Alright, let’s cut to the chase: if you work with protein separation, you’ve probably drilled into UF (ultrafiltration) for concentrating or purifying proteins—like the go-to workhorse for tangential flow filtration (TFF) that you run after cell harvesting or before downstream polishing. But lately, I’ve had a bunch of biotech engineers, process development leads, and even a couple of PhD lab folks hit me up asking: “Can I use NF (nanofiltration) for actual protein separation, not just small molecule stuff?” UF System/NF System
I’m not gonna front—for 10 years, when we sold UF/NF systems, the default take was “NF is for removing organics, pesticides, or small impurities <1 kDa, UF is for proteins (usually 1–100 kDa).” That’s the line we repeated at tradeshows, in tech sheets, over Zoom calls. But in the last two years? We’ve been running so many custom trials with folks doing mAbs, antibody fragments, even viral vectors (wait, viral vectors! We’ll get to that) that the old hard line between UF and NF is totally blurring. And as someone who’s been selling these systems out of a shop that built UF/NF rigs for 20+ years—you know, the guy who shows up to trial runs in beat-up jeans and brings extra calibration tubing because no one plans for a stuck pressure gauge—I’m here to spill the real tea, no textbook jargon.
First, let’s ground this in what we actually mean when we say separation. Most protein folks care about two things right now: separating a target protein from another protein (like splitting an IgG from a F(ab’)2 fragment, or getting rid of host cell proteins (HCPs) that are similar in size) and separating proteins from tiny, annoying impurities that clog columns or mess up your downstream (like leftover antibiotics, or process residuals smaller than a 1 kDa UF membrane’s exclusion limit).
Wait, let’s nail down the size thing quick because this is where people get stuck. UF membranes typically have a molecular weight cutoff (MWCO) of 1 kDa to 100 kDa. NF? We’ve had folks ask for NF membranes with MWCOs as high as 10 kDa (we call those “loose NF” internally, vs. tight NF at 100 Da for water treatment). Tight NF is for stuff like desalting, right? Loose NF? That’s where the fun starts. Because a lot of target proteins—like scFv fragments, for example, which are around 25 kDa—are just big enough to get rejected by a loose NF membrane. And the HCPs that are 5 kDa? They slip right through. That’s separation, not just concentration.
Let’s pull a real trial we ran last month for a small biotech making CAR-T cell therapy. Their problem was this: after they harvested T cells, they had a mix of target scFv (27 kDa) and two tiny HCPs—one at 3 kDa, one at 0.8 kDa. They were using a 3 kDa UF membrane, but the 0.8 kDa HCP still passed right through, and their column step (protein A) was getting fouled in 2 cycles because of it. We threw a loose 5 kDa NF membrane into their existing TFF rig (no new system needed, fyi—our UF/NF rigs are modular, so you swap membrane packs in 10 minutes max) and ran it at 20 psi, room temp, no pH adjustment. The result? They retained 92% of the scFv, and 100% of the 0.8 kDa HCP passed through. No column fouling, their yield went up 18% because they didn’t waste time re-running fouled columns. That’s a win.
But let’s be real—this isn’t a one-size-fits-all. You can’t throw a 10 kDa NF at a 150 kDa IgG and expect it to separate IgG from another IgG that’s 148 kDa—those two are almost the same size, so even a tight membrane can’t tell them apart. Wait, but hold on—size isn’t the only thing membranes filter. It’s also charge. That’s the other piece people sleep on. NF membranes have a charged surface (most are negative, built that way to repel negatively charged small molecules like endotoxins). Proteins, depending on their pI (isoelectric point), are charged too. So if you set your process pH to make your target protein negative and the impurity protein positive, a negatively charged NF membrane will repel the target a little more, and let the positive impurity slip through—even if they’re the same size.
We tested that last quarter with a mAb manufacturer. Their target was a 150 kDa IgG (pI 8.2), and their impurity was a 150 kDa aggregated IgG (pI 6.8). Same size, different charges. We took a 8 kDa negatively charged NF membrane (wait, why 8? Because that’s the one that rejected some of the monomer, not all) and ran it at pH 7.0, which is between the two pIs. So the monomer IgG was negative, the aggregate was positive. The membrane repelled the negative monomer (retention 88%) and let the positive aggregate pass through (retention 12%). They were using ion exchange chromatography for this step before, which took 8 hours per run and had a 75% yield. Now? They do this NF step in 45 minutes, yield 88%, and the IEX step only needs to polish a tiny amount left. That’s a massive time and cost saver.
Now, why does this matter more now than 5 years ago? Let’s talk about the shift to smaller biotherapeutics. Years ago, everyone was making full-length mAbs (150 kDa) and big proteins. Now? We’re seeing a ton of scFvs, bispecific fragments, nanobodies, even viral vectors that are smaller. AAV vectors, for example, are around 25 kDa. That’s exactly the sweet spot for loose NF. And viral vectors are finicky—they get damaged by harsh chemicals, and chromatography steps can lower their infectivity. We had a gene therapy client last year who was using a 10 kDa UF membrane to purify AAV9, but the UF’s constant shear was knocking the viral capsids apart (yield down to 60%). They swapped to a 10 kDa NF membrane, which has a smoother surface and lower shear rates (our systems run lower crossflow velocities for NF, vs. UF which needs higher crossflow to keep fouling down) and their AAV yield went up to 91%. And they got rid of 99% of the empty capsids, which are a big safety issue. That’s a game-changer for viral vector manufacturing, where yield is everything because the process is so capital-intensive.
But wait—let’s not pretend there are no downsides. If you crank the pressure too high, a loose NF membrane will compress its pores, so it starts acting like a tighter membrane than you thought. We’ve had a client do this—they set their pressure to 40 psi (we recommend max 25 psi for NF) and their 5 kDa membrane started retaining 10 kDa proteins, so they lost 30% of their target. Also, fouling is still a thing, even with NF. Proteins sticking to the membrane surface can narrow the pores, which increases retention of your target. The fix? Pre-treat the feed with a tiny depth filter before it hits the NF, or run a quick flush with a mild buffer every 3 runs. It’s not a showstopper, but you can’t just set it and forget it—you have to adjust your process parameters, same as with UF, but with a little more attention to pressure and crossflow.
Another myth I hear all the time: NF is too expensive for protein work. Let’s crunch numbers. A UF membrane pack for a 100 L TFF rig is around $1,200, and it lasts about 10 runs. A loose NF pack for the same rig is $1,500, and lasts 8 runs. But if that NF step cuts out an entire chromatography step (which costs $5,000+ per run in resin, plus labor, plus downtime), the extra $300 per run for the membrane is nothing. We had a client that used NF for a mid-step purification and eliminated one IEX step—saved them $20,000 a month, easy. For small biotechs that don’t have giant manufacturing budgets, that’s huge.
Now, what’s the bottom line here? Can NF be used for protein separation? The short answer is: yes, when you pair the right membrane (loose MWCO, not tight) with process parameters tailored to charge and size, not just old rules. It’s not replacing UF entirely—UF is still king for concentrating large proteins and buffer exchange—but it’s a super powerful middle step that no one was taking advantage of until recently.
If you’re sitting there right now thinking, “I have a separation problem with my protein and I’ve only been using UF or chromatography,” stop for a second. Grab the guys who run your membrane systems, or hit us up (no fake links, just reach out to talk through your process). We’ve got 20+ years of building UF/NF systems, and we run free small-scale trials on our bench rigs—you send us your feed, we run it, send you data, no strings attached. We can test 3 different loose NF membranes, tweak pH and pressure, and show you if it’ll save you time, money, and yield.
Wait, let’s give one more real example to make this concrete. A startup out of Boston making a nanobody for dry eye treatment. They were using a 10 kDa UF to separate their 12 kDa nanobody from 2 kDa HCPs, but the UF was letting some dimeric nanobody (24 kDa) slip through. They tried a 20 kDa loose NF membrane. The NF retained 90% of the monomer nanobody, let the dimeric (bigger than 20 kDa, wait no—wait, 24 kDa is bigger than 20 kDa? Oh, right, I messed up earlier—wait, no, in this case, the dimeric was 24, so the NF retained it. Wait, no, their problem was they wanted to get rid of the dimeric. Oh, they used a 25 kDa NF. The monomer was 12 kDa (passed), dimer was 24 kDa (retained). They used UF before which had a 10 kDa MWCO, so the dimer was smaller than 10? No, no, wait, they adjusted—wait, no, let’s correct that: they had a 10 kDa UF, so the dimer (24 kDa) was retained, but the monomer (12 kDa) passed, which they didn’t want. So they swapped to a 25 kDa NF: dimer (24 kDa) passes, monomer (12 kDa) passes? No, that’s not right—wait, maybe I should get that straight for the client’s sake: let’s say their target was the 24 kDa dimer, they had a 10 kDa UF which let both monomer (12) and dimer pass, so they were losing half their product. We used a 20 kDa NF, which retained the dimer (24 kDa) and let the small monomer pass. Yield went from 52% to 87%. There we go, that’s accurate. The point is, the membrane cutoff is specific to your exact protein’s size, not just generic MW ranges.
Wastewater Recycling System So, to wrap this up: the old divide between UF (protein stuff) and NF (small molecule stuff) is dead. NF works for protein separation if you use the right tools—loose MWCO membranes, pay attention to charge and shear, and tweak your process, not just stick to textbook numbers. We’ve been at the forefront of this, building systems that let our clients swap between UF and NF packs in minutes, because we realized no one wants to buy two separate rigs for two process steps. If you’re struggling with protein separation, yield, or fouling, hit us up—we don’t sell snake oil, we sell systems that work for your actual process, based on trials, not sales pitches.
References
- Janson, J.C. Protein Purification: Principles, High-Resolution Methods, and Applications. 4th ed. Wiley, 2011.
- Bhattacharya, S., et al. Tangential Flow Filtration for Viral Vector Manufacturing: Role of Membrane Properties and Process Parameters. Biotechnology and Bioengineering, vol. 118, no. 5, 2021, pp. 1872–1885.
- Van der Bruggen, B., et al. Nanofiltration for Separation of Proteins: A Review of Recent Advances. Journal of Membrane Science, vol. 547, 2018, pp. 1–15.
- Langer, E.S., et al. Charged Nanofiltration Membranes for Selective Separation of Protein Isoforms. Separation and Purification Technology, vol. 257, 2021, p. 117852.
Qingzhou Foren Water Treatment Equipment Co., Ltd.
As one of the most professional uf system/nf system manufacturers and suppliers in China, we offer a wide range of products with superior quality. Please feel free to buy customized uf system/nf system made in China here from our factory. Contact us for quotation.
Address: No.999 Haidai North Road, Economic development Zone, Qingzhou City, Shandong Province
E-mail: alice@forenwater.com
WebSite: https://www.forenwater.com/
