Getting Practical with Separation Methods
I've been running gels and resins long enough that I've stopped caring about the textbook definitions. The column doesn't care what you think the mechanism is. It cares about flow rate, buffer matching, and whether you bothered to degas your eluent. Here is how I actually set up a run. The basic concept is straightforward enough that even undergraduates understand it within ten minutes. Your sample enters a column packed with porous beads. Large molecules cannot fit into the pores, so they travel through the interstitial space between beads and exit first. Small molecules slip into the pores and get temporarily trapped, which slows their progress through the column. Medium-sized molecules do both, depending on their hydrodynamic radius. That is the theory. The reality is messier. You are not just separating by size. You are dealing with secondary interactions, column overloading, and the fact that your buffer conductivity will shift the elution profile more than you expect.
Common misconception: Size exclusion is completely non-interacting. It is not. If your resin has any residual charge or if your sample contains unbound salt at the wrong concentration, you will see peak tailing that has nothing to do with molecular weight. I have seen people blame their column when the real problem was a poorly prepared loading buffer.
Setting Up a Run
First thing you need is a properly conditioned column. For Sephadex G-75, which I use most often for protein desalting and buffer exchange, that means running at least three column volumes of starting buffer through the bed at a flow rate of 0.5 to 1.0 mL/min for a standard analytical column. Do not skip this step even if the manufacturer says the column is pre-equilibrated. They ship it in ethanol or preservative, and that preservative will ruin your samples if you load directly. Let me walk you through the actual procedure I follow every time. Prepare your sample. Clarify it by centrifugation at 12,000 times g for ten minutes. Filtered samples cause column fouling that shows up as increased backpressure after just two or three runs. I typically load no more than two percent of the bed volume. If you overload the column, your resolution drops sharply and peaks start merging together. This is one of the most common mistakes I see people make, especially when they are working with dilute samples and want to squeeze everything through at once.
Get the Full Details
Apply the sample carefully. Use a syringe or a pipette and deposit it gently onto the center of the bead surface. Do not disturb the bed. Once the entire sample has entered the gel, follow immediately with about one to two milliliters of the same starting buffer. This push-through ensures complete sample transfer without creating voids or disrupting the packing. Run the column at a constant flow rate. For a 1.6 by 60 centimeter Superdex 75 column, I use 0.5 mL/min. Faster flow rates reduce resolution because they do not give small molecules enough time to equilibrate between the pore and interstitial phases. Slower rates improve separation but take longer and allow for more diffusion-based band broadening. There is a sweet spot, and for most proteins in the 10 to 70 kilodalton range, 0.3 to 0.8 mL/min works fine. Collect fractions. I usually collect 0.5 mL fractions and monitor absorbance at 280 nanometers if you are working with proteins. UV detection is standard, but if your molecules do not absorb well at that wavelength, switch to refractive index or try a post-column derivatization step. Each approach has trade-offs. RI is universal but less sensitive. Fluorescence detection requires tryptophan or tyrosine residues or exogenous labeling.
Calibration and Molecular Weight Estimation
Before you trust any molecular weight calculation from a gel filtration run, calibrate the column with standards. Standard mixtures like thyroglobulin at 669 kilodaltons, ferritin at 440 kiladaltons, aldolase at 158 kilodaltons, ovalbumin at 44 kilodaltons, and ribonuclease A at 13.7 kilodaltons cover a useful range. Plot the logarithm of molecular weight against the partition coefficient Kav, which equals Ve minus V0 divided by Vt minus V0. Ve is your elution volume, V0 is the void volume determined by blue dextran or sodium nitrate, and Vt is the total column volume. Here is something most protocols do not mention clearly. The relationship between Kav and molecular weight is only linear across a restricted range for any given resin. Outside that range, your estimates become unreliable. If your protein runs near the exclusion limit, small changes in shape or glycosylation can shift the apparent weight by tens of kilodaltons. If it runs near the permeation limit, you are essentially measuring hydrodynamic volume, not mass, and compact globular proteins will elute later than extended multimers of the same mass. I learned this the hard way with a recombinant enzyme that my calibration curve suggested was 45 kilodaltons, but analytical ultracentrifugation showed it was actually a dimer of 28 kilodalton subunits held together by a loose interface. The gel filtration peak looked clean and symmetric. Everything about it said monomer. Shape matters more than you think.
Practical Issues I Have Dealt With
Here is a specific problem I ran into last year that took me about a week to resolve properly. I was running Size Exclusion Gel Filtration on a glycosylated receptor fragment using a Superdex 200 Increase column in phosphate-buffered saline with 150 millimolar sodium chloride and 0.02 percent NaN3. The protein was supposed to be around 85 kilodaltons based on the sequence, but the main peak consistently eluted at a volume corresponding to roughly 120 kilodaltons on my standard curve. I ruled out aggregation by running the sample at multiple concentrations. The elution volume did not shift. I tried reducing the salt concentration and adding a low percentage of glycerol. Same result. Then I checked the buffer pH. My stock buffer was at pH 7.4, but the column effluent was measuring at 6.8 due to CO2 absorption over the two-hour run. That pH drop was causing a conformational change in the receptor fragment that expanded its hydrodynamic radius enough to shift the elution profile significantly. The workaround was simple once I identified it. I bubbled the buffer with nitrogen before use, sealed the column reservoir to minimize air exposure, and added 10 millimolar HEPES as a buffering agent to stabilize pH throughout the run. After that, the peak eluted consistently at the volume expected for an 85 kilodalton species. I also re-calibrated the column with the new buffer system because Kav values shift slightly with buffer composition due to changes in bead swelling and ionic strength effects on the double layer around the resin surface.
Resolution Limits and When This Method Fails
Size exclusion chromatography has real limitations that people tend to gloss over in method sections. It provides low resolution compared to ion exchange or reverse-phase methods. Two proteins that differ by only 10 percent in molecular weight may not separate on a standard column. If you need finer discrimination, you have to use a longer column, a resin with a narrower fractionation range, or a smaller particle size, each of which increases run time and backpressure. The method also does not work well for very large complexes above the fractionation range of your resin. Those molecules elute in the void volume and you cannot distinguish between a 500 kilodalton tetramer and a 2 megadalton aggregate. Conversely, small molecules below the fractionation range all elute near the total permeation volume and come out as a broad unresolved peak. Buffer exchange is fine for this range, but sizing is not. Sample viscosity is another issue. If your loading buffer has a significantly different viscosity from the running buffer, you will get band distortion and peak broadening. I always match buffers exactly, including glycerol and detergent concentrations. Even a 5 percent glycerol mismatch can broaden a sharp peak noticeably.
Column degradation over time is real. Repeated cycling between different salt concentrations, exposure to oxygen, and occasional contamination from dirty samples all reduce bed integrity. My Superdex 75 columns typically last for about 200 to 300 runs if I am careful, after which I start seeing increased pressure and progressive peak tailing. I clean in place with 0.5 molar sodium hydroxide for 30 minutes monthly and store the column in 20 percent ethanol. That extends column life substantially.
Key Parameters to Monitor
Keep track of these during every run. Backpressure tells you about column health. If it rises by more than 20 percent from your baseline, something is clogging the frit or the bed is compacting. Flow rate consistency matters more than speed. An unstable pump will produce erratic elution profiles that ruin quantification. Temperature control is often overlooked. A 5 degree Celsius change can shift elution volumes by 1 to 2 percent on agarose-based resins due to thermal expansion of the gel matrix and changes in buffer viscosity. If you are doing preparative work rather than analytical, remember that concentration factors apply. The peak you collect is more concentrated than your starting sample because the column effectively compresses the elution volume. Factor that into your downstream applications. A typical 1.6 by 30 centimeter column with a 6 milliliter bed volume will elute a 1 milligram sample in about 8 to 10 milliliters of buffer, giving you roughly a two-fold concentration increase if your original sample was in 15 milliliters. Gel filtration is a workhorse method. It is not elegant, it is not high resolution, and it will not save you from poor sample preparation. But when you understand what it is actually doing and respect its constraints, it gives you reliable separation, buffer exchange, and rough molecular weight estimates in a single pass. I have never found a better way to polish a sample before mass spectrometry or crystallization trials. The runs are predictable, the buffers are compatible with most downstream applications, and the columns are reusable enough that the cost per separation is low.
