bchE (P26168, Rhodobacter capsulatus SB1003) — curation notes

Provenance note. No deep-research API keys exist in this container, so
just deep-research was not run and no -deep-research-<provider>.md file
exists for this gene. All literature below was read directly: the cached
publications/PMID_*.md files, PubMed E-utilities, and WebSearch. GONUTS
(gowiki.tamu.edu), the CACAO source wiki, was re-checked on 2026-11-01 and is
now permanently retired: https://gowiki.tamu.edu/wiki/index.php/Main_Page
returns HTTP 410 and the server serves a notice titled "GONUTS has been
retired" ("The GO wiki that was at gowiki.tamu.edu is no longer online."). The
archived copy at web.archive.org is blocked by this container's egress policy.
Search-engine snippets of the archived RHOCB:BCHE page were visible and are
noted below as third-hand context only, never cited as evidence.

What the gene is

P26168 / RCAP_rcc00669 is the anaerobic (oxygen-independent)
Mg-protoporphyrin IX monomethyl ester (MPE) cyclase, BchE: the enzyme that
installs the C13^1 oxo group and closes the isocyclic fifth (E) ring, taking
MPE to 3,8-divinyl protochlorophyllide a in bacteriochlorophyll a
biosynthesis.

The chemistry is a cobalamin-dependent radical-SAM six-electron oxidation that
takes its oxygen from water, EC 1.21.98.3 / RHEA:49096
[file:RHOCB/bchE/bchE-uniprot.txt "Catalyzes the anaerobic formation of the
isocyclic ring (E-ring) in Mg-protoporphyrin monomethyl ester (MPE) to yield
protochlorophyllide a (PChlide a) via a six-electron oxidation and the
formation of an oxo group at position C13 using oxygen from a water molecule"].

This is the step where misassignment has historically happened. The same
overall conversion is performed in oxygenic and aerobic organisms by the
structurally and mechanistically unrelated di-iron oxygenase AcsF/CHL27/XanL,
EC 1.14.13.81, which consumes O2 and NADPH. GO has a term for the aerobic
reaction only — GO:0048529 "magnesium-protoporphyrin IX monomethyl ester
(oxidative) cyclase activity", whose definition is "Catalysis of the reaction:
magnesium protoporphyrin IX 13-monomethyl ester + 3 NADPH + 3 H+ + 3 O2 =
divinylprotochlorophyllide + 3 NADP+ + 5 H2O" with xrefs EC:1.14.13.81 and
RHEA:33235, and which is_a GO:0016709 (incorporation of molecular oxygen).
There is no GO term for EC 1.21.98.3 / RHEA:49096: QuickGO ontology search
for "1.21.98.3", "RHEA:49096" and "anaerobic magnesium-protoporphyrin" returns
nothing matching. Nonetheless GO:0048529 is propagated by IEA to bchE
orthologs (e.g. UniProtKB:A0A087S6L6 bchE, GO:0048529 IEA GO_REF:0000003),
which is the misassignment the project flagged. P26168 itself does not carry
GO:0048529, so this review has nothing to retract on that score; instead the
gap shows up as a bare GO:0003824 catalytic activity row, and this review
proposes the missing term.

R. capsulatus has no AcsF at all, which is why BchE is the sole route to
PChlide in this organism PMID:42748141.

Evidence that bchE is the MPE cyclase in R. capsulatus

• Yang & Bauer 1990 disrupted three PGC reading frames; F1025 (later named
  bchE) blocked the MPE → PChlide step PMID:2203738. The abstract names
  the ORF, not the gene symbol; the identification of that lesion as bchE is
  stated independently by Gough et al. PMID:10841582.
• Bollivar et al. 1994 (the CACAO reference) is a cluster-wide directed
  mutational analysis: "we have undertaken a systematic directed mutational
  analysis of 12 open reading frames in the photosynthesis gene cluster to
  evaluate whether individual open reading frames have a role in photopigment
  biosynthesis", with "mutations constructed in seven open reading frames
  resulted in a loss of bacteriochlorophyll biosynthesis, concomitant with the
  accumulation of specific intermediates in the Mg-tetrapyrrole biosynthetic
  pathway" PMID:8158642. Cached abstract only (full_text_available: false).
• Gough et al. 2000 showed the cyclase reaction itself has an absolute
  cobalamin requirement, using B12-auxotroph bluE/bluB mutants and an in
  vivo assay PMID:10841582. Note
  this paper worked in R. capsulatus 37b4 ("Cell growth of capsule-less
  wild-type R. capsulatus 37b4"), not SB1003, and it never touched purified
  BchE protein. Its [4Fe-4S] statement is explicitly a proposal: "This
  cysteine arrangement resembles somewhat an iron–sulfur center as found in
  biotin synthase ( 31 )". UniProt's cofactor and BINDING lines for the
  cluster are correspondingly ECO:0000305 (curator inference), while the
  adenosylcobalamin cofactor is ECO:0000269.
• Wiesselmann et al. 2020 is the strongest gene-specific work: an in vitro
  assay on R. capsulatus BchE validated by H2^18O label transfer onto the
  C13^1-oxo group, plus 24 site-directed mutants scored by in vivo
  complementation, which identified "Potential ligands of the [4Fe-4S] cluster
  (Cys204, Cys208, Cys211), of SAM (Phe210, Glu308 and Lys320) and of the
  proposed cobalamin cofactor (Asp248, Glu249, Leu29, Thr71, Val97)"
  PMID:33211085. Those three cysteines are exactly UniProt's BINDING
  positions, and in the P26168 sequence they are the canonical radical-SAM
  CxxxCxxC motif (residues 195-215 read VAIPNMARGCPFTCSFCSQWK, i.e.
  C204-x3-C208-x2-C211). Abstract-only in the cache.
• York, Zhang & Booker 2026 achieved the first purification and in vitro
  reconstitution, confirming both cofactors spectroscopically: "The enzyme was
  purified anaerobically and characterized by UV–Vis and electron paramagnetic
  resonance spectroscopy to confirm the presence of Cbl and a [Fe4S4]
  cluster." This was not the R. capsulatus protein: "The enzyme, BchE,
  from Rubrivivax pictus was heterologously expressed in either Escherichia
  coli BL21(DE3) or a modified ΔbtuR knockout strain of BL21(DE3)", after "we
  attempted heterologous expression and purification of BchE homologs from
  eight different organisms; only three yielded any soluble protein, and of
  these, only BchE from Rubrivivax pictum catalyzed the formation of PChlide"
  PMID:42748141. So for P26168 the cofactor identities are ortholog evidence
  plus the gene's own conserved motif and mutagenesis, not direct
  spectroscopy. The same paper does contain R. capsulatus whole-cell work
  replicating Gough.

Localization

The TreeGrafter GO:0005829 cytosol row is the weakest annotation on this
entry. P26168 has no transmembrane segment and no signal peptide in UniProt,
so a cytosol default is the expected machine output — but direct subcellular
fractionation of R. capsulatus says otherwise: "Subcellular fractionation
experiments revealed that membrane-localized BchE requires an additional,
heat-sensitive cytosolic component for activity" PMID:33211085. The protein
is also notoriously insoluble when expressed recombinantly PMID:42748141. The honest reading
is that BchE is membrane-associated in its native host and partners with a
small (3-10 kDa) soluble component; asserting cytosol as the location
overstates what is known, and asserting a membrane term would overstate it in
the other direction from an abstract alone. Flagged as a knowledge gap and a
suggested experiment.

CACAO row and the cross-isolate hypothesis (screen 2)

The predicted failure shape was: PMID:8158642 is cited by CACAO rows under two
different NCBI taxon ids that both mean R. capsulatus, suggesting a result
from one isolate was annotated onto a protein filed under a sibling taxon id
without re-grounding.

QuickGO (reference=PMID:8158642, 2026-11-01) returns exactly four
annotations, all CACAO, all GO:0030494 IMP involved_in:

The two-taxon-id signature is real but it is not on bchE. The 1061 row is
bchM, P26236, whose UniProt entry is the older species-level record
BCHM_RHOCA (OS Rhodobacter capsulatus (Rhodopseudomonas capsulata).,
OX NCBI_TaxID=1061;), dating from the 1984 Youvan sequencing paper. bchE,
bchF and bchJ are all on the modern strain-level SB1003 entries (272942). The
split therefore reflects the age and scope of the UniProt entries the students
annotated, not a transfer across isolates — and in any case a strain-level id
within one species would not by itself be a defect.

The neighbour-mis-attribution alternative is also absent. Bollivar et al. 1994
is a cluster-wide analysis of 12 PGC reading frames, so three adjacent genes
(bchF, bchE, bchJ) citing one reference is exactly what that paper supports;
it is not a sign that one gene's result was spread across its neighbours. The
gene's lesion is independently corroborated: bchE mutants accumulate MPE,
reported both by Yang & Bauer and restated by Gough. (Search-engine snippets
of the archived GONUTS RHOCB:BCHE page describe a table entry for a bchE
mutant strain with a 590 nm peak attributed to MPE accumulation — consistent
with the bchE transposon strain DB575 that Gough et al. used — but the page
itself is unreachable, so this is context, not evidence.)

Verdict: the predicted cross-isolate transfer shape is ABSENT for bchE.
What the screen actually detected is a taxon-id heterogeneity across one
reference's annotation set, driven by UniProt entry vintage. That is worth
adding to the taxonomy as a distinct, benign pattern, because it will keep
firing on any multi-gene pathway paper annotated across a mixture of
species-level and strain-level entries.

Term-level conclusions

• GO:0003824 catalytic activity (IEA, InterPro rSAM signatures) — true but
  root-level, and it is standing in for a term GO does not have. MODIFY to a
  proposed new term for EC 1.21.98.3.
• GO:0030494 bacteriochlorophyll biosynthetic process (IMP, CACAO) — the
  claim is right and the paper supports it; the entry already carries the more
  specific GO:0036070, of which GO:0030494 is an ancestor (QuickGO
  ancestors of GO:0036070 include GO:0030494). MODIFY to GO:0036070.
• GO:0031419 cobalamin binding and GO:0051536 iron-sulfur cluster binding
  (IEA, InterPro B12-binding and rSAM domains) — both independently supported
  beyond family membership: residue-level mutagenesis on this protein
  PMID:33211085, in vivo cofactor dependence in this species PMID:10841582
  and spectroscopy on a purified ortholog PMID:42748141. ACCEPT.
• GO:0046872 metal ion binding (IEA, InterPro cobalamin-binding domain) —
  this is the one chemistry claim that is inferred from family membership
  alone and is not entailed by the specific terms. QuickGO ancestors of
  GO:0051536 are GO:0051540 (metal cluster binding), GO:0036094, GO:0005488 —
  not GO:0046872; ancestors of GO:0031419 are GO:0046906 (tetrapyrrole
  binding), GO:0019842 (vitamin binding), GO:0036094, GO:0005488 — again not
  GO:0046872. GO deliberately files both cofactors away from "metal ion
  binding", because the cobalt is held inside the corrin ring and the irons
  inside a cluster; neither is a discrete metal ion site. No free metal ion
  site is reported for BchE. MARK_AS_OVER_ANNOTATED (not REMOVE: cobalt and
  iron are genuinely present, so the term is not flatly false).
• GO:0005829 cytosol (IEA, TreeGrafter PTN000228879) —
  MARK_AS_OVER_ANNOTATED, see Localization above.