ADAM17/TACE (P78536) feasibility workup
Recommendation up front: produce the prodomain + catalytic zymogen (residues 18–474) as a secreted CHO transient build (primary), with E. coli cell-free, same construct as the hedge. The catalytic domain alone folds in E. coli PUBMED 9520379, but the prodomain is required for secretion of functional enzyme in cells PUBMED 15581423, and ADAM17's N-glycans measurably change its activity and inhibitor response PUBMED 24361716 — so for "assay-ready," CHO is the route that gives you the molecule your client's assay expects.
Two caveats before the detail, stated plainly:
- A self-caught error in my own data. My first computation run used a transcribed sequence that came back 825 aa with 34 Cys; UniProt's canonical sequence is 824 aa. I re-pulled the authoritative FASTA from UniProt REST mid-turn and confirmed 824 aa / 35 Cys. The biophysical table below was computed on the first (wrong) sequence, so MW values may be off by ~0.1 kDa and residue indices past ~226 shift by one. Everything residue-indexed below is corrected against UniProt's own feature annotation, which I did retrieve in full. No decision in this report turns on the discrepancy.
- The chart you asked for in §7 was not produced. This turn hit its processing limit before I could render it. See "What's outstanding" at the end.
1. Sequence and domain reality
Source record: UniProt P78536 (ADA17_HUMAN), canonical isoform A, 824 aa, 93.0 kDa, EC 3.4.24.86, MEROPS M12.217 UNIPROT P78536.
| Feature | Residues | Notes |
|---|
| Signal peptide | 1–17 | type I membrane protein |
| Prodomain (propeptide) | 18–214 | PROPEP, evidence at protein level |
| Mature chain start | 215 | furin/PC cleavage at the 214/215 junction |
| Catalytic domain (peptidase M12B / ZnMc) | 223–474 | UniProt domain call |
| Disintegrin | 475–563 | Pfam PF00200, SMART DISIN |
| Cysteine-rich / MPD + "crambin-like" | 565–671 | UniProt annotates REGION 603–671 as crambin-like |
| CANDIS (juxtamembrane) | ~655–671 | required for shedding activity |
| Transmembrane | 672–692 | |
| Cytoplasmic tail | 693–824 | disordered (732–824); 2 SH3 motifs (731–738, 741–748) |
UNIPROT P78536
Constructs I would actually express
- (a) Prodomain + catalytic, single chain: residues 18–474. This is the zymogen — prodomain 18–214 fused to catalytic 215–474 with the native 214/215 junction intact, so it can be activated by furin in situ or in vitro. If the client wants the prodomain plus the whole ectodomain, that is 18–671.
- (b) Catalytic domain alone: residues 215–474. Note that the successful crystallographic constructs were slightly wider — 219–474 (1BKC), 218–474 (2DDF), 215–477 (multiple), and 212–493 (2I47/3G42) UNIPROT P78536. If 215–474 expresses poorly I would widen to 215–477 to match what has demonstrably folded.
Catalytic zinc site. The motif is HEXXH at His405–Glu406-X-X-His409, with His415 as the third zinc ligand; Glu406 is the catalytic residue (UniProt ACT_SITE 406) UNIPROT P78536. This is the classic adamalysin/ADAM zinc environment — the 2.0 Å structure showed a fold and zinc coordination resembling the snake-venom metalloproteinases PUBMED 9520379. One zinc per subunit UNIPROT P78536.
Cysteine switch (contested — see §6). UniProt annotates a cysteine-switch motif at 182–189 whose Cys184 binds the catalytic zinc in the inhibited zymogen UNIPROT P78536. Treat that as annotation, not settled mechanism.
2. Structure
Two cards were rendered this turn:
- "ADAM17/TACE catalytic domain — X-ray, 2.0 Å (PDB 1BKC)" — residues 219–474, co-crystallised with a hydroxamic-acid inhibitor and the catalytic Zn²⁺; the landmark structure that defined the ADAM fold, with large insertion loops generating unique surface features PUBMED 9520379. I chose this over purely higher-resolution entries because the active site is inhibitor-occupied and therefore interpretable. For a clean apo/high-resolution view, 2DDF (1.70 Å) and 3L0V (1.75 Å) in the same entry are better UNIPROT P78536.
- "Full-length human ADAM17/TACE — AlphaFold model (P78536)" — predicted, not experimental. Ectodomain folds as a compact module; the TM helix and cytoplasmic tail are low-confidence, consistent with UniProt's MobiDB disorder call on 732–824 UNIPROT P78536.
Better experimental options exist for the full molecule. UniProt cross-references cryo-EM entries 8SNN (2.32 Å, 215–824), 8SNO (2.78 Å, 215–824), 8SNL (2.78 Å, full-length 1–824), 8SNM (the zymogen: chain A 1–214, chain B 215–824), 9O54 (3.50 Å, 1–477) and 9O58 (3.53 Å, 28–824) UNIPROT P78536. The zymogen-plus-iRhom2 cryo-EM structure resolved how the prodomain and the therapeutic antibody MEDI3622 inhibit the enzyme PUBMED 40512800. If your client cares about the autoinhibited state, 8SNM is the structure to look at.
What the structures imply for folding difficulty
- Disulfides — this is the dominant risk. The catalytic domain carries 3 disulfides, all experimentally determined: C225–C333, C365–C469, C423–C453 UNIPROT P78536 PUBMED 9520379. The ectodomain adds 4 more that are annotated only "by similarity", not experimentally determined: C534–C555, C573–C582, C578–C591, C593–C600 UNIPROT P78536. And here is the number that should drive construct choice: residues 475–671 contain 26 cysteines, but UniProt annotates only 8 of them (4 disulfides) in that span. 18 cysteines in the disintegrin/CRD region have no annotated partner. The 2.32 Å cryo-EM reconstruction exists, but the deposited annotation does not give you a bond-by-bond map of that region. For a contract job, that is the single most important caveat: anything past ~477 is a disulfide-connectivity gamble, not a characterised module.
- Zinc. One Zn²⁺ per subunit UNIPROT P78536; a zinc-free or zinc-misloaded apoenzyme is inactive and will fail the client's activity assay while passing SDS-PAGE. Zinc must be in the buffers and verified by ICP-MS.
- Glycosylation. Nine N-glycosylation sites are annotated on the full-length protein: N103, N157, N174, N264, N452, N498, N539, N551, N594 UNIPROT P78536 — three in the prodomain, two inside the catalytic domain (N264, N452), four in the disintegrin/CRD. GlyGen (via UniProt) reports 9 N-linked sites with 57 glycans; GlyConnect reports 18 N-linked glycans across 5 sites UNIPROT P78536. The catalytic-domain glycans sit on the surface, not at the active site, which is why the domain crystallised despite them — but occupancy changes enzyme behaviour (see §4) PUBMED 24361716.
3. Hard numbers (computed)
Computed with Biopython ProteinAnalysis + molecular_weight on the P78536 sequence; region-defining annotations from UniProt UNIPROT P78536. ε280 is the cystine (fully oxidised) value. Instability index is the Guruprasad heuristic.
| Construct | Residues | aa | MW (kDa) | pI | ε280 (M⁻¹cm⁻¹) | A280 (1 g/L) | GRAVY | Instab. | Cys | Annot. SS | N-glyc sequons | Provenance |
|---|
| Prodomain alone | 18–214 | 197 | 22.92 | 6.44 | 19,940 | 0.87 | −0.683 | 45.4 | 1 | 0 | 3 (N103/157/174) | computed UNIPROT P78536 |
| Cat domain alone | 215–474 | 260 | 29.34 | 5.66 | 35,215 | 1.20 | −0.665 | 42.2 | 6 | 3 | 2 (N264, N452) | computed UNIPROT P78536 |
| Pro + cat zymogen | 18–474 | 457 | 52.24 | 5.97 | 55,155 | 1.056 | −0.673 | 44.9 | 7 | 3 | 5 | computed UNIPROT P78536 |
| Cat + disintegrin | 215–563 | 349 | 38.85 | 5.26 | 38,945 | 1.002 | −0.703 | 43.5 | 19 | 4 | 5 | computed UNIPROT P78536 |
| Ectodomain | 215–671 | 457 | 51.13 | 5.18 | 46,810 | 0.916 | −0.638 | 41.7 | 32 | 7 | 6 | computed UNIPROT P78536 |
| Pro + ectodomain | 18–671 | 654 | 74.05* | — | 66,750* | 0.901* | — | — | 33* | 7 | 9* | summed from fragments above |
| Full-length (reference) | 1–824 | 824 | 92.99 | 5.50 | 73,865 | 0.794 | −0.574 | 45.8 | 34† | 7 | 10 | computed UNIPROT P78536 |
* 18–671 and its ε280/A280/Cys/sequon counts are exact sums of two contiguous, non-overlapping fragments (18–214 plus 215–671); MW and extinction coefficient are additive, pI is not and is therefore not stated. † Full-length Cys shows 34 because the erroneous 825-residue input truncated at 824; UniProt-verified total is 35 cysteine — the C-terminal Cys is C824.
What these numbers actually tell you
- GRAVY is negative for every construct (−0.57 to −0.70). These are hydrophilic proteins. The aggregation risk here is not hydrophobic collapse — it is disulfide mis-pairing. That is a different failure mode with a different fix (redox control, not detergent/solubility tags).
- Instability index flags "unstable" for all constructs (>40). Treat this as weak evidence — the heuristic is built on short-lived cytoplasmic proteins and penalises exactly the Cys-rich, charge-rich composition found here. It should not drive the go/no-go. Its one useful signal is that the smallest construct (cat alone, 42.2) scores marginally better than the full-length (45.8).
- pI 5.2–6.4 — anionic at pH 7.4; expect to bind anion exchange poorly and to be well-behaved on cation exchange below pH 6.5.
- A280 of 1.20 for the catalytic domain is the number your QC group needs for concentration; using a generic 1.0 mg/mL-per-A280 assumption would under-report by 20%.
- Cysteine load is the construct-selection lever. Cat alone = 6 Cys / 3 disulfides, all experimentally determined. Every construct beyond 477 multiplies the problem: 19 Cys (cat+disintegrin) → 32 Cys (ectodomain), with 18 of those CRD-region cysteines having no annotated partner.
4. Expression-system call
| Axis | Cell-free (E. coli extract) | E. coli refolding | CHO |
|---|
| Folding fidelity | Moderate; open system, redox tunable, no ER chaperones | Low–moderate; must be validated, not assumed | Highest; full ER machinery, PDI, prodomain-assisted |
| Disulfide handling | Tunable — no reducing agent, GSH/GSSG, PDI/DsbC supplementation; 0.3–0.7 mg/mL purified correctly disulfide-bonded eukaryotic protein demonstrated PUBMED 22776321; dedicated disulfide-competent CFPS systems exist pubmed:23291178 | Redox refold from IB; workable for 3 disulfides, increasingly hostile at 19–32 | Native, correct |
| Glycosylation | None in prokaryotic extract; limited in plant-based ALiCE PUBMED 40885962 | None | Yes (CHO-type; note CHO lacks α2,6-sialylation) |
| Yield | Highest speed-to-material; ~7 mg active protein from a 10 mL fed-batch demonstrated for a well-behaved enzyme PUBMED 14763830; disulfide-containing targets lower PUBMED 22776321 | IB expression usually high, refold recovery is the bottleneck PUBMED 32729411 | Typically single-digit mg/L transient |
| Timeline | Days to ~3–4 weeks | ~6–10 weeks incl. refold screening | ~6–8 weeks transient |
| Specific failure mode for THIS protein | Multi-Cys CRD aggregates or scrambles; apoenzyme (zinc not loaded) PUBMED 17146616; aglycosyl product is a different molecule from native PUBMED 24361716 | Scrambled disulfides → soluble-but-dead enzyme; loses the prodomain chaperone if you use cat-alone; validated precedent for an ADAM pro+cat domain in E. coli is ADAM8, purified active via autocatalytic prodomain removal PUBMED 16542157 | ER retention/low secretion if the prodomain is absent or the construct mis-traffics PUBMED 40512800; autolysis of a shed ectodomain; clonal variability |
Primary route — CHO, secreted, construct 18–474, C-terminal His8 or Twin-Strep. Reasons, in order of weight: (i) it is the only route that gets you the glycosylated molecule, and glycosylation of ADAM17 changes both activity and inhibitor potency PUBMED 24361716; (ii) the prodomain is retained in cis, which is what the prodomain chaperone data require PUBMED 15581423; (iii) correct disulfides with no refold gamble; (iv) secreting the zymogen mitigates autolysis — you harvest an inactive proenzyme and activate on demand.
Hedge — E. coli cell-free, same 18–474 construct. Gets you active, aglycosyl material in days rather than weeks, which is the right instrument for early assay development and for a client who needs a reagent now. Precedents: cell-free synthesis of zinc-binding proteins PUBMED 17146616, cell-free systems engineered specifically for disulfide-containing proteins PUBMED 23291178 PUBMED 22776321, and a comparison study of E. coli cell-based vs. wheat-germ cell-free for exactly this "which platform for a hard eukaryotic target" question PUBMED 25854603.
E. coli inclusion-body refolding — I would not lead with it. With 6 cysteines it is defensible; with 19–32 it is a project, not a task. Keep it as a third option only if the client accepts a long refold-development phase.
Planning estimates — these are my professional estimates, not literature values, and they should be read as such
| Route | Weeks to purified mg-scale | Expected mg | Assumptions |
|---|
| CHO transient | 6–8 | 1–5 mg from 1–3 L | gene synthesis 2 wk; transfection + pool 2–3 wk; expression 1 wk; 2-step purification 1 wk |
| E. coli cell-free | 3–4 | 0.2–1 mg from a 5–10 mL reaction | anchored on the published 0.3–0.7 mg/mL purified benchmark for disulfide-containing targets PUBMED 22776321 |
| E. coli refold | 6–10 | 1–5 mg if refold succeeds | includes ≥2 rounds of refold-condition screening; meaningful probability of never reaching active monomer at 19+ Cys |
5. Codon and construct engineering
| Change | Rationale |
|---|
| C-terminal His8 or Twin-Strep; no N-terminal tag on the mature enzyme | The mature N-terminus (215–) sits at the active-site cleft; an N-terminal tag there interferes with the cysteine-switch region and with furin processing. For the zymogen, a tag on the prodomain N-terminus is tolerable but offers no advantage. |
| Keep the native 214/215 furin junction in the zymogen construct | Lets you activate in vitro with furin rather than committing to an active enzyme during expression. |
| Be aware of the second, upstream PC site inside the prodomain | Maturation of ADAM9/10/17 requires processing at a newly identified upstream proprotein-convertase site that is prerequisite to the canonical RXXR site — so the canonical site alone is not the whole activation story PUBMED 25795784. Design your activation readout accordingly. |
| Use the native signal peptide (1–17) for CHO; a heterologous high-secretion SP is the fallback | Native SP preserves trafficking; if secretion is poor, swapping the SP is the cheapest single variable to change. |
| No signal peptide for cell-free | Cell-free has no secretory pathway; a SP adds nothing and can hurt. |
| Cysteine handling: oxidising extract, no reducing agent, GSH/GSSG, PDI or DsbC supplementation | The demonstrated route to native disulfides in E. coli CFPS PUBMED 22776321 PUBMED 23291178. For cell-based E. coli, use SHuffle (cytoplasmic DsbC) or periplasmic expression — never standard BL21 cytoplasm. |
| Add 50–100 µM ZnCl₂ to all buffers; EDTA-free throughout | Zinc metalloprotease; a zinc-free prep passes SDS-PAGE and fails activity UNIPROT P78536. |
| Low-temperature expression (16–18 °C) for cell-free and E. coli | Slows folding relative to translation, reduces aggregation; standard for disulfide-rich targets. |
| E406A catalytic-dead variant as a parallel negative control | Gives you a matched "active-site-dead" reference for the client's assay and prevents autolysis during production. |
| Cap the construct at 477 unless the client's assay needs exosites | The noncatalytic domains of ADAM17 do affect substrate binding (largely by steric hindrance) PUBMED 23779109, and the CANDIS juxtamembrane segment is essential for shedding activity PUBMED 26348730 — but those matter for cell-surface substrate shedding, not for a peptide-substrate assay. Do not take on 18 uncharacterised cysteines for a capability the assay does not test. |
| Do not build the ectodomain construct for a peptide assay | 26 Cys in 475–671 with only 4 annotated disulfides = an unvalidated folding problem UNIPROT P78536. |
| N264Q / N452Q only as crystallography/control reagents, never as the delivered material | Glycosylation of ADAM17 affects its activity and inhibition PUBMED 24361716; deglycosylating the deliverable changes the client's biology. |
6. Evidence — prodomain requirement and production precedents
Prodomain is required for functional secretion (the load-bearing claim)
- The prodomain of TACE is essential for secretion of the functional enzyme; a TACE truncate was not secreted without it and was degraded intracellularly, and the prodomain could rescue secretion in trans — an intramolecular-chaperone function. Full-length TACE was also degraded when expressed without the prodomain PUBMED 15581423 (abstract; open access, full text available).
- The cysteine-switch mutants (C184A) resembled wild-type functionally, and zymogens with C184A were more susceptible to intracellular degradation — i.e. the switch is not the simple on/off that the model predicts PUBMED 15581423.
- Key sequence determinants of prodomain inhibition map to two short regions where TACE and ADAM10 diverge from other ADAMs; ADAM10's prodomain inhibits TACE as potently as TACE's own PUBMED 18707126 (abstract only).
- The amino terminus of the TACE prodomain contributes to keeping TACE inactive, independently of the cysteine switch PUBMED 15749738 (abstract only).
- A dedicated re-examination of the "cysteine switch" model in the TACE zymogen exists PUBMED 16679521 (metadata only — no abstract snippet returned).
- Cryo-EM of the ADAM17 zymogen bound to iRhom2 resolved the inhibitory mechanism of the prodomain (and of MEDI3622) PUBMED 40512800 (abstract only).
Where the literature disagrees — flag this to the client
- Is the TACE prodomain a general autoinhibitor? Potent inhibition is reported by the Milla lab PUBMED 18707126, but recombinant mouse TACE prodomain did not inhibit its own enzyme, while ADAM10's prodomain was a potent competitive inhibitor (Kᵢ = 48 nM) and ADAM10-selective PUBMED 17895248 (abstract only). That is an unresolved contradiction in what I retrieved.
- Is the cysteine switch the mechanism? UniProt annotates it as the mechanism UNIPROT P78536; the primary data say the switch is not required PUBMED 15581423 PUBMED 15749738. Practical consequence: do not engineer around C184.
- Which protease matures ADAM17? Furin is the annotated enzyme UNIPROT P78536, but furin activity alone was not sufficient for effective TACE maturation — PACE-4, PC5/PC6, PC1 and PC2 also cleave it PUBMED 14623079 (abstract only), and a prerequisite upstream PC site within the prodomain was identified later PUBMED 25795784 (abstract only).
Successful recombinant productions
- Yeast — success. Correctly processed, active human TACE catalytic domain expressed and purified from Saccharomyces cerevisiae PUBMED 9631522 (metadata only — no abstract snippet was returned; I am citing the title and journal).
- E. coli — success, but in a different ADAM. A soluble ADAM8 pro- plus metalloprotease-domain construct was expressed in E. coli and purified as active protease owing to autocatalytic prodomain removal PUBMED 16542157 (abstract only). This is the closest published analogue to the primary route I'm recommending and is the strongest single piece of evidence that route (a) can work in a bacterial host.
- Crystallography — success. The ADAM17 catalytic domain was produced for the 2.0 Å structure PUBMED 9520379, and ~25 subsequent TACE catalytic-domain crystal structures are deposited UNIPROT P78536.
Cell-free zinc metalloproteases / zinc-binding proteins
- Cell-free synthesis of zinc-binding proteins PUBMED 17146616 (metadata only).
- Cell-free systems engineered for disulfide-containing proteins PUBMED 23291178 (metadata only), PUBMED 22776321, PUBMED 17626291 (abstract only).
Literature gaps I could not close. I did not retrieve a primary production-methods paper for the ADAM17 ectodomain from CHO or baculovirus, nor a dedicated E. coli refolding paper for the TACE catalytic domain specifically. My searches on those two topics returned generic disulfide-refolding and protease-expression hits rather than ADAM17 production papers, and I am not going to dress those analogues up as ADAM17 precedent. The E. coli and mammalian-cell claims above rest on the ADAM8 analogue PUBMED 16542157, the yeast precedent PUBMED 9631522, and the general disulfide-protein CFPS benchmarks PUBMED 22776321.
One further precedent worth naming with its evidence depth flagged: the iRhom2 requirement for TACE ER exit is described in a Faculty Opinions commentary DOI 10.3410/f.14262973.15775084 — that is a secondary source; I did not retrieve the primary papers this turn. Its content matters for CHO: if ADAM17 trafficking depends on iRhom partners, an over-expressed secreted construct may need the prodomain (and possibly membrane-proximal context) to leave the ER.
7. Assay readiness and QC
Standard activity assay: a continuous, quenched fluorogenic (FRET) peptide assay. A short peptide spanning the substrate cleavage site, flanked by a fluorophore/quencher pair (Mca/Dpa or ACC), cleaved to release fluorescence. The specificity of TACE versus ADAM10 was mapped with combinatorial peptide libraries, which is where the TACE-preferred substrate sequences come from PUBMED 19715556 (abstract only). Practical protocols: FRET peptides for metalloproteinases PUBMED 20135296, and the standard Mca-peptide protocol with a worked ADAM/MMP example PUBMED 18429258. For a biological readout, an ADAM17 catalytic-domain construct has been used to cleave a physiological substrate fragment PUBMED 23416080 (abstract only), and multiplexed ADAM10/17 activity measurement in lysates is established PUBMED 29187866. ACC as a leaving group gives ~3-fold higher quantum yield than AMC, useful if you are measuring low-activity lots PUBMED 10869434 (abstract only).
QC package I would ship with each lot
- Purity — reducing and non-reducing SDS-PAGE ≥95%, plus SEC-HPLC monomer ≥95% (single peak, correct elution volume).
- Identity — intact mass by LC-MS against the computed MW from §3 (52.24 kDa for the 18–474 zymogen, 29.34 kDa for cat alone — these differ enough that mass alone tells you which construct and whether the prodomain has been clipped).
- Concentration — A280 using the computed ε: 55,155 M⁻¹cm⁻¹ (A280 1.056) for 18–474, or 35,215 M⁻¹cm⁻¹ (A280 1.20) for cat alone; state the value used on the CoA.
- Zinc content — ICP-MS, stoichiometry to protein (expect ~1 Zn/subunit) UNIPROT P78536.
- Free thiol — Ellman or maleimide labelling; near-zero free thiol confirms complete disulfide formation.
- Disulfide connectivity — peptide mapping by LC-MS/MS with non-reduced digest; report which disulfides are confirmed. Be explicit that the 3 catalytic-domain pairs are the experimentally established ones PUBMED 9520379 and that pairs beyond 477 are inferred UNIPROT P78536.
- Specific activity — initial-rate kcat/Km against the FRET substrate; report with the assay buffer, pH, Zn²⁺ and temperature. Include an inhibitor check (a hydroxamic acid such as batimastat, or MEDI3622 PUBMED 40512800) to prove the signal is active-site-dependent.
- Specificity/selectivity — report activity against an ADAM10-preferred substrate so the client can see the TACE/ADAM10 discrimination of their lot PUBMED 19715556.
- Active-site occupancy — because a zinc-free or partially occupied prep passes 1–3 and fails 4/7.
- Endotoxin (<0.1 EU/µg) and sterility if the material goes anywhere near cells.
- Glycosylation profile (CHO lots only) — intact mass, PNGase-F shift, and site occupancy at N264/N452 UNIPROT P78536; flag CHO versus human sialylation differences.
The chart was not produced this turn — I'm not going to describe a chart I didn't render. When you want it, the honest version is a chromatographic + purity release-criteria panel: SEC-HPLC trace with the monomer/aggregate/fragment integration and acceptance cut-offs, plus a representative progress-curve of FRET fluorescence against time for released lot versus the E406A dead-enzyme control. Both of those are drawn from real QC data, not invented points.
Assumptions I made, stated explicitly
- Isoform A (canonical, P78536-1, 824 aa), human sequence, no tags specified by the client.
- "Ectodomain plus its prodomain" = the client wants the prodomain retained in the molecule; I interpreted this as 18–474 for an assay reagent and 18–671 only if native substrate context is required.
- Residue-numbering throughout is full-length P78536 numbering (signal peptide included), not "mature after signal peptide." Crystallographic papers sometimes use other conventions — confirm before ordering the gene.
- Buffer conditions (pH, Zn²⁺, redox) for the yield estimates assume standard practice for secreted metalloproteases; no ADAM17-specific optima are asserted.
- The week-counts and mg-scale figures in §4 are my planning estimates derived from project structure, anchored where possible to published CFPS benchmarks PUBMED 22776321 — they are not measured yields for this protein, and no one has published a yield for this exact construct that I could retrieve.
What's outstanding
Three things were cut off by this turn's processing limit: (1) the release/activity chart in §7; (2) re-running the §3 property table on the now-verified 824-aa sequence to remove the ~0.1 kDa ambiguity; and (3) a targeted literature push for the ADAM17-specific E. coli-refolding and CHO/baculovirus production papers, which my searches did not surface.
Ask me to continue and I'll pick up from the verified sequence, render the QC chart, and hunt those two production papers specifically.