Can a cytidine deaminase biodegrader sensitise cancers to deoxycytidine analogues?

CDA inactivates the drug before it can act

Deoxycytidine (dC) analogues such as gemcitabine and cytarabine are prodrugs: they must enter a cell and be activated to kill it. Cytidine deaminase (CDA) works against that logic. It converts cytidine and deoxycytidine into uridine and deoxyuridine, and in doing so deaminates dC analogues into inactive metabolites. The paper states that up to 90% of gemcitabine can be inactivated by CDA, which is why the enzyme is treated as a resistance node rather than a bystander.

The obvious fix would be to inhibit CDA's catalytic site, and small molecules such as tetrahydrouridine and zebularine do that. The paper's premise is that these inhibitors have shown limited efficacy and/or off-target effects, so the authors pursue a different mechanism: removing the CDA protein itself rather than blocking its active site.

That distinction matters for how the results should be read. A catalytic inhibitor competes with substrate at an enzyme; a degrader recruits the cell's ubiquitin-proteasome system to eliminate the enzyme. The paper's central claim is that this protein-level removal is specific enough and potent enough to make tumour cells more sensitive to dC analogues.

A two-step screen pairs an anti-CDA VHH with SPOP

The construct is an intracellular antibody fragment, a VHH, selected to bind CDA, fused to the SPOP E3 ubiquitin ligase domain. The VHH provides target recognition; SPOP provides the degradation machinery. The paper calls the resulting molecule a biodegrader, part of the broader bioPROTAC family in which an E3 ligase is attached to a binding module so that the target is ubiquitinated and degraded.

Finding a working pair was not trivial, and the paper describes a two-step cell-based screen. First, the authors tested which E3 ligase, fused to an anti-GFP VHH, could deplete a GFP-CDA fusion in cells; only the SPOP-based version led to GFP-CDA degradation. Second, they subcloned a polyclonal sub-library of anti-CDA VHHs, obtained after four rounds of phage display, in frame with the SPOP domain, and screened 200 biodegrader plasmids by transient transfection in HeLa S3 cells stably expressing GFP-CDA. Degradation was quantified by flow cytometry as GFP median fluorescence intensity in transfected (mCherry-positive) cells, normalised to a negative-control biodegrader.

Thirteen positive hits were confirmed, and the three best biodegraders were tested against endogenously expressed CDA by Western blot. The paper reports that CDA biodegrader 1 binds CDA with affinity in the pM range and biodegrader 2 in the nM range, measured by surface plasmon resonance, which the authors use to explain why biodegrader 1 is the stronger of the two.

CDA falls while the rest of the proteome holds

The mechanistic case rests on three checks in AsPC-1 pancreatic cancer cells, using a doxycycline-inducible lentiviral system. First, CDA protein depletion was dose-dependent with CDA biodegraders 1 and 2 but not with the control biodegrader. Second, biodegrader expression was detectable from 5 h after doxycycline induction and persisted to 72 h, accompanied by complete CDA degradation; CDA mRNA levels were unaffected, locating the effect at the protein level. Third, the proteasome inhibitor epoxomicin blocked CDA depletion, tying the effect to proteasomal degradation rather than to altered transcription or translation.

Specificity was assessed by whole-cell mass spectrometry in AsPC-1 cells expressing CDA biodegrader 1 for 10 h. The paper reports that CDA was the only significantly detected protein showing more than 50% depletion (P < 0.05), and that SPOP was the most enriched protein. No proteins were significantly downregulated when the control biodegrader was expressed under the same conditions, and the few proteins that increased in either condition argue against an off-target effect of doxycycline itself.

This is the evidence that separates the approach from a catalytic inhibitor in the paper's own framing: the claim is not merely that CDA activity falls, but that CDA protein is removed with little detectable collateral depletion across the proteome at the time point measured.

Sensitisation does not track baseline CDA levels

The authors extended the work across pancreatic, lung and acute myeloid leukaemia cell lines, first mapping CDA protein expression by Western blot. In pancreatic lines, Capan-2 and AsPC-1 expressed more CDA than BxPC-3, while MIA PaCa-2 had no detectable CDA and served as a control. In lung lines, H441 expressed the most CDA, followed by PC-9 and HCC4006, and the non-cancerous lung lines BEAS-2B and MRC-5 had no detectable CDA.

Two results here shape how the strategy should be understood. Cells that did not express CDA were neither sensitised to dC analogues nor affected in baseline viability, which the authors present as confirmation of specificity. And across the panel, baseline CDA expression did not directly correlate with IC50 values for dC analogues. The paper connects this to prior observations, including the GemPred signature, that CDA expression alone cannot predict gemcitabine sensitivity, and points to additional factors such as the hENT1 transporter.

The authors' conclusion from this comparison is that CDA-targeted degradation is sufficient to sensitise cancer cell lines to dC analogues regardless of their baseline CDA levels. That is a claim about sufficiency within the tested panel, not a claim that CDA level is irrelevant to which tumours respond.

Tumour growth effects split by model, not by proliferation

In vivo, the CDA biodegrader combined with gemcitabine inhibited pancreatic and lung tumour growth. But the paper also reports a split that readers should not smooth over: the CDA biodegrader alone suppressed BxPC-3 CDX tumour growth, yet had no effect on H441 CDX tumours, despite no measurable differences in 2D proliferation between the two lines.

The authors interpret this as context dependence, suggesting that tumour growth inhibitory effects of CDA degradation may reflect cell line-specific gene regulatory programmes that emerge under 3D growth conditions. They draw an explicit parallel to KRAS signalling, where phenotypic effects are often more pronounced in 3D than in 2D systems and vary across cell lines.

For a reader weighing the strategy, this means the in vivo case is strongest where the biodegrader is paired with a dC analogue. Single-agent activity of the degrader was model-specific in the paper's own comparisons, so the combination setting is the one the evidence most directly supports.

VLP-delivered mRNA sensitises tumours transiently

Because biodegraders cannot cross cell membranes, the paper needed a delivery route. The authors used non-integrative virus-like particles (VLPs) carrying mRNA encoding the biodegrader, built on the FlashRNA vector. In PC-9 and AsPC-1 cells, VLP-delivered CDA biodegrader degraded CDA, while control VLPs did not.

The kinetics constrain how the delivery should be judged. After a single transduction, the biodegrader was expressed and CDA was depleted for 2 days in PC-9 cells and 3 days in AsPC-1 cells, after which CDA levels gradually returned to baseline; the authors state that an additional transduction would be required for experiments beyond 2 to 3 days. Accordingly, the sensitisation experiment transduced cells on day 1 and day 3, treated with gemcitabine on day 1, and counted cells on day 6. Gemcitabine alone, or with control VLPs, reduced cell number by 50% versus untreated controls; with CDA biodegrader VLPs the reduction was 85% in AsPC-1 and 90% in PC-9 cells, a significant improvement over gemcitabine with control VLPs. Apoptosis increased significantly in PC-9 but not in AsPC-1.

In mice, intratumoural FLuc-mRNA VLPs produced bioluminescent signal exclusively in tumours, following a dose-response across 2.43, 6.08 and 12.2 µg p24, with no signal in lung, liver, kidneys, spleen or heart; signal peaked at 6 h and had almost disappeared by 48 h. In the PC-9 CDX model, gemcitabine plus CDA biodegrader VLPs significantly inhibited tumour growth compared with all other groups, with roughly 40 to 60% CDA degradation in tumour tissue and no body-weight effect versus vehicle or gemcitabine alone.

Intratumoural dosing is the binding constraint

The paper is explicit that this is a proof-of-concept study, and the delivery design carries most of the translational weight. VLPs were injected intratumourally, which the authors note is not applicable in patients with metastatic disease, although they also point to a prior phase I/II clinical gene therapy trial of intratumoural injection in pancreatic cancer as support for the route's feasibility.

Dosing frequency is a second constraint the authors name themselves: mice received VLPs twice weekly, a schedule they describe as unlikely to be clinically practical, with once every two weeks more feasible. Reaching that schedule would depend on extending the half-life of both the delivered mRNA and the biodegrader protein, and repeated dosing raises immunogenicity questions about the vehicle and its cargo that the paper says will require careful evaluation.

The authors also state what they do not expect: that clinical implementation would alter dC analogue dosing regimens or directly expand their therapeutic index. The intended benefit is enhanced tumour sensitivity to those agents. That framing is the most useful way to read the paper's priority, because it places the value in the combination rather than in the degrader as a stand-alone drug.

Combination sensitisation is the result to carry forward

Across the paper's own comparisons, the consistent signal is the combination: CDA biodegrader plus gemcitabine inhibited pancreatic and lung tumour growth in vivo, and VLP-delivered biodegrader mRNA raised gemcitabine-induced cell-number reduction from 50% to 85% in AsPC-1 and 90% in PC-9 cells. The single-agent tumour-growth effect, by contrast, appeared in BxPC-3 but not H441 CDX models.

The paper's own proposed next steps follow from that split: optimising the biodegrader, potentially with E3 ligases more active than SPOP, and refining delivery, either by increasing intratumoural VLP concentration or by turning to platforms such as nanoparticles. It also notes that dC analogues are used in ovarian and bladder cancer and decitabine in myelodysplastic syndromes, and suggests the approach could extend there depending on CDA expression levels, which the paper presents as a possibility rather than a demonstrated result.

The practical reading is therefore narrow and specific: targeted CDA degradation is a validated sensitisation mechanism in the cell lines and mouse models tested, strongest when paired with a dC analogue, and still gated by a delivery route and dosing schedule that the authors themselves describe as not yet clinically practical.

Sources used in this answer