Is your protein not sticking to the ion exchange column when you thought it should have? As frustrating as this result is, every experienced protein purifier runs into this problem sooner or later.
In a previous article we discussed how online programs calculate a protein’s isoelectric point, and how that information informs on conditions that should work well for purifying your protein with ion exchange chromatography.
Technically speaking, however, a protein’s isoelectric point (pI) is almost always different than its calculated value. Usually, this difference is subtle enough that the calculated pI is still a useful approximation for purposes such as ion exchange purification.
However, a protein’s isoelectric point (pI) will occasionally deviate substantially from calculated values due to posttranslational modifications, asymmetric distribution of charged residues, or significant differences in the acid dissociation constants of amino acids compared to reference values.
Now, before we get into the details for these discrepancies, I want to emphasize that estimating the isoelectric point is still a great first step when working with a new protein that you have never purified before. In my personal experience, this approach works most of the time (greater than 90%).
This article assumes that you are familiar with the basics of calculating protein charge and how that charge depends on the pH of the surrounding environment. If you’re feeling a little fuzzy on these concepts such as acid dissociation constants (Ka), closely related pKa values, and protein isoelectric points (pI), head on over to this related article.
Ok, if you’re feeling good about these terms, let’s dive deeper into three common reasons that pI calculators may miss on your protein:
- pKa of one or more ionizable residues deviates from reference values
- Asymmetric distribution of charged residues
- Posttranslational modifications
In this article:
Amino acid pKas Deviate from Reference Values
Asymmetric Distribution of Charged Residues
Posttranslational Modifications
Amino Acid pKas Deviate from Reference Values
Protein isoelectric point calculators treat amino acids as independent variables. That is to say, they combine the pKa values for all of the ionizable residues in the input protein sequence with little regard for the ordering or the structural context of where those ionizable residues are located on the actual protein residue.
A visual way of thinking about this is the calculator essentially treats your protein like a bowl of alphabet soup with individual amino acids floating around on their own (Figure 1). In reality, the ordering and positioning of amino acids within a protein is extremely important, including for that protein’s actual charge.
Figure 1.
Protein sequence and structure (PDB: 5ILS) influences the pKa
values of individual amino acids. However, the pI calculator’s view of a
protein sequence is more like an alphabet soup of independent amino acids. In
this view, the amino acids that make up a protein are free floating, and they
do not impact each other’s p
Ka values.
To put it succinctly, the pKa values of ionizable residues are not independent. The local environment of an amino acid, which largely depends on which other amino acids that are nearby in 3D space, has a large impact on pKa values. In fact, for individual amino acids, the pKa values can differ from the reference values by greater than 6 pH units depending on their local environment (Harris & Turner, 2002). Remember, pH is a log scale meaning that a 6 pH unit difference corresponds to a million-fold difference!
Since individual pKa residues can be wrong by so much, you might be wondering why pI calculators work well enough for so many cases.
One reason is that while individual residues can have large differences in pKa values, the majority of ionizable residues will have pKa values close to their reference values (Harris & Turner, 2002). These close-to-reference residues will average out the one, or few, residues with big pKa differences that any individual protein might have.
Another reason is that we intentionally leave some room for small deviations in how different we make the buffer pH compared to the estimated protein pI.
I described hypothetical ion exchange conditions in this article using buffers that are 3 pH units more acidic or basic compared to the protein’s pI. This example is a little overzealous in how far we changed the pH. In general, it is good practice to use buffers that are at least 1-2 pH units different than the protein’s isoelectric point. This difference will leave sufficient room for our purification strategy to still work even if the pI estimate is a little bit off.
Asymmetric Distribution of Charged Residues
By now you can probably appreciate that while a pI calculator will predict the net charge of a protein, it won’t tell us where those charges are on the protein.
Let’s consider two boundary cases to illustrate this point. First, imagine a situation where charge, both positive and negative, is uniformly distributed throughout a protein (Figure 2, left panel). Alternatively, consider a protein where all positive charges are clustered on one part of a protein, and all negative charges are clustered on the opposite side of the protein (Figure 2, right panel). Both examples have the same net charge, yet as you can see their surface charge properties will be very different!
Figure 2. Both proteins have a net charge of +4, but in one of the proteins, both positive and negative charges are distributed throughout the protein (left), and in the other protein, positive and negative charges are clustered on opposite sides of the protein (right).
In this example, I arbitrarily assigned both proteins four more positive charges than negative proteins.
For the protein with a homogenous charge distribution, we would definitely want to use cation-exchange chromatography since the protein is net positively charged.
For the protein with extreme asymmetric charge distribution, however, we could likely use either cation-exchange or anion-exchange chromatography since the protein has both a positive patch and a negative patch on its surface.
Clustering negative charges often makes the pKa values for negative residues more basic, however (Harris & Turner, 2002). Therefore, if we were doing anion-exchange chromatography for this protein, we would want to give a little bit more of a gap than the normal 1-2 pH unit difference discussed above.
Posttranslational Modifications
Some posttranslational modifications change the charge of a protein. As the name indicates, posttranslational modifications are chemical changes that occur to the side chains of select amino acids after the protein has been translated. Two common posttranslational modifications that modify the charge of a protein are phosphorylation and acetylation (Figure 3).
Phosphorylation occurs on a number of different amino acids, with serine, threonine, and tyrosine residues being the most frequently phosphorylated residues (Figure 3). The addition of a phosphate group changes these residues from being neutral to having a charge of -2. As a single change, this might not seem like much of a difference, but proteins often have several phosphorylation sites, so the negative charges can add up quickly (Hardman et al, 2019).
Acetylation is another posttranslational modification that changes a protein’s charge. Acetylation most often happens on lysine residues, which converts them from being positively charged to being neutral. Acetylation happens on many different types of proteins but may be most well characterized for the changes in gene expression due to acetylation (and deacetylation) on histone proteins.
Previously, I covered the relationship between histone acetylation and gene expression, how this modification becomes mis-regulated in many types of cancers and how proteins that write, erase, and read acetylation marks are attractive drug targets. Check out that article for more information about this fascinating modification!
Posttranslational modifications such as phosphorylation and acetylation usually do not happen when expressing a eukaryotic protein recombinantly in Escherichia coli, unless you are using a slick expression setup to intentionally generate these modifications. However, if you are expressing and purifying a protein from a eukaryotic source such as yeast, insect, or mammalian cells, then you’ll need to keep in mind that your protein might come preloaded with posttranslational modifications.
Figure 3.
Phosphorylation (left) and acetylation (right) are two common posttranslational
modifications that change a protein’s charge.
Practical Suggestions
So, what should you do if your protein doesn’t stick to the ion exchange column when you thought it should have?
One option is to adjust the pH buffer further away from the estimated protein pI. So, for example, if your protein’s pI is 6 and you used a protein purification buffer with pH 7 thinking your protein would stick to an anion exchange column, try making the pH of your buffer more basic, like 8 or 8.5, and try again with anion exchange. This would be the right step to take if you think you were too close with your buffer’s pH relative to the protein’s pI.
Alternatively, keep the buffers the same and try a different column. Extending the same example from the previous paragraph, this would mean keeping your buffers at pH 7 and seeing if your protein will bind to a cation exchange column instead. This might work if your protein is negatively charged overall, but the charges are asymmetrically distributed, and the protein has a high density of positive charges on one part of its surface (Figure 2). Are there any structural data or predictions for your protein that could help you decide if this might be the case?
You can always switch both the buffers and column. Continuing our example, switch the buffer pH to 5 or 4.5 and try binding your protein to a cation exchange column. Sometimes even in cases when your protein does stick to one ion exchange column, tinkering with another column type and buffer conditions can help improve purity.
The advantage of these approaches is that they don’t require much thinking, and if you have an automated protein purification system, they also don’t require much work on your part. If it’s the end of a long hard day in the lab, and you don’t know why your protein didn’t stick to that dang column, just switch columns, buffers, or both really quick and set it up to run overnight while you go home and enjoy your dinner.
If those strategies don’t work, you unfortunately might have to do a little more thinking and a little more work.
Can you validate that it is indeed your protein that you’re purifying? For example, if your protein binds to DNA, or has an enzymatic activity, take your protein that won’t stick to the column and test to see if it possesses that activity.
Mass spectrometry can be helpful here to confirm that the protein you are working with has the same molecular weight as your target protein. Mass spec may also identify if your protein has a posttranslational modification that you don’t know about. Alternatively, many proteins have antibodies that recognize specific phosphorylation sites, and there are also reagents such as Phos-tagTM that broadly recognize phosphorylation on a protein (Kinoshita et al, 2006).
Which of these troubleshooting strategies will work best in your case? Carefully consider your protein, what you know about it, and how you have tried to purify it so far. Then carefully prioritize which approach to try first based on your intuition and your schedule.
As you can see there are quite a few ways that protein isoelectric point calculations can go wrong! You’ll probably agree with me that we should think of these tools more as pI estimators rather than calculators.
However, don’t forget that these estimators work really well most of the time. When you run into a fringe case where it doesn’t, come back here to remember why that might be.
References
Berman, H. M., Westbrook, J., Feng, Z., Gilliland, G., Bhat, T. N., Weissig, H., Shindyalov, I. N., & Bourne, P. E. (2000). The Protein Data Bank. Nucleic acids research, 28(1), 235–242. https://doi.org/10.1093/nar/28.1.235
Berman, H., Henrick, K., & Nakamura, H. (2003). Announcing the worldwide Protein Data Bank. Nature structural biology, 10(12), 980. https://doi.org/10.1038/nsb1203-980
Currie, S. L., Lau, D. K. W., Doane, J. J., Whitby, F. G., Okon, M., McIntosh, L. P., & Graves, B. J. (2017). Structured and disordered regions cooperatively mediate DNA-binding autoinhibition of ETS factors ETV1, ETV4 and ETV5. Nucleic acids research, 45(5), 2223–2241. https://doi.org/10.1093/nar/gkx068
Hardman, G., Perkins, S., Brownridge, P. J., Clarke, C. J., Byrne, D. P., Campbell, A. E., Kalyuzhnyy, A., Myall, A., Eyers, P. A., Jones, A. R., & Eyers, C. E. (2019). Strong anion exchange-mediated phosphoproteomics reveals extensive human non-canonical phosphorylation. The EMBO journal, 38(21), e100847. https://doi.org/10.15252/embj.2018100847
Harris, T. K., & Turner, G. J. (2002). Structural basis of perturbed pKa values of catalytic groups in enzyme active sites. IUBMB Life 53, 85–98.
Kinoshita, E., Kinoshita-Kikuta, E., Takiyama, K., & Koike, T. (2006). Phosphate-binding tag, a new tool to visualize phosphorylated proteins. Molecular & cellular proteomics : MCP, 5(4), 749–757. https://doi.org/10.1074/mcp.T500024-MCP200
The PyMOL Molecular Graphics System, Version 2.5.2 Schrödinger, LLC.
.jpg?t=1714409158)
