(Received for publication, November 26, 1996, and in revised form, February 13, 1997)
From the Division of Bone Marrow Transplantation and
Stem Cell Biology, Departments of Internal Medicine and Genetics,
Washington University Medical School, St. Louis, Missouri 63110-1093, § Khepri Pharmaceuticals, South San Francisco,
California 94080, the ** Departments of Medicine and Surgery, University
of California, San Francisco, California 94143, and the
Laboratory of Experimental Carcinogenesis, Molecular
Cytogenetics Section, NCI, National Institutes of Health,
Bethesda, Maryland 20892-4255
Dipeptidyl peptidase I (DPPI) is a lysosomal cysteine protease that catalyzes the sequential removal of dipeptides from the amino termini of various protein substrates. We have isolated a cDNA coding for murine DPPI from mouse thymus and spleen cDNA libraries. The deduced amino acid sequence codes for a protein of 462 amino acid residues; comparison of this deduced sequence with that of rat and human DPPI revealed 90.1% and 77.8% identity, respectively. Using DPPI cDNA, we obtained two BAC (Bacterial Artificial Chromosome) clones that contained the murine DPPI locus. The DPPI gene consists of seven exons and 6 introns, and spans approximately 20 kilobases. Using fluorescence in situ chromosome hybridization, we localized murine DPPI to chromosome 7D3-E1.1. We determined that DPPI protein is widely distributed in mouse tissues, although its relative abundance varies from tissue to tissue. In contrast to previous reports, we show here that DPPI mRNA and protein levels and enzymatic activity are unchanged during in vitro T cell activation, implying that this enzyme is not rate-limiting for granzyme processing.
Dipeptidyl peptidase I (DPPI,1 or cathepsin C) is a lysosomal cysteine protease that belongs to the papain superfamily of proteases (1, 2). DPPI is capable of sequentially removing dipeptides from the amino termini of various peptides and protein substrates (3-7) and, along with other cysteine proteases, is thought to play an important role in intracellular protein degradation and turnover (4, 8, 9). This enzyme has other postulated functions, including involvement in cell growth (10), neuraminidase activation (11), and platelet factor XIII activation (12). Several reports have also suggested that DPPI activity, although present in a wide variety of tissues, is significantly higher in cytotoxic lymphocytes and myeloid cells (13-15); in these cells, DPPI localizes to the secretory granule compartment and may be the major enzyme responsible for processing serine proteases from the proenzyme form to the catalytically active form (14, 15). Finally, two different reports have shown that DPPI can directly cleave the activation prodipeptides of granzymes A and B (16, 17), serine proteases that are found in the secretory granules of activated cytotoxic lymphocytes (CTL) and natural killer cells (18).
DPPI has been purified from several species (human, rat, bovine, porcine, and goat) (4, 19-23), and the rat and human cDNA sequences have been described (1-2, 24). The amino acid sequences deduced from these cDNAs encode a protein of 462 amino acids (rat) or 463 amino acids (human) with three functional domains: a signal peptide (28 versus 24 residues in rat and human, respectively), an unusually long propeptide (201 versus 206 residues), and the mature region (233 residues in both species). The mature proteins share 87.5% identity with each other, and 28.5-39.5% identity with other members of the papain family (cathepsins H, L, B, O, and K, and papain).
DPPI is initially synthesized as a 55-kDa proenzyme, and is rapidly converted to the two-chain form of the mature enzyme (the enzymes responsible for the proteolytic processing of DPPI are currently unknown). The two chains have molecular masses of 25 kDa (heavy chain) and <10 kDa (light chain) (25). Unlike most other cysteine proteases, which are small monomeric proteins with molecular masses of 20-30 kDa, DPPI exists as an oligomeric enzyme of about 200 kDa, as determined by gel filtration (4, 6, 26, 27). There are differing reports concerning the subunit composition of the enzyme, ranging from dimers to hexamers. Most studies suggest that the monomeric subunit is composed of a heavy chain and light chain (both originating from the mature region of the protein); however, a recent report presented evidence that active human DPPI consists of four identical subunits, each composed of three different polypeptide chains, two of them disulfide-linked (28). Two of the chains showed pronounced similarity to the heavy and light chains of the other papain-like cysteine proteases, whereas the third chain corresponds to the proregion of the enzyme. Thus, it appears that a substantial portion of the proregion remains bound to the mature complex. The regulation of DPPI processing and assembly are unknown.
In this paper, we report the sequence of murine DPPI cDNA, and the genomic organization and chromosomal localization of the gene. We determined that murine DPPI protein is widely distributed, with the highest levels found in the spleen, lung, liver, and small and large intestines. Additionally, our studies revealed that DPPI expression and activity are unchanged following in vitro T cell activation.
Total RNAs from mouse thymus and spleen were isolated by
the guanidinium thiocynate method as described previously (29). The
first strand of cDNA was obtained with reverse transcriptase (Promega) using random hexamer primers. Double-stranded cDNAs were
amplified by PCR using Taq polymerase (Perkin-Elmer) using two sets of primers: 1) 5-AACTGCACTTACCCTGACCTG-3
and
5
-GGCTTCATTGCAGCCACCATAG-3
(to amplify the 5
portion of the
cDNA); 2) 5
-TGCCCAAGGTTGTGATGGTGGA-3
and
5
-CTACAATTTAGGAATCGGTATGGC-3
(to amplify the 3
portion of the
cDNA). These primers were derived from the rat cDNA sequence. The amplified products were subcloned into the pCR2.0 vector
(Invitrogen) and transformed into JM109. Plasmids with appropriate
insert lengths were sequenced on an Applied Biosystems model 373 automated sequencer, using vector-based external primers and
cDNA-specific primers constructed on the basis of previously
obtained sequences.
To obtain the 5-untranslated region and leader sequence, a mouse
thymus cDNA library in
ZAP was screened using PCR, with primers
based on
ZAP vector sequences and the reverse primer 5
-CGGAGGCAATTCTCCCTTGGT-3
, based on the murine cDNA sequence. The
amplified products were subcloned into pCR2.0 (Invitrogen) and
sequenced.
To obtain the full length genomic DPPI clone, the first 833 bp of the mouse cDNA sequence was used to screen a murine BAC library (Genome Systems, St. Louis, MO). Two clones (BAC 7930 and BAC 7931) hybridized strongly with the cDNA probe. DNA obtained from these clones were digested with EcoRI, and four fragments were subcloned into pUC19 based vectors. These plasmids were sequenced using a series of oligonucleotide primers defined initially by the cDNA sequences and subsequently by derived sequences. To establish the distance between exons, primers flanking the intron/exon borders were designed and PCR amplifications were carried out on the BAC clones using KlenTaqLA (Wayne Barnes, Washington University, St. Louis, MO) and prolonged (5-10 min) extension cycles. Single specific bands derived from each set of primers were directly subcloned into pCR2.0 (Invitrogen) and sequenced to confirm the intron/exon borders.
Southern AnalysisDNA obtained from 129/SvJ ES cells, mouse
tails, or from BAC clones was digested with EcoRI and
BamHI, electrophoresed on 1% agarose gels, transferred onto
Hybond (Amersham) membranes, and probed with either a 833-bp 5
cDNA or a 645-bp 3
cDNA probe, as described previously
(30).
Chromosomes obtained from normal mouse spleen cultures were used for fluorescence in situ hybridization (FISH). The DPPI probe (BAC 7930) was labeled with biotin or digoxigenin using the Random-Prime DNA labeling kit (Boehringer Mannheim). The FISH protocol was described in detail elsewhere (31, 32). Slides were pretreated with RNase, denatured in 2 × SSC, 70% (v/v) formamide for 2 min at 70 °C. The DNA probes (200 ng), mouse Cot-1 DNA (Life Technologies, Inc.) in 2 × SSC-50% (v/v) formamide, 10% (w/v) dextran sulfate, 2 × Denhardt's solution, 1% Tween 20 (v/v) were denatured for 5 min at 70 °C, reannealed for 2 h at 37 °C, and hybridized in a humid environment for 18 h at 37 °C. A post-hybridization final wash was performed in 0.1 × SSC at 60 °C. Biotin- and digoxigenin-labeled DNA was detected by fluorescein isothiocyanate-conjugated avidin DCS (Vector Laboratories) and rhodamine-conjugated antidigoxigenin (Boehringer Mannheim), respectively.
Chromosomes were counterstained with prodidium iodide or 4,6-diamino-2-phenylindole (DAPI) and examined with a Zeiss Axiophot epifluorescent microscope with a 100-watt mercury lamp. Digital images of selected metaphase spreads were recorded using cooled charge-couple device camera CH250 (Photometrics, Tuscon, AZ) and a filter system consisting of a triple-band-pass beam splitter and emission filters. Excitation of each of fluochromes used was accomplished by single-band-pass excitation filters in a computer-controlled motorized filter wheel. This made it possible to acquire sequential, registration shift-free gray-scale images of two or three fluorochromes (DAPI, fluorescein isothiocyanate, and/or rhodamine). Images were processed and analyzed on an Apple Power Macintosh 8100/100 computer using Oncor's recording and the analytic program Image, as well as NIH's Image and Yale University's Gene Join. To identify individual chromosomes and to assign the location of signal to specific chromosome regions, a method for direct visualization of fluorescent spots on look-up-table (LUT)-inverted digital images of DAPI banded chromosomes was used (33). To confirm the identity of chromosomes, preparations were hybridized (34) with a mouse chromosome 7 painting probe (Cambio, Cambridge, United Kingdom) and previously observed labeled metaphases were rerecorded.
RNA Preparations and S1 Nuclease Protection AssaysRNA was
purified from adult 129/SvJ mouse organs or from activated splenocytes
using a guanidinium thiocyanate miniprep protocol, as described (29).
Since the first exon of DPPI did not contain any unique restriction
sites that would permit the construction of a specific S1 probe, we
designed a PCR probe spanning the 5-flanking sequence and the putative
first exon, using the primer set 5
-CCCAAAGGCCACTTTCCTTC-3
(forward)
and 5
-GAACTTCTAGGGCCCACCTGG-3
(reverse). The 478-bp product was
gel-purified and end-labeled using [
-32P]ATP and
polynucleotide kinase as described previously (35). 20 µg of total
RNA was hybridized with 100,000 cpm of this labeled probe, and an S1
nuclease protection assay was performed as described previously (35).
The specific probes for murine granzyme B and murine
2
microglobulin (m
2M) have been described previously (36).
The reverse primer
5-CAGGAGTGTCGGAGCGCACGGTGCAGACTCCCAAAAGCAC-3
(directed against
sequences in exon 1) was end-labeled with [
-32P]ATP
and polynucleotide kinase. 50-100 µg of total cellular RNA was
lyophilized with 100,000 cpm of the labeled oligonucleotide, and primer
extension was carried out as described previously (37).
Splenocytes were either activated with 5 µg/ml ConA (Sigma) supplemented with 50 units/ml recombinant human IL-2 or in a one-way mixed lymphocyte cultures as described previously (38). For the generation of Ad-Lak cells, splenocytes were cultured in the presence of high dose IL-2 (1000 units/ml) for 10 days as described previously (38).
Western AnalysisOrgans from 129/SvJ mice and splenocytes
harvested from IL-2/ConA activation were suspended in
phosphate-buffered saline, and then an equal volume of 2 × extraction buffer (2 M NaCl, 50 mM Tris, pH
8.0, 0.2% Triton X-100) was added. The extract was sonicated, and
crude cellular debris was removed by centrifugation at 15,000 rpm at
4 °C × 15 min. Thirty µl of this extract was used in each
lane of an SDS-PAGE analysis using reducing conditions. Immunoblotting
was carried out using a rabbit anti-murine DPPI. Murine anti-DPPI was
generated by immunizing rabbits (EL Labs, Soquel, CA) with recombinant
DPPI protein (amino acids 230-462) made in Escherichia
coli. The IgG fraction of the antiserum was purified and was shown
to react specifically with recombinant DPPI protein (data not shown)
and tissue extracts. The antibody recognizes the 25-kDa mature form of
the DPPI protein, in agreement with a previous report (2). Murine
granzyme B antiserum (38) and murine -actin antiserum (Sigma) were
used in control blots. Antibody-protein complexes were detected with a
peroxidase-labeled anti-rabbit (DPPI and granzyme B) or anti-mouse
(
-actin) antibody, followed by enhanced chemiluminescence (ECL,
Amersham).
Bovine DPPI, hydroxylamine hydrochloride, Gly-Phe amide, 2-mercaptoethylamine hydrochloride, trichloroacetic acid, and ferric chloride solutions (all from Sigma) were prepared according to manufacturer's directions. Protein concentrations of cell lysates were assayed using the Bio-Rad method (39) and bovine serum albumin as a standard. DPPI activity was assayed in a 0.50-ml reaction mix (pH 6.8) containing 400 mM hydroxylamine, 50 mM Gly-Phe amide, 25 mM 2-mercaptoethylamine, and 100 µl of cell lysate. The reaction was incubated at 37 °C overnight and stopped by the addition of 0.50 ml of 20% trichloroacetic acid and 0.50 ml of 5% ferric chloride solution. The absorbance values at 510 nm for control and test lysates were recorded, and units of DPPI activity/mg of protein were calculated using purified bovine DPPI (Sigma) (assayed simultaneously by the same method over a concentration range of 0.0001-0.1 units) as a standard.
We used a
series of primers based on the rat cDNA sequence and RT-PCR to
isolate the cDNA for murine DPPI. We were unable to obtain a
full-length cDNA using primers that encompassed the entire mature
rat coding region. However, using additional primer sets based on the
rat sequence, we were able to amplify several 5 and 3
fragments from
mouse and thymus cDNA, which together encompass most of the mature
coding region. PCR fragments from at least three separate reactions for
each primer set were subcloned and sequenced to confirm their identity.
To obtain the 5
-untranslated region and leader sequence, we used
primers based on
ZAP sequences and primers derived from the mouse
sequence to screen a mouse thymus cDNA library. Fig.
1A shows the mouse cDNA and deduced amino
acid sequences. The cDNA contains an open reading frame of 1386 nucleotides. The cDNA sequence isolated did not contain a poly(A)
tail; however, a putative polyadenylation signal (AATAAA) is found 385 bp from the TAG stop codon in the genomic sequence (Fig.
2A). The mouse sequence in the coding region
shares 92.8% and 82.1% identity with the rat and human cDNAs,
respectively. The percentage of identity for all three species is
highest in the region coding for the mature protein (Table
I). The initiation codon was assigned to the first
in-frame ATG at position 25; the nucleotide sequence around it
(AGCATGG) is nearly a perfect match with the Kozak consensus sequence.
The deduced amino acid sequence of the open reading frame encodes 462 amino acids, yielding a calculated molecular mass of 52.41 kDa.
Alignment of the deduced amino acid sequence of the mouse DPPI with rat
and human DPPI revealed 90.1% and 77.8% sequence identity,
respectively. The three amino acid residues that form part of the
active site of all papain family members (Cys25,
His159, and Asn197, according to the papain
numbering system) are conserved in the mouse sequence and across
species (Fig. 1B). The deduced protein sequences
24DTPANCTYPD33,
230LPESWDWRN238, and
394DPFNPFELTN403 (which in human and rat
correspond to NH2-terminal sequences of the proregion,
mature, and light chains; Ref. 28) are also well conserved in the mouse
cDNA clone and align exactly with the human and rat sequences. A
search for protein modification sites in the deduced amino acid
sequence revealed three potential N-glycosylation sites (see
Fig. 1B): two in the proregion (at residues 29 and 53) and
one in the mature peptide (at residue 275). The positions and sequences
of these sites are also conserved in all three species, suggesting that
these sites may be glycosylated, potentially serving as recognition
signals for targeting the enzyme to lysosomes.
|
We used the first 833 bp
of the coding sequence to screen a mouse BAC library derived from
129/SvJ mice and obtained two clones (BAC 7930 and 7931) that
hybridized strongly with the cDNA probe. An EcoRI and
BamHI digest and Southern analysis of these two clones indicated that these BAC clones contain the entire coding sequence (data not shown). EcoRI-cleaved genomic DNA and BAC DNA
probed with the 5 cDNA fragment revealed three hybridizing
fragments (5, 3.8, and 2.3 kb). These fragments were subcloned into
pUC19 vectors and subjected to DNA sequencing using a series of
oligonucleotides based on the cDNA sequence. The 5-kb
EcoRI fragment contains the 5
-flanking sequence, the
5
-untranslated sequence, and the ATG start codon with exons 1 and 2 (see Fig. 2). The 2.3-kb EcoRI fragment contains exon 3. The
3.8-kb EcoRI fragment contains exons 4 and 5. A 3
cDNA
probe identified a 5-kb EcoRI fragment. This fragment
contains the 3
portion of exons 6 and 7, and the 3
-untranslated region. Exon 7 contains a single in-frame TAG stop codon, and an AATAAA
polyadenylation motif. To confirm the 5
portion of exon 6, we used
forward primer derived from exon 5 and a reverse primer starting from
the internal EcoRI site in exon 6, and obtained a single PCR
fragment of 5 kb. This fragment contains intron 5 and the remainder of
exon 6 coding region. Once the intron/exon borders were established, we
designed primers that flanked the rest of the exon borders and
subjected the BAC clones to PCR to establish the intron sizes. The
results are shown in Fig. 2; all the intron donor and acceptor sites
conform to the consensus rules (40). The entire gene therefore consists
of 7 exons and 6 introns, and spans over 20 kb (Fig.
2B).
The DPPI gene was
localized by FISH using chromosomes prepared from mouse spleen
cultures. Symmetrical fluorescent signals on sister chromatids of a
medium-size chromosome was observed in 180 out of 200 metaphases
hybridized with DPPI probe (Fig. 3, A and
C). The probe had a high specificity for this site, since symmetrical signals were not observed on other chromosomes. In 35 metaphases with satisfactory DAPI G-like banding, the FISH signals were
localized relative to the chromosome bands, and the DPPI gene locus was
assigned to chromosome 7D3-E1.1 (Fig. 3B). Since the
resolution of DAPI banded mouse chromosomes is not always satisfactory,
recorded metaphases were also rehybridized with mouse chromosome 7 painting probe. All the chromosomes with the specific hybridization
signal were also positive for painting (Fig. 3D). This
region is homologous with human chromosomes 15q, 6p, and 11q
(41-43).
Defining the 5
To define the
transcription initiation site of DPPI gene, primer extension analysis
and S1 nuclease protection assays were performed (Fig.
4). For primer extension analysis, a 40-mer
complementary to mRNA (located in the middle of exon 1) was
end-labeled and annealed with 50-100 µg of total cellular RNA
derived from mouse spleen or lung tissues. The primer extension product
revealed a major band of 150 nucleotides (nt). The strongest band
corresponds to the G residue 68 nt upstream from the ATG site, and was
designated residue +1 (Fig. 4A). To confirm the location of
the 5 end of mRNA, an S1 nuclease protection assay was performed.
For a DPPI-specific S1 probe, we generated a double-stranded DNA
fragment by PCR using a reverse primer derived from exon 1 and a
forward primer derived from the 5
-flanking region. This probe was
end-labeled and annealed to 20 µg of total cellular RNA derived from
mouse spleen. A probe fragment of 212 nt was protected from S1
digestion by DPPI mRNA. A DNA sequencing ladder, generated using
the same exon 1 reverse primer, confirmed that the initiation site is
located at the same G residue identified by primer extension (Fig.
4B). These results imply that there are no additional exons
and no alternatively spliced 5
-untranslated exons.
DPPI Is Constitutively Expressed in Many Tissues, and Is Not Up-regulated following Splenocyte Activation
We next examined
DPPI protein levels in adult mouse organs using Western blot analysis
(Fig. 5) with a polyclonal rabbit anti-DPPI antibody
generated against recombinant murine DPPI (see "Experimental Procedures"). The Western blots showed that the 25-kDa mature form of
DPPI is detected in almost all tissues, but at varying levels. DPPI is
most abundant in lung, liver, spleen, and small and large intestines;
the abundance is intermediate in bone marrow, thymus, and stomach, but
low in kidney. Heart and brain have very little detectable DPPI
protein. S1 nuclease protection assays revealed a similar wide tissue
distribution of DPPI mRNA (data not shown).
To determine whether the expression of DPPI is regulated following
in vitro lymphocyte activation, we examined DPPI mRNA
levels in resting, day 10 Ad-Lak cells (obtained from splenocyte
cultures in the presence of high dose Il-2), day 1-5 MLR (Fig.
6), as well as day 2 and day 4 CTL derived from
IL-2/ConA-activated splenocytes (Fig. 7A).
These RNAs were analyzed using S1 nuclease protection assays with the
exon 1 DPPI probe described above, a granzyme B probe (36) (as a marker
of lymphocyte activation), and a m2M probe (as a control
for quality and content of mRNA). The results indicate that
granzyme B mRNA is up-regulated following T cell activation, as
expected; however, the level of DPPI mRNA is either unchanged
(compared with resting levels) or actually decreased following
activation.
To extend these results, the same day 0, 2, and 4 IL-2/ConA cells were used to make whole cell lysates that were subjected to Western blot analysis. Fig. 7B shows these blots sequentially probed for DPPI, granzyme B, and actin (as a control for protein loading). DPPI protein is present in resting splenocytes, but the level did not change substantially with activation. Granzyme B, on the other hand, was not detected in resting splenocytes, but was markedly up-regulated following T cell activation, as expected.
Finally, DPPI activity in these lysates was determined by examining levels of hydrolysis using the specific substrate Gly-Phe amide. IL-2/ConA-activated splenocyte lysates were assayed for DPPI activity and compared with standards prepared from commercially available purified bovine DPPI. As shown in Fig. 7C, approximately equal amounts of DPPI activity were present in all three lysates.
In this report, we characterized the murine DPPI gene and determined that this gene is highly conserved among species. The murine DPPI gene is localized on chromosome 7 and is a single copy gene. DPPI protein is widely distributed in mouse tissues; its expression is not up-regulated by mitogens or allogeneic stimuli known to activate T lymphocytes.
DPPI exhibits a number of similarities with other lysosomal papain-like cysteine proteases. First, in the mature region of the enzyme, the amino acid sequence of DPPI shares significant homology with the other members of the papain family of cysteine proteases (up to 39% in the case of cathepsin H). It also retains the three amino acid residues that form the active sites of all papain-like proteases. However, unlike other cysteine proteases, DPPI contains a very long proregion, the function of which is still unknown. Although most other cysteine proteases are small monomeric proteins with molecular masses of 20-30 kDa, DPPI exists as an oligomeric enzyme of about 200 kDa. DPPI therefore represents a distinct class of papain-like proteases.
In this report, we have defined the genomic organization of the murine DPPI gene. Southern blot analysis and chromosomal analysis by FISH revealed that DPPI represents a single locus in the mouse genome. The gene was localized by FISH to chromosome 7 at region D3-E1.1. This region has limited homology to human chromosome 15q 11-ter (41-43), but synteny with this region has not yet been established. Several genes have been localized within the same region of chromosome 7. The oncogene product of feline sarcoma virus and the gene for mitochondrial isocitrate dehydrogenase have been localized to the 7 B3-E1 region; their human homologs were mapped to the distal segment of chromosome 15. The genes for malic enzyme and tyrosinase were localized at 7 E1-E2; their human counterparts are located on chromosome 6p25-24 and 11q21, respectively (41-43). Finally, the mouse hepatic fusion mesoderm development and taupe genes (41-43) are localized within the same region as DPPI.
The murine DPPI gene has seven exons and spans more than 20 kb.
Cathepsins B and H both also span more than 20 kb, and comprise 10 and
more than 12 exons, respectively (44, 45). In contrast, cathepsin L is
a relatively compact gene of 8.5 kb, comprising 8 exons (46). The
intron-exon organization of cysteine proteases is only partially
conserved (Fig. 8). DPPI, and cathepsins H and L share
the common intron insertion position that isolates the exon containing
the active site cysteine residue (47). This highly conserved domain is
split by an intron in cathepsin B (44). DPPI also shares one common
intron insertion site with cathepsin B (intron 6-exon 7 for DPPI and
intron 5-exon 6 for cathepsin B). In all four genes, the active site
histidine and asparagine residues are found within one exon, although
the exon on which they reside, and the positions of the intron-exon
insertion sites, differ from gene to gene (44, 47). These similarities
suggest that these cysteine proteases arose from a common ancestral
gene, but that their structures have evolved to include intron losses and insertions.
Several studies have suggested that DPPI activity is enriched in in vitro and in vivo activated cytotoxic lymphocytes (14). In addition, DPPI activity was found to colocalize with serine protease activity (such as granzyme A tryptase activity) in the specialized granule fraction of CTL (14). Several studies have also demonstrated the ability of DPPI to directly cleave the prodipeptide of various serine proteases, thereby activating them (16, 17). To examine whether the expression of DPPI is regulated during in vitro lymphocyte activation, we examined DPPI mRNA and protein levels, and DPPI activity (using a defined synthetic substrate) in resting splenocytes, or splenocytes activated with various protocols (Figs. 6 and 7). To our surprise, we found that DPPI mRNA were not up-regulated in Ad-Lak cells, MLR-derived cells, or Il-2/ConA-activated cells. Moreover, the protein levels and enzymatic activities seem to correlate with mRNA expression.
The discrepancy between these findings and previously published data (13-15) (measuring DPPI enzymatic activity only) may be due to several factors. In our study, we used unfractionated splenocytes; activated CD8+ T cells could potentially contain higher levels of DPPI activity than resting splenocytes or activated CD4+ T cells and B lymphocytes. However, we were also unable to find a significant increase in DPPI levels in either MLR or Ad-Lak cells, both of which contain relatively pure populations of activated cells. Also, in one study, the elevated levels of DPPI activity were not detected in all activated populations of CD8+ T cells (14). In fact, the level of DPPI activity in in vitro activated CD8+ T cells was unchanged compared with resting CD8+ T cells, which agrees with our data. Furthermore, despite the wide variations in levels of DPPI activity in various populations of in vitro and in vivo activated CTLs, these T cells all exhibited similar levels of cytolytic activity (14). These results suggest that DPPI is not up-regulated during the activation of T cells, but that its expression is constitutive.
These data do not alter the hypothesis that DPPI plays an important role in the posttranslational processing of serine proteases. Lytic granules (and the serine proteases contained within them) are not found in resting T cells, but are synthesized after T cell activation. The serine proteases undergo proteolytic processing during their intracellular transport and are packaged into secretory granules (with the prodipeptides presumably already cleaved). DPPI therefore may interact with the serine proteases shortly after they leave the Golgi apparatus, although the exact subcellular compartment where this processing takes place is not yet established.
The availability of DPPI probes and antisera will enable us to further study the functions of DPPI in normal and pathological conditions. Expression of purified rDPPI protein will also help to establish the role of this protease in granzyme processing. Finally, the creation of a loss of function mutation in mice, using homologous recombination techniques, should clarify the importance of this enzyme in vivo.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBankTM/EMBL Data Bank with accession number(s) U89269[GenBank].
We thank Nancy Reidelberger for the excellent preparation of the manuscript.