A novel 2D-based approach to the discovery of candidate
substrates for the metalloendopeptidase meprin
Daniel Ambort
1
, Daniel Stalder
2
, Daniel Lottaz
1
, Maya Huguenin
1
, Beatrice Oneda
1
, Manfred Heller
2
and Erwin E. Sterchi
1
1 Institute of Biochemistry and Molecular Medicine, University of Berne, Switzerland
2 Department of Clinical Research, University Hospital, Berne, Switzerland
The astacin-like zinc-dependent metalloendopepti-
dase human meprin (hmeprin) (EC 3.4.24.18) was
first discovered in 1982 for its ability to hydrolyze
N-benzoyl-l-tyrosyl-p-aminobenzoic acid, a chymo-
trypsin substrate used for assessing exocrine pancreas
function [1]. N-benzoyl-l-tyrosyl-p-aminobenzoic acid
hydrolase (PPH) was subsequently purified and charac-
terized from human small intestinal mucosa [2]. At the
same time, PPH orthologs, called meprin (metal endo-
peptidase from renal tissue) or endopeptidase-2, were
found in mouse and rat kidney, respectively [3,4]. Two
similar subunits, termed meprinaand meprinb, with
molecular masses of 95 and 105 kDa, respectively,
were identified. Human meprin cDNA was expressed
in Madin–Darby canine kidney (MDCK) cells, a well-
established cell system for polarized epithelial cells. To
date, no such thoroughly characterized model system
exists for human epithelial cells. Hmeprinais secreted
into the culture medium of MDCK cells as inactive
homodimers, whereas hmeprinbis primarily mem-
brane-bound [5]. Hence, heterodimers of hmeprina⁄b
allowed for localization of the a-subunit to the plasma
membrane [6]. Inactive zymogens of hmeprinaand b
are processed by limited proteolysis with trypsin into
their active forms [5,6]. Hmeprina, but not b, may
alternatively be activated by plasmin [7,8].
A first step towards the elucidation of the biological
function of meprin was achieved by testing putatively
cleavable polypeptide substrates. A variety of pro-
tein and peptide substrates were processed in vitro;
Keywords
astacin family; image analysis; Madin–Darby
canine kidney cells; meprin; protease
proteomics
Correspondence
E. E. Sterchi, Institute of Biochemistry and
Molecular Medicine, University of Berne,
Bu
¨hlstrasse 28, CH-3012 Berne, Switzerland
Fax: +41 31 631 3737
Tel: +41 31 631 4199
E-mail: erwin.sterchi@mci.unibe.ch
(Received 14 April 2008, revised 8 July
2008, accepted 10 July 2008)
doi:10.1111/j.1742-4658.2008.06592.x
In the past, protease-substrate finding proved to be rather haphazard and
was executed by in vitro cleavage assays using singly selected targets. In the
present study, we report the first protease proteomic approach applied to
meprin, an astacin-like metalloendopeptidase, to determine physiological
substrates in a cell-based system of Madin–Darby canine kidney epithelial
cells. A simple 2D IEF ⁄SDS ⁄PAGE-based image analysis procedure was
designed to find candidate substrates in conditioned media of Madin–
Darby canine kidney cells expressing meprin in zymogen or in active form.
The method enabled the discovery of hitherto unkown meprin substrates
with shortened (non-trypsin-generated) N- and C-terminally truncated
cleavage products in peptide fragments upon LC-MS ⁄MS analysis. Of 22
(17 nonredundant) candidate substrates identified, the proteolytic process-
ing of vinculin, lysyl oxidase, collagen type V and annexin A1 was analysed
by means of immunoblotting validation experiments. The classification of
substrates into functional groups may propose new functions for meprins
in the regulation of cell homeostasis and the extracellular environment, and
in innate immunity, respectively.
Abbreviations
ADAM, a disintegrin and metalloprotease; BMP-1, bone morphogenetic protein 1; CID, collision-induced dissociation; ECM, extracellular
matrix; hmeprin, human meprin (EC 3.4.24.18); ICAT, isotope-coded affinity tag; MDCK, Madin–Darby canine kidney; MMP, matrix
metalloproteinase; PPH, N-benzoyl-L-tyrosyl-p-aminobenzoic acid hydrolase; TLD, tolloid.
4490 FEBS Journal 275 (2008) 4490–4509 ª2008 The Authors Journal compilation ª2008 FEBS
biologically active peptides [2,9], as well as gastrointes-
tinal peptides and extracellular matrix (ECM) com-
ponents, such as collagen type IV, fibronectin and
laminin-nidogen [10–12]. These findings suggest that
meprin may be involved in processes such as renal clear-
ance of vasoactive peptides from blood plasma, regula-
tion of cell movement, secretory activity and growth of
intestinal tract, and tissue remodelling. In addition,
marked differences between a- and b-subunits in sub-
strate and peptide bond specificity point to distinct func-
tions for the two forms [10]. Meprinaselects for small
(e.g. serine, alanine and threonine) or hydrophobic (e.g.
phenylalanine) residues in the P1 and P1¢sites and pro-
line in the P2¢position. Meprinbprefers acidic amino
acids in the P1 and P1¢sites and selects against basic res-
idues at P2¢and P3¢. In conclusion, protease-substrate
discovery executed by these in vitro cleavage assays was
rather haphazard. Thus, meprin and its substrate reper-
toire may be studied in a complex biological context to
identify physiologically relevant substrates.
The introduction of protease proteomics enabled
identification of protease and protease-substrate reper-
toires on an organism-wide scale by means of proteomic
techniques [13]. Using different cell-based systems [14–
16] a variety of hitherto unkown substrates were found
in conditioned media for the metzincin metalloendopep-
tidases, a disintegrin and metalloprotease (ADAM)-17
and matrix metalloproteinase (MMP)-14. Human
plasma was also used to identify substrates for recombi-
nant MMP-14 in a cell-free system [17]. Two methodo-
logical platforms were successfully applied for protein
separation: LC-MS ⁄MS and 2D IEF ⁄SDS ⁄PAGE
[14–17]. These standard techniques were used in com-
bination with lectin-affinity pre-fractionation and
quantitative tags such as isotope-coded affinity tags
(ICAT) or cyanine dyes for differential in-gel electro-
phoresis. From these protease proteomic studies, it
became obvious that metalloendopeptidases are key
modulators of diverse signalling pathways and not
merely ECM degrading entities [18]. For example, the
major role of the MMP family is the control of cellular
responses critical to homeostatic regulation of the extra-
cellular environment and the immune response [19,20].
We decided to apply protease proteomics to identify
novel physiologic substrates for meprin, aiming to
elucidate its key functions at the cellular level. For
the above described techniques, some conceptual prob-
lems may arise: first, ICAT-based approaches compare
pairs of peptides, and therefore it is not possible to
discover cleaved protein fragments with shortened
(non-trypsin-generated) N- or C-termini; second,
nonglycosylated proteins and fragments escape from
lectin-affinity purification. We thus designed a simple
2D IEF ⁄SDS ⁄PAGE-based protease proteomic appro-
ach that remedied these limitations and circumvented
complicated quantitative and statistical evaluation.
Hmeprina⁄bwas transfected into MDCK cells and
activated in situ by limited trypsin treatment at conflu-
ent cell stage. Conditioned media of meprin activated
and non-activated cells were concentrated with ultrafil-
tration and then separated by 2D IEF ⁄SDS ⁄PAGE. A
simple 2D IEF ⁄SDS ⁄PAGE-based image analysis pro-
cedure allowed for detection of protein spots unique to
2D gels produced from conditioned media of meprin
activated cells. LC-MS ⁄MS analysis of candidate
substrates confirmed the validity of this protease prote-
omic approach for the discovery of shortened (non-
trypsin-generated) N- and C-terminally truncated
cleavage products in peptide fragments.
Results
Design and application of a simple 2D IEF/SDS/
PAGE-based protease proteomic approach
in substrate finding
Traditionally, 2D IEF ⁄SDS ⁄PAGE-based image analy-
sis is performed on two sets of gels and protein spots
are matched to the same reference gel within one single
analysis. Statistical tools are then applied to quanti-
tatively assess subtle but significant changes in peak
volumes to find up- or down-regulated protein spots.
Unfortunately, error-prone matching to wrong refer-
ence spots is often underestimated, making quantita-
tive statistical information useless. Hence, annotations
of interesting candidate spots to wrong spots in the
reference gel leads to misinterpretation of the data set
and protein spots unique to only one specific condition
are then not properly displayed in the corresponding
reference gel. A remedy to false-positive data interpre-
tation is the stepwise reduction in complexity of such
an analysis. Therefore, we designed a simple image
analysis procedure in which digitized 2D gels were cut
into four parts or quadrant sections. This procedure
enabled the performance of four independent image
analyses in which the gel parts of each corresponding
quadrant were used to construct four independent ref-
erence gels instead of one. The corresponding quadrant
sections were grouped into sets of gels termed level 1
match-sets for each condition (activated meprin versus
non-activated meprin) and then into supersets of level
1 match-sets (higher-level match-sets) (Fig. 1). The
four level 1 match-sets are the reference gels of the
respective quadrants from the 2D gel sections of each
condition and the four higher-level match-sets are the
reference gels of the two different conditions (activated
D. Ambort et al. Meprin protease proteomics
FEBS Journal 275 (2008) 4490–4509 ª2008 The Authors Journal compilation ª2008 FEBS 4491
meprin versus non-activated meprin). This procedure
allowed for subsequent matching of protein spots first
to reference gels of the same condition and thereafter
to reference gels common to both conditions. The step-
wise annotation of protein spots to two independent
levels of reference gels allowed for detection of unique
spots in the final higher-level match-sets (Fig. 2). These
differential spots were unique to one specific condition
and absent in the other or vice versa. Applying the
above procedure to conditioned media of MDCKa⁄b
cells revealed that, among 817 protein spots displayed,
35 were unique to media of cells expressing activated
meprina⁄band 40 to media of cells with non-activated
meprina⁄b(Table 1). These unique protein spots were
therefore absent in the corresponding other condition.
Thus, unique spots were indicative of proteins released
into or proteolytically cleaved in the extracellular
milieu by hmeprina⁄b. We then hypothesized that,
upon LC-MS ⁄MS analysis of candidate substrates, it
may be feasible to find shortened (non-trypsin-gener-
ated) N- and C-termini in peptide fragments. Such
potential N- or C-terminally truncated cleavage prod-
ucts can be identified in protein spots unique to condi-
tioned media of trypsin activated MDCKa⁄bcells, as
shown below.
Fig. 1. Simple 2D IEF ⁄SDS ⁄PAGE-based image analysis procedure.
The procedure is based on qualitative differences among reference
gels (level 1 match-sets) of each group of five gel replicates (three
pooled biological gel replicates and two more technical gel repli-
cates). Gel replicates of each group (activated meprin versus
non-activated meprin) were cut virtually into four equally spaced
quadrants for four independent image analyses. Reference gels of
each group were then clustered into a new set for higher-level image
analysis. The spot matching features of PDQUEST (version 7.3.1)
allowed for detection of unique protein spots. The combined higher-
level match-set is the final fusion of all annotated unique spots into
one big 2D reference map.
Fig. 2. Application of a simple 2D IEF ⁄SDS ⁄PAGE-based protease proteomic approach in substrate finding. A representative image analysis
of the first quadrant is shown. Two hundred and fifty micrograms of conditioned medium protein from trypsin activated and non-activated
MDCKa⁄bcells was separated by IEF in a 24 cm long IPG pH 3–10 NL strip. Vertical separation was according to mass in a 12.5% SDS gel.
Optimized Ruthenium staining: for each condition (activated meprin versus non-activated meprin), three pooled biological gel replicates (from
18 dishes per pooled sample) and two more technical gel replicates (of one pooled sample) were produced for subsequent image analysis.
Unique protein spots are labelled in level 1 and higher-level match-sets with SSP assigned by the image analysis software.
Meprin protease proteomics D. Ambort et al.
4492 FEBS Journal 275 (2008) 4490–4509 ª2008 The Authors Journal compilation ª2008 FEBS
Protein identification by means of LC-MS/MS,
PHENYX-based and BLASTP-based protein database
searching
By visual inspection, the 35 protein spots unique to
media of trypsin activated MDCKa⁄bcells could be
reduced to 33 putative candidates. The redundancy of
two spots present in more than one quadrant from
each set of 2D gels analysed prompted correction
(Fig. 2; see Fig. S1). On colloidal Coomassie stained
preparative 2D gels, 24 protein spots of interest were
detectable. These spots could be rematched to putative
candidates found in fluorescence stained analytical
gels (data not shown). Gel plugs were then prepared,
in-gel digested with trypsin and peptides thereof
separated ⁄fragmented by LC-MS ⁄MS. Collision-
induced dissociation (CID) spectra interpretation with
phenyx (version 2.1) against the uniprot-SwissProt
protein database (release 48.8) led to 22 (17 nonredun-
dant) protein identifications (Fig. 3 and Table 2). The
taxonomic search space was restricted to Mammalia
(40 084 sequence entries). To double-check significant
hits, the same spectra were interpreted with the web-
based search engine mascot (version 2.1) against the
same database and parameter settings (data not
shown) [21]. The identification of nucleophosmin (pro-
tein spot SSP 2102; Table 2) was accepted because the
peptide VDNDENEHQLSR and its in-source pro-
duced fragment DNDENEHQLSLR were unambigu-
ously identified with good scores by phenyx and
mascot. In addition, the whole tryptic peptide
MSVQPTVSLGGFEITPPVVLR was identified by
phenyx and mascot as first ranking identification, but
with scores below the chosen acceptance criteria
(Table 2 and data not shown). Beside six positive hits
for dog, other species (e.g. rat, human, rabbit and
mouse) were predominantly represented. The current
release (51.3) of the uniprot-SwissProt protein data-
base lists 664 sequence entries for dog and thus may
explain the poor representation in this species.
Recently, the dog genome was sequenced to comple-
tion [22]. Peptide sequence tags deciphered from our
previous analysis permitted search with blastp (version
2.2.16) against the 33 527 dog RefSeq protein sequence
entries of the NCBI [21]. All top scoring significant
hits corresponded to predicted dog protein sequence
entries. Finally, all equivocal uniprot-SwissProt protein
database searches were successfully matched to pre-
dicted dog protein orthologs (Table 3).
Discovery of shortened (non-trypsin-generated)
N- and C-terminally truncated cleavage products
in peptide fragments
phenyx offers the remarkable feature to search for
non-tryptic peptides (i.e. half-cleaved peptides). In-gel
tryptic digestion of proteins contained within gel plugs
produces peptide fragments terminating C-terminally
with a lysine or arginine residue. Trypsin cleavage
specificity is then fixed to the N- or C-terminus.
Table 1. Protein spot matching statistics. 2D IEF ⁄SDS ⁄PAGE-based image analysis was performed with PDQUEST (version 7.3.1) on five gel
replicates (three biological replicates, two technical replicates) of conditioned media from trypsin activated and non-activated MDCKa⁄bcells.
Qualitative spot matching differences among reference gels (level 1 match-sets) are expressed as unique spots (% of each corresponding
quadrant section).
Condition Quadrant
Gel replicates Reference gel
Unique
spots (%)
Replicate 1 Replicate 2 Replicate 3 Replicate 4 Replicate 5
Level 1
match-set
Higher-level
match-set
Activated meprin 1 315 313 318 315 316 318 2 (0.6)
Non-activated meprin 1 334 333 334 332 332 334 18 (5.4)
In total 1 336
Activated meprin 2 221 219 221 215 218 222 10 (4.4)
Non-activated meprin 2 217 212 218 216 217 218 6 (2.6)
In total 2 228
Activated meprin 3 106 103 116 115 117 122 12 (9.2)
Non-activated meprin 3 107 110 113 103 104 119 9 (6.9)
In total 3 131
Activated meprin 4 105 107 115 108 105 115 11 (9.0)
Non-activated meprin 4 110 108 109 102 104 111 7 (5.7)
In total 4 122
Activated meprin All 777 35 (4.3)
Non-activated meprin All 782 40 (4.9)
In total All 817
D. Ambort et al. Meprin protease proteomics
FEBS Journal 275 (2008) 4490–4509 ª2008 The Authors Journal compilation ª2008 FEBS 4493
In silico digestion of the theoretical full-length protein
product with trypsin enables the determination of all
tryptic peptides terminating with a lysine or arginine
residue. Hence, peptide fragments not featuring a
lysine or arginine residue in the C-terminal ends or
truncated in the N-termini by some amino acids rela-
tive to the preceding in silico-generated tryptic frag-
ments are candidates for proteolytically processed
(non-trypsin-derived) cleavage products. In a protease
proteomic approach, this option facilitates the discov-
ery of shortened (non-trypsin-generated) N- or C-ter-
minally truncated cleavage products defined by meprin
protease activity. To determine new peptide ends other
than lysine or arginine, peptides must not be identified
either C- or N-terminal to the truncated peptide. We
applied this strategy to all protein database searches
performed with phenyx. Several shortened half-cleaved
peptides (not full-length tryptic peptides) were detected
(Table 2). Half-cleaved peptides may also originate
from in-source fragmentation of intact tryptic peptides
during the ionization process. Accordingly, the follow-
ing half-cleaved peptides co-eluted with corresponding
intact tryptic peptides after chromatographic separa-
tion: TDGNSEHLKR and DGNSEHLKR from pro-
tein spots SSP 602 ⁄9602 and SSP 1602, respectively;
PGPVFGSK from protein spot SSP 1602; and DNDE-
NEHQLSLR from protein spot SSP 2102. The half-
cleaved peptides derived from the sequence stretching
over amino acids 159–182 of clusterin (IDSLLENDR-
QQTHALDVMWDSFNR) found in protein spots SSP
502 and SSP 1502 were chromatographically separated
and thus may not refer to in-source fragmentation
products. Those half-cleaved products are most proba-
bly related to in-gel digestion artefacts because cleav-
age within this protein sequence stretch by meprin
must be excluded due to an overall amino acid
sequence coverage of this protein that exceeded amino
acid 182. In addition, the two half-cleaved peptides
DQAVSDTELQEMSTEGSK (residues 23–40) and
DTELQEMSTEGSK (residues 28–40) in SSP 502 and
1502 were chromatographically separated and were not
in-source fragmentation products generated during the
ionization process. The former peptide represented the
mature N-terminus of clusterin (aspartate at position
23) and hence was not generated by meprin activity.
The latter peptide was presumably produced by
meprinbwith acidic amino acids preferred in the P1¢
position and selecting against basic amino acids in the
P2¢and P3¢positions [10]. The leguminous lectin-like
VIP36 was present in two different protein spots (SSP
1602 and SSP 602 ⁄9602) and also met our criteria for
shortened (non-trypsin-generated) C-terminally trun-
cated cleavage products in peptide fragments. In both
spots, the truncated peptide LFQLMVEH (residues
273–280) was identified with no further peptides
towards the C-terminal end (not ending with a lysine
A
B
C
Fig. 3. Two-dimensional reference maps on protein identifications.
Representative 2D gel images of conditioned medium protein from
MDCKa⁄bcells. (A) 2D gel of trypsin activated meprin. (B) 2D gel
of non-activated meprin. Unique protein spots were labelled with
SSP defined by image analysis software. (C) Close-up view of one
representative protein spot, namely, SSP 7006. LC-MS ⁄MS analy-
sis of candidate substrates confirmed the validity of this protease
proteomic approach for the discovery of shortened (non-trypsin-
generated) N- and C-terminally truncated cleavage products in
peptide fragments (Table 2).
Meprin protease proteomics D. Ambort et al.
4494 FEBS Journal 275 (2008) 4490–4509 ª2008 The Authors Journal compilation ª2008 FEBS
