For Bispecific Ab Programs, Measure Twice And Outsource Once
By Jeffrey S. Buguliskis, PhD, Deputy Chief Editor, Outsourced Pharma
You know that moment during a home renovation, whether in your own house or your favorite HGTV show, when the walls are opened up, and someone voices a loud, “Uh-oh!” Typically, it’s because whatever is behind it bears very little resemblance to what you thought or what the blueprints show. Now, suddenly, what was to be a “simple renovation” is a hodgepodge of bad wiring, questionable plumbing, and something structural being held together by what appears to be duct tape and optimism. That's when the flashbacks from the hit comedy The Money Pit start swirling through your head, and you begin to understand why Tom Hanks’ character went off the deep end.
So, why the home improvement analogy, you ask?
Well, bispecific antibody development has its own version of that moment, and it goes something like this: The biology looks seemingly terrific. The molecule is potent. A candidate has been selected. Management wants to move. Now it’s time to find a CDMO to make it.
Except that inevitably leads the CDMO to open the walls…
And that’s usually when the surprises start. Scale, purification, formulation, and analytical testing often expose liabilities that weren’t obvious when it still looked solid on the surface, before anyone pulled back the drywall.
Bispecifics benefit from the robust manufacturing foundation built for standard monoclonal antibodies, a fact explicitly recognized in FDA development guidance. However, regulatory agencies also emphasize format-specific hazards: product heterogeneity, potency variances, aggregates, and mispaired species. In short, a reliable playbook is available, but you cannot assume your molecule will automatically follow it.
Before you hand your bispecific to a CDMO, run it through the pre-outsourcing inspection below to ensure you aren't passing off hidden structural flaws:
Make Sure Your Molecule Can Survive Manufacturing
A molecule can behave beautifully at the discovery scale and still become a completely different animal under real manufacturing conditions.
One particularly instructive study followed a bispecific that appeared perfectly normal during discovery, only to precipitate during 15-liter chemistry, manufacturing, and controls production after agitation. The ultimate fix wasn’t a clever process adjustment. Researchers revisited the sequence and engineered the molecule to reduce surface hydrophobicity and improve conformational stability.
That distinction matters.
A CDMO can optimize a process. It can’t make an inherently unstable molecule stable through sheer force of project management.
Stress testing, aggregation propensity, thermal stability, solubility, and other developability assessments shouldn’t be boxes checked shortly before technology transfer. They’re evidence that the thing you’re outsourcing is actually ready to be manufactured.
Understand Exactly What Your Format Is Asking The Cell To Do
A “bispecific antibody” describes a therapeutic concept, not a single manufacturing platform.
Changing the orientation, sequence, or placement of domains can affect how efficiently a molecule is expressed. Researchers comparing different configurations of the same bispecific format in Chinese hamster ovary cells found that sequence and chain configuration directly affected cell-line construction and productivity.
That means a sponsor should understand why its architecture was selected beyond “This one had the best activity.”
How many chains need to assemble? What unwanted species can that architecture generate? How efficiently does the desired molecule express? What engineering is being used to favor correct pairing?
Those questions belong in the candidate selection process, not in the kickoff meeting with your CDMO.
Know What Else You’re Making
Here’s an uncomfortable question: When your cells produce your bispecific, how much of what comes out is actually the molecule you wanted?
With some immunoglobulin G-like bispecific designs, multiple heavy and light chains can create incorrectly paired products that closely resemble the desired molecule. A 2025 study in Nature Biotechnology described how four-chain co-expression can produce numerous mispaired species with similar biophysical characteristics, making them difficult to separate downstream.
That turns molecular assembly into a purification problem, an analytical problem, and potentially a yield problem. Before outsourcing, sponsors should have a clear picture of the product-related impurity landscape and why those impurities form.
Otherwise, downstream development becomes less like purification and more like molecular whack-a-mole.
Bispecific antibodies can generate closely related product variants that are difficult to separate downstream, making molecular design as important as purification strategy.
Decide What Needs To Be Engineered Out, Not Purified Out
There’s a natural temptation, when a molecule produces troublesome variants, to assume that downstream processing will clean everything up. Sometimes it will, but often you’re asking chromatography to solve a fundamental protein-engineering problem.
Because mispaired species often share physical properties with the intended bispecific, relying solely on downstream separation hits yield and increases costs. Researchers are increasingly turning to new upstream techniques aimed at preventing unwanted species from being secreted in the first place.
That leads to an important pre-outsourcing decision: Which liabilities are acceptable process-development challenges, and which are signals that the molecule should go back to the engineering stage?
Redesigning a molecule hurts. Redesigning it after process development, scale-up, and clinical manufacturing have begun hurts considerably more.
Build The Analytical Strategy Before You Need It
You can’t control what you can’t adequately measure.
The FDA specifically calls for characterization of attributes including antigen specificity, affinity, avidity when relevant, potency, stability, aggregates, fragments, homodimers, and other mispaired species. It also notes that potency-assay design depends on the attributes of the individual product.
For a bispecific, that can mean proving more than simply, “Yes, it binds.”
Does each arm function appropriately? Does simultaneous engagement matter to the mechanism of action? Can your methods distinguish the desired molecule from closely related variants? Can they detect changes created during processing or storage?
If analytical development is trailing process development by six months, the problem isn’t just that the assays are late. Your team may be making process decisions without fully understanding how those decisions affect the product.
Establishing robust characterization assays early prevents biopharmas from making critical process decisions in the dark.
Give The CDMO A Development Package, Not A Mystery Box
Be realistic about what technology transfer actually transfers.
A useful tech transfer package is far more than a sequence, a preliminary protocol, and a pitch deck showcasing impressive bioactivity. Your CDMO needs the molecule’s full medical history.
That means documenting which variants cropped up under stress, which formulations failed, which analytical assays are mature (and which are held together by hope), and, most importantly, where the candidate has already behaved strangely.
Nobody enjoys writing up the experiments that bombed. But your CDMO will enjoy billing you to repeat them even less.
The goal isn't to show up at your kick-off meeting with every problem solved. Early-stage biotech rarely affords that luxury. The goal is to clearly categorize your challenges: know which problems belong to the molecule, which belong to the process, and which ones you are consciously paying your external partner to solve.
Outsourcing buys you expertise, capacity, and specialized infrastructure. What it cannot do is outsource the fundamental biology and physics of your molecule.
Eventually, someone is going to open the wall.
Better to know what’s behind it before the contractor arrives.