Selection Antibiotics for Mammalian Cell Lines

Selection antibiotics do the opposite job to the penicillin–streptomycin in a maintenance medium. Culture antibiotics are used at concentrations mammalian cells tolerate, to suppress bacteria. Selection antibiotics are used at concentrations that kill mammalian cells, to remove every cell in the dish that has not acquired a specific resistance gene.

The workflow is always the same. A resistance gene is delivered on the same construct as your gene of interest — by plasmid transfection, lentiviral or retroviral transduction, or a targeted knock-in. A few days later the selection agent goes into the medium. Cells expressing the resistance gene survive; everything else dies. What remains is a polyclonal resistant population, which can then be single-cell cloned into a stable line.

Two things determine whether this works: choosing an agent whose resistance gene is on your construct, and using the right concentration. The second is where most selections fail, and it is why the kill curve below is not optional.

A note on nomenclature that causes real confusion in ordering: G418, geneticin and G-418 sulfate are the same aminoglycoside. Zeocin and phleomycin D1 are the same glycopeptide from the bleomycin family. Blasticidin and blasticidin S are the same peptidyl nucleoside.

G418 / geneticin is an aminoglycoside from Micromonospora rhodorangea. It blocks polypeptide synthesis by interfering with 80S ribosomal function, so it is toxic to eukaryotic as well as prokaryotic cells. Resistance comes from neo (also called nptII), an aminoglycoside 3'-phosphotransferase from transposon Tn5, which phosphorylates and inactivates the drug. G418 is the slowest of the five: expect 7–14 days for complete selection. One practical trap — powdered G418 is sold with a stated potency in µg of active drug per mg of powder, typically well under 1000 µg/mg, and it varies by lot. Concentrations must be calculated on the active basis, not on powder weight. Ready-made solutions (for example 50 mg/mL) remove this problem entirely.

Hygromycin B is an aminoglycoside from Streptomyces hygroscopicus that inhibits protein synthesis by disrupting translocation on the 80S ribosome and promoting mistranslation. Resistance comes from hph (also written hyg), hygromycin B phosphotransferase, originally from E. coli. Working range is broad — 50–1000 µg/mL — and selection typically takes 7–14 days.

Blasticidin S is a peptidyl nucleoside from Streptomyces griseochromogenes that inhibits peptide bond formation on both 70S and 80S ribosomes. Two resistance genes are in common use: bsd, a blasticidin S deaminase from Aspergillus terreus, and bsr, a deaminase from Bacillus cereus. Both convert blasticidin to a non-toxic deaminated form. Blasticidin works at by far the lowest concentrations of the five — 1–20 µg/mL, commonly 2–10 — and kills quickly, with morphological changes often visible within 48 hours and selection complete in about 3–10 days.

Zeocin (phleomycin D1) is a copper-chelated glycopeptide of the bleomycin family from Streptomyces verticillus. Unlike the other four it does not target the ribosome: it intercalates into DNA and cleaves it, killing both dividing and non-dividing cells. Resistance comes from Sh ble from Streptoalloteichus hindustanus, whose product binds the drug stoichiometrically rather than enzymatically — one protein molecule sequesters one drug molecule. That mechanism means resistance saturates, so overshooting the concentration kills resistant cells too. Zeocin's activity is reported to be reduced in high-salt media and at pH above about 7.5, so keep the selection medium near pH 7.0–7.5 and note that a failed zeocin selection is often a medium chemistry problem rather than a concentration problem. Handle it as a DNA-damaging agent and protect solutions from light.

Puromycin is an aminonucleoside from Streptomyces alboniger. It is a structural analogue of the 3' end of aminoacyl-tRNA, so it enters the A site of the ribosome, is incorporated into the growing chain, and causes premature chain termination. Resistance comes from pac, puromycin N-acetyltransferase, which acetylates and inactivates it. Puromycin is the fastest agent — 0.5–10 µg/mL, commonly 1–5, with untransfected cells dying in 2–4 days — which is why it is the default for lentiviral workflows where speed matters.

In practice the resistance gene on your construct decides for you. When you do have a choice:

• Speed: puromycin (2–4 days) < blasticidin (3–10 days) < zeocin and hygromycin B (7–14 days) ≈ G418 (7–14 days).
• Cost per litre of selection medium: blasticidin and puromycin are used at microgram-per-mL concentrations, so despite higher per-mg prices they are often cheaper in use than G418 or hygromycin B at hundreds of micrograms per mL.
• Dual and triple selection: the five agents have distinct mechanisms and distinct resistance genes, so they stack. A common pattern is puromycin for the first construct and blasticidin for the second, since neither cross-resists the other.
• Expression level: published comparisons of drug selection systems in the same cells report that the choice of selectable marker can shift average transgene expression several-fold and changes how homogeneous expression is across the population. If uniform expression matters for your assay, the marker is worth choosing deliberately rather than by convenience.
• Non-dividing cells: zeocin is the only one of the five that kills non-dividing as well as dividing cells, which matters for slowly-cycling primary cultures.

Published ranges span an order of magnitude or more because sensitivity genuinely varies that much between cell types. HEK293 cells may need 400 µg/mL G418 where a primary fibroblast line needs 1000. There is no way to get this right from a catalogue, and the two failure modes are equally expensive: too low and untransfected cells survive, giving a background of non-expressing cells in your "stable" line; too high and you kill the resistant cells too, especially with zeocin, where resistance saturates.

Run the kill curve on the untransfected parental line, in the same medium, at the same seeding density, in the same plate format you will use for the real selection. Density in particular matters — a confluent monolayer resists selection far better than a sparse one, so a kill curve run at the wrong density gives an answer that will not transfer.

Protocol

1. Seed the untransfected parental cells into a 24-well plate at 0.5 mL per well. For adherent lines use 0.8–3.0 × 10⁵ cells/mL; for suspension lines use 2.5–5.0 × 10⁵ cells/mL. Aim for 50–80% confluence at the time the agent is added — the same density at which you will run the real selection. Use duplicate wells for every concentration.
2. Allow the cells to attach and resume growth, normally overnight for adherent lines.
3. Prepare a dilution series of the agent in complete growth medium, including a zero-antibiotic control well. Worked series for each agent are given below.
4. Replace the medium in each well with the corresponding antibiotic-containing medium. Handle the zero control identically, with a medium change and no drug — it controls for the handling, not just the drug.
5. Replace the antibiotic-containing medium every 2–3 days for the whole selection period. This is the step most often skipped, and skipping it invalidates the curve: the agents degrade in medium at 37 °C, and dead-cell debris accumulates.
6. Examine every well daily under the microscope and record the approximate percentage of dead or detached cells. A simple daily table of well versus percentage kill is enough.
7. Run the curve for the agent's normal selection window — up to 4 days for puromycin, up to 10 days for blasticidin, and 10–14 days for G418, hygromycin B and zeocin.
8. Read the result. The working concentration is the lowest concentration that kills 100% of the cells within the agent's normal window. If every concentration kills everything on day 2, the series was too high — repeat with a lower range. If the top concentration leaves survivors, extend the range upward.
9. Confirm on the resistant population. Once you have transfected cells, verify that they tolerate the chosen concentration. If they do not, the marker is under-expressed and the construct or the promoter, not the concentration, is the problem.

Worked concentration series (all in complete growth medium, duplicate wells, zero control included):

• Puromycin: 0, 0.25, 0.5, 1, 2, 4, 6, 8, 10 µg/mL. From a 10 mg/mL stock, 1 µg/mL is a 1:10,000 dilution — make an intermediate 100 µg/mL working dilution rather than pipetting sub-microlitre volumes.
• Blasticidin S: 0, 1, 2, 4, 6, 8, 10, 15, 20 µg/mL. From a 10 mg/mL stock, 10 µg/mL is 1:1000.
• G418 / geneticin: 0, 100, 200, 400, 600, 800, 1000, 1500, 2000 µg/mL. From a 50 mg/mL solution, 800 µg/mL is 1:62.5 — that is 8 mL into 500 mL of medium.
• Hygromycin B: 0, 50, 100, 200, 300, 400, 600, 800, 1000 µg/mL. From a 50 mg/mL solution, 300 µg/mL is 1:167.
• Zeocin: 0, 50, 100, 200, 300, 400, 600, 800, 1000 µg/mL. From a 100 mg/mL solution, 200 µg/mL is 1:500. Check and record the medium pH, since zeocin activity falls above about pH 7.5.

Repeat the kill curve whenever you change cell line, base medium, serum concentration or plate format. Serum concentration in particular shifts the answer, because serum protein binds a fraction of the drug.

These are two different numbers and conflating them causes avoidable losses.

The selection concentration is the kill-curve value — the dose that eliminates 100% of untransfected cells. Use it from the start of selection until the resistant population is established and expanding, typically for the full selection window plus a few passages.

The maintenance concentration keeps the resistance gene under enough pressure to prevent silencing or loss, without continuously stressing an established line. It is conventionally about half the selection concentration, and many labs go further: they culture the established stable line without any agent and reapply the maintenance concentration for one passage every four to six weeks, or before expanding for a critical experiment.

Withdraw the agent completely before functional assays wherever possible. Puromycin at selection concentrations inhibits translation in every cell to some degree; zeocin is a DNA-damaging agent; the aminoglycosides affect mitochondrial translation. None of these is a variable you want inside a phenotype measurement.

• The concentration was never determined for this line. The commonest cause. A number taken from a paper on a different cell type in a different medium is a guess.
• The cells were too dense. Confluent cultures resist selection. Re-run at the density you will actually use.
• The medium was not replaced often enough. These agents degrade at 37 °C. Every 2–3 days, without exception.
• Serum concentration changed. Higher serum binds more drug and raises the effective concentration needed.
• Zeocin in the wrong medium. High salt or pH above about 7.5 reduces activity. This looks exactly like an under-dose.
• Selection started too early. Allow 24–72 hours after transfection or transduction for the resistance protein to be expressed before applying the agent. Applying it immediately kills cells that would have become resistant.
• The resistance marker is poorly expressed. If transfected cells die at a concentration untransfected cells survive, suspect promoter silencing, a truncated cassette, or an internal ribosome entry site that expresses the marker weakly.
• Stock handling. Aliquot on first thaw, store as directed — typically −30 to −5 °C and protected from light for G418, blasticidin, zeocin and puromycin, and 2–8 °C for hygromycin B — and avoid repeated freeze–thaw cycles.