AP1S2 (sigma1B) sequence analysis

Reproduce with:

cd genes/human/AP1S2/AP1S2-bioinformatics
uv run --no-project --with biopython python sigma_cargo_site.py

All sequences are fetched live from the UniProt REST API at run time; the script asserts that
the accession returned is the accession requested, that the sequence length matches the length
UniProt declares, and that each comparator still carries the residue the cited paper reported at
the position it reported. Nothing is hardcoded. The captured run is in sigma_cargo_site.out
and the per-site table in sigma_cargo_site.tsv.

Question 1 — does sigma1B retain the dileucine-signal binding residues?

Mattera et al. 2011 (PMID:21097499) mapped the (D/E)XXXL(L/I) sorting-signal binding site onto
the sigma side of the AP-1 gamma/sigma1, AP-2 alpha/sigma2 and AP-3 delta/sigma3 hemicomplexes by
mutagenesis, and showed that the gamma1-sigma1B hemicomplex binds the Nef, tyrosinase and
LIMP-II signals with avidity comparable to gamma1-sigma1A. The paper reports the sigma-side
residue numbers for sigma1A, sigma2 and sigma3A but not for sigma1B. This analysis maps them.

Global pairwise alignment (Biopython PairwiseAligner, BLOSUM62, gap open -11, extend -1):

Conclusion. All five sigma1A positions PMID:21097499 names in its text are present in
AP1S2, at R14, A62, V87, L100 and I102 in
AP1S2's own numbering (UniProt sequence version 1). That includes all three whose substitution
abolished the gamma1-sigma1A interaction outright — the paper reports it "was abolished only
by V88D and I103S (for Nef) and also by A63D (for tyrosinase)" — which map to V87, I102 and
A62. A62 is the one tied specifically to the tyrosinase signal, which is also the signal where
the gamma2-sigma1B hemicomplex's fine specificity differs from gamma1-sigma1B.

The two positions where AP1S2 differs from sigma2 are AP-1 subfamily states, not sigma1B
losses. The paper states it tested the sigma1A counterparts of the sigma2 residues wholesale
("The sigma2 residues that participate in the interaction with (D/E)XXXL(L/I) signals are
conserved on sigma1A and sigma3A ... We therefore tested the effect of mutating these sigma1A
and sigma3A residues") without naming every outcome in the text, so the sigma2 set is scanned
here too. AP1S2 differs from sigma2 at exactly two of those positions, and the script computes
the sigma1A counterpart of each rather than assuming it:

At the first of those two, sigma2's Asn is the outlier of the three complexes rather than AP-1's
Asp: sigma3A D98 maps to the same AP1S2 position 91 and is also Asp. The scan therefore covers
every position PMID:21097499 names for any of the three sigma subunits — sigma1A R15/A63/V88/
L101/I103, sigma2 R15/A63/V88/N92/E100/L101/L103, sigma3A R15/V94/D98/L107/L109 — and finds no
sigma1B-specific difference at any of them.

So neither difference is sigma1B-specific. The N92 position is in any case a tolerant one in
sigma2 itself: "the N92A and L101A mutations had no effect on the interaction with the Nef
signal but decreased the interaction with the tyrosinase signal". And sigma1A — which was shown
experimentally to bind all three test signals — carries the AP-1 state at both positions.

There is therefore no residue-level basis for arguing that sigma1B has lost the cargo-signal
binding site; the sequence evidence agrees with the functional result in PMID:21097499.

Caveat on R14 specifically: in AP-2 and AP-3 the sigma-side Arg15 is load-bearing, but in
AP-1 it is not. PMID:21097499 reports that "interaction with gamma1-sigma1A depends mainly on
gamma1 Arg 15", and lists sigma1A Arg15 among the residues "which can be substituted with
relatively little impact on the ability of gamma1-sigma1A to recognize (D/E)XXXL(L/I) signals".
Retention of R14 in sigma1B is therefore corroborative rather than decisive; A62, V87 and I102
are the positions that carry the argument, since their sigma1A counterparts (A63D, V88D, I103S)
are the three substitutions that abolish binding. L100 is in the same corroborative class as
R14: PMID:21097499 names sigma1A Leu101 alongside Arg15 as substitutable "with relatively little
impact".

Caveat: this establishes that the site is intact, not that the pocket has the same fine
specificity. PMID:21097499 itself shows fine specificity differs between hemicomplexes
(gamma2-sigma1B binds the tyrosinase signal but not the LIMP-II signal), and that difference is
not read off these five positions.

Question 2 — pairwise identity within the family

Global alignment identity to AP1S2 (P56377, 157 aa), over aligned columns:

The 87.3% sigma1B/sigma1A identity reproduces the "87% amino acid identity" reported when
sigma1B was first described (PMID:9733768), which is an independent check that the right pair of
accessions is being compared.

Question 3 — how far can mouse sigma1B phenotypes be transferred to human?

Human AP1S2 and mouse Ap1s2 align with zero mismatched columns over all 157 human residues;
the mouse protein differs only by a 3-residue insertion at mouse positions 143-145. The mouse
sigma1B knockout phenotypes (PMID:20203623, PMID:25128028, PMID:24928897, PMID:27411398) are
therefore about a protein that is identical in sequence to the human one at every position the
two share. This is the strongest possible sequence basis for an ISS/ISO transfer, and it is why
the mouse work is treated here as sequence-similarity evidence for human AP1S2 rather than being
set aside as "mouse only".

Question 4 — the sigma1 C-terminal extension

Against sigma2 (AP2S1, 142 aa), AP1S2 has 16 positions with no aligned sigma2 residue, of which
13 are contiguous at the C terminus (AP1S2 142-154, with 133-135 a separate short insertion).
This is the "C-terminal extension" that PMID:20203623 notes distinguishes the sigma1 isoforms
from sigma2, and it is the region in which the three sigma1 isoforms differ most from one
another. It is not covered by the dileucine-binding site analysed above, so this analysis says
nothing about what that extension does.