Protein Expression and Purification Protocol: Practical Steps

Producing a purified recombinant protein involves a chain of interdependent decisions, from which organism will make the protein to how you store the final product. Each step shapes what is possible at the next, so the protocol is less a fixed recipe and more a decision tree that branches depending on the protein’s properties, the host cell, and how you plan to use the purified material. What follows is a walkthrough of the practical steps, the reasoning behind common choices, and the places where things tend to go wrong.

Picking the Right Expression Host

Your first decision is which organism will manufacture your protein. Escherichia coli remains the default for many labs because it grows fast, is cheap to culture, and has decades of optimized tooling behind it. But not every protein folds correctly in bacteria. If the protein requires complex sugar modifications or disulfide bonds to function, you may need yeast, insect cells, or mammalian cell lines. A practical decision scheme walks researchers through four key questions to match the target protein’s characteristics with the best-fit expression system.¹ The most common mistake at this stage is defaulting to E. coli out of habit when the protein’s biology demands a eukaryotic host, or jumping to an expensive mammalian system when bacteria would work fine.

Construct Design and Codon Optimization

Once you have a host, you need a DNA construct that tells the host’s cellular machinery how to build your protein. Two design decisions matter most here: codon optimization and the choice of purification tag.

Every organism has preferred codons for each amino acid. When you take a gene from, say, a malaria parasite and try to express it in E. coli, the bacterial machinery stumbles over codons it rarely uses. Redesigning the gene to use the host’s preferred codons can boost yields dramatically. In one case, codon optimization of a Plasmodium falciparum vaccine candidate gene increased protein expression at least threefold in E. coli.² The approach matters, though. A study comparing two codon-optimization strategies found that a randomization-based method produced significantly more protein than the native sequence, with one gene showing a 70% increase, while a simpler one-amino-acid-one-codon method showed no significant improvement at all.³ Outside E. coli, similar principles apply. Codon optimization for actinobacteria improved expression in 75% of tested genes at one threshold setting, and for all five genes that originally showed low or undetectable expression, optimized sequences rescued production entirely.⁴

The second construct-level choice is which affinity tag to fuse onto your protein. Tags are short peptide sequences or small protein domains added to the gene so the resulting fusion protein can be grabbed out of a cell extract using a matching resin. The polyhistidine tag (His-tag) is the most popular because the resins are cheap and high-capacity, but it delivers only moderate purity from E. coli extracts and performs worse with yeast or mammalian extracts. Epitope-based tags like FLAG and HPC produce the highest purity across extract types but use expensive, low-capacity resins. The Strep II tag has been flagged as a reasonable middle ground, offering strong purification at moderate cost.⁵ The tag’s identity and its position on the protein (attached at the front or the back) both affect expression levels, solubility, and purification yield, and these effects vary depending on the protein itself, so there is no universal best answer.⁶

Growing Cells and Inducing Expression

With a construct in hand and transformed into your host, the next step is growing enough cells and triggering them to make the protein. For E. coli, the standard approach uses IPTG as an inducer, but auto-induction media offer a hands-off alternative. In auto-induction, the medium itself contains a mix of carbon sources that first support growth and then automatically switch the cells into protein-production mode. One study found that optimizing auto-induction conditions, specifically lowering the growth temperature and using an enriched medium with trace elements, increased biologically active protein production by at least threefold.⁷

Temperature during induction deserves extra attention. Higher temperatures push cells to make more protein faster, but the protein often misfolds and aggregates. Lowering the temperature to 18–25°C slows production but gives the folding machinery more time to get the structure right. That same reteplase study saw a 60% jump in active protein when the temperature dropped to 18°C compared to 37°C. If you are making a protein that needs to be correctly folded and functional, running a cooler induction is one of the simplest improvements available.

Breaking Cells Open

Your protein is now inside the cells, and you need to get it out. The main options are mechanical disruption (high-pressure homogenization, sonication) and gentler approaches (enzymatic lysis, osmotic shock, freeze-thaw). The aggressive methods generally give higher yields at lower cost and scale up more easily.⁸ Sonication works well at bench scale. You place your cell suspension on ice and pulse ultrasound through it. The main risk is overheating the sample, which denatures protein. Keeping the tip submerged, using short pulses with rest intervals, and keeping everything cold handles that.

After lysis, a centrifugation step separates the soluble fraction (where your correctly folded protein should be) from the insoluble pellet. If your protein ended up soluble, you proceed to purification. If not, you have inclusion bodies on your hands.

When Your Protein Forms Inclusion Bodies

Inclusion bodies are dense aggregates of misfolded protein that collect in the insoluble pellet. They are frustrating but not necessarily a dead end. The aggregated protein molecules often retain native-like secondary structure, which means mild solubilization can sometimes unfold them just enough to let them refold correctly.⁹ The classical approach uses strong denaturants like urea or guanidinium chloride to fully dissolve the aggregates, followed by a slow refolding step where the denaturant is gradually removed.¹⁰

A newer line of work uses detergents instead of harsh denaturants. Ionic detergents like sodium dodecyl sulfate, sarkosyl, and lauroyl-glutamate can solubilize inclusion bodies effectively, and slow or stepwise removal of the detergent during refolding appears critical to keeping the protein soluble as it regains structure.¹¹ Refolding is still the bottleneck, though. Yields are protein-dependent, and screening multiple buffer conditions is standard practice. If you are spending weeks trying to refold a stubborn protein from inclusion bodies, it may be time to revisit the expression conditions (lower temperature, different host, solubility-enhancing fusion partners) rather than continuing to optimize refolding.

Affinity Capture

The first real purification step for most workflows is affinity chromatography, where your tagged protein binds a resin and everything else washes through. For a His-tagged protein, this means running the cell lysate over a nickel or cobalt resin column, washing to remove loosely bound contaminants, and eluting with imidazole, which competes for the metal-binding sites. The entire capture step can take under an hour and typically gives you protein that is 80–90% pure, depending on the tag and the extract. As noted earlier, the purity you get varies with the tag system and the source organism.

If the downstream application requires untagged protein, you can cleave the tag after affinity capture using a site-specific protease. TEV protease is the most widely used, but newer proteases are dramatically faster. One study reported engineered tag-cleaving proteases that process a 50-fold to 10,000-fold molar excess of substrate per hour at 0°C, making them up to 1,000-fold more active than TEV, with some variants working efficiently even in buffers containing 1 M salt.¹² After cleavage, a second pass over the affinity resin (reverse purification) lets the cleaved tag and any uncleaved fusion protein stick while the clean target flows through.

Polishing with Ion Exchange and Size Exclusion

Affinity capture gets you most of the way, but remaining host-cell proteins, degradation products, and aggregates often demand a second or third chromatography step. Ion exchange chromatography separates proteins by surface charge. You pick a cation or anion exchange resin depending on whether your protein is positively or negatively charged at the buffer pH you choose.¹³ A useful shortcut for basic proteins expressed in E. coli: because the host’s own proteins are mostly acidic, setting the buffer pH just one unit below the target protein’s isoelectric point causes the recombinant protein to bind a cation exchanger while most host proteins flow through.¹⁴

Size exclusion chromatography separates molecules by size. It is widely used both as a polishing step and as an analytical tool for assessing aggregation in biopharmaceuticals.¹⁵ Running your purified sample through a size exclusion column also doubles as a buffer exchange step, swapping the protein into whatever final buffer you need. The trade-off is that size exclusion dilutes your sample, so you may need a concentration step afterward.

Endotoxin Removal

If you are purifying from E. coli and the protein will be used in cell-based assays or animal studies, endotoxin contamination is a real problem. Endotoxins are lipopolysaccharide fragments from the bacterial outer membrane, and even trace amounts provoke strong immune responses. Standard affinity purification does not reliably remove them.

One efficient approach combines affinity chromatography with a non-ionic detergent wash step, reducing endotoxin levels to below 0.2 endotoxin units per milligram of protein while recovering close to 100% of the target protein.¹⁶ Phase separation with the detergent Triton X-114 is another well-established method that can reduce endotoxin levels by more than 99% with protein recovery above 90%.¹⁷ A more recent protocol for endotoxin-free human tau protein demonstrated on-column Triton X-114 depletion integrated directly into the purification workflow, combining capture and endotoxin removal into a single chromatographic run.¹⁸ Whichever method you use, always measure endotoxin in the final product with a LAL (Limulus amebocyte lysate) assay or equivalent. Assuming your purification removed it is not enough.

Measuring Concentration and Assessing Quality

After purification, you need to know how much protein you have and whether it is what you think it is. Concentration measurement sounds simple but has pitfalls. A systematic comparison of spectrophotometric and colorimetric assays found that the Bradford method was the most sensitive for purified protein, with the widest range of detectability, while lysis buffers commonly used in protein preparation dramatically interfered with all tested assays.¹⁹ For modified or cross-linked proteins, dye-binding assays can underestimate actual concentration by 50–60%, so using at least two independent methods is a good idea.²⁰

Beyond concentration, you want to confirm identity (mass spectrometry), purity (SDS-PAGE, analytical size exclusion), and stability. The thermal shift assay, which uses a fluorescent dye to monitor how a protein unfolds with increasing temperature, has become a quick screen for assessing how stable a protein is in different buffers, salt concentrations, and with various additives.²¹ Running this assay on your purified batch takes less than an hour and can flag problems before you commit the protein to a weeks-long experiment.

Storage and Stabilization

You have clean protein. Now you need to keep it that way. Proteins degrade, aggregate, and lose activity over time, and how you store them matters enormously. For short-term use (days to a couple of weeks), 4°C in an appropriate buffer is usually fine. For longer storage, flash-freezing in liquid nitrogen and storing at −80°C is standard.

Freezing itself is a stress event. Ice crystals form, concentrating solutes at the interfaces and exposing protein to destabilizing conditions. Controlling freezing and thawing rates and adding cryoprotectants minimizes this damage.²² Sucrose and glycerol consistently outperform other cryoprotectants. Glycerol also showed the best performance for heat-induced reversibility of aggregates after freeze-thaw cycling.²³ Adding 10–20% glycerol or 5–10% sucrose to your storage buffer before freezing is a simple step that many protocols skip, and it makes a real difference in recovered activity after thawing. Aliquoting into single-use volumes avoids repeated freeze-thaw cycles, which compound the damage.

Glycerol serves double duty: beyond cryoprotection, it stabilizes proteins in liquid solution by shifting the native ensemble toward more compact states and inhibiting aggregation during refolding.²⁴ If your protein is aggregation-prone even before freezing, glycerol in the working buffer can help.

Membrane Proteins Are a Different Game

Everything discussed so far assumes a soluble protein. Membrane proteins live embedded in lipid bilayers, and pulling them out of that environment without destroying their structure requires detergents. The challenge is finding a detergent that solubilizes the protein but does not strip away the structural lipids it needs. High-throughput detergent screening has shown that maltose neopentyl glycol detergents, particularly LMNG (Lauryl Maltose Neopentyl Glycol), stabilize a range of transporter proteins, producing positive shifts in both thermal melting temperature and aggregation temperature. PEG-family detergents, by contrast, showed a major destabilizing effect on the same proteins, with some apparently denaturing them at room temperature.²⁵ The practical message is that detergent choice is not a minor detail: the wrong detergent will give you aggregated, inactive material even if every other step in the protocol is optimized.

Expressing Toxic Proteins

Some proteins kill or inhibit their host cells even at trace levels. Standard expression systems, which always leak a small amount of protein before induction, can cause plasmid loss or cell death during overnight growth. For these targets, conventional transcriptional control is not tight enough. Dual transcriptional-translational control systems add a second layer of suppression at the translation step, reducing leakage expression to near zero.²⁶ An alternative called “restrained expression” exploits the natural behavior of the Lac repressor, limiting transcription initiation frequency by keeping the repressor bound in the absence of inducer and restricting T7 RNA polymerase levels using a tightly regulated arabinose promoter.²⁷ If you are struggling with overnight cultures that keep losing your plasmid, toxicity is worth investigating even if the protein is not supposed to be toxic. Low-level leakage can select against the expression plasmid without obvious signs of cell death.

Scaling Through Automation

When you need to screen many constructs, optimize conditions, or just produce a large panel of proteins, doing everything by hand becomes impractical. Automated pipetting robots have been adapted for parallel expression and purification in 96-well microplate format, yielding up to 90 micrograms of purified protein per milliliter of culture with 96 proteins purifiable in roughly three hours.²⁸ For antibody discovery programs, automated liquid handling workflows have pushed throughput to over 2,000 purified antibodies per day from one-milliliter cultures.²⁹

Even without a full robotic setup, borrowing the logic of high-throughput screening improves manual workflows. Expressing a panel of constructs (varying tags, codon variants, truncations) in parallel at small scale, then committing to large-scale production only for the winners, saves time and reagents. The small-scale screen does not need to reproduce the final purification protocol exactly. It just needs to tell you which constructs express soluble protein and which do not.³⁰