Naked mole-rat Cd44 (A0AAX6R0R7): isoform architecture and hyaluronan-binding site

Question

Two things about the naked mole-rat (NMR) CD44 reference protein could not be settled
from the UniProt record alone:

1. The TrEMBL entry A0AAX6R0R7 is 701 aa and carries no ALTERNATIVE PRODUCTS or
   VAR_SEQ annotation, so the record itself says nothing about which CD44 splice form
   the RefSeq gene model XP_012930388.1 represents. CD44 is one of the most heavily
   alternatively spliced genes in the genome (human P16070 has 19 annotated isoforms),
   and the distinction between the short standard form (CD44s) and the long
   variant-exon-containing forms (CD44v) matters for several GO annotations.
2. The entry carries a CAUTION from the Link-domain ProRule:
   Lacks conserved residue(s) required for the propagation of feature annotation
   (PROSITE-ProRule PRU00323). Read carelessly this could be taken to mean the
   hyaluronan-binding module is degenerate, which would undercut
   GO:0005540 hyaluronic acid binding — the gene's core molecular function.

Method

cd44_isoform_architecture.py performs a global Needleman-Wunsch alignment (BLOSUM62,
gap open -11, extend -1, Biopython PairwiseAligner) of A0AAX6R0R7 against human
P16070-1 (742 aa, the full variant-exon-containing canonical form), then maps
UniProt-annotated human landmark positions through the alignment. Landmarks are read
from the two UniProt flat files, not assumed:

The variant-exon test is: if the NMR protein were a standard CD44s form, human
residues 223–535 (the segment removed in the short human isoform 11 / CD44R2) would
have no aligned counterpart. Sequences are fetched from the UniProt REST API and
cached beside the script; results are written to cd44_isoform_architecture.json.

Reproduce with:

uv run --with biopython python cd44_isoform_architecture.py

Results

human_length                              742
hetga_length                              701
percent_identity_over_aligned            77.3
human_link_domain_32_120_aligned        89/89
human_TM_650_670_aligned                21/21
human_variant_insert_223_535_aligned   270/313  (86.3%)
percent_identity_link_domain_32_120      92.1
percent_identity_variant_insert_223_535  74.1
percent_identity_cytoplasmic_tail        97.2
hetga_ectodomain_len_after_signal         588
human_ectodomain_len_after_signal         629

Residue-level correspondence:

Interpretation

1. The reference protein is a variant-exon-containing (CD44v-like) gene model, not
CD44s. 86% of the human alternatively spliced insert (223–535) has an aligned NMR
counterpart, and the NMR ectodomain is 588 residues against 629 for the full-length
human canonical form. A standard CD44s model would be several hundred residues shorter.
This is a statement about the RefSeq gene model that UniProt happens to have selected as
the reference protein for this gene; it is not evidence about which splice forms NMR
tissues actually express, which no cached study reports. The variant region is also the
least conserved part of the protein (74.1% identity), as expected for a mucin-like,
heavily O-glycosylated stem.

2. The hyaluronan-binding module is intact. All four UniProt-annotated
hyaluronan-contact residues of human CD44 and all six cysteines forming the three
Link-region disulfides have direct counterparts in the NMR sequence, and the Link domain
is 92.1% identical to human — the most conserved region of the ectodomain. The PRU00323
CAUTION is a feature-propagation flag (the rule declines to auto-transfer its
DISULFID features when its profile-position conditions are not met); it is not a
finding that the hyaluronan-binding site is degenerate, and the residues that matter are
present.

3. The cytoplasmic tail is the single most conserved region (97.2% identity). That
is the segment through which CD44 couples to ERM proteins and to NF2/merlin, the
interaction through which naked mole-rat hyaluronan signalling arrests proliferation
(PMID:23783513).

Limitations

This is a sequence-architecture analysis of one predicted gene model against one human
reference sequence. It establishes that the residues required for hyaluronan binding are
present; it does not measure binding, and it says nothing about affinity, about the
splice repertoire expressed in NMR tissue, or about post-translational modification
(glycosylation state gates CD44 hyaluronan binding and is not addressed here).