We showed that RING finger protein 13 RNF13, the murine homolog of C-RZF, is a type I integral membrane protein localized in the endosomal⁄ lysosomal system.. Abbreviations APP, Alzheimer
Trang 1an endosomal integral membrane E3 ubiquitin ligase
whose RING finger domain is released to the cytoplasm
by proteolysis
Jeffrey P Bocock1, Stephanie Carmicle1, Saba Chhotani1, Michael R Ruffolo1, Haitao Chu2and Ann H Erickson1
1 Department of Biochemistry and Biophysics, University of North Carolina, Chapel Hill, NC, USA
2 Department of Biostatistics, University of North Carolina, Chapel Hill, NC, USA
Proteins of the PA-TM-RING family have a
protease-associated (PA) domain and a RING finger domain
separated by a transmembrane (TM) domain PA
domains are 120–210 amino acid sequences located in the noncatalytic regions of diverse proteases [1,2] They are found in multiple members of MEROPS peptidase
Keywords
E3 ubiquitin ligase; neurite outgrowth;
protease-associated domain; proteolysis;
RNF13
Correspondence
A Erickson, Department of Biochemistry
and Biophysics, CB 7260 GM, University of
North Carolina, Chapel Hill, NC 27599, USA
Fax: +1 929 966 2852
Tel: +1 919 966 4694
E-mail: ann_erickson@med.unc.edu
(Received 1 November 2008, revised 23
December 2008, accepted 20 January 2009)
doi:10.1111/j.1742-4658.2009.06913.x
PA-TM-RING proteins have an N-terminal protease-associated domain, a structure found in numerous proteases and implicated in protein binding, and C-terminal RING finger and PEST domains Homologous proteins include GRAIL (gene related to anergy in leukocytes), which controls T-cell anergy, and AtRMR1 (receptor homology region-transmembrane domain-RING-H2 motif protein), a plant protein storage vacuole sorting receptor Another family member, chicken RING zinc finger (C-RZF), was identified as being upregulated in embryonic chicken brain cells grown in the presence of tenascin-C Despite algorithm predictions that the cDNA encodes a signal peptide and transmembrane domain, the protein was found in the nucleus We showed that RING finger protein 13 (RNF13), the murine homolog of C-RZF, is a type I integral membrane protein localized in the endosomal⁄ lysosomal system By quantitative real-time RT-PCR analysis, we demonstrated that expression of RNF13 is increased
in adult relative to embryonic mouse tissues and is upregulated in B35 neu-roblastoma cells stimulated to undergo neurite outgrowth We found that RNF13 is very labile, being subject to extensive proteolysis that releases both the protein-associated domain and the RING domain from the mem-brane By analyzing microsomes, we showed that the ectodomain is shed into the lumen of vesicles, whereas the C-terminal half, which possesses the RING finger, is released to the cytoplasm This C-terminal fragment of RNF13 has the ability to mediate ubiquitination Proteolytic release of RNF13 from a membrane anchor thus provides unique spatial and tempo-ral regulation that has not been previously described for an endosomal E3 ubiquitin ligase
Abbreviations
APP, Alzheimer’s precursor protein; AtRMR1, Arabidopsis thaliana receptor homology region-transmembrane domain-RING-H2 motif protein; CHO, Chinese hamster ovary; C-RZF, chicken RING zinc finger; CTF, cytoplasmic C-terminal fragment; EEA1, early endosomal antigen 1; ER, endoplasmic reticulum; GRAIL, gene related to anergy in leukocytes; HA, hemagglutinin; HAF, hemagglutinin and 3· FLAG epitopes; HRP, horseradish peroxidase; ICD, intracellular domain; LAMP2, lysosomal-associated membrane protein 2; MPR, mannose 6-phosphate receptor; MVB, multivesicular body; NLS, nuclear localization signal; PA, protease-associated; PDI, protein disulfide isomerase; PNGase F, peptide: N-glycosidase F; RNF13, RING finger protein 13; TM, transmembrane.
Trang 2families [3], including the transferrin receptor, a
cata-lytically inactive protease, prostate-specific membrane
antigen [4], the human Golgi⁄ endosomal signal
pepti-dase peptipepti-dase-like proteins SPPL2a and SPPL2b [5],
and streptococcal C5a peptidase [6] PA domains have
been proposed to serve as substrate or ligand
recogni-tion domains [1] or as protease regulatory regions [2],
yet they have been functionally characterized only in
plant proteins The BP-80 receptor, which targets
pro-teases to the plant lytic vacuole through recognition of
the NPIR sorting determinant, contains a PA domain
Binding of vacuolar proteases requires the PA domain
as well as other regions of the BP-80 luminal domain
[7]
RING finger proteins constitute a subfamily of the
proteins that possess a pattern of cysteine and histidine
residues that chelate zinc ions The RING subfamily is
thought to function exclusively in protein–protein
interactions rather than protein–nucleic acid
interac-tions [8] Many RING finger proteins are E3 ubiquitin
ligases [9] The ubiquitination system functions in a
variety of cellular processes, including protein
degra-dation and protein trafficking
PA-TM-RING proteins that combine these two
domains have been identified in plants, Xenopus,
Drosophilaand mammals, but not in yeast The
Arabi-dopsis thaliana PA-TM-RING receptor homology
region–transmembrane domain–RING-H2 motif
pro-tein (AtRMR1) was found to colocalize with a propro-tein
storage vacuole membrane marker and was predicted to
be a receptor mediating targeting to the plant storage
vacuole [10] This organelle is a multivesicular body
(MVB) containing segregated compartments of lytic and
storage activity [11,12] AtRMR1 was subsequently
determined to be responsible for sorting the bean
stor-age protein phaseolin to the protein storstor-age vacuole
[13] and was shown to bind to C-terminal vacuolar
sorting determinants on tobacco chitinase and barley
lectin [14], establishing that in plants the PA domain can
serve as a ligand-binding domain
The best-characterized mammalian PA-TM-RING
family member is RNF128⁄ gene related to anergy in
lymphocytes (GRAIL) GRAIL was first identified in a
screen for genes upregulated in anergic CD4+ T-cells,
which are unresponsive to antigen rechallenge [15] It
was further characterized as an E3 ubiquitin ligase that
localizes to recycling endosomes, and was later
con-firmed to be necessary for induction of T-cell anergy
[16,17] RING finger protein 13 (RNF13) was first
designated chicken RING zinc finger (C-RZF), a
pro-tein upregulated when chicken embryo brain cells were
treated with the extracellular matrix component
tenas-cin-C [18] The protein was also upregulated in basilar
papilla when chickens were exposed to acoustic trauma [19] A truncated splice variant that lacks a complete RING-H2 domain was additionally identified in mice [19] but was not characterized On the basis of immu-nofluorescence microscopy and nuclear fractionation experiments, Tranque et al [18] reported that RNF13
is a nuclear protein, even though the tmpred algo-rithm [20] predicts that it has a TM domain A recent study established that RNF13 is an E3 ubiquitin ligase whose expression is increased in pancreatic ductal ade-nocarcinoma tissues, suggesting that the protein may participate in pancreatic cancer development [21]
We show that RNF13 is synthesized as an endoso-mal integral membrane protein rather than a soluble nuclear protein, consistent with other members of the PA-TM-RING family We demonstrate that RNF13 mRNA is upregulated following initiation of neurite outgrowth, thus expanding on an array study that found RNF13 expression to be sufficient to induce neurite outgrowth [22] We show that RNF13 is sub-ject to unexpected proteolysis that releases both the
PA domain and the RING domain from the mem-brane, providing a biochemical basis for understanding the regulation of this family of multimodular endo-somal membrane E3 ubiquitin ligases
Results
Domain structure of RNF13 RNF13 contains a number of protein domains likely
to regulate its localization and function (Fig 1) The first 34 amino acid residues at the N-terminus are pre-dicted by the algorithm signalp v.3.0 [23] to function
as a transient signal peptide, suggesting that the newly synthesized polypeptide is translocated across the endoplasmic reticulum (ER) membrane cotranslation-ally Residues 56–162 (numbering based on alignment
in Fig S1) have a high degree of sequence identity to
PA domains netnglyc 1.0 [24] predicts that this domain contains two N-linked glycosylation sites, resi-dues 43 and 88 Consistent with synthesis on the ER, the program tmpred [20] predicts that residues 182–
203 comprise a 22-residue integral membrane sequence, indicating that RNF13 might be a type I membrane protein predictnls [25] predicts that RNF13 has a nuclear localization signal (NLS) (RRNRLRKD) at residues 214–221, in the cytoplasmic half of the pro-tein, near the membrane psort [26] also predicts that RNF13 has an NLS (PVHKFKK) but at residues 227–233, a site C-terminal to that identified by predictnls Residues 240–292 form a RING-H2 domain Contiguous with the RING-H2 domain is a
Trang 324-residue sequence (residues 284–307) predicted, with
a high probability score of 14.33 (significant if > 5),
by the algorithm pestfind [27] to be a PEST domain
PEST domains, defined as hydrophilic stretches of at
least 12 amino acids having a high concentration of
proline, glutamic acid, serine, and threonine, are
pro-tein domains that direct rapid degradation and thus
are usually found in proteins with a short half-life [28]
The remainder of the C-terminal region is rich in
ser-ine residues, similar to transcription factor activation
domains Multiple phosphorylation sites are predicted
in the cytoplasmic half of the protein both by
netph-osk1.0 [29] and by group-based phosphorylation
scoring(GPS) 1.1 [30,31]
Sequence alignment of RNF13 with other
PA-TM-RING proteins
Three PA-TM-RING proteins, plant AtRMR1, mouse
GRAIL and mouse RNF13, exhibit little overall
sequence identity, as shown in the alignment in Fig S1
Only approximately 12% of the amino acids are
identical between the three proteins, as determined by
tcoffeealignment [32] Most of the conserved residues
(gray boxes) lie within either the PA domain or the
RING-H2 domain
RNF13 is an E3 ubiquitin ligase
RING finger sequences frequently mediate ubiquitin
ligase activity [9]; however, at least three distinct roles
have been described for RING domains [33] We
there-fore investigated whether the RING domain in the
cytoplasmic half of RNF13 was capable of catalyzing
polyubiquitination The cytosolic domain of
RNF13D1–205 comprising residues 206–381, and thus
the entire RING-H2 domain, was expressed in bacteria
with or without the point mutation C266A This
muta-tion was designed to inactivate E3 ubiquitin ligase
activity of the RING-H2 domain, as does mutation of the same conserved cysteine in the RING domain of the E3 c-Cbl [34] The expressed proteins, which con-tained N-terminal 6· His epitope tags, were purified
on Ni2+–nitrilotriacetic acid affinity columns, eluted, and resolved by SDS⁄ PAGE An antibody against 6· His recognized two proteins in a western blot of each eluate (Fig S2A, lanes 1 and 2), establishing that both bands contained the N-terminal epitope tag The lower band could result from early termination, but the two discrete bands were reproducibly equally intense Thus,
it is more likely that C-terminal cleavage of the protein
by a bacterial enzyme produces the lower band The size difference of 2 kDa indicates that only approxi-mately 18 residues are missing from the C-terminus
As the RING domain of RNF13 is composed of resi-dues 240–292 out of 381, both protein bands should contain an intact RING-H2 domain The truncated RNF13 proteins eluted from the affinity columns were resolved on polyacrylamide gels that were stained with Coomassie Blue R250 to assess purity (Fig S2B) Eluted protein was assayed for ubiquitin ligase activity without further purification
When RNF13D1–205 was added to an in vitro ubiq-uitination reaction mixture including ubiquitin, puri-fied commercial E1 enzyme, and a commercial E2 enzyme, either UbcH5a, UbcH5c, or UbcH6, it was able to catalyze the formation of polyubiquitin chains,
as shown by the appearance of a high molecular mass ladder of protein bands (Fig S2C, lanes 1–3) As there were only four proteins present in this in vitro assay, and one of them was ubiquitin itself, these data sug-gest that, like many E3 ubiquitin ligases, RNF13 can ubiquitinate itself All three E2s assayed interacted with RNF13, but UbcH6 appeared to produce more polyubiquitination (Fig S2C, lane 3) As expected by analogy with c-Cbl, purified RNF13D1–205 C266A was unable to catalyze polyubiquitination when added
to a similar assay (Fig S2C, lanes 4–6), as seen by the
PA TM NLS RING PEST Ser-Rich
Fig 1 RNF13 is a PA-TM-RING protein composed of several domains that might regulate other proteins RNF13 is predicted by TMPRED [20]
to be a TM protein, with the hydrophobic TM domain falling in the middle of the amino acid sequence (residues 182–203) Additional major domains include a predicted signal peptide (residues 1–34), a luminal PA domain (residues 56–162), and a cytoplasmic RING-H2 domain (resi-dues 240–292) The protein is also predicted to have an NLS (resi(resi-dues 214–221 or 227–233), a PEST sequence (resi(resi-dues 284–307), and a serine-rich region predicted to be phosphorylated (residues 309–381) We prepared expression constructs containing one or more of the following epitope tags: an HA tag at position 38, a FLAG tag at position 377, or a 3· FLAG tag at position 381.
Trang 4failure to produce the characteristic polyubiquitin
ladder This indicates that catalysis of polyubiquitin
chains is specific to the RING-H2 domain of purified
RNF13, as a single point mutant of a conserved
cyste-ine can abrogate E3 ligase activity Failure to catalyze
polyubiquitination was also seen when assay mixtures
were prepared that lacked any E2 enzyme (Fig S2C,
lanes 7 and 8) These data show that the RING finger
of RNF13 requires active E2 enzyme to function as an
E3 ubiquitin ligase As expected, when any of the
other essential components of the reaction, including
the E1 or E3 enzyme, ATP, or ubiquitin, was not
included in the reaction, polyubiquitination did not
occur (data not shown)
RNF13 is an endosomal protein RNF13 is predicted to have a TM domain and signal peptide, suggesting that it is an integral membrane protein in the secretory pathway C-RZF was localized
to the nucleus in chicken embryo heart cells [18], but RNF13 was recently reported to be present in the ER and Golgi when expressed transiently in MiaPaca-2 pancreatic cancer cells [21] As no other PA-TM-RING protein has been found in the nucleus or the
ER, we performed immunofluorescence experiments to determine the subcellular localization of mouse RNF13 (Figs 2 and 3)
C
E
Fig 2 Endogenous, transiently expressed and stably expressed RNF13 all show punctate staining consistent with localization to endoso-mal–lysosomal vesicles (A, B) Primary cortical neurons prepared from embryonic day 14.5 mouse embryos were treated with MG132 for
12 h Endogenous RNF13 was detected with antibodies directed against the 14 amino acid C-terminal peptide of mouse RNF13 Staining was observed with the use of secondary donkey anti-rabbit Alexa Fluor 488 serum The size bar in (B) represents 10 lm (C) PC12 cells sta-bly expressing RNF13 were treated with MG132 for 12 h RNF13 expression was detected with mouse anti-FLAG serum and, as secondary antibody, donkey anti-mouse Alexa Fluor 568 serum (D–F) COS cells were transiently transfected with the RNF13 expression plasmid pSG5X-RNF13 FLAG377 RNF13 (D) was detected with rabbit anti-FLAG serum and, as secondary antibody, anti-rabbit Texas Red serum Cells were counterstained with mouse antibodies raised against PDI (E) and donkey anti-rabbit Alexa Fluor 488 serum These panels are merged in (F) The size bar in (D–F) represents 20 lm (G–I) HeLa cells stably expressing RNF13 were treated with MG132 for 12 h RNF13 (G) was stained with mouse FLAG and donkey mouse Alexa Fluor 488 sera Calnexin staining (H) was observed with rabbit anti-calnexin and goat anti-rabbit Alexa Fluor 568 sera These panels were merged in (I) The size bar in (G–I) represents 5 lm RNF13 did not colocalize with either of the two ER markers.
Trang 5RNF13 observed in embryonic mouse cortical
neu-rons using an antiserum specific for the C-terminal 14
amino acids of RNF13 showed punctate, non-nuclear
staining characteristic of endosomes and lysosomes
(Fig 2A,B) To facilitate detection of RNF13 by
immunofluorescence and to enable us to determine the
origin of the biosynthetic forms detected by western
blotting, we constructed vectors to express RNF13
with an N-terminal hemagglutinin (HA) epitope and a
C-terminal FLAG tag Stably expressed,
epitope-tagged RNF13 exhibited punctate staining in PC12
cells, which are derived from a pheochyromocytoma of
the rat adrenal medulla and are frequently used as a model for neuronal differentiation (Fig 2C) The same pattern was observed when epitope-tagged RNF13 was expressed either transiently in COS cells (Fig 2D–F)
or stably in HeLa cells (Fig 2G–I) Thus, ectopic expression from the vectors utilized in this study does not appear to alter the localization of RNF13 relative
to the endogenous protein
RNF13 was recently reported to be localized in the
ER, on the basis of transient expression in pancreatic tumor cells [21] In contrast, we found that the protein
is not present in the ER, as it failed to colocalize with
MPR
Golgin 97
K
F E
D
C
G
LAMP2
CD63
EEA1
Marker
Fig 3 RNF13 is localized in MVBs and endosomes COS cells (A–L) or HeLa cells (M–O) were transiently transfected with RNF13-FLAG377, which was detected using rabbit anti-FLAG sera (B, E, H, K, N) Cells were costained with mouse anti-human Golgin 97 (A) serum, mouse anti-human LAMP2 serum (D), mouse anti-human CD63 serum (G), mouse anti-human MPR serum (J) or mouse anti-human EEA1 serum (M) Primary antibodies were visualized with the secondary antibodies donkey anti-mouse AlexaFluor 488 serum (A, D, G, J), goat anti-rabbit Texas Red serum (B, E, H, K), donkey anti-rabbit AlexaFluor 488 serum (N) and goat anti-mouse AlexaFluor 568 serum (M) RNF13 colocalized with LAMP2 (F), CD63 (I), and MPR, (L), but not with Gol-gin 97 (C) or EEA1 (O) Images were obtained with a Zeiss LSM 210 confocal microscope The size bars represent 10 lm (A–C, J–L) and 20 lm (D–I).
Trang 6two different ER chaperone proteins RNF13 did not
colocalize with endogenous protein disulfide isomerase
(PDI) when expressed transiently in COS cells
(Fig 2D–F) Similarly, RNF13 expressed stably in
HeLa cells did not colocalize with calnexin (Fig 2G–
I) Consistent with this, RNF13 did not accumulate
with the trans-Golgi network protein golgin 97
(Fig 3A–C), indicating that our ectopically expressed,
epitope-tagged RNF13 is able to traverse the secretory
pathway efficiently
Our immunofluorescence confocal microscopy
studies indicated that RNF13 is localized in the
endo-somal–lysosomal system (Fig 3) RNF13 showed
significant colocalization with lysosomal-associated
membrane protein 2 (LAMP2), which localizes to the
membranes of endosomes and lysosomes (Fig 3D–F)
RNF13 also partially colocalized with CD63
(Fig 3G–I), a tetraspanin that localizes to
multivesic-ular endosomes [35], and with mannose 6-phosphate
receptors (MPRs) (Fig 3J–L), which are enriched in
late endosomes RNF13 failed to colocalize with the
early endosomal tether early endosomal antigen 1
(EEA1) (Fig 3M–O) at several planes of depth in the
cell Consistent with this, RNF13 did not colocalize
with fluorescently labeled transferrin internalized for
either 7.5 or 30 min by receptor-mediated endocytosis
(data not shown)
No accumulation of RNF13 in the nucleus could be detected at steady-state by immunofluorescence stain-ing of primary neurons or of cells expressstain-ing the pro-tein either stably or transiently (Figs 2 and 3) Similarly, nuclear RNF13 was not observed in pancre-atic cancer cells transiently expressing RNF13 [21]
RNF13 undergoes extensive post-translational proteolysis
To characterize the biosynthetic processing of RNF13,
we constructed viral expression vectors encoding mouse RNF13 with an HA epitope at position 38 and
a 3· FLAG epitope at position 381 (RNF13-HAF) that we used to infect Chinese hamster ovary (CHO) cells to produce the CHO-RNF13-HAF cell line, which stably expresses RNF13 FLAG-positive RNF13-spe-cific bands were not detected by western blot analysis
of cells expressing empty vector (Fig 4A, lane 1) Sur-prisingly, RNF13-specific FLAG-positive bands were barely detectable in cell lysate when cells stably expressing RNF13 were treated with dimethylsulfoxide vehicle for 8 h (Fig 4A, lane 2) When these cells were incubated with the protease inhibitor MG132 in dimethylsulfoxide for 8 h prior to harvest, however, a specific RNF13 banding pattern indicative of extensive post-translational modification was reproducibly
45” NS
1 2
1 2 3
-
-
-
1
2
3
97
37
54
kDa
Anti-FLAG
1 2 3 4 5
RNF13 DMSO MG132
RNF13 – + + + +
+ – + + + + – – +
– – –
+ + –
+ + + +
– – – + +
DMSO MG132 Epoxomicin
Anti-FLAG
Fig 4 RNF13 undergoes extensive post-translational proteolysis (A) CHO cells (lane 1) or CHO cells stably expressing RNF13-HAF (lanes 2 and 3) were treated, as indicated, with dimethylsulfoxide (DMSO) or MG132 for 8 h Equal quantities of cellular protein were resolved on a 12% polyacrylamide gel Biosynthetic forms of RNF13 were visualized on a western blot with mouse anti-FLAG serum Prestained molecular mass markers are indicated on the left (B) CHO cells (lane 1) or CHO cells stably expressing RNF13-HAF (lanes 2–5) were treated, as indi-cated, with dimethylsulfoxide, MG132 or epoxomicin for 10 h RNF13 was visualized with anti-HA serum (C) CHO cells transiently express-ing pSG5X-RNF13-HAF were pulse-labeled with [35S]methionine for 45 min RNF13 was immunoprecipitated with anti-FLAG serum and resolved on a 12% polyacrylamide gel (lane 1) To detect nonspecific protein bands, normal whole serum (NS) was substituted for specific affinity-purified anti-FLAG serum (lane 2).
Trang 7detected (Fig 4A, lane 3) The pattern included a
het-erogeneous collection of proteins of approximately
80 kDa, a second series of proteins that occasionally
resolved into four discrete bands at approximately
65 kDa (e.g Fig 5A, lane 2), three protein bands of
approximately 45 kDa, and one protein band of
approximately 36 kDa As all these proteins were
visu-alized with antiserum that recognizes the 3· FLAG
epitope at residue 381, they all contain the C-terminus
of RNF13 An identical protein pattern was obtained
when RNF13-HAF was expressed stably in B35 rat
neurons (data not shown)
We next utilized antiserum specific for the
N-termi-nal HA epitope tag (residue 38) to determine which of
the RNF13 proteins in the banding pattern contain the
N-terminus Cells were treated as indicated (Fig 4B)
The specific proteasome inhibitor epoxomicin
stabi-lized RNF13 (Fig 4B, lane 5), as did MG132 (Fig 4B,
lane 4) Both the heterogeneous bands at 80 kDa
and the group of bands at 65 kDa were recognized
by the anti-HA serum (Fig 4B, lanes 4 and 5) As
these proteins are also recognized by the anti-FLAG
serum, they must possess both residues 38 and 381 and
therefore be close to full-length RNF13 The lower
molecular mass bands around 45 kDa and at 36 kDa
were not recognized with anti-HA serum, suggesting that the N-terminal portion of the protein containing the HA epitope was lost from these proteins by pro-teolysis
To determine which of the RNF13 bands is the ini-tial biosynthetic product, we pulsed transiently trans-fected CHO cells expressing RNF13-HAF with [35S]methionine and immunoprecipitated RNF13 using antibodies specific for the FLAG epitope (Fig 4C) The major protein detected after a 45 min pulse migrated at 65 kDa (Fig 4C, lane 1) This protein band was absent upon immunoprecipitation with nor-mal serum as a negative control (Fig 4C, lane 2)
RNF13 acquires carbohydrate modification
As we observed forms of RNF13 that migrated more slowly on polyacrylamide gels than the 43 kDa form predicted by the primary sequence alone, we assayed the protein for sugar modification The netnglyc 1.0 algorithm predicts that RNF13 possesses two sequences in the N-terminal domain that could acquire N-linked carbohydrate Transiently expressed RNF13 was immunoprecipitated and treated with peptide: N-glycosidase F (PNGase F), which removes both asparagine-linked high-mannose and complex oligosac-charides [36] The 65 kDa region resolved, on this gel, into four distinct proteins in the absence of endogly-cosidase treatment (Fig 5A, lane 2) After endoglycosi-dase treatment, the two upper protein bands disappeared, with a concomitant increase of the lowest band Identical results were obtained with drug prepa-rations from two different suppliers (Fig 5A, lanes 3 and 4) To confirm this result, CHO cells transiently expressing FLAG-tagged RNF13 were cultured in the presence of tunicamycin, an antibiotic that inhibits transfer of N-acetylglucosamine 1-phosphate to doli-cholmonophosphate [37], thus blocking the synthesis
of asparagine-linked oligosaccharide chains on glyco-proteins Tunicamycin treatment reproducibly reduced the amount of the upper band and resulted in loss of the middle band These results confirm that two N-linked sugar chains can be removed from RNF13, supporting the predictions made by netnglyc 1.0
As a percentage of certain integral membrane pro-teins, such as the Alzheimer’s precursor protein (APP) [38] and the immunoglobulin invariant chain [39,40], acquire chondroitin sulfate glycosaminoglycan chains,
we also assayed RNF13 for this modification RNF13 possesses one potential Ser-Gly dipeptide acceptor sequence [41] in its luminal domain When immuno-precipitated RNF13 was treated with chondroitin-ase ABC, the intensity of the diffusely staining bands
1 2 3 4
A
B
5
– + Chondroitinase
+ + –
+ + –
+ – –
– – –
65 kDa
65 kDa
~80 kDa
Fig 5 RNF13 is modified with N-linked sugars and chondroitin
sulfate (A) pSG5X-RNF13-FLAG377 was expressed transiently in
CHO cells Immunoprecipitated RNF13 was treated with PNGase F
from two different manufacturers (lanes 3 and 4) Transfected cells
were incubated overnight with tunicamycin to block high-mannose
sugar addition (lane 5) Cellular proteins were resolved on a 12%
gel, and RNF13 was identified by western blotting using an
antise-rum specific for the FLAG epitope (B) Mouse RNF13-HAF was
expressed transiently in CHO cells, and immunoprecipitated with
antiserum specific for the FLAG epitope The immunoprecipitate
was split into two equal parts, which were incubated overnight in
the absence (lane 1) or presence (lane 2) of chondroitinase ABC
prior to resolution on a 12% polyacrylamide gel.
Trang 8at 80 kDa dramatically decreased, whereas the 65–
70 kDa bands increased in intensity (Fig 5B) This
result indicates that at least a proportion of the
RNF13 protein is modified with chondroitin sulfate
Proteolysis releases N-terminal and C-terminal
fragments of RNF13 from the membrane
To further characterize the 36 kDa FLAG-positive
RNF13 band observed in cell lysates, CHO cells were
transfected transiently with a construct that encodes
only the C-terminal half of RNF13 This variant
(RNF13D1–204) is initiated a few residues beyond the
putative TM sequence and retains the FLAG epitope
It was found to comigrate with the 36 kDa protein in
cell lysates (Fig 6A, lane 2 versus lane 4), suggesting
that the 36 kDa band is derived from full-length
RNF13 by proteolysis at or near the TM sequence An
HA-positive protein of approximately the same size
was reproducibly detected when blots probed with
anti-HA serum were overexposed (Fig 6B, lane 5)
This protein band always appeared fuzzy, consistent
with the presence of carbohydrate Detection of this
protein suggests that the N-terminal domain of
RNF13, like the C-terminal domain, is released by
proteolysis from the TM anchor localized
approxi-mately in the middle of the protein
RNF13 is a type I integral membrane protein
By isolating microsomes and stripping them of periph-eral proteins, we confirmed the prediction of Mahon & Bateman [1] that RNF13 is synthesized as an integral membrane, not a nuclear, protein We prepared micro-somal membranes, by Dounce homogenization in the presence of sucrose to maintain microsome integrity, from a postnuclear supernatant of MG132-treated CHO-RNF13-HAF cells Proteins in both the postnu-clear supernatant, which contains soluble cytoplasmic proteins, and in the pelleted microsomes were resolved
on a polyacrylamide gel (Fig 7A) A western blot was probed for both the FLAG and HA epitopes The
36 kDa protein, which comigrated with the expressed soluble C-terminal cytoplasmic half of the protein (Fig 6A), was detected in the postnuclear superna-tant⁄ cytoplasmic fraction (Fig 7A, lane 1), establish-ing that the FLAG-tagged C-terminal half of RNF13
is released from the membrane by proteolysis and thus resembles the intracellular domain (ICD) of other inte-gral membrane proteins such as APP and Notch Recovery of the FLAG-tagged C-terminal fragment in the cytoplasmic fraction also indicates that RNF13 is
a type I membrane protein that has its PA domain either in the lumen of vesicles or on the cell surface and its C-terminal half in the cell cytoplasm All other biosynthetic forms of RNF13, including the N-termi-nal HA-tagged domain (Fig 7B, lane 3), were present
in the microsome fraction, indicating they are either embedded in the microsomal membrane or present inside vesicles
To confirm that RNF13 is an integral, not a periph-eral membrane protein, we isolated microsomes from CHO-RNF13-HAF cells and lysed them in high-pH carbonate buffer (Fig 7B) Freezing and thawing microsomes in pH 11.5 buffer lyses vesicles and solu-bilizes peripheral membrane proteins not embedded in the membrane bilayer [42,43] The luminal lysosomal protease cathepsin L, detected as a control, was present in the soluble fraction, confirming that soluble content proteins are released by carbonate treatment (data not shown)
All forms of RNF13 present in microsomes and recognized by the anti-FLAG serum were present in the membrane fraction and were not solubilized when vesi-cles were lysed at high pH, indicating they are integral, not peripheral, membrane proteins (Fig 7B, lane 2) The HA-tagged 36 kDa fragment of RNF13 was also detectable within microsomes (Fig 7B, lane 3), sug-gesting that the luminal domain is shed within endo-somes With this protocol, an additional HA-positive protein in the RNF13 pattern was reproducibly
Anti-FLA
54
38
Fig 6 RNF13 undergoes proteolysis on both sides of its
trans-membrane domain (A) B35-Con cells (lane 1), B35-RNF13-HAF
cells (lanes 3 and 4) or CHO cells transiently expressing RNF13D1–
204, the ICD (lane 2), were treated, as indicated, with
dimethylsulf-oxide (DMSO) or with MG132 for 8 h Equal quantities of protein
(600 lg) were loaded in lanes 1, 3 and 4, and 100 lg of protein
was loaded in lane 2 Biosynthetic forms of RNF13 were visualized
on a western blot of a 10% polyacrylamide gel with anti-FLAG-HRP
serum (B) CHO-RNF13-HAF cells were treated with MG132 for
9 h, and RNF13 was visualized on a blot of a 12% gel with anti-HA
serum.
Trang 9able in the membrane fraction below 65 kDa
(Fig 7A,B, arrow) This protein, which lacks a
FLAG-tag and thus presumably has lost its C-terminal
sequences, was readily detectable in intact microsomes
(Fig 7A), but was less apparent once microsomes were
lysed (Fig 7B) It can often be visualized in whole cell
extracts after long exposure of the western blot to film,
suggesting that it corresponds to an authentic
biosyn-thetic form of RNF13 that is stable in intact
micro-somes (data not shown)
A model summarizing the major biosynthetic forms
of RNF13 and their relationship to membranes, based
on the observed molecular mass and presence or
absence of the epitope tags, is presented in Fig 7C
The three cytoplasmic C-terminal fragments (CTFs) or
‘stubs’ remaining after loss of the PA domain could be
generated by multiple proteases or could result from
multiple cleavages by one enzyme Other biosynthetic
intermediates may be present but not detectable by our
gel system Additionally, the ratio of the forms may
vary with the cell type and the metabolic condition of
the cells expressing RNF13
Inhibiting the vacuolar ATPase only partially
stabilizes RNF13
Since RNF13 localizes to the endosomal–lysosomal
membrane system, we treated cells with two inhibitors
that raise the pH of vesicles in an attempt to inhibit lysosomal proteolysis of RNF13 Bafilomycin A1 inhibits the vacuolar ATPase [44,45], and ammonium chloride raises the pH of lysosomes and blocks the light–heavy chain cleavage of lysosomal cathepsin L [46] Although MG132 is commonly employed as a proteasome inhibitor, it has also been reported to inhibit lysosomal cathepsins [47,48], calpains [47], and BACE1 [49] Stably expressed RNF13 was barely detectable in cell extracts unless the cells were pretreated with MG132 (Fig 8A, lane 1 versus lane 2) Inhibition
of the vacuolar ATPase with bafilomycin A1 or by treating cells with ammonium chloride (Fig 8A, lanes
3 and 4, respectively) stabilized biosynthetic forms of RNF13, but not as efficiently as did MG132 treatment
of cells The data suggest that other proteases primar-ily mediate the turnover of RNF13 in vesicles distinct from mature lysosomes
Ectopically expressed RNF13 ICD does not localize in the nucleus when expressed transiently from a plasmid
RNF13 is predicted to have one or two NLSs (Fig 1), but we detected only RNF13 in punctate structures by confocal microscopy (Figs 2 and 3) Similarly, we were unable to detect the FLAG-tagged 36 kDa ICD in preparations of purified nuclei (data not shown) We
Anti-FLAG
1 2 1 2
Anti-HA
Cyto Mb Cyto Mb
Cell fractionation
A B Lysed microsomes stripped of C
peripheral proteins
Sol Mb Sol Mb Microsomes
1 2 1 2 Anti-FLAG Anti-HA
Model for RNF13 Biosynthetic Forms
HA
70
FLA
Full-length CTF ICD ~kDa: 63
97-
54-
37
-97
54
3
-95 72 55
36
HA
Fig 7 Proteolysis releases a C-terminal fragment of RNF13 into the cytoplasm Microsomes were prepared by Dounce homogenization of CHO cells stably expressing RNF13-HAF treated with MG132 (A) RNF13 in the soluble cytoplasmic fraction (lane 1, Cyto) and in the pelleted microsome fraction (lane 2, Mb) was visualized by probing a blot with antibodies specific for the C-terminal FLAG or N-terminal HA epitope,
as indicated (B) Pelleted microsomes were lysed and stripped of peripheral membrane proteins by resuspension and incubation in pH 11.5 carbonate buffer RNF13 was visualized in the soluble (Sol) and membrane (Mb) fractions by probing a blot of a minigel with antibodies specific for the C-terminal FLAG or N-terminal HA epitope, as indicated The arrows mark an HA-tagged form that markedly decreases in intensity when microsomes are lysed (B) This protein is present, but more difficult to resolve, on commercial minigels [(B), all lanes except lane 3] For the minigel, 100 lg of protein was resolved in each lane (C) A model of the epitope-tagged protein bands detected is presented The three short horizontal lines on the full-length protein represent chrondroitin sulfate modification.
Trang 10therefore transiently expressed RNF13D1–204 3·
FLAG381 (Fig 9), containing only the C-terminal half
of RNF13 including the putative NLS, to determine
whether RNF13 could be observed in the nucleus
when expression of the ICD was high Figure 9A,
showing cells treated with dimethylsulfoxide alone,
establishes the specificity of the anti-FLAG serum
Despite the presence of two sequences predicted to be NLSs, RNF13D1–204 3· FLAG381 localized to the cytoplasm (Fig 9C) This is in agreement with our cell fractionation data, which also indicated that it is local-ized to the cytosol (data not shown)
RNF13 expression is higher in adult than in embryonic tissues
Genome sequencing suggests that RNF13 is ubiqui-tously expressed This is supported by expression data from the Stanford Microarray Database, which show RNF13 to be widely expressed in many cell types, including throughout tissues of the human immune and nervous systems [50] However, initial northern blot analysis of expression of C-RZF, the chicken homolog of RNF13, showed that the protein was expressed in embryonic heart and brain, but not in liver [18] We therefore analyzed mouse RNF13 expression by quantitative real-time RT-PCR, isolating mRNA from both embryonic and adult mouse tissues The oligonucleotides used in this assay were specifi-cally designed to bind only the full-length RNF13 transcript Expression of RNF13 in adult heart tissue was similar to that in spleen We observed fold increases of 5.7, 2.6 and 1.9 for adult kidney, liver and brain, respectively, relative to spleen (Table 1; see Table S3 for statistical analysis) The PA-TM-RING family member GRAIL has been found to have similar expression in mouse tissues, using northern blots [15], but it has been primarily studied in T-cells We also observed that RNF13 expression levels in adult tissues were higher than in the corresponding embryonic tissue For example, there was a four-fold increase in adult brain as compared to embryonic brain after 14.5
or 16.5 days of development (Table 1) Our analysis of embryonic tissue showed similar expression of RNF13
1 2 3 4
+MG132+Bafilom
ycin +DM
SO
+NH4 Cl
Anti-α Tubulin
Anti-A
B
54
Fig 8 Inhibitors stabilize RNF13 cleavage fragments (A) CHO
cells stably expressing RNF13-HAF were treated, as indicated, with
dimethylsulfoxide (DMSO) (lane 1), MG132 (lane 2) or
bafilo-mycin A1 (lane 3) for 8 h, or with NH 4 Cl for 24 h (lane 4)
Biosyn-thetic forms of RNF13 were visualized on a western blot of a 12%
polyacrylamide gel with mouse anti-FLAG serum Migration of
pre-stained molecular mass markers is indicated on the right All lanes
shown are derived from the same exposure of the same blot (B)
Equal quantities of total cellular protein were loaded in each lane,
as verified by blotting for a-tubulin.
RNF13 +DMSO RNF13 +MG132 RNF13 ICD
Fig 9 Expressed RNF13 ICD is not localized in the nucleus (A, B) CHO cells stably expressing RNF13-HAF were plated on coverslips, incu-bated for 10 h with either dimethylsulfoxide (DMSO) vehicle (A) or MG132 (B), and stained with anti-FLAG serum, as indicated (C) CHO cells were transiently transfected with a plasmid encoding RNF13D1–204 3· FLAG381, the soluble ICD of RNF13, and stained with anti-FLAG serum RNF13 was visualized by confocal microscopy The size bar in (C) represents 10 lm.