Crystal Structure of Firefly Luciferase Throws Light on a Superfamily of Adenylate-Forming Enzymes

• The first crystal structure of any firefly luciferase, and the first of any ANL-superfamily enzyme. Photinus pyralis luciferase solved at 2.0 Å resolution from frozen crystals grown by microbatch with PEG, deposited as PDB 1LCI. This is the apo structure, no luciferin, no luciferyl-AMP, no ATP bound, which turned out to be both a limitation (no substrate-binding mode visible) and a revelation (the bare architecture exposed a problem the authors then turned into a mechanistic proposal). Every subsequent structural and computational study of firefly luciferase, and of every other ANL-family enzyme, traces back to this paper as the structural reference frame.
• The canonical two-domain ANL fold was defined here. The 62 kDa protein folds into a large N-terminal domain (~residues 1 to 436) and a small C-terminal domain (~residues 440 to 550) separated by a wide solvent-accessible cleft. The N-terminal domain has a distinctive αβαβα five-layered architecture, a central β-barrel flanked by two β-sheets, all sandwiched by α-helices, that has since been seen in every other ANL-family structure (acyl-CoA synthetases, NRPS adenylation domains, 4-coumarate:CoA ligases, beetle luciferases). The C-terminal domain is structurally distinct and connected to the N-terminal domain by a single hinge-like loop. Naming this fold and recognizing it as a superfamily framework was the paper's lasting contribution to structural biology.
• The conserved-residue paradox forced the domain-closure mechanism. Sequence analysis across acyl-CoA synthetases, peptide synthetases, and 4CL had identified ~25 residues conserved across the superfamily. Conti, Franks, and Brick mapped these onto the structure and discovered they sat on the surfaces of both domains, on either side of the cleft, but at distances far too great to bind ATP and the carboxylate substrate simultaneously in the open conformation observed. Their proposal: the two domains must close around the substrates during catalysis, bringing the two halves of the active site together. This was a structural prediction made before any closed-form structure existed. Subsequent work (Nakatsu 2006 with luciferyl-adenylate analog; Gulick 2009 review of conformational dynamics across the family) confirmed the prediction in detail, domain rotation by ~140° is now established as the canonical ANL catalytic motion, and Branchini's mutagenesis work showing the partial reactions are catalyzed by different conformations of the C-terminal domain is the direct experimental follow-up to this proposal.
• Implications for the firefly luciferase mechanism that subsequent work then nailed down. The cleft architecture predicted that adenylation (substrate carboxylate + ATP → adenylate intermediate + PPi) and oxidation (adenylate + O₂ → oxyluciferin + AMP + CO₂ + photon) would be catalyzed by different conformations of the same enzyme, an unusual feature for an enzyme of this size, since most enzymes use a single closed conformation for one chemistry. This is now understood as the defining mechanistic feature of the entire ANL superfamily: a single active site that switches between two functional states by rotation of the C-terminal domain. For luciferase specifically, the implication is that the substrate-binding pocket and the oxidation environment are not the same physical site; they are the same residues seen from two different domain-orientation states. Engineering one without considering the other is the trap that has caught many subsequent mutagenesis efforts.
• What this structure does not show. No luciferin, no D-luciferyl-AMP, no Mg-ATP, no oxyluciferin. The substrate-binding residues had to be inferred from later structures, Nakatsu 2006 (Luciola cruciata with DLSA, the non-hydrolysable luciferyl-adenylate analog, PDB 2D1S/2D1R) and the Photinus pyralis DLSA structures from Branchini's group, to know which residues actually contact the luciferin. So while Conti 1996 is the foundational citation for the architecture and the domain-closure mechanism, it is not the right citation for substrate-binding-pocket residues; that role belongs to Nakatsu 2006 and the Branchini mutagenesis papers (3RIX, etc., which are the structures used for the ESM2 deep mutational scanning work in the project).

Bottom line for the project: This is the foundational citation for the structural framework that every computational and engineering decision in the project rests on, even when the project is using more recent structures (3RIX for DMS, AlphaFold predictions for variant analysis, ProteinMPNN for inverse folding). The two-domain architecture explains why mutations distant from the active site can have outsized catalytic effects, they alter domain-closure dynamics rather than direct substrate contacts, and explains why the active-site divergence reported in Watkins 2018 (Arachnocampa luciferase has rebuilt the luciferin-binding pocket but kept the ATP-binding pocket and the conserved adenylation-half lysine) maps cleanly onto the bipartite logic of this structure: adenylation chemistry is conserved across the superfamily because the N-terminal domain is conserved; substrate identity is rebuildable because the C-terminal domain and the cleft surface are evolutionarily plastic. Cite Conti 1996 alongside Nakatsu 2006 and Gulick 2009 wherever the structural rationale for any luciferase engineering decision is being made, the apo structure here, the closed substrate-bound structures from Nakatsu, and the conformational-dynamics review from Gulick are the canonical three-paper structural foundation. For TU1 specifically, this paper is what tells you that the SKL peroxisomal targeting tag at the C-terminus is unlikely to perturb the active site: the C-terminal domain's last residues sit on the surface of the small domain, distant from the cleft, and adding three residues there has not measurably affected luciferase activity in any of the published studies that used C-terminally tagged luciferase variants.