Cloning and Characterization of human Vps26, Vps29, and Vps35
Recently, several yeast proteins (Vps5p, Vps17p, Vps26p, Vps29p,
and Vps35p) have been shown to form a complex involved in retrograde
trafficking between the prevacuolar endosome and the TGN (
Horazdovsky
et al., 1997 
;
Seaman et al., 1997 ,
1998 ). The
retromer complex, as it has been termed, may act as a coat for
retrograde vesicles leaving the prevacuolar compartment. We have cloned
full-length cDNAs of the human homologs of Vps26p, Vps29p, and Vps35p
(Figure ). Human Vps26, Vps29, and Vps35
are hydrophilic proteins containing 327, 182, and 796 amino acids,
respectively. There are no recognizable protein motifs or domains in
the molecules. Furthermore, all three molecules are predicted to be
soluble proteins. Homology searches have identified sequences from
multiple species that exhibit a high degree of amino acid identity with
the human sequences reported here (Figure
). hVps26 shows a high degree of
end-to-end homology with proteins from several species, including
C. elegans,
Drosophila melanogaster,
Arabidopsis thaliana, and
S. cerevisiae. Vps35 is
also well conserved throughout evolution. Based on the amino acid
identity among the various Vps35 orthologs, the protein contains at
least three conserved regions (
Nothwehr et al., 1999 ).
Interestingly, in addition to having orthologs in
C.
elegans, and
S. cerevisiae, Vps29 is also found in one
thermophilic bacterium (
Thermatoga maritima) and several
species of Archae (
Methanococcus jannaschii,
Methanobacterium thermoautotrophicum,
Pyrococcus
abyssi, and
Pyrococcus horikoshii). Northern blot
analyses show that hVps26, hVps29, and hVps35 mRNAs are ubiquitously
expressed, with hVps26 and hVps35 mRNAs being most highly expressed in
heart, skeletal muscle, kidney, liver, and placenta. In addition to
these tissues, hVps29 mRNA is also highly expressed in spleen, thymus,
and peripheral leukocytes (Figure ).
Western blot analyses of extracts from various rat organs showed that
Vps35 and Vps26 protein are expressed at similar levels in all tissues
examined (Figure A), whereas
endogenous rat Vps29 is not readily detected in tissue extracts with
the reagents currently available (our unpublished results). Rat Vps35
and Vps26 are expressed at low levels in heart and skeletal muscle,
moderate levels in brain and lung, and the highest levels in liver,
spleen, kidney, and testes. Based on the deduced amino acid sequences
of Vps35, Vps29, and Vps26, the molecules are predicted to be soluble
proteins. However, analyses of cytosolic and sedimentable fractions
from rat liver (Figure B) and COS7 cells (our unpublished results)
showed that the proteins were detected in both cytosolic and
sedimentable fractions as judged by immunoblotting with
polyclonal antibodies generated against GST fusion proteins of each
molecule. In addition, sedimentation of a rat liver microsomal fraction
through a linear sucrose gradient showed that both Vps35 and Vps26
entered the gradient and were enriched in the early fractions
(F6–F12). These fractions also contain EEA1, an early endosomal
marker, and GM130, a Golgi marker. Small amounts of Vps35 and Vps26
were also detected in regions of the gradient containing EGF receptor,
a plasma membrane marker, and calnexin, an endoplasmic reticulum marker
(F14–F18). No detectable Vps35 and Vps26 were found in the heavier
fractions containing LAMPI, a late endosomal and lysosomal marker.
These data suggest that the mammalian retromer proteins exist in both a
membrane-associated and cytosolic pools. However, because large protein
aggregates may also sediment partially into a sucrose gradient loaded
from the top, we performed a floatation analysis. Liver homogenate was
prepared and made dense. The homogenate was then overlaid with a linear
sucrose gradient and centrifuged. Western blot analyses of the material
that pelleted (P), the material remaining in the load following
centrifugation (L), and the material floating into the gradient
(fractions 1–18) showed that 22–30% of the loaded Vps35 and Vps26
floated into the gradient and thus was membrane associated (Figure C).
We also determined the percentages of various markers that floated
under our conditions: Golgi membranes containing GM130, ~80%; ER
membranes containing calnexin, ~ 50%; and plasma membrane-derived
vesicles containing the EGF receptor, ~45%. These data suggest that
at least ~25% of rat liver Vps35 and Vps26 are membrane associated,
whereas the remainder is in the cytosol in large complexes (see below).
However, given that 45–80% of known organelle marker proteins floated
under the conditions used rather than 100% of each marker, it is
possible that as much as ~50% of the cellular Vps35 and Vps26 are
truly membrane associated.
Associations of SNX1 and SNX2 with Human Vps26, Vps29, and Vps35 in
Mammalian Cells
In yeast, the retromer complex contains at least five proteins:
Vps35p, Vps17p, Vps26p, Vps29p, and Vps5p, the ortholog of mammalian
SNX1 and SNX2 (
Horazdovsky et al., 1997 ;
Seaman et
al., 1998 ). We inquired whether a similar complex is formed in
mammalian cells. Total cell homogenates were made in the absence of
detergent from COS7 cells coexpressing various combinations of
myc-tagged SNX1, hVps35, hVps29, and hVps26, and incubated with
antibodies generated against the various molecules. Interestingly,
antibodies to hVps26 immunoadsorbed hVps35, hVps29, and SNX1, as well
as hVps26 (Figure A, lanes 3, 5, and 7).
In contrast, antibodies to hVps35 brought down no detectable hVps26,
hVps29, or SNX1, and only small amounts of hVps35 (seen on darker
exposures), (Figure A, lanes 2, 4, and 6). These data suggest that
recombinant mammalian retromer proteins can form complexes in COS7
cells. These data, however, do not demonstrate whether the complexes
are formed in the cytoplasm, on membranes, or both. To address this
point, we immunoadsorbed a rat liver microsomal fraction with either
anti-Vps35 or anti-Vps26 antibodies (Figure B). As seen for the
transfected proteins, antibodies to Vps26 immunoadsorbed endogenous
Vps35, as well as endogenous Vps26 (Figure B, lane 2), whereas
antibodies to Vps35 brought down only small amounts of Vps35 (seen on
darker exposures) and no detectable Vps26 (Figure B, lane 1). Similar
results were obtained in immunoadsorptions from cytosol and from light
fractions (1.06–1.11 g/ml) obtained after linear sucrose gradient
fractionation of microsomes (our unpublished results).
To further investigate the type(s) of complex(es) that is formed in the
cytoplasm by the various mammalian retromer proteins, a cytosolic
fraction prepared from COS7 cells was separated on a Sephacryl S-300
gel filtration column. Proteins in each fraction were precipitated with
trichloroacetic acid and analyzed by SDS-PAGE followed by
immunoblotting with polyclonal antibodies against the
various retromer proteins. Little, if any, of the endogenous proteins
present in COS7 cell cytosol was detected in column fractions at the
size predicted for monomers. On the immunoblot, anti-hVps35
antibodies detect two bands. The faster migrating band elutes in the
range corresponding to ~200–300 kDa, whereas the slower migrating
band elutes from the Sephacryl S-300 column at somewhat smaller size
(Figure A). It is not clear why Vps35
appears as a doublet, but it is possible that the faster migrating band
may be a proteolytic fragment of the protein generated during the
analysis, or the result of variable splicing or posttranslational
modification. The majority of endogenous SNX2, Vps26, and Vps29 have
overlapping elution profiles with endogenous Vps35. In addition, a
small amount of Vps26 elutes from the column at a small size
corresponding to the predicted size for monomeric Vps26. Similar
results were found with endogenous proteins in rat liver cytosol (our
unpublished results). Gel filtration analyses of cytosolic fractions
prepared from COS7 cells overexpressing myc-tagged hVps35, hVps29,
hVps26, and SNX2 also showed that the transfected proteins formed large
complexes. However, the complexes tended to be larger with cytosol
prepared from COS7 cells overexpressing the various myc-tagged proteins
(Figure B). In addition, a portion of the myc-SNX2 and myc-hVps26 also
eluted in fractions at the size predicted for monomers. Interestingly,
the evidence suggests that the complexes (containing endogenous or
recombinant proteins) are heterogeneous with respect to stoichiometry
and composition.
To gain insight into how the various retromer proteins might assemble
into the cytosolic complexes, we used the COS7 cell overexpression
system as our model. We expressed various combinations of myc-tagged
SNX1, hVps35, hVps29, and hVps26 in COS7 cells and then examined the
elution profiles of each protein to define the requirements for
assembly. When myc-tagged hVps29 and hVps26 were expressed alone (our
unpublished results) or in combination (Figure
A), both molecules eluted in pool V at
positions consistent with the deduced masses of the monomers. Moreover,
when hVps35 was expressed alone or in combination with hVps26 (Figure
B), hVps35 eluted at ~220–300 kDa (Figure B, pool II), whereas
hVps26 eluted as a monomer (Figure B, pool V). However, under these
conditions very little hVps35 was detected in the cytosolic fraction.
In contrast, when hVps35 was coexpressed with hVps29 (Figure C), the
level of hVps35 in the cytosolic fraction was substantially higher, and
both proteins eluted with a distribution centered at ~220 kDa, a size
much larger than either protein's predicted monomeric weight. These
findings suggest that when hVps35 and hVps29 are coexpressed they are
part of a multimeric complex in the cytosol (Figure C, pools III and
IV). Lastly, when hVps35, hVps29, and hVps26 were expressed together,
all three molecules coeluted at a position corresponding to high
molecular masses (~280–440 kDa), suggesting that they form a
multimeric protein complex (Figure D, pool II). Interestingly, when
SNX1 was expressed alone (Figure A) or
in combination with one of the other retromer proteins, it eluted
exclusively as large oligomers (~440–158 kDa). The expression of
recombinant SNX1 along with various pairs of the other retromer
proteins also had no effect on the elution profiles of the recombinant
molecules (our unpublished results). However, coexpression of all four
molecules together shifted the elution position of some of the SNX1 to
a smaller size. Most of the SNX1 eluted in pool II when expressed by
itself. In contrast, most of the SNX1 eluted in pools III and IV when
expressed in the presence of hVps26, hVps29, and hVps35. The coelution
of all four molecules in pools II and III suggests that they may all be
present in the same multimeric complexes. However, the ratios of the
various retromer proteins are strikingly different in the two pools,
suggesting that the complexes may be heterogeneous with respect to the
stoichiometry of subunit composition.
Associations among Human Retromer Proteins in the Yeast Two-Hybrid
System
We next used the yeast two-hybrid system to define the binary
interactions between the retromer proteins and thereby elucidate the
interactions leading to assembly of the retromer complex. Two-hybrid
analyses confirmed previous coimmunoprecipitation studies (
Haft
et al., 1998 ), demonstrating that SNX1 and SNX2 associate to
form both homo- and heterodimers (Figure
). In addition, hVps35 participates in
binding interactions with SNX1, hVps26, and hVps29. These data suggest
that hVps35 can serve as the nucleus for a multimeric complex
containing SNX1, hVps26, and hVps35. We also observed several weak
binary interactions that may provide additional driving forces to
stabilize the complex: hVps26 with hVps29, hVps29 with SNX1, and hVps29
with SNX2.
We next designed experiments to map the binding sites on hVps35 for
other retromer proteins. Homology searches using the human Vps35 amino
acid sequence showed that hVps35 has at least three regions that are
well conserved throughout evolution (Figure ): region I (amino acid
residues 54–307), region II (residues 307–603), and region III
(residues 603–753). We hypothesized that the conserved sequences
corresponded to functional domains, and used this information to carry
out deletional analyses of the hVps35 molecule. Two-hybrid analyses
performed using these deletion constructs showed that hVps26 interacts
strongly with amino acid residues 1–172 of hVps35 (Figure
A), but only weakly with other parts
of the hVps35 molecule. Further deletion of the 1–172 fragment
abolished the interaction with hVps35. In contrast, hVps29 shows only a
weak interaction with the N-terminal portion of hVps35 (amino acids
1–307), but shows a strong interaction with amino acid residues
307–796 (Figure B). Additional truncation of hVps35 from either end
abolished interactions with hVps29. Lastly, SNX1 interacted strongly
with two regions of hVps35: amino acid residues 1–53 and 307–796 of
hVps35 (Figure C). Furthermore, deletional analyses of SNX1 showed
that amino acid residues 1–53 and 307–796 of hVps35 interacted with
the C-terminal region of SNX1, predicted to possess three coiled-coils.
In contrast, neither the N-terminal nor the phox homology domains of
SNX1 showed detectable associations with hVps35 or its deletion mutants
(our unpublished results).
This article has been 
