Insect Cell Culture for Baculovirus Expression: Sf9, Sf21, and Hi5 Cell Line Selection, Culture Optimization, and Troubleshooting

The baculovirus expression vector system (BEVS) is one of the most versatile eukaryotic platforms for recombinant protein production, and the insect cell host you pick before you ever open a vial of virus does more to determine your final titer than almost any downstream optimization. Sf9, Sf21, and Hi5 are not interchangeable — they differ in growth rate, maximum achievable density, virus production capacity, secretion efficiency, and even the glycan structures they attach to your protein. Get the host wrong and no amount of MOI or harvest-time tuning will close the gap.

This guide covers how to choose between Sf9, Sf21, and Hi5, how to culture each one in serum-free suspension, how to design a seed train from a cryovial to a production bioreactor, how to prepare and titer baculovirus stock, and how to set the multiplicity of infection (MOI), cell density, and harvest timing that actually determine your yield. It closes with a troubleshooting table for the failure modes that account for most lost campaigns.

What Are Sf9, Sf21, and Hi5 Insect Cell Lines?

Sf9 and Sf21 are both derived from Spodoptera frugiperda (fall armyworm) pupal ovarian tissue, first established by Vaughn and colleagues in 1977. Sf9 is a subclone of the parental Sf21 line, selected for faster growth, smaller and more uniform cell size, and better attachment/suspension behavior — which is why Sf9 became the default workhorse for both virus amplification and routine protein expression, while Sf21 is used less often for production but remains valuable for plaque assays and virus titration because its larger, flatter cells produce sharper, more countable plaques.

Hi5 cells (also written High Five, official designation BTI-TN-5B1-4) come from a different species entirely: Trichoplusia ni (cabbage looper) egg cell homogenates, established by Granados and colleagues in the mid-1980s and commercialized in the 1990s. Hi5 cells are not simply a faster or slower version of Sf9 — they process and secrete proteins through a fundamentally more efficient pathway, which is the single biggest reason process developers reach for Hi5 when the target protein needs to leave the cell.

Physically, the three lines are distinguishable under a standard inverted microscope. Sf9 cells run 13–17 µm in diameter, round to slightly oval, and grow in loose suspension with minimal clumping once adapted. Sf21 cells are visibly larger (15–20 µm) and more irregular in shape, which is part of why they are easier to score in a plaque assay but harder to keep in tight suspension at very high density. Hi5 cells overlap Sf21 in size (15–20 µm) but have a distinct granular cytoplasm and, in early adaptation, a stronger tendency to clump — suspension-adapted Hi5 lines from reputable vendors have largely engineered this out, but freshly adapted stocks still benefit from extra shear protection.

The glycosylation row deserves a callout because it surprises people coming from mammalian cell culture: all three insect lines produce paucimannose-type N-glycans rather than the complex, sialylated glycans that CHO or HEK293 cells add. Insect cells lack the enzymes (sialyltransferases, most branching GlcNAc transferases) needed to build human-type complex glycans, and they express an endogenous fucosyltransferase that can attach immunogenic core α1,3-fucose. For research reagents, structural biology targets, and many vaccine antigens this is a non-issue or even an advantage — simpler, more homogeneous glycoforms crystallize better. For a therapeutic candidate destined for human dosing, glyco-engineered insect lines (e.g., SfSWT or Mimic lines expressing mammalian glycosyltransferases) or a post-purification glycan remodeling step are usually required.

The virus production gap between Sf9 and Hi5 is the other number worth internalizing early: Sf9 cultures routinely yield baculovirus titers 10–100-fold higher than Hi5 cultures infected under identical conditions. This is why almost every BEVS protocol, regardless of which cell line is used for final production, amplifies virus stock (P1 through P3) in Sf9 and only switches to Hi5 at the production infection step.

How to Choose Between Sf9, Sf21, and Hi5 for Your Target Protein

The right cell line is determined almost entirely by where your protein ends up: inside the cell, secreted into the medium, embedded in the plasma membrane, or assembled into a virus-like particle. Use the decision tree below as a starting point, then confirm with a small-scale screen if your protein doesn't fit neatly into one category.

Intracellular proteins (enzymes, structural biology targets, most research reagents) generally do best in Sf9. The cells are robust, well characterized, easy to culture at high density, and because the protein never has to cross a membrane, Hi5's secretion advantage doesn't apply. Sf9 is also simply better documented — decades of published protocols, more predictable batch-to-batch behavior, and it's the same cell line you're already using to amplify virus, so there's no second adaptation to manage.

Secreted glycoproteins — antibody fragments, receptor ectodomains, growth factors, most vaccine antigens — are where Hi5 earns its reputation. Hi5's secretory pathway processes and exports protein more efficiently than Sf9's, and published head-to-head comparisons routinely show 5–10× higher volumetric titer for the same construct and infection conditions. The trade-off is a host that's somewhat less forgiving: Hi5 is more shear-sensitive in stirred systems, needs more careful suspension adaptation, and produces far less baculovirus per cell, so you'll still amplify your virus stock in Sf9 and only switch hosts for the production infection.

Virus-like particles split between the two hosts depending on the VLP's biology. Enveloped VLPs that bud from the plasma membrane (influenza HA/NA-based, some retroviral Gag-based particles) often assemble and bud better in Sf9, where membrane composition and budding efficiency are well characterized. Non-enveloped, capsid-only VLPs (many papillomavirus- and parvovirus-based particles) that rely on efficient intracellular assembly and, in some cases, partial secretion can favor Hi5. Run both in a small screen if you don't already have precedent for your specific VLP architecture — the difference is large enough (often 1.5–2×) to be worth the extra week.

Membrane proteins for structural or functional studies are almost always expressed in Sf9. It's the platform most cryo-EM and X-ray structure papers use, which means more troubleshooting precedent, more compatible detergent-screening protocols, and better-characterized lipid composition for reconstitution work downstream.

Plaque assays and virus titration are the one place Sf21 has a durable advantage over Sf9: its larger, flatter cell morphology produces sharper, more countable plaques under an agarose or methylcellulose overlay. Many labs maintain a small Sf21 monolayer culture purely for titration while doing all suspension work in Sf9.

Culture Conditions and Serum-Free Media Selection

Insect cell culture is, in several important ways, easier to run than mammalian cell culture: no CO₂ incubator, no humidity control, and a wider pH tolerance. But the details that matter — osmolality, shear protection, and glutamine stability — still need to be right, and getting them wrong is a common source of the "my cells just won't grow" support ticket.

Temperature should be held at 27–28°C, with 27°C generally considered optimal for both growth rate and baculovirus replication. Standard tissue-culture incubators at 37°C will kill or badly stress insect cells within hours — this is the single most common new-user mistake. Insect cells also don't need CO₂ supplementation or humidified air, because their native environment (soil, plant surfaces, insect hosts) has none of the atmospheric CO₂ control mammalian cells evolved to expect. A standard, non-humidified 27°C incubator or shaking incubator is sufficient; sealed shake flasks with vented caps handle gas exchange passively.

pH should be maintained around 6.2–6.5 — noticeably more acidic than the pH 7.0–7.4 mammalian cell culture targets. Commercial insect cell media are buffered for this range out of the bottle; don't try to "correct" it toward neutral pH, which will stress the culture. A slow upward pH drift during infection (toward 6.6+) is often an early sign of extensive cell lysis, worth tracking as a harvest-timing cue.

None of these formulations require fetal bovine serum for established, adapted lines — that's the whole point of "serum-free" insect media, and it's one of the practical advantages BEVS holds over serum-dependent mammalian platforms: lower cost, no serum lot-to-lot variability, and a cleaner regulatory story. If you're starting from an old serum-dependent stock, budget 10–15 passages of gradual serum weaning before you trust the suspension-adapted culture for production work.

Pluronic F-68 at 0.1% (w/v) is standard practice for any suspension culture in shake flasks, wave bags, or stirred vessels. It's a non-ionic surfactant that coats the cell membrane and dramatically reduces shear- and bubble-associated cell damage, which matters more for insect cells than most people expect — they lack a rigid cell wall and are genuinely more shear-sensitive than their robust growth curves suggest. Most commercial media already include Pluronic F-68; check your certificate of analysis before adding more.

L-glutamine is included in most insect media formulations but degrades in solution over time (spontaneous hydrolysis to pyrrolidone carboxylic acid and ammonia), so a bottle sitting at 4°C for a month has meaningfully less usable glutamine than a freshly supplemented one. Keep a concentrated L-glutamine stock (typically 200 mM) at −20°C and add it fresh at each media bottle opening if your supplier recommends supplementation, rather than relying on a pre-mixed bottle that's been open for weeks.

Storage: complete, supplemented media should be stored at 4°C and used within about one month — degradation of glutamine and other labile components accelerates after that window, and you'll see it as sluggish growth before you see it on a certificate of analysis. Concentrated L-glutamine stock keeps for roughly a year at −20°C in single-use aliquots (avoid repeated freeze-thaw).

Seed Train Design and Passage Management

A disciplined seed train is what separates a reproducible BEVS campaign from one where every production run behaves a little differently. The goal is to take a cryopreserved vial through a predictable sequence of expansions that lands the culture in exponential growth, at the right density, on the day you need it for infection — without letting passage number creep past the point where expression starts to drift.

Thaw and early expansion: thaw one cryovial directly into a T-25 flask as a static or lightly rocked monolayer culture, then progress to T-75 and T-175 flasks as the culture recovers and expands, before transitioning to a shake flask for true suspension growth. This staged approach gives freshly thawed cells, which are always a little stressed, time to recover attachment and viability before you ask them to handle shear in suspension.

Splitting cadence: once in steady suspension culture, split at a 1:3 to 1:5 ratio every 3–4 days. Maintain the seeding density between 0.3–0.5 × 10⁶ cells/mL after each split — never let it fall below roughly 0.2 × 10⁶ cells/mL, where cultures can stall in a long lag phase, and never let it run past 6–8 × 10⁶ cells/mL in maintenance culture, where nutrient depletion and accumulated metabolic waste start eroding both growth rate and downstream expression capacity even before viability visibly drops.

Passage limits: keep Sf9 (and, by extension, Sf21 and Hi5) below passage 50–60 from the original cryopreserved stock. Expression levels decline progressively past passage 50 — a mix of genetic drift in the immortalized line and reduced susceptibility to baculovirus infection — so a well-run lab tracks passage number on every flask label and retires lines proactively rather than waiting for a failed campaign to notice. Viability should stay above 95% at every passage; a culture that's dipping into the low 90s or high 80s between splits is telling you something is already wrong (stale media, overcrowding, contamination) well before growth rate visibly slows.

Baculovirus Stock Preparation and Titer Assessment

Every production infection depends on a virus stock whose titer you actually trust, and that trust is built through a disciplined amplification series — not by infecting a huge flask once and using whatever comes out. The standard progression runs P0 (the transfection or initial recombination event that generates virus) through P1, P2, and P3, with each passage amplifying virus volume while watching carefully for the passage effect: repeated high-MOI serial passage accumulates defective interfering particles (DIPs) that reduce titer and, eventually, expression capacity in downstream infections.

Amplification MOI should be kept low — 0.01 to 0.1 — specifically to avoid this DIP accumulation. A low-MOI amplification infects only a small fraction of cells on first contact; those cells burst and release progeny virus that infects the rest of the culture over the following 24–48 hours, which naturally selects against defective genomes that need co-infection with a helper virus to replicate. High-MOI amplification, by contrast, guarantees every cell (including ones carrying defective genomes) gets co-infected, and DIPs propagate unchecked.

Cell density at infection for amplification runs should sit at 1.5–2.0 × 10⁶ cells/mL — the same standard-batch range used for production infections, for the same reason: infecting into the cell density effect (see the production infection section below) reduces per-cell virus output just as it reduces per-cell protein output.

Harvest timing for virus stock differs from a protein-production harvest. You're not chasing peak protein titer; you're chasing peak infectious virus titer, which typically means harvesting P1 through P3 amplifications at 48–72 hpi, once viability has dropped to roughly 70–80%. Harvesting earlier leaves virus still budding into the medium; harvesting much later risks virus particle degradation from extracellular proteases released during extensive lysis.

Titer assessment methods

• Plaque assay — the gold standard, measuring genuinely infectious particles as plaque-forming units (pfu/mL). Slow (4–7 days to read plaques) but the reference method every other assay is calibrated against.
• BacPAK qPCR (or equivalent quantitative PCR titer kit) — measures viral genome copies, returning a result in about 4 hours. Fast enough to inform same-day infection decisions, but it counts total genomes (including defective and non-infectious particles), so it tends to read higher than a plaque assay and needs a correlation established for your specific virus.
• Flow cytometry with a GFP reporter — if your baculovirus backbone co-expresses GFP, flow cytometry against a dilution series gives a same-day estimate of the infectious fraction without waiting on plaques, though it requires the reporter to be present in your construct.

For routine process work, many labs run qPCR for fast go/no-go decisions and confirm with a plaque assay for any stock that will be used across multiple production campaigns. Storage matters as much as the initial titer: stocks kept at 4°C protected from light hold their titer reasonably well for weeks, but for anything beyond a few months, store at −80°C, where degradation slows dramatically. Typical well-prepared P2 stocks titer at 1–5 × 10⁸ pfu/mL — if a fresh titer comes back an order of magnitude below that, re-check your amplification MOI, cell density, and harvest timing before assuming the construct itself is the problem.

Production Infection: Optimal Cell Density, MOI, and Harvest Timing

The production infection is where cell line choice, virus stock quality, and culture health all get tested at once. Three parameters — cell concentration at infection (CCI), multiplicity of infection (MOI), and time of harvest — interact strongly enough that optimizing them one at a time usually leaves yield on the table; treat them as a system.

Cell density at infection: the standard target for batch infection is 1.5–2.0 × 10⁶ cells/mL. This range balances volumetric productivity against the cell density effect — the well-documented decline in specific (per-cell) protein productivity that sets in above roughly 2–3 × 10⁶ cells/mL in unsupplemented batch culture, driven by nutrient depletion and shifts in central metabolism. High-density infection strategies (10–20 × 10⁶ cells/mL) are used in more advanced processes specifically to push past this ceiling, but they require a medium exchange or feed step at or near the time of infection to restore the nutrient environment a standard batch culture can't sustain at that density — recent published work using this approach has reported titers in the range of 263 mg/L for a GFP reporter construct, well above what unsupplemented high-density batch infection achieves.

Multiplicity of infection: an MOI of 1–5 pfu/cell delivers synchronous infection — nearly every cell is infected within a narrow window, which tightens the harvest timing and makes the campaign more reproducible run to run. Higher MOI gets there faster but consumes proportionally more virus stock. Low MOI (0.01–0.1) relies on the same secondary-infection mechanism used for stock amplification, uses far less virus, but adds roughly one to two days to reach peak expression because the culture has to wait for progeny virus to spread through the population. As a starting point, use MOI 1–5 for Sf9 and MOI 0.5–2 for Hi5 — Hi5 cells are more readily infected, so a somewhat lower MOI still achieves synchronous infection.

Harvest timing splits by product location: intracellular product should be harvested at 48–72 hpi, while it's still accumulating inside cells that haven't yet lysed and released host proteases into the medium. Secreted product tolerates a longer window — 48–96 hpi — because it has already left the cell and continues to accumulate in the medium even as the producing cells begin to decline. In both cases, track viability alongside a calendar: harvest when viability has dropped to roughly 70–80%, which is a much more reliable signal than hours-post-infection alone, since infection kinetics shift with MOI, cell health, and construct.

Read the diagnostic chart as a pattern-matching tool. A successful infection shows VCD gradually declining as cells commit to virus production and eventually lyse, paired with viability sliding from the high 90s down toward 50% by day 7 — the classic signature you're aiming for. A failed infection at too-low MOI (or with a dead virus stock) looks like uninfected growth: VCD keeps climbing past the point where infected cultures should be declining, while viability stays stubbornly above 90% — if you see this pattern, suspect virus stock titer or a pipetting error before anything else. Contamination announces itself fast and hard: VCD and viability both crash within the first 24–48 hours, far faster than any infection kinetics, usually accompanied by visible turbidity or an off smell at harvest.

For a deeper treatment of how MOI, cell density, and harvest time interact — including a DOE protocol for optimizing all three together — see our companion guide on baculovirus MOI and harvest optimization.

Scale-Up to Stirred-Tank and Wave Bioreactors

Moving a BEVS process from shake flasks to bioreactor scale is more forgiving than the equivalent mammalian cell scale-up — insect cells don't need CO₂ or pH cascade control — but shear sensitivity and oxygen transfer still need deliberate engineering as volume grows.

Shake flasks remain the workhorse from process development through pilot screening, typically 50–500 mL working volume, agitated at 100–130 rpm on an orbital shaker platform, held at 27°C. Wave (rocking motion) bioreactors extend this gentle-agitation approach to 1–50 L, using a rocking platform rather than an impeller to mix and aerate — a good fit for shear-sensitive Hi5 cultures or early-stage GMP work where single-use bags simplify changeover between campaigns. Stirred-tank bioreactors scale from 1 L bench systems up to 2000 L manufacturing vessels, typically using pitched-blade or marine impellers rather than the higher-shear Rushton turbines common in microbial fermentation, specifically to protect the wall-less insect cell membrane.

Dissolved oxygen control should maintain at least 30% air saturation. Insect cells have a lower specific oxygen uptake rate (OUR) than mammalian cells at equivalent density, but that OUR climbs 2–3× once infection sets in and cells shift into high-output protein synthesis, so a DO setpoint that comfortably supports growth-phase culture can become marginal during the production infection window — size sparging and agitation with the post-infection OUR in mind, not just the pre-infection number. Our OTR / kLa Estimator is built for exactly this kind of pre- and post-infection oxygen demand check.

Agitation ranges roughly 80–150 rpm depending on vessel scale and impeller geometry, with the real design constraint being impeller tip speed rather than rpm alone — keep tip speed below about 1.5 m/s to stay clear of shear damage thresholds reported for suspension insect cell lines. Pluronic F-68 at 0.05–0.2% remains standard practice at every scale for exactly this reason — more aggressive agitation and sparging at larger scale increases the shear and bubble-burst exposure the surfactant is protecting against.

Because insect cell culture doesn't need CO₂ stripping or a pH control cascade the way mammalian culture does, scale-up criteria simplify to matching either constant kLa (oxygen transfer coefficient) or constant P/V (power input per unit volume) between scales, rather than juggling the multiple, sometimes conflicting scale-up rules mammalian processes require. Whichever criterion you choose, validate it with a small confirmation run at the new scale before committing a full production campaign — our Scale-Up Calculator handles both kLa- and P/V-matched geometry across vessel types.

Troubleshooting Common Insect Cell Culture Problems

Most BEVS process failures fall into a handful of recognizable patterns. Work through the table below systematically before assuming the construct itself is at fault — the majority of "low expression" tickets trace back to culture handling, not biology.

A practical habit that prevents most of this list: log passage number, seeding density, and viability on every flask at every split, and log CCI, MOI, and harvest viability on every infection. When something does go wrong, that log turns a guessing exercise into a two-minute diagnosis.

Frequently Asked Questions

Related Tools

• Cell Seeding Density Chart — Reference seeding densities for Sf9, Sf21, Hi5, and mammalian cell lines
• Scale-Up Calculator — Match kLa or P/V when moving from shake flask to bioreactor
• Fermentation Economics — Estimate media and process costs for your BEVS campaign

References

1. Kwiatkowska J, Stein E, Romanenko A, et al. A beginners guide to Sf9 and Sf21 insect cell line culture and troubleshooting. Scientific Reports. 2025;15:19907. doi:10.1038/s41598-025-99812-0
2. Wilde M, Klausberger M, Palmberger D, Ernst W, Grabherr R. Tnao38, High Five and Sf9 — evaluation of host–virus interactions in three different insect cell lines: baculovirus production and recombinant protein expression. Biotechnology Letters. 2014;36(4):743–749. (Online 2013.) doi:10.1007/s10529-013-1429-6
3. Hwang SW, Jung SH, Kim HJ, Roh JY. Recent innovations in Baculovirus–insect cell expression systems for therapeutic protein production. Entomological Research. 2024;e70008. doi:10.1111/1748-5967.70008
4. Kaiser SC, Eibl D, Bock A, Tscherrig D, Meier K, Rhiel M. Scaling-up of an insect cell-based virus production process in a novel single-use bioreactor with flexible agitation. Chemie Ingenieur Technik. 2022;94(12):1950–1959. doi:10.1002/cite.202200103
5. van Oers MM, Pijlman GP, Vlak JM. Thirty years of baculovirus–insect cell protein expression: from dark horse to mainstream technology. Journal of General Virology. 2015;96(Pt 1):6–23. doi:10.1099/vir.0.067108-0