ABSTRACT
Levels of most nonsense mRNAs are normally reduced in prokaryotes and eukaryotes
when compared with that of corresponding functional mRNAs. Genes encoding
polypeptides that selectively reduce levels of nonsense mRNA have so far only
been identified in simple eukaryotes. We have now cloned a human cDNA whose
deduced amino acid sequence shows the highest degree of homology to that of
UPF1, a
bona fide
Saccharomyces
cerevisiae
group I RNA helicase required for accelerated degradation of nonsense mRNA.
Based on the total sequence of the shorter yeast UPF1 protein, the overall
identity between the human protein and UPF1 is 51%. Besides NTPase and other
RNA helicase consensus motifs, UPF1 and its human homolog also share similar
putative zinc finger motifs that are absent in other group I RNA helicases.
Northern blot analysis with the human cDNA probe revealed two transcripts in
several human cell lines. Further, antibodies raised against a synthetic
peptide of the human polypeptide detected a single 130 kDa polypeptide on
Western blots from human and mouse cells. Finally, immunofluorescence and
Western blot analyses revealed that the human and mouse polypeptides, like
yeast UPF1, are expressed in the cytoplasm, but not in the nucleus. We have
thus identified the first mammalian homolog of yeast UPF1, a protein that
regulates levels of nonsense mRNA, and we tentatively name this protein human
HUPF1 (for human
h
omolog of
UPF1
).
An mRNA with a premature translational stop codon (nonsense mRNA) can originate
from mistakes during post-transcriptional events such as RNA editing and RNA splicing (reviewed in
1
). Alternatively, nonsense mRNA can be transcribed from a germline or
somatically mutated gene, from a pseudo-gene (
2
) or from non- productively rearranged immunoglobulin (Ig) and T cell receptor (TCR)
genes (reviewed in
3
). A nonsense mRNA encodes a truncated polypeptide that, if it accumulates, can
affect normal cellular processes in a dominant-negative or gain-of-function fashion. Thus, high levels of nonsense mRNA might
influence the growth, differentiation or other physiological functions of a
cell (
4
-
6
).
Cytoplasmic levels of most nonsense mRNAs are normally reduced in higher and
simple eukaryotes, as well as prokaryotes, when compared with that of their
corresponding functional mRNAs (reviewed in
1
,
7
-
9
). Genes that encode polypeptides required to reduce levels of nonsense mRNA
have so far only been identified in the yeast
Saccharomyces cerevisiae
(
10
-
13
) and
Caenorhabditis elegans
(
6
). One example of such a gene is the
S.cerevisiae
UPF1
gene (for
up
-
f
rameshift mutation
1
;
10
), also known as the
NAM7
gene (for
n
uclear
a
ccommodation of
m
itochondria;
14
). However, a mammalian gene encoding a protein that controls levels of nonsense
mRNA has not previously been identified or cloned.
Using combined comparative genomics and cDNA library screenings, we isolated and
characterized a human cDNA clone that encodes a structural homolog of yeast
UPF1. Hence, we suggest naming this protein human HUPF1 (for human
h
omolog of
UPF1
).
All cell lines were grown in complete RPMI medium as described (
15
). VXH is a murine B cell hybridoma line (
16
). Human cell lines used in this study are the plasmacytoma line MC/CAR
(American Type Culture Collection no. CRL8083), the heart muscle line HA-VSMC (ATCC no. CRL-1999), the glioma line Cla (established from a grade 4 neuroblastoma by Dr Len Erickson, Indiana
University), the T lymphoma line Jurkat (ATCC no. TIB152), the B lymphoma line
Raji (ATCC no. CCL86) and the monocyte line U-937 (ATCC no. CRL 1593).
About 2.5 * 105
recombinant [lambda] phages from each cDNA library were plated and screened with 32
P-nick-translated DNA probes as described (
17
). A 1.5 kb
Hin
dIII-
Not
I fragment from EST clone R13609, which was obtained from the IMAGE Consortium
through Genome Systems Inc (St Louis, MO), was used to isolate clone 3.6 from
an amplified oligo(dT)-primed human HeLa cDNA library in the excisable phage vector [lambda]YES (
18
). A 1.4 kb
Xho
I-
Sal
I fragment from clone 3.6 was used to isolate clone 5.5 from an amplified random
hexamer/oligo(dT)-primed human Jurkat cDNA library in the excisable phage vector [lambda]ZapII (Stratagene, La Jolla, CA).
Double-stranded nucleotide sequencing was performed on two identical inserts by
the Sanger dideoxy chain termination method (
19
) and primer walking (
20
). Percent nucleic acid and amino acid identities and similarities were
determined using the BLAST program (
21
).
Total RNA was prepared with the RNeasy Total RNA Isolation kit from Qiagen Inc.
(Chatsworth, CA) and 10 [mu]g RNA was analyzed by Northern blotting as described (
22
). Bands were detected by autoradiography.
An anti-human HUPF1 peptide serum was generated by immunizing a rabbit with a
synthetic human HUPF1 peptide (codons 106-123 in Fig.
1
A) coupled to keyhole limpet hemocyanin. Anti-human HUPF1 peptide antibodies were purified on a peptide affinity column.
The affinity-purified antibodies were used at a 1:200 dilution on Western blots and at
a 1:25 dilution in immunofluorescence analysis. The generation of rabbit anti-mouse BiP antibodies was as previously described (
23
). Anti-proliferating cell nuclear antigen (PCNA) antibody was purchased from
Novocastra Laboratories (Burlingame, CA), horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG from BioRad (Richmond, CA), HRP-conjugated goat anti-mouse IgG from Southern Biotechnolgy (Birmingham, AL)
and fluorescein (FITC)-conjugated goat anti-rabbit (IgG) antibodies from Gibco-BRL (Gaithersburg, MD).
Nuclei were prepared according to the citric acid/Triton X-100 method (
24
). Cells were collected at 150
g
for 5 min and washed twice in ice-cold PBS. Cells were resuspended in 5 ml ice-cold 25 mM citric acid, 1% Triton X-100, allowed to swell on ice for 5 min and subsequently
homogenized (10 strokes in a type-B dounce homogenizer) on ice. The homogenate was layered on top of a 0.88
M sucrose, 25 mM citric acid cushion. Nuclei were pelleted at 800
g
for 5 min at 3oC, resuspended in 5 ml ice-cold 25 mM citric acid, 0.25 M sucrose and pelleted again through a
0.88 M sucrose, 25 mM citric acid cushion. Nuclei were then resuspended in 25%
glycerol, 5 mM magnesium acetate, 0.1 mM EDTA and 50 mM Tris, pH 8.0. The
integrity of nuclei and the absence of cytoplasmic tags were confirmed by phase
contrast microscopy. Prior to subjecting the SDS-PAGE, nuclei were lysed in 3* SDS sample buffer (0.2 M Tris, pH 6.8, 30% glycerol, 15% [beta]-mercaptoethanol, 0.006% bromophenol blue and 7.5% SDS)
and boiled for 3 min. To prevent protein degradation, all buffers contained 1 [mu]M leupeptin, 1 mM phenylmethylsulfonyl fluoride (PMSF), 1 [mu]M pepstatin, 100 [mu]M EDTA and 0.072 [mu]g/[mu]l aprotinin.
Cytoplasmic immunofluorescence was performed as described (
15
). Proteins were separated on a 10% SDS-polyacrylamide gel and analyzed by Western blotting as described (
15
), with the exception that cells were lysed in the presence of the protease
inhibitors leupeptin (1 [mu]M), PMSF (1 mM), pepstatin (1 [mu]M), EDTA (100 [mu]M) and aprotinin (0.072 [mu]g/[mu]l) and protein-antibody complexes were visualized with HRP-conjugated goat anti-rabbit IgG antibodies (1:10,000; BioRad,
Hercules, CA) and an Enhanced Chemiluminescence kit from Boehringer Mannheim
(Indianapolis, IN).
A 3.5 kb
Xho
I fragment containing the complete coding region of human
HUPF1
was isolated from clone 3.6 and cloned into the
Xho
I site of plasmid pGEM-7Zf(-) (Promega Corp., Madison, WI). The linearized plasmid was subjected to an
in vitro
coupled transcription-translation reaction in the presence of Tran35
S label (ICN Biochemicals, Costa Mesa, CA) using the TnT® Coupled Reticulocyte Lysate System from Promega (Madison, WI). Translation
products were separated on a reducing 10% SDS-polyacrylamide gel and detected by fluorography.
To clone a mammalian
UPF1
homolog, we used a comparative genomics approach, provided by XREFdb (
26
) through the World Wide Web site http://www.ncbi.nlm.nih.gov/xrefdb. When we
searched the human, mouse and rat database subsets of expressed sequence tags
(ETS) of random cDNA clones (dbEST;
27
) with the complete UPF1 amino acid sequence, XREFdb identified a 1.5 kb EST
cDNA fragment with sequence homology to
UPF1
. The 5' `single pass' 396 nt sequence of the 1.5 kb cDNA fragment (GenBank
accession no. R13609) showed 63% identity and 71% similarity at the amino acid
level to that of the RNA helicase region of yeast UPF1 (data not shown). Using
the 1.5 kb R13609 fragment, we isolated a [lambda] phage containing a 3.6 kb
Eco
RI insert (clone 3.6) from a human HeLa cDNA library and determined the complete
nucleic acid sequence of its insert (Fig.
1
A). Clone 3.6 contains from nt 176-3529 a 3354 bp open reading frame (ORF). The first N-terminal ATG codon of clone 3.6 from nt 176-178 is very likely the
true translation initiation codon, because it is flanked by sequences that
predict strong translational initiation (
28
) and preceded by an in-frame TAG stop codon at nt 107-109. The ORF encodes a polypeptide of 1118 amino acids with a
calculated molecular mass of 123 kDa and a calculated pI of 6.06. When we
analyzed
in vitro
translated products from a template of clone 3.6 by SDS-PAGE, we detected a single band with an apparent molecular mass of ~130 kDa (Fig.
2
). This suggests that clone 3.6 encodes a 130 kDa polypeptide. The difference
between predicted (123 kDa) and apparent (130 kDa) molecular weight might be
due to the fact that acidic polypeptides run aberrantly in SDS-polyacrylamide gels (
23
). Finally, we found at the 3'-end of clone 3.6 one classical polyadenylation signal (AATAAA) from
nt 3592-3597. However, residual adenine residues could not be detected at the 3'-terminal end, suggesting that clone 3.6 lacks part of the 3'-untranslated region.
When we searched the translated non-redundant GenBank, PDB, SwissProt and PIR sequence databases (release date
July 22, 1996) with the deduced amino acid sequence of clone 3.6 using the
BLAST program (
21
), we found that the human polypeptide showed the highest degree of homology to
the UPF1 (NAM7) protein of
S.cerevisiae
(
10
,
14
) and a UPF1-like hypothetical 105.6 kDa protein whose ORF was identified in a cosmid
clone of
Schizosaccharomyces
pombe
(SwissProt accession no. W02865). Based on the complete amino acid sequence of the shorter yeast UPF1 protein,
the overall identity between the human and the yeast UPF1 proteins is 51%. The
human polypeptide and UPF1, as well as the UPF1-like hypothetical 105.6 kDa protein from
S.pombe
(which we abbreviate to UPF1-like
S.pombe
protein), each contain putative zinc finger motifs at their N-terminal ends and central RNA helicase regions (Figs
1
and
3
). The human coding sequence from codon 121 to codon 918, which contains the
zinc finger and RNA helicase regions and represents 71.3% of its ORF, is 60%
identical and 74% similar to the corresponding
UPF1
(
NAM7
) coding region from codon 60 to codon 856.
When we performed an XREFdb search with the complete amino acid sequence of
yeast UPF1 and HUPF1, another human EST cDNA probe (GenBank accession no.
F06433), whose deduced amino acid sequence showed 98 and 54% identity to the
RNA helicase regions of human HUPF1 and UPF1, respectively, was identified
(data not shown). This XREFdb search also revealed that the F06433 probe was
selected by the XREF project to localize the mouse homolog of
UPF1
(mouse
HUPF1
) to chromosome 8 (XREFdb mouse map report for D8Xrf83, unpublished results).
This inferred that the human homolog is located on chromosome 19. Indeed, the
same probe (F06433) hybridized to a somatic hybrid cell line containing only
human chromosome 19 (XREFdb human map report for D8Xrf83, unpublished results).
Finally, the XREFdb search using the human HUPF1 amino acid sequence revealed
two overlapping human EST clones (GenBank accession nos H13969, H13971), both
of which have been isolated by exon trapping from a cosmid clone containing
part of human chromosome 19 (19p12-p13.1;
33
). The deduced amino acid sequences of the two clones (195 and 187 amino acid
residues, respectively), differed from that of the corresponding region in
human HUPF1 by only one and two amino acids, respectively (data not shown). These findings and the results from
unpublished XREFdb mapping reports strongly suggest that the human
HUPF1
gene is located on chromosome 19.
If HUPF1 is required to reduce levels of nonsense mRNAs, we would expect that
the
HUPF1
gene is expressed in every tissue, because any cell can potentially generate
nonsense mRNA, for example via imprecise pre-mRNA splicing. When we used the human 3.6 kb
HUPF1
probe (clone 3.6 in Fig.
4
B) to perform a Northern blot analysis of total RNA from several human cell
lines representing various human tissues, we detected a predominant ~5.5 kb and a minor ~3.7 kb transcript in all analyzed samples (Fig.
4
A), suggesting that
HUPF1
is indeed expressed in many tissues.
To isolate a
HUPF1
cDNA that represents the 5.5 kb transcript, we screened a human Jurkat [lambda] phage cDNA library with a 5' probe of clone 3.6 (
Xho
I-
Sal
I fragment in Fig.
4
B). We identified one clone with a ~5.5 kb
Eco
RI insert (clone 5.5), whose size corresponds very well with that of the 5.5 kb transcript in Figure
4
A. Restriction mapping and partial DNA sequencing of clone 5.5 revealed that clone
3.6 overlaps entirely with clone 5.5. For example, we detected in clone 5.5 all
restriction sites that are present in clone 3.6 (Fig.
4
B and data not shown). Moreover, we verified by partial DNA sequencing in clone
5.5 sequences of clone 3.6 extending from nucleotide positions 1 to 190 (Figs
1
A and
4
B, contains the translational start codon) and from nucleotide positions 3514 to
3601 (Figs
1
A and
4
B, contains the translational stop codon and the polyadenylation site). We also
detected at the 3'-end of clone 5.5 an additional classical polyadenylation signal
site (AATAAA), however, a poly(A) tail could not be detected (data not shown).
Therefore, clones 3.6 and 5.5 contain the same ORF, but clone 5.5 has a longer 3'-untranslated region with an extra polyadenylation site at its 3'-end. Thus, we conclude that clone 5.5 represents a
partial clone of the 5.5 kb transcript, whereas clone 3.6 represents either a
partial clone of the 5.5 kb transcript or a full-length clone of the 3.7 kb transcript. However, we cannot exclude the
possibility that the 3.7 kb transcript is encoded by another RNA helicase gene
with considerable sequence homology to the human
HUPF1
sequence.
Western blot analysis of human cell extracts with anti-HUPF1 peptide antibodies revealed a single band with an apparent molecular
mass of 130 kDa (Fig.
4
C, lane 2). The size of this band corresponds with the 130 kDa band that was
detected in products translated
in vitro
from a template of clone 3.6 (Fig.
2
). Because we also detected a 130 kDa band in mouse cells (Fig.
4
C, lane 1), the 130 kDa polypeptide is, at least in part, conserved between
mouse and man. Thus, we conclude that both the human and mouse
HUPF1
genes encode a single 130 kDa polypeptide.
UPF1, the yeast homolog of human HUPF1, is located in the cytoplasm but not in
the nucleus (
34
). Immunofluorescence analysis with anti-HUPF1 peptide antibodies revealed a similar expression pattern of the
HUPF1 protein in ethanol-fixed mouse VXH cells (Fig.
5
A) as well as human Raji and U937 cells (data not shown), i.e. we detected a
typical cytoplasmic staining, but little or no nuclear staining. The dim
nuclear staining in some cells is very likely due to non-specific staining or staining that resulted from cytoplasm squashed on top
of the nuclei during the mounting procedure, because a similar nuclear staining
pattern was detected with antibodies against immunoglobulins (data not shown).
The absence of nuclear HUPF1 in mouse VXH cells was confirmed by Western blot
analysis of purified nuclei using the anti-HUPF1 peptide antibodies (Fig.
5
B, lane 2). However, HUPF1 could, as expected, be easily detected in post-nuclear extracts of these cells (lane 1). The same results were obtained
when the human monocytic cell line U-937 was examined by Western blotting (data not shown). To isolate nuclei
from cultured cell lines, we used the citric acid/Triton X-100 method that yields intact nuclei lacking the outer nuclear membrane
with attached endoplasmic reticulum and polysomes. That the nuclei prepared by
this method were intact and free of contaminating endoplasmic reticulum was confirmed by the presence of PCNA protein (Fig.
5
B, lane 2) and the absence of the 72 kDa BiP protein (
35
), an endoplasmic reticulum-resident chaperone (Fig.
5
B, lane 2), respectively.
In summary, we have cloned a putative human RNA helicase that represents the
first mammalian structural homolog of yeast UPF1, a protein that is required
for accelerated degradation of nonsense mRNA. These conclusions are based on
three observations. First, HUPF1 shows the highest degree of sequence homology
at the amino acid level to the
bona fide
yeast RNA helicase UPF1. Second, both proteins exhibit a very similar overall
polypeptide domain structure, i.e. each contains several putative zinc finger
motifs and seven motifs characteristic of group I RNA helicases. Putative zinc
finger and RNA helicase consensus motifs have so far only been found together
on the same polypeptide chain in the group I RNA helicase UPF1 (NAM7) of
S.cerevisiae
and the UPF1-like protein of
S.pombe
, as well as the group II DEAD box RNA helicase GLH-1 of
C.elegans
(
36
). Finally, HUPF1, like UPF1, is expressed in the cytoplasm, but not within the
nucleus.
Evidence that HUPF1 is not only a structural but very likely also a functional
homolog of the nonsense mRNA-reducing yeast UPF1 protein comes from a study published during the review
of this manuscript (
37
). In this study, Perlick and colleagues reported the cloning and sequencing of
the
Rent1
(for
re
gulator of
n
onsense
t
ranscripts) gene, which is identical to our
HUPF1
gene. Using a modified allosupression growth assay these authors also showed
that a chimeric yeast UPF1/human RENT1 protein restored the UPF1+
phenotype in UPF1-deficient yeast
.
RNA helicase motifs.
UPF1 and HUPF1 contain seven conserved motifs that are characteristic of RNA
helicases (Figs
1
and
3
A). RNA helicases are enzymes that recognize and unwind double-stranded RNA regions. These enzymes are involved in gene transcription and
recombination, DNA replication and repair, RNA processing, transport and
stability, as well as protein translation (reviewed in
38
). To facilitate the displacement of paired RNA strands, RNA helicases must
contain an ATPase activity and single-stranded (ss) RNA binding sites (discussed in
39
). Additionally, extra RNA binding sites might be required to confer substrate
specificity to the RNA helicase. Some of these activities, such as ATPase, RNA
binding and helicase activities, were previously identified in affinity-purified UPF1 (
32
). The RNA helicase consensus motifs I and II, which are also commonly found as A box and B box elements in NTPases (
40
), are critical for ATP hydrolysis and helicase activity of UPF1 (
32
), as well as of other RNA helicases (
41
). The helicase motif VI, which is critical for
in vivo
function of UPF1 (
10
), might be a candidate for a ss RNA binding site because mutations in this
motif abolished ATPase, helicase and RNA binding activities of yeast UPF1 (
42
).
Potential zinc finger motifs.
Other nucleic acid binding sites might be formed by putative zinc finger motifs
that can be located in the cysteine/histidine-rich region of HUPF1, UPF1 and the UPF1-like
S.pombe
protein (Fig.
3
B). Putative zinc finger motifs can be partially aligned with the zinc
finger/knuckle motif, CX2-5
CX4-12
C/HX2-4
C/H, found in some transcription factors as well as RNA binding proteins (
43
). For example, two potential zinc finger motifs within the cysteine/histidine-rich region of HUPF1 (starting at codon 123 and codon 183, respectively),
as well as of UPF1 (starting at codon 62 and codon 122, respectively) and the
UPF1-like
S.pombe
protein (starting at codon 44 and codon 104, respectively), can be written as
CX2
CX28
HX3
H (C indicates a cysteine, H a histidine and X any amino acid residue) and CX2
CX22
CX3
C (Fig.
3
B) respectively. The distances between these two motifs are preserved in all
three polypeptides (Fig.
3
B).
The first zinc finger motif (CX2
CX28
HX3
H), which was not recognized in UPF1 (NAM7) by other researchers (
10
,
14
), resembles the motif of a CC/HH zinc finger protein (reviewed in
44
), such as the 5S RNA binding proteins TFIIIA (
45
,
46
) and p43 (
47
). Two additional cysteine residues in this motif are conserved in HUPF1, UPF1
and the UPF1-like
S.pombe
protein (Fig.
3
B), indicating that these residues might also be critical for protein function.
The second zinc finger motif (CX2
CX25
CX3
C), which was previously recognized in UPF1 (NAM7) (
10
,
14
), approximates a motif characteristic of CC/CC zinc finger proteins, such as
glucocorticoid receptors (
48
) and the eukaryotic translational initiation factor eIF-2[beta] (
49
). In eIF-2[beta], the zinc finger motif CX2
CX18
CX2
C is required to recognize the AUG start codon during the ribosome-mediated translational scanning process. Therefore, it is tempting to
speculate that the two zinc finger motifs of HUPF1 and UPF1 interact with ds or
ss RNA regions and assist in the recognition of nonsense codons. For example,
the zinc fingers might facilitate interaction between UPF1 or HUPF1 and
ribosomes. Once a nonsense codon is recognized by the translocating ribosome, a
conformational change of the zinc fingers might activate the helicase domain,
which results in unwinding of ds RNA regions and, ultimately, in accelerated
decapping and degradation of the nonsense transcript (
50
). Alternatively, the zinc fingers might facilitate recognition of an mRNA
instability sequence located downstream of the nonsense codon (reviewed in
8
), which also leads to unwinding of ds RNA regions, decapping and degradation.
A third potential zinc finger motif (CX2
CX6
CX3
C, starting at codon 62), which fits the zinc finger/knuckle consensus motif CX2-5
CX4-12
C/HX2-4
C/H very well, was previously recognized in UPF1 (
10
,
14
). The same motif is present in HUPF1 (starting at codon 123), but absent from
the UPF1-like
S.pombe
protein and, therefore, might not be critical for function (Fig.
3
B)
Other motifs.
Both yeast UPF1 and human HUPF1 contain at their N-terminal ends short acidic amino acid stretches without much sequence
homology (Fig.
1
A). Acidic amino acid stretches of limited homology are also found in many
transcription factors and might be involved in homodimer formation (discussed
in
14
and reviewed in
51
). Similarly, the acidic stretches in UPF1 and HUPF1 could facilitate formation
of a multicomponent nonsense mRNA-reducing complex. It has recently been shown that UPF1 interacts with UPF2
(
11
,
12
), another UPF factor required for nonsense codon-mediated degradation of mRNA in yeast (
10
).
HUPF1 contains two additional putative structural and functional motifs that are
absent in UPF1. First, three tandemly repeated glycine-rich motifs (PGGXG, Fig.
1
A) are located in the N-terminal tail. Glycine-rich regions have been proposed to facilitate protein-protein (
52
-
54
) or protein-RNA interactions (reviewed in
43
). Second, a short positively charged 27 amino acid residue stretch (Fig.
1
A, codons 1003-1029), which contains seven basic (arginine or lysine) and 11 small amino
acid residues (glycine or alanine), can be arranged into one conserved and two
degenerated RGG motifs. Such motifs are found in some RNA binding domains
(reviewed in
43
).
There is strong experimental evidence that UPF1-dependent reduction of nonsense mRNA occurs in the cytoplasm of yeast
cells (reviewed in
8
). For example, nonsense mRNA is associated with polysomes (
55
), a dominant-negative form of UPF2 interferes with the function of UPF1 in yeast cells
when it is targeted to the cytoplasm, but not when it is targeted to the
nucleus (
12
), and yeast UPF1 is located in the cytoplasm, co-purifies with polysomes and is not detected in the nucleus (
34
).
The finding that HUPF1 is expressed in the cytoplasm, but not within the nucleus
of mammalian cells, suggests that HUPF1, like yeast UPF1, exerts its function
in the cytoplasm. Therefore, HUPF1, in concert with other HUPF factors, might
only induce degradation of nonsense mRNAs, such as nonsense Ig [mu] (Li and Jäck, in preparation) and adenine phosphoribosyl transferase (APRT)
mRNAs (
56
), once these transcripts have reached the cytoplasm. However, our analysis does
not exclude the possibility that HUPF1 is also associated with the outer
nuclear membrane or the nuclear pore complex and that small amounts of HUPF1,
which escaped our detection methods, could be transported into the nucleus.
Therefore, HUPF1 might also induce translation-dependent reduction of nucleus-associated nonsense mRNAs, such as nonsense triosephosphate
isomerase (TPI) mRNA (
57
,
58
), and interfere with the splicing of nonsense pre-mRNA (
59
,
60
,
61
; reviewed in
1
). The intracellular localization of HUPF1 by immunoelectron microscopy, as well as the analysis of cytoplasmic and nuclear nonsense mRNA levels and turnover rates in mammalian cells that lack
the HUPF1 gene, will distinguish between these possibilities.
The isolation of a human gene whose deduced amino acid sequence has a striking homology to the nonsense mRNA-reducing UPF1 factor implies a critical role of nonsense mRNA-reducing factors in mammalian organisms. For example, HUPF proteins
reduce cytoplasmic levels of nonsense mRNA that could compete with functional
mRNA for RNA processing and regulation factors as well as for translating
ribosomes. Additionally, reducing nonsense mRNA levels prevents the
accumulation of high levels of truncated polypeptides that can interfere with
the function of the corresponding full-length polypeptide chain. For example, a shorter form of the myosin heavy
chain, which is translated from stable nonsense mRNA, causes, even in the presence of its corresponding full-length chain, abberant development of muscle cells in
C.elegans
(
6
). Within the immune system, HUPF proteins might be critical to reduce levels of
nonsense Ig and TCR transcripts in lymphocytes, which are predominantly encoded
by non-productively rearranged Ig and TCR genes (
3
). Because non-productively rearranged Ig and TCR genes are often generated during
somatic rearrangement of Ig and TCR gene segments, lymphocytes that contain on
one chromosome a productive and on the other chromosome a non-productively rearranged Ig or TCR gene can easily be detected in lymphoid
organs (reviewed in
3
). If nonsense Ig and TCR transcripts were not removed from a lymphocyte, the
accumulating shorter Ig or TCR polypeptides could compete with their
corresponding functional chains for other polypeptides that are required for
the assembly of signal transducing Ig or TCR complexes. Because signaling
through Ig and TCR complexes is obligatory for either a lymphocyte precursor to
develop into an antigen-responsive mature lymphocyte or for a mature lymphocyte to respond to foreign antigen (reviewed in
64
), an intact nonsense mRNA surveillance system should be critical for B and T
cell development. Thus, it will be interesting to see whether deleting the
HUPF1
gene in a mouse affects lymphocyte development.
We thank Laura Hartwell for help in preparing rabbit antisera, Drs G.Napolitano,
C.Blobel, G.Weskamp, P.Tucker, G.Wu and S.Elledge for genomic and cDNA
libraries, Drs T.Ellis, R.Pieper and R.Shankar for human cell lines, Dr M.Wabl
for discussions and Drs I.Haas and S.Amero for critical reading of the
manuscript. This work was supported in part by the National Institutes of
Health (R29CA56772-01A1), the Cancer Society of America and the Tobacco Research Council of
America.
*To whom correspondence should be addressed. Tel.: +1 708 216 5816; Fax: +1 708
216 9574; Email: hjaeck@wpo.it.luc.edu
+
Present address: Department of Pathology, University of Uppsala, Uppsala, Sweden
REFERENCES
Return