Article pubs.acs.org/ac
Exploring the Possibilities of Cryogenic Cooling in Liquid Chromatography for Biological Applications: A Proof of Principle Hamed Eghbali,*,† Koen Sandra,‡ Bart Tienpont,‡ Sebastiaan Eeltink,† Pat Sandra,‡ and Gert Desmet† †
Department of Chemical Engineering, Vrije Universiteit Brussel, Pleinlaan 2, B-1050 Brussels, Belgium Research Institute for Chromatography, Kennedypark 26, B-8500 Kortrijk, Belgium
‡
ABSTRACT: The possibilities to use cryogenic cooling to trap components in liquid chromatography was investigated. In a first step, van ‘t Hoff plots were measured with a reversed-phase column using the temperature control unit of a conventional high performance liquid chromatography (HPLC) system to gain insight in the retention behavior of proteins at low temperatures. It was estimated that retention factors in the range of k = 104 could be achieved at T = −20 °C for lysozyme, indicating that temperature is a usable parameter to trap components in LC. In a next step, trapping experiments were carried out on a nano-LC system, equipped with a UV-detector, using a commercial reversed-phase column. An in-house built setup, allowing cooling of a segment of the column down to temperatures below T = −20 °C, was used to trap components. Experiments were conducted under isocratic and gradient conditions with methanol as organic solvent. It is demonstrated that, by thermally trapping and elution of components, an enhanced S/N ratio and decreased peak widths can be obtained. At the same time, a significant increase in pressure drop occurs during the cooling process. Limitations and benefits of the technique are further discussed.
T
very limited.16 To investigate this temperature range in more detail, and more specifically, the increase in retention that can be realized in this range, we presently report on a preliminary study which aims at exploring the potential use of cryogenic cooling in reversed-phase liquid chromatography for biological applications (proteins and peptides). Potential applications could be online sample enrichment prior to analysis, detector preconcentration, heart cutting, and a modulation system for two-dimensional LC (2D-LC). Using an in-house built setup, the possibility to trap (focus) biological components (proteins and peptides) by rapidly cooling and heating up a predefined column segment located near the end of the analytical column is explored. The reason to apply this operation near the end of the column can be supported by the fact that the effect of the trapping (focusing) would be much more noticeable as the injected peaks have already broadened before they reach the trapping zone. The influence of the cooling process on the different system parameters (S/N ratio, pressure drop, etc.) is investigated. The experiments are carried out on a nano-LC system using a commercial reversed-phase (C18) column having an outer diameter OD = 360 μm and an inner diameter ID = 75 μm. Methanol is applied as organic solvent for the mobile phase. Both isocratic and gradient conditions are considered. Prior to the separation and trapping experiments, a series of van ‘t Hoff plots were measured on a conventional high performance liquid
emperature is an important variable in liquid chromatography (LC) and has led to the concept of “high temperature liquid chromatography” which has attracted major attention during the last years since an increased chromatographic speed and a reduced back pressure can be achieved when operating at elevated temperatures.1−11 In addition, higher temperatures also lower the retention factors in reversed-phase mode. This allows a reduction of the organicphase content of the mobile phase resulting in a “greener” and more cost-effective approach.12 Detectors that are not compatible with organic solvents such as a flame ionization detector (FID) could also be applied.13 To meet the demand of the rapidly emerging research fields from the life sciences (genomics, proteomics, metabolomics, etc.) where one has to deal with limited sample amounts, nanoLC systems have been introduced and commercialized.14 The reduced column dimensions of these systems lead to shorter heat transfer distances, offering the possibility to rapidly alter the radial temperature profile of the separation column in very short times whenever required. Very recently, Collins et al. described an experimental setup for capillary and microbore liquid chromatography columns allowing a simultaneous temporal and spatial control. In their setup, the temperature could be varied over a range between 15 and 200 °C.15 On the one hand, they used their system to apply axial and temporal temperature gradients. On the other hand, they were also able to demonstrate in-capillary synthesis of monolithic stationary phases. Whereas the vast majority of the temperature studies in LC focuses on an increased temperature, the number of reports where the effect of decreased temperatures (subzero) and possible accompanying applications are investigated has been © 2012 American Chemical Society
Received: December 8, 2011 Accepted: January 17, 2012 Published: January 17, 2012 2031
dx.doi.org/10.1021/ac203252u | Anal. Chem. 2012, 84, 2031−2037
Analytical Chemistry
Article
chromatography (HPLC) system to gain insight in the retention behavior of analytes at low temperatures.
■
EXPERIMENTAL SECTION Chemicals. HPLC grade water and methanol were obtained from Biosolve (Valkenswaard, The Netherlands). Spectrophotometric trifluoroacetic acid (TFA), ribonuclcease A (from bovine pancreas), lysozyme (from chicken egg white), and carbonic anhydrase (from bovine erythrocytes) were purchased from Sigma−Aldrich (Bornem, Belgium). Tryptic digest of bovine cytochrome c was obtained from Dionex (Amsterdam, The Netherlands). LC Instrumentation. van ’t Hoff plots were recorded using a conventional HPLC system (UltiMate 3000 Rapid Separation LC (RSLC)) (Dionex, Munich, Germany) provided with a UVdiode array detector and equipped with a temperature controlling unit (column oven). Solvents A and B consisted of water (0.1% TFA) and methanol (0.1% TFA), respectively. A narrow bore column (Agilent 300 SB-C8, 2.1 mm × 50 mm, dp = 3.5 μm) (Agilent Technologies, Little Falls, US) was used at a flow rate of 0.2 mL/min. The amount of sample that was injected was 1.5 μg. The trapping experiments were conducted using an UltiMate 3000 RSLC nano-LC system (Dionex, Munich, Germany) including a splitless nanopump and a variable-wavelength detector (λ = 214 nm; 280 nm) equipped with a 3 nL flow cell. A nano-LC column (Acclaim PepMap 300 C18, 75 μm × 15 cm, dp = 5 μm) (Dionex, Amsterdam, The Netherlands) operated at a flow rate of 300 nL/min was directly coupled to the injection valve comprising a 1 μL injection loop. The nanoLC column was connected to the detector capillary (L = 30 cm, ID = 20 μm) through an additional transfer capillary (L = 27 cm, ID = 20 μm) (Figure 1). The connections were established by means of zero-dead volume unions (Upchurch Scientific, Oak Harbor, US). Linear gradients were conducted for different gradient times followed by column regeneration and equilibration at a temperature of 24 °C. The sample amounts (proteins and peptides) that were injected ranged between 2 and 11 ng. Chromeleon software (Dionex) was used for data acquisition and analysis. Cooling Setup. The in-house built prototype setup used nitrogen gas as carrier fluid to generate a cold or warm gas flow (Figure 1). More specifically, dry ice (T = −78 °C) was employed as cooling agent whereas boiling water was used as heating agent to cool down and warm up the nitrogen gas flow, respectively. Both gas circuits (cold and warm) were connected to two valves which could be separately controlled by two external switches enabling an alternation between a cold and warm gas flow. A special interface (“cooling chip”) (Figure 1) was designed to allow an efficient temperature alternation over a predefined column segment, located 3 cm upstream of the end of the analytical column. The chip, fabricated in plexy glass, consisted of two different parts (main substrate and cover plate) which could be assembled by means of two screws (not shown in Figure 1). The nano-LC column could be embedded in a small central groove running over the entire length of the main substrate. This groove was connected to the cold and warm gas circuit through a T-connector. The cover plate, which was fixed on top of the main substrate, was provided with two different circular openings (a distance of 4 mm remote from each other and each opening having a diameter of 1 mm). These two openings were directly connected to the groove of the main
Figure 1. (a) Schematic representation of the experimental setup used to cool down and warm up a predefined column segment near the end of the analytical column with (1) two external switches to operate the cold and warm circuit; (2) gas valves; (3) flow regulators; (4,5) two insulated containers for the cooling and heating agent, respectively. (b) Detailed schematic drawing of the “cooling chip” showing the main substrate with the groove and the cover plate with the two circular openings (blue); the T-connector (red); the two plugs (green) to close the side openings of the “cooling chip”. The dashed arrows (white) indicate the direction of the convective gas flow inside the “cooling chip” coming from the cold (dashed blue arrow) and warm (dashed red arrow) circuits.
substrate wherein the nano-LC column was placed. The two remaining openings that emerged from the groove on the sides of the “cooling chip” were closed with two plugs (green) to prevent the cooling gas flow to leak out. As a consequence, the convective gas flow (indicated with dashed arrows in Figure 1b) was limited to the groove between these two circular openings in order to minimize the formation of an axial temperature gradient along the column during the cooling process. Therefore, this specific area (spanning a column length of 4 mm) defined the specific segment of the column that was cooled or warmed up. An Almemo Type 2290-4 temperature sensor (Ahlborn, Holzkirchen, Germany) equipped with a NiCr-Ni temperature probe and having an operative range between T = −200 °C and T = 500 °C was used. The probe had a diameter of 0.5 mm and a length of 1 cm. The value of the time constant of the sensor 2032
dx.doi.org/10.1021/ac203252u | Anal. Chem. 2012, 84, 2031−2037
Analytical Chemistry
Article
was 0.8 s. The temperature sensor was connected to a computer, and the temperature was recorded digitally as a function of time at a rate of 1 data point/s through the accompanying Almemo View software. In order to measure the temperature profile inside the “cooling chip”, the nano-LC column was replaced by the probe of the temperature sensor. After recording the temperature profile, the cooling setup was left intact while the probe was removed. Subsequently, the column was repositioned into the groove of the “cooling chip” followed by the experiments that were performed under the same conditions as the temperature was measured inside the chip.
components (tubing, chip, etc.) of the system have a nonzero time constant, implying that the temperature of all components must first reach a steady state before it can plateau for the column. This assessment had its implications with respect to temperature accuracy. In general, the average target temperature could be imposed within a deviation of ΔT = 5 °C. Furthermore, through a modification of the setup, faster temperature changes than the one in the current study would be possible either by reducing the heat capacity of all components of the system or by using a cooling/heating agent with higher heat capacity and better heat transfer properties (e.g., a liquid). As was described in the Cooling Setup section, the “cooling chip” was designed to prevent the formation of an axial temperature gradient along the column and to cool very locally. This strategy had the purpose to enchance the peak focusing effect at the transition zone where the analytes, that migrate through the column at ambient temperature (T = 24 °C), suddenly experience the cold spot. However, during prolonged cooling experiments, tiny ice crystals could be observed on the capillary wall outside the “cooling chip”. This indicates that, despite the fact that the “cooling chip” perfectly restricts the application of the hot and cold gas flow to its internal volume, it cannot be prevented that part of the added or removed heat also influences the temperature in the vicinity of the “cooling chip”, due to the flow of the cold mobile phase inside the capillary and also partly, but less pronounced, by heat conduction occurring along the mantle of the separation capillary. van ‘t Hoff Analysis. Prior to exploring the possibility to trap components with the in-house built cooling setup, a van ‘t Hoff analysis was carried out with a conventional LC system (allowing one to cool the complete column down to a temperature of T = 5 °C) to gain more insight in the retention behavior of proteins at lower temperatures. For these experiments, lysozyme was used as test component. The van ‘t Hoff equation describes the relationship between the retention factor k and the absolute temperature T (K):17
■
RESULTS AND DISCUSSION Cooling Setup. Figure 2 shows an example of the temperature profile in the “cooling chip” that can be obtained
Figure 2. Temperature profile in the groove of the “cooling chip” illustrating cooling and heating rates that can be obtained with the experimental setup. Five consecutive cycles where the temperature is rapidly alternated from T ≈ −40 °C to T ≈ 25 °C can be observed. One complete cycle takes approximately 20 s.
when the temperature is alternated with the current setup. Five different temperature cycles ranging between T ≈ −40 °C and T ≈ 25 °C can be clearly observed while demonstrating the cooling and heating rates (one cycle taking approximately 20 s) that are currently possible. The temperature range of the setup as well as the cooling and heating rates depended on several setup parameters such as the gas flow rate and the length of the tubing that connected the cooling and heating reservoirs with the “cooling chip” (Figure 1). As a consequence, the temperature range of the setup as well as the cooling and heating rates could be imposed as desired, by tuning the flow rate using the flow regulators that were integrated in the setup and/or by adjusting the length of the tubing. Yet, it should be mentioned that these setup characteristics could be varied within certain limits. Considering, for example, the temperature range of the setup, it is already obvious that the lower and upper temperature limits were theoretically determined by the temperature of the cooling and heating agent which is T = −78 °C for dry ice and T = 100 °C for boiling water. In addition, it is also noteworthy in Figure 2 that the curve shape at the temperature extremities (T ≈ −40 °C and T ≈ 25 °C) during each cycle is not perfectly flat, as it would be in the ideal case. This can be explained by the fact that the different
ln(k) = −
ΔH 0 ΔS 0 + + ln(φ) RT R
(1)
where ΔH0 (J/mol) and ΔS0 (J/(K·mol)) are the retention enthalpy and entropy, respectively, R is the universal gas constant, and φ represents the phase ratio. Figure 3 shows van ’t Hoff plots (ln(k) vs 1/T) measured with a narrow bore column within a temperature range close to room temperature. The lower temperature limit for the experiments was T = 5 °C which was imposed by the temperature controlling unit (column oven) of the LC system. Experiments were conducted for two different mobile-phase compositions (63/37 and 65/35 (v/v) (B/A)) (Figure 3). Within the considered range of temperatures, a clear linear trend (R2> 0.99) can be assessed for both data sets which is consistent with eq 1 (Figure 3). Assuming the same linear correlation will persist when the temperature is further decreased, the retention constant at lower temperatures can be estimated. As an illustration, it can be calculated that, for a mobile-phase composition of 63/37 (v/v) (B/A), lysozyme would display a retention constant of k = 2 × 104 at a temperature of T = −20 °C. This value is four magnitudes larger than the retention constant (k = 3) obtained at room temperature (T = 25 °C) (Figure 3). It would be very 2033
dx.doi.org/10.1021/ac203252u | Anal. Chem. 2012, 84, 2031−2037
Analytical Chemistry
Article
Figure 3. van ‘t Hoff plots of lysozyme obtained with a conventional HPLC system using a narrow bore column (2.1 mm × 50 mm, dp= 3.5 μm) at two different mobile-phase compositions (63/37 (v/v) (B/A) (squares); 65/35 (v/v) (B/A) (diamonds)). All data points result from triplicate measurements. Relative standard deviations were less than 2%. The full line curves are linear regressions of the experimental data.
Figure 4. Trapping experiment of lysozyme showing a UVchromatogram (λ = 214 nm) where the protein is eluted under isocratic conditions (67/33 (v/v) (B/A)) at T = 24 °C (red curve) using a nano-LC system and a UV-chromatogram obtained after implementation of the depicted temperature program (the blue zone and the red zone correspond to T ≈ −20 °C and T ≈ 30 °C, respectively) where lysozyme is thermally trapped and released (purple curve). A partial-loop injection (lysozyme (10 ng; V = 0.1 μL)) was used.
impractical to try to verify this extrapolation, given the extremely long retention times that would occur under this condition. Nevertheless, the result indicates that the trapping of large MW compounds should be possible, especially considering that lysozyme can be regarded as a relatively small protein, composed of only 128 amino acids (MW = 14.3 kDa). More concretely, the sensitivity of a component to changes in temperature is eventually determined by the retention enthalpy (ΔH0), which can be derived from the slope of the van ‘t Hoff plot (Figure 3) according to eq 1. From the slope of the curves shown in Figure 3, it can be calculated that the retention enthalpy of lysozyme has a value of about ΔH0 = −102 kJ/mol. This value is of the same magnitude as the value that can be derived for lysozyme from Figure 18 in ref 18. For smaller molecules, lower values in the range of ΔH0 = −10 kJ/ mol are typical,19 so that the presently proposed temperaturebased trapping approach will probably be insufficient for small MW compounds. In this case, additional retention capacity will need to be added, e.g., by packing the trapping segment with C30-derivatized particles. Trapping under Isocratic Conditions. The red curve in Figure 4 shows a chromatogram of lysozyme under isocratic conditions exhibiting a retention constant of k = 0.8. This experiment was established with the nano-LC setup shown in Figure 1 at ambient temperature (T = 24 °C). It can be noted that the obtained peak is relatively broad (peak width ≈ 2.5 min) due to the employed isocratic conditions. To investigate whether the sample could be trapped for a certain time at the end of the analytical column, a temperature program (Figure 4, blue and red zone), based on the chromatogram obtained under reference conditions (Figure 4, red curve), was imposed to the chip. The purple curve in Figure 4 shows the result of this operation. It can be seen that, whereas the reference chromatogram displays a clear peak around t = 10 min, the new chromatogram remains at baseline level during the entire cooling step (Figure 4, blue zone). This clearly indicates that the injected sample does not break through the column and, hence, is trapped. After the start of the warming-up step (Figure 4, red zone), it takes approximately t = 1.7 min before a sharp peak reaches the
detector showing the possibilities to release the trapped and concentrated component. The observed delay can be mainly explained by two different contributions. On the one hand, it should be taken into account that the cooling setup needs some time to deliver the desired temperature in the chip groove (magnitude of 10 s) after the initiation of the warming-up step. On the other hand, after the column segment has been warmed up, the peak has yet to pass the final part of the separation column followed by the transfer capillary and the detector capillary (interconnected by two connection pieces) (Figure 1a) to finally reach the detector. Knowing the flow rate (300 nL/min), the dimensions of the postcolumn tubing (detector capillary (L = 30 cm, ID = 20 μm), transfer capillary (L = 27 cm, ID = 20 μm), and the zero-dead time of the column (t0,column = t0,(column+tubing) − t0,tubing), it can be calculated that the trapped component has an approximate travel time of 1.5 min after the “cooling chip”. Consequently, this latter issue is the main contributor to the overall delay time that is observed before the trapped peak reaches the detector, as is depicted in Figure 1a. The same relatively long postchip trajectory also leads to additional band broadening. Therefore, it can be assumed that even sharper peaks can be obtained than the one obtained in the present study (peak width ≈ 1 min) by reducing the postcolumn connection tubing and the number of connection pieces and by positioning the “cooling chip” closer to the column end. Nevertheless, these issues are going to be addressed in the future. Although isocratic separations of proteins are not often used in practice, the obtained results directly demonstrate that a decreased temperature can be applied to trap and concentrate proteins. This cooling approach could for instance be advantageous to enhance online sample enrichment in the proteomics field where the separation is typically preceded by a preconcentration step in a separate enrichment column.20 More specifically, by cooling the enrichment column at the exterior, one could concentrate the sample more efficiently prior to injection in the analytical column. A similar yet different 2034
dx.doi.org/10.1021/ac203252u | Anal. Chem. 2012, 84, 2031−2037
Analytical Chemistry
Article
The purple curve in Figure 5a shows the same gradient separation but now obtained by applying a subsequent set of cooling and heating steps to focus the three different proteins on-column during the separation run. The employed timebased temperature program imposed to the chip is depicted in Figure 5a as well (cf., the blue and red colored zones). The resulting chromatogram clearly shows the significant increase of the S/N ratio of all three proteins. The UV-signal of the proteins now strongly exceeds the noise level, thus demonstrating the possibilities of the trapping chip as a generic detector preconcentration technique for large MW analytes. When considering mass spectrometry detection as an example, it is generally known that the reliability of component identification can strongly depend on the measured signal intensity. Thermal peak focusing could, therefore, be a helpful tool for the identification of unknown components as it can enhance the signal intensity. Besides an improved sensitivity, a better peak shape of all three peaks can be observed as well (Figure 5a). All in all, much narrower peaks are obtained compared with the separation (red curve in Figure 5a) where no cooling was applied. In a next step, the influence of the thermal process on the pressure drop was investigated since it is well-known that the physicochemical properties of the mobile phase such as the viscosity are affected by temperature. The relationship between the column pressure drop (ΔP (Pa)) and mobile-phase viscosity (η (Pa·s)) is giving by Darcy’s law:
approach has already been demonstrated for small MW compounds by Holm et al.16 Ultimately, the potential use of decreased temperatures for sample concentration purposes can, among others, be supported by the success of cold fiber solid phase microextraction (CFSPME), a sample preparation technique where a decreased temperature is used to increase the adsorption of components on a solid material.21,22 However, the practical operation of this technique differs significantly from the above-described concept since it usually consists of a syringe-like device and, on top of that, CFSPME has mainly been applied for small molecule components in combination with GC. In the present study, the focus is on LC in combination with large biological molecules. Trapping under Gradient Conditions. Protein Trapping. The red curve in Figure 5a shows a gradient separation of
ΔP =
u·η·L Kv
(2)
where u (m/s) is the mobile-phase velocity and L (m) and Kv (m2) are the column length and column permeability, respectively. Figure 5b shows the pressure drop profiles of the two separation runs (red curve: without cooling; purple curve: with cooling) shown in Figure 5a. Comparing both profiles with each other, it is striking that the pressure drop profile correlates perfectly with the temperature program of the focusing process. A significant pressure drop increase over the whole column is observed during each cooling step whereas the pressure drop decreases to the same value as the separation run without cooling (red curve) during each warming-up step. On the other hand, both curves (red and purple curve) perfectly coincide over the rest over the separation run including column regeneration and equilibration. It is also quite remarkable that, although only a short segment of the column is cooled, a significant increase in the overall pressure drop can be observed. This could be very likely attributed to a local increase of the viscosity according to eq 2 caused by the cooling process. Estimating that we cool about 10% of the total packed column and assuming this column represents the main source of pressure drop, it can be estimated that the roughly 20% increase in pressure drop (volume averaged over the three temperature zones, each corresponding to a different mobilephase composition) observed when cooling down to T = −20 °C corresponds to a local viscosity increase of the liquid of about 200%. The increase of the local viscosity caused by the cooling process can also be approached using the empirical
Figure 5. Trapping experiment of proteins (ribonuclease A (rib. A), lysozyme (lys.), carbonic anhydrase (c.a.)) under gradient conditions (50% B to 100% B in tg = 30 min) with (a) a UV-chromatogram (λ = 280 nm) showing a separation of the 3 proteins at T = 24 °C (red curve) and a UV-chromatogram where the 3 proteins are thermally focused (= trapped and released) (purple curve) according to the temperature program depicted on the graph (blue zone (T ≈ −20 °C) and red zone (T ≈ 30 °C)); (b) the pressure drop profiles of the separation runs shown in (a). A full-loop injection (ribonuclease A (5 ng), lysozyme (2 ng), and carbonic anhydrase (1 ng)) was used.
three standard proteins recorded at ambient temperature (T = 24 °C). Although the three components are completely resolved, the peak shapes are rather poor, especially for ribonuclease A. The observed signal intensity can be considered to be relatively low, as it just exceeds the noise level. This was, to some extent, done on purpose, to emphasize the possibility to use the trapping as a means to enhance the S/N ratio. 2035
dx.doi.org/10.1021/ac203252u | Anal. Chem. 2012, 84, 2031−2037
Analytical Chemistry
Article
obtained fractions are introduced to the second column. A similar concept already exists in GC, which is referred to as “cryofocusing”, where subambient temperatures obtained by a cryogenic liquid or gas are used to concentrate a sample in a small plug.24,25 Trapping Limitations under Gradient Conditions. When dealing with gradient separations in combination with cooling, the retention factor of the different components is mainly determined by two different factors. One factor is the temperature of which the relationship with the retention factor is giving by the van ‘t Hoff equation eq 1 as was already discussed extensively in the van ‘t Hoff Analysis section. The other factor is the mobile-phase composition where the volume fraction of the organic modifier (φ) is related to the retention factor (k) according to ref 17:
correlation giving by eq 3 which relates the mobile-phase composition and temperature with the viscosity.6 2 2 ηΦ,T = 10(−2.429 + (714/T) − 1.859Φ+ (912Φ /T) + 1.8586Φ − 968(Φ /T))
(3)
Despite that eq 3 is used well outside the temperature range it was established for, it predicts a viscosity increase of 215% for the presently considered temperature (T = −20 °C) and average mobile-phase composition (75/25 (v/v) (MeOH/ H2O)), which is in very good agreement with the estimated value (200%). This indicates that it is plausible to assume that the local viscosity increase during the cooling process is responsible for the significant increase in the overall pressure drop over the column. Peptide Trapping. Next to proteins, the trapping of peptides was also explored. The red curve in Figure 6 shows a separation
ln(k) = k w − s ·φ
(4)
with s as a constant and kw representing the logarithmic retention factor with the mobile phase containing no organic modifier. Figure 3, which was partly discussed in the van ‘t Hoff Analysis, represents two van ‘t Hoff plots at two different mobile-phase compositions (63/37 and 65/35 (v/v) (B/A)). Considering the retention factor at T = 20 °C for both cases, it can be assessed that only 2% increase of organic modifier already results in significantly different (5 fold difference) retention factors (Figure 3). This observation, emphasizing the major influence of the mobile-phase composition on the retention factor, has important consequences on the trapping process since the thermally trapped components experience a continuously changing mobile-phase composition, i.e., becoming less retained during a gradient separation. This was investigated by injecting lysozyme in the column and attempting to continuously trap lysozyme through cooling. It was found that, even if the column segment was continuously cooled, the component always eluted from the column after a certain period of time. As was speculated earlier, this can be explained by the fact that, from a certain moment on, the elution strength of the mobile phase eq 4 will start to dominate over the retention induced by the decrease of the temperature eq 1, resulting in the elution of the component from the column. This implies that it will not be possible to trap a component for an extended period of time under gradient conditions, which can rather be seen as a limitation of the present trapping technique. Defining the maximum trapping time (MTT) of a component as the difference between the retention time when continuous cooling is applied and the retention time without cooling under the same gradient conditions, the relation between the MMT and gradient time tg was studied. This is illustrated in Figure 7, showing a linear trend (R2 = 0.998), in agreement with the fact that linear gradients are applied. More explicitly, this implies that, the larger the gradient time, the longer it takes before the elution of the trapped component would take place. A good approach to circumvent this restriction on the trapping time would be to modify the linear shape of the gradient profile by, for example, incorporating periods of constant mobile-phase composition (below a concentration of organic modifier at which the component would elute) in the gradient program. During the passage of this specific mobilephase composition, the component can be kept trapped, hence, extending the trapping time. However, to establish such an
Figure 6. Trapping experiment of peptides under gradient conditions (0% B to 100% B in tg = 60 min) showing a UV-chromatogram (λ = 214 nm) where a tryptic digest of cytochrome c (11 ng) is separated at T = 24 °C (red curve) and a UV-chromatogram where partially resolved peptides, tagged with an asterisk, are thermally focused (= trapped and released) (purple curve) according to the temperature program depicted on the graph (blue zone (T ≈ −20 °C) and red zone (T ≈ 30 °C)). A full-loop injection was used.
of a trypic digest of cytochrome c. However, during this study, it was not attempted to thermally focus one particular component, as was illustrated in the previous paragraph, but to focus multiple components into one single peak. A temperature program (blue and red zone in Figure 6) was imposed to focus a broadly eluting peak containing the partially resolved peptides, tagged with an asterisk in the red chromatogram. The purple curve in Figure 6 shows the result. It can be seen that the different components are focused into one single sharp peak with an increased S/N ratio. This result offers perspectives to use cooling as a tool to combine different components in one concentrated plug, which is actually the reversed operation of a separation. This could be, among others, applied in LC for “heart cutting”, a technique commonly applied in GC, where multiple unresolved components from one separation are isolated and placed as a new mixture on a second column to undergo a superior separation.23 Moreover, the same principle could also be extended to a 2D-LC modulation system where the (partial) separation of the first column is fractionated and focused through repetitive cooling and warming-up cycles after which the different 2036
dx.doi.org/10.1021/ac203252u | Anal. Chem. 2012, 84, 2031−2037
Analytical Chemistry
■
Article
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS H.E. and S.E. gratefully thank the FWO (Fonds voor Wetenschappelijk Onderzoek) for a research grant. The authors gratefully acknowledge Rik Slosse, Gerd Vanhoenacker, and Frank David for the technical support and the fruitful discussions.
■
Figure 7. Maximum trapping time (MTT) as a function of the gradient time (tg) with lysozyme as test component. All data points result from triplicate measurements. Relative standard deviations were less than 8%. The full line curve is a linear regression of the experimental data.
approach, a good synchronization between the cooling and gradient program would be required.
■
REFERENCES
(1) Chen, M. H.; Horvath, C. J. Chromatogr., A 1997, 788, 51−61. (2) Lestremau, F.; Cooper, A.; Szucs, R.; David, F.; Sandra, P. J. Chromatogr., A 2006, 1109, 191−196. (3) Vanhoenacker, G.; Sandra, P. J Sep Sci. 2006, 29, 1822−1835. (4) Smith, R. M.; Burgess, R. J. J. Chromatogr., A 1997, 785, 49−55. (5) Wiese, S.; Teutenberg, T.; Schmidt, T. C. Anal. Chem. 2011, 83, 2227−2233. (6) Guillarme, D.; Heinisch, S.; Rocca, J. L. J. Chromatogr., A 2004, 1052, 39−51. (7) Guillarme, D.; T.-T. Nguyen, D.; Rudaz, S.; Veuthey, J.-L. J. Chromatogr., A 2007, 1149, 20−29. (8) Heinisch, S.; Rocca, J. L. J. Chromatogr., A 2009, 1216, 642−658. (9) Heinisch, S.; Puy, G.; Barrioulet, M.-P.; Rocca, J. L. J. Chromatogr., A 2006, 1118, 234−243. (10) Teutenberg, T. High Temperature Liquid Chromatography; RSC publishing: Cambridge, 2010. (11) Thompson, J. D.; Carr, P. W. Anal. Chem. 2002, 74, 4150−4159. (12) Sandra, P.; Sandra, K.; Pereira, A.; Vanhoenacker, G.; David, F. LC GC Eur. 2010, 23, 242−259. (13) Smith, R. M. J. Chromatogr., A 2008, 1184, 441−455. (14) Hernandez-Borges, J.; Aturki, Z.; Rocco, A.; Fanali, S. J. Sep. Sci. 2007, 30, 1589−1610. (15) Collins, D.; Nesterenko, E.; Connolly, D.; Vasquez, M.; Macka, M.; Brabazon, D.; Paull, B. Anal. Chem. 2011, 83, 4307−4313. (16) Holm, A.; Molander, P.; Lundanes, E.; Greibrokk, T. J. Sep. Sci. 2003, 26, 1147−1153. (17) Neue, U. D. HPLC Columns: Theory, Technology, and Practice; Wiley: New York, 1997. (18) Horvath, C.; Chen, H. J. Chromatogr., A 1995, 705, 3−20. (19) Greibrokk, T.; Andersen, T. J. Sep. Sci. 2001, 24, 899−909. (20) Sandra, K.; Moshir, M.; D’hondt, P.; Verleysen, K.; Kas, K.; Sandra, P. J. Chromatogr., B 2008, 866, 48−63. (21) Zhang, Z.; Pawliszyn, J. Anal. Chem. 1995, 57, 34−43. (22) Chen, Y.; Pawliszyn, J. Anal. Chem. 2006, 78, 5222−5226. (23) Deans, D. R. J. Chromatogr., A 1981, 203, 19−28. (24) Harynuk, J.; Gorecki, T. J. Chromatogr., A 2003, 1019, 53−63. (25) Marriott, P. J.; Kinghorn, R. M. J. Chromatogr., A 2000, 866, 203−212. (26) Zarzycki, P.; Woodarczyk, E.; Lou, D.; Jinno, K. Anal. Sci. 2006, 22, 453−456.
CONCLUSIONS AND FUTURE PERSPECTIVES
Up to now, the use of very low temperatures (below room temperature) has been left nearly unexplored in LC. The present study delivers a proof of principle for the fact that cooling can be used for analyte trapping (focusing) in LC. A prototype setup, capable of delivering subzero temperatures reaching below −20 °C, was constructed to trap components by cooling a small segment located at the end of a reversedphase nano-LC column. The main focus was on protein samples, although the principle was also applied to a peptide sample. These large biological components have typically large retention enthalpies, so that temperature has a major influence on their retention behavior. It was illustrated that thermal peak focusing can be applied to achieve a better peak shape and an increased S/N ratio of the different components in a separation. This could offer perspectives to use thermal focusing as a detector preconcentration technique. The same concept was furthermore used to combine different components in one concentrated band which is similar to the “cryofocusing”-technique in GC. This opens the road to use this principle in LC for equivalent purposes as in GC, including heart cutting, 2D-LC modulation, and dual-stage cooling. However, the cooling process also caused a significant increase in the overall pressure drop due to the local increase of the viscosity which can be rather seen as a drawback. In the near future, the current experimental setup is going to be optimized with respect to robustness and temperature control. The influence of the positioning distance of the “cooling chip” along the column is going to be investigated with respect to the trapping process. In the present study, methanol−water mobile phases were applied. It would, hence, also be interesting to explore ACN−water mobile phases which are commonly used in the proteomics field. However, there is one major restriction when dealing with cooling in LC, which is that the freezing points of the used mobile phases should be respected.26 The applicability of this technique for other components besides proteins and peptides is also going to be considered. 2037
dx.doi.org/10.1021/ac203252u | Anal. Chem. 2012, 84, 2031−2037
