https://www.nature.com/articles/s41598-018-19991-x
Amadoriases, also known as fructosyl amine oxidases (FAOX), are enzymes that catalyze the de-glycosylation of fructosyl amino acids. As such, they are excellent candidates for the development of enzyme-based diagnostic and therapeutic tools against age- and diabetes-induced protein glycation.
Retromer-dependent defects in the sorting of these cargos impair autophagy and chaperone-mediated-autophagy (Kaushik and Cuervo, 2012) and decrease the content of hydrolases in the lysosome lumen (Seaman, 2004; Kaushik and Cuervo, 2012; McMillan et al., 2017). These hydrolases are necessary for aggregate and damaged organelle disposal via autophagy pathways
The Segura Lab at Duke, $50,000, to continue work on materials that promote healthy tissue regrowth after stroke. They say their experiments are difficult to fund because regrowing dead brain tissue is a long shot that requires a lot of out of the box thinking and is hard to explain. If you want to learn more about their work, check out http://seguralab.duke.edu. If you’re a stroke survivor and want to share your story, they’d like you to check out their Patient Connection page. They’re also looking for help spreading their ideas. If you have knowledge of both science and writing/visual communication, apply to work with them here; if you want to donate, you can do so here.
Even the best-studied genomes (those of humans, mice, nem-
atodes, fruit flies and budding yeast), have numerous genes ofunknown function. Notably, many of more recently functionally
characterized genes code for repair proteins115. Virtually all enzymes
exhibit side activities that often represent <0.1 % of the classical
activity and even sometimes <10−6for the most specific enzymes.
The products generated by side activities have been neglected
until recently when it was realized that a new category of enzymes,
metabolite repair and clearance enzymes serve to destroy the most
important side products and avoid their accumulation, which might
otherwise be toxic, causing disease.An example of a side activity is the production of l-2-hydroxy-
glutarate by l-malate dehydrogenase and lactate dehydrogenase, twoabundant enzymes. Their apparently tiny side activity (<10−6
com-
pared to the regular activity) leads to the daily production of gramsof l-2-hydroxyglutarate in humans. An FAD-linked mitochondrial
enzyme reconverts l-2-hydroxyglutarate to α-ketoglutarate, avoid-
ing its accumulation, which is toxic particularly to the brain. Themetabolic disease l-2-hydroxyglutaric aciduria, which is due to
inactivating mutations in the repair enzyme, leads to progressive
neurodegeneration and increased incidence of brain tumors.
In glycolysis, there seems to be at least as many distinct repair
reactions as the 11 classical reactions of glycolysis116. This hugediversity of side products related to glycolysis suggests that hun-
dreds and probably thousands of different side products may beformed in cells. It is likely that only some of them are eliminated by
repair and clearance enzymes. But at least for those that are elimi-
nated, it is possible to evaluate their potential toxicity in cell-basedexperiments. Such experiments have shown that some of the side
products are indeed extremely toxic.
methylguanine methyltransferase (MGMT) as an inducible degron for protein fusions. MGMT is a suicide protein that removes alkyl groups from the O6 position of guanine (O6G)åa
LYSOSOMAL REPAIR - A phosphoinositide signalling pathway mediates rapid lysosomal repair | Nature
https://www.genengnews.com/…/longevity-depends-on…
“Universal” pathway behind cell recycling offers clues to combat aging
https://newatlas.com/…/universal-pathway-cell-recycling…
A phosphoinositide signalling pathway mediates rapid lysosomal repair
(Sept 2022) https://doi.org/10.1038/s41586-022-05164-4
The PITT (phosphoinositide-initiated membrane tethering and lipid transport) pathway may have therapeutic implications for a wide range of age-dependent diseases characterized by impaired lysosomal function.
Lysosomal dysfunction has been increasingly linked to disease and normal ageing1,2. Lysosomal membrane permeabilization (LMP), a hallmark of lysosome-related diseases, can be triggered by diverse cellular stressors3. Given the damaging contents of lysosomes, LMP must be rapidly resolved, although the underlying mechanisms are poorly understood. Here, using an unbiased proteomic approach, we show that LMP stimulates a phosphoinositide-initiated membrane tethering and lipid transport (PITT) pathway for rapid lysosomal repair. Upon LMP, phosphatidylinositol-4 kinase type 2α (PI4K2A) accumulates rapidly on damaged lysosomes, generating high levels of the lipid messenger phosphatidylinositol-4-phosphate. Lysosomal phosphatidylinositol-4-phosphate in turn recruits multiple oxysterol-binding protein (OSBP)-related protein (ORP) family members, including ORP9, ORP10, ORP11 and OSBP, to orchestrate extensive new membrane contact sites between damaged lysosomes and the endoplasmic reticulum. The ORPs subsequently catalyse robust endoplasmic reticulum-to-lysosome transfer of phosphatidylserine and cholesterol to support rapid lysosomal repair. Finally, the lipid transfer protein ATG2 is also recruited to damaged lysosomes where its activity is potently stimulated by phosphatidylserine. Independent of macroautophagy, ATG2 mediates rapid membrane repair through direct lysosomal lipid transfer. Together, our findings identify that the PITT pathway maintains lysosomal membrane integrity, with important implications for numerous age-related diseases characterized by impaired lysosomal function.
https://www.upmc.com/media/news/090222-cell-repair-longevity
Intrepid biohacker gives himself infrared night vision, but at what price? - ExtremeTech this is not even remotely precise, but shows how some interventions are a #lowbar
Last April, an interesting paper was published that demonstrated the use of deglycating enzymes for treating presbyopia. Glycation is an important process in aging, and it’d be very good if we could reverse it - I believe Revel and Lento (@Dr. Kris Barnes) are working on this. Glycation is a multi-step process (with some steps taking hours and other steps taking months), and the earlier steps are generally easier to reverse. The authors from April’s paper took cadaveric human lens (a tissue that is especially hard-hit by glycation), soaked them with a weird microbial enzyme, and apparently reversed some early glycation products, as measured by lens transparency, autofluorescence, and optical power. This is a first for me! I’m unaware of any other papers that have taken human tissue and macroscopically, obviously reversed signs of glycation with an enzyme.Their “weird microbial enzyme” is fructosyl amino acid oxidase. There is actually a human enzyme, fructosamine 3-kinase, that catalyzes a similar deglycating reaction, but (1) I’m unaware of any studies that have used F3K on human tissue, (2) F3K’s ATP-dependence makes it a less plausible candidate for topical use, and (3) F3K might need to operate in tandem with some as-yet unknown players to work (F3K doesn’t break break glycation products, it just adds a phosphate - presumably some other downstream process causes the desired breakage).
CEO at Revel Pharmaceuticals
2h •
Today our team at Revel published a paper we’ve been working towards for several years: the first enzymatic reversal of an aging hallmark in human tissue.
A quick way to think about it:
Only long-lived things age. In the body, that means the structural molecules that hold us together — and over decades they accumulate damage from oxidation, glycation and crystals.
At Revel, we build first-in-class therapies for age related diseases on a simple path: Identify the damage that builds up as we age → Reverse it.
Our new paper — Reversal of Protein Aging by Enzymatic Deglycation — is the proof this approach is viable. We engineered an enzyme that reverses a form of damage the field considered irreversible for 40 years, in tissue from elderly donors.
And it’s just the start. Our first crystal-degrading enzyme is moving towards human trials as a functional cure for gout next year. Paper + coverage in the comments 
Grateful to the Revel team that made this real, and to everyone building toward a future where aging is something we can treat.
The user is asking me to explain each of the enzymes involved. Let me give a clear, substantive breakdown of the four template enzymes I repurposed — what they are, what they naturally do, their catalytic machinery, and their role in the design. This is a knowledge/explanation question grounded in the work already done, so I can answer directly without tools. I should ground the mechanistic claims in the citations already established.
Let me be precise and cite the established reference indices.
Here are the four natural enzymes I used as design templates — one per substrate class. Each is a real human enzyme with a known catalytic mechanism; the designs are new scaffolds built around each enzyme’s intact catalytic residues.
Full name: SprT-like N-terminal domain protein (gene SPRTN, also called DVC1/C1orf124). Human protein 
UniProt.
What it is: A DNA-dependent metalloprotease — a zinc-containing enzyme that cuts protein backbones. It is the master protease that resolves DNA–protein crosslinks (DPCs) during DNA replication, chopping up the crosslinked protein so the replication fork can pass [15, 17, 18].
Catalytic machinery (the SprT domain): It is a di-zinc enzyme — two zinc ions with distinct jobs:
Catalytic zinc — held by three histidines (His111, His115, His130). Glu112 sits nearby (~5.3 Å away, not touching the zinc) and acts as the general base: it activates a water molecule to attack and hydrolyze the peptide bond.
Structural zinc — held by a Cys₃His cluster (Cys169, Cys173, His201, Cys205); it holds the fold together rather than doing chemistry.
Why it’s the flagship: DPCs were your primary target, and SPRTN is the enzyme nature uses for exactly that. It also gave the strongest computational result — the designs spontaneously rebuilt the correct di-zinc site with Glu112 correctly poised as a general base, and sprtn_cons_11 docked a peptide substrate right into the active site.
Full name: Tyrosyl-DNA phosphodiesterase 2 (gene TDP2, also TTRAP/EAPII). Human protein UniProt.
What it is: A phosphodiesterase — it cuts the phosphate–oxygen bond linking a DNA end to a protein. Its specific job is repairing topoisomerase-2 DPCs: when topoisomerase II gets trapped covalently attached to DNA, it is stuck through a 5′-phosphotyrosyl bond (the enzyme’s active-site tyrosine linked to the DNA 5′-phosphate). TDP2 hydrolyzes that bond, releasing clean DNA [2, 5].
Catalytic machinery (the EEP domain): Magnesium-dependent. Glu162 directly coordinates the catalytic Mg²⁺ ion, and Asn130 completes the active-site dyad. The metal positions and activates the water that attacks the scissile phosphate.
Role in the design: Covers the topoisomerase-crosslink flavor of DPC — chemically distinct from SPRTN (phosphodiester cleavage vs. peptide-bond cleavage). The designs correctly recruited Mg²⁺ onto Glu162 unaided.
Full name: Alkyladenine DNA glycosylase (gene MPG; also called AAG or ANPG). Human protein UniProt.
What it is: A DNA glycosylase — it initiates base-excision repair by snipping a damaged base off the DNA sugar-phosphate backbone. AAG recognizes a broad range of alkylated and etheno (ring-fused) purine base adducts, flips the damaged base out of the double helix into a pocket, and hydrolyzes the N-glycosidic bond to release it, leaving an abasic site for downstream repair [6, 7, 8].
Catalytic machinery: Metal-free (unlike T1/T2). A single catalytic residue, Glu125, activates a water molecule as the nucleophile; the mechanism relies heavily on the base-flipping motion.
Role in the design: Covers DNA base adducts — the direct chemical modification of a base rather than a crosslink. Because it’s metal-free and has one catalytic residue, it validated differently (clean single-chain folds; I measured Glu125’s recovered position, 1.2–1.9 Å from native).
Caveat I flagged: the most common acrolein/4-HNE-derived guanine adducts are actually cleared substantially by nucleotide excision repair (NER), not glycosylases [11, 12] — so AAG is most directly relevant to the alkyl/etheno adducts it natively excises.
Full name: Aldo-keto reductase family 1 member C1 (gene AKR1C1). Human protein UniProt.
What it is: An NADPH-dependent oxidoreductase (reductase). It doesn’t cleave a bond — it reduces reactive carbonyls (aldehydes/ketones) to less-reactive alcohols. It is induced by oxidative stress and Michael acceptors, and protects cells against the toxicity of reactive lipid-peroxidation aldehydes like 4-HNE by reducing them [20, 21, 23, 24].
Catalytic machinery: A catalytic tetrad — Asp50, Tyr55, Lys84, His117 — working with the NADPH cofactor. Tyr55 is the catalytic acid/base; Lys84 lowers Tyr55’s pKa; Asp50 lowers Lys84’s pKa; the nicotinamide of NADPH delivers the hydride to the substrate carbonyl.
Why it’s speculative: Your target was cleaving Michael adducts. Nature largely does not run the reverse (retro-Michael) reaction to undo adducts; instead it detoxifies the reactive aldehyde before it forms an adduct, or reduces the carbonyl. So T4 explores a related but different chemistry — reductive detoxification — and I validated it structurally only (it had the highest fold confidence in the whole project, but that says nothing about whether it could act on a conjugated adduct).
| Track | Enzyme | Type | Natural reaction | Metal/cofactor |
|---|---|---|---|---|
| T1 | SPRTN | Metalloprotease | Cuts peptide bond of crosslinked protein (DPC) | 2× Zn²⁺ |
| T2 | TDP2 | Phosphodiesterase | Cuts 5′-phosphotyrosyl bond (Topo-2 DPC) | Mg²⁺ |
| T3 | AAG/MPG | DNA glycosylase | Cuts N-glycosidic bond of adducted base | none |
| T4 | AKR1C1 | Aldo-keto reductase | Reduces reactive aldehyde (not cleavage) | NADPH |
The first two are the direct DPC-cutting enzymes (the heart of your request), T3 handles the base-adduct case, and T4 is the speculative “detoxify the electrophile” angle.
This is a knowledge question grounded in enzymology. I should answer from what’s actually known about these enzymes’ native substrate scope, and be honest about (a) what’s established in the literature vs. (b) what’s speculative for the designs. I have citations already established for these enzymes. Let me think carefully about each enzyme’s real promiscuity/breadth.
Let me reason through each:
T1 SPRTN — This is the big one for breadth. SPRTN is famously a broad-spectrum, largely sequence-nonspecific protease. It cleaves DPCs regardless of the identity of the crosslinked protein — it degrades the protein component of essentially any DPC. Key nuance: it requires the substrate to be bound to DNA (ssDNA/dsDNA junctions activate it), and it cleaves the protein rather than the DNA. So its “breadth” is across protein partners, not across chemistry types. It does NOT cleave DNA, does NOT do phosphodiester or glycosidic bonds. There’s also the point that it’s a protease so it can in principle attack many peptide substrates but its activity is switched on by DNA binding. So: yes, broad across DPC protein partners; no, not across the other three chemistries.
T2 TDP2 — Native scope: primarily 5’-tyrosyl-DNA (topo II). It has a strong preference for the 5’ phosphotyrosyl linkage. It has some activity on 3’-blocking too but much weaker; TDP1 is the 3’-specialist. TDP2 also has been shown to process some other 5’ adducts. It’s a phosphodiesterase — could in principle hydrolyze other phosphodiester-type blocking groups at DNA 5’ ends. It won’t cut peptide bonds or glycosidic bonds. There’s crosstalk with TDP1/SPRTN in histone-DPC repair (ref 3 mentions TDP1 and SPRTN repair histone DPCs). So some overlap.
T3 AAG/MPG — This is a genuinely promiscuous glycosylase. It removes a WIDE range of alkylated and deaminated purines: 3-methyladenine, 7-methylguanine, 1,N6-ethenoadenine (εA), hypoxanthine, 1,N2-ethenoguanine, etc. So within its class (N-glycosidic bond of damaged purines) it is broad. But it’s monofunctional (no AP lyase), metal-free, single residue. It can’t cleave DPCs or phosphodiester bonds. Its breadth is across base adduct chemistries. The etheno adducts are exactly Michael-type/exocyclic adducts — relevant to the user’s Michael-adduct theme. Also note: the glycosylase FAMILY collectively covers a lot (NEIL1/2/3, OGG1, etc.) — NEIL3 even unhooks interstrand crosslinks. I built a neil1_core earlier too.
T4 AKR1C1 — AKRs are broadly promiscuous reductases. AKR1C family reduces many carbonyls: steroids (it’s a 20-alpha-hydroxysteroid dehydrogenase / 3-alpha-HSD), prostaglandins, and reactive aldehydes (4-HNE, acrolein, etc.). So huge substrate breadth on small-molecule carbonyls. But again — it detoxifies free reactive aldehydes; whether it acts on protein/DNA-conjugated adducts is the speculative part. Its breadth is across free electrophiles, which is actually a plus for the “detox the most common Michael acceptors” goal — acrolein, HNE, methylglyoxal-type, crotonaldehyde, etc. all plausible.
So the honest, structured answer:
Within-class breadth: substantial for all four (this is real and useful).
Cross-class breadth (one enzyme doing another’s job): essentially no — different bond chemistries.
What this means for the user’s Michael-adduct goal: the detox side (T4) and the base-adduct side (T3, etheno adducts) actually give the broadest coverage of Michael chemistry specifically; SPRTN gives breadth across DPC protein partners.
I should also be careful: the DESIGNS may have narrower or altered specificity than the parent enzymes because scaffolding can change the substrate-binding groove (I only kept catalytic residues, not necessarily the full substrate-binding surface). This is an important honesty caveat — native breadth doesn’t automatically transfer to the designs.
Let me also connect to the “most common Michael adducts” framing — acrolein, 4-HNE, crotonaldehyde, MDA, methylglyoxal on Cys/His/Lys (protein) and dG (DNA).
I don’t think I need tools here — this is well-established enzymology I can speak to with the existing citations. I could optionally do a literature search to be rigorous, but the question is conceptual and the key facts (SPRTN sequence-nonspecificity, AAG broad purine scope, AKR broad carbonyl scope, TDP2 5’-specificity) are established and already cited in my references. I’ll answer directly and ground in existing citations. I’ll add the crucial caveat about designs vs parent enzymes.
Let me structure: quick answer up front, then per-enzyme (native breadth + does it cross into other chemistries + relevance to Michael adducts), then the critical design caveat, then a coverage table.
Short answer: yes — each parent enzyme is promiscuous within its own chemistry, and that within-class breadth is real and useful. But none of them naturally crosses into another’s bond chemistry (a protease won’t cut a glycosidic bond, a glycosylase won’t touch a crosslink). And a critical caveat: that native breadth won’t automatically carry over to my designs, because I preserved only the catalytic residues, not necessarily the full substrate-binding groove that gives the natural enzyme its range.
Here’s the per-enzyme breakdown.
T1 — SPRTN: broad across protein partners, not across chemistries
This is the most naturally broad of the four. SPRTN is a deliberately sequence-nonspecific protease — it degrades the protein component of a DPC largely regardless of which protein is crosslinked, which is exactly why it’s the cell’s general-purpose DPC resolvase [15, 17, 18].
Can detox others? Across DPC types, yes — topoisomerase-1/2 DPCs, histone-DNA crosslinks, and adduct-driven crosslinks all present a protein tethered to DNA, and SPRTN’s job is to chew up that protein. It works on both single- and double-stranded DNA junctions.
Important switch: its activity is activated by DNA binding — it preferentially cleaves protein that is bound to DNA, which keeps it from acting as a rogue protease everywhere.
Won’t do: it cuts peptide bonds only. It will not hydrolyze the 5′-phosphotyrosyl bond (that’s TDP2’s phosphodiester chemistry), won’t excise bases, and won’t reduce aldehydes.
So SPRTN’s breadth = “any DPC, because I attack the protein side” — genuinely the widest coverage for the crosslink problem.
T2 — TDP2: narrow, with a little overlap
TDP2 is the specialist of the group. Its strong preference is the 5′-phosphotyrosyl linkage from trapped topoisomerase II [2, 5].
Can detox others? Somewhat. As a phosphodiesterase it can process certain other 5′-blocking phosphoadducts on DNA ends, and it participates alongside TDP1 and SPRTN in clearing histone-DPCs [3]. Its 3′-end activity is weak — that end is TDP1’s specialty (a different enzyme I did not template).
Won’t do: no peptide-bond cleavage, no base excision, no reduction.
So TDP2 adds a specific capability (the phosphotyrosyl bond) that SPRTN’s protease route handles differently — they’re complementary rather than redundant.
T3 — AAG/MPG: genuinely broad within base adducts — and this is the most “Michael-relevant” breadth
AAG is a promiscuous glycosylase by design. Natively it excises a wide panel of damaged purines [6]:
3-methyladenine, 7-methylguanine (alkylation damage)
1,N⁶-ethenoadenine and 1,N²-ethenoguanine — these etheno adducts are exactly the exocyclic, Michael-type DNA adducts formed by reactive aldehydes like 4-HNE and by vinyl-chloride metabolites
hypoxanthine (deaminated adenine)
Can detox others? Within the base-adduct class, yes, quite broadly — that promiscuity is intrinsic to the flip-and-excise mechanism with a single catalytic Glu125.
Won’t do: no crosslinks, no phosphodiester bonds, no reduction. It’s also monofunctional (it excises the base but doesn’t itself nick the backbone — APE1 does that next).
Family bonus: the broader glycosylase family covers even more — for example NEIL3 unhooks interstrand DNA crosslinks [8] — so if you wanted wider DNA-lesion coverage, other glycosylase templates are natural extensions (I actually prepped a NEIL1 core earlier that we didn’t push through).
For the Michael-adduct theme specifically, AAG’s etheno-adduct activity is the most direct hit on DNA.
T4 — AKR1C1: the broadest small-molecule detoxifier — best coverage of Michael acceptors themselves
AKRs are famously promiscuous reductases. AKR1C1 reduces a wide range of carbonyls [20, 21, 23, 24]:
Reactive lipid-peroxidation aldehydes: 4-HNE, acrolein, crotonaldehyde, and related α,β-unsaturated aldehydes — i.e. the very Michael acceptors that go on to form adducts
steroids and prostaglandins (its “day job” as a hydroxysteroid dehydrogenase)
Can detox others? For free reactive electrophiles, yes — very broadly. This is arguably the widest substrate scope of the four, and it’s well-matched to “neutralize the most common Michael acceptors.” The strategic difference: it acts upstream — reducing the electrophile before it forms an adduct — rather than cleaving an adduct that already formed.
The speculative part (unchanged): whether it can act on a carbonyl that is already conjugated to a protein or DNA base is unproven, and nature doesn’t generally run retro-Michael. So T4’s breadth is real for free aldehydes, speculative for conjugated ones.
The critical caveat: native breadth ≠ design breadth
This matters a lot for interpretation. My designs preserved the catalytic residues and rebuilt a new scaffold around them — but I did not necessarily preserve the full substrate-binding groove that gives each natural enzyme its range. Consequences:
The parent enzyme’s promiscuity is a property of its whole binding surface, not just the catalytic residues. A redesigned scaffold could end up narrower (a tighter, more selective pocket) or differently specific than the parent.
The substrate docking I ran tested one representative substrate per track, not the full panel. So the breadth described above is the parent enzymes’ documented scope, and a hypothesis for the designs — not something I’ve computationally verified for each design across many substrates.
Testing actual design breadth would mean docking (and ultimately assaying) each design against a panel of substrates.
Coverage summary
Track Within-class breadth (parent enzyme) Crosses into other chemistries? Best-covered Michael targets T1 SPRTN Broad — most DPCs, any protein partner No (peptide bonds only) DPCs (protein side), regardless of crosslinker T2 TDP2 Narrow — mainly 5′-phosphotyrosyl; minor other 5′-adducts No Topo-2 crosslinks T3 AAG Broad — alkyl + etheno purine adducts, deaminated bases No DNA base adducts, incl. HNE/vinyl-type etheno-dG T4 AKR1C1 Very broad — many free reactive carbonyls No (reduces, doesn’t cleave) 4-HNE, acrolein, crotonaldehyde (free electrophiles)
Bottom line: collectively the four cover a lot of Michael-adduct chemistry — SPRTN for the crosslink/protein side, AAG for DNA base adducts (especially etheno adducts), and AKR1C1 for neutralizing the free electrophiles upstream. But each is broad only within its own bond chemistry, and I’d want to dock/assay each design against a substrate panel before claiming the designs inherit that breadth.
If it’s useful, I can dock a substrate panel against the top design in each track (e.g. SPRTN against different DPC peptide sequences, AAG against εA / 3-methyladenine / HNE-dG, AKR1C1 against acrolein / 4-HNE / crotonaldehyde) to actually test the cross-substrate breadth computationally.
four designed enzyme chemistries can plausibly engage the widest range of HNE-derived Michael adducts and crosslinks, using three independent methods so the answer doesn’t rest on one modeling assumption. All three measure geometric/pose plausibility of reaching the catalytic anchor — not catalytic turnover.
The ranking
AKR1C1 > SPRTN > AAG ≫ TDP2 (geometric ≤5 Å gate):
Enzyme Coverage Passing / target-bond-present Median reactive→anchor Native breadth AKR1C1 0.76 19 / 25 3.57 Å broad promiscuous reductase SPRTN 0.67 12 / 18 3.23 Å broad, sequence-independent AAG 0.50 6 / 12 5.00 Å broad among base lesions TDP2 — (0/12) 0 target-bond-present n/a narrow specialist
This computational ordering is concordant with the enzymes’ documented native substrate breadth (three broad enzymes vs. one narrow specialist).
The two findings that matter most for honesty
AKR1C1 is “broadest” only by docking proximity. Under the mechanistically-strict reduction gate (carbonyl within hydride distance of NADP-C4N and oriented to the Tyr55 proton donor), its coverage collapses from 0.76 to 0.20 (5/25). The limiting factor is Tyr55 geometry, not the hydride distance — many adducts reach the nicotinamide but don’t simultaneously present the carbonyl to the acid. So its docking breadth is wide while its productive-reduction breadth is narrow — and, per the earlier boundary probe, it offers no crosslink-release chemistry at all.
TDP2’s 0/12 is a genuine mechanistic negative, not a design failure. No HNE or base lesion carries a 5′-phosphotyrosyl bond, so none presents TDP2’s target bond — exactly matching its specialized native role.
Read mechanistically, SPRTN is the enzyme best positioned to engage the widest range of bona fide crosslinks (as opposed to free/residual carbonyls): its sequence-independent protease chemistry reaches the scissile bond for both small electrophiles (in the top-scoring pose) and bulky HNE-DPCs (in a sub-optimal pose within ~0.7 kcal/mol).