Here is the standard story about ADA2 deficiency: patients carry mutations in the ADA2 gene, the protein fails to get secreted, enzyme activity in the blood drops to near zero, and inflammation runs unchecked. Treat with TNF inhibitors. Done.
Except that story has a gap. A big one. What is the protein actually doing inside the cell before it ever reaches the bloodstream? Nobody had looked carefully at the intracellular life of ADA2 in primary patient cells. When this team did, they found something nobody expected: healthy macrophages carry two distinct forms of ADA2. Patient macrophages carry only one.
The missing form is not the secreted protein. It is a specific, stable, partially glycosylated version that lives inside the cell. Its absence turns out to be a consistent cellular hallmark of the disease, present across every pathogenic missense variant tested in this cohort of ten DADA2 patients.
The discovery came from a straightforward experiment: differentiate patient monocytes into macrophages, run a western blot, and look at what ADA2 protein is actually present. In healthy control macrophages, two bands appear at roughly 60 kD and 57 kD. The lower, LMW band is actually the more abundant of the two intracellularly. In DADA2 patient macrophages, only the upper 60 kD band shows up. The LMW band is gone.
Worth pausing on the monocyte data too. Patient monocytes have normal ADA2 mRNA levels but almost no detectable protein. Macrophage differentiation partially rescues mutant ADA2 expression, which is what made the two-band comparison possible in the first place. Without that differentiation step, the signal is too weak to see anything.
CD14+ monocytes were isolated by magnetic positive selection and differentiated over ten days with GM-CSF (20 ng/mL) or M-CSF (50 ng/mL). ADA2 was detected by western blot using the anti-ADA2 clone EPR25430-131 (abcam #ab288296), which the authors note is superior at detecting the LMW form specifically.
result
A 3 kD size difference between two forms of the same protein points straight at post-translational modification. ADA2 has four N-glycosylation sites, so glycans were the obvious suspect. The team stripped glycans off using PNGase F, which removes all N-linked sugars, and watched what happened to the two bands.
After complete deglycosylation, both the HMW and LMW forms collapsed into a single band at 53 kD. The HMW form lost 7 kD of glycan mass. The LMW form lost less, but critically, it did not produce a band smaller than the deglycosylated HMW form. That rules out the LMW form being glycan-free ADA2. LMW-ADA2 is partially glycosylated, not unglycosylated.
Two additional experiments nail this down. Tunicamycin, which blocks the very first step of N-glycosylation, shifts the HMW band downward but leaves the LMW band completely unchanged in size. Castanospermine, which inhibits ER glucosidases I and II, causes the HMW band to shift upward in both healthy and DADA2 macrophages, but again the LMW band is unaffected. The LMW form has already finished its glycan processing before any of these inhibitors can touch it. That is the signature of a stable, mature protein, not a biosynthetic intermediate on its way to being secreted.
The ~57 kD LMW-ADA2 form is a stable, partially glycosylated intracellular protein. Its absence in DADA2 macrophages is not a side effect of low expression. It is a specific, consistent feature of pathogenic ADA2 variants, present across all ten patients in this cohort.
result
So what happens to pathogenic ADA2 variants inside the cell? They get stuck. Cycloheximide chase assays, which block new protein synthesis and let you watch existing protein disappear over time, show that mutant ADA2 persists inside cells far longer than wild-type. The variants with the most complete secretion block, p.T129P and p.R169Q, show the highest intracellular levels 24 hours after protein synthesis stops. Blocking secretion of wild-type ADA2 with brefeldin A phenocopies this accumulation, confirming that impaired secretion is the primary driver of intracellular buildup.
The accumulation has a structural consequence. Pathogenic variants form intracellular dimers, visible as high-molecular-weight bands on non-reducing western blots. This was previously reported for only one variant (p.L351Q). Here it shows up across multiple pathogenic variants, and the degree of dimer formation correlates with how severely secretion is impaired.
The most provocative finding in this section: when p.R169Q is co-expressed with wild-type ADA2, it impairs secretion of the wild-type protein more than the transfection ratio would predict. A dominant-negative effect on trafficking. Serum ADA2 protein levels in heterozygous carriers, who carry one mutant and one wild-type copy, are only ~25% of healthy control levels, while intracellular protein in their monocytes sits at ~75% of healthy control. The math suggests the mutant protein is actively dragging the wild-type down with it.
Pathogenic ADA2 variants don't just fail to get secreted. They form intracellular aggregates that appear to impair secretion of wild-type ADA2, potentially explaining why even heterozygous carriers show serum protein levels at only ~25% of healthy controls.
result
Misfolded proteins that cannot be secreted have to go somewhere. Interactome analysis of the pathogenic variant p.R169Q, pulled down by immunoprecipitation and identified by mass spectrometry, shows a clear enrichment of ER quality control proteins compared to wild-type ADA2. The same enrichment appears for wild-type ADA2 when N-glycosylation is blocked with tunicamycin, which is a clean confirmation that glycosylation is what keeps ADA2 out of the folding machinery's hands under normal conditions.
When ER-to-Golgi transport is blocked with brefeldin A, mutant ADA2 gets degraded rather than accumulating. In HEK293T cells, that degradation can be rescued by the proteasome inhibitor delanzomib or the lysosome inhibitor chloroquine, implicating both pathways. In primary patient macrophages, neither inhibitor rescues the degradation, which the authors attribute to the two pathways compensating for each other when one is blocked. That discrepancy between the overexpression system and primary cells is worth flagging: HEK293T cells are not macrophages, and the behavior of endogenous mutant ADA2 in its native cellular context differs in ways that matter for interpreting mechanism.
The ER stress angle, which might seem like an obvious consequence of all this misfolding, does not pan out in primary cells. BiP levels are not elevated in DADA2 patient monocytes compared to healthy controls. Some pathogenic variants do trigger BiP and CHOP upregulation in HEK293T overexpression, but that effect disappears when transfected DNA amounts are titrated down to near-physiological levels. Supraphysiological overexpression artifacts are a real concern here, and the authors are appropriately cautious about it.
result
The absence of LMW-ADA2 is now a cellular hallmark of DADA2, consistent across all pathogenic missense and deletion variants tested. That is a clean, reproducible phenotype in primary patient cells, which is rarer than it sounds in rare disease research.
The bigger question the paper opens is what LMW-ADA2 actually does. Its glycan structures are sensitive to Endo H, suggesting high-mannose oligosaccharides, which is consistent with a potential lysosomal localization. One prior study proposed ADA2 functions as a lysosomal DNase. If LMW-ADA2 is the lysosomal form, its absence in DADA2 macrophages would point to an intracellular function that current diagnostic assays, which measure secreted enzyme activity in serum, would miss entirely.
The cystic fibrosis parallel the authors draw is not just rhetorical. The most common CFTR variant, p.deltaF508, is almost entirely cleared by ER-associated degradation, and small-molecule correctors that stabilize its folding have transformed patient outcomes. If ADA2 misfolding follows similar logic, correctors that stabilize mutant ADA2 and restore its intracellular processing could be a tractable therapeutic direction. The data here provide the mechanistic foundation to start asking that question seriously.
result