): vol =

4 ⎛ a ⎞⎛ b ⎞⎛ c ⎞ π ⎜ ⎟⎜ ⎟⎜ ⎟ 3 ⎝ 2 ⎠⎝ 2 ⎠⎝ 2 ⎠

(1)

where a and b correspond to the major and minor axis of the droplet’s cross section, respectively. Assuming b ≈ c (being c the droplet’s axis which lies parallel to the camera lens’ principal axis), the flow rate could be then calculated (eq 2) by estimating the change in the droplet volume at two (or more) time points. flow rate =

Δvol Δt

(2)

. To test this hypothesis, video recordings of droplets formed at 100, 200, 300, and 400 nL/min (nominal flow rates) were acquired and stored for image analysis. For comparison purposes, manual flow rate measurements were also obtained under identical conditions. A summary of the experimental setup described in this paper is shown in Figure 1. Initially, manual flow rates were empirically measured using a graduated capillary and timer, as commonly described in the literature.6 The ionization voltage was set at 0 V to allow the formation of the droplet at the end of the nanospray emitter. Flow rates were determined at 120 and 240 s intervals for each nominal flow rate setting. The results are shown in Figure S-1. Interestingly, the values obtained both at 120 and 240 s were significantly lower than the nominal flow rate produced by the LC pumps. When a linear regression is applied to the experimental data, the extrapolated (y-intercept) flow rates were closer to the nominal flow rate values. The discrepancies in the measured values could partially and presumably be caused by evaporation. To estimate the evaporation rate under 7477

DOI: 10.1021/acs.analchem.6b01523 Anal. Chem. 2016, 88, 7476−7480

Technical Note

Analytical Chemistry

Furthermore, this correction factor will certainly be influenced by the temperature and atmospheric pressure conditions in different laboratories. Once the manual flow rates were established as reference values, we proceeded to acquire video images of the droplets formed on the nano-ESI emitter tips. As a proof of concept using the digital image approach, the video footage was first studied using ImageJ. The videos for each nominal flow rate were split into 1 s frames. Each frame was then analyzed by fitting the contour of each droplet using the ellipsoid selection tool (Figure 2A). The image pixels were calibrated to distance (in micrometers) using the capillary’s outer diameter (Figure 2B) as specified by the vendor. Figure S-2A shows the flow rates obtained for the 100 nL/ min video recording. The dispersion of the data is likely generated by the errors introduced while manually and visually fitting the images at 1 s intervals. An interesting result is observed when the time intervals used for image analysis are increased to 2, 4, 8, 16, and 32 s (Figure S-2B−F). The lower dispersion of the data points clearly shows that measurement errors introduced in fitting the droplet are reduced as the changes in droplet size increase. Another striking observation is the clear drop in measured flow rate (negative slope) as the droplet increases its size. This seems counterintuitive, as the LC pump technically delivers a constant nominal flow rate throughout the experiment. We speculate that the decrease in

Figure 1. Experimental setup summary.

our experimental conditions, 1 μL of mobile phase (3% acetonitrile with 0.1% formic acid) was deposited on the tip of the nano-ESI emitter. The droplet evaporated within 9 min and 40 s, equivalent to an evaporation rate of approximately 103 nL/min at room temperature (laboratory thermostat set at 22 °C). Regardless, we are cautious on how to apply this correction factor, as the evaporation rate might differ between a static droplet and that of a dynamic droplet “in-formation”.

Figure 2. (A) Image analysis of droplet video footage using ImageJ software. The video files were separated into 1 s interval frames. The contours of the droplets were manually fitted using ImageJ’s ellipsoid tool. The major and minor axes of the drop cross section were measured in pixel length. (B) The lengths measured in pixels were calibrated using the outer diameter (360 μm) of the capillary (0.102 pixel/μm). (C) Results obtained by DIAS at 100 nL/min of nominal flow rate. The images correspond to the frames extracted at 2, 13, and 23 s. The left column shows raw images, obtained from the video recordings. The right column shows the boundaries of the droplet determined by DIAS superimposed on the same images. 7478

DOI: 10.1021/acs.analchem.6b01523 Anal. Chem. 2016, 88, 7476−7480

Technical Note

Analytical Chemistry

Figure 3. Boxplot comparing the data collected for the different flow rate analysis methods. Manual collection of droplets using graduated capillaries at 2 and 4 min are labeled as M2 and M4, respectively. These results correspond to four technical replicates. Digital image analysis using ImageJ is labeled as iJ. The results for both iJ and DIAS are represented by flow rate data points obtained at 2 s intervals for approximately 60 s. The asterisk (*) symbol represents the outliers. These outliers are observations that are at least 1.5 times the interquartile range (Q3−Q1) from the edge of the box.

Figure 4. Integration of DIAS into LC−MS/MS acquisition software.

measured flow rate is likely due to an error introduced in the assumption of the droplet shape and dimensions. As the drop grows in size, the weight of the drop distorts its shape. Consequently, the length of the axis occluded to the camera (c axis in eq 1) is underestimated. Figure S-3 shows the overlap of the six graphs depicted in Figure S-2. In addition to the negative slope, this graph also shows a subtle difference in each of the regression lines. We speculate that these differences may likely be caused by evaporation: flow rates measured at longer time intervals would be more susceptible to evaporation. Regardless, all the extrapolated regression lines (y-intercept) intersect approximately at t = 0 s, and as such we believe this to be the most appropriate estimator of the measured flow rates. This value

should minimize the effects caused by both evaporation and distortion of the droplet shape. We next set out to automate the image analysis processing and reduce the errors introduced by manual estimation of the drop size. Here we developed Droplet Image Analysis Software (DIAS), which automatically identifies the boundaries and axes of the droplet. These parameters are used to calculate the measured flow rates based on the droplet size at each time frame. The same video files analyzed by ImageJ manually were processed using DIAS. Figure 2C illustrates the ability of DIAS to accurately identify the droplet boundaries. An external white LED light was used to improve the contrast of the images. Figure 3 shows output generated by DIAS compared to the results obtained by ImageJ and the traditional manual methods at 100, 200, 300, and 400 nL/min nominal flow rates. The 7479

DOI: 10.1021/acs.analchem.6b01523 Anal. Chem. 2016, 88, 7476−7480

Analytical Chemistry

■

average flow rates measured manually using calibrated capillaries (at both 2 and 4 min readings) were considerably lower than the nominal flow rates. The results from the digital image analyses (both ImageJ and DIAS data) are represented by individual flow rate data points obtained at 2 s intervals for approximately 60 s (considerably lower than the 2 and 4 min required for the manual method). Interestingly, the results obtained by ImageJ consistently overestimated the nominal flow rates. The average data points obtained by DIAS were 58.6, 204.5, 312.3, and 405.2 for nominal flow rates of 100, 200, 300, and 400 nL/min, respectively. All the data (averaged measured flow rates and CV%) obtained from the different methods is summarized in Table S-1. At very low flow rates (∼100 nL/min), the measured flow rates underestimate the nominal values, regardless of the method used. At flow rates typically used in capillary LC−nano-ESI-MS/MS (between 200 and 400 nL/min), DIAS outperformed all other approaches. Although the CV% in DIAS is clearly higher than the manual method, the reported averages were consistently closer to the nominal values.

■

Technical Note

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.6b01523. Manual determination of flow rates (Figure S-1), flow rate measured by image analysis (Figure S-2), superimposed data and regression lines (Figure S-3), and experimental average flow rates and CV% (Table S-1) (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: 902-494-8359. Fax: 902-494-1394. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This project was funded in part by an NSERC Discovery Grant (J.P.F.), and equipment was purchased through a Canadian Foundation for Innovation and Nova Scotia Research and Innovation Trust program (J.P.F.). J.P.F. is a Tier 2 CRC in Brain Repair.

CONCLUSIONS

Postcolumn leaks in nano-LC−ESI-MS/MS often proceed unnoticed and are difficult to diagnose. These leaks result in a major loss of analytical sensitivity, as part of the LC eluent is diverted away from the mass spectrometer. Here we have established a way to determine the flow rate at the tip of the nanospray emitters as a means to confirm the presence of such leaks. Manual determination of flow rates using calibrated capillaries is time-consuming and cumbersome. Furthermore, it interrupts data acquisition and often requires the ion source to be removed and disconnected from the inlet of the mass spectrometer. In this paper we present DIAS, a computational approach by which the flow rate can be measured by analyzing the image of a droplet formed at the tip of the ESI emitter. According to our experiments, this approach is as accurate as manual methods for measuring flow rates by means of graduated capillaries. DIAS has the potential to be run in parallel (or integrated into) to LC−MS/MS data acquisition software (Figure 4). This setup would allow a quick (under 5 s) estimation of the flow rate in real time without human intervention. Ideally, this measurement could take place prior to each LC injection by disabling the ESI voltage for a short period. An automated decision algorithm could be integrated into the acquisition batch to determine whether to proceed or abort the run, avoiding the loss of valuable sample material if the flow rate does not meet an expected value. The estimation of the flow rate by digital analysis would benefit from a second video camera placed orthogonally to the first one so that it can capture the dimension lost by using a single camera. This would allow a better assessment of the droplet shape, thus improving the accuracy of the flow rate. Furthermore, increased image resolution of the CCD camera would improve the estimation of the axes’ lengths and pixel-todistance calibration of the image. By addressing these limitations, DIAS could become a standard feature in future LC−MS/MS acquisition programs.

■

REFERENCES

(1) Smith, R. D.; Shen, Y.; Tang, K. Acc. Chem. Res. 2004, 37 (4), 269−278. (2) Manisali, I.; Chen, D. D. Y.; Schneider, B. B. TrAC, Trends Anal. Chem. 2006, 25 (3), 243−256. (3) McCormack, A. L.; Schieltz, D. M.; Goode, B.; Yang, S.; Barnes, G.; Drubin, D.; Yates, J. R. Anal. Chem. 1997, 69 (4), 767−776. (4) Shen, Y.; Zhao, R.; Berger, S. J.; Anderson, G. A.; Rodriguez, N.; Smith, R. D. Anal. Chem. 2002, 74 (16), 4235−4249. (5) Noga, M.; Sucharski, F.; Suder, P.; Silberring, J. J. Sep. Sci. 2007, 30 (14), 2179−2189. (6) Technical Note PT-7, 2005. New Objective Incorporated Web site. http://www.newobjective.com/downloads/technotes/PT-7.pdf. (7) Schneider, C. A.; Rasband, W. S.; Eliceiri, K. W. Nat. Methods 2012, 9 (7), 671−675. (8) Tomar, S. Linux J. 2006, 2006 (146), 10. (9) Wu, X.-F.; Dzenis, Y. a. Acta Mech. 2006, 185 (3−4), 215−225. (10) Du, J.; Michielsen, S.; Lee, H. J. Langmuir 2010, 26 (20), 16000−16004.

7480

DOI: 10.1021/acs.analchem.6b01523 Anal. Chem. 2016, 88, 7476−7480