Plasmid transfection into CHO cells, antibiotic/metabolic selection, and formation of a polyclonal producer pool ahead of single-cell cloning
Stable cell line development begins with an expression plasmid: a circular DNA construct carrying the gene of interest (GOI) under a strong constitutive promoter, alongside a selectable marker gene that will later let survivors be identified from a vast excess of cells. Delivering that plasmid efficiently into millions of Chinese Hamster Ovary (CHO) cells — without killing them — is the first bottleneck of the entire cell line development timeline.
A production-grade CHO expression plasmid is a compact piece of genetic engineering, typically containing:
• Heavy and light chain (or single GOI) expression cassettes: each with its own strong mammalian promoter (CMV, EF-1α, or a hybrid CHEF1 promoter), a Kozak sequence, the coding sequence, and a polyadenylation signal • A selectable marker cassette: DHFR (dihydrofolate reductase), GS (glutamine synthetase), or an antibiotic-resistance gene (neo/G418, puromycin-N-acetyltransferase, hygromycin phosphotransferase) driven by a deliberately weak promoter (SV40 early, or an attenuated/mutant promoter) so that only high-copy or well-integrated cells express enough marker to survive selection • A bacterial origin of replication and a bacterial selection marker (ampicillin or kanamycin resistance) used solely for plasmid propagation in E. coli during manufacturing — silent in the mammalian host • Optional elements: matrix attachment regions (MARs), ubiquitous chromatin opening elements (UCOE), or insulators that reduce positional silencing after random genomic integration
The deliberate weakness of the marker promoter is a design choice, not an oversight: it raises the bar so that only clones with a favorable, transcriptionally active integration site — often co-integrated near the GOI itself — survive selection, indirectly enriching for high GOI expression.
Two delivery chemistries dominate CHO transfection at both small (96-well screening) and large (bioreactor seed) scale:
Electroporation: cells and plasmid DNA are mixed in a conductive buffer, then subjected to a brief high-voltage electrical pulse that transiently destabilizes the plasma membrane, opening aqueous pores through which plasmid DNA diffuses. Modern flow-electroporation instruments (e.g., MaxCyte STx) can process 10⁸–10⁹ cells in a single run with high, consistent uptake efficiency — the workhorse method for GMP-track cell line development because it avoids animal-derived lipid reagents and scales predictably.
Lipofection: cationic lipid nanoparticles complex with the negatively charged plasmid backbone, and the resulting lipoplex fuses with the cell membrane or is taken up by endocytosis, releasing DNA into the cytoplasm. Lipofection is gentler on cell viability and remains popular for small-scale, high-throughput screening (96- or 384-well transfection panels comparing vector designs) but scales less predictably to bioreactor volumes.
Regardless of method, only a fraction of the DNA that enters the cytoplasm survives cytoplasmic nucleases and successfully traffics into the nucleus, where it can either be transcribed episomally (transiently) or, far more rarely, become integrated into the host genome by non-homologous recombination.
Every plasmid design decision in this stage is made with the next stage already in mind. Because random integration is rare and its genomic location is uncontrolled, the population emerging from transfection is enormously heterogeneous: most cells hold no integrated copy at all, some hold a single low-expressing copy in transcriptionally quiet chromatin, and a small minority integrate multiple copies in an actively transcribed locus. The selectable marker is the only lever available at this point to computationally-blindly, but biologically-elegantly, filter that heterogeneous population down to genuine, productive stable integrants — which is exactly the job of the next two stages.
Immediately after transfection, CHO cells enter a window of intense, unstable protein expression. Episomal plasmid — DNA sitting free in the nucleus, never integrated into a chromosome — is transcribed avidly by the host machinery, driving a strong but temporary expression peak. This is the same phenomenon exploited deliberately in transient transfection production (used for early-stage material), but here it is a passing phase that must be waited out before the real selection experiment can begin.
Plasmid DNA that reaches the nucleus but never integrates behaves as an episome: a free, non-replicating circular molecule. It has no centromere and no mechanism to be faithfully partitioned to daughter cells at mitosis, so each cell division halves the average copy number across the population — roughly analogous to serial dilution. Because episomal copy number starts high (often hundreds of copies per cell immediately after transfection), transcription in the first 48–96 hours can be very strong, even stronger on a per-cell basis than what a single stably integrated locus will later produce.
Genomic integration is a rare, essentially accidental event mediated by the cell's own double-strand break repair machinery (non-homologous end joining), which occasionally captures a linearized or nicked plasmid fragment and stitches it into a chromosome. Only cells in which this event happens — typically well under 5% of the transfected population, and often far lower — will retain the transgene, and the marker gene, permanently through every subsequent division.
This is the central, sometimes counterintuitive fact of stable cell line development: the great majority of the brightly-expressing cells visible during the transient window will contribute nothing to the final stable pool. Their high expression is a temporary artifact of transfection, not a sign of genomic integration. If selection pressure were applied too early — before episomal plasmid has had time to dilute out — cells with no stable copy at all could survive on borrowed episomal marker expression long enough to be mistaken for genuine integrants, contaminating the pool with false survivors.
For this reason, cell line development protocols typically insert a short recovery period (24–72 hours) after transfection before selection agent is added, allowing the culture to re-establish healthy growth, and sometimes a longer window is used deliberately to let episomal signal decay before assessing which cells are "truly" stable.
If a cell divides with probability 0.5 of each daughter inheriting a copy of a given episomal plasmid molecule (a simplification, but directionally accurate for non-replicating, non-partitioned DNA), then after n divisions the fraction of the original plasmid pool retained scales roughly as (0.5)ⁿ per lineage, while the population itself grows as 2ⁿ. In practice this means that by 7–10 days post-transfection (roughly 7–10 CHO doublings), non-integrated plasmid is essentially undetectable, and any expression still visible in the culture is coming almost exclusively from the rare cells that achieved stable genomic integration in stage one.
With the transient signal cleared, the culture is exposed to a selective agent matched to the plasmid's marker gene. Cells without a stably integrated, functional copy of that marker cannot survive the challenge and are progressively eliminated — typically over one to three weeks — leaving behind a small but genuine population of stable producers. This is the step that converts "DNA got into the cell" into "DNA is now a permanent part of this cell's genome and is being actively expressed."
Two families of selection dominate industrial CHO cell line development, both exploiting metabolic auxotrophy rather than simple drug resistance, because they additionally allow gene amplification:
DHFR / methotrexate (MTX): dihydrofolate reductase is essential for nucleotide (thymidine, purine) biosynthesis. Using a DHFR-deficient CHO host (e.g., CHO DXB11 or DG44) transfected with a plasmid carrying both the GOI and a DHFR minigene, cells lacking a functional integrated DHFR copy cannot synthesize nucleotides de novo and die in nucleoside-free medium. Critically, the antifolate drug methotrexate can then be titrated in stepwise increasing concentrations against cells that already carry the marker: MTX inhibits DHFR, and cells survive only by amplifying their DHFR gene copy number — and because the GOI is physically linked to DHFR in the integration locus, GOI copy number (and often expression) co-amplifies, sometimes 10- to 100-fold, across successive rounds.
GS / MSX: glutamine synthetase converts glutamate and ammonia into glutamine, an amino acid many cell lines require from the medium. CHO cells have some endogenous GS activity, so the glutamine synthetase inhibitor methionine sulfoximine (MSX) is used to suppress background activity, raising the bar so that only cells with an integrated, overexpressed GS transgene (again typically linked to the GOI) can synthesize enough glutamine to survive in glutamine-free medium. GS selection is popular because it needs no specialized DHFR-deficient host and reaches usable stable pools somewhat faster.
Antibiotic resistance markers (neo/G418, puromycin, hygromycin, zeocin, blasticidin) work more simply: the marker gene encodes an enzyme that detoxifies the antibiotic directly. These are faster and simpler to apply but do not support gene amplification the way DHFR/MTX or GS/MSX do, so they are often used for research-grade or transient-adjacent stable pools rather than high-titer production lines.
The concentration of selection agent is a dial, not a switch, and turning it has real consequences in both directions:
Higher stringency (more MTX, more MSX, more antibiotic) kills a larger fraction of the population, including some genuine but weak integrants, leaving a smaller, more purified pool enriched for cells with either high marker copy number or a favorable integration site — both of which correlate, imperfectly but usefully, with high GOI expression. The cost is a longer recovery time, since a smaller surviving population takes longer to re-establish confluent growth, and a real risk of losing genuine but modestly-expressing clones that might otherwise have been useful.
Lower stringency preserves more of the population and shortens time-to-pool, but yields a less enriched pool with a wider spread of expression levels, including cells that barely cleared the bar and contribute little to bulk titer — pushing more of the enrichment burden downstream onto single-cell cloning.
Most cell line development groups tune stringency empirically for each new construct and host cell bank, often running a small dose-response (kill-curve) experiment first to find the minimum concentration that reliably eliminates untransfected control cells within the target selection window.
| Product | Indication | Trial Design | Key Result |
|---|---|---|---|
| DHFR / Methotrexate | DHFR-deficient CHO (DXB11, DG44) | Nucleotide biosynthesis blocked; MTX enables stepwise gene amplification | Highest achievable copy number and titer |
| GS / MSX | Wild-type or GS-knockout CHO-K1 | Glutamine synthesis blocked; MSX suppresses endogenous GS background | Faster pool, no specialized host required |
| Puromycin (pac) | Any CHO host | Acetylates and inactivates puromycin directly | Fast, simple, good for research-grade pools |
| G418 / Neomycin (neo) | Any CHO host | Phosphotransferase inactivates aminoglycoside | Widely available, well-characterized kinetics |
Once selection pressure has done its work, the surviving cells — a genetically heterogeneous mixture of hundreds to thousands of independent integration events, each with its own copy number, integration site, and expression level — are allowed to recover and expand into a bulk stable pool. This pool is not yet a cell line in the strict sense (no single clone has been isolated), but it is stable, expandable, and productive enough to be measured, banked, and used to make an early go/no-go decision.
The defining feature of a stable pool is heterogeneity. Because integration occurs at essentially random genomic locations and copy numbers, every surviving cell is, strictly speaking, genetically distinct at its integration locus — some sit in transcriptionally silent heterochromatin and express poorly despite carrying an intact transgene, others integrated near strong endogenous enhancers and express robustly, and a subset may have suffered partial gene truncation or rearrangement during the integration event itself and express a non-functional or truncated product.
Bulk pool titer is therefore a weighted average across this whole distribution, dominated in practice by whichever subpopulations grow fastest and produce the most under the specific culture conditions used — which is not necessarily the same ranking that single-cell cloning will later reveal.
As the pool recovers from selection and reaches a stable, exponentially growing state, cell line development teams typically track:
• Viable cell density (VCD) and doubling time, to confirm the pool has fully recovered normal CHO growth kinetics after the stress of selection • Viability by trypan blue or automated imaging exclusion, which should climb back above 90% as dead and dying non-integrants are diluted out of the culture • Bulk titer, usually measured by a short fed-batch shake-flask or ambr®-scale mini-bioreactor run (typically 10–14 days), quantified by Protein A HPLC (for antibodies) or ELISA • Product quality attributes (aggregation, charge variants, glycosylation) at a coarse level, mainly to flag any gross process red flags before investing further
These measurements together form the pool's "productivity fingerprint" and are the primary evidence used to decide whether the construct and host cell line combination is worth carrying forward.
A characterized, stable, healthy pool is the deliverable of this whole workflow — and the gatekeeper for everything that follows. Single-cell cloning (by limiting dilution or single-cell FACS sorting into 96- or 384-well plates, followed by weeks of expansion and re-screening of hundreds of clones) is dramatically more labor- and resource-intensive than pool generation. Investing in it before confirming the pool itself is productive risks weeks of wasted downstream effort on a construct or cell bank that was never going to yield a commercially viable clone.
Single-cell cloning campaigns commit real time and resources: hundreds of individual wells must each be expanded, fed, imaged for single-cell-derived confirmation (often via early-stage imaging to document clonality for regulatory purposes), and screened for productivity — typically over 4–8 additional weeks before even a shortlist of top clones exists. Running that full campaign on a pool that turns out, on closer bulk-scale measurement, to be a poor producer wastes that entire investment.
By measuring bulk pool titer, viability, and growth rate before committing to cloning, teams can compare multiple candidate constructs, vector designs, or even host cell banks cheaply and in parallel, and only advance the strongest pool (or pools) into the expensive single-cell cloning stage. This "fail fast, fail cheap" gate is one of the highest-leverage checkpoints in the entire cell line development timeline.
End to end, a typical stable pool generation campaign — from plasmid transfection to a fully characterized, bankable pool — takes roughly two to four weeks: a few days of transient expression and recovery, one to three weeks of active selection depending on stringency and marker system, and one to two more weeks of expansion and titer assessment. Programs under time pressure sometimes accept lower selection stringency to shorten this timeline, banking a less-enriched but faster pool and pushing more of the productivity enrichment burden onto cloning; programs where clone quality matters more than speed lean toward higher stringency (and, with DHFR/MTX, additional amplification rounds) to hand cloning a smaller but stronger starting population.
Either way, the pool that emerges from this stage is not the end product — it is best understood as a curated, pre-enriched sampling universe from which single-cell cloning will subsequently draw out the one or few clones with the growth, stability, and productivity profile needed for manufacturing.