Lactose Toxicity Has Quietly Been Capping HMO Yields — A New Patent From OligoScience Solves It

INFANT NUTRITION

Harleen Singh

6/17/20266 min read

This is an AI generated illustration and does not represent actual product or service.

If you've followed the human milk oligosaccharide (HMO) space over the past few years, you've seen the headline numbers: dozens of structurally distinct HMOs now in commercial or pre-commercial development, growing infant formula adoption, and a widening set of claimed health benefits extending into adult nutrition and even animal feed. What gets less attention is the unglamorous fermentation chemistry that determines whether any of this is actually affordable to produce.

A newly published international patent application from OligoScience Biotechnology GmbH (WO 2026/115113 A1) goes straight at one of the most persistent — and least discussed — bottlenecks in lactose-based fermentation: a phenomenon researchers have called "lactose killing" since the 1960s. The patent doesn't propose a new enzyme, a new pathway, or a new HMO structure. It proposes a different lactose transporter. And the data behind it suggest the difference between a mediocre fermentation run and a genuinely industrial one.

The problem: lactose is both the fuel and the poison

Most HMOs are built starting from lactose — it's cheap, abundant (a dairy byproduct, in fact), and serves as either a carbon source, a precursor, or the acceptor molecule that gets decorated with fucose, sialic acid, or other sugars to form the final oligosaccharide.

The catch is how lactose gets into the cell. In E. coli — the workhorse organism for most HMO fermentation — lactose uptake is handled by LacY, the lactose permease encoded by the lac operon. LacY is a proton-lactose symporter: for every lactose molecule it pulls in, it also pulls in a proton. Under normal, low-expression conditions this is fine. But industrial HMO production requires LacY to be overexpressed well beyond its natural levels, and the medium typically contains far more lactose than E. coli would ever encounter in nature.

When you combine high LacY expression with high lactose concentration, the proton influx overwhelms the cell. The proton motive force — the electrochemical gradient cells rely on to make ATP — collapses. ATP-starved cells respond by shutting down protein synthesis almost entirely (a survival mechanism called the stringent response), stop dividing, and start elongating into long, distorted filaments instead of dividing normally. The patent includes microscopy showing exactly this: wild-type LacY-expressing cells turn into stringy, non-dividing filaments under lactose stress, while the mutant cells stay short and rod-shaped.

For HMO production specifically, this energy crisis has a very direct downstream consequence: biosynthesis of the glycosyltransferase enzymes that actually build the HMO — in this case FutC, the α-1,2-fucosyltransferase used for 2'-fucosyllactose (2'-FL) — drops sharply. So you end up in a bind. Add more lactose to drive the reaction forward, and you choke off the very enzyme that's supposed to use it.

What earlier patents tried — and why it wasn't enough

The application does something useful for readers trying to map this space: it walks through three prior approaches and explains, in the inventors' framing, why each falls short.

One approach (WO 2012/112777) tried to sidestep the problem by separating growth and production phases — build biomass first on a different carbon source, then switch to lactose. This can work for biomass accumulation, but according to the application it doesn't fix the underlying energetic collapse once lactose-driven production actually starts.

A second approach (EP 3 218 509 B1) tackled it by simply dialing down LacY expression — less transporter, less proton influx, less toxicity. That's a reasonable engineering instinct, but it's a trade-off: you're deliberately throttling the very protein responsible for getting your substrate into the cell. The application notes this still leaves adverse growth effects, even while retaining roughly half of normal lactose influx.

The third (WO 2022/136568 A1, from Chr. Hansen HMO GmbH — now part of DSM's HMO portfolio) is the most directly comparable, and it's worth dwelling on because OligoScience's own examples use it as the benchmark. That patent introduced mutations at LacY positions 292, 293, and 294, which reduce the transporter's specific lactose transport activity. The claimed result was a roughly 20% improvement in HMO yield. Useful, but modest — and again, the mechanism is "make the transporter less active."

OligoScience's reframing: don't slow the transporter down, decouple it

The conceptual move in this patent is to stop treating "less LacY activity" and "less proton influx" as the same thing. They're not. LacY's proton coupling and its substrate (lactose) transport are mechanistically linked but separable functions, and decades of structural biology on LacY — cited extensively in the application, including work by Kaback, Smirnova, Johnson, and Franco — had already mapped which residues are involved in proton translocation versus substrate binding.

What nobody had apparently done, according to the application's account of the prior art, was take mutations that knock out or reduce proton coupling specifically (while leaving lactose-binding and translocation intact), and then overexpress that mutant above normal wild-type levels in an industrial cultivation context. Earlier structural papers studied these "uncoupled" or "uniporter" mutants purely for mechanistic insight — nobody had tested whether a cell full of them, dosed with industrial lactose concentrations, would actually grow and produce better.

The headline variant in the patent is LacY-H322R — a single substitution converting a histidine at position 322 (one of the residues directly involved in proton translocation, per the application's own structural model) into arginine. The effect, as described: the mutant lactose transporter becomes a passive, concentration-gradient-driven carrier rather than an active proton-coupled pump. It still moves lactose into the cell — just without dragging protons along for the ride. The proton motive force stays intact, ATP generation continues, FutC biosynthesis is restored, and cell division proceeds normally even at lactose concentrations that would otherwise be lethal to a wild-type-overexpressing strain.

The numbers, and the caveat that comes with them

In shake-flask comparisons, the H322R and E325A variants substantially outperformed both wild-type LacY and the L293R reference mutation (the one from the Chr. Hansen/DSM patent) across 24, 48, and 72-hour timepoints for both growth (OD600) and 2'-FL titer.

The more striking figure comes from a fed-batch fermentation comparison: a strain carrying lacY-H322R reached a final 2'-FL concentration of 75 g/L within roughly 65 hours, against 18.5 g/L for the strain expressing native lacY under the same batchwise lactose-feeding regime — roughly a fourfold difference.

A few things are worth flagging for anyone reading these numbers as more than illustrative. This is a single comparison from the patent's own examples, in a specific genetic background (an engineered BL21-derived 2'-FL strain with the lacZYA operon knocked out and Helicobacter pylori futC expressed from a plasmid). It's not an independently verified, peer-reviewed result, and patent examples are written to support claims, not to serve as a representative industrial benchmark. Still, a fourfold uplift in titer for what amounts to a single point mutation plus an expression-cassette swap is the kind of number that gets attention in strain engineering circles — if it generalizes.

The operationally interesting part: this could simplify the plant, not just the strain

The part of this patent that I think deserves more attention than the titer numbers is what it implies for process control.

Conventional lactose-fed fermentations for HMOs typically require careful management of lactose concentration — feed it too fast or let it accumulate, and you trigger the toxicity cascade described above. That means monitoring, feedback-controlled dosing, and often a "priming" pre-culture step to acclimate cells to lactose before the main production run.

The application explicitly claims that none of this is necessary with the uncoupled-permease strains: no pre-cultivation in low-lactose medium to acclimate the cells, no real-time monitoring or controlled feeding of lactose during production, and tolerance for lactose concentrations reportedly up to around 100 g/L without adverse effects (with claimed working ranges in the process claims spanning 10–90 g/L, and broader exploratory ranges discussed up to 150–250 g/L for specialized feeds). It also notes a practical rationale: concentrated lactose feed solutions are prone to precipitating and clogging feed lines, so batch-wise addition — which temporarily spikes local lactose concentration — is often the more practical approach anyway, and this strain design is specifically tolerant of those spikes.

If that holds up at scale, the value isn't just "more product per batch" — it's fewer analytical instruments, simpler control loops, and more forgiving operator behavior. For contract manufacturers and smaller HMO producers without Glycom/DSM-scale process automation, that's potentially a meaningful reduction in both capital and operating complexity, separate from the yield improvement itself.

What to watch

For R&D and strategy teams tracking the HMO production space, three things from this filing are worth keeping on the radar. First, whether OligoScience or others publish independently verifiable fermentation data at pilot or demonstration scale (the application's own definition of "industrial scale" starts at 100 L) — shake-flask and single fed-batch comparisons are a long way from a validated GMP process. Second, how the claims evolve during prosecution given the dense prior art landscape, particularly relative to DSM's overlapping permease patents — that will tell you a lot about how much real freedom-to-operate space exists here for third parties. And third, whether the "no monitoring, no priming, tolerant of batch lactose spikes" claims hold up as a practical operating advantage for HMOs beyond 2'-FL — the application gestures at a long list of other oligosaccharide targets (3-FL, LNT, 6'-SL, and many others), but the actual experimental data is built around a single 2'-FL strain.

The broader takeaway is less about this specific patent and more about where the real engineering leverage in HMO fermentation now sits. The pathway enzymes that build these molecules have received enormous attention. The transporter that gets the raw material into the cell in the first place has been comparatively overlooked — and on the evidence presented here, it may have been quietly capping yields the whole time.