This Month's Feature

Separating Proteins by pI-Values - Can 2D LC Replace 2D GE? (continued)

Alternatives to 2D GE Depending on the type of sample and the amount of protein available, there are at least two main alternative routes. With plenty of sample at hand, off-line separations are simpler and often an advantage as the first dimension. Isoelectric focusing can separate proteins according to pI values and has many advantages as a prefractionation technique (7). However, with very limited sample amounts, on-line techniques allow transfer of the whole fraction to the second dimension, and with miniaturized columns, little band broadening is obtained. Unfortunately, isoelectric focusing is not easily included in an on-line 2D instrumentation with microbore columns. In samples containing relatively few proteins, the "shotgun" approach is an alternative, by enzymatic degradation of the mixture of proteins to peptides and coupled LC–MS-MS or MALDI-TOF MS analyses of the peptide mixtures for protein identification. Even with some overlap of peptides, the protein databases have the potential of identifying the majority of the proteins, depending upon the concentration. Samples containing more complex sources of proteins always benefit from a preseparation of the proteins prior to enzymatic degradation, either in one dimension or in two dimensions. With a 2D system, the two separation principles should be orthogonal, such as a combination of pI separation and reversed-phase chromatography or hydrophobic interaction chromatography. For peptides, a combination of cation exchange and reversed phase is common today. The separation of proteins on cation exchangers with salt gradients has some disadvantages; foremost that there is no direct relationship between the pI and the salt concentration and the retention time. Eluting proteins by a pH gradient would be much more advantageous, because the pH of each eluting peak in principle can be measured directly, resulting in a direct link between eluting pH and pI of the proteins. Unfortunately, common pH gradients always have been known for low repeatability and nonlinearity. Thus, an important question is whether a separation according to pI can be implemented as the first dimension of a 2D LC separation. Separations According to pI  In the late 1970s, Sluyterman and colleagues (8,9) described a pH-gradient ion exchange chromatography (IEC) method that used the buffering capacity of the column to generate linear outlet pH-gradients without external mixing of the buffers, which they termed chromatofocusing (CF). CF is most commonly performed by titration of a weak anion-exchange column, (e.g., DEAE(diethylaminoethyl)-functionalized cellulose), which initially has been equilibrated with a start buffer at a pH higher than the pI of the most basic proteins, and is eluted with a buffer of low pH. To obtain as linear pH-gradients as possible, and larger peak capacity, the buffers usually contain mixtures of polyampholytes with different pKa values to get an even buffering capacity over the chosen pH range. However, CF has several limitations; limited reproducibility because of batch-to-batch variability in the chemical composition of polyampholyte mixtures, formation of association complexes with proteins (10), high background absorption with UV detection, and little flexibility in choice of buffer concentrations. While CF generally exploits the buffering capacity of the ion exchanger to obtain a retained intracolumn pH-gradient, the opposite is true for pH-gradient ion exchange. Liu and Anderson (11) and Shan and Anderson (12) used pH-gradient ion exchange chromatography (IEC), which they termed gradient CF, to overcome many of the shortcomings of CF. They used ion-exchange materials with small buffering capacities in the applied pH range in combination with buffer components that are not adsorbed on the ion-exchange column, (e.g., amine buffering species on anion-exchange materials). Under such conditions, the contribution from the column itself to the delay of the pH-gradient is minimal. Consequently, the outlet pH-gradient will roughly reflect the pump gradient settings, which gives enhanced flexibility in controlling the slope of the pH-gradient (11,12). In contrast to CF, pH-gradient IEC also allows the use of higher buffer concentrations without affecting the slope of the pH-gradient. Accordingly, improved chromatographic performance can be obtained with only a few common buffer components, while still attaining the characteristic focusing effect of the protein bands. Recently, Andersen and colleagues developed a pI separation of proteins on packed capillary columns using a strong anion exchanger, which improved the peak shape of the proteins while maintaining the focusing effect (Figure 1) (13). In the weakly basic to acidic pH range, a linear pH curve was obtained (Figure 2), which could be used to determine b-lactoglobulins in skimmed milk (Figure 3).  The pH monitoring was performed by a flow-through cell with a pH electrode (13). Unfortunately, the cell was not sufficiently robust in the long run. Another smaller electrode was built into a housing for microlitre flow but the response time of the pH electrode was much too long. Thus, we decided that a larger volume pH electrode was needed in order to give fast response, requiring 1–2 mm i.d. ion exchange columns in order to avoid too much peak broadening. It is a basic fact of this kind of pH gradients that a linear pH curve is obtained only in a range outside the buffering area of the ion exchanger. Thus, starting with strongly basic pH led to a nonlinear pH curve because of titration of the ion exchanger. This is, however, not necessarily a problem, as long as it is reproducible (Figure 4). A mixture of acidic and basic proteins in a wide range of pI values could be separated easily in two groups by a wide range pH gradient (Figure 5). In order to avoid precipitation of hydrophobic proteins at their pI values and to reduce potential losses by adsorption, trifluoroethanol (14) was added to the mobile phase in 2–20%, but with disastrous results. The peak shape of the proteins was strongly malformed with strong peak broadening. Replacing trifluoroethanol with ethanol, however, gave more symmetric and narrow peak shapes. The addition of 20% or more of ethanol dehydrated the pH electrode, but 10% could be used without problems, as shown in Figure 5. (continued)