Engineering Guide · Vendor-Neutral

ATF vs TFF for Perfusion Cell Retention: Which Should You Pick?

ATF vs TFF perfusion cell retention side-by-side comparison Hollow fibre Diaphragm forward reverse harvest 88-95% sieve ATF · Diaphragm pump Bidirectional · self-cleaning Low shear · low fouling VS Hollow fibre Peristaltic one-way loop harvest 52-95% sieve TFF · Peristaltic pump One-directional · cross-flow Higher shear · fouling decay
Figure 1: ATF (left) uses a diaphragm pump to pulse cell suspension back and forth through a hollow fibre filter, alternating forward and reverse strokes to self-clean the membrane each cycle. TFF (right) uses a one-directional peristaltic pump that maintains continuous cross-flow shear over the membrane; the pump itself is the dominant source of cell lysis, which drives the lower 52% sieving coefficient seen in 14-day runs.
Quick Verdict

For CHO mAb perfusion at 30-100 million cells/mL, pick ATF (Repligen XCell ATF) for sieving stability and lower lysis. The alternating diaphragm flow continuously backwashes the hollow fibre, holding mAb sieving at 88-95% across 14-30 day runs. Pick TFF when you need a vendor-independent filter housing, when integrating directly into a continuous downstream skid, or when you can drive the loop with a low-shear centrifugal pump such as Levitronix instead of a peristaltic head.

Key differences at a glance

Side-by-side comparison

FactorATFTFF
Flow directionAlternating bidirectional (diaphragm)One-directional cross-flow (peristaltic or centrifugal pump)
Self-cleaning mechanismReverse stroke backwashes lumen every 1-2 sCross-flow shear only; no backwash
Mature mAb sieving coefficient88-95% maintained across 14-30 daysDecays 95% to 52% over 14 days (peristaltic pump)
Cell lysis / shearLower; diaphragm imposes lower peak stressHigher with peristaltic; comparable to ATF with centrifugal pump (Wang 2017)
Filter fouling rateSlow; self-cleaning preserves fluxFaster with peristaltic-pump-driven lysis; mitigated by wide-pore membrane (Pinto and Brower 2020)
Demonstrated max cell density~1.3 x 10^8 cells/mL in WAVE (Clincke 2013); >1.5 x 10^8 in stirred-tank2.14 x 10^8 cells/mL in WAVE (Clincke 2013)
Vendor flexibilityRepligen XCell ATF only for diaphragm housingHollow fibre cartridges sourced from any supplier
Single-use capital cost (per system)~80-200k USD (XCell ATF 2-10)~80-180k USD (Repligen KrosFlo KPS or Cytiva FlexFactory loop)
Scale range0.5 L to 5000 L (ATF 1 to ATF 10)2 L to ~2000 L
Integration with continuous downstreamPossible but typically batched into surge tankDirect loop feed into capture skid (Pinto 2020, 29-day run)

Values reflect typical published specifications and peer-reviewed comparisons. Your vendor's current datasheet takes precedence.

ATF in detail

Alternating tangential flow filtration (ATF) is the dominant cell retention technology in commercial CHO monoclonal antibody perfusion. The Repligen XCell ATF system, originally developed by Refine Technology and acquired by Repligen in 2014, is deployed in over 40 commercial mAb processes and operates at scales from 0.5 L bench (XCell ATF 1) up to 5000 L production (XCell ATF 10).

How it works

An ATF device consists of a hollow-fibre filter housing connected to a diaphragm pump via a single port on the bioreactor. A pneumatic actuator drives the diaphragm upward and downward inside the pump head with a cycle time of one to two seconds. On the downstroke, cell suspension is pushed from the diaphragm chamber through the lumen of the hollow fibres and back into the bioreactor; on the upstroke, suspension is pulled from the bioreactor through the same fibres back into the diaphragm chamber. Permeate is drawn from the shell side of the hollow fibre continuously, while the alternating direction of the lumen flow continuously backwashes any deposited debris from the membrane surface. The single-use XCell ATF SU devices use 0.2 micron polyethersulfone (PES) fibres; stainless-steel housings offer 0.2 micron, 0.5 micron, or 50 kDa molecular weight cutoff options.

Commercial vendor presence is concentrated. Repligen XCell ATF is essentially the only mature ATF platform, sold either as a stand-alone system with its own controller or as the embedded XCell ATF module on the integrated Sartorius Biostat STR + XCell ATF platform launched in 2018 for 50-2000 L scale. Cytiva sells the predecessor Wave perfusion configurations with ATF; Thermo Fisher HyPerforma and Getinge / Applikon single-use bioreactors are also routinely paired with XCell ATF devices in process development.

When ATF wins

ATF wins decisively in three scenarios. First, commercial CHO mAb perfusion. The combination of stable 88-95% mAb sieving and low cell lysis means a 14-day run holds steady-state harvest titre with minimal high-molecular-weight species enrichment in the bioreactor. Pappenreiter et al. (2023) reported ATF sieving holding at 88-95% versus a decline to 52% in conventional peristaltic-pump TFF. Second, N-1 seed train intensification. Running an ATF-equipped N-1 reactor to 50-100 million cells/mL lets you seed the production bioreactor at 10x the conventional density, compressing the seed train by 15-30 days. Third, extended viral vector and vaccine campaigns. The reverse stroke prevents the filter from fouling on the high-DNA, high-debris broth typical of lentiviral and AAV production, extending filter life beyond what TFF can sustain on the same feedstock.

TFF in detail

Tangential flow filtration (TFF) drives cell suspension through a hollow-fibre cartridge in a single direction using an external pump, with permeate drawn through the membrane perpendicular to the bulk flow. The same physics that underpins TFF as a downstream concentration and diafiltration tool, covered in our TFF membrane sizing guide, is what makes it an option for upstream cell retention. The critical engineering question is which pump drives the cross-flow loop.

How it works

A TFF cell retention loop takes suspension out of the bioreactor through a recirculation port, pushes it across a hollow-fibre cartridge with a peristaltic or centrifugal pump, and returns the retentate back to the bioreactor. The cross-flow shear at the membrane surface prevents most cells from depositing onto the fibre wall, and permeate (cell-free supernatant containing product) is collected from the shell side. The dominant single-use systems are the Repligen KrosFlo KPS TFF (sold separately from XCell ATF), the Cytiva TFF perfusion configurations (FlexFactory recirculation loops), and Cytiva and Sartorius hollow-fibre cartridges (typically 0.2 micron PES or polysulfone). Hollow-fibre cartridges from Meissner and Merck Millipore are commonly substituted into vendor-agnostic skids.

The pump choice is the defining decision. A peristaltic pump pinches the tubing repeatedly with rollers, generating a pulsatile flow with high local shear at the roller contact points. Wang et al. (2017, Journal of Biotechnology) demonstrated that the peristaltic pump itself, not the absence of alternating flow, is the dominant source of cell lysis in TFF perfusion. Replacing the peristaltic pump with a low-shear centrifugal pump such as a Levitronix magnetically-levitated centrifugal pump restored cell viability, particle concentration, and product sieving to ATF-equivalent levels. This finding has reshaped TFF perfusion engineering since 2017, and most new TFF perfusion installations now specify a centrifugal pump rather than a peristaltic head.

When TFF wins

TFF wins in three contexts. First, very high cell density. Clincke et al. (2013) achieved 2.14 x 10^8 cells/mL with TFF in a WAVE bioreactor versus 1.32 x 10^8 cells/mL with ATF; the ATF ceiling was driven by diaphragm vacuum struggling against the viscous high-density broth rather than by ATF itself, but the difference is real at the extreme end of the density envelope. Second, integrated continuous bioprocessing. The TFF permeate loop can feed directly into a downstream capture chromatography skid without a surge tank intermediate, an architecture demonstrated by Pinto and Brower (2020, Biotechnology and Bioengineering) running 29 days uninterrupted with a wide-surface-pore microfiltration membrane. Third, vendor-independent hardware. TFF loops use commodity peristaltic or centrifugal pumps and off-the-shelf hollow-fibre cartridges, avoiding the Repligen-specific diaphragm housing of XCell ATF.

Pros and cons

ATF (alternating tangential flow)

Advantages

  • Stable mAb sieving at 88-95% across full perfusion run; no decay from progressive fouling
  • Lower hydrodynamic shear and cell lysis than peristaltic-pump TFF (Karst et al. 2016)
  • Mature commercial cGMP track record: 40+ commercial mAb processes and 500+ deployed sites
  • Linear scale-up from XCell ATF 1 (0.5 L) to ATF 10 (5000 L) without changing physics

Disadvantages

  • Single-vendor lock-in on diaphragm hardware and controller (Repligen XCell ATF)
  • Diaphragm pneumatic vacuum can struggle with viscous broth above 150 million cells/mL
  • Single-port connection to bioreactor adds risk of gas bubble accumulation in dip tube
  • Less natural fit for continuous-downstream integration than TFF permeate loop

TFF (tangential flow filtration)

Advantages

  • Vendor-independent hardware: any commodity peristaltic or centrifugal pump plus any hollow-fibre cartridge
  • Reached the highest published cell density in head-to-head WAVE comparison (Clincke 2013: 2.14 x 10^8 cells/mL)
  • Centrifugal-pump TFF (Levitronix) brings sieving and lysis to ATF-equivalent levels (Wang 2017)
  • Permeate loop integrates directly into continuous downstream capture skids

Disadvantages

  • Classic peristaltic-pump configuration suffers progressive sieving decay (95% to 52% over 14 days)
  • Higher cell lysis from peristaltic pump generates more debris, accelerating fouling
  • HMW species accumulate in the bioreactor when product is partially retained, complicating quality
  • Centrifugal-pump upgrade adds Levitronix-class hardware cost and validation effort

Which should you choose?

Pick based on the dominant constraint in your process. For most CHO mAb perfusion teams the answer is ATF; the TFF cases are real but specific.

CHO mAb perfusion, 50-1000 L

Steady-state 30-100 million cells/mL, 14-30 day production runs, commercial cGMP path. ATF gives stable sieving with low lysis; Repligen XCell ATF is the proven default with a 40+ commercial process track record.

Choose ATF

Integrated continuous downstream

Direct loop feed from upstream permeate into Protein A capture, no surge tank. The TFF permeate stream connects naturally; Pinto and Brower (2020) ran 29 days continuous from bioreactor to chromatography with a wide-pore TFF membrane.

Choose TFF

Vendor-independent procurement

Sourcing flexibility, multi-supplier risk reduction, or a process licensed to a CDMO using non-Repligen hardware. TFF with centrifugal-pump architecture lets you swap pumps and cartridges across vendors.

Choose TFF

Viral vector / vaccine perfusion

Lentiviral, AAV, or vaccine campaigns with high-DNA, high-debris broth. The reverse stroke on ATF prevents the filter fouling that limits TFF runs on the same feedstock; ATF extends filter life by 2-4x in lentiviral campaigns.

Choose ATF

Real-world use cases

Typical setups where bioprocess teams have converged on one cell retention device or the other.

CHO mAb, commercial perfusion
XCell ATF 6 on Biostat STR 200

Steady-state 60 million cells/mL, 21-day production, 1 vessel volume per day perfusion rate. Stable 90% mAb sieving from day 3 to day 21. The Sartorius + Repligen integrated platform is the default at 50-2000 L scale.

CHO mAb, N-1 intensification
XCell ATF 4 on N-1 seed

5-day N-1 perfusion intensification to 50 million cells/mL, seeded into a 2000 L fed-batch production reactor at 5 million cells/mL. Compresses seed train by 15-30 days versus conventional N-1 batch.

Lentiviral vector, single-use
Centrifugal-pump TFF, 50 L

HEK293 suspension producing LV at 1-2 million cells/mL with continuous harvest into downstream concentration. Levitronix centrifugal pump on a Cytiva XDR50 with 0.2 micron PES cartridge. Daily filter swap on heavily debris-laden broth.

Integrated continuous mAb
Wide-pore TFF + Protein A capture

500 L perfusion bioreactor with Repligen KrosFlo KPS TFF on a centrifugal pump driving permeate directly into a Cadence BioSMB Protein A skid. 29-day integrated continuous campaign with no surge tank.

Sizing the perfusion loop and bleed for your retention device?

Run cell-specific perfusion rate (CSPR), vessel volumes per day (VVD), bleed rate, and steady-state VCD calculations for ATF or TFF cell retention efficiencies (95-99%+).

Open the Perfusion Calculator

Cost and lifecycle considerations

Total cost of ownership has four moving parts

Capital cost of the retention device and pump, recurring single-use consumables (hollow-fibre cartridges, tubing assemblies, single-use ATF housings), product loss in the retentate from sub-100% sieving, and labor/validation overhead. ATF wins on product loss and labor; TFF wins on procurement flexibility.

At 200 L production scale, capital cost is broadly similar between ATF and TFF. A single-use Repligen XCell ATF 6 controller plus device package lands in the 80-150k USD range; an equivalent Repligen KrosFlo KPS TFF or a centrifugal-pump TFF skid built on a Levitronix BPS-200 falls in the 80-180k USD range depending on whether the pump is commodity or magnetic-levitation. Both technologies use comparable hollow-fibre cartridge consumable economics (1-3k USD per cartridge, with cartridge lifetimes of 14-30 days).

The dominant cost differentiator is product loss. A peristaltic-pump TFF system holding mAb sieving at 70% on average across a 21-day run loses 10-15% more product than an ATF holding sieving at 92%. For a 200 L CHO perfusion at 1 g/L titre and 1 VVD, that is 200 g/day x 22% lost yield equals roughly 900 g of mAb across a 21-day run. At about 1000 USD per gram manufactured value, this single line item dwarfs the capital and consumable difference between the technologies.

Cost component (200 L scale, single-use)ATFTFF (peristaltic)
Capital: controller + device~80-150k USD~80-180k USD
Consumables / 21-day run~2-4k USD (single-use cartridge + tubing)~1.5-3k USD (cartridge + tubing)
Operator labor / monthLow (single-port, integrated controller)Medium (separate pump + sampler integration)
Product loss penalty (21-day run, 1 g/L)~5-10% (sieving 88-95%)~20-35% (sieving decay 95% to 52%)
3-year TCO (10 runs/yr, with product loss)~1.4M USD~1.7M USD (peristaltic) / ~1.45M USD (centrifugal)

Vendor landscape

Major vendors in each camp, with one-line positioning notes.

ATF vendors

TFF vendors

Frequently asked questions

What is the main difference between ATF and TFF for perfusion?
ATF uses a diaphragm pump that pulses the cell suspension back and forth through a hollow fibre filter, so the alternating flow continuously self-cleans the membrane. TFF uses a one-directional peristaltic or centrifugal pump that pushes the suspension across the membrane in a single direction, relying on cross-flow shear to control fouling. The practical consequences are different sieving coefficients (88-95% for ATF, declining to 52% over time for peristaltic-pump TFF), different shear histories, and different vendor cost structures.
Which gives higher cell density, ATF or TFF?
Both routinely reach 80-150 million cells/mL in commercial CHO perfusion. Clincke et al. (2013) demonstrated 2.14 x 10^8 cells/mL with TFF and 1.32 x 10^8 cells/mL with ATF in a WAVE bioreactor, but the ATF ceiling was driven by dip-tube gas-bubble interference, not by ATF itself. For typical 30-100 million cells/mL steady-state operation, both technologies deliver equivalent biomass.
Why does TFF retain product (low sieving) and ATF does not?
The dominant cause is cell lysis from the peristaltic pump traditionally used to drive TFF cross-flow. Wang et al. (2017) showed that replacing the peristaltic pump with a low-shear centrifugal pump (e.g. Levitronix) restored TFF sieving to ATF-equivalent levels. Membrane fouling from lysed-cell debris, DNA, and host-cell protein binds product and drops sieving from 95% to 50% over a 14-day run. The fix is either the alternating flow of ATF or a centrifugal-pump TFF; the membrane chemistry matters less than the pump.
Is ATF gentler on cells than TFF?
Yes, by a meaningful margin. The diaphragm pump on the ATF imposes less hydrodynamic stress than a peristaltic pump on a TFF loop, and the alternating direction reduces residence time of any cell in the highest-shear region. Karst et al. (2016) measured smaller maximum stress in ATF at matched mixing and kLa. A centrifugal-pump TFF closes much of the gap, but classic peristaltic-pump TFF runs measurably higher lysis.
Can you scale ATF and TFF to commercial mAb manufacturing?
Yes for both. The Repligen XCell ATF range covers 0.5 L to 5000 L bioreactors, and over 40 commercial mAb processes are deployed on ATF cell retention. Sartorius and Repligen co-developed an integrated Biostat STR + XCell ATF system at 50-2000 L. KrosFlo and Cytiva TFF perfusion systems scale similarly. Filter housings are single-use up to about 1000 L; stainless-steel housings dominate above that scale.
Does ATF or TFF foul faster?
Peristaltic-pump TFF fouls fastest because lysed-cell debris coats the inside of the hollow fibres and is not backwashed off. ATF fouls more slowly because the reverse-flow stroke partially backwashes the lumen each cycle. A centrifugal-pump TFF with low cell lysis fouls at a comparable rate to ATF. Wide-surface pore microfiltration membranes (Pinto and Brower 2020) further reduce TFF sieving decay by minimising biomass deposition at the membrane surface.
What is the price difference between ATF and TFF cell retention?
Capital cost is broadly similar at 80-200k USD per system at production scale, but the cost structure differs. ATF systems carry a Repligen-specific diaphragm housing and dedicated controller; TFF systems can use general-purpose peristaltic or centrifugal pumps with off-the-shelf hollow fibre cartridges. ATF single-use consumables tend to be more expensive per run; TFF consumables are commoditised. 3-year TCO at scale is roughly within 15-20% between the two when product loss is included.
Is TFF ever preferred over ATF?
Yes, in three scenarios. First, very-high-density runs (above 150 million cells/mL) where the ATF diaphragm vacuum struggles with viscous broth. Second, processes where filter housings need to be vendor-independent and you want to source hollow fibre cartridges from any supplier. Third, integrated continuous bioprocesses where the TFF cell retention loop also feeds directly into a downstream capture skid, simplifying the connection. For most CHO mAb perfusion at 30-100 million cells/mL, ATF is the default.

Resources and references