J. Sep. Sci. 2007, 30, 2179 – 2189
M. Noga et al.
2179
Marek Noga1 Filip Sucharski1 Piotr Suder1 Jerzy Silberring1, 2
Review
1
Capillary LC is one of the most powerful analytical tools available for separation scientists. Its unique analytical properties are associated with numerous technical issues that may cause operation of such systems to be somehow troublesome. Because of that, a good experience in capillary LC troubleshooting is required to keep the system in shape and, in effect, to obtain reliable results. In this paper, we summarize the most important issues of the capillary systems, including void and dead volumes, leakages, sample injection, and a multidimensional LC approach. The aim of this paper was to provide practical advise on system diagnosis, and to present solutions to problems discussed. Also, several exemplary nano-LC separations are included to demonstrate some typical problems encountered in our daily work.
Department of Neurobiochemistry, Faculty of Chemistry and Regional Laboratory, Jagiellonian University, Krakow, Poland 2 Centre for Polymer Chemistry, Polish Academy of Sciences, Zabrze, Poland
A practical guide to nano-LC troubleshooting
Keywords: Capillary chromatography / Column / Proteomics / Separation / Troubleshooting / Received: May 23, 2007; revised: July 2, 2007; accepted: July 4, 2007 DOI 10.1002/jssc.200700225
1 Introduction Principles of the electrospray process were described by Dole et al. in 1968 [1] who proposed the theoretical basis of the behavior of charged and evaporated droplets in a strong electric field. Their work inspired Fenn to design a modern electrospray ion source. Fenn [2] presented his ideas at the Annual Conference of American Society of MS in San Francisco in 1988. From that moment, electrospray ion sources started to be widely used in analytical chemistry and life sciences, especially in medicine, biochemistry, and biotechnology. For his invention, Fenn was awarded a Nobel Prize in chemistry in 2002. Due to its capabilities, ESI is seen as one of the most versatile techniques in MS, allowing for analyses of compounds dissolved in liquids. Thus, the method is ideal for the measurements of biological samples, because they can be introduced, as they tend to occur in the nature. In contrast to most ion sources, ESI effectively ionizes high molecular mass molecules, and works under atmospheric pressure, which allows for direct introduction of liquids into the ionization area. This property is especially appreciated by the scientists working with LC and other separation techniques. Before the “mass spectrometry era”, detectors applied in HPLC instruments were only able to detect and partially characterize samples eluted from the chromatographic column, providing the very limited possibilities for identification. Typical detecCorrespondence: Professor Jerzy Silberring, Department of Neurobiochemistry, Faculty of Chemistry, Jagiellonian University, Ingardena St. 3, 30-060 Krakow, Poland E-mail:
[email protected] Fax: +48-12-6340515
i
2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
tors, such as those based on absorbance, conductivity, thermal conductivity, or fluorescence of eluting molecules, were sufficient when the separated sample contained ca. 10 – 30 components. Based on, e. g., absorbance profiles and retention times of standards, the identification of the eluted compounds was possible. Unfortunately, in the case of samples containing hundreds or even thousands of compounds (e. g., “proteomic samples”), detection with the use of conventional detectors became quite useless. Mass spectrometer equipped with an electrospray ion source appeared as a detector of choice, due to its analytical capabilities. In contrast to any other detector, mass spectrometer provides information, which can unambiguously identify a molecule eluted from the LC column, including its molecular mass and structural characterization (e. g., amino acid sequence). Moreover, it is indifferent for the mass spectrometer whether the molecule leaves a column as a separate peak or coelutes with another compounds. In such case, all coeluting molecules may be identified simultaneously. About a decade ago, the high-pressure liquid chromatograph linked to the ESI-equipped mass spectrometer was a top-trendy analytical instrument, desirable in almost every laboratory involved in separation sciences. Typical LC-MS systems, equipped with analytical or microbore columns, were almost perfect instruments. They were astonishingly sensitive, as compared to other detectors, and provided a far more specific information. In the middle of the last decade of the 20th century, Wilm and Mann described a revolutionary modification of a typical electrospray ion source. In their first paper, they called the finding as microelectrospray [3]. Their www.jss-journal.com
2180
M. Noga et al.
idea was to apply a very narrow, fused-silica capillary as an ESI emitter, instead of the routinely used, relatively wide, steel or fused-silica capillaries. Reduction of the internal diameter of the tip from about 100 lm to 1 – 20 lm led to a decrease in the initial size of liquid droplets, smaller sample consumption, and an increase of the ionization yield. As a result, much higher sensitivity of the system could be achieved. Almost simultaneously, in 1994 Caprioli’s group published an article describing the capabilities of a modified electrospray ion source, where the fused-silica capillary filled with the RP material worked as an emitter needle [4]. The next article from the same group showed sensitivity of the nanospray, where the authors demonstrated a successful analysis of a peptide at a concentration of about 500 zmol/lL [5]. Data presented by both groups clearly showed capabilities of this new technique as one of the most powerful analytical tools, combining extremely high sensitivity with low sample consumption. Other improvements made by Davis et al. [6] and Gatlin et al. [7] were based on the integration of column packing with the emitter tip, which resulted in a system, free from the after-column void volumes. Reduction of the void volumes to an absolute minimum plays an extremely important role in micro- and nanoscale LC. Design of the miniaturized interface resulted in a decrease in the size of the Taylor cone, which has few consequences. First of all, the size of the droplets that are produced during ionization, is significantly decreased. Moreover, the smaller the size of a Taylor cone, the smaller is the ion emission region, which results in the production of droplets with a smaller volume-to-surface ratio. Under such conditions, the vaporization process advances so fast that an application of the drying gas is not necessary. Furthermore, the properly shaped Taylor cone is created at a lower electric potential allowing for closer positioning of the emitter tip toward the hot capillary inlet, thus providing good signal stability. A major obstacle was that nanoelectrospray could not be connected to the typical, analytical HPLC system because of a much higher flow rate of the mobile phase. Application of the postcolumn flow splitters was helpful but appeared to be useless as it strongly affected sensitivity. The only practical solution to the problems associated with linking the LC to the nanospray-MS was to design capillary, high performance chromatography. That was the way to receive one of the most sensitive and versatile analytical instruments based on the hyphenated techniques: capillary LC-nano-ESI-MS instrument. With such an advanced technique, the scale of all problems associated with the use of such system may be astonishingly high, and the payoffs for an effective application of the capillary chromatographs linked online to the mass spectrometers equipped with the nano-ESI ion sources will be explained below.
i
2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
J. Sep. Sci. 2007, 30, 2179 – 2189
2 General characteristics of a typical nanoLC system In theory, everything is simple. It is sufficient to purchase the capillary HPLC system, turn it on, connect to a nano-ESI-MS instrument, inject a sample, and wait for exciting data. But between the sample and good final results, there is an analytical instrument able to lead almost every scientist to fury. The major issue is to understand the difference between analytical HPLC and the capillary setup, though the general principles of their operation are identical (column, pumps, gradient, solvents, etc.). We will make an attempt to briefly clarify the most important aspects of this, sometimes troublesome, system.
2.1 Typical columns, flow rates, analysis times In the nano-LC system, everything is much smaller than in conventional, analytical chromatographs. Typical analytical instruments usually provide flow rates from 1 lL/ min to 10 mL/min when a single (isocratic) pump is used. For the gradient runs, analysis is more dependent on our system; we may use two or more pumps to form a desired gradient. Flow rates usually start from 50 lL/min. A typical analytical RP column (4.6 mm id, 25 cm long) is eluted at a flow rate between 0.5 and 1 mL/min. In the case of the capillary LC, chromatographic columns are very narrow. The standard diameters offered by a majority of manufacturers are: 75 – 500 lm id. The most commonly used are 75 lm id. Columns are made of the fused-silica or carbon, titanium, or stainless steel tubings. The last three materials are used for the 150 lm and higher diameter items. The 75 lm columns are manufactured exclusively of fused-silica. The stationary phase is usually prepared from the best achievable beads of diameters 5 lm or lower. Alternatively, monolithic columns are fabricated. In contrast to the analytical (and microbore) columns, terminators are prepared as porous filling, polymerized in the tubing outlet. Such sophisticated construction may easily be damaged by the few, most commonly occurring events: (i) The presence of even submicroscopic solids in a sample may effectively clog the column. Clogging may occur in every segment of the column – starting from the top of the stationary phase down to the terminator, which due to the very small pores in its structure is particularly exposed. Column, after clogging, is permanently damaged as there is no possibility to backflush it. This is due to the fact that the stationary phase is not protected (no frit) from the inlet side. In contrast to standard columns, where backflushing or frit replacement is a routine, such manipulations usually lead to the loss of the stationary phase. www.jss-journal.com
J. Sep. Sci. 2007, 30, 2179 – 2189
(ii) Too high flow rate and, as an effect, too high pressure on the column often leads to the terminator expulsion, which destroys a column. Monolithic columns are more resistant than the packed ones as they do not contain any frits (stationary phase is covalently bound to the capillary walls), and the overall pressure is significantly lower. Neither capillary nor standard columns accept sudden pressure changes but, according to the flow rates applied and the possibility of their regulation, capillary columns seems to be more exposed for some adverse effects of fast changes in the eluent pressure. (iii) The common problem with small diameters is column overloading. In fact, it does not lead to a permanent damage but, when the sample is in excess, it would be necessary to perform few “blank” runs to remove the remaining material. This is a very annoying process because of a very long analysis time (low flow rates). The problems mentioned above are also typical for the standard, analytical HPLC systems. But in the case of capillary chromatography, the scale of the problem is different and, even more important, the margin of tolerance for improper handling is much smaller. It is important to admit that operation of the capillary LC demands a patient scientist. In a typical LC system, time of analysis varies between 5 and 30 – 40 min. Very short columns (1 cm or slightly longer) may give reasonable separation even within less than 5 min. In capillary chromatography, nothing is going so fast. Because of the time necessary for sample introduction, separation, and column equilibration using low flow rates, a typical separation usually lasts from 40 to 90 min. In general, for the fully automated instruments, it is possible to perform up to 24 separations a day. But including blank run after each separation, which is highly recommended, only 12 samples per 24 h may be analyzed. Table 1 summarizes typical sizes of chromatographic columns and their corresponding flow rates.
2.2 Split versus splitless systems – construction and maintenance The older technique to create low flow rates delivery to a capillary column utilizes splitters as the devices dividing eluent flow. In the capillary LC, typical HPLC pumps can also be used. Just after the pump(s), the splitter is mounted. This device divides the flow into two streams. The major part of the flow (ca. 90%) is directed to the detector or to the waste, and a small percent is introduced into a capillary column. The splitter is usually made of a few meters long capillary of a diameter between 20 and 50 lm, the length of which should be estimated empirically. This solution has one advantage and many drawbacks. The advantage is a lower cost of the typical HPLC instrument. The entire system may be dismantled and
i
2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Liquid Chromatography
2181
Table 1. Typical id versus flow rate and application for LC columns Column ID
Common name
Typical flow rate
75 lm 150 lm, 250 lm 500 lm 0.75, 0.8 mm 1.1 mm, 2.1 mm 4.6 mm 8 – 20 mm
Capillary Capillary Capillary/narrowbore Narrowbore/microbore Microanalytical Analytical Semi-preparative
100 – 300 nL/min 300 – 500 nL/min 0.5 – 10 lL/min 10 – 100 lL/min 100 – 500 (1000) lL/min 1 – 2.5 mL/min A2 mL/min
inspected when something goes wrong. A number of disadvantages is listed below: (i) Reproducibility of the retention times (and, thus chromatograms) is slightly worse in comparison to the systems without splitters, due to the possible pressure fluctuations, especially during gradient runs. (ii) Increasing concentration of the organic solvent (lower viscosity as compared to water) during gradient run causes a decrease in the internal pressure and, thus, a decrease in the flow rate. (iii) More than 99% of an ultrapure, expensive eluent goes to the waste. (iv) System has too many internal connections, which may cause leakage, void volumes, and other maintenance problems. (v) System is not designed for rapid changes of the flow rates or other separation parameters. That is why every change in the gradient program takes at least a few minutes. (vi) Calculation of the flow rate is calibrated using the most typical eluents. Application of a rarely used eluent, viscosity of which significantly differs from the average, leads to an improper flow rate introduced into the column. This may lead to a disturbed or unsuccessful separation or even column damage. During recent years, few manufacturers introduced splitless nano-LC instruments. Apart from their costs, it seems that almost all disadvantages of splitters were eliminated. The newly designed instruments use the socalled microfluidic flow control where electronic devices monitor, with high frequency and accuracy, the actual backpressure in the capillary column and, relying on these data, adjust necessary parameters. Additionally, flow rates from every source are monitored and, if necessary, corrected to receive optimal eluents delivery to receive a desired gradient. The pumping device uses electroosmotic force instead of standard pistons to evoke liquid flow. This strategy allows to achieve separation parameters unattainable by conventional systems. Flow rate may be precisely set, starting from 1 nL/min, gradient formation is stable and reproducible, even from 20 nL/min. Pressures are comparable to the conventional analytical syswww.jss-journal.com
2182
M. Noga et al.
tems, and are dependent on microfluidic cells used for the pumps construction. Last but not least, the system's response is very fast.
3 Troubleshooting 3.1 Void volumes Low flow rates used in the capillary LC make everything in a much slower motion than in the case of standard, analytical HPLC. Even the smallest tubing and connection must be filled with the solvent. Void volume, representing the total volume of the system starting from the gradient former up to the detector, determines the time that needs to pass until the system does respond. Not only in the chromatograph but in capillary LC the crucial rule is to minimize all volumes in the entire system. That is why every tubing applied after split or in the splitless instruments is made of the fused-silica or PEEKsil of an internal diameter of 50 lm or, in the case of 75 lm id column, even 20 lm. This allows for the reduction of void volumes to an absolute minimum. Also, application of the zero-dead volume fittings and/or Teflon sleeves reduces the problem. Why is it so critical? Let us assume a tubing which connects the column outlet with the nanospray needle and has a length of 200 mm, and an internal diameter of 100 lm. Its internal volume is equal to 1.57 mm3 (i. e., 1.57 lL). With the flow rate of 200 nL/min applied, this tubing will be filled with the fluid after ca. 8 min. After this time, compounds we would like to separate on the column will diffuse in the capillary and mix together. Under similar conditions, a tubing of 20 lm id and identical length has an internal volume of 0.0628 mm3 (i. e., 63 nL), and this volume will be filled by the eluent in 19 s. Usually, the total void volume of the capillary system, starting from the gradient mixer up to the detector, is about 3 lL. This volume would not even be noticeable in the case of an analytical HPLC system. But in case of the capillary LC operating at a flow rate of 200 nL/min, such void volume causes a 15 min gradient delay. It should be noted that every improper connection between capillaries evokes huge chambers of dead volumes where the separated compounds are mixed together. Therefore, each connection should be carefully inspected before the separation starts. It is a good practice to perform such inspections every few hours of the analysis.
3.2 Leakages Leakages consist of a yet another problem for the operators working with capillary LC-MS. Leakages may appear at every place and at every moment. As the flow rates are
i
2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
J. Sep. Sci. 2007, 30, 2179 – 2189
very low and the ambient temperature in the LC-MS laboratory is ca. 238C, those tiny droplets can evaporate even before they are spotted. Such an “invisible” leakage is very difficult to detect, and only an experienced operator can properly diagnose the problem. The most annoying is that the system seems to work properly and an important sample has already been injected. As mentioned above, a nanoflow below 200 nL/min makes all possible leakages in the system very difficult to localize. Within this range, the long analysis time may pass before the leak will cause even a tiny droplet to emerge, making leakage sensors quite useless. Without careful examination of all connections prior to analysis, it may happen that the sample will be lost before the operator will even mark the problem. When the leakage is suspected (it should always be!), the first, routine step is examination of the column pressure. If the leakage occurs before the column, it is very likely that mobile phase will leave the system, through the connector and the column pressure will drop significantly. Also, after few minutes, a small droplet of liquid will emerge around the connection. This kind of leakage is simplest to find, as no liquid is delivered into the ion source, and thus no signal in the mass spectrometer is observed, not even the background noise. The effect is so pronounced because the leakage occurs before the column, which mostly contributes to the high pressure of the system. Leakages that occur between the column and the ion source are definitely more difficult to identify. First of all, in most cases, the column pressure remains similar to the standard pressure. Moreover, the leak is often only partial. While part of the mobile phase still reaches the ion source, a portion of the liquid leaks out. Hence, the high pressure is still maintained and the leakage does not cause any drop in its level. The user may experience sensitivity loss and/or unstable operation of the ion source caused by the smaller amount of liquid forming the spray. The above problems are very common in the case of the frequently used low-pressure connections with two capillaries linked by a Teflon sleeve. Despite the important advantages of these sleeves (as they do not introduce any additional dead volume into the system, and are fast to prepare), such connections show a tendency to leak. The inner surface of the Teflon tubing tends to be damaged during frequent insertions and removals of the capillaries. Therefore, Teflon tubings should often be replaced. Also, inserting a capillary of a slightly different outer diameter may cause malfunctioning of the connection. If the inserted capillary is too wide, connection may be troublesome and will eventually widen the soft tubing (scraped Teflon pieces may clog the tubing). If the inserted capillary is too narrow, the tubing may be too wide to provide connection tight enough. Teflon sleeves www.jss-journal.com
J. Sep. Sci. 2007, 30, 2179 – 2189
can only be used at the low-pressure side, i. e., after the column. They do not withstand high pressures and the system will not function properly. We should also note that capillaries with the same nominal id obtained from various companies may have different outside diameters (od). A special care should be taken when replacing the capillary with the other obtained from another manufacturer. For example, inserting 280 lm od (real size) capillary in the tubing, where previously 285 lm od (real size) capillary was installed, will most probably cause leakage. Despite small differences in od, capillaries can still be connected together using teflon tubing, provided some precautions will be taken: (i) Internal diameter of the Teflon sleeve needs to be adjusted to fit the narrower of the two capillaries, (ii) Installation of the wider of the two capillaries will require more strength but care should be taken not to break the capillary. (iii) Inserting a too wide capillary may scrape small layer of teflon from the internal surface, resulting in a clog. It should be removed by pushing a narrow metal wire through the tubing. (iv) Tubing should not be reversed and should be used only once. Researchers commonly connect 280 and 285 lm id capillaries using a 250 lm id Teflon tubing. Respecting the rules mentioned above, it is possible to obtain safe, leak-free connections. Similar to other connections described above, a leakage from the Teflon sleeve appears as a tiny droplet formed at one of the ends of the Teflon tubing. Due to the low flow rates, such droplets became evident only after several minutes (or after much longer time). They might be spotted earlier using a dry paper napkin. The unfolded edge of the towel should be placed, so its edge will touch the connection between the capillary and tubing. If this connection leaks, the edge of the paper towel will soak with the mobile phase. Leakages in the low-pressure region, between the column outlet and a nanoelectrospray ion source are most often caused by clogging of the nanoelectrospray needle or one of the connections between the leaking one and the spray needle. It is also important to note that if the liquid path is blocked, the pressure in the system rises, and Teflon tubing connectors are not resistant to high pressures (see above). Even the best connections of this kind will leak under such circumstances. Before replacing the connection the operator is, therefore, advised to diagnose the reason of the leakage. In some cases, the blockage of the nanoelectrospray emitter causes a slight increase in the column pressure, especially when the Teflon sleeve connection is tight. Removal of the clogged connection should cause the pressure to return to the typical values. Because of the
i
2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Liquid Chromatography
2183
low flow rates in the system, the pressure changes very slowly, and it may take few minutes to reach the new value. Patience is recommended again, as in all other cases.
3.3 Flow estimation It is important to know the actual flow rate in the system. Flow rate estimation is the most important diagnostic tool for the systems utilizing splitters, and such a procedure should be performed occasionally during daily operation. The procedure is always recommended when any unexpected system behavior is observed (i. e., significant delay in retention times, unstable gradient, unstable spray operation, etc.). Modern commercial nano-LC systems are equipped with the digital flow rate sensors which can be used for accurate measurements, but in the case of older equipment and in-house built nano-LC instruments, the more straightforward approaches must be utilized. The measurements will not be as accurate as those performed with specific devices, but in most cases will provide accurate enough information to localize the problem. The simplest way to measure the flow rate is to collect the liquid from the system outlet for a given time and measure its volume using a calibrated micropipette or a Hamilton syringe. In order to do this, it is convenient to use a bare capillary outlet. Just before start of the time counting, the capillary end should be wiped using a paper napkin to make sure that no additional liquid is present. Then, the capillary end should be gently pressed against the bottom of a plastic tube for a given period of time. A somewhat more accurate way is the application of the Hamilton syringe. A connecting union is installed at the end of the capillary and the other end of this union holds the needle of the Hamilton syringe. As the mobile phase fills in the syringe, one can measure the volume after a certain time interval. The piston should be removed from the syringe before the operation.
3.4 Dead volumes Dead volumes are the “spaces” in the system that are not swept by the mobile phase. They can be compared with mixing chambers – the empty spaces where consecutive segments of the liquid mix together instead of being separated. A good example of a dead volume is a T-connector used as a union with two entrances used, and the third being sealed. The volume of the unused part of the T-connector is still filled with liquid and it acts as a mixing chamber. Another example of a mixing chamber is a nut – ferrule connection where there is some empty space left between the end of the tubing and the port wall. Both dead volumes and mixing chambers hamper the performance of chromatographic separation, causing www.jss-journal.com
2184
M. Noga et al.
J. Sep. Sci. 2007, 30, 2179 – 2189
Figure 1. Separation of 1 pmol of BSA digest using the Ultimate nano-LC system (LC Packings/Dionex) with 5 cm long 75 lm id C-18 column (LC Packings/Dionex) at the flow rate of 300 nL/min, and detection with an Esquire 3000 IT instrument (Bruker Daltonics). The sample was injected using a C-18 trap column. The upper panel shows the analysis with dead volume on one of the precolumn's connectors, while the bottom panel shows the same analysis repeated after the removal of the problem. Chromatograms plotted are base peak chromatograms (range 400 – 1400 m/z) showing base peak intensity on the y-axis.
peak broadening and prolonging elution times. Though these problems are common to all types of the LC, they are particularly destructive for the nanoscale separations. A small inaccuracy in the preparation of the ferrule – nut fitting, forming a mixing chamber of 500 nL that would not even be noticed during the separation on the analytical scale (1 mL/min flow rate), will completely ruin a nanoscale separation (200 nL/min). A typical example of the impact of a huge dead volume on the quality of chomatographic separation is shown in Fig. 1. A small inaccuracy in the preparation of one of the fittings in the system caused so severe mixing inside the system that no peaks were detected at all. We would like to point out once again that all fittings used in the nanochromatographic system should be prepared with special care. As manual cutting of the tubings may result in an uneven surface, the use of the commercially prepared tubings is encouraged. Also, the number of fittings should be reduced to a minimum, as each and every connection may contribute to the overall dead volume.
i
2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
It is also desirable to prepare all connections in the easiest possible way. Typical high-pressure connection in the capillary LC consists of the fused-silica capillary inserted in a Teflon or PEEK sleeve tightening connection, which is mounted in the port using nut and a ferrule or an integrated PEEK or KelF fingertight fittings. Such connections are very reliable and pressure resistant. However, preparation of such a link may be prone to errors. First of all, there are two tubings in the port: the sleeve and a capillary, so there are two possibilities to create a mixing chamber by leaving a small space between the port wall and tubing. Moreover, wrench-tight fittings may cause this tubing to be sealed too tight, thus permanently fixing the capillary in the sleeve. This makes both removal and reinstallation of the capillary even more impossible, as any further adjustment of the capillary position in relation to the sleeve is difficult and may lead to the creation of a dead volume. To simplify the preparation of the capillary connection, manufacturers of the LC systems suggest the use of PEEKSil capillaries. These precut tubings consist of the www.jss-journal.com
J. Sep. Sci. 2007, 30, 2179 – 2189
Liquid Chromatography
2185
Figure 2. Effect of the l100 nL mixing chamber located directly before the nano-ESI spray needle. Upper panel shows chromatogram under standard conditions, while the lower panel shows the presence of a mixing chamber. Apart from shifting the retention times, the peak shape and resolution were not affected by the presence of the dead volume. The reason is that larger dead volume existed earlier in the flow path, thus reducing the impact of the examined mixing chamber on the chromatogram observed. Analysis was performed using the Ultimate nano-LC system (LC Packings/Dionex) with a 5 cm long 75 lm id C-18 column (LC Packings/Dionex) at the flow rate of 300 nL/min with the Esquire 3000 IT instrument (Bruker Daltonics). Chromatograms plotted are base peak chromatograms (range 400 – 1400 m/z) showing base peak intensity on the y-axis.
fused-silica capillary, permanently fixed inside a Teflon tubing. Such a connection is not only much more reliable, but also much more expensive. It should also be noted that PEEKSil tubings are often used with the PEEK fingertight fittings that are less pressure resistant as compared with steel fittings, and small mixing chambers tend to be created over time, especially if the tubing is subjected to sudden pressure drops (i. e., if the tubing connects injection loop and/or precolumn cartridge). On the other hand, steel fittings will permanently remain on the capillaries and any replacement is impossible. The only solution is to shorten the capillary. When additional detectors are used before the mass spectrometer, such as UV – Vis or fluorescence detector, care should be taken to ensure that the flow cells would not function as mixing chambers. Typical low-volume, Zshaped flow cells have a volume of around 5 nL and width of around 20 m, which is similar to a typical id of a capillary transferring liquid from the column to the ion
i
2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
source of the mass spectrometer. If such a nano-LC cell will be replaced with a 250 nL cell for microbore chromatography, severe peak broadening would arise, causing a tremendous sensitivity drop. In the case of highly concentrated samples, good sensitivity of the nanoelectrospray may still allow for the detection of the most concentrated compounds, but such a system is unable to deal with small quantities of the less abundant components. At the end, it should be emphasized that sometimes, neither dead volumes nor mixing chambers affect the separation in any way. The peak will be as broad as it is caused by the biggest mixing chamber inside, i. e., if there is a large mixing chamber just after the column, removal of the small dead volume present between the UV – Vis detector and MS may not change the separation performance at all. Figure 2 shows comparison of the exemplary separations with and without l100 nL mixing chamber created intentionally just before the ESI needle. In both www.jss-journal.com
2186
M. Noga et al.
J. Sep. Sci. 2007, 30, 2179 – 2189
Figure 3. Injection using the 1 cm 100 lm id precolumn (upper panel) versus direct injection with sample loop (lower panel). Separation obtained after direct injection shows narrower peaks and better resolution. The reason is that the connection between the trap column and analytical precolumn acts as a mixing chamber. Analysis was performed using the Ultimate nano-LC system (LC Packings/Dionex) and a 5 cm long 75 lm id C-18 column (LC Packings/Dionex) at the flow rate of 300 nL/min with the Esquire 3000 IT instrument (Bruker Daltonics). Chromatograms plotted are base peak chromatograms (range 400 – 1400 m/z) showing base peak intensity on y-axis.
cases, a much bigger mixing chamber already existed in the flow-path, therefore, no clear difference between these chromatograms is observed.
3.5 Trap columns versus injection loops The need to reduce total volume of the system and shorten the analysis time encouraged introduction of the sample injection system that is alternative to the use of the injection loop. The smallest, commonly used HPLC injection loop having a volume of 1 lL is almost 1.5-times larger than the total volume of a typical capillary LC column (15 cm length, 75 lm id, estimated volume l660 nL). At the flow rate of 200 nL/min, it takes just 5 min to pump the liquid through the injection loop. A commonly used alternative is the use of additional precolumns, also called trap columns, positioned in the system just before the analytical column. The sample is first loaded from the injection loop or autosampler onto the trap column, at a higher flow (i. e., 20 – 30 lL/min). The
i
2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
trap column contains the same packing material as the analytical column, and the mobile phase composition is adjusted to retain all the molecules of interest, and to wash off all impurities. In the most commonly used RP separations, the loading buffer typically consists of water with small concentration of organic acid such as TFA or formic acid. Under such conditions, organic compounds such as peptides and proteins are adsorbed on the stationary phase, whereas impurities, such as salts are washed away to the waste. After sample loading, the flow path is altered, and the low-flow gradient is delivered through the trap column into the analytical capillary column. Previously adsorbed compounds are transferred into the separation column and then separated. As a result, the sample is purified, preconcentrated, desalted, and the analysis time is significantly reduced. This setup does not differ from that used in the analytical systems. The problem is the scale of analysis. Figure 3 shows chromatograms from the separation of 1 pmol of the BSA digest using 5 cm length, 75 lm id RP www.jss-journal.com
J. Sep. Sci. 2007, 30, 2179 – 2189
C-18 column (LC Packings/Dionex, Amsterdam, The Netherlands). In the first case, the sample was directly injected using a 1 lL sample loop; in the second method, a 1 cm, 100 lm id precolumn (with the same stationary phase) was used to transfer the sample into the system. Surprisingly, the chromatogram obtained using direct injection with the sample loop looks much better with the well-separated, sharp peaks. In the case of the precolumn, the peaks are wider and the effective sensitivity and resolution of the analysis are reduced. The explanation of this phenomenon is that the connection between the end of the precolumn and the front of the analytical column forms a small mixing chamber that causes peak broadening. Does it mean that introduction of the loading columns should be discouraged? By all means no! Even if a direct injection using sample loops has many advantages, simple construction and minimized dead volumes, such an approach has multiple drawbacks. First of all, capillary columns are very prone to damage, especially by the solid particles that may be present in the sample, and most likely such solids are present in biological samples. As such impurities pose a serious threat when sample is directly injected onto analytical column, the precolumn acts as a guard column, thus protecting the main column from damage. Moreover, manual injection using sample loop consumes at least 2 – 3 times the volume of the sample loop during sample transfer from the assay tube to the chromatographic column. In contrast, modern autosamplers are able to perform the injection without any sample loss, but require much larger loop volumes (10 – 20 lL) that cannot be used for direct injection onto the capillary LC system. All these problems are minimized when the trap columns are utilized. Practical advantages of the application of the loading columns (column protection, no sample loss, adjustable sample volumes, automation) are often more important for a successful analysis than slightly decreased resolution and sensitivity. In most cases, the mass spectrometer is sensitive enough to detect and fragment most of the compounds of interest, even if certain peak broadening occurs. Sometimes, wider peaks even lead to a better analysis, especially when data-dependent fragmentation is used. The fragmentation process is often time-consuming, and a broader peak leaves more time for fragmentation in the MS system. For this reason, the final identification scores provided better results for the analysis shown in Fig 3. despite the worse chromatographic resolution. There are two possible approaches to install the precolumn in the system: (i) frontflush and (ii) backflush. In the first case, the flow of a solvent through the precolumn has the same direction as the nanoflow used later to transfer the sample onto the analytical column. In the backflush mode, these flow directions are opposite. In
i
2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Liquid Chromatography
2187
the case of the frontflush approach, the sample has to pass through the entire precolumn first, before it is transferred onto the analytical column. In the backflush setup, the compounds are retained mostly at the very front of the precolumn. Then, the flow direction is reversed and the sample is desorbed. Typically, the backflush mode allows for better chromatographic resolution than the frontflush approach. However, the backflush mode has several disadvantages that limit its application. First of all, the backflush mode requires that the precolumn is equipped with frits on both ends of a bed. Moreover, the stationary phase must be tightly packed between both frits to avoid any displacement of the packing material during reversal of the flow direction. Capillary precolumns fulfilling such requirements are difficult to manufacture. It should be noted that both loading flow and the analytical nanoflow are delivered at high pressure. Reversing the flow within a short time-frame leads to a huge pressure drop. During the experiments performed in our laboratory, we encountered situations when a 120 bar pressure generated by the loading pump was reversed with a nanoflow with a pressure of 80 bar in the reverse direction, resulting in a total of 200 bar pressure drop in less than a second. Such pressure falls are very dangerous for the precolumn and cause very high stress to the system, thus causing leakages and sample loss. Yet another disadvantage of the backflush mode is that it does not protect the analytical column against solid particles that are abundant in the biological samples. Large particles will be stopped on the precolumn's inlet frit, but after the flow has been reversed, they will inevitably move toward the analytical column. In the frontflush setup, the precolumn acts as a guard column and solid particles do not reach the inlet of the analytical column. In our laboratory, we tested both the backflush and frontflush modes. Initially, our setup operated in the backflush mode. Facing the problems described above, we were forced to abandon this idea and switched to the frontflush mode. In our case, an increased system reliability and integrity was more important than a slightly decreased performance. Therefore, all results presented in this paper were acquired using the frontflush setup.
3.6 Sudden pressure drop It should be noted that issues caused by pressure falls in the system are not limited to the leakages or precolumn setup, but may also occur every time the flow path is altered during the system operation, e. g., during valve switching. Although we did not encounter any serious problems caused by this element of instrumentation, we believe that malfunctions of the valves (either manual or automatic) may contribute to the overall quality of sepwww.jss-journal.com
2188
M. Noga et al.
arations. In general, pressure drops caused, e. g., by an improper dismantling or assembling of the valve parts, may lead to, e. g., the leakage of a damaged rotor seal. Also, capillary columns may be destroyed by the occasional transient pressure shocks. The latter often happens with the use of manual valves, when they are switched too slowly. The flow is then obscured and the pressure increases.
3.7 When UV matters Whereas UV – Vis detector has very limited analytical capabilities as compared to the mass spectrometer, it might be anticipated that the flow-cell of a UV – Vis detector used in the capillary LC system is by no means useless. In contrast, the UV detector may be utilized as a powerful tool for system diagnostics and troubleshooting. First of all, as the absorbance of solvents applied for gradient separation differs in time, it is possible to use such a detector to measure the void volume of the system. For example, ACN has higher absorbance than water at 230 nm, so the linear gradient with increasing concentration of ACN versus water will result in a linear increase in the absorbance detected by the UV detector. If the column was equilibrated with water for certain time before gradient start, the emerging chromatogram will consist of a flat baseline (water), followed by an absorbance increase. This will form a raising line. The time of the intercept of this line with the baseline corresponds to the delay time. After multiplying by the actual flow rate, the void volume of the system is calculated. Second, an even more important way to utilize the UV detector is the detection of the dead volumes. Let us assume that the system with a loading pump and a precolumn is employed for sample injection, and the loading solvent has an absorbance different from those of the mobile phases used for separation (i. e., 0.1% TFA in water as a loading solvent and 0.1% formic acid in water/ACN used to form the gradient). After the sample has been loaded, the precolumn is filled with 0.1% TFA solution. Just after the valve has been switched, this liquid segment is injected into the flow path through the UV detector. As hydrophobic compounds will be retained on the columns, ionic species like TFA molecules will pass through them freely. If there is no mixing and no diffusion, the TFA segment will be seen in the UV detector as a rectangle-shaped “negative” peak (Fig. 4). Unavoidable diffusion causes its corners to be rounded. If there is a dead volume in the flow path, the front slope of the peak will remain sharp because the liquid fills the space completely. As the liquid in the chamber mixes, it will be eluted slower than usual, and the TFA concentration will gradually decrease. Instead of a sharp slope of the TFA peak, a smooth curve will be visible as the absorbance slowly returns to the initial value. The broadening of this
i
2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
J. Sep. Sci. 2007, 30, 2179 – 2189
TFA plug can be used to estimate how severe the dead volumes within the system are. Figure 4 shows UV chromatograms of four different nano-LC setups. The upper one shows the severe void volume caused by the faulty mounting of the precolumn. Lower chromatograms describe two precolumn cartridges of different lengths, installed using PEEKsil tubings and fingertight fittings. Some broadening is still visible and the TFA plug is not symmetric. The bottom chromatogram shows the same analysis performed on a fused-silica precolumn mounted directly with just two stainless steel fittings. In this case, the TFA plug is symmetrical with both slopes sharp and with no traces of mixing; but some traces of diffusion can be seen. If the analysis is unsuccessful due to the created mixing chamber, the use of a UV detector allows for an easy detection of this problem. In the presence of such severe dead volume, the mass spectrometer will not detect any peaks. MS chromatogram from Fig. 1 is recorded during the same analysis as the topmost chromatogram in Fig. 4. Whereas chromatogram in Fig. 1 shows only the unsuccessful analysis, the one in Fig. 4 gives an insight into the causes of the problem. The same approach may be used to troubleshoot the systems when sample loop has been applied for injection. The only requirement is that the solvent forming the segment injected onto the analytical column has different absorbance values compared to the mobile phase.
4 Multidimensional LC Other LC MS techniques, such as multidimensional capillary chromatography linked to MS, are out of the scope of this article. However, as it becomes popular, we would like to add at least few words describing this approach. In general, multidimensional LC MS is a powerful method, leading to a better sensitivity and higher throughput, as compared to the LC MS method. On the other hand, as many columns and valves are involved, all methodological problems described here are multiplied. One of the basic principles of mounting such a system is that id of the following columns should decrease, i. e., the first column in the system has the largest internal diameter, and the final purification/separation should be performed on the column with the smallest internal diameter. This prevents sample dilution or peak broadening. As complexity of the system increases with the number of dimensions (number of columns), it is often advisable to perform rough separation in the first dimension with the aid of the offline setup. Collected fractions are concentrated and subsequently analyzed with the LC-MS system. This strategy still maintains the main advantages of the multidimensional LC system (such as high resoluwww.jss-journal.com
J. Sep. Sci. 2007, 30, 2179 – 2189
Liquid Chromatography
2189
Figure 4. Exemplary UV chromatograms used for the dead volume troubleshooting. All chromatograms were taken at 215 nm (peptide bond) and after injection of 1 pmol of BSA digest. Analyses were performed using the Ultimate nano-LC system (LC Packings/Dionex) using a 5 cm long 75 lm id C-18 column (LC Packings/Dionex) and an injection using loading precolumns. The top chromatogram was obtained during the analysis with the 5 mm long 300 lm id precolumn with faulty connection and a large mixing chamber completely ruining the analysis. Second chromatogram results form repetition of the previous analysis after removal of the dead volume. Third chromatogram shows analysis using 1 mm long 300 lm id precolumn and the fourth shows the analysis obtained by using a 1 cm long 100 lm id precolumn. All precolumns were filled with the same packing material as the analytical column. Flow rate was 300 nL/min.
tion) but is much simpler than dealing with all columns linked together.
Italy (grant no. CRP/POL05/02). M. N. was supported by a research grant from the Polish Ministry of Science and Higher Education (grant no. N204 136 32/3396).
5 Conclusions Operation of a nano-LC system is often subjected to far more methodological problems than an “ordinary”, analytical scale HPLC. It combines some of the most sophisticated analytical tools available for the separation scientist. Unfortunately, although capillary LC coupled to the mass spectrometer combines most of the important features of both HPLC and MS, it also combines their problems, adding a few new as a free bonus. The work of a separation scientist using the nano-LC system is full of unexpected challenges. We hope this guide will help to solve at least a few of them.
6 References [1] Dole, M., Mach, L. L., Hines, R. L., Mobley, R. C., Ferguson, L. D., Alice, M. B., J. Chem. Phys. 1968, 49, 2240 – 2247. [2] Fenn, J. B., Proc 36th Annual Conferrence, Am.Soc. Mass Spectrom. San Francisco, 1988, 772. [3] Wilm, M., Mann, M., Int. J. Mass Spectrom. 1994, 136, 167 – 180. [4] Emmett, M. R., Caprioli, R. M., J. Am. Soc. Mass Spectrom. 1994, 5, 605 – 613. [5] Andren, P. E., Emmett, M. R., Caprioli, R. M., J. Am. Soc. Mass Spectrom. 1994, 5, 867 – 869. [6] Davis, M. T., Stahl, D. C., Hefta, S. A., Lee, T. D., Anal. Chem. 1995, 67, 4549 – 4556. [7] Gatlin, C. L., Kleeman, G. R., Hays, L. G., Link, A. J., Yates, J. R., Anal. Biochem. 1998, 263, 93 – 101.
We would like to acknowledge the grant from the International Centre for Genetic Engineering and Biotechnology (ICE), Trieste,
i
2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.jss-journal.com
